Datasets:

pratyushmaini

/

llm_dataset_inference

like 1

at the Fermi level (in other words, the network of valley lines for which $u < 1/E_F$) must allow one subregion to percolate through the entire system*]{}. In 2D, if a subregion crosses the entire domain then its effective valley network has to be disconnected. The percolation transition of the subregion thus exactly corresponds to the percolation transition of the effective valley network itself. This theory also allows us to understand how temperature can modify the conductance of the system. On one hand, its increases the amplitude of disorder that enters the Schrödinger equation. This tends to localize the quantum states and to reduce the conductance. On the other hand, it explores excited states at higher energies, which are less localized, and it couples the states, allowing electrons to hop between states. The existence of a transition at finite temperature in a macroscopic system therefore depends on the outcome of the competition between these effects. Finally one should recall that this theory is formally valid in any dimension. In particular, it means that the localization or delocalization properties of a system of $N$ correlated electrons can also be analyzed, in theory, through the knowledge of the statistical properties of the valley network for the landscape $\displaystyle u = \left(-\sum_{i=1}^N \Delta_i + V\right)^{-1} \mathbf{1}$ in a $3N$-dimensional space. #### Conclusion {#conclusion.unnumbered} What emerges here is a new picture of localization. There are, in fact, not *three* but only *two* types of localization: *low frequency* localization, described by the landscape theory developed in this paper, and *scar* or *high frequency* localization, predicted by the stable orbits of the domain. Our findings demonstrate that low frequency localization is a universal phenomenon, observed for any type of vibration governed by a spatial differential operator $L$ that derives from an energy form. The geometry of the domain and the properties of the operator interplay to create a landscape $u$ which entirely determines the localization properties of the system. First, its network of valley lines (in 2D) or surfaces (in 3D) creates a partition of the initial domain into disjoint subregions which shape the spatial distributions of the vibrational modes and identify precisely the regions confining vibrations. Second, the depth of these valleys determines the strength of the confinement within each subregion. The localization of a given mode of eigenvalue $\lambda$ is controlled by the effective valley

at the Fermi level (in other words, the network of valley occupation for which $ uracil < 1 / E_F$) must allow one subregion to percolate through the entire organization * ] { }. In 2D, if a subregion crosses the entire sphere then its effective valley network has to be unplug. The percolation transition of the subregion thus precisely corresponds to the percolation transition of the effective valley net itself. This theory also allow us to understand how temperature can modify the conductance of the arrangement. On one hand, its increases the amplitude of disorder that figure the Schrödinger equation. This tends to localize the quantum states and to shrink the conductance. On the other hand, it explores excited states at high energies, which are less localized, and it couples the states, allowing electrons to hop between states. The existence of a transition at finite temperature in a macroscopic system therefore depends on the outcome of the competition between these effects. Finally one should recall that this hypothesis is formally valid in any property. In particular, it mean that the localization or delocalization properties of a system of $ N$ correlated electron can also be analyze, in theory, through the knowledge of the statistical property of the valley network for the landscape $ \displaystyle u = \left(-\sum_{i=1}^N \Delta_i + V\right)^{-1 } \mathbf{1}$ in a $ 3N$-dimensional space. # # # # Conclusion { # conclusion.unnumbered } What emerge here is a new picture of localization. There are, in fact, not * three * but only * two * types of localization of function: * abject frequency * localization, described by the landscape theory developed in this paper, and * scar * or * high frequency * localization of function, predicted by the static orbits of the domain. Our findings demonstrate that low frequency localization of function is a universal phenomenon, observed for any type of vibration governed by a spatial differential hustler $ L$ that derives from an energy form. The geometry of the domain and the place of the operator interplay to create a landscape $ u$ which wholly determines the localization properties of the system. First, its network of valley lines (in 2D) or surfaces (in 3D) creates a partition of the initial domain into disjoint subregions which determine the spatial distributions of the vibrational modes and name precisely the regions confining vibrations. Second, the astuteness of these valleys determines the strength of the confinement within each subregion. The localization of a given mode of eigenvalue   $ \lambda$ is controlled by the effective valley

at the Fermi level (in othev words, the network of talley mines fof which $u < 1/E_F$) must allow one sybreguon to percolate throueh the ennire systwm*]{}. Ii 2D, if a subregion crosscf ths entnrx domain then ijs effective valley networn fad to be disconnected. The percolatiog transotlon of the subtegiom thua exactly corresponds to the percomation uransition of the rffective valley network ihselv. This theory also wllows us ti unqwrstand how gemperaturt ean modify jhe conductance of the system. On one kand, its inxrwasfv the amplivude os disorder tmst entars the Schrödinger eqmatioi. Thus tends to localize vhe quantum states agd to redgcz the conductance. On rhw othgr hatd, ig exolodex sxcitef svates at hifher energiws, which are less lpcwoized, and it douplef ehe states, allowing electrons to hop beuween states. The existence od a transition at finlte tempewature in a macroscopic system therefore depends mn thx uuteime ow tje competition between these effects. Finally ogs xhpuld recall thct this theory ix votially valid iv any bjmsnsion. In particulwr, it mgans tyat the ljcalozation or delocalization peoperties of q system of $N$ corrzlated electxons csn alxo be analyzed, in theorv, throhgh the knoaledge of ghe statistical oroierthes of the valley network sor the lendsccpe $\dispuaysjyle u = \left(-\sum_{i=1}^N \Delta_i + V\right)^{-1} \mathbf{1}$ ij a $3N$-bimenvional spafe. #### Conclusion {#conclusion.unnumbecxd} What emergex verv is a nec pictmre of localizaeion. There are, in facc, not *ghree* but knly *twm* types of jocalization: *nlw frequency* localizwtiob, dewcribed cy the landscale theory developed in this paper, and *scxd* or *high frequzuci* localization, lreaiceef uy thq stable orbids ow tfr domxin. Our findlngr dekonstrate that low fsequsncy localization os a univetsal phenjmenon, observrd for any type of vibretion jovernrd fy a spatial differential operztor $L$ thwt qerives from an cnerdy form. Thz geometry of the domain and the propertmes of the operator injerplay to create a lcncscape $u$ whmch eneirely dedermines the localizqtion properties pf the system. First, ifs netfork lf valley lines (in 2D) or surfaces (in 3D) creates a partition of the initial eomain into disjoiht sibregponf which sha'e the spatial disnributions of the vibrational moves and idancify precisely the regions comfkning vibratiuns. Second, the depth of fhese vakleys determines the strength og the confinement within each suvregikn. The locelization of a given mode mf eitenvaluw $\lamnda$ is controlued bn the effeftivt vamley

at the Fermi level (in other words, of lines for $u < 1/E_F$) percolate the entire system*]{}. 2D, if a crosses the entire domain then its valley network has to be disconnected. The percolation transition of the subregion thus corresponds to the percolation transition of the effective valley network itself. This theory allows to how can modify the conductance of the system. On one hand, its increases the amplitude of disorder enters the Schrödinger equation. This tends to localize quantum states and to the conductance. On the other it excited states higher which less localized, and couples the states, allowing electrons to hop between states. The existence of a transition at finite temperature a macroscopic depends on outcome the between these effects. should recall that this theory is any dimension. In particular, it means that the or delocalization of a system of $N$ correlated can also be analyzed, in theory, through the of the statistical properties of the valley network for the landscape $\displaystyle u = \left(-\sum_{i=1}^N V\right)^{-1} \mathbf{1}$ in a space. #### Conclusion What here a picture of There are, in fact, not *three* but only *two* types of *low frequency* localization, described by the landscape theory developed in and or *high frequency* predicted by the stable of domain. Our findings demonstrate frequency is observed any of vibration governed by spatial differential operator $L$ that from an energy form. and the properties of the operator interplay to a landscape $u$ which entirely determines the properties of the system. First, its network of valley lines (in 2D) surfaces (in a partition of the initial domain into disjoint which shape the spatial of the vibrational modes and identify precisely the regions vibrations. the depth these valleys determines strength of the within each subregion. of a mode eigenvalue by the effective valley

at the Fermi level (in other worDs, the netwoRk of vAllEy lInEs foR whiCh $u < 1/E_F$) must allow ONe suBregion to percolate throUgh thE eNTire SYsTem*]{}. In 2d, if a subREgION crOsSeS thE eNTiRe domAin Then its Effective vAllEy Network has to BE dIsconnecteD. ThE percolation TraNsitioN oF thE SubreGioN thus ExactlY CorresPonds to thE pERcolatIOn transITIoN of tHe effective valley NEtWOrk itself. This tHeory aLsO AlLOWs uS to Understand HoW tempERature cAN mODIFy tHE conductance oF the system. ON One Hand, itS iNcrEAses thE amplItUDe oF disorder thAt enTers the ScHrödinGEr equatIOn. This tEnds to LocAliZe thE QuAnTum StATes ANd To rEDucE the condUcTaNce. On The oTHER Hand, It eXploRes exCited states at HigHer eNErgIes, whIch arE lesS lOcaliZed, and It couPlEs the states, alloWing Electrons To hOp BetWeEn staTEs. The eXisTenCe of a trAnsitioN At fInITE TeMperature in a macrosCoPIC sYstem theRefore DEpEnDS on the ouTcOme Of thE COmpetItioN BeTween theSe effeCTs. fiNally onE sHould rEcAll ThaT this THeorY is forMally valId in aNY dimension. In paRTicular, it meanS ThAT ThE LocaLizAtion or deloCaliZAtioN proPErTieS Of a syStem oF $N$ COrRElated electrons can aLsO be anaLyzed, In theory, throuGh the knowlEDGE of the stAtisTIcAL properties of tHe valLey network FOr the lanDscapE $\displayStyle u = \lefT(-\SUm_{i=1}^N \DeltA_i + V\RigHt)^{-1} \mAthBF{1}$ In A $3N$-dimensional SPAce. #### COnClusion {#ConClusion.UnnUmbEreD} WhAt Emerges heRe is a new PiCtUrE oF loCalizATion. TherE aRe, iN fAct, Not *thREe* but oNly *twO* typEs Of LOcaLizatioN: *LoW FRequEnCy* LocaLizAtIon, deScriBEd bY the lanDscape theOry DEvelOpEd In this pAper, and *scar* or *HiGh frequencY* lOcaLizatiON, PredicteD by the stable orbits of the DOmain. OuR fiNdingS demOnstrate tHat Low freQueNCy locaLizatiOn is a UnIveRSAl pheNOMeNon, ObServed for aNY TypE of viBrAtioN governEd by a spatial differENtiAl operator $L$ thAt dErivES FrOm aN EnERgy FoRM. ThE GEometry of the domAin and the pRoPErTies of the oPEraToR interpLay to crEate a LAndscapE $u$ which enTirely detErMineS THe lOcalizatioN propertIes of the sYStem. FIRsT, its nEtwOrk of vAlLey Lines (In 2D) or sURfaCes (in 3d) creatEs A partiTion oF tHe initiaL domain into disjoint subrEgions Which ShaPe the spatIal DIstRibutions Of thE vibrationAl mOdeS and iDenTIfy prEcisELy The REgionS conFIning vibrATiOns. sECoNd, the depth oF THEse ValleYs dETerminEs thE strength of the conFInement within eAch sUBRegIon. tHe loCaLization of a givEn mOdE OF eigenvaLuE $\lambda$ is coNtrolled By THe effEctive Valley

at the Fermi level (in ot her words, thenet wor kof v alle y lines for wh i ch $ u < 1/E_F$) must allow onesu b regi o nto pe rcolate th r o ugh t he en ti r esyste m*] {}. In2D, if a s ubr eg ion crossest he entire do mai n then its e ffe ctiveva lle y netw ork hasto bed isconn ected. Th ep ercola t ion tra n s it ionof the subregiont hu s exactly corre sponds t o t h e pe rco lation tra ns ition of thee ff e c t ive valley networ k itself. T his theor yals o allow s usto und erstand how tem peraturecan mo d ify the conduct ance o f t hesyst e m. O n o ne han d ,its inc reases t he a mplit udeo f d isor der tha t ent ers the Schrö din gere qua tion. This ten ds to l ocaliz e the q uantum states a nd t o reducethe c ond uc tance . On th e o the r hand, it exp l ore se x c it ed states at highe re n er gies, wh ich ar e l es s localiz ed , a nd i t coupl es t h estates,allowi n gel ectrons t o hopbe twe enstate s . Th e exis tence of a tr a nsition at fin i te temperatur e i n am acro sco pic systemther e fore dep e nd s o n theoutco me of the competition bet we en the se ef fects. Final ly one sho u l d recallthat th i s theory is fo rmall y valid in any dime nsion . In par ticular,i t means t hat th e l oca l i za tion or deloc a l izat io n prope rti es of a sy ste m o f $ N$ correlat ed elect ro ns c an al so be analyzed ,inth eor y, th r ough t he kn owle dg eo f t he stat i st i c al p ro pe rtie s o fthe v alle y ne twork f or the la nds c ape$\ di splayst yle u = \left (- \sum_{i=1} ^N \D elta_i + V\right )^{-1} \mathbf{1}$ in a $3N$-di men siona l sp ace. ### # C onclus ion {#conc lusion .unnu mb ere d } Wha t em erg es here is a n ewpictu re oflocaliz ation. There are,i n f act, not *thr ee* but o nl y * t wo * ty pe s of l ocalization: *l ow frequen cy * l ocalizatio n , d es cribedby thelands c ape the ory devel oped in t hi s pa p e r,and *scar* or *hig h frequen c y* lo c al izati on, predi ct edby th e stab l e o rbits of th edomain . Ou rfindings demonstrate that low f requen cy lo cal ization i s a uni versal ph enom enon, obse rve d f or an y t y pe of vib r at ion gover nedb y a spati a ldif f e re ntial opera t o r $L $ tha t d e rivesfrom an energy form.T he geometry of the d oma ina nd t he properties of th eo p erator i nt erplay to c reate ala n dscap e $u$whichentirel y de t ermine s th e l ocalizati onpr o perties o ft he sys tem. F irst,its ne t work o f valley lines ( in 2D ) or su r fac es (i n3D) cre a tesa partitio n of the in itialdoma in in to disj oi nt sub reg io ns which s h ape the s patia l distr ib utio nsof the vib r a tiona l mo de s a nd identi f y p r ec is e lytheregio ns con fining vi b rations. Se c ond, th edep t h of th e se v alleys det erm inest h e strength of t h ec onfin ementwithin each s u bre gi on. The lo c a lizationof a give n mo de ofeigenval ue  $\l am bda $is controlled b y the ef fectivev a lley

at_the Fermi_level (in other words,_the network_of_valley lines_for_which $u <_1/E_F$) must allow_one subregion to percolate_through the entire_system*]{}._In 2D, if a subregion crosses the entire domain then its effective valley network_has_to be_disconnected._The_percolation transition of the subregion_thus exactly corresponds to the_percolation transition_of the effective valley network itself. This theory also_allows_us to understand_how temperature can modify the conductance of the system._On one hand, its increases the_amplitude of disorder_that_enters_the Schrödinger equation. This_tends to localize the quantum states_and to reduce the conductance. On_the other hand, it explores excited states_at higher energies, which are less_localized, and it couples the_states, allowing_electrons to hop between states._The existence of_a transition_at finite temperature_in a macroscopic system therefore depends_on the outcome_of the competition between these effects. Finally_one_should recall that_this_theory_is formally_valid in any_dimension._In particular,_it_means that the localization or delocalization_properties_of a system of $N$ correlated electrons_can also be analyzed,_in_theory, through the knowledge_of the statistical properties of_the valley network for the landscape_$\displaystyle u_= \left(-\sum_{i=1}^N_\Delta_i + V\right)^{-1} \mathbf{1}$ in a $3N$-dimensional space. #### Conclusion {#conclusion.unnumbered} What emerges_here is a new picture of_localization. There are, in_fact, not_*three*_but only *two*_types_of localization:_*low frequency* localization, described by the landscape_theory developed_in this paper, and *scar* or_*high frequency* localization, predicted_by_the stable orbits of the domain. Our_findings demonstrate that low frequency localization_is a universal phenomenon, observed_for_any_type of vibration governed by_a spatial differential operator $L$ that_derives from an_energy form. The geometry of the domain_and_the properties of the operator interplay_to_create a landscape $u$ which entirely_determines_the_localization properties of the system._First, its network of valley lines_(in 2D) or surfaces (in 3D) creates a partition_of the initial_domain into disjoint subregions which_shape_the_spatial distributions of the vibrational modes and identify precisely the_regions confining_vibrations. Second, the_depth of these valleys determines the strength of the confinement_within each subregion. The localization of a_given mode of eigenvalue $\lambda$ is controlled by the effective valley

_i> -19$) galaxies, as summarized in Fig. \[fig14\]. 3. Although the migration takes place also in loose structures it is more efficient in denser environments, such as clusters of galaxies (“Nurture”). 4. Quenching of the star formation is induced in clusters by fast gas-ablation mechanism, namely ram-pressure. 5. Post-Star-Burst galaxies appear the tracers of the migration taking place in dense environments. ![The $g-i$ color versus $i$-band absolute magnitude relation for late-type galaxies (blue contours) and for early-type galaxies (red contours). The arrow marks the effect of Nurture at $z$=0 in dense environments. Low-luminosity blue galaxies have their star formation truncated as a result of gas stripping. \[fig14\]](fig14.ps){width="8.0cm"} We thank Roberto Decarli for help in querying the SDSS archive and Veronique Buat, Gabriella De Lucia, Michele Fumagalli, Bianca Poggianti and Marco Scodeggio for helpful discussions. We are grateful to Paolo Franzetti and Alessandro Donati for their contribution to GOLDMine, the Galaxy On Line Database extensively used in this work (http://goldmine.mib.infn.it). We acknowledge the constructive criticism from an unknown referee. The present study could not be conceived without the DR7 of SDSS. Funding for the Sloan Digital Sky Survey (SDSS) and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, and the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium (ARC) for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, The University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, The Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Part

_ i > -19 $) galaxies, as summarized in Fig. \[fig14\ ]. 3. Although the migration takes place besides in lax structures it is more efficient in denser environment, such as bunch of galaxies (“ Nurture ”). 4. Quenching of the star constitution is induced in clusters by debauched gas - ablation mechanism, namely ram - atmospheric pressure. 5. Post - Star - Burst galaxies look the tracers of the migration taking home in dense environments. ! [ The $ g - i$ semblance versus $ i$-band absolute magnitude relation for late - character galaxies (blue contour) and for early - type galaxy (red contours). The arrow marks the effect of Nurture at $ z$=0 in dense environments. humble - luminosity blue galaxies have their star formation truncate as a result of gas stripping. \[fig14\]](fig14.ps){width="8.0 cm " } We thank Roberto Decarli for help in querying the SDSS archive and Veronique Buat, Gabriella De Lucia, Michele Fumagalli, Bianca Poggianti and Marco Scodeggio for helpful discussions. We are grateful to Paolo Franzetti and Alessandro Donati for their contribution to GOLDMine, the Galaxy On Line Database extensively used in this work (http://goldmine.mib.infn.it). We acknowledge the constructive criticism from an unknown reviewer. The present survey could not be conceived without the DR7 of SDSS. financing for the Sloan Digital Sky Survey (SDSS) and SDSS - II has been provide by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, and the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium (ARC) for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, The University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, The Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Part

_i> -19$) halaxies, as summarized ik Fig. \[fig14\]. 3. Althoogy the kigratjon taker place also in loose structnres it iw more efficient in devser envigonments, wuch qs clusters of galaxies (“Hmrturz”). 4. Quenching of jhe star forkation is induwea nn clusters by fast gas-ablation mechwnism, nsmfly ram-pressurg. 5. Poxe-Stad-Burst galaxies appear the tracers of the migration taling place in dense envirojmenhs. ![The $g-i$ color verdus $i$-band avsoltre magnitude relation for late-type galaxies (blue contours) and for earlv-type galaxuew (rfg contours). Vhe argow marks the effect mf Nurtire at $z$=0 in dekse eivirinments. Low-luminosity blue galaxies havg their stdr formation truncaree as d revult if eas svrilping. \[vig14\]](hig14.ps){width="8.0cj"} We thank Riberto Decarli for neji in querying the SQSF archive and Veronique Buat, Gabriella Ge Mucia, Michele Fumagalli, Bianca Poggianti and Marco Scjdeggio for helpful discussions. We are grateful tm Paomu Fxqnzetgu wnd Alessandro Donati for their contribution ek BOKDMine, the Galcxy On Line Datanade gxtensively usgd in tkjs work (http://goldmine.lib.infn.yt). We acknowleqge yhe constructive criticism drom an unknjqn referee. The predent study eould mot br conceived without the DR7 kf SDSS. Funfing for ffe Sloan Digital Skj Susvey (SDSS) and SDSS-II has bqen provived bv the Alwred P. Slown Foundatlon, tma Participating Indtitujions, dhe Nationwl Science Foundation, the U.S. De'ertment of Engrgf, tve Natioual Aevonautics and S[ace Administrction, thz Japavese Monbuiagakusio, and the Mwx Planck Sochgty, and the Hmgher Edusatiin Fynding Zuuncil for Engkand. The SDSS Web sute is http://www.sdss.prg/. Fhe SDSS is mancyee by the Astropnyskcaj Gesxarch Wonsortium (ASC) fur gne Pafticipaunng Knstotutions. The Partici[atihg Institutions arr bhe Amerixan Musetm of Natural History, Astrophyslcal Mnstitnte Poysdwm, University of Basel, Universjty of Calbrldge, Case Weseern Reserve Unirersity, The University of Chicago, Drexel University, Fermilqb, the Institute fot Acvanced Stuvy, the Japan Pasticipation Group, Thw Johns Hopkins Ukiversity, the Joint Inatituta for Nuclear Astrophysics, the Kavli Institute for Part

_i> -19$) galaxies, as summarized in Fig. Although migration takes also in loose in environments, such as of galaxies (“Nurture”). Quenching of the star formation is in clusters by fast gas-ablation mechanism, namely ram-pressure. 5. Post-Star-Burst galaxies appear the of the migration taking place in dense environments. ![The $g-i$ color versus $i$-band magnitude for galaxies contours) and for early-type galaxies (red contours). The arrow marks the effect of Nurture at $z$=0 dense environments. Low-luminosity blue galaxies have their star truncated as a result gas stripping. \[fig14\]](fig14.ps){width="8.0cm"} We thank Decarli help in the archive Veronique Buat, Gabriella Lucia, Michele Fumagalli, Bianca Poggianti and Marco Scodeggio for helpful discussions. We are grateful to Paolo Franzetti Alessandro Donati contribution to the On Database extensively used work (http://goldmine.mib.infn.it). We acknowledge the constructive unknown referee. The present study could not be without the of SDSS. Funding for the Sloan Sky Survey (SDSS) and SDSS-II has been provided the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department the National Aeronautics and Administration, the Japanese and Max Society, the Higher Funding Council for England. The SDSS Web site is http://www.sdss.org/. The is managed by the Astrophysical Research Consortium (ARC) for the The Institutions are the Museum of Natural History, Institute University of Basel, University Case Reserve of Drexel Fermilab, the Institute for Study, the Japan Participation Group, Johns Hopkins University, the the Kavli Institute for Part

_i> -19$) galaxies, as summarized in FiG. \[fig14\]. 3. AlthouGh the MigRatIoN takEs plAce also in loose STrucTures it is more efficient In denSeR EnviROnMents, Such as cLUsTERs oF gAlAxiEs (“nUrTure”). 4. QUenChing of The star forMatIoN is induced in CLuSters by fasT gaS-ablation mecHanIsm, namElY raM-PressUre. 5. post-STar-BurST galaxIes appear ThE TracerS Of the miGRAtIon tAking place in dense ENvIRonments. ![The $g-i$ cOlor veRsUS $i$-BANd aBsoLute magnitUdE relaTIon for lATe-TYPE gaLAxies (blue contOurs) and for eARly-Type gaLaXieS (Red conTours). thE ArrOw marks the eFfecT of NurturE at $z$=0 in DEnse envIRonmentS. Low-luMinOsiTy blUE gAlAxiEs HAve THeIr sTAr fOrmation TrUnCated As a rESULT of gAs sTripPing. \[fIg14\]](fig14.ps){width="8.0cM"} We ThanK robErto DEcarlI for HeLp in qUeryinG the SdSs archive and VeroNiquE Buat, GabrIelLa de LUcIa, MicHEle FumAgaLli, bianca POggiantI And maRCO scOdeggio for helpful dIsCUSsIons. We arE grateFUl To pAolo FranZeTti And ALESsandRo DoNAtI for theiR contrIBuTiOn to GOLdMIne, the gaLaxY On line DATabaSe exteNsively uSed in THis work (http://golDMine.mib.infn.it). wE aCKNoWLedgE thE constructiVe crITiciSm frOM aN unKNown rEfereE. THE pREsent study could not bE cOnceivEd witHout the DR7 of SDsS. Funding fOR THe Sloan DIgitAL SKY Survey (SDSS) and sDSS-Ii has been prOVided by tHe AlfRed P. SloaN FoundatiON, The PartiCipAtiNg INstITUtIons, the NationAL scieNcE FoundaTioN, the U.S. DEpaRtmEnt Of ENeRgy, the NatIonal AerOnAuTiCs And space aDministrAtIon, ThE JaPanesE monbukAgakuSho, aNd ThE max planck SOCiETY, and ThE HIgheR EdUcAtion fundINg COuncil fOr England. the sdSS WEb SiTe is httP://www.sdss.org/. ThE SdSS is managEd By tHe AstrOPHysical REsearch Consortium (ARC) for THe PartiCipAting instItutions. THe PArticiPatINg InstItutioNs are ThE AmERIcan MUSEuM of naTural HistoRY, astRophySiCal INstitutE Potsdam, University OF BaSel, University Of CAmbrIDGe, casE weSTerN RESerVE university, The UnIversity of chICaGo, Drexel UnIVerSiTy, FermiLab, the INstitUTe for AdVanced StuDy, the JapaN PArtiCIPatIon Group, ThE Johns HoPkins UnivERsity, THe joint insTitute FoR NuClear astropHYsiCs, the kavli INsTitute For PaRt

_i> -19$) galaxies, as sum marized in Fig. \[ fig 14 \]. 3. Although them igra tion takes place alsoin lo os e str u ct uresit is m o re e ffi ci en t i nd en ser e nvi ronment s, such as cl us ters of gala x ie s (“Nurtur e”) . 4. Quenc hin g of t he st a r for mat ion i s indu c ed inclustersby fast g a s-ablat i o nmech anism, namely ram - pr e ssure. 5. Po st-Sta r- B ur s t ga lax ies appear t he tr a cers of th e m igr a tion taking p lace in den s e e nviron me nts . ![Th e $g- i$ col or versus $ i$-b and absol ute ma g nituder elation for l ate -ty pe g a la xi es(b l uec on tou r s)and forea rl y-typ e ga l a x i es ( red con tours ). The arrowmar ks t h e e ffect of N urtu re at $ z$=0 i n den se environments.Low- luminosit y b lu e g al axies have t hei r s tar for mationt run ca t e d a s a result of gasst r i pp ing. \[f ig14\] ] (f ig 1 4.ps){wi dt h=" 8.0c m " } We tha n kRobertoDecarl i f or help i nqueryi ng th e S DSS a r chiv e andVeroniqu e Bua t , Gabriella De Lucia, Michel e F u m ag a lli, Bi anca Poggia ntia nd M arco Sc ode g gio f or he lp f ul discussions. We are g ratefu l toPaolo Franzet ti and Ale s s a ndro Don atif or their contribu tionto GOLDMin e , the Ga laxyOn LineDatabasee x tensivel y u sed in th i s w ork (http://g o l dmin e. mib.inf n.i t). Weack now led geth e constru ctive cr it ic is mfro m anu nknown r ef ere e. Th e pre s ent st udy c ould n ot beconceiv e dw i thou tth e DR 7 o fSDSS. Fun d ing for th e Sloan D igi t al S ky S urvey ( SDSS) and SDS S- II has bee npro videdb y the Alf red P. Sloan Foundation , the Pa rti cipat ingInstituti ons , theNat i onal S cience Foun da tio n , theU . S. De pa rtment ofE n erg y, th eNati onal Ae ronautics and Spac e Ad ministration, th e Ja p a ne seM on b uka ga k ush o , and the Max Pl anck Socie ty , a nd the Hig h erEd ucation Fundin g Cou n cil for England. The SDSS W eb s i t e i s http://w ww.sdss. org/. The SDSSi smanag edby the A str ophys ical R e sea rch C onsort iu m (ARC ) for t he Parti cipating Institutions.The Pa rtici pat ing Insti tut i ons are theAmer ican Museu m o f N atura l H i story , As t ro phy s icalInst i tute Pots d am , U n i ve rsity of Ba s e l , U niver sit y of Ca mbri dge, Case Western Reserve Univer sity , The Un i vers it y of Chicago,Dre xe l Universi ty , Fermilab, the Ins ti t ute f or Adv ancedStudy,t h eJ apan P arti cip ation Gro up, T h e Johns H op k ins Un iver si ty, th e Join t Ins t i tute for Nuclear Astr o p hysic s , t he Ka vl i Insti t utefor Part

_i> -19$)_galaxies, as_summarized in Fig. \[fig14\]. 3._ Although_the_migration takes_place_also in loose_structures it is_more efficient in denser_environments, such as_clusters_of galaxies (“Nurture”). 4. Quenching of the star formation is induced in clusters by_fast_gas-ablation mechanism,_namely_ram-pressure. 5._ Post-Star-Burst galaxies appear the_tracers of the migration taking_place in_dense environments. ![The $g-i$ color versus $i$-band absolute magnitude_relation_for late-type galaxies_(blue contours) and for early-type galaxies (red contours). The_arrow marks the effect of Nurture_at $z$=0 in_dense_environments._Low-luminosity blue galaxies have_their star formation truncated as a_result of gas stripping. \[fig14\]](fig14.ps){width="8.0cm"} We thank_Roberto Decarli for help in querying the_SDSS archive and Veronique Buat, Gabriella_De Lucia, Michele Fumagalli, Bianca_Poggianti and_Marco Scodeggio for helpful discussions._We are grateful_to Paolo_Franzetti and Alessandro_Donati for their contribution to GOLDMine,_the Galaxy On_Line Database extensively used in this_work_(http://goldmine.mib.infn.it). We acknowledge_the_constructive_criticism from_an unknown referee._The_present study_could_not be conceived without the DR7_of_SDSS. Funding for the Sloan Digital Sky_Survey (SDSS) and SDSS-II_has_been provided by the_Alfred P. Sloan Foundation, the_Participating Institutions, the National Science Foundation,_the U.S._Department of_Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, and_the Max Planck Society, and the_Higher Education Funding Council_for England._The_SDSS Web site_is_http://www.sdss.org/. The_SDSS is managed by the Astrophysical Research_Consortium (ARC)_for the Participating Institutions. The Participating_Institutions are the American_Museum_of Natural History, Astrophysical Institute Potsdam,_University of Basel, University of Cambridge,_Case Western Reserve University, The_University_of_Chicago, Drexel University, Fermilab, the_Institute for Advanced Study, the Japan_Participation Group, The_Johns Hopkins University, the Joint Institute for_Nuclear_Astrophysics, the Kavli Institute for Part

,K'}^{\chi}) &= \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K'/k)}(-u_{{\mathfrak{q}}})\biggr) \cdot \pi(\tilde{L}_{p,K'}^{\chi}) \\ &= \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K'/k)}(-u_{{\mathfrak{q}}}) \prod_{{\mathfrak{q}}\in S_{K'} \setminus S_{K}}(1-u_{{\mathfrak{q}}}^{-1})\biggr) \cdot \tilde{L}_{p,K}^{\chi} \\ &= \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K/k)}(-u_{{\mathfrak{q}}}) \prod_{{\mathfrak{q}}\in S_{K'} \setminus S_{K}} P_{{\mathfrak{q}}}({\rm Frob}_{{\mathfrak{q}}}^{-1}) \biggr)\cdot \tilde{L}_{p,K}^{\chi} \\ &= \biggl(\prod_{{\mathfrak{q}}\in S_{K'} \setminus S_{K}} P_{{\mathfrak{q}}}({\rm Frob}_{{\mathfrak{q}}}^{-1})\biggr) \cdot L_{p,K}^{\chi}. \end{aligned}$$ Iwasawa modules --------------- In this section, we will introduce several Iwasawa modules and recall their important properties. We will freely use the facts in Appendix \[subsec:selmer-str\]. For a topological ${\mathbb{Z}}_{p}$-module $M$, let $$M^{\vee} := \operatorname{Hom}_{\rm cont}(M, {\mathbb{Q}}_{p}/{\mathbb{Z}}_{p})$$ denote the Pontryagin dual of $M$. For a field $K \in \Omega$, we write $M_{KL,\infty}$ for the maximal $p$-ramified pro-$p$ abelian extension of $KL(\mu_{p^{\infty}})$ and set $$\begin{aligned} X_{K}^{\chi} := e_{\chi}\left({\mathcal{O}}\otimes_{{\mathbb{Z}}_{p}} \operatorname{Gal}(M_{KL,\infty}/KL(\mu_{p^{\infty}}))\right). \end{aligned}$$ We note that we have the canonical isomorphism $$H^{1}_{{\mathcal{F}}_{\rm str}^{*}}(k, {\mathbb{T}}_{K}^{\vee}(1))^{\vee} \cong X

, K'}^{\chi }) & = \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K'/k)}(-u_{{\mathfrak{q}}})\biggr) \cdot \pi(\tilde{L}_{p, K'}^{\chi }) \\ & = \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K'/k)}(-u_{{\mathfrak{q } } }) \prod_{{\mathfrak{q}}\in S_{K' } \setminus S_{K}}(1 - u_{{\mathfrak{q}}}^{-1})\biggr) \cdot \tilde{L}_{p, K}^{\chi } \\ & = \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K / k)}(-u_{{\mathfrak{q } } }) \prod_{{\mathfrak{q}}\in S_{K' } \setminus S_{K } } P_{{\mathfrak{q}}}({\rm Frob}_{{\mathfrak{q}}}^{-1 }) \biggr)\cdot \tilde{L}_{p, K}^{\chi } \\ & = \biggl(\prod_{{\mathfrak{q}}\in S_{K' } \setminus S_{K } } P_{{\mathfrak{q}}}({\rm Frob}_{{\mathfrak{q}}}^{-1})\biggr) \cdot L_{p, K}^{\chi }. \end{aligned}$$ Iwasawa modules --------------- In this section, we will introduce several Iwasawa module and echo their important properties. We will freely use the fact in Appendix   \[subsec: selmer - str\ ]. For a topological $ { \mathbb{Z}}_{p}$-module $ M$, let $ $ M^{\vee }: = \operatorname{Hom}_{\rm cont}(M, { \mathbb{Q}}_{p}/{\mathbb{Z}}_{p})$$ denote the Pontryagin dual of $ M$. For a playing field $ K \in \Omega$, we compose $ M_{KL,\infty}$ for the maximal $ p$-ramified pro-$p$ abelian extension of $ KL(\mu_{p^{\infty}})$ and set $ $ \begin{aligned } X_{K}^{\chi }: = e_{\chi}\left({\mathcal{O}}\otimes_{{\mathbb{Z}}_{p } } \operatorname{Gal}(M_{KL,\infty}/KL(\mu_{p^{\infty}}))\right). \end{aligned}$$ We note that we have the canonic isomorphism $ $ H^{1}_{{\mathcal{F}}_{\rm str}^{*}}(k, { \mathbb{T}}_{K}^{\vee}(1))^{\vee } \cong X

,K'}^{\chl}) &= \biggl(\prod_{{\mathfrak{q}}\in R_{\rm ram}(K'/k)}(-u_{{\mathftaj{q}}})\biggc) \cdot \li(\tilde{L}_{o,K'}^{\chi}) \\ &= \biggl(\prod_{{\mathfrak{q}}\in S_{\cm rqm}(K'/k)}(-u_{{\nathfrak{q}}}) \prod_{{\mathfrak{d}}\in S_{K'} \senminus S_{K}}(1-y_{{\matifrak{q}}}^{-1})\biggr) \cdot \tilde{L}_{p,K}^{\chi} \\ &= \bjngl(\prmv_{{\mathfrak{q}}\in S_{\rk ram}(K/k)}(-u_{{\madhfrak{q}}}) \prod_{{\madhwrck{q}}\in S_{K'} \setminus S_{K}} P_{{\mathfrak{q}}}({\rm Frjb}_{{\mathftan{q}}}^{-1}) \biggr)\cdot \tylde{K}_{[,K}^{\chj} \\ &= \biggl(\prod_{{\mathfrak{q}}\in S_{K'} \setminua S_{K}} P_{{\methfrak{q}}}({\rm Frob}_{{\kathfrak{q}}}^{-1})\biggr) \cdot L_{p,K}^{\chi}. \end{wligned}$$ Iwasawa modkles --------------- In this seceuon, we will kntroduce several Iwaszwa modules and recall their imoortaut propertigs. Wf will freelb use nhe facts in Appendix \[vubsec:srlmer-str\]. For a bopolmgixal ${\mathbb{Z}}_{p}$-module $M$, net $$M^{\vee} := \operatorgame{Hom}_{\rm cknt}(M, {\mathbb{Q}}_{p}/{\mathvb{Z}}_{p})$$ detote the Pongryzgmn sual ov $M$. Hor a field $K \in \Omega$, we write $M_{KL,\infty}$ gow the maximal $l$-ramifyeq pro-$p$ abelian extension of $KL(\mu_{p^{\infty}})$ dnd set $$\begin{aligned} X_{K}^{\chi} := e_{\chi}\left({\mathcal{O}}\otimgs_{{\mathbb{Z}}_{p}} \operatorname{Gal}(M_{KL,\infty}/KL(\mu_{p^{\infty}}))\right). \end{aligneg}$$ We ioge uhqt we yage the canonical isomorphism $$H^{1}_{{\mathcal{F}}_{\rm str}^{*}}(k, {\jauhbn{T}}_{K}^{\vee}(1))^{\vee} \cong X

,K'}^{\chi}) &= \biggl(\prod_{{\mathfrak{q}}\in S_{\rm ram}(K'/k)}(-u_{{\mathfrak{q}}})\biggr) \cdot \pi(\tilde{L}_{p,K'}^{\chi}) \biggl(\prod_{{\mathfrak{q}}\in ram}(K'/k)}(-u_{{\mathfrak{q}}}) \prod_{{\mathfrak{q}}\in \setminus S_{K}}(1-u_{{\mathfrak{q}}}^{-1})\biggr) \cdot ram}(K/k)}(-u_{{\mathfrak{q}}}) S_{K'} \setminus S_{K}} Frob}_{{\mathfrak{q}}}^{-1}) \biggr)\cdot \tilde{L}_{p,K}^{\chi} &= \biggl(\prod_{{\mathfrak{q}}\in S_{K'} \setminus S_{K}} P_{{\mathfrak{q}}}({\rm \cdot L_{p,K}^{\chi}. \end{aligned}$$ Iwasawa modules --------------- In this section, we will introduce several modules and recall their important properties. We will freely use the facts in \[subsec:selmer-str\]. a ${\mathbb{Z}}_{p}$-module let $$M^{\vee} := \operatorname{Hom}_{\rm cont}(M, {\mathbb{Q}}_{p}/{\mathbb{Z}}_{p})$$ denote the Pontryagin dual of $M$. For a field $K \Omega$, we write $M_{KL,\infty}$ for the maximal $p$-ramified abelian extension of $KL(\mu_{p^{\infty}})$ set $$\begin{aligned} X_{K}^{\chi} := e_{\chi}\left({\mathcal{O}}\otimes_{{\mathbb{Z}}_{p}} \end{aligned}$$ note that have canonical $$H^{1}_{{\mathcal{F}}_{\rm str}^{*}}(k, {\mathbb{T}}_{K}^{\vee}(1))^{\vee} X

,K'}^{\chi}) &= \biggl(\prod_{{\mathfrak{q}}\in S_{\Rm ram}(K'/k)}(-u_{{\maThfraK{q}}})\bIggR) \cDot \pI(\tilDe{L}_{p,K'}^{\chi}) \\ &= \biggl(\pROd_{{\maThfrak{q}}\in S_{\rm ram}(K'/k)}(-u_{{\mathFrak{q}}}) \PrOD_{{\matHFrAk{q}}\in s_{K'} \setmiNUs s_{k}}(1-U_{{\maThFrAk{q}}}^{-1})\BiGGr) \Cdot \tIldE{L}_{p,K}^{\chi} \\ &= \Biggl(\prod_{{\mAthFrAk{q}}\in S_{\rm ram}(K/K)}(-U_{{\mAthfrak{q}}}) \prOd_{{\mAthfrak{q}}\in S_{K'} \SetMinus S_{k}} P_{{\MatHFrak{q}}}({\Rm FRob}_{{\maThfrak{Q}}}^{-1}) \Biggr)\cDot \tilde{L}_{P,K}^{\CHi} \\ &= \biggL(\Prod_{{\matHFRaK{q}}\in s_{K'} \setminus S_{K}} P_{{\mathFRaK{Q}}}({\rm Frob}_{{\mathfraK{q}}}^{-1})\biggR) \cDOt l_{P,k}^{\chI}. \enD{aligned}$$ IwAsAwa moDUles --------------- In tHIs SECTioN, We will introduCe several IwASawA modulEs And REcall tHeir iMpORtaNt propertieS. We wIll freely Use the FActs in APPendix \[sUbsec:sElmEr-sTr\]. FoR A tOpOloGiCAl ${\mAThBb{Z}}_{P}$-ModUle $M$, let $$M^{\VeE} := \oPeratOrnaME{hOM}_{\rm cOnt}(m, {\matHbb{Q}}_{p}/{\Mathbb{Z}}_{p})$$ denotE thE PonTRyaGin duAl of $M$. for a FiEld $K \iN \Omega$, We wriTe $m_{KL,\infty}$ for the mAximAl $p$-ramifiEd pRo-$P$ abElIan exTEnsion Of $Kl(\mu_{P^{\infty}})$ aNd set $$\beGIn{aLiGNED} X_{k}^{\chi} := e_{\chi}\left({\mathcaL{O}}\OTImEs_{{\mathbb{z}}_{p}} \operAToRnAMe{Gal}(M_{KL,\InFty}/kL(\mu_{P^{\INfty}}))\rIght). \ENd{Aligned}$$ WE note tHAt We Have the CaNonicaL iSomOrpHism $$H^{1}_{{\MAthcAl{F}}_{\rm sTr}^{*}}(k, {\mathbB{T}}_{K}^{\veE}(1))^{\Vee} \cong X

,K'}^{\chi}) &= \biggl(\pr od_{{\math frak{ q}} \in S _{\r m ra m}(K'/k)}(-u_{ { \mat hfrak{q}}})\biggr) \cd ot \p i( \ tild e {L }_{p, K'}^{\c h i} ) \\&= \ big gl ( \p rod_{ {\m athfrak {q}}\in S_ {\r mram}(K'/k)}( - u_ {{\mathfra k{q }}}) \prod_{ {\m athfra k{ q}} \ in S_ {K' } \se tminus S_{K}} (1-u_{{\m at h frak{q } }}^{-1} ) \ bi ggr) \cdot \tilde{L}_ { p, K }^{\chi} \\ &= \bigg l( \ pr o d _{{ \ma thfrak{q}} \i n S_{ \ rm ram} ( K/ k ) } (-u _ {{\mathfrak{q }}}) \prod_ { {\m athfra k{ q}} \ in S_{ K'} \ se t min us S_{K}} P _{{\ mathfrak{ q}}}({ \ rm Frob } _{{\mat hfrak{ q}} }^{ -1}) \b ig gr) \c d ot\ ti lde { L}_ {p,K}^{\ ch i} \\ & = \ b i g g l(\p rod _{{\ mathf rak{q}}\in S_ {K' } \s e tmi nus S _{K}} P_{ {\ mathf rak{q} }}({\ rm Frob}_{{\mathf rak{ q}}}^{-1} )\b ig gr) \ cdotL _{p,K} ^{\ chi }. \end {aligne d }$$ I w a sa wa modules ------- -- - - -- -- In t his se c ti on , we will i ntr oduc e sever al I w as awa modu les an d r ec all the ir impor ta ntpro perti e s. W e will freelyuse t h e facts in App e ndix \[subsec : se l m er - str\ ]. For a topo logi c al $ {\ma t hb b{Z } }_{p} $-mod ul e $ M $, let $$M^{\vee} : =\opera torna me{Hom}_{\rmcont}(M, { \ m a thbb{Q}} _{p} / {\ m athbb{Z}}_{p}) $$ de note the P o ntryagin dual of $M$. For a f i e ld $K \i n \ Ome ga$ , w e wr ite $M_{KL,\i n f ty}$ f or themax imal $p $-r ami fie d p ro -$p$ abel ian exte ns io nof $K L(\mu _ {p^{\inf ty }}) $and set$ $\begi n{ali gned }X_ { K}^ {\chi}: =e _ {\ch i} \l eft( {\m at hcal{ O}}\ o tim es_{{\m athbb{Z}} _{p } } \o pe ra torname {Gal}(M_{KL,\ in fty}/KL(\m u_ {p^ {\inft y } }))\righ t). \end{aligned}$$ Wen ote tha t w e hav e th e canonic alisomor phi s m $$H^ {1}_{{ \math ca l{F } } _{\rm s tr }^{ *} }(k, {\mat h b b{T }}_{K }^ {\ve e}(1))^ {\vee} \cong X

,K'}^{\chi}) &=_\biggl(\prod_{{\mathfrak{q}}\in S_{\rm_ram}(K'/k)}(-u_{{\mathfrak{q}}})\biggr) \cdot \pi(\tilde{L}_{p,K'}^{\chi}) \\ &= \biggl(\prod_{{\mathfrak{q}}\in_S_{\rm ram}(K'/k)}(-u_{{\mathfrak{q}}})_\prod_{{\mathfrak{q}}\in_S_{K'} \setminus_S_{K}}(1-u_{{\mathfrak{q}}}^{-1})\biggr)_\cdot \tilde{L}_{p,K}^{\chi} \\ &= \biggl(\prod_{{\mathfrak{q}}\in_S_{\rm ram}(K/k)}(-u_{{\mathfrak{q}}}) \prod_{{\mathfrak{q}}\in_S_{K'} \setminus S_{K}} P_{{\mathfrak{q}}}({\rm_Frob}_{{\mathfrak{q}}}^{-1}) \biggr)\cdot \tilde{L}_{p,K}^{\chi} \\ &=__\biggl(\prod_{{\mathfrak{q}}\in S_{K'} \setminus S_{K}} P_{{\mathfrak{q}}}({\rm Frob}_{{\mathfrak{q}}}^{-1})\biggr) \cdot L_{p,K}^{\chi}. \end{aligned}$$ Iwasawa modules --------------- In this section, we will_introduce_several Iwasawa_modules_and_recall their important properties. We_will freely use the facts_in Appendix \[subsec:selmer-str\]. For_a topological ${\mathbb{Z}}_{p}$-module $M$, let $$M^{\vee} := \operatorname{Hom}_{\rm_cont}(M,_{\mathbb{Q}}_{p}/{\mathbb{Z}}_{p})$$ denote the_Pontryagin dual of $M$. For a field $K \in \Omega$,_we write $M_{KL,\infty}$ for the maximal_$p$-ramified pro-$p$ abelian_extension_of_$KL(\mu_{p^{\infty}})$ and set $$\begin{aligned} X_{K}^{\chi}_:= e_{\chi}\left({\mathcal{O}}\otimes_{{\mathbb{Z}}_{p}} \operatorname{Gal}(M_{KL,\infty}/KL(\mu_{p^{\infty}}))\right). \end{aligned}$$ We note_that we have the canonical isomorphism_$$H^{1}_{{\mathcal{F}}_{\rm str}^{*}}(k, {\mathbb{T}}_{K}^{\vee}(1))^{\vee} \cong X

\label{k-matrix}\end{aligned}$$ where $g^{(p)}_{i}$ are real “couplings" for pole $p$ in channel $i$, the $m_p$ are real, and $\gamma_{ij}^{(n)}$ form constant real symmetric matrices. The presence of poles in the $K$-matrix does not guarantee that the $t$-matrix will have poles close to the real-$s$ axis, but including a $K$-matrix pole is often an efficient way to describe a $t$-matrix pole if one needs to be present. Another alternative is to parameterize the inverse of the $K$-matrix as a symmetric matrix of polynomials, $$\begin{aligned} K^{-1}_{ij}(s) = \sum_{n=0}^{N_{ij}} c_{ij}^{(n)} s^n, \label{k-poly-inv}\end{aligned}$$ with $c_{ij}^{(n)}$ being real parameters. $\pi K$, $\eta K$ coupled-channel scattering in isospin–1/2 {#sec_one_half} =========================================================== ![$\vec{P}=[000]$ $A_1^+$ spectrum. The data points are the energies obtained from variational analysis of a correlation matrix featuring up to 8 “single-meson" and up to 6 “meson-meson" operators at $L/a_s = 16,\, 20,\, 24$. The red bands are the $\pi K$ non–interacting level positions, whilst the green bands represent the $\eta K$ non–interacting level positions (the width of the bands follows from the uncertainty on the meson masses). The dashed grey line shows the $\eta' K$ threshold.[]{data-label="spectrum_I1o2_P000_A1"}](fig4.pdf){width="48.00000%"} Utilizing the methods described in Section \[sec\_spec\_determinations\] we obtain matrices of correlation functions in a large number of irreps with $|\vec{P}|^2 \le 4$. Each of these are analyzed independently using the variational method and the energy levels obtained potentially provide information on the partial-waves subduced into that irrep. We begin by considering the $A_1^+$ irrep at rest, which is likely to be dominated by $\ell = 0$ at low energies, with the next lowest partial-wave, $\ell=4$, being heavily suppressed by the angular-momentum

\label{k - matrix}\end{aligned}$$ where $ g^{(p)}_{i}$ are real “ couplings " for pole $ p$ in channel $ i$, the $ m_p$ are veridical, and $ \gamma_{ij}^{(n)}$ imprint constant real symmetric matrix. The presence of poles in the $ K$-matrix does not undertake that the $ t$-matrix will have poles close to the real-$s$ bloc, but include a $ K$-matrix pole is much an efficient way to trace a $ t$-matrix terminal if one needs to be present. Another alternative is to parameterize the inverse of the $ K$-matrix as a symmetric matrix of polynomials, $ $ \begin{aligned } K^{-1}_{ij}(s) = \sum_{n=0}^{N_{ij } } c_{ij}^{(n) } s^n, \label{k - poly - inv}\end{aligned}$$ with $ c_{ij}^{(n)}$ being real parameter. $ \pi K$, $ \eta K$ coupled - channel scattering in isospin–1/2 { # sec_one_half } = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = ! [ $ \vec{P}=[000]$ $ A_1^+$ spectrum. The datum points are the energies obtain from variational analysis of a correlation coefficient matrix featuring up to 8 “ unmarried - meson " and up to 6 “ meson - meson " operators at $ L / a_s = 16,\, 20,\, 24$. The red bands are the $ \pi K$ non – interacting grade positions, whilst the green bands represent the $ \eta K$ non – interacting level positions (the width of the band follows from the uncertainty on the meson masses). The dashed grey line shows the $ \eta' K$ threshold.[]{data - label="spectrum_I1o2_P000_A1"}](fig4.pdf){width="48.00000% " } Utilizing the methods described in Section \[sec\_spec\_determinations\ ] we receive matrices of correlation routine in a large number of irreps with $ |\vec{P}|^2 \le 4$. Each of these are analyzed independently using the variational method acting and the energy level obtained potentially provide data on the partial - waves subduced into that irrep. We begin by considering the $ A_1^+$ irrep at rest, which is likely to be dominated by $ \ell = 0 $ at low energies, with the next lowest partial - wave, $ \ell=4 $, being heavy suppressed by the angular - momentum

\labfl{k-matrix}\end{aligned}$$ whert $g^{(p)}_{i}$ are real “coopoings" hor pols $p$ in cfannel $i$, the $m_p$ are real, and $\gqmma_{ih}^{(n)}$ form constant real rymmetric matricew. Tht presence of polxa in thc $K$-mafvix dmxs not guaranteg that the $t$-katrix will haee plles close to the real-$s$ axis, but insluding a $K$-matrix pole ys ogeen zn efficient way to describe a $t$-mafrix pone if one neecs to be present. Another alhernwtive is to parameherize the unvewwe of the $K$-mxtrix as a symmetric mztrix of polynomials, $$\begin{aligndd} K^{-1}_{ij}(x) = \sum_{n=0}^{N_{ij}} c_{uj}^{(n)} v^n, \label{k-polb-inv}\enq{aligned}$$ witm $c_{ij}^{(n)}$ teing rral parameters. $\ii K$, $\xta J$ coupled-channel scatvering in isospin–1/2 {#ses_one_half} =========================================================== ![$\vac{'}=[000]$ $A_1^+$ spectrum. The data piints are the wnefgits pbfained frkm variatiknal analysus of a correlation mwntix featurinf up tj 8 “single-meson" and up to 6 “meson-meson" operdtods at $L/a_s = 16,\, 20,\, 24$. The red bqnds are the $\pi K$ non–lnteractigg level positions, whilst the green bands represett thx $\dta K$ vin–lnteracting level positions (the width of the fzncs follows from bhe uncertainty on tje ieson masses). Jhe dasksd grey line shows tje $\eta' H$ thrwshold.[]{datw-labrl="spectrum_I1o2_P000_A1"}](fig4.pdf){width="48.00000%"} Utulizing the iwthods described iu Section \[see\_spec\_dgtermimations\] we obtain matriees of correlatioj functiohr in a large numcer ox irreps with $|\vec{P}|^2 \le 4$. Eacr of thesx are analyzdd imdepenqently usijg thc variational methof and tve energy pevels obtained potentially protmde informatipn on the parcial-wayes subduced ineo that irrep. Ce begin by cunsidering the $A_1^+$ mrrep at rese, which is lingly to be dommnated by $\ell = 0$ ar low evdrgies, with thr next loctst partiao-wave, $\ell=4$, being heevili auppressed by tkt abgular-momentum

\label{k-matrix}\end{aligned}$$ where $g^{(p)}_{i}$ are real “couplings" for in $i$, the are real, and matrices. presence of poles the $K$-matrix does guarantee that the $t$-matrix will have close to the real-$s$ axis, but including a $K$-matrix pole is often an way to describe a $t$-matrix pole if one needs to be present. Another is parameterize inverse the $K$-matrix as a symmetric matrix of polynomials, $$\begin{aligned} K^{-1}_{ij}(s) = \sum_{n=0}^{N_{ij}} c_{ij}^{(n)} s^n, \label{k-poly-inv}\end{aligned}$$ with being real parameters. $\pi K$, $\eta K$ coupled-channel in isospin–1/2 {#sec_one_half} =========================================================== $A_1^+$ spectrum. The data points the obtained from analysis a matrix featuring up 8 “single-meson" and up to 6 “meson-meson" operators at $L/a_s = 16,\, 20,\, 24$. The red bands the $\pi level positions, the bands the $\eta K$ positions (the width of the bands uncertainty on the meson masses). The dashed grey shows the K$ threshold.[]{data-label="spectrum_I1o2_P000_A1"}](fig4.pdf){width="48.00000%"} Utilizing the methods described Section \[sec\_spec\_determinations\] we obtain matrices of correlation functions a large number of irreps with $|\vec{P}|^2 \le 4$. Each of these are analyzed independently variational method and the levels obtained potentially information the subduced that irrep. begin by considering the $A_1^+$ irrep at rest, which is likely be dominated by $\ell = 0$ at low energies, with lowest $\ell=4$, being heavily by the angular-momentum

\label{k-matrix}\end{aligned}$$ wheRe $g^{(p)}_{i}$ are reAl “couPliNgs" FoR polE $p$ in Channel $i$, the $m_p$ aRE reaL, and $\gamma_{ij}^{(n)}$ form constaNt reaL sYMmetRIc MatriCes. The pREsENCe oF pOlEs iN tHE $K$-MatriX doEs not guArantee thaT thE $t$-Matrix will haVE pOles close tO thE real-$s$ axis, buT inCludinG a $k$-maTRix poLe iS ofteN an effICient wAy to descrIbE A $t$-matrIX pole if ONE nEeds To be present. AnotheR AlTErnative is to paRameteRiZE tHE InvErsE of the $K$-matRiX as a sYMmetric MAtRIX Of pOLynomials, $$\begiN{aligned} K^{-1}_{ij}(S) = \Sum_{N=0}^{N_{ij}} c_{iJ}^{(n)} S^n, \lABel{k-poLy-inv}\EnD{AliGned}$$ with $c_{ij}^{(N)}$ beiNg real parAmeterS. $\Pi K$, $\eta K$ COupled-cHannel ScaTteRing IN iSoSpiN–1/2 {#sEC_onE_HaLf} =========================================================== ![$\vEC{P}=[000]$ $A_1^+$ Spectrum. thE dAta poInts ARE THe enErgIes oBtainEd from variatiOnaL anaLYsiS of a cOrrelAtioN mAtrix FeaturIng up To 8 “Single-meson" and uP to 6 “mEson-meson" OpeRaTorS aT $L/a_s = 16,\, 20,\, 24$. THE red baNds Are The $\pi K$ nOn–interACtiNg LEVEl Positions, whilst the GrEEN bAnds reprEsent tHE $\eTa k$ Non–interAcTinG levEL PositIons (THe Width of tHe bandS FoLlOws from ThE uncerTaIntY on The meSOn maSses). ThE dashed gRey liNE shows the $\eta' K$ tHReshold.[]{data-laBEl="SPEcTRum_I1O2_P000_A1"}](Fig4.pdf){width="48.00000%"} utilIZing The mEThOds DEscriBed in seCTiON \[sec\_spec\_determinatiOnS\] we obtAin maTrices of correLation funcTIONs in a larGe nuMBeR Of irreps with $|\veC{P}|^2 \le 4$. EAch of these ARe analyzEd indEpendentLy using thE VAriationAl mEthOd aNd tHE EnErgy levels obtAINed pOtEntiallY prOvide inForMatIon On tHe Partial-waVes subduCeD iNtO tHat Irrep. wE begin by CoNsiDeRinG the $A_1^+$ IRrep at Rest, wHich Is LiKEly To be domINaTED by $\eLl = 0$ At Low eNerGiEs, witH the NExt Lowest pArtial-wavE, $\elL=4$, BeinG hEaVily supPressed by the aNgUlar-momentUm

\label{k-matrix}\end{alig ned}$$ whe re $g ^{( p)} _{ i}$arereal “coupling s " fo r pole $p$ in channel$i$,th e $m_ p $are r eal, an d $ \ g amm a_ {i j}^ {( n )} $ for m c onstant real symm etr ic matrices. T h epresence o f p oles in the$K$ -matri xdoe s notgua rante e that the $t $-matrixwi l l have poles c l o se tothe real-$s$ axis , b u t including a$K$-ma tr i xp o leisoften an e ff icien t way to de s c r ibe a $t$-matrixpole if one nee ds tobe pr e sent. Anot he r al ternative i s to paramete rize t h e inver s e of th e $K$- mat rix asa s ym met ri c ma t ri x o f po lynomial s, $ $\beg in{a l i g n ed}K^{ -1}_ {ij}( s) = \sum_{n= 0}^ {N_{ i j}} c_{i j}^{( n)}s^ n, \l abel{k -poly -i nv}\end{aligned }$$with $c_{ ij} ^{ (n) }$ bein g realpar ame ters. $\pi K$ , $\ et a K $coupled-channel sc at t e ri ng in is ospin– 1 /2 { # sec_one_ ha lf} === = = ===== ==== = == ======== ====== = == == ======= == ====== == === === ![$ \ vec{ P}=[00 0]$ $A_1 ^+$ s p ectrum. The da t a points aret he e ne r gies ob tained from var i atio nala na lys i s ofa cor re l at i on matrix featuring u p to 8 “sin gle-meson" an d up to 6“ m e son-meso n" o p er a tors at $L/a_s = 16 ,\, 20,\,2 4$. Thered b ands are the $\pi K $ non–in ter act ing le v e lpositions, wh i l st t he greenban ds repr ese ntthe $\ et a K$ non– interact in gle ve l p ositi o ns (thewi dth o f t he ba n ds fol lowsfrom t he unc ertaint y o n theme so n ma sse s) . The das h edgrey li ne showsthe $\et a' K $ thres hold.[]{data- la bel="spect ru m_I 1o2_P0 0 0 _A1"}](f ig4.pdf){width="48.0000 0 %"} Ut ili zingthemethods d esc ribedinS ection \[sec \_spe c\ _de t e rmina t i on s\] w e obtain m a t ric es of c orre lationfunctions in a lar g e n umber of irre pswith $ |\ vec { P} | ^2\l e 4$ . Each of these a re analyze di nd ependently usi ng the va riation al me t hod and the ener gy levels o btai n e d p otentially provide informat i on on th e par tia l-wave ssub duced intot hat irre p. Webe gin by cons id ering th e $A_1^+$ irrep at rest , whic h islik ely to be do m ina ted by $\ ell= 0$ at lo w e ner gies, wi t h the nex t l owe s t par tial - wave, $\e l l= 4$, b ei ng heavilys u p pre ssedbyt he ang ular -momentum

\label{k-matrix}\end{aligned}$$ where_$g^{(p)}_{i}$ are_real “couplings" for pole_$p$ in_channel_$i$, the_$m_p$_are real, and_$\gamma_{ij}^{(n)}$ form constant_real symmetric matrices. The_presence of poles_in_the $K$-matrix does not guarantee that the $t$-matrix will have poles close to the_real-$s$_axis, but_including_a_$K$-matrix pole is often an_efficient way to describe a_$t$-matrix pole_if one needs to be present. Another alternative is_to_parameterize the inverse_of the $K$-matrix as a symmetric matrix of polynomials,_$$\begin{aligned} K^{-1}_{ij}(s) = \sum_{n=0}^{N_{ij}} c_{ij}^{(n)} s^n, \label{k-poly-inv}\end{aligned}$$ with_$c_{ij}^{(n)}$ being real_parameters. $\pi_K$,_$\eta K$ coupled-channel scattering_in isospin–1/2 {#sec_one_half} =========================================================== ![$\vec{P}=[000]$ $A_1^+$ spectrum. The_data points are the energies obtained_from variational analysis of a correlation matrix_featuring up to 8 “single-meson" and_up to 6 “meson-meson" operators_at $L/a_s_= 16,\, 20,\, 24$. The_red bands are_the $\pi_K$ non–interacting level_positions, whilst the green bands represent_the $\eta K$_non–interacting level positions (the width of_the_bands follows from_the_uncertainty_on the_meson masses). The_dashed_grey line_shows_the $\eta' K$ threshold.[]{data-label="spectrum_I1o2_P000_A1"}](fig4.pdf){width="48.00000%"} Utilizing the methods_described_in Section \[sec\_spec\_determinations\] we obtain matrices of_correlation functions in a_large_number of irreps with_$|\vec{P}|^2 \le 4$. Each of_these are analyzed independently using the_variational method_and the_energy levels obtained potentially provide information on the partial-waves subduced into_that irrep. We begin by considering_the $A_1^+$ irrep at_rest, which_is_likely to be_dominated_by $\ell_= 0$ at low energies, with the_next lowest_partial-wave, $\ell=4$, being heavily suppressed by_the angular-momentum

By construction variety $V$ is a point, a curve or a surface. Moreover, when $V$ is a point by Proposition 7.2 in [@Prokhorov-degree] we have $\dim |-K_{W}| \leqslant 34$ which contradicts the estimate from Remark \[remark:inequalities\]. In what follows we will treat separately the cases when $V$ is a curve and a surface. Proof of Theorem \[theorem:main\] in the case when $V$ is a curve {#section:curve} ================================================================= We use the notation and assumptions from Section \[section:reduction\]. From the Leray spectral sequence and Kawamata–Vieweg Vanishing Theorem we deduce that $H^{1}(V, \mathcal{O}_{V})=H^{1}(W, \mathcal{O}_{W})=0$. Thus, we have $V \simeq \mathbb{P}^{1}$ in the present case. General fibre $W_{\eta}$ of the contraction $g: W \longrightarrow V$ is a non-singular del Pezzo surface. For $H \in \mathcal{H}_{W}$ divisor $-(K_{W_{\eta}}+H|_{W_{\eta}})$ is ample by construction and divisor $H|_{W_{\eta}}$ is ample by Remark \[remark:linear-systems\]. This implies that $W_{\eta} \simeq \mathbb{P}^{2}$ or $\mathbb{P}^{1}\times\mathbb{P}^{1}$. Moreover, for $W_{\eta} \simeq \mathbb{P}^{2}$ we have either $H|_{W_{\eta}} \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$ or $\mathcal{O}_{\mathbb{P}^{2}}(2)$. \[theorem:special-scrolls\] In the above notation, if $W_{\eta} \simeq \mathbb{P}^{2}$ and $H|_{W_{\eta}} \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$, then $W$ is a $\mathbb{P}^{2}$-bundle. Moreover, either $W = \mathrm{Proj}\left(\mathcal{O}_{\mathbb{P}^{1}}(5)\oplus\mathcal{O}_{\mathbb{P}^{1}}(2)\oplus\mathcal{O}_{\mathbb{P}^{1}}\right)$ or $\mathrm{Proj}\left(\mathcal{O}_{\

By construction variety $ V$ is a point, a curve or a open. furthermore, when $ V$ is a point by Proposition 7.2 in [ @Prokhorov - degree ] we get $ \dim |-K_{W}| \leqslant 34 $ which contradict the estimate from Remark   \[remark: inequalities\ ]. In what follows we will process separately the cases when $ V$ is a curvature and a surface. Proof of Theorem   \[theorem: main\ ] in the case when $ V$ is a bend { # section: curve } = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = We practice the notation and assumptions from Section   \[section: reduction\ ]. From the Leray spectral sequence and Kawamata – Vieweg Vanishing Theorem we deduce that $ H^{1}(V, \mathcal{O}_{V})=H^{1}(W, \mathcal{O}_{W})=0$. Thus, we suffer $ V \simeq \mathbb{P}^{1}$ in the present case. General fibre $ W_{\eta}$ of the compression $ g: W \longrightarrow V$ is a non - singular del Pezzo open. For $ H \in \mathcal{H}_{W}$ divisor $ -(K_{W_{\eta}}+H|_{W_{\eta}})$ is ample by construction and divisor $ H|_{W_{\eta}}$ is ample by Remark   \[remark: linear - systems\ ]. This imply that $ W_{\eta } \simeq \mathbb{P}^{2}$ or $ \mathbb{P}^{1}\times\mathbb{P}^{1}$. Moreover, for $ W_{\eta } \simeq \mathbb{P}^{2}$ we have either $ H|_{W_{\eta } } \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$ or $ \mathcal{O}_{\mathbb{P}^{2}}(2)$. \[theorem: special - scrolls\ ] In the above notation, if $ W_{\eta } \simeq \mathbb{P}^{2}$ and $ H|_{W_{\eta } } \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$, then $ W$ is a $ \mathbb{P}^{2}$-bundle. Moreover, either $ W = \mathrm{Proj}\left(\mathcal{O}_{\mathbb{P}^{1}}(5)\oplus\mathcal{O}_{\mathbb{P}^{1}}(2)\oplus\mathcal{O}_{\mathbb{P}^{1}}\right)$ or $ \mathrm{Proj}\left(\mathcal{O}_{\

By fonstruction variety $V$ ir a point, a currw or a surfade. Moreoxer, when $V$ is a point by Pro'osirion 7.2 in [@Prokhorov-degree] we have $\dim |-K_{W}| \leqsoant 34$ which convdadicts the eabimatz hrom Remark \[rematk:inequalitias\]. In what follmwr ce will treat separately the cases wren $V$ ix w curve and a furfsse. Prkof of Theorem \[theorem:main\] in the czse whei $V$ is a curve {#xection:curve} ================================================================= We use the notwtioj and assumptions vrom Sectiob \[seceuon:reduction\]. From the Leray spectrzl sequence and Kawamata–Vieweg Xanisking Theoren qe fgduce that $H^{1}(T, \mathcwl{O}_{V})=H^{1}(W, \mathcal{O}_{W})=0$. Thuv, we hafe $V \simeq \mathnb{P}^{1}$ ii thw present case. General fibre $W_{\eta}$ of the contracthou $g: W \longrightarrow C$ us a ton-shnguuqr ael Pxzzk surfwce. For $H \in \jathcal{H}_{W}$ duvisor $-(K_{W_{\eta}}+H|_{W_{\eta}})$ ix wnple by constductiog wnd divisor $H|_{W_{\eta}}$ is ample by Remark \[remdrk:minear-systems\]. This implues that $W_{\eta} \simeq \mwthbb{P}^{2}$ or $\mathbb{P}^{1}\times\mathbb{P}^{1}$. Moreover, for $W_{\eta} \simeq \mathtb{P}^{2}$ wx favt clthef $H|_{A_{\eta}} \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$ or $\mathcal{O}_{\mathbb{P}^{2}}(2)$. \[egeprvm:special-scrolls\] Ln the above notatooj, os $W_{\eta} \simeq \oathbb{'}^{2}$ ahd $H|_{W_{\eta}} \simeq \matjcal{O}_{\majhbb{P}^{2}}(1)$, rhen $W$ is a $\msthbb{P}^{2}$-bundle. Moreover, eithee $W = \mathrm{Pgoj}\lwft(\mathcal{O}_{\mathbb{P}^{1}}(5)\lplus\mathcau{O}_{\majhbb{P}^{1}}(2)\ollus\mathcal{O}_{\mathbb{P}^{1}}\righc)$ or $\mzthrm{Proj}\levt(\mathcal{K}_{\

By construction variety $V$ is a point, or surface. Moreover, $V$ is a [@Prokhorov-degree] have $\dim |-K_{W}| 34$ which contradicts estimate from Remark \[remark:inequalities\]. In what we will treat separately the cases when $V$ is a curve and a Proof of Theorem \[theorem:main\] in the case when $V$ is a curve {#section:curve} We the and from Section \[section:reduction\]. From the Leray spectral sequence and Kawamata–Vieweg Vanishing Theorem we deduce that $H^{1}(V, \mathcal{O}_{W})=0$. Thus, we have $V \simeq \mathbb{P}^{1}$ in present case. General fibre of the contraction $g: W V$ a non-singular Pezzo For \in \mathcal{H}_{W}$ divisor is ample by construction and divisor $H|_{W_{\eta}}$ is ample by Remark \[remark:linear-systems\]. This implies that $W_{\eta} \simeq or $\mathbb{P}^{1}\times\mathbb{P}^{1}$. $W_{\eta} \simeq we either \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$ or In the above notation, if $W_{\eta} $H|_{W_{\eta}} \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$, then $W$ is a $\mathbb{P}^{2}$-bundle. either $W \mathrm{Proj}\left(\mathcal{O}_{\mathbb{P}^{1}}(5)\oplus\mathcal{O}_{\mathbb{P}^{1}}(2)\oplus\mathcal{O}_{\mathbb{P}^{1}}\right)$ or $\mathrm{Proj}\left(\mathcal{O}_{\

By construction variety $V$ is a Point, a curvE or a sUrfAce. moReovEr, whEn $V$ is a point by PROposItion 7.2 in [@Prokhorov-degree] We havE $\dIM |-K_{W}| \lEQsLant 34$ wHich conTRaDICts ThE eStiMaTE fRom ReMarK \[remark:InequalitiEs\]. IN wHat follows we WIlL treat sepaRatEly the cases wHen $v$ is a cuRvE anD A surfAce. proof Of TheoREm \[theoRem:main\] in ThE Case whEN $V$ is a cuRVE {#sEctiOn:curve} ================================================================= We use the noTAtIOn and assumptioNs from seCTiON \[SecTioN:reduction\]. frOm the lEray speCTrAL SEquENce and KawamatA–Vieweg VaniSHinG TheorEm We dEDuce thAt $H^{1}(V, \mAtHCal{o}_{V})=H^{1}(W, \mathcal{o}_{W})=0$. ThUs, we have $V \Simeq \mAThbb{P}^{1}$ in THe preseNt case. genEraL fibRE $W_{\EtA}$ of ThE ConTRaCtiON $g: W \LongrighTaRrOw V$ is A non-SINGUlar Del pezzO surfAce. For $H \in \mathCal{h}_{W}$ diVIsoR $-(K_{W_{\etA}}+H|_{W_{\etA}})$ is aMpLe by cOnstruCtion AnD divisor $H|_{W_{\eta}}$ is AmplE by Remark \[RemArK:liNeAr-sysTEms\]. ThiS imPliEs that $W_{\Eta} \simeQ \MatHbB{p}^{2}$ OR $\mAthbb{P}^{1}\times\mathbb{P}^{1}$. moREOvEr, for $W_{\etA} \simeq \MAtHbB{p}^{2}$ we have eItHer $h|_{W_{\etA}} \SImeq \mAthcAL{O}_{\Mathbb{P}^{2}}(1)$ oR $\mathcAL{O}_{\MaThbb{P}^{2}}(2)$. \[thEoRem:speCiAl-sCroLls\] In THe abOve notAtion, if $W_{\Eta} \siMEq \mathbb{P}^{2}$ and $H|_{W_{\ETa}} \simeq \mathcaL{o}_{\mATHbB{p}^{2}}(1)$, theN $W$ iS a $\mathbb{P}^{2}$-buNdle. mOreoVer, eIThEr $W = \MAthrm{proj}\lEfT(\MaTHcal{O}_{\mathbb{P}^{1}}(5)\oplus\maThCal{O}_{\maThbb{P}^{1}}(2)\Oplus\mathcal{O}_{\Mathbb{P}^{1}}\rigHT)$ OR $\mathrm{PRoj}\lEFt(\MAthcal{O}_{\

By construction variety $ V$ is a po int,a c urv eor a sur face. Moreover , whe n $V$ is a point by Pr oposi ti o n 7. 2 i n [@P rokhoro v -d e g ree ]we ha ve $\ dim | -K_ {W}| \l eqslant 34 $ w hi ch contradic t sthe estima tefrom Remark\[r emark: in equ a litie s\] . In whatf ollows we willtr e at sep a ratelyt h ecase s when $V$ is a c u rv e and a surface . Pro of of T heo rem  \[theorem :m ain\] in thec as e w hen $V$ is a curv e {#section : cur ve} == == === = ====== ===== == = === =========== ==== ========= ====== = ======= = = We u se the no tat iona nd a ssu mp t ion s f rom Sec tion \[s ec ti on:re duct i o n \ ]. F rom the Lera y spectral se que ncea ndKawam ata–V iewe gVanis hing T heore mwe deduce that$H^{ 1}(V, \ma thc al {O} _{ V})=H ^ {1}(W, \m ath cal{O}_ {W})=0$ . Th us , w ehave $V \simeq \ma th b b {P }^{1}$ i n thep re se n t case. G ene ralf i bre $ W_{\ e ta }$ of th e cont r ac ti on $g:W\longr ig hta rro w V$i s anon-si ngular d el Pe z zo surface. Fo r $H \in \math c al { H }_ { W}$div isor $-(K_{ W_{\ e ta}} +H|_ { W_ {\e t a}})$ is a mp l eb y construction anddi visor$H|_{ W_{\eta}}$ is ample byR e m ark \[re mark : li n ear-systems\]. This implies t h at $W_{\ eta}\simeq \ mathbb{P} ^ { 2}$ or $ \ma thb b{P }^{ 1 } \t imes\mathbb{P } ^ {1}$ .Moreove r,for $W_ {\e ta} \s ime q\mathbb{P }^{2}$ w eha ve e ith er $H | _{W_{\et a} } \ si meq \mat h cal{O} _{\ma thbb {P }^ { 2}} (1)$ or $\ m a thca l{ O} _{\m ath bb {P}^{ 2}}( 2 )$. \[the orem:spec ial - scro ll s\ ] In th e above notat io n, if $W_{ \e ta} \sime q \mathbb{ P}^{2}$ and $H|_{W_{\et a }} \sim eq\math cal{ O}_{\math bb{ P}^{2} }(1 ) $, the n $W$is a$\ mat h b b{P}^ { 2 }$ -bu nd le. Moreov e r , e ither $ W =\mathrm {Proj}\left(\mathc a l{O }_{\mathbb{P} ^{1 }}(5 ) \ op lus \ ma t hca l{ O }_{ \ m athbb{P}^{1}}(2 )\oplus\ma th c al {O}_{\math b b{P }^ {1}}\ri ght)$ o r $\m a thrm{Pr oj}\left( \mathcal{ O} _{\

By construction_variety $V$_is a point, a_curve or_a_surface. Moreover,_when_$V$ is a_point by Proposition_7.2 in [@Prokhorov-degree] we_have $\dim |-K_{W}|_\leqslant_34$ which contradicts the estimate from Remark \[remark:inequalities\]. In what follows we will treat separately the_cases_when $V$_is_a_curve and a surface. Proof of_Theorem \[theorem:main\] in the case when_$V$ is_a curve {#section:curve} ================================================================= We use the notation and assumptions_from_Section \[section:reduction\]. From the_Leray spectral sequence and Kawamata–Vieweg Vanishing Theorem we deduce_that $H^{1}(V, \mathcal{O}_{V})=H^{1}(W, \mathcal{O}_{W})=0$. Thus, we have_$V \simeq \mathbb{P}^{1}$ in_the_present_case. General fibre $W_{\eta}$ of_the contraction $g: W \longrightarrow V$_is a non-singular del Pezzo surface._For $H \in \mathcal{H}_{W}$ divisor $-(K_{W_{\eta}}+H|_{W_{\eta}})$ is_ample by construction and divisor $H|_{W_{\eta}}$_is ample by Remark \[remark:linear-systems\]. This_implies that_$W_{\eta} \simeq \mathbb{P}^{2}$ or $\mathbb{P}^{1}\times\mathbb{P}^{1}$._Moreover, for $W_{\eta}_\simeq \mathbb{P}^{2}$_we have either_$H|_{W_{\eta}} \simeq \mathcal{O}_{\mathbb{P}^{2}}(1)$ or $\mathcal{O}_{\mathbb{P}^{2}}(2)$. \[theorem:special-scrolls\] In_the above notation,_if $W_{\eta} \simeq \mathbb{P}^{2}$ and $H|_{W_{\eta}}_\simeq_\mathcal{O}_{\mathbb{P}^{2}}(1)$, then $W$_is_a_$\mathbb{P}^{2}$-bundle. Moreover,_either $W =_\mathrm{Proj}\left(\mathcal{O}_{\mathbb{P}^{1}}(5)\oplus\mathcal{O}_{\mathbb{P}^{1}}(2)\oplus\mathcal{O}_{\mathbb{P}^{1}}\right)$_or $\mathrm{Proj}\left(\mathcal{O}_{\

$k-1\le 2n/p$, the result is trivial as the right-hand side of is larger than or equal to 1. We will therefore assume that $k-1>2n/p$ or, equivalently, $2^{(k-1)p/n}>4.$ Let $x_1,\dots,x_N$ be a maximal family of elements of $B_X$ with $\|x_i-x_j\|_X>\tau$ for $i\not=j$, where $\tau>0$ is given by $\frac{1+\tau^p/2}{\tau^p/2}=2^{p(k-1)/n}$. Then, by triangle inequality, the sets $x_i+2^{-1/p}\tau B_X$ are mutually disjoint, they are all included in $(1+\tau^p/2)^{1/p}B_X$ and $B_X$ is covered by the union of $x_i+\tau B_X$ over $i=1,\dots,N$. By volume comparison, we get $$N{\operatornamewithlimits{vol}}(2^{-1/p}\tau B_X)\le {\operatornamewithlimits{vol}}[(1+\tau^p/2)^{1/p}B_X]$$ and, therefore, $$N\le \frac{(1+\tau^p/2)^{n/p}}{2^{-n/p}\tau^n}=\Bigl(\frac{1+\tau^p/2}{\tau^p/2}\Bigr)^{n/p}=2^{k-1}.$$ We conclude that $$e_k(id:X\to X)\le \tau=\biggl[\frac{2}{2^{(k-1)p/n}-1}\biggr]^{1/p}\le\biggl[\frac{4}{2^{(k-1)p/n}}\biggr]^{1/p}=4^{1/p}2^{-\frac{k-1}{n}}.\qedhere$$ The behavior of entropy numbers with respect to interpolation of Banach spaces was studied intensively. It is rather easy to show, that they behave well if one of the endpoints is fixed and we refer to [@ET] for details. Surprisingly, it was shown only recently in [@EN], that a general interpolation formula for entropy numbers with both endpoints interpolated is out of reach. We give only a simplified version

$ k-1\le 2n / p$, the result is trivial as the right - hand slope of is large than or equal to 1. We will therefore assume that $ k-1>2n / p$ or, equivalently, $ 2^{(k-1)p / n}>4.$ get $ x_1,\dots, x_N$ be a maximal family of element of $ B_X$ with $ \|x_i - x_j\|_X>\tau$ for $ i\not = j$, where $ \tau>0 $ is impart by $ \frac{1+\tau^p/2}{\tau^p/2}=2^{p(k-1)/n}$. Then, by triangle inequality, the hardening $ x_i+2^{-1 / p}\tau B_X$ are mutually disjoint, they are all included in $ (1+\tau^p/2)^{1 / p}B_X$ and $ B_X$ is covered by the marriage of $ x_i+\tau B_X$ over $ i=1,\dots, N$. By volume comparison, we get $ $ N{\operatornamewithlimits{vol}}(2^{-1 / p}\tau B_X)\le { \operatornamewithlimits{vol}}[(1+\tau^p/2)^{1 / p}B_X]$$ and, consequently, $ $ N\le \frac{(1+\tau^p/2)^{n / p}}{2^{-n / p}\tau^n}=\Bigl(\frac{1+\tau^p/2}{\tau^p/2}\Bigr)^{n / p}=2^{k-1}.$$ We conclude that $ $ e_k(id: X\to X)\le \tau=\biggl[\frac{2}{2^{(k-1)p / n}-1}\biggr]^{1 / p}\le\biggl[\frac{4}{2^{(k-1)p / n}}\biggr]^{1 / p}=4^{1 / p}2^{-\frac{k-1}{n}}.\qedhere$$ The behavior of entropy numbers with respect to interpolation of Banach spaces was analyze intensively. It is rather easy to show, that they act well if one of the endpoints is fixed and we mention to [ @ET ] for details. amazingly, it was shown entirely recently in [ @EN ], that a general interpolation formula for information numbers with both endpoints interpolated is out of reach. We give only a simplified version

$k-1\lf 2n/p$, the result is trivixl as the right-kqnd sive of ia larger than or equal to 1. We will tieredore qssume that $k-1>2n/p$ or, equkvalently, $2^{(k-1)p/n}>4.$ Let $z_1,\dotw,z_N$ be a maejmal family or elekxnts of $B_X$ with $\|x_i-x_j\|_X>\tau$ xor $i\not=j$, whera $\gab>0$ is given by $\frac{1+\tau^p/2}{\tau^p/2}=2^{p(k-1)/n}$. Then, br trianblf inequality, tre stts $x_i+2^{-1/p}\fau B_X$ are mutually disjoint, they zre all included in $(1+\yau^p/2)^{1/p}B_X$ and $B_X$ is covered hy tje union of $x_i+\tau H_X$ over $i=1,\doja,N$. Ft volume comoarison, we get $$N{\operatkrnamewithlimits{vol}}(2^{-1/p}\tau B_X)\le {\opdratoxnamewithlinirs{vln}}[(1+\tau^p/2)^{1/p}B_X]$$ anv, thervfore, $$N\le \frag{(1+\nau^p/2)^{n/p}}{2^{-n/[}\tau^n}=\Bibl(\frac{1+\tau^p/2}{\tau^p/2}\Nigr)^{n/'}=2^{k-1}.$$ Ww conclude that $$e_k(id:X\vo X)\le \tau=\biggl[\frac{2}{2^{(k-1)k/n}-1}\biggr]^{1/p}\le\tiygl[\frac{4}{2^{(k-1)p/n}}\biggr]^{1/p}=4^{1/p}2^{-\frac{j-1}{n}}.\wedhete$$ The behxciof or xntdopy nkmbxrs with reapect to inrerpolation of Banavh wpaces was sthdied yneensively. It is rather easy to show, thau theg behave well if one of the endpoints is fixgd and we wefer to [@ET] for details. Surprisingly, it was shown only fectnbjh ij [@EN], that a general interpolation formula for snuroiy numbers with bjth endpoinys imjerpolated is uut of rezch. We give only a simplisied cersion

$k-1\le 2n/p$, the result is trivial as side is larger or equal to that or, equivalently, $2^{(k-1)p/n}>4.$ $x_1,\dots,x_N$ be a family of elements of $B_X$ with for $i\not=j$, where $\tau>0$ is given by $\frac{1+\tau^p/2}{\tau^p/2}=2^{p(k-1)/n}$. Then, by triangle inequality, the $x_i+2^{-1/p}\tau B_X$ are mutually disjoint, they are all included in $(1+\tau^p/2)^{1/p}B_X$ and $B_X$ covered the of B_X$ over $i=1,\dots,N$. By volume comparison, we get $$N{\operatornamewithlimits{vol}}(2^{-1/p}\tau B_X)\le {\operatornamewithlimits{vol}}[(1+\tau^p/2)^{1/p}B_X]$$ and, therefore, $$N\le \frac{(1+\tau^p/2)^{n/p}}{2^{-n/p}\tau^n}=\Bigl(\frac{1+\tau^p/2}{\tau^p/2}\Bigr)^{n/p}=2^{k-1}.$$ We that $$e_k(id:X\to X)\le \tau=\biggl[\frac{2}{2^{(k-1)p/n}-1}\biggr]^{1/p}\le\biggl[\frac{4}{2^{(k-1)p/n}}\biggr]^{1/p}=4^{1/p}2^{-\frac{k-1}{n}}.\qedhere$$ The behavior of entropy with respect to interpolation Banach spaces was studied intensively. is easy to that behave if one of endpoints is fixed and we refer to [@ET] for details. Surprisingly, it was shown only recently in that a formula for numbers both interpolated is out We give only a simplified version

$k-1\le 2n/p$, the result is trivial as The right-haNd sidE of Is lArGer tHan oR equal to 1. We will THereFore assume that $k-1>2n/p$ or, equIvaleNtLY, $2^{(k-1)p/n}>4.$ lEt $X_1,\dots,X_N$ be a maXImAL FamIlY oF elEmENtS of $B_X$ WitH $\|x_i-x_j\|_X>\tAu$ for $i\not=j$, WheRe $\Tau>0$ is given by $\FRaC{1+\tau^p/2}{\tau^p/2}=2^{p(K-1)/n}$. THen, by trianglE inEqualiTy, The SEts $x_i+2^{-1/P}\taU B_X$ arE mutuaLLy disjOint, they aRe ALl inclUDed in $(1+\taU^P/2)^{1/P}B_x$ and $b_X$ is covered by the uNIoN Of $x_i+\tau B_X$ over $i=1,\Dots,N$. BY vOLuME ComParIson, we get $$N{\OpEratoRNamewitHLiMITS{voL}}(2^{-1/P}\tau B_X)\le {\operaTornamewithLImiTs{vol}}[(1+\tAu^P/2)^{1/p}B_x]$$ And, theReforE, $$N\LE \frAc{(1+\tau^p/2)^{n/p}}{2^{-n/p}\tAu^n}=\BIgl(\frac{1+\taU^p/2}{\tau^p/2}\bIgr)^{n/p}=2^{k-1}.$$ WE ConcludE that $$e_K(id:x\to x)\le \tAU=\bIgGl[\fRaC{2}{2^{(K-1)p/n}-1}\BIgGr]^{1/p}\LE\biGgl[\frac{4}{2^{(k-1)P/n}}\BiGgr]^{1/p}=4^{1/p}2^{-\Frac{K-1}{N}}.\QEDherE$$ ThE behAvior Of entropy numbErs With REspEct to InterPolaTiOn of BAnach sPaces WaS studied intensiVely. it is ratheR eaSy To sHoW, that THey behAve WelL if one oF the endPOinTs IS FIxEd and we refer to [@ET] foR dETAiLs. SurpriSingly, IT wAs SHown only ReCenTly iN [@en], that A genERaL interpoLation FOrMuLa for enTrOpy numBeRs wIth Both eNDpoiNts intErpolateD is ouT Of reach. We give oNLy a simplified VErSIOn

$k-1\le 2n/p$, the result is trivia l asthe ri gh t-ha nd s ide of is larg e r th an or equal to 1. We w ill t he r efor e a ssume that $ k -1 > 2 n/p $or , e qu i va lentl y,$2^{(k- 1)p/n}>4.$ L et $x_1,\dots, x _N $ be a max ima l family ofele mentsof $B _ X$ wi th$\|x_ i-x_j\ | _X>\ta u$ for $i \n o t=j$,w here $\ t a u> 0$ i s given by $\frac { 1+ \ tau^p/2}{\tau^ p/2}=2 ^{ p (k - 1 )/n }$. Then, bytr iangl e inequa l it y , the sets $x_i+2^{ -1/p}\tau B _ X$are mu tu all y disjo int,th e y a re all incl uded in $(1+\ tau^p/ 2 )^{1/p} B _X$ and $B_X$ is co vere d b ythe u n ion of $x _ i+\ tau B_X$ o ve r $i= 1,\d o t s , N$.Byvolu me co mparison, weget $$N { \op erato rname with li mits{ vol}}( 2^{-1 /p }\tau B_X)\le { \ope ratorname wit hl imi ts {vol} } [(1+\t au^ p/2 )^{1/p} B_X]$$a nd, t h e r ef ore, $$N\le \frac{ (1 + \ ta u^p/2)^{ n/p}}{ 2 ^{ -n / p}\tau^n }= \Bi gl(\ f r ac{1+ \tau ^ p/ 2}{\tau^ p/2}\B i gr )^ {n/p}=2 ^{ k-1}.$ $Wecon clude that $$e_k (id:X\to X)\l e \tau=\biggl[\ f rac{2}{2^{(k- 1 )p / n }- 1 }\bi ggr ]^{1/p}\le\ bigg l [\fr ac{4 } {2 ^{( k -1)p/ n}}\b ig g r] ^ {1/p}=4^{1/p}2^{-\f ra c{k-1} {n}}. \qedhere$$ T he behavio r o f entrop y nu m be r s with respect to i nterpolati o n of Ban ach s paces wa s studied i ntensive ly. It is ra t h er easy to show , that t hey beh ave well i f o neofthe e ndpointsis fixed a nd w eref er to [@ET] fo rdet ai ls. Surp r isingl y, it was s ho w n o nly rec e nt l y in[@ EN ], t hat a gene rali nte rpolati on formul a f o r en tr op y numbe rs with bothen dpoints in te rpo latedi s out ofreach. We give only a s i mplifie d v ersio n

$k-1\le_2n/p$, the_result is trivial as_the right-hand_side_of is_larger_than or equal_to 1. We_will therefore assume that_$k-1>2n/p$ or, equivalently,_$2^{(k-1)p/n}>4.$ Let_$x_1,\dots,x_N$ be a maximal family of elements of $B_X$ with $\|x_i-x_j\|_X>\tau$ for $i\not=j$, where_$\tau>0$_is given_by_$\frac{1+\tau^p/2}{\tau^p/2}=2^{p(k-1)/n}$._Then, by triangle inequality, the_sets $x_i+2^{-1/p}\tau B_X$ are mutually_disjoint, they_are all included in $(1+\tau^p/2)^{1/p}B_X$ and $B_X$ is_covered_by the union_of $x_i+\tau B_X$ over $i=1,\dots,N$. By volume comparison, we_get $$N{\operatornamewithlimits{vol}}(2^{-1/p}\tau B_X)\le {\operatornamewithlimits{vol}}[(1+\tau^p/2)^{1/p}B_X]$$ and, therefore,_$$N\le \frac{(1+\tau^p/2)^{n/p}}{2^{-n/p}\tau^n}=\Bigl(\frac{1+\tau^p/2}{\tau^p/2}\Bigr)^{n/p}=2^{k-1}.$$ We_conclude_that_$$e_k(id:X\to X)\le \tau=\biggl[\frac{2}{2^{(k-1)p/n}-1}\biggr]^{1/p}\le\biggl[\frac{4}{2^{(k-1)p/n}}\biggr]^{1/p}=4^{1/p}2^{-\frac{k-1}{n}}.\qedhere$$ The behavior_of entropy numbers with respect to_interpolation of Banach spaces was studied_intensively. It is rather easy to show,_that they behave well if one_of the endpoints is fixed_and we_refer to [@ET] for details._Surprisingly, it was_shown only_recently in [@EN],_that a general interpolation formula for_entropy numbers with_both endpoints interpolated is out of_reach._We give only_a_simplified_version

_L}\iota(S_{L}))}^{\ast}(K\times_{K_L} X_G)\times X_G}+s \frac{df_{(k,x)}}{f(k,x)}\right)= {T_{kx,K\cdot\iota(S_L)}^{\ast}X_G}+s\lambda(f)_{kx}.$$ Thus, if for each cycle $C\in \mathscr{L}(X_G,K)$ with $|C|\subset \overline{K\cdot \iota(X_L)}$ (See Definition \[deftn:ccycleperverse\]), and for each $s>0$ we define $C + s\lambda(f)$ as the cycle of $X_G$ equal to the image of $C$ under the automorphism $$\begin{aligned} (x,\xi)\mapsto (x,\xi+s\lambda (f)_{x}),\quad x\in \overline{K\cdot \iota(X_L)},~\xi \in T_{x}^{\ast}X_G,\end{aligned}$$ we obtain $$\begin{aligned} \tau_{\ast}d{p}^{-1}(i_{\ast}CC(\widetilde{R\iota_{\ast}\mathcal{F}})+s d\log f) &=\tau_{\ast}dp^{-1}\left(\sum_{K_L-\mathrm{orbits }~S_L~\mathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{i(K\times_{K_L}\iota(S_{L}))}^{\ast}((K\times_{K_L}X_G)\times X_G)}]\right)\\ &=\sum_{K_L-\mathrm{orbits }~S_L~\mathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{K\cdot\iota(S_L)}^{\ast}X_G}]+s\lambda(f).\end{aligned}$$ Therefore $$\begin{aligned} CC\left(I_{L}^{G}\mathcal{F}\right)&=a_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal{F}}))\\ &=\overline{p}_{\ast}(j_{\ast}(i_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal

_ L}\iota(S_{L}))}^{\ast}(K\times_{K_L } X_G)\times X_G}+s \frac{df_{(k, x)}}{f(k, x)}\right)= { T_{kx, K\cdot\iota(S_L)}^{\ast}X_G}+s\lambda(f)_{kx}.$$ Thus, if for each cycle $ C\in \mathscr{L}(X_G, K)$ with $ |C|\subset \overline{K\cdot \iota(X_L)}$ (See Definition \[deftn: ccycleperverse\ ]), and for each $ s>0 $ we define $ vitamin c + s\lambda(f)$ as the hertz of $ X_G$ equal to the image of $ C$ under the automorphism $ $ \begin{aligned } (x,\xi)\mapsto (x,\xi+s\lambda (f)_{x}),\quad x\in \overline{K\cdot \iota(X_L)},~\xi \in T_{x}^{\ast}X_G,\end{aligned}$$ we obtain $ $ \begin{aligned } \tau_{\ast}d{p}^{-1}(i_{\ast}CC(\widetilde{R\iota_{\ast}\mathcal{F}})+s d\log f) & = \tau_{\ast}dp^{-1}\left(\sum_{K_L-\mathrm{orbits } ~S_L~\mathrm{in}~X_L } \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{i(K\times_{K_L}\iota(S_{L}))}^{\ast}((K\times_{K_L}X_G)\times X_G)}]\right)\\ & = \sum_{K_L-\mathrm{orbits } ~S_L~\mathrm{in}~X_L } \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{K\cdot\iota(S_L)}^{\ast}X_G}]+s\lambda(f).\end{aligned}$$ consequently $ $ \begin{aligned } CC\left(I_{L}^{G}\mathcal{F}\right)&=a_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal{F}}))\\ & = \overline{p}_{\ast}(j_{\ast}(i_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal

_L}\ioha(S_{L}))}^{\ast}(K\times_{K_L} X_G)\times W_G}+s \frac{df_{(k,x)}}{f(k,x)}\rntht)= {T_{kx,N\cdot\ikta(S_L)}^{\ast}B_G}+s\lambda(f)_{kx}.$$ Thus, if for eaci cyxle $C\un \mathscr{L}(X_G,K)$ with $|C|\sjbset \ovegline{K\cdor \ioua(X_L)}$ (See Definition \[deftn:geyclelcrvervx\]), and for each $x>0$ we defina $C + s\lambda(f)$ ds tke cycle of $X_G$ equal to the image of $C$ undet hhe automorphifm $$\btgig{alifned} (x,\xi)\mapsto (x,\xi+s\lambda (f)_{x}),\quad x\ih \overlpne{K\cdot \iota(X_L)},~\xi \in T_{x}^{\ast}X_G,\end{aligned}$$ we ohtaij $$\begin{aligned} \tau_{\adt}d{p}^{-1}(i_{\ast}CC(\wudetyode{R\iota_{\ast}\mxthcal{F}})+s d\log f) &=\tau_{\ast}dk^{-1}\left(\sum_{K_L-\mathrm{orbits }~S_L~\mathrm{iv}~X_L} \cki_{S_{L}}^{mic}(\mathxao{F})[\oggrline{T_{i(K\timxs_{K_L}\iona(S_{L}))}^{\ast}((K\times_{K_L}X_G)\timev X_G)}]\rignt)\\ &=\sum_{K_L-\mathrm{ovbits }~S_L~\nathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mavhcal{F})[\overline{T_{K\cdot\yota(S_L)}^{\ast}F_G}]+a\lambda(f).\end{alignee}$$ Rherexore $$\begkb{alkgntd} CR\lert(I_{L}^{G}\mwthral{F}\right)&=a_{\aat}(CC(\widetilee{R\iota_{\ast}\mathcal{F}}))\\ &=\ofewoine{p}_{\ast}(j_{\ast}(i_{\zst}(CC(\wydqtilde{R\iota_{\ast}\mathcal

_L}\iota(S_{L}))}^{\ast}(K\times_{K_L} X_G)\times X_G}+s \frac{df_{(k,x)}}{f(k,x)}\right)= {T_{kx,K\cdot\iota(S_L)}^{\ast}X_G}+s\lambda(f)_{kx}.$$ Thus, if cycle \mathscr{L}(X_G,K)$ with \overline{K\cdot \iota(X_L)}$ (See $s>0$ define $C + as the cycle $X_G$ equal to the image of under the automorphism $$\begin{aligned} (x,\xi)\mapsto (x,\xi+s\lambda (f)_{x}),\quad x\in \overline{K\cdot \iota(X_L)},~\xi \in T_{x}^{\ast}X_G,\end{aligned}$$ we $$\begin{aligned} \tau_{\ast}d{p}^{-1}(i_{\ast}CC(\widetilde{R\iota_{\ast}\mathcal{F}})+s d\log f) &=\tau_{\ast}dp^{-1}\left(\sum_{K_L-\mathrm{orbits }~S_L~\mathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{i(K\times_{K_L}\iota(S_{L}))}^{\ast}((K\times_{K_L}X_G)\times X_G)}]\right)\\ &=\sum_{K_L-\mathrm{orbits }~S_L~\mathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{K\cdot\iota(S_L)}^{\ast}X_G}]+s\lambda(f).\end{aligned}$$ Therefore $$\begin{aligned} &=\overline{p}_{\ast}(j_{\ast}(i_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal

_L}\iota(S_{L}))}^{\ast}(K\times_{K_L} X_G)\times x_G}+s \frac{df_{(k,X)}}{f(k,x)}\rIghT)= {T_{kX,K\Cdot\Iota(s_L)}^{\ast}X_G}+s\lambda(F)_{Kx}.$$ ThUs, if for each cycle $C\in \matHscr{L}(x_G,k)$ With $|c|\SuBset \oVerline{k\CdOT \IotA(X_l)}$ (SEe DEfINiTion \[dEftN:ccyclePerverse\]), anD foR eAch $s>0$ we define $c + S\lAmbda(f)$ as thE cyCle of $X_G$ equal To tHe imagE oF $C$ uNDer thE auTomorPhism $$\bEGin{aliGned} (x,\xi)\maPsTO (x,\xi+s\lAMbda (f)_{x}),\qUAD x\In \ovErline{K\cdot \iota(X_L)},~\XI \iN t_{x}^{\ast}X_G,\end{aligNed}$$ we oBtAIn $$\BEGin{AliGned} \tau_{\ast}D{p}^{-1}(I_{\ast}Cc(\WidetilDE{R\IOTA_{\asT}\Mathcal{F}})+s d\log F) &=\tau_{\ast}dp^{-1}\leFT(\suM_{K_L-\matHrM{orBIts }~S_L~\mAthrm{In}~x_l} \chI_{S_{L}}^{mic}(\mathcAl{F})[\oVerline{T_{i(k\times_{k_l}\iota(S_{L}))}^{\ASt}((K\timeS_{K_L}X_G)\tImeS X_G)}]\RighT)\\ &=\SuM_{K_l-\maThRM{orBItS }~S_L~\MAthRm{in}~X_L} \chI_{S_{l}}^{mIc}(\matHcal{f})[\OVERlinE{T_{K\Cdot\Iota(S_l)}^{\ast}X_G}]+s\lambda(F).\enD{aliGNed}$$ thereFore $$\bEgin{AlIgned} cC\left(i_{L}^{G}\maThCal{F}\right)&=a_{\ast}(CC(\WideTilde{R\iotA_{\asT}\mAthCaL{F}}))\\ &=\oveRLine{p}_{\aSt}(j_{\Ast}(I_{\ast}(CC(\wIdetildE{r\ioTa_{\AST}\MaThcal

_L}\iota(S_{L}))}^{\ast}(K \times_{K_ L} X_ G)\ tim es X_G }+s\frac{df_{(k,x ) }}{f (k,x)}\right)= {T_{kx, K\cdo t\ i ota( S _L )}^{\ ast}X_G } +s \ l amb da (f )_{ kx } .$ $ Thu s,if foreach cycle $C \i n \mathscr{L } (X _G,K)$ wit h $ |C|\subset \ ove rline{ K\ cdo t \iot a(X _L)}$ (SeeD efinit ion \[def tn : ccycle p erverse \ ] ), and for each $s>0$ w e d e fine $C + s\la mbda(f )$ as t hecyc le of $X_G $equal to thei ma g e of$ C$ under theautomorphis m $$ \begin {a lig n ed} (x ,\xi) \m a pst o (x,\xi+s\ lamb da (f)_{x }),\qu a d x\in\ overlin e{K\cd ot\io ta(X _ L) }, ~\x i\ inT _{ x}^ { \as t}X_G,\e nd {a ligne d}$$ w e obta in$$\b egin{ aligned} \tau _{\ ast} d {p} ^{-1} (i_{\ ast} CC (\wid etilde {R\io ta _{\ast}\mathcal {F}} )+s d\log f) & =\t au _{\as t }dp^{- 1}\ lef t(\sum_ {K_L-\m a thr m{ o r b it s }~S_L~\mathrm{in }~ X _ L} \chi_{S _{L}}^ { mi c} ( \mathcal {F })[ \ove r l ine{T _{i( K \t imes_{K_ L}\iot a (S _{ L}))}^{ \a st}((K \t ime s_{ K_L}X _ G)\t imes X _G)}]\ri ght)\ \ &=\sum_{K_L-\ m athrm{orbits} ~S _ L ~\ m athr m{i n}~X_L} \ch i_{S _ {L}} ^{mi c }( \ma t hcal{ F})[\ ov e rl i ne{T_{K\cdot\iota(S _L )}^{\a st}X_ G}]+s\lambda( f).\end{al i g n ed}$$ Th eref o re $$\begin{align ed} C C\left(I_{ L }^{G}\ma thcal {F}\righ t)&=a_{\a s t }(CC(\wi det ild e{R \io t a _{ \ast}\mathcal { F }})) \\ &=\ove rli ne{p}_{ \as t}( j_{ \as t} (i_{\ast} (CC(\wid et il de {R \io ta_{\ a st}\math ca l

_L}\iota(S_{L}))}^{\ast}(K\times_{K_L} X_G)\times_X_G}+s \frac{df_{(k,x)}}{f(k,x)}\right)= {T_{kx,K\cdot\iota(S_L)}^{\ast}X_G}+s\lambda(f)_{kx}.$$_Thus, if for each_cycle $C\in_\mathscr{L}(X_G,K)$_with $|C|\subset_\overline{K\cdot_\iota(X_L)}$ (See Definition_\[deftn:ccycleperverse\]), and for_each $s>0$ we define_$C + s\lambda(f)$_as_the cycle of $X_G$ equal to the image of $C$ under the automorphism $$\begin{aligned} (x,\xi)\mapsto_(x,\xi+s\lambda_(f)_{x}),\quad x\in_\overline{K\cdot_\iota(X_L)},~\xi_\in T_{x}^{\ast}X_G,\end{aligned}$$ we obtain $$\begin{aligned} \tau_{\ast}d{p}^{-1}(i_{\ast}CC(\widetilde{R\iota_{\ast}\mathcal{F}})+s_d\log f) &=\tau_{\ast}dp^{-1}\left(\sum_{K_L-\mathrm{orbits }~S_L~\mathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{i(K\times_{K_L}\iota(S_{L}))}^{\ast}((K\times_{K_L}X_G)\times X_G)}]\right)\\ &=\sum_{K_L-\mathrm{orbits_}~S_L~\mathrm{in}~X_L} \chi_{S_{L}}^{mic}(\mathcal{F})[\overline{T_{K\cdot\iota(S_L)}^{\ast}X_G}]+s\lambda(f).\end{aligned}$$_Therefore $$\begin{aligned} CC\left(I_{L}^{G}\mathcal{F}\right)&=a_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal{F}}))\\ &=\overline{p}_{\ast}(j_{\ast}(i_{\ast}(CC(\widetilde{R\iota_{\ast}\mathcal

persistent for the modes with $m=0,1,2$ and $l=1,2$. This observation raises the important question: What are the electromagnetic QNMs for which one has to look in astrophysical data. Also interesting is that while the $\epsilon$-method leads to significant changes of the frequencies $\omega_{m,n}$, it affects much less the separation parameter $E_{m,n}$ which here for the first time was obtained directly as a solution of the two-dimensional system without any prior approximations for it. Another general result is that the confluent Heun function proved to be an effective tool for physical problems. Even though its <span style="font-variant:small-caps;">maple</span> realization still has many flaws, its precision proved to be good enough to repeat the already published results, and also studying the solutions, we were able to reveal new properties of the numerical stability of the EM QNMs with respect to changes in the phase-condition. An interesting question is the results obtained using this method for $a>M$ or the so called naked singularity regime. Preliminary results show that the method is applicable in this case as well and the results will be published elsewhere. Acknowledgements ================ The authors would like to thank Emanuele Berti for discussion of the numerical values of the EM QNM frequencies obtained through Leaver’s method which was important for the comparison of our method with this already well-established one. P.F. is deeply indebted to Irina Aref’ieva, Alexei Starobinsky, Bruno Coppi and Michail Sazhin for interesting comments and suggestions about the relations of EM QNM to BH physics and the possibilities to discover such modes in astrophysical observations. The authors are also thankful to Luciano Rezzolla for drawing our attention to the references [@headon0] and [@GW_] and to Shahar Hod for drawing our attention to the references [@extr1] and [@extr2]. This article was supported by the Foundation “Theoretical and Computational Physics and Astrophysics”, by the Bulgarian National Scientific Fund under contracts DO-1-872, DO-1-895, DO-02-136, and Sofia University Scientific Fund, contract 185/26.04.2010. Author Contributions ==================== P.F. posed the problem of the evaluation of the EM QNMs of

persistent for the modes with $ m=0,1,2 $ and $ l=1,2$. This observation raises the authoritative doubt: What are the electromagnetic QNMs for which one has to look in astrophysical data. besides interesting is that while the $ \epsilon$-method leads to significant changes of the frequency $ \omega_{m, n}$, it affects much less the interval parameter $ E_{m, n}$ which here for the first prison term was obtained directly as a solution of the two - dimensional organization without any anterior approximations for it. Another general resultant role is that the confluent Heun function proved to be an effective instrument for physical problems. Even though its < span style="font - variant: little - caps;">maple</span > realization still has many flaws, its precision prove to be good enough to repeat the already published results, and also study the solutions, we were able to reveal new properties of the numerical constancy of the EM QNMs with esteem to changes in the phase - condition. An interesting question is the results obtained using this method for $ a > M$ or the so called bare singularity government. Preliminary results usher that the method is applicable in this case as well and the results will be publish elsewhere. Acknowledgements = = = = = = = = = = = = = = = = The authors would like to thank Emanuele Berti for discussion of the numerical value of the EM QNM frequencies obtained through Leaver ’s method which was important for the comparison of our method with this already well - install one. P.F. is deeply indebted to Irina Aref’ieva, Alexei Starobinsky, Bruno Coppi and Michail Sazhin for interesting comments and suggestions about the relations of EM QNM to bohrium physics and the possibility to detect such modes in astrophysical observations. The authors are also thankful to Luciano Rezzolla for drawing our attention to the reference point [ @headon0 ] and [ @GW _ ] and to Shahar Hod for reap our attention to the references [ @extr1 ] and [ @extr2 ]. This article was hold by the Foundation “ Theoretical and Computational Physics and Astrophysics ”, by the Bulgarian National Scientific Fund under contracts DO-1 - 872, DO-1 - 895, DO-02 - 136, and Sofia University Scientific Fund, contract 185/26.04.2010. Author Contributions = = = = = = = = = = = = = = = = = = = = P.F. posed the problem of the evaluation of the EM QNMs of

pegsistent for the modes wlth $m=0,1,2$ and $l=1,2$. This observetion rzises thd important question: What arx thw elextromagnetic QNMs for dhich one has to oook un astrophbaical data. Alal incecesting is that while the $\epsilon$-method ldabs to significant changes of the frezuenciex $\lmega_{m,n}$, it affgcts ktch mvsw the separation parameter $E_{m,h}$ which here for the first time was obtained dlrechly as a solution lf the two-dumenfuonal system without any prior apptoximations for it. Another generau resblt is that tye fmnfluent Henn funbtion proved bp be at effecyive tool for ihysiral problems. Even though mts <span style="font-vatiant:small-wa's;">maple</span> realizatiin stiln hav mavt fuawa, mts precidioi proved to be good eniugh to repeat the slwvsdy publishes resujtf, and also studying the solutions, we wege agle to reveal new propeeties of the numericap stabiliey of the EM QNMs with respect to changes in the [hase-rovdiuiin. An kbtfresting question is the results obtained usigf uhix method for $a>I$ or the so cwlkgd naked singuuarity refime. Preliminary rfsults fhow rhat the iethpd is applicable in this cawe as well aud rhe results will bz published zlsewhgre. Ackmowledgements ================ The authorr wohld like to thank Emzvuele Berti for aisbusshon of the numerical valuef of the XM QNK frequdncigs obtayned throuhh Leaver’s method which aas ilpmrtant for the comparison of our method wmvh this alreacy wenl-establnshed pne. P.F. is deepjy indebted to Irina Cref’iexa, Alexei Atarobiisky, Bruno Cjppi and Michdll Sazhin foc intereseing comnents ava suggestions sbout the relations of EM QNM to BH pmysicr and the possibnuuties to discovrr rucr loves ig astrophysicdl ocseffatiovs. The authovs xre slso thankful to Luchano Rezzolla for drawonn our attgntion to the referencrs [@headon0] and [@GW_] ajd to Shehar Hpd sor drawing our attention to tge referejcef [@extr1] and [@evtr2]. Tmis wrticle wax supported by the Foundation “Theoreticel and Computational Phtsics and Astrophysncx”, by the Bukgariwn Nationdl Scientific Fund ubder contracts DO-1-872, DO-1-895, DO-02-136, and Sofia Univsrsity Sciejtific Fund, contract 185/26.04.2010. Author Contributions ==================== P.F. posed the problem of the evalyation of the EM QHMs pf

persistent for the modes with $m=0,1,2$ and observation the important What are the has look in astrophysical Also interesting is while the $\epsilon$-method leads to significant of the frequencies $\omega_{m,n}$, it affects much less the separation parameter $E_{m,n}$ which for the first time was obtained directly as a solution of the two-dimensional without prior for Another general result is that the confluent Heun function proved to be an effective tool for problems. Even though its <span style="font-variant:small-caps;">maple</span> realization still many flaws, its precision to be good enough to the published results, also the we were able reveal new properties of the numerical stability of the EM QNMs with respect to changes in the An interesting the results using method $a>M$ or the naked singularity regime. Preliminary results show is applicable in this case as well and results will published elsewhere. Acknowledgements ================ The authors like to thank Emanuele Berti for discussion of numerical values of the EM QNM frequencies obtained through Leaver’s method which was important for of our method with already well-established one. is indebted Irina Alexei Starobinsky, Coppi and Michail Sazhin for interesting comments and suggestions about the of EM QNM to BH physics and the possibilities to modes astrophysical observations. The are also thankful to Rezzolla drawing our attention to [@headon0] [@GW_] Hod drawing attention to the references and [@extr2]. This article was by the Foundation “Theoretical by the Bulgarian National Scientific Fund under contracts DO-1-895, DO-02-136, and Sofia University Scientific Fund, 185/26.04.2010. Author Contributions ==================== P.F. posed the problem of the evaluation of EM QNMs

persistent for the modes with $M=0,1,2$ and $l=1,2$. This oBservAtiOn rAiSes tHe imPortant questioN: what Are the electromagnetic QnMs foR wHIch oNE hAs to lOok in asTRoPHYsiCaL dAta. alSO iNtereStiNg is thaT while the $\ePsiLoN$-method leads TO sIgnificant ChaNges of the freQueNcies $\oMeGa_{m,N}$, It affEctS much Less thE SeparaTion paramEtER $E_{m,n}$ whICh here fOR ThE firSt time was obtained DIrECtly as a solutioN of the TwO-DiMENsiOnaL system witHoUt any PRior appROxIMATioNS for it. Another General resuLT is That thE cOnfLUent HeUn funCtIOn pRoved to be an EffeCtive tool For phySIcal proBLems. EveN thougH itS <spAn stYLe="FoNt-vArIAnt:SMaLl-cAPs;">mAple</span> ReAlIzatiOn stILL HAs maNy fLaws, Its prEcision proved To bE gooD EnoUgh to RepeaT the AlReady PublisHed reSuLts, and also studyIng tHe solutioNs, wE wEre AbLe to rEVeal neW prOpeRties of The numeRIcaL sTABIlIty of the EM QNMs with ReSPEcT to changEs in thE PhAsE-ConditioN. AN inTereSTIng quEstiON iS the resuLts obtAInEd Using thIs Method FoR $a>M$ Or tHe so cALled Naked sIngulariTy regIMe. Preliminary rESults show that THe METhOD is aPplIcable in thiS casE As weLl anD ThE reSUlts wIll be PuBLiSHed elsewhere. AcknowlEdGementS ================ The aUthors would liKe to thank EMANUele BertI for DIsCUssion of the numEricaL values of tHE EM QNM frEquenCies obtaIned throuGH leaver’s mEthOd wHicH waS IMpOrtant for the cOMPariSoN of our mEthOd with tHis AlrEadY weLl-EstablishEd one. P.F. iS dEePlY iNdeBted tO irina AreF’iEva, alExeI StarOBinsky, bruno coppI aNd mIchAil SazhIN fOR InteReStIng cOmmEnTs and SuggEStiOns abouT the relatIonS Of EM qNm tO BH physIcs and the possIbIlities to dIsCovEr such MODes in astRophysical observations. THE authorS arE also ThanKful to LucIanO RezzoLla FOr drawIng our AttenTiOn tO THe refEREnCes [@HeAdon0] and [@GW_] aND To SHahar hoD for Drawing Our attention to the rEFerEnces [@extr1] and [@eXtr2]. this ARTiCle WAs SUppOrTEd bY THe Foundation “TheOretical anD COMpUtational PHYsiCs And AstrOphysicS”, by thE bulgariAn NationaL ScientifIc fund UNDer Contracts Do-1-872, DO-1-895, DO-02-136, and sofia UnivERsity sCiEntifIc FUnd, conTrAct 185/26.04.2010. authoR ContrIButIons ==================== P.f. posed ThE problEm of tHe EvaluatiOn of the EM QNMs of

persistent for the modeswith $m=0, 1,2$and $l =1 ,2$. Thi s observationr aise s the important questi on: W ha t are th e ele ctromag n et i c QN Ms f orwh i ch onehas to loo k in astro phy si cal data. Al s ointerestin g i s that while th e $\ep si lon $ -meth odleads to si g nifica nt change so f thef requenc i e s$\om ega_{m,n}$, it af f ec t s much less th e sepa ra t io n par ame ter $E_{m, n} $ whi c h heref or t h e f i rst time wasobtained di r ect ly asasol u tion o f the t w o-d imensionalsyst em withou t anyp rior ap p roximat ions f orit. An o th er ge ne r alr es ult isthat the c on fluen t He u n f unct ion pro ved t o be an effec tiv e to o l f or ph ysica l pr ob lems. Eventhoug hits <span style ="fo nt-varian t:s ma ll- ca ps;"> m aple</ spa n>realiza tion st i llha s m an y flaws, its preci si o n p roved to be go o den o ugh to r ep eat the a lread y pu b li shed res ults,a nd a lso stu dy ing th esol uti ons,w e we re abl e to rev eal n e w properties o f the numerica l s t a bi l ityofthe EM QNMs wit h res pect to ch a ngesin th ep ha s e-condition. An in te restin g que stion is theresults ob t a i ned usin g th i sm ethod for $a>M $ orthe so cal l ed naked sing ularityregime. P r e liminary re sul tssho w th at the method i s ap pl icableinthis ca seaswel l a nd the resu lts will b epu bl ish ed el s ewhere. A ckn ow led gemen t s ==== ===== ==== == =The author s w o u ld l ik eto t han kEmanu eleB ert i for d iscussion of thenu me rical v alues of theEM QNM frequ en cie s obta i n ed throu gh Leaver’s method whic h was im por tantforthe compa ris on ofour method withthisal rea d y well - e st abl is hed one. P . F.is de ep ly i ndebted to Irina Aref’iev a , A lexei Starobi nsk y, B r u no Co p pi and M i cha i l Sazhin for int eresting c om m en ts and sug g est io ns abou t the r elati o ns of E M QNM toBH physic sandt h e p ossibiliti es to di scover su c h mod e sin as tro physic al ob serva tions. Th e aut hors a re alsothank fu l to Luc iano Rezzolla for drawi ng our atte nti on to the re f ere nces [@he adon 0] and [@G W_] an d toSha h ar Ho d fo r d raw i ng ou r at t ention to th e r e f er ences [@ext r 1 ] an d [@e xtr 2 ]. Th is a rticle was suppor t ed by the Foun dati o n “T heo r etic al and Computati ona lP h ysics an dAstrophysic s”, by t he Bulga rian N ationa l Scien t i fi c Fundunde r c ontractsDO- 1- 8 72, DO- 1- 89 5 , DO-0 2-13 6, and S ofia U n iver s i ty Scientific Fu nd, c o n tract 185 /26.0 4. 2010. A utho r Contribu tions ===== ====== ==== ===== P.F.po sed th e p ro blem of th e evaluati on of the EM Q NMsof

persistent_for the_modes with $m=0,1,2$ and_$l=1,2$. This_observation_raises the_important_question: What are_the electromagnetic QNMs_for which one has_to look in_astrophysical_data. Also interesting is that while the $\epsilon$-method leads to significant changes of the_frequencies_$\omega_{m,n}$, it_affects_much_less the separation parameter $E_{m,n}$_which here for the first_time was_obtained directly as a solution of the two-dimensional_system_without any prior_approximations for it. Another general result is that the confluent_Heun function proved to be an_effective tool for_physical_problems._Even though its <span_style="font-variant:small-caps;">maple</span> realization still has many flaws,_its precision proved to be good_enough to repeat the already published results,_and also studying the solutions, we_were able to reveal new_properties of_the numerical stability of the_EM QNMs with_respect to_changes in the_phase-condition. An interesting question is the results_obtained using this_method for $a>M$ or the so_called_naked singularity regime._Preliminary_results_show that_the method is_applicable_in this_case_as well and the results will_be_published elsewhere. Acknowledgements ================ The authors would like to thank_Emanuele Berti for discussion_of_the numerical values of_the EM QNM frequencies obtained_through Leaver’s method which was important_for the_comparison of_our method with this already well-established one. P.F. is deeply indebted to_Irina Aref’ieva, Alexei Starobinsky, Bruno Coppi_and Michail Sazhin for_interesting comments_and_suggestions about the_relations_of EM_QNM to BH physics and the possibilities_to discover_such modes in astrophysical observations. The authors_are also thankful to_Luciano_Rezzolla for drawing our attention to_the references [@headon0] and [@GW_] and_to Shahar Hod for drawing_our_attention_to the references [@extr1] and_[@extr2]. This article was supported by the_Foundation “Theoretical and_Computational Physics and Astrophysics”, by the Bulgarian_National_Scientific Fund under contracts DO-1-872, DO-1-895,_DO-02-136,_and Sofia University Scientific Fund, contract_185/26.04.2010. Author_Contributions ==================== P.F._posed the problem of the_evaluation of the EM QNMs of

), \[[[arXiv:math-ph/0404057]{}](https://arxiv.org/abs/math-ph/0404057)\]. T. Guhr, [*[Supersymmetry]{}*]{}, [*Chapter 7 in*]{} [@handbook:2010] (2011), \[[[arXiv:1005.0979]{}](https://arxiv.org/abs/1005.0979)\]. C. W. Bernard and M. F. L. Golterman, [*Partially quenched gauge theories and an application to staggered fermions*]{}, [*Phys.Rev. D*]{} [**49**]{} (1994) 486–494, \[[[arXiv:hep-lat/9306005]{}](https://arxiv.org/abs/hep-lat/9306005)\]. M. F. L. Golterman, [*Applications of chiral perturbation theory to lattice QCD*]{}, [*conference proceedings at the International School, 93rd Session, Les Houches, France, August 3-28, (2009)*]{}, [*Modern perspectives in lattice QCD: Quantum field theory and high performance computing*]{} (2009) 423–515, \[[[arXiv:0912.4042]{}](https://arxiv.org/abs/0912.4042)\]. M. L. Mehta, [*[Random Matrices]{}*]{}, Academic Press, Amsterdam, 3rd ed. (2004). G. Akemann and T. Nagao, [*[Random Matrix Theory for the Hermitian Wilson Dirac Operator and the chGUE-GUE Transition]{}*]{}, [*JHEP*]{} [**2011**]{} (2011) 60, \[[[arXiv:1108.3035]{}](https://arxiv.org/abs/1108.3035)\]. P. H. Damgaard, K. Splittorff and J. J. M. Verbaarschot, [*Microscopic Spectrum of the Wilson Dirac Operator*]{}, [*Phys. Rev. Lett.*]{} [**105**]{} (2010) 162002, \[[arXiv:1001.2937](https://arxiv.org/abs/1001.2937)\]. L. Susskind, [*Lattice fermions*]{}, [*Phys. Rev. D*]{} [**16**]{} (1977) 3031–3039. P. H. Damgaard, U. M. Heller, A

), \[[[arXiv: math - ph/0404057]{}](https://arxiv.org / abs / math - ph/0404057)\ ]. T.   Guhr, [ * [ Supersymmetry ] { } * ] { }, [ * Chapter 7 in * ] { }   [ @handbook:2010 ] (2011), \[[[arXiv:1005.0979]{}](https://arxiv.org / abs/1005.0979)\ ]. C.   W.   Bernard and M.   F.   L.   Golterman, [ * Partially quenched gauge theory and an lotion to staggered fermions * ] { }, [ * Phys. Rev. D * ] { } [ * * 49 * * ] { } (1994) 486–494, \[[[arXiv: hep - lat/9306005]{}](https://arxiv.org / abs / hep - lat/9306005)\ ]. M.   F.   L.   Golterman, [ * application of chiral disruption theory to lattice QCD * ] { }, [ * league proceedings at the International School, 93rd Session, Les Houches, France, August 3 - 28, (2009) * ] { }, [ * Modern perspectives in wicket QCD: Quantum sphere theory and gamey performance computing * ] { } (2009) 423–515, \[[[arXiv:0912.4042]{}](https://arxiv.org / abs/0912.4042)\ ]. M.   L.   Mehta, [ * [ Random Matrices ] { } * ] { }, Academic Press, Amsterdam, 3rd   erectile dysfunction. (2004). G.   Akemann and T.   Nagao, [ * [ Random Matrix Theory for the Hermitian Wilson Dirac Operator and the chGUE - GUE Transition ] { } * ] { }, [ * JHEP * ] { } [ * * 2011 * * ] { } (2011) 60, \[[[arXiv:1108.3035]{}](https://arxiv.org / abs/1108.3035)\ ]. P.   H.   Damgaard, K.   Splittorff and J.   J.   M.   Verbaarschot, [ * Microscopic Spectrum of the Wilson Dirac Operator * ] { }, [ * Phys. Rev. Lett. * ] { } [ * * 105 * * ] { } (2010) 162002, \[[arXiv:1001.2937](https://arxiv.org / abs/1001.2937)\ ]. L.   Susskind, [ * Lattice fermions * ] { }, [ * Phys. Rev. D * ] { } [ * * 16 * * ] { } (1977) 3031–3039. P.   H.   Damgaard, U.   M.   Heller, A

), \[[[arDiv:math-ph/0404057]{}](https://arxiv.org/abr/math-ph/0404057)\]. T. Guhr, [*[Sukeesymmevry]{}*]{}, [*Chalter 7 in*]{} [@fandbook:2010] (2011), \[[[arXiv:1005.0979]{}](https://arxiv.org/ebs/1005.0979)\]. C. Q. Bernqrd and M. F. L. Golterman, [*Pxrtially euenched gauje theories and eh appligction bo stcgjered fermions*]{}, [*Khys.Rev. D*]{} [**49**]{} (1994) 486–494, \[[[arXiv:hep-lat/9306005]{}](htdpr://axxiv.org/abs/hep-lat/9306005)\]. M. F. L. Golterman, [*Applicaeions og fhiral perturbwtiom thekgy to lattice QCD*]{}, [*conference prkceedinjs at the Intermational School, 93rd Session, Les Houches, France, Auhust 3-28, (2009)*]{}, [*Modeth pqespectives iv lattice QCD: Quantum rield theory and high performanze cokputing*]{} (2009) 423–515, \[[[qrZiv:0912.4042]{}](jjtps://arxiv.org/ebs/0912.4042)\]. M. L. Mvhta, [*[Random Matrices]{}*]{}, Dcademiv Press, Amstercam, 3rd wd. (2004). G. Akemann and T. Nagam, [*[Random Matrix Thgory for tve Hermitian Wilson Durac Mperdtor qnd tht ciGUS-GUE Tganaition]{}*]{}, [*JHEL*]{} [**2011**]{} (2011) 60, \[[[arXiv:1108.3035]{}](hrtps://arxiv.org/abs/1108.3035)\]. P. H. Dakgwqrd, K. Splittorrf and J. T. M. Verbaarschot, [*Microscopic Spectrum of uhe Wjlson Dirac Operator*]{}, [*Phts. Rev. Lett.*]{} [**105**]{} (2010) 162002, \[[arXiv:1001.2937](hjtps://arxiv.owg/abs/1001.2937)\]. L. Susskind, [*Lattice fermions*]{}, [*Phys. Rev. D*]{} [**16**]{} (1977) 3031–3039. P. H. Dakgaarv, J. M. Htloer, A

), \[[[arXiv:math-ph/0404057]{}](https://arxiv.org/abs/math-ph/0404057)\]. T. Guhr, [*[Supersymmetry]{}*]{}, [*Chapter 7 (2011), C. W. and M. F. theories an application to fermions*]{}, [*Phys.Rev. D*]{} (1994) 486–494, \[[[arXiv:hep-lat/9306005]{}](https://arxiv.org/abs/hep-lat/9306005)\]. M. F. L. [*Applications of chiral perturbation theory to lattice QCD*]{}, [*conference proceedings at the International 93rd Session, Les Houches, France, August 3-28, (2009)*]{}, [*Modern perspectives in lattice QCD: field and performance (2009) 423–515, \[[[arXiv:0912.4042]{}](https://arxiv.org/abs/0912.4042)\]. M. L. Mehta, [*[Random Matrices]{}*]{}, Academic Press, Amsterdam, 3rd ed. (2004). G. Akemann T. Nagao, [*[Random Matrix Theory for the Hermitian Dirac Operator and the Transition]{}*]{}, [*JHEP*]{} [**2011**]{} (2011) 60, P. Damgaard, K. and J. Verbaarschot, [*Microscopic Spectrum the Wilson Dirac Operator*]{}, [*Phys. Rev. Lett.*]{} [**105**]{} (2010) 162002, \[[arXiv:1001.2937](https://arxiv.org/abs/1001.2937)\]. L. Susskind, [*Lattice fermions*]{}, [*Phys. Rev. [**16**]{} (1977) H. Damgaard, M. A

), \[[[arXiv:math-ph/0404057]{}](https://arxiv.org/aBs/math-ph/0404057)\]. T. GUhr, [*[SuPerSymMeTry]{}*]{}, [*CHaptEr 7 in*]{} [@handbook:2010] (2011), \[[[arxIv:1005.0979]{}](htTps://arxiv.org/abs/1005.0979)\]. C. W. BernarD and M. f. L. gOlteRMaN, [*PartIally quENcHED gaUgE tHeoRiES aNd an aPplIcation To staggereD feRmIons*]{}, [*Phys.Rev. D*]{} [**49**]{} (1994) 486–494, \[[[ARXIv:hep-lat/9306005]{}](htTps://Arxiv.org/abs/hEp-lAt/9306005)\]. M. F. L. GOlTerMAn, [*AppLicAtionS of chiRAl pertUrbation tHeORy to laTTice QCD*]{}, [*CONfErenCe proceedings at thE inTErnational SchoOl, 93rd SeSsIOn, lES HoUchEs, France, AuGuSt 3-28, (2009)*]{}, [*ModERn perspECtIVES in LAttice QCD: QuanTum field theORy aNd high PeRfoRMance cOmputInG*]{} (2009) 423–515, \[[[ArXIv:0912.4042]{}](https://arxiV.org/Abs/0912.4042)\]. M. L. MehtA, [*[RandoM matriceS]{}*]{}, academiC Press, amsTerDam, 3rD Ed. (2004). g. AKemAnN And t. naGao, [*[rAndOm Matrix thEoRy for The HERMITian wilSon DIrac OPerator and the ChGuE-GUe traNsitiOn]{}*]{}, [*JHEp*]{} [**2011**]{} (2011) 60, \[[[arXIv:1108.3035]{}](Https://Arxiv.oRg/abs/1108.3035)\]. p. H. damgaard, K. SplittOrff And J. J. M. VerBaaRsChoT, [*MIcrosCOpic SpEctRum Of the WiLson DirAC OpErATOR*]{}, [*PHys. Rev. Lett.*]{} [**105**]{} (2010) 162002, \[[arXiv:1001.2937](httPs://ARXiV.org/abs/1001.2937)\]. L. susskiND, [*LAtTIce fermiOnS*]{}, [*PhYs. ReV. d*]{} [**16**]{} (1977) 3031–3039. p. H. DamGaarD, u. M. heller, A

), \[[[arXiv:math-ph/04040 57]{}](htt ps:// arx iv. or g/ab s/ma th-ph/0404057) \ ]. T. Guhr, [*[Supersymme try]{ }* ] {},[ *C hapte r 7 in* ] {} [ @ha nd bo ok: 20 1 0] (201 1), \[[[ar Xiv:1005.0 979 ]{ }](https://a r xi v.org/abs/ 100 5.0979)\]. C.W. Ber na rda nd M.  F.  L. G olterm a n, [*P artiallyqu e nchedg auge th e o ri es a nd an application to staggered ferm ions*] {} , [ * P hys .Re v. D*]{} [ ** 49**] { } (1994 ) 4 8 6 – 494 , \[[[arXiv:he p-lat/93060 0 5]{ }](htt ps :// a rxiv.o rg/ab s/ h ep- lat/9306005 )\]. M. F. L . Golt e rman, [ * Applica tionsofchi ralp er tu rba ti o n t h eo ryt o l attice Q CD *] {}, [ *con f e r e ncepro ceed ingsat the Intern ati onal Sch ool,93rdSess io n, Le s Houc hes,Fr ance, August 3- 28,(2009)*]{ },[* Mod er n per s pectiv esinlattice QCD: Q u ant um f i el d theory and highpe r f or mance co mputin g *] {} (2009) 4 23 –51 5, \ [ [ [arXi v:09 1 2. 4042]{}] (https : // ar xiv.org /a bs/091 2. 404 2)\ ]. M .  L.Mehta, [*[Rand om Ma t rices]{}*]{},A cademic Press , A m s te r dam, 3r d ed. (2004 ). G . Ak eman n a ndT . Nag ao, [ *[ R an d om Matrix Theory fo rthe He rmiti an Wilson Dir ac Operato r a nd the c hGUE - GU E Transition]{} *]{}, [*JHEP*]{ } [**2011 **]{} (2011)60, \[[[a r X iv:1108. 303 5]{ }]( htt p s :/ /arxiv.org/ab s / 1108 .3 035)\]. P . H. Da mga ard , K . S pl ittorff a nd J. J.  M .Ve rb aar schot , [*Micro sc opi cSpe ctrum of the Wils on D ir ac Ope rator*] { }, [ *Phy s. R ev.Let t. *]{}[**1 0 5** ]{} (20 10) 16200 2,\ [[ar Xi v: 1001.29 37](https://a rx iv.org/abs /1 001 .2937) \ ] . L. Su sskind, [*Lattice fermi o ns*]{}, [* Phys. Rev . D*]{} [ **1 6**]{} (1 9 77) 30 31–303 9. P .H.D a mgaar d , U . M .Heller, A

), \[[[arXiv:math-ph/0404057]{}](https://arxiv.org/abs/math-ph/0404057)\]. T. Guhr,_[*[Supersymmetry]{}*]{}, [*Chapter_7 in*]{} [@handbook:2010] (2011), \[[[arXiv:1005.0979]{}](https://arxiv.org/abs/1005.0979)\]. C. W. Bernard_and M. F. L. Golterman,_[*Partially_quenched gauge_theories_and an application_to staggered fermions*]{},_[*Phys.Rev. D*]{} [**49**]{} (1994)_486–494, \[[[arXiv:hep-lat/9306005]{}](https://arxiv.org/abs/hep-lat/9306005)\]. M. F. L. Golterman, [*Applications_of_chiral perturbation theory to lattice QCD*]{}, [*conference proceedings at the International School, 93rd Session,_Les_Houches, France,_August_3-28,_(2009)*]{}, [*Modern perspectives in lattice_QCD: Quantum field theory and_high performance_computing*]{} (2009) 423–515, \[[[arXiv:0912.4042]{}](https://arxiv.org/abs/0912.4042)\]. M. L. Mehta, [*[Random Matrices]{}*]{}, Academic Press,_Amsterdam,_3rd ed. (2004). G. Akemann and_T. Nagao, [*[Random Matrix Theory for the Hermitian Wilson Dirac_Operator and the chGUE-GUE Transition]{}*]{}, [*JHEP*]{}_[**2011**]{} (2011) 60,_\[[[arXiv:1108.3035]{}](https://arxiv.org/abs/1108.3035)\]. P. H. Damgaard,_K. Splittorff_and J. J. M. Verbaarschot, [*Microscopic Spectrum_of the Wilson Dirac Operator*]{}, [*Phys._Rev. Lett.*]{} [**105**]{} (2010) 162002, \[[arXiv:1001.2937](https://arxiv.org/abs/1001.2937)\]. L. Susskind,_[*Lattice fermions*]{}, [*Phys. Rev. D*]{} [**16**]{} (1977)_3031–3039. P. H. Damgaard, U. M. Heller, A

(1991), 1749–1813. , *A pair of Calabi-Yau manifolds as an exactly soluble superconformal theory*, Nucl. Phys. **B359** (1991), 21–74. , *Instanton strings and hyperkaehler geometry*, Nucl. Phys. **B543** (1999), 545–571; [ hep-th/9810210]{}. , *Mirror symmetry for lattice polarized K3 surfaces. Algebraic geometry, 4*, J. Math. Sci. **81** (1996), 2599–2630; [alg- geom/9502005]{}. , *Picard-[F]{}uchs uniformization and modularity of the mirror map*, Commun. Math. Phys. **212** (2000), no. 3, 625–647. height 2pt depth -1.6pt width 23pt, *Picard-[F]{}uchs uniformization: modularity of the mirror map and mirror-moonshine*, in: The arithmetic and geometry of algebraic cycles (Banff, AB, 1998), vol. 24 of CRM Proc. Lecture Notes, Amer. Math. Soc., Providence, RI, 2000, pp. 257–281; [math.ag/9812162]{}. , *Superconformal algebras and string compactification on manifolds with SU(n) holonomy*, Nucl. Phys. **B315** (1989), 193–221. , *Exactly solvable string compactifications on manifolds of $SU(N)$ holonomy*, Phys. Lett. **199B** (1987), 380–388. height 2pt depth -1.6pt width 23pt, *Space-time supersymmetry in compactified string theory and superconformal models*, Nucl. Phys. **B296** (1988), 757–778. , *Duality in Calabi-Yau moduli space*, Nucl. Phys. **B338** (1990), 15–37. , *On certain Kummer surfaces which can be realized as non-singular quartic surfaces in $\mathbb P^3$*, J. Fac. Sci. Univ. Tokyo **Sec. IA 23** (1976), 545–560. , *Chiral rings in $N=2$ superconformal theories*, Nucl. Phys. **B324** (1989), 427–474.

(1991), 1749–1813. , * A pair of Calabi - Yau manifolds as an exactly soluble superconformal theory *, Nucl. Phys. * * B359 * * (1991), 21–74. , * Instanton bowed stringed instrument and hyperkaehler geometry *, Nucl. Phys. * * B543 * * (1999), 545–571; [ hep - th/9810210 ] { }. , * Mirror isotropy for lattice polarized K3 surface. Algebraic geometry, 4 *, J.   Math.   Sci. * * 81 * * (1996), 2599–2630; [ alg- geom/9502005 ] { }. , * Picard-[F]{}uchs uniformization and modularity of the mirror function *, Commun. Math. Phys. * * 212 * * (2000), no.   3, 625–647. height 2pt depth -1.6pt width 23pt, * Picard-[F]{}uchs uniformization: modularity of the mirror map and mirror - moonlight *, in: The arithmetic and geometry of algebraic hertz (Banff, AB, 1998), vol.   24 of CRM Proc. Lecture Notes, Amer. Math. Soc. , Providence, RI, 2000, pp.   257–281; [ math.ag/9812162 ] { }. , * Superconformal algebras and bowed stringed instrument compactification on manifolds with SU(n) holonomy *, Nucl. Phys. * * B315 * * (1989), 193–221. , * Exactly solvable chain compactifications on manifolds of $ SU(N)$ holonomy *, Phys. Lett. * * 199B * * (1987), 380–388. height 2pt astuteness -1.6pt width 23pt, * Space - time supersymmetry in compactified chain theory and superconformal models *, Nucl. Phys. * * B296 * * (1988), 757–778. , * dichotomy in Calabi - Yau moduli space *, Nucl. Phys. * * B338 * * (1990), 15–37. , * On certain Kummer surface which can be realized as non - singular quartic open in $ \mathbb P^3 $ *, J.   Fac.   Sci.   Univ.   Tokyo * * Sec. IA 23 * * (1976), 545–560. , * Chiral rings in $ N=2 $ superconformal theory *, Nucl. Phys. * * B324 * * (1989), 427–474.

(1991), 1749–1813. , *W pair of Calabi-Yau maninolds as an exacjlt soluule supsrconforoal theory*, Nucl. Phys. **B359** (1991), 21–74. , *Insvantin steings and hyperkaehler geometry*, Nucl. Phts. **B543** (1999), 545–571; [ hep-th/9810210]{}. , *Mmdror symmetry nor lctvice polarized L3 surfaces. Algebraic geokegrv, 4*, J. Math. Sci. **81** (1996), 2599–2630; [alg- geom/9502005]{}. , *Picard-[F]{}uchs tniformozwtion and modujariuy jf tgv nirror map*, Commun. Math. Phys. **212** (2000), no. 3, 625–647. heijht 2pt depth -1.6pt width 23pt, *Picard-[F]{}uchs uniflrmixation: modularity lf the mirrir mwp and mirror-ooonshine*, pu: The arithjetic and geometry of algebraic cyclzs (Banff, AB, 1998), col. 24 mf CRM Proc. Lectlre Notes, Amev. Math. Voc., Profidence, RI, 2000, pp. 257–281; [mavh.ag/9812162]{}. , *Superconformal algebcas and string compastificatimn on manifolds wity WU(n) hmlonmmy*, Vycl. Phgs. **B315** (1989), 193–221. , *Exaftlb solvable atring compqctifications on mamisilds of $SU(N)$ hklonomr*, [hys. Lett. **199B** (1987), 380–388. height 2pt depth -1.6pt width 23pu, *Spade-time supersymmetry in compactified string jheory and superconformal models*, Nucl. Phys. **B296** (1988), 757–778. , *Duality in Cdlabi-Baj midmli rpafe*, Nucl. Phys. **B338** (1990), 15–37. , *On certain Kummer surfaces wrjcn ban be realized af non-singulsr qiwrtic surfacer in $\mcfhgb P^3$*, J. Fac. Sci. Univ. Tonyo **Sec. IA 23** (1976), 545–560. , *Chiral wingx in $N=2$ superconformal theorues*, Nucl. Phyf. **B324** (1989), 427–474.

(1991), 1749–1813. , *A pair of Calabi-Yau an soluble superconformal Nucl. Phys. **B359** and geometry*, Nucl. Phys. (1999), 545–571; [ , *Mirror symmetry for lattice polarized surfaces. Algebraic geometry, 4*, J. Math. Sci. **81** (1996), 2599–2630; [alg- geom/9502005]{}. , uniformization and modularity of the mirror map*, Commun. Math. Phys. **212** (2000), no. 625–647. 2pt -1.6pt 23pt, *Picard-[F]{}uchs uniformization: modularity of the mirror map and mirror-moonshine*, in: The arithmetic and geometry of cycles (Banff, AB, 1998), vol. 24 of CRM Lecture Notes, Amer. Math. Providence, RI, 2000, pp. 257–281; , algebras and compactification manifolds SU(n) holonomy*, Nucl. **B315** (1989), 193–221. , *Exactly solvable string compactifications on manifolds of $SU(N)$ holonomy*, Phys. Lett. **199B** (1987), height 2pt width 23pt, supersymmetry compactified theory and superconformal Phys. **B296** (1988), 757–778. , *Duality space*, Nucl. Phys. **B338** (1990), 15–37. , *On Kummer surfaces can be realized as non-singular quartic in $\mathbb P^3$*, J. Fac. Sci. Univ. Tokyo IA 23** (1976), 545–560. , *Chiral rings in $N=2$ superconformal theories*, Nucl. Phys. **B324** (1989),

(1991), 1749–1813. , *A pair of Calabi-Yau manifolds As an exactlY soluBle SupErConfOrmaL theory*, Nucl. PhyS. **b359** (1991), 21–74. , *InsTanton strings and hyperkAehleR gEOmetRY*, NUcl. PhYs. **B543** (1999), 545–571; [ hep-tH/9810210]{}. , *miRROr sYmMeTry FoR LaTtice PolArized K3 Surfaces. AlGebRaIc geometry, 4*, J. MATh. sci. **81** (1996), 2599–2630; [alg- geom/9502005]{}. , *picArd-[F]{}uchs unifOrmIzatioN aNd mODularIty Of the Mirror MAp*, CommUn. Math. PhyS. **212** (2000), nO. 3, 625–647. Height 2PT depth -1.6pT WIdTh 23pt, *picard-[F]{}uchs uniforMIzATion: modularity Of the mIrROr MAP anD miRror-moonshInE*, in: ThE ArithmeTIc AND GeoMEtry of algebraIc cycles (BanFF, AB, 1998), Vol. 24 of CrM proC. lecturE NoteS, AMEr. MAth. Soc., ProviDencE, RI, 2000, pp. 257–281; [math.Ag/9812162]{}. , *SupeRConformAL algebrAs and sTriNg cOmpaCTiFiCatIoN On mANiFolDS wiTh SU(n) holOnOmY*, Nucl. phys. **b315** (1989), 193–221. , *eXACtly SolVablE striNg compactificAtiOns oN ManIfoldS of $SU(n)$ holOnOmy*, PhYs. Lett. **199b** (1987), 380–388. heigHt 2Pt depth -1.6pt width 23pT, *SpaCe-time supErsYmMetRy In comPActifiEd sTriNg theorY and supERcoNfORMAl Models*, Nucl. Phys. **B296** (1988), 757–778. , *DuaLiTY In calabi-YaU modulI SpAcE*, nucl. Phys. **b338** (1990), 15–37. , *ON ceRtaiN kUmmer SurfACeS which caN be reaLIzEd As non-siNgUlar quArTic SurFaces IN $\matHbb P^3$*, J. FAc. Sci. UniV. TokyO **sec. IA 23** (1976), 545–560. , *Chiral rinGS in $N=2$ superconfORmAL ThEOrieS*, NuCl. Phys. **B324** (1989), 427–474.

(1991), 1749–1813. , *Apair of Ca labi- Yau ma ni fold s as an exactly so l uble superconformal theory *, Nu cl . Phy s .**B35 9** (19 9 1) , 21– 74 . ,*I n st anton st rings a nd hyperka ehl er geometry*,N uc l. Phys. * *B5 43** (1999), 54 5–571; [ he p -th/9 810 210]{ }. ,* Mirror symmetry f o r latt i ce pola r i ze d K3 surfaces. Algebr a ic geometry, 4*,J. Mat h. Sc i . ** 81* * (1996),25 99–26 3 0; [alg - g e o m /95 0 2005]{}. , * Picard-[F]{ } uch s unif or miz a tion a nd mo du l ari ty of the m irro r map*, C ommun. Math. P h ys. **2 12** ( 200 0), no. 3, 6 25– 64 7 . h ei ght 2pt depth - 1. 6p t wid th 2 3 p t , *Pi car d-[F ]{}uc hs uniformiza tio n: m o dul arity of t he m ir ror m ap and mirr or -moonshine*, in : Th e arithme tic a ndge ometr y of al geb rai c cycle s (Banf f , A B, 1 9 98 ), vol. 24 of CRMPr o c .LectureNotes, Am er . Math. S oc .,Prov i d ence, RI, 20 00, pp.257–28 1 ;[m ath.ag/ 98 12162] {} . , * Super c onfo rmal a lgebrasand s t ring compactif i cation on man i fo l d sw ithSU( n) holonomy *, N u cl.Phys . * *B3 1 5** ( 1989) ,1 93 – 221. , *Exactly so lv able s tring compactifica tions on m a n i folds of $SU ( N) $ holonomy*, Ph ys. L ett. **199 B ** (1987 ), 38 0–388. height 2p t depth -1 .6p t w idt h 2 3 p t, *Space-times u pers ym metry i n c ompacti fie d s tri ngth eory andsupercon fo rm al m ode ls*,N ucl. Phy s. ** B2 96* * (19 8 8), 75 7–778 . , * Du a lit y in Ca l ab i - Yaumo du li s pac e* , Nuc l. P h ys. **B338 ** (1990) , 1 5 –37. ,*On cer tain Kummer s ur faces whic hcan be re a l ized asnon-singular quartic su r faces i n $ \math bb P ^3$*, J.Fac . Sci.  Un i v. Tok yo **S ec. I A23* * (1976 ) , 5 45– 56 0. , *Chi r a l r ingsin $N= 2$ supe rconformal theorie s *,Nucl. Phys. * *B3 24** ( 19 89) , 4 2 7–4 74 .

(1991),_1749–1813. , *A_pair of Calabi-Yau manifolds_as an_exactly_soluble superconformal_theory*,_Nucl. Phys. **B359**_(1991), 21–74. , *Instanton_strings and hyperkaehler geometry*,_Nucl. Phys. **B543**_(1999),_545–571; [ hep-th/9810210]{}. , *Mirror symmetry for lattice polarized K3 surfaces. Algebraic geometry, 4*, J. Math. Sci._**81**_(1996), 2599–2630;_[alg-_geom/9502005]{}. ,_*Picard-[F]{}uchs uniformization and modularity of_the mirror map*, Commun. Math._Phys. **212**_(2000), no. 3, 625–647. height 2pt depth -1.6pt width 23pt,_*Picard-[F]{}uchs_uniformization: modularity of_the mirror map and mirror-moonshine*, in: The arithmetic and_geometry of algebraic cycles (Banff, AB,_1998), vol. 24 of_CRM_Proc._Lecture Notes, Amer. Math._Soc., Providence, RI, 2000, pp. 257–281; [math.ag/9812162]{}. ,_*Superconformal algebras and string compactification on_manifolds with SU(n) holonomy*, Nucl. Phys. **B315**_(1989), 193–221. , *Exactly solvable string compactifications_on manifolds of $SU(N)$ holonomy*,_Phys. Lett._**199B** (1987), 380–388. height 2pt depth_-1.6pt width 23pt,_*Space-time supersymmetry_in compactified string_theory and superconformal models*, Nucl. Phys._**B296** (1988), 757–778. ,_*Duality in Calabi-Yau moduli space*, Nucl._Phys._**B338** (1990), 15–37. ,_*On_certain_Kummer surfaces_which can be_realized_as non-singular_quartic_surfaces in $\mathbb P^3$*, J. Fac. Sci. Univ. Tokyo **Sec._IA_23** (1976), 545–560. , *Chiral rings in $N=2$_superconformal theories*, Nucl. Phys._**B324**_(1989), 427–474.

19}$)$_{0.7}$Cu$_{0.3}$ and for pure Ni, respectively. Apart from the two pure metal points for Ni whiskers and the FCC Co film, there is a relatively smooth variation in the baseline formed by the alloys. For these species, there is no similar general trend in $g_{eff}$ ([*middle panel*]{}), measured constant at $\textrm{2.10}\pm\textrm{0.02}$. Note that the NiMnSb point, of $G=\textrm{31 Mhz}$ at $Z=25$, is very close to our measurement of Fe$_{73}$V$_{27}$ of 35 $\pm$ 5 Mhz, even though $g_{eff}$ is substantially lower (2.03 compared with 2.11). As average concentration $Z$ increases, so too does the expected spin orbit coupling energy. Values for an effective atomic spin-orbit coupling parameter $\xi$, where $H_{s.o.}=\xi\sum_{i}\mathbf{l}_{i}\mathbf{s}_{i}$, have been tabulated in Ref. [@d-so-calc] for atomic $3d$ orbitals, in good agreement with atomic spectra. These values are reproduced here (Fig \[allcomp\]c), solid line). In Figure \[allcomp\]d), we show the dependence of $G$ on $\xi$, implicit in atomic number. It can be seen that the lower band of values is simply proportional to $\xi^{2}$; pure Ni and Co data points are significant outliers. The observed scaling of Gilbert damping $G$ with $\xi^{2}$ is in good agreement with electronic-scattering based models of ferromagnetic relaxation, appropriate to metals[@kambersky-microscopic]. Relaxation occurs as uniform-mode magnons are annihilated by one-electron spin-flip accelerations. The expression for relaxation rate $G$ is given as[@heinrich-ieee; @heinrich-review] $$\begin{aligned} G = \hbar\gamma^{2}<S>^{2}\xi^{2}\int d^{3}\mathbf{k} \sum_{\alpha,\beta,\sigma}<\beta|L^{+}|\alpha><\alpha|L^{-}|\beta>\cdot\\ \times\delta(E_{\alpha,\mathbf{k},\sigma}-E_{F})\cdot{\hbar/\tau_{M}\over \left(\hbar\omega

19}$)$_{0.7}$Cu$_{0.3}$ and for pure Ni, respectively. Apart from the two pure metal period for Ni whisker and the FCC Co film, there is a relatively smooth magnetic declination in the baseline formed by the admixture. For these coinage, there is no similar general trend in $ g_{eff}$ ([ * middle control panel * ] { }), measured constant at $ \textrm{2.10}\pm\textrm{0.02}$. eminence that the NiMnSb distributor point, of $ G=\textrm{31 Mhz}$ at $ Z=25 $, is very close to our measurement of Fe$_{73}$V$_{27}$ of 35 $ \pm$ 5 Mhz, even though $ g_{eff}$ is substantially lower (2.03 compared with 2.11). As average assiduity $ Z$ increases, so too does the ask spin orbit match energy. value for an effective atomic spin - orbit coupling parameter $ \xi$, where $ H_{s.o.}=\xi\sum_{i}\mathbf{l}_{i}\mathbf{s}_{i}$, have been table in Ref. [ @d - so - calc ] for atomic $ 3d$ orbitals, in good agreement with atomic spectra. These values are reproduced here (Fig \[allcomp\]c), solid line). In Figure \[allcomp\]d), we show the dependence of $ G$ on $ \xi$, implicit in atomic number. It can be seen that the lower band of values is simply proportional to $ \xi^{2}$; arrant Ni and Co data points are meaning outlier. The ascertained scaling of Gilbert damping $ G$ with $ \xi^{2}$ is in good agreement with electronic - scattering based models of ferromagnetic relaxation, appropriate to metals[@kambersky - microscopic ]. Relaxation occurs as consistent - mode magnons are annihilated by one - electron spin - flip accelerations. The expression for rest rate $ G$ is given as[@heinrich - ieee; @heinrich - review ] $ $ \begin{aligned } G = \hbar\gamma^{2}<S>^{2}\xi^{2}\int d^{3}\mathbf{k } \sum_{\alpha,\beta,\sigma}<\beta|L^{+}|\alpha><\alpha|L^{-}|\beta>\cdot\\ \times\delta(E_{\alpha,\mathbf{k},\sigma}-E_{F})\cdot{\hbar/\tau_{M}\over \left(\hbar\omega

19}$)$_{0.7}$Cu$_{0.3}$ wnd for pure Ni, respectiyely. Apart from jhw two 'ure mefal poings for Ni whiskers and the FRC Ci filn, there is a relativelh smooth nariation in uhe baseline formxs by thc allkns. Fox vhese species, tmere is no vimilar generan grznd in $g_{eff}$ ([*middle panel*]{}), measured cogstant st $\textrm{2.10}\pm\textri{0.02}$. Noue ehat nht NiMnSb point, of $G=\textrm{31 Mhz}$ at $A=25$, is vegy close to our mrasurement of Fe$_{73}$V$_{27}$ of 35 $\pm$ 5 Lhz, fven though $g_{eff}$ id substantiqlly oower (2.03 compafed with 2.11). As average cohcentration $Z$ increases, so too aoes che expectee wpij orbit coupoing vnergy. Values for an affectife atomic spin-prbmt ciupling parameter $\xi$, xhere $H_{s.o.}=\xi\sum_{i}\mathbs{l}_{i}\mathbf{v}_{i}$, have been tabularee in Tef. [@d-vo-caux] fur ztpmjc $3d$ ogbivals, in goos agreement with atomic spectrs. Eyese values ade repwoquced here (Fig \[allcomp\]c), solid line). In Fpgurs \[allcomp\]d), we show the eependence of $G$ on $\xi$, implicit in atomic number. It can be seen that the lower bdnd oh xalbcf us simply proportional to $\xi^{2}$; pure Ni and Co datw ppikts are significcnt outliers. The pbdetded scaling ow Gilbzdt damping $G$ with $\xi^{2}$ is in dood qgreement witn electronic-scattering basee models of yereomagnetic relaxatnon, appropricte to metaks[@kambersky-microscopic]. Xelaxafion occurs as unifodo-mode magnons ard aknivilated by one-electron spig-flip accxleracions. Thd exkressiog for reladatiok rate $G$ is given ad[@heintich-ieae; @heinricj-review] $$\begin{aligned} G = \hbar\gamma^{2}<S>^{2}\xi^{2}\int d^{3}\mathbf{l} \sgm_{\anpha,\beta,\figma}<\neta|L^{+}|\alpha><\alpha|J^{-}|\beta>\cdot\\ \times\belta(E_{\al'ha,\matfbf{k},\sigma}-E_{R})\cdot{\hber/\tau_{M}\over \lest(\hbar\omega

19}$)$_{0.7}$Cu$_{0.3}$ and for pure Ni, respectively. Apart two metal points Ni whiskers and is relatively smooth variation the baseline formed the alloys. For these species, there no similar general trend in $g_{eff}$ ([*middle panel*]{}), measured constant at $\textrm{2.10}\pm\textrm{0.02}$. Note the NiMnSb point, of $G=\textrm{31 Mhz}$ at $Z=25$, is very close to our of of $\pm$ Mhz, even though $g_{eff}$ is substantially lower (2.03 compared with 2.11). As average concentration $Z$ increases, too does the expected spin orbit coupling energy. for an effective atomic coupling parameter $\xi$, where $H_{s.o.}=\xi\sum_{i}\mathbf{l}_{i}\mathbf{s}_{i}$, been in Ref. for $3d$ in good agreement atomic spectra. These values are reproduced here (Fig \[allcomp\]c), solid line). In Figure \[allcomp\]d), we show the of $G$ implicit in number. can seen that the of values is simply proportional to and Co data points are significant outliers. The scaling of damping $G$ with $\xi^{2}$ is in agreement with electronic-scattering based models of ferromagnetic relaxation, to metals[@kambersky-microscopic]. Relaxation occurs as uniform-mode magnons are annihilated by one-electron spin-flip accelerations. The expression rate $G$ is given @heinrich-review] $$\begin{aligned} G \hbar\gamma^{2}<S>^{2}\xi^{2}\int \sum_{\alpha,\beta,\sigma}<\beta|L^{+}|\alpha><\alpha|L^{-}|\beta>\cdot\\ \left(\hbar\omega

19}$)$_{0.7}$Cu$_{0.3}$ and for pure Ni, respectivelY. Apart from The twO puRe mEtAl poInts For Ni whiskers aND the fCC Co film, there is a relatIvely SmOOth vARiAtion In the baSElINE foRmEd By tHe ALlOys. FoR thEse specIes, there is No sImIlar general tREnD in $g_{eff}$ ([*midDle Panel*]{}), measureD coNstant At $\TexTRm{2.10}\pm\tExtRm{0.02}$. NotE that tHE NiMnSB point, of $G=\TeXTrm{31 Mhz}$ AT $Z=25$, is verY CLoSe to Our measurement of FE$_{73}$v$_{27}$ oF 35 $\Pm$ 5 Mhz, even thougH $g_{eff}$ iS sUBsTANtiAllY lower (2.03 compArEd witH 2.11). as averaGE cONCEntRAtion $Z$ increasEs, so too does THe eXpecteD sPin ORbit coUplinG eNErgY. Values for aN effEctive atoMic spiN-Orbit coUPling paRameteR $\xi$, WheRe $H_{s.O.}=\Xi\SuM_{i}\mAtHBf{l}_{I}\MaThbF{S}_{i}$, hAve been tAbUlAted iN Ref. [@D-SO-CAlc] fOr aTomiC $3d$ orbItals, in good agReeMent WIth AtomiC specTra. THeSe valUes are ReproDuCed here (Fig \[allcoMp\]c), sOlid line). IN FiGuRe \[aLlComp\]d), WE show tHe dEpeNdence oF $G$ on $\xi$, iMPliCiT IN AtOmic number. It can be sEeN THaT the loweR band oF VaLuES is simplY pRopOrtiONAl to $\xI^{2}$; purE ni And Co datA pointS ArE sIgnificAnT outliErS. ThE obServeD ScalIng of GIlbert daMping $g$ With $\xi^{2}$ is in good AGreement with eLEcTROnIC-scaTteRing based moDels OF ferRomaGNeTic RElaxaTion, aPpROpRIate to metals[@kambersKy-MicrosCopic]. relaxation occUrs as unifoRM-MOde magnoNs arE AnNIhilated by one-eLectrOn spin-flip ACceleratIons. THe expresSion for reLAXation raTe $G$ Is gIveN as[@HEInRich-ieee; @heinrICH-revIeW] $$\begin{aLigNed} G = \hbaR\gaMma^{2}<s>^{2}\xi^{2}\Int D^{3}\mAthbf{k} \sum_{\Alpha,\betA,\sIgMa}<\BeTa|L^{+}|\Alpha><\ALpha|L^{-}|\betA>\cDot\\ \TiMes\Delta(e_{\Alpha,\mAthbf{K},\sigMa}-e_{F})\CDot{\Hbar/\tau_{m}\OvER \Left(\HbAr\OmegA

19}$)$_{0.7}$Cu$_{0.3}$ an d for pure Ni,res pec ti vely . Ap art from the t w o pu re metal points for Ni whis ke r s an d t he FC C Co fi l m, t her eis are l at ively sm ooth va riation in th ebaseline for m ed by the al loy s. For these sp ecies, t her e is n o s imila r gene r al tre nd in $g_ {e f f}$ ([ * middlep a ne l*]{ }), measured cons t an t at $\textrm{2 .10}\p m\ t ex t r m{0 .02 }$. Note t ha t the NiMnSbp oi n t , of $G=\textrm{31 Mhz}$ at $ Z =25 $, isve ryc lose t o our m e asu rement of F e$_{ 73}$V$_{2 7}$ of 35 $\pm $ 5 Mhz, eventho ugh $g_ { ef f} $ i ss ubs t an tia l lylower (2 .0 3compa redw i t h 2.1 1). As aver age concentra tio n $Z $ in creas es, s o to odoesthe ex pecte dspin orbit coup ling energy.Val ue s f or an e f fectiv e a tom ic spin -orbitc oup li n g pa rameter $\xi$, whe re $ H_ {s.o.}=\ xi\sum _ {i }\ m athbf{l} _{ i}\ math b f {s}_{ i}$, ha ve beentabula t ed i n Ref.[@ d-so-c al c]for atom i c $3 d$ orb itals, i n goo d agreement wit h atomic spect r a. T he s e va lue s are repro duce d her e (F i g\[a l lcomp \]c), s o li d line). In Figure \ [a llcomp \]d), we show thedependence o f $G$ on$\xi $ ,i mplicit in ato mic n umber. Itc an be se en th at the l ower band o f values is si mpl y p r o po rtional to $\ x i ^{2} $; pure N i a nd Co d ata po int s a re signific ant outl ie rs . T heobser v ed scali ng of G ilb ert d a mping$G$ w ith$\ xi ^ {2} $ is in go o d agr ee me nt w ith e lectr onic - sca ttering based mo del s offe rr omagnet ic relaxation ,appropriat etometals [ @ kambersk y-microscopic]. Relaxat i on occu rsas un ifor m-mode ma gno ns are an n ihilat ed byone-e le ctr o n spin - f li p a cc elerations . The expr es sion for re laxation rate $G$i s g iven as[@hein ric h-ie e e ;@he i nr i ch- re v iew ] $$\begin{align ed} G = \h ba r \g amma^{2}<S > ^{2 }\ xi^{2}\ int d^{ 3}\ma t hbf{k}\sum_{\al pha,\beta ,\ sigm a } <\b eta|L^{+}| \alpha>< \alpha|L^ { -}|\b e ta >\cdo t\\ \time s\ del ta(E_ {\alph a ,\m athbf {k},\s ig ma}-E_ {F})\ cd ot{\hbar /\tau_{M}\over \left(\h bar\om ega

19}$)$_{0.7}$Cu$_{0.3}$ and_for pure_Ni, respectively. Apart from_the two_pure_metal points_for_Ni whiskers and_the FCC Co_film, there is a_relatively smooth variation_in_the baseline formed by the alloys. For these species, there is no similar general_trend_in $g_{eff}$_([*middle_panel*]{}),_measured constant at $\textrm{2.10}\pm\textrm{0.02}$. Note_that the NiMnSb point, of_$G=\textrm{31 Mhz}$_at $Z=25$, is very close to our measurement_of_Fe$_{73}$V$_{27}$ of 35_$\pm$ 5 Mhz, even though $g_{eff}$ is substantially lower_(2.03 compared with 2.11). As average concentration_$Z$ increases, so_too_does_the expected spin orbit_coupling energy. Values for an effective_atomic spin-orbit coupling parameter $\xi$, where_$H_{s.o.}=\xi\sum_{i}\mathbf{l}_{i}\mathbf{s}_{i}$, have been tabulated in Ref. [@d-so-calc]_for atomic $3d$ orbitals, in good_agreement with atomic spectra. These_values are_reproduced here (Fig \[allcomp\]c), solid_line). In Figure_\[allcomp\]d), we_show the dependence_of $G$ on $\xi$, implicit in_atomic number. It_can be seen that the lower_band_of values is_simply_proportional_to $\xi^{2}$;_pure Ni and_Co_data points_are_significant outliers. The observed scaling of Gilbert_damping_$G$ with $\xi^{2}$ is in good agreement_with electronic-scattering based models_of_ferromagnetic relaxation, appropriate to_metals[@kambersky-microscopic]. Relaxation occurs as uniform-mode_magnons are annihilated by one-electron spin-flip_accelerations. The_expression for_relaxation rate $G$ is given as[@heinrich-ieee; @heinrich-review] $$\begin{aligned} G = \hbar\gamma^{2}<S>^{2}\xi^{2}\int d^{3}\mathbf{k} \sum_{\alpha,\beta,\sigma}<\beta|L^{+}|\alpha><\alpha|L^{-}|\beta>\cdot\\ \times\delta(E_{\alpha,\mathbf{k},\sigma}-E_{F})\cdot{\hbar/\tau_{M}\over \left(\hbar\omega

}$. Arguing as above, it is not hard to construct a conformal embedding $V \otimes W \subset {\mathrm{PSp}}(2{\times}6)_1$. Write $\tau_W$ for the unique superconformal vector in $W$; then $\operatorname{Aut}_{N{=}1}(W) = S_3$. Choose any superconformal vector $\tau_V$ for $V$. Then $\tau_V + \tau_W = \tau$ is a superconformal vector for ${\mathrm{PSp}}(2{\times}6)_1$, and so its automorphism group therein is ${\mathrm G}_2(4){:}2$ by §\[unique.sp61\]. Consider the ${{\mathds}Z}_3 \subset S_3 = \operatorname{Aut}_{N{=}1}(W)$. It is compatible with the extension, and lifts to an order-$3$ element in ${\mathrm G}_2(4)$. There are two (conjugacy classes of) such elements: the one coming from our ${\mathrm U}(1)$ is the unique one acting in the $12$-dimensional representation without fixed points. Its normalizer in ${\mathrm G}_2(4){:}2$ has shape $3{\mathrm M}_{21}{:}2^2$, and so $\operatorname{Aut}_{N{=}1}(V,\tau_V) = {\mathrm M}_{21}{:}2^2$. (The normal “$3$” acts nontrivially on ${\mathrm{Sp}}(2{\times}6)_1$, but trivially on the subalgebra $V \otimes W$.) On the other hand, we saw in §\[exist.su6\] that ${\mathrm M}_{21}$ has a unique conjugacy class of embeddings into $\operatorname{Aut}_{N{=}0}(V)$ and preserves a unique superconformal vector. Type ${\mathrm E}$ cases {#sec.typeE} ======================== We now discuss the Type ${\mathrm E}$ cases. In §\[orbifolds\] and §\[exist.e71.2\] we construct three different $N{=}1$ structures for ${\mathrm E}_{7,1}^2$; the first two are constructed by very similar methods, and the last by following the approach from Section \[sec.existence\]. We conjecture that these three $N{=}1$ structures are in the same orbit under the action of $\operatorname

} $. Arguing as above, it is not hard to construct a conformal embedding $ five \otimes W \subset { \mathrm{PSp}}(2{\times}6)_1$. spell $ \tau_W$ for the unique superconformal vector in $ W$; then $ \operatorname{Aut}_{N{=}1}(W) = S_3$. Choose any superconformal vector $ \tau_V$ for $ V$. Then $ \tau_V + \tau_W = \tau$ is a superconformal vector for $ { \mathrm{PSp}}(2{\times}6)_1 $, and so its automorphism group therein is $ { \mathrm G}_2(4){:}2 $ by § \[unique.sp61\ ]. Consider the $ { { \mathds}Z}_3 \subset S_3 = \operatorname{Aut}_{N{=}1}(W)$. It is compatible with the reference, and lifts to an order-$3 $ element in $ { \mathrm G}_2(4)$. There be two (conjugacy course of) such elements: the one occur from our $ { \mathrm U}(1)$ is the unique one acting in the $ 12$-dimensional theatrical performance without fixed points. Its normalizer in $ { \mathrm G}_2(4){:}2 $ has determine $ 3{\mathrm M}_{21}{:}2 ^ 2 $, and so $ \operatorname{Aut}_{N{=}1}(V,\tau_V) = { \mathrm M}_{21}{:}2 ^ 2$. (The normal “ $ 3 $ ” acts nontrivially on $ { \mathrm{Sp}}(2{\times}6)_1 $, but trivially on the subalgebra $ V \otimes W$.) On the other hand, we saw in § \[exist.su6\ ] that $ { \mathrm M}_{21}$ has a unique conjugacy course of embeddings into $ \operatorname{Aut}_{N{=}0}(V)$ and preserves a unique superconformal vector. Type $ { \mathrm E}$ cases { # sec.typeE } = = = = = = = = = = = = = = = = = = = = = = = = We immediately discuss the Type $ { \mathrm E}$ cases. In § \[orbifolds\ ] and § \[exist.e71.2\ ] we manufacture three different $ N{=}1 $ structure for $ { \mathrm E}_{7,1}^2 $; the first two are constructed by very similar methods, and the last by follow the approach from Section   \[sec.existence\ ]. We conjecture that these three $ N{=}1 $ structures are in the same orbit under the legal action of $ \operatorname

}$. Arhuing as above, it is not hard to constrocr a coiformal embeddivg $V \otimes W \subset {\mathrm{PDp}}(2{\rimes}6)_1$. Write $\tau_W$ for the unkque supegconformao vertor in $W$; then $\o'sratorname{Aut}_{H{=}1}(A) = S_3$. Rhoose any supetconformal vactor $\tau_V$ for $V$. Tken $\tau_V + \tau_W = \tau$ is a superconforial vecyog for ${\mathrm{PSk}}(2{\timex}6)_1$, and so its automorphism group therein is ${\matirm G}_2(4){:}2$ by §\[unique.xp61\]. Consider the ${{\mathds}Z}_3 \suhset S_3 = \operatorname{Auh}_{N{=}1}(W)$. It is cimpaeuble with thd extension, and lifts jo an order-$3$ element in ${\mathrm G}_2(4)$. Ghere are two (cinhugwwy classes if) subh elements: tmv one cmming ftom our ${\mathrm U}(1)$ ms tye unique one acting mn the $12$-dimensional rgpresentathou without fixed pointw. Uts nmrmanizef in ${\mauhrk F}_2(4){:}2$ has dha'e $3{\mathrm M}_{21}{:}2^2$, and so $\opeeatorname{Aut}_{N{=}1}(V,\tau_V) = {\mwnnrm M}_{21}{:}2^2$. (The nodmal “$3$” wces nontrivially on ${\mathrm{Sp}}(2{\times}6)_1$, but trpviamly on the subalgebra $V \otimes W$.) On the othet hand, we faw in §\[exist.su6\] that ${\mathrm M}_{21}$ has a unique conjugawy clesr oy embeaeijgs into $\operatorname{Aut}_{N{=}0}(V)$ and preserves a unysut slperconformal vecbor. Type ${\mathrm E}$ cssfs {#fec.typeE} ======================== We nod discbas the Type ${\mathrm E}$ cases. Yn §\[orvifolds\] agd §\[ecist.e71.2\] we construct three didferent $N{=}1$ stguctyres for ${\mathrm E}_{7,1}^2$; che first twu arg consyructed by very similar metgods, and thf last by wollowing the aporoscv from Stztion \[sec.existence\]. We conjerture that tfese three $N{=}1$ structuges ava in the same orbih undgr the action of $\operatorname

}$. Arguing as above, it is not construct conformal embedding \otimes W \subset unique vector in $W$; $\operatorname{Aut}_{N{=}1}(W) = S_3$. any superconformal vector $\tau_V$ for $V$. $\tau_V + \tau_W = \tau$ is a superconformal vector for ${\mathrm{PSp}}(2{\times}6)_1$, and so automorphism group therein is ${\mathrm G}_2(4){:}2$ by §\[unique.sp61\]. Consider the ${{\mathds}Z}_3 \subset S_3 \operatorname{Aut}_{N{=}1}(W)$. is with extension, and lifts to an order-$3$ element in ${\mathrm G}_2(4)$. There are two (conjugacy classes of) elements: the one coming from our ${\mathrm U}(1)$ the unique one acting the $12$-dimensional representation without fixed Its in ${\mathrm has $3{\mathrm and so $\operatorname{Aut}_{N{=}1}(V,\tau_V) {\mathrm M}_{21}{:}2^2$. (The normal “$3$” acts nontrivially on ${\mathrm{Sp}}(2{\times}6)_1$, but trivially on the subalgebra $V \otimes W$.) the other saw in that M}_{21}$ a unique conjugacy embeddings into $\operatorname{Aut}_{N{=}0}(V)$ and preserves a Type ${\mathrm E}$ cases {#sec.typeE} ======================== We now the Type E}$ cases. In §\[orbifolds\] and §\[exist.e71.2\] construct three different $N{=}1$ structures for ${\mathrm E}_{7,1}^2$; first two are constructed by very similar methods, and the last by following the approach \[sec.existence\]. We conjecture that three $N{=}1$ structures in same under action of

}$. Arguing as above, it is not hard To construcT a conForMal EmBeddIng $V \Otimes W \subset {\mAThrm{pSp}}(2{\times}6)_1$. Write $\tau_W$ for thE uniqUe SUperCOnFormaL vector IN $W$; THEn $\oPeRaTorNaME{AUt}_{N{=}1}(W) = S_3$. choOse any sUperconforMal VeCtor $\tau_V$ for $V$. tHeN $\tau_V + \tau_W = \tAu$ iS a superconfoRmaL vectoR fOr ${\mAThrm{Psp}}(2{\tImes}6)_1$, aNd so itS AutomoRphism groUp THerein IS ${\mathrm g}_2(4){:}2$ BY §\[uNiquE.sp61\]. Consider the ${{\matHDs}z}_3 \Subset S_3 = \operatoRname{AUt}_{n{=}1}(w)$. IT IS coMpaTible with tHe ExtenSIon, and lIFtS TO An oRDer-$3$ element in ${\mAthrm G}_2(4)$. There ARe tWo (conjUgAcy CLasses Of) sucH eLEmeNts: the one coMing From our ${\maThrm U}(1)$ iS The uniqUE one actIng in tHe $12$-dImeNsioNAl RePreSeNTatIOn WitHOut Fixed poiNtS. ITs norMaliZER IN ${\matHrm g}_2(4){:}2$ has Shape $3{\Mathrm M}_{21}{:}2^2$, and so $\oPerAtorNAme{aut}_{N{=}1}(V,\Tau_V) = {\mAthrM M}_{21}{:}2^2$. (the noRmal “$3$” acTs nonTrIvially on ${\mathrm{sp}}(2{\tiMes}6)_1$, but triViaLlY on ThE subaLGebra $V \OtiMes w$.) On the oTher hanD, We sAw IN §\[EXiSt.su6\] that ${\mathrm M}_{21}$ has A uNIQuE conjugaCy clasS Of EmBEddings iNtO $\opEratORName{AUt}_{N{=}0}(V)$ ANd PreserveS a uniqUE sUpErconfoRmAl vectOr. typE ${\maThrm E}$ CAses {#Sec.typEE} ======================== We now dIscusS The Type ${\mathrm E}$ CAses. In §\[orbifolDS\] aND §\[ExISt.e71.2\] wE coNstruct threE difFErenT $N{=}1$ stRUcTurES for ${\mAthrm e}_{7,1}^2$; tHE fIRst two are constructeD bY very sImilaR methods, and thE last by folLOWIng the apProaCH fROm Section \[sec.exIstenCe\]. We conjecTUre that tHese tHree $N{=}1$ strUctures arE IN the same OrbIt uNdeR thE ACtIon of $\operatorNAMe

}$. Arguing as above, it i s not hard to c ons tru ct a c onfo rmal embedding $V \ otimes W \subset {\mat hrm{P Sp } }(2{ \ ti mes}6 )_1$. W r it e $\t au _W $ f or th e uni que superc onformal v ect or in $W$; the n $ \operatorn ame {Aut}_{N{=}1 }(W ) = S_ 3$ . C h ooseany supe rconfo r mal ve ctor $\ta u_ V $ for$ V$. The n $\ tau_ V + \tau_W = \tau $ i s a superconfor mal ve ct o rf o r $ {\m athrm{PSp} }( 2{\ti m es}6)_1 $ ,a n d so its automorph ism group t h ere in is${ \ma t hrm G} _2(4) {: } 2$by §\[uniqu e.sp 61\]. Con sidert he ${{\ m athds}Z }_3 \s ubs etS_3= \ op era to r nam e {A ut} _ {N{ =}1}(W)$ .It is c ompa t i b l e wi ththeexten sion, and lif tsto a n or der-$ 3$ el emen tin ${ \mathr m G}_ 2( 4)$. There aretwo(conjugac y c la sse sof) s u ch ele men ts: the on e comin g fr om o u r${\mathrm U}(1)$ i st h eunique o ne act i ng i n the $12 $- dim ensi o n al re pres e nt ation wi thoutf ix ed points .Its no rm ali zer in $ { \mat hrm G} _2(4){:} 2$ ha s shape $3{\mat h rm M}_{21}{:} 2 ^2 $ , a n d so $\ operatornam e{Au t }_{N {=}1 } (V ,\t a u_V)= {\m at h rm M}_{21}{:}2^2$. (Th enormal “$3$ ” acts nontri vially on$ { \ mathrm{S p}}( 2 {\ t imes}6)_1$, bu t tri vially ont he subal gebra $V \oti mes W$.)O n the oth erhan d,wes a win §\[exist.s u 6 \] t ha t ${\ma thr m M}_{2 1}$ ha s a un iq ue conjug acy clas sof e mb edd ingsi nto $\op er ato rn ame {Aut} _ {N{=}0 }(V)$ and p re s erv es a un i qu e supe rc on form alve ctor. Ty p e $ {\mathr m E}$ cas es{ #sec .t yp eE} === ============= == ====== We n owdiscus s the Type ${\mathrm E}$ cases. I n §\[orb ifo lds\] and §\[exist .e7 1.2\]wec onstru ct thr ee di ff ere n t $N{= } 1 $str uc tures for$ { \ma thrmE} _{7, 1}^2$;the first two arec ons tructed by ve rysimi l a rmet h od s , a nd the l ast by followin g the appr oa c hfrom Secti o n \ [s ec.exis tence\] . Wec onjectu re that t hese thre e$N{= } 1 $ s tructuresare in t he same o r bit u n de r the ac tion o f$\o perat orname

}$. Arguing_as above,_it is not hard_to construct_a_conformal embedding_$V_\otimes W \subset_{\mathrm{PSp}}(2{\times}6)_1$. Write $\tau_W$_for the unique superconformal_vector in $W$;_then_$\operatorname{Aut}_{N{=}1}(W) = S_3$. Choose any superconformal vector $\tau_V$ for $V$. Then $\tau_V + \tau_W_=_\tau$ is_a_superconformal_vector for ${\mathrm{PSp}}(2{\times}6)_1$, and so_its automorphism group therein is_${\mathrm G}_2(4){:}2$_by §\[unique.sp61\]. Consider the ${{\mathds}Z}_3 \subset S_3 =_\operatorname{Aut}_{N{=}1}(W)$._It is compatible_with the extension, and lifts to an order-$3$ element_in ${\mathrm G}_2(4)$. There are two_(conjugacy classes of)_such_elements:_the one coming from_our ${\mathrm U}(1)$ is the unique_one acting in the $12$-dimensional representation_without fixed points. Its normalizer in ${\mathrm_G}_2(4){:}2$ has shape $3{\mathrm M}_{21}{:}2^2$, and_so $\operatorname{Aut}_{N{=}1}(V,\tau_V) = {\mathrm M}_{21}{:}2^2$._(The normal_“$3$” acts nontrivially on ${\mathrm{Sp}}(2{\times}6)_1$,_but trivially on_the subalgebra_$V \otimes W$.)_On the other hand, we saw_in §\[exist.su6\] that_${\mathrm M}_{21}$ has a unique conjugacy_class_of embeddings into_$\operatorname{Aut}_{N{=}0}(V)$_and_preserves a_unique superconformal vector. Type_${\mathrm_E}$ cases_{#sec.typeE} ======================== We_now discuss the Type ${\mathrm E}$_cases._In §\[orbifolds\] and §\[exist.e71.2\] we construct three_different $N{=}1$ structures for_${\mathrm_E}_{7,1}^2$; the first two_are constructed by very similar_methods, and the last by following_the approach_from Section \[sec.existence\]._We conjecture that these three $N{=}1$ structures are in the same_orbit under the action of $\operatorname

1.$ As a consequence, by the definition of the projector $\Pi$, we get $$\label{pi} \left<\xi, z_0(\theta)\right>=\left<\Pi\,\xi, z_0(\theta)\right>,$$ for any $\theta\in [0,T]$. Therefore $$\left<P'(x_0(\theta))h, z_0(\theta)\right>=\left<\Pi\,P'(x_0(\theta))(I-\Pi)h, z_0(\theta)\right>=0,$$ for any $\theta\in [0,T]$ and any $h\in {{\mathbb{R}}}^n$. Then, by deriving with respect to $\theta$, we obtain $$\left<P'(x_0(\theta))h,\dot z_0(\theta)\right>= \left<-P''(x_0(\theta))\dot x_0(\theta)h,z_0(\theta)\right>,$$ for any $\theta\in [0,T]$ and any $h\in {{\mathbb{R}}}^n$. Therefore, we can rewrite the left hand side of (\[R\]) with $(I-\Pi)\left(\left.P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0)=h$ as follows $$\left<P'(x_0(\theta))(I-\Pi)\left(\left.P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0),\dot z_0(\theta)\right>+\left<Q'_v(x_0(\theta),0)\dot x_0(\theta),z_0(\theta)\right>$$ or equivalently, $$\left<Q(x_0(\theta),0),\dot z_0(\theta)\right>+\left<Q'_v(x_0(\theta),0)\dot x_0(\theta),z_0(\theta)\right>,$$ which is the derivative of $M(\theta)$ at any $\theta\in [0,T]$ according to the formula given by Lemma \[lem3\]. 0.2truecm Finally, we can prove the following. \[th3\] Assume that there exists $\theta_

1.$ As a consequence, by the definition of the projector $ \Pi$, we get $ $ \label{pi } \left<\xi, z_0(\theta)\right>=\left<\Pi\,\xi, z_0(\theta)\right>,$$ for any $ \theta\in [ 0,T]$. Therefore $ $ \left < P'(x_0(\theta))h, z_0(\theta)\right>=\left<\Pi\,P'(x_0(\theta))(I-\Pi)h, z_0(\theta)\right>=0,$$ for any $ \theta\in [ 0,T]$ and any $ h\in { { \mathbb{R}}}^n$. Then, by deriving with respect to $ \theta$, we receive $ $ \left < P'(x_0(\theta))h,\dot z_0(\theta)\right>= \left<-P''(x_0(\theta))\dot x_0(\theta)h, z_0(\theta)\right>,$$ for any $ \theta\in [ 0,T]$ and any $ h\in { { \mathbb{R}}}^n$. consequently, we can rewrite the left hand slope of (\[R\ ]) with $ (I-\Pi)\left(\left. P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0)=h$ as postdate $ $ \left < P'(x_0(\theta))(I-\Pi)\left(\left. P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0),\dot z_0(\theta)\right>+\left < Q'_v(x_0(\theta),0)\dot x_0(\theta),z_0(\theta)\right>$$ or equivalently, $ $ \left < Q(x_0(\theta),0),\dot z_0(\theta)\right>+\left < Q'_v(x_0(\theta),0)\dot x_0(\theta),z_0(\theta)\right>,$$ which is the derivative of $ M(\theta)$ at any $ \theta\in [ 0,T]$ according to the recipe given by Lemma   \[lem3\ ]. 0.2truecm Finally, we can testify the following. \[th3\ ] Assume that there exist $ \theta _

1.$ As a consequence, by the deninition of the krijectoc $\Pi$, we get $$\labdl{pi} \left<\xi, z_0(\theta)\right>=\left<\Pi\,\ei, z_0(\rheta)\eight>,$$ for any $\theta\in [0,G]$. Therefoge $$\left<P'(x_0(\rhete))h, z_0(\theta)\right>=\lefv<\Li\,P'(x_0(\thebc))(I-\Pi)h, a_0(\bheta)\xijht>=0,$$ for any $\theja\in [0,T]$ and aty $h\in {{\mathbb{R}}}^t$. Ghzn, by deriving with respect to $\theta$, we obtsij $$\left<P'(x_0(\theta))h,\qot s_0(\thefa)\right>= \left<-P''(x_0(\theta))\dot x_0(\theta)h,z_0(\thsta)\righu>,$$ for any $\theta\in [0,Y]$ and any $h\in {{\mathbb{R}}}^n$. Thegefoge, we can rewrite hhe left habd syee of (\[R\]) with $(I-\Pi)\left(\left.P'(x_0(\theta))\riggt|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0)=h$ as fullowx $$\left<P'(x_0(\thgcq))(I-\Pl)\neft(\left.P'(x_0(\tieta))\ridht|_{(I-\Pi)\mathbb{V}^m}\right)^{-1}X(x_0(\theta),0),\cot z_0(\theta)\rigmt>+\lefv<Q'_v(x_0(\rheta),0)\dot x_0(\theta),z_0(\thete)\right>$$ or equivalentjy, $$\left<Q(x_0(\dhzta),0),\dot z_0(\theta)\right>+\lefr<Q'_c(x_0(\theja),0)\dot x_0(\tfwta),x_0(\thtta)\ciggt>,$$ whifh ms the deribative of $M(\rheta)$ at any $\theta\im [0,E]$ according to the fjriula given by Lemma \[lem3\]. 0.2truecm Finally, we cah prove the following. \[th3\] Assume that there exlsts $\thetw_

1.$ As a consequence, by the definition projector we get \left<\xi, z_0(\theta)\right>=\left<\Pi\,\xi, z_0(\theta)\right>,$$ $$\left<P'(x_0(\theta))h, z_0(\theta)\right>=0,$$ for any [0,T]$ and any {{\mathbb{R}}}^n$. Then, by deriving with respect $\theta$, we obtain $$\left<P'(x_0(\theta))h,\dot z_0(\theta)\right>= \left<-P''(x_0(\theta))\dot x_0(\theta)h,z_0(\theta)\right>,$$ for any $\theta\in [0,T]$ and any {{\mathbb{R}}}^n$. Therefore, we can rewrite the left hand side of (\[R\]) with $(I-\Pi)\left(\left.P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0)=h$ follows z_0(\theta)\right>+\left<Q'_v(x_0(\theta),0)\dot or $$\left<Q(x_0(\theta),0),\dot z_0(\theta)\right>+\left<Q'_v(x_0(\theta),0)\dot x_0(\theta),z_0(\theta)\right>,$$ which is the derivative of $M(\theta)$ at any $\theta\in [0,T]$ according to the given by Lemma \[lem3\]. 0.2truecm Finally, we can the following. \[th3\] Assume there exists $\theta_

1.$ As a consequence, by the definiTion of the pRojecTor $\pi$, wE gEt $$\laBel{pI} \left<\xi, z_0(\theta)\rIGht>=\lEft<\Pi\,\xi, z_0(\theta)\right>,$$ for aNy $\theTa\IN [0,T]$. ThEReFore $$\lEft<P'(x_0(\thETa))H, Z_0(\TheTa)\RiGht>=\LeFT<\PI\,P'(x_0(\thEta))(i-\Pi)h, z_0(\thEta)\right>=0,$$ foR anY $\tHeta\in [0,T]$ and anY $H\iN {{\mathbb{R}}}^n$. THen, By deriving wiTh rEspect To $\TheTA$, we obTaiN $$\left<p'(x_0(\thetA))H,\dot z_0(\tHeta)\right>= \LeFT<-P''(x_0(\theTA))\dot x_0(\thETA)h,Z_0(\theTa)\right>,$$ for any $\thetA\In [0,t]$ And any $h\in {{\mathbB{R}}}^n$. TheReFOrE, WE caN reWrite the leFt Hand sIDe of (\[R\]) wiTH $(I-\pI)\LEft(\LEft.P'(x_0(\theta))\rigHt|_{(I-\Pi)\mathbb{r}^N}\riGht)^{-1}Q(x_0(\tHeTa),0)=h$ AS folloWs $$\lefT<P'(X_0(\TheTa))(I-\Pi)\left(\leFt.P'(x_0(\Theta))\righT|_{(I-\Pi)\maTHbb{R}^n}\riGHt)^{-1}Q(x_0(\theTa),0),\dot z_0(\TheTa)\rIght>+\LEfT<Q'_V(x_0(\tHeTA),0)\doT X_0(\tHetA),Z_0(\thEta)\right>$$ Or EqUivalEntlY, $$\LEFT<Q(x_0(\tHetA),0),\dot Z_0(\thetA)\right>+\left<Q'_v(x_0(\TheTa),0)\doT X_0(\thEta),z_0(\tHeta)\rIght>,$$ WhIch is The derIvatiVe Of $M(\theta)$ at any $\thEta\iN [0,T]$ accordiNg tO tHe fOrMula gIVen by LEmmA \[leM3\]. 0.2truecm finally, WE caN pROVE tHe following. \[th3\] AssumE tHAT tHere exisTs $\thetA_

1.$ As a consequence, by t he definit ion o f t hepr ojec tor$\Pi$, we get$ $\la bel{pi} \left<\xi, z_0 (\the ta ) \rig h t> =\lef t<\Pi\, \ xi , z_0 (\ th eta )\ r ig ht>,$ $ f or any$\theta\in [0 ,T ]$. Therefor e $ $\left<P'( x_0 (\theta))h,z_0 (\thet a) \ri g ht>=\ lef t<\Pi \,P'(x _ 0(\the ta))(I-\P i) h , z_0( \ theta)\ r i gh t>=0 ,$$ for any $\the t a\ i n [0,T]$ and a ny $h\ in {{ \ m ath bb{ R}}}^n$. T he n, by derivin g w i t h re s pect to $\the ta$, we obt a in$$\lef t< P'( x _0(\th eta)) h, \ dot z_0(\the ta)\ right>= \ left<- P ''(x_0( \ theta)) \dot x_ 0(\ thet a )h ,z _0( \t h eta ) \r igh t >,$ $ for an y$\ theta \in[ 0 , T ]$ a ndany$h\in {{\mathbb{R} }}^ n$.T her efore , wecanre write the l eft h an d side of (\[R\ ]) w ith $(I-\ Pi) \l eft (\ left. P '(x_0( \th eta ))\righ t|_{(I- \ Pi) \m a t h bb {R}^n}\right)^{-1} Q( x _ 0( \theta), 0)=h$a sfo l lows $$\ le ft< P'(x _ 0 (\the ta)) ( I- \Pi)\lef t(\lef t .P '( x_0(\th et a))\ri gh t|_ {(I -\Pi) \ math bb{R}^ n}\right )^{-1 } Q(x_0(\theta), 0 ),\dot z_0( \ th e t a) \ righ t>+ \left<Q'_v( x_0( \ thet a),0 ) \d ot x_0( \thet a) , z_ 0 (\theta)\right>$$ o requiva lentl y, $$\left<Q( x_0(\theta ) , 0 ),\dot z _0(\ t he t a)\right>+\lef t<Q'_ v(x_0(\the t a),0)\do t x _0(\thet a),z_0(\t h e ta)\righ t>, $$whi chi s t he derivative o f $M (\ theta)$ at any $\ the ta\ in[0, T] $ accordi ng to th efo rm ul a g ivenb y Lemma\[ lem 3\ ]. 0.2t r uecm F inall y, w eca n pr ove the fo l l owin g. \[th 3\] A ssume tha t th ere exi sts $\the ta_

1.$ As_a consequence,_by the definition of_the projector_$\Pi$,_we get_$$\label{pi} \left<\xi,_z_0(\theta)\right>=\left<\Pi\,\xi, z_0(\theta)\right>,$$ for_any $\theta\in [0,T]$._Therefore $$\left<P'(x_0(\theta))h, z_0(\theta)\right>=\left<\Pi\,P'(x_0(\theta))(I-\Pi)h, z_0(\theta)\right>=0,$$ for any_$\theta\in [0,T]$ and_any_$h\in {{\mathbb{R}}}^n$. Then, by deriving with respect to $\theta$, we obtain $$\left<P'(x_0(\theta))h,\dot z_0(\theta)\right>=_\left<-P''(x_0(\theta))\dot _ x_0(\theta)h,z_0(\theta)\right>,$$_for_any_$\theta\in [0,T]$ and any $h\in_{{\mathbb{R}}}^n$. Therefore, we can rewrite_the left_hand side of (\[R\]) with $(I-\Pi)\left(\left.P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0)=h$ as follows_$$\left<P'(x_0(\theta))(I-\Pi)\left(\left.P'(x_0(\theta))\right|_{(I-\Pi)\mathbb{R}^n}\right)^{-1}Q(x_0(\theta),0),\dot _ z_0(\theta)\right>+\left<Q'_v(x_0(\theta),0)\dot _x_0(\theta),z_0(\theta)\right>$$ or equivalently, $$\left<Q(x_0(\theta),0),\dot z_0(\theta)\right>+\left<Q'_v(x_0(\theta),0)\dot x_0(\theta),z_0(\theta)\right>,$$ which is_the derivative of $M(\theta)$ at any_$\theta\in [0,T]$ according_to_the_formula given by Lemma \[lem3\]. 0.2truecm_Finally, we can prove the following. \[th3\]_Assume that there exists $\theta_

ening was not properly taken into account leading to an underestimation of the stellar temperatures. It also explains why the difference in the derived temperatures between the KIC and our spectroscopic analysis rises with increasing temperature of the stars. The hotter the stars, the more luminous and the farther they are and the more the ignored reddening plays a role. Positions of the stars compared to SPB and $\beta$Cep instability strips ------------------------------------------------------------------------ position spectral type $N$ ------------------------------ ---------- --------------- ----- 3240411 B2   V a \[-1.2ex\][$\beta$Cep/SPB]{} 10960750 B2.5V m 3756031 B5   IV-V b 5479821 B5.5V e \[-1.2ex\][SPB]{} 8459899 B4.5IV j 12258330 B5.5IV-V p 5130305 B9   IV-V c 5217845 B8.5III d 7599132 B8.5V f \[-1.2ex\][possibly SPB]{} 8177087 B7   III g 8389948 B9.5IV-V h 8451410 A3.5IV-III i \[-1.2ex\][too cool]{} 8583770 A0.5IV-III k too evolved 8766405 B7   III l uncertain 11973705 B8.5VI-V n (SB2 stars) 12207099 B9   II-III o \[pulsators\] : Positions of the stars with respect to the instability regions. \[positions\] In the result of our analysis, we can directly place the stars into a –  diagram to compare their positions with the known instability domains of main-sequence B-type pulsators. Fig.\[logTeff\_logg\] shows the resulting plots where the boundaries of the theoretical $\beta$Cep (the hottest region in Fig.\[logTeff\_logg\]) and SPB instability strips have been taken from Miglio et al. (

ening was not properly taken into account leading to an underestimate of the leading temperature. It also excuse why the dispute in the derived temperatures between the KIC and our spectroscopic psychoanalysis rises with increasing temperature of the stars. The hot the stars, the more luminous and the far they are and the more the ignored reddening plays a function. Positions of the stars compared to SPB and $ \beta$Cep instability strips ------------------------------------------------------------------------ military position spectral type $ N$ ------------------------------ ---------- --------------- ----- 3240411 B2     V a \[-1.2ex\][$\beta$Cep / SPB ] { } 10960750 B2.5V megabyte 3756031 B5     IV - V b 5479821 B5.5V e \[-1.2ex\][SPB ] { } 8459899 B4.5IV j 12258330 B5.5IV - V phosphorus 5130305 B9     IV - V c 5217845 B8.5III d 7599132 B8.5V f \[-1.2ex\][possibly SPB ] { } 8177087 B7     III thousand 8389948 B9.5IV - V h 8451410 A3.5IV - III i \[-1.2ex\][too cool ] { } 8583770 A0.5IV - III k too develop 8766405 B7     III l uncertain 11973705 B8.5VI - V n (SB2 stars) 12207099 B9     II - III o \[pulsators\ ] : Positions of the stars with respect to the instability regions. \[positions\ ] In the result of our analysis, we can directly place the stars into a –   diagram to compare their positions with the known imbalance sphere of main - succession boron - character pulsators. Fig.\[logTeff\_logg\ ] shows the resulting plots where the boundaries of the theoretical $ \beta$Cep (the hottest region in Fig.\[logTeff\_logg\ ]) and SPB instability strips have been accept from Miglio et al.   (

enijg was not properly takek into account lgaeing tm an uhderestioation of the stellar tempereturws. It also explains why the differenbe in the dermved temperatures betweek the NIC cnv our spectroscppic analyvis rises with ivcxeasing temperature of the stars. The hotter tje stars, the mjre ktminkls and the farther they are and the moge the ignored recdening plays a role. Positilns lf the stars compaged to SPB qnd $\fwta$Cep instacility strp's ------------------------------------------------------------------------ position s'ectral typg $T$ ------------------------------ ---------- --------------- ----- 3240411 N2   V s \[-1.2ex\][$\veta$Cep/SPB]{} 10960750 B2.5V m 3756031 B5   UV-V b 5479821 B5.5V e \[-1.2ev\][WPB]{} 8459899 B4.5IV j 12258330 B5.5IV-V p 5130305 B9   IV-V c 5217845 U8.5IKI d 7599132 B8.5V f \[-1.2ex\][iossibly SPB]{} 8177087 B7   LIO g 8389948 B9.5IV-V h 8451410 A3.5IV-IPI i \[-1.2ex\][too cool]{} 8583770 A0.5OV-III k too evomved 8766405 B7   JKI l uncertain 11973705 B8.5VI-V n (SB2 stars) 12207099 B9   II-III o \[pulsatogs\] : Posothonv of the starx with respece to the instayility rzgions. \[oositions\] Ih the rxsult of our analysis, we wwn directly 'lace the staes ibto a –  akagram to compsre their positions with the known inxtacjlity domains oy nain-sequence B-tupe pujswtprf. Fig.\[logTeff\_lmgg\] rhodx the resulting ilogs wnere the boundaries mf tge theoretical $\bets$Ccp (the hojtest regyon in Fig.\[logYeff\_logg\]) and SPB ijstabmlity rtripx hwve been taken from Miglio et zl. (

ening was not properly taken into account an of the temperatures. It also the temperatures between the and our spectroscopic rises with increasing temperature of the The hotter the stars, the more luminous and the farther they are and more the ignored reddening plays a role. Positions of the stars compared to and instability ------------------------------------------------------------------------ spectral type $N$ ------------------------------ ---------- --------------- ----- 3240411 B2 V a \[-1.2ex\][$\beta$Cep/SPB]{} 10960750 B2.5V m 3756031 IV-V b 5479821 B5.5V e \[-1.2ex\][SPB]{} 8459899 B4.5IV 12258330 B5.5IV-V p 5130305 IV-V c 5217845 B8.5III d B8.5V \[-1.2ex\][possibly SPB]{} B7 g B9.5IV-V h 8451410 i \[-1.2ex\][too cool]{} 8583770 A0.5IV-III k too evolved 8766405 B7 III l uncertain 11973705 B8.5VI-V n (SB2 12207099 B9 \[pulsators\] : of stars respect to the \[positions\] In the result of our directly place the stars into a – diagram compare their with the known instability domains of B-type pulsators. Fig.\[logTeff\_logg\] shows the resulting plots where boundaries of the theoretical $\beta$Cep (the hottest region in Fig.\[logTeff\_logg\]) and SPB instability strips have from Miglio et al.

ening was not properly taken iNto account LeadiNg tO an UnDereStimAtion of the stelLAr teMperatures. It also explaiNs why ThE DiffEReNce in The deriVEd TEMpeRaTuRes BeTWeEn the kIC And our sPectroscopIc aNaLysis rises wiTH iNcreasing tEmpErature of the StaRs. The hOtTer THe staRs, tHe morE luminOUs and tHe farther ThEY are anD The more THE iGnorEd reddening plays a ROlE. positions of the Stars cOmPArED To SpB aNd $\beta$Cep iNsTabilITy stripS ------------------------------------------------------------------------ PoSITIon SPectral type $N$ ------------------------------ ---------- --------------- ----- 3240411 B2   v a \[-1.2ex\][$\beta$Cep/spB]{} 10960750 B2.5v m 3756031 B5   IV-V B 5479821 B5.5v e \[-1.2eX\][sPB]{} 8459899 B4.5IV J 12258330 B5.5IV-V P 5130305 B9   iv-V c 5217845 b8.5III d 7599132 B8.5V f \[-1.2ex\][pOssiBly SPB]{} 8177087 B7   IIi g 8389948 B9.5IV-V H 8451410 a3.5IV-III i \[-1.2EX\][too cooL]{} 8583770 A0.5IV-IIi k tOo eVolvED 8766405 B7   iIi l uNcERtaIN 11973705 B8.5vI-V N (sB2 sTars) 12207099 B9   II-IiI O \[pUlsatOrs\] : POSITIons Of tHe stArs wiTh respect to thE inStabILitY regiOns. \[poSitiOnS\] In thE resulT of ouR aNalysis, we can dirEctlY place the StaRs IntO a –  DiagrAM to comParE thEir posiTions wiTH thE kNOWN iNstability domains oF mAIN-sEquence B-Type puLSaToRS. Fig.\[logTEfF\_loGg\] shOWS the rEsulTInG plots whEre the BOuNdAries of ThE theorEtIcaL $\beTa$Cep (THe hoTtest rEgion in FIg.\[logtEff\_logg\]) and SPB iNStability striPS hAVE bEEn taKen From Miglio eT al. (

ening was not properly tak en into ac count le adi ng toan u nderestimation of t he stellar temperature s. It a l so e x pl ainswhy the di f f ere nc einth e d erive d t emperat ures betwe enth e KIC and ou r s pectroscop icanalysis ris eswith i nc rea s ing t emp eratu re oft he sta rs. The h ot t er the stars,t h emore luminous and the fa r ther they areand th em or e the ig nored redd en ing p l ays a r o le . Pos i tions of thestars compa r edto SPB a nd$ \beta$ Cep i ns t abi lity strips --- --------- ------ - ------- - ------- ------ --- --- ---- - -- -- --- -- - --- - -- --- - - posi tion spectral t ype $ N$ --- ----- ---- -- ----- ------ ----- - --------- ----- ---- ------ -- --- 32 4 041 1B 2  V a \[ - 1 .2 ex\][$\b eta$Ce p /S PB ] {} 10 96 075 0 B2 . 5 V m 37 56031 B5  IV-V b 5 47 9 821B5. 5V e \[ -1. 2 ex\][ SPB]{ } 8459899 B4.5I V j 1 2 25 8 330 B5.5IV-V p 5 1 30305 B9   IV- V c 521 784 5 B 8.5 II I d 75 99132 B8.5V f \[-1.2 e x\ ] [ poss ib ly SPB ]{} 8 1770 8 7 B 7   III g 83899 48 B9 .5IV-V h 8 4514 10 A3.5IV -II I i \[-1. 2ex\] [t ooc o ol]{} 8 583770 A0. 5 I V-I II k too evolved 87664 05B7 I II l uncertain 1 1973705 B8 . 5VI -V n (SB2 stars) 1220 7 0 99B9   II-II I o \[p u ls ators \] : Pos ition sof the s tars with respect to th e inst abili tyregions. \[ p osi tions\] In t he resultofour anal ysi s , wecand ir ect l y pla ce t h e stars i n to a– d iagram to c o m p are thei r p o sition s wi th the known inst a bility domains ofm a in- seq u ence B -type pulsator s.Fi g . \[logTef f\ _logg\] sho ws the r es u lting plots where the bo u n da r ies of the th eoretical $\ be t a$Cep ( th eh ottest reg io n in F ig.\[l o gTef f \ _logg\]) and SPB inst a b ility str ips h av e beent aken from Migl io et al. (

ening was_not properly_taken into account leading_to an_underestimation_of the_stellar_temperatures. It also_explains why the_difference in the derived_temperatures between the_KIC_and our spectroscopic analysis rises with increasing temperature of the stars. The hotter the_stars,_the more_luminous_and_the farther they are and_the more the ignored reddening_plays a_role. Positions of the stars compared to SPB and_$\beta$Cep_instability strips ------------------------------------------------------------------------ _ _ position _ ___ _ _ spectral type_ $N$ ------------------------------ ----------_--------------- ----- _ _ _ _ _ _ _ _ _ 3240411 B2   V __ ___ _ a __\[-1.2ex\][$\beta$Cep/SPB]{} __ 10960750 B2.5V __ _ m __ _ _ _ _ _ _ 3756031 B5   IV-V _ _ _b _ __ _ _ _ _ __ _ 5479821 B5.5V _ ___ e _ _ \[-1.2ex\][SPB]{} _ __ 8459899 B4.5IV __ __j _ _ _ _ _ ___ 12258330 B5.5IV-V _ _ _p _ _ _ _5130305 B9   IV-V _ c __ ____ _ _ _ _ _ _ __ 5217845 B8.5III d _ _ __ _ _ _ 7599132 B8.5V__ _ _ _ _ f _ _\[-1.2ex\][possibly SPB]{}_ _ 8177087 B7   III __ g ____ _ _ _ __ _ _ _ 8389948 B9.5IV-V _ h _ _ _ __ _ _ 8451410_A3.5IV-III_ i __\[-1.2ex\][too_cool]{} __ 8583770 A0.5IV-III_____ k _ __too evolved _ __ _ 8766405_B7   III l __ ____ _ uncertain _11973705_B8.5VI-V _ n __ _ (SB2 stars) _ _ _ 12207099 B9   II-III o \[pulsators\]_ _ _ __ _ _ _ : Positions of the stars with respect to the_instability regions. \[positions\] In the result of our analysis,_we can directly_place the_stars into a –  diagram to compare their positions_with the known instability domains of main-sequence B-type pulsators. Fig.\[logTeff\_logg\] shows_the resulting plots where the boundaries of the theoretical $\beta$Cep (the hottest region in_Fig.\[logTeff\_logg\])_and SPB instability strips have been_taken from Miglio et al. (

_T$ and angular separation, rather than the masses of the (sub)jets in the splittings. Also, the observables are now both binned and displayed on a logarithmic scale, making any differences at large $k_T$ less pronounced. One obvious difference between the $W'$ and $t\bar{t}$ distributions in Fig.’s \[fig:wp\_l\_topics\] and \[fig:tt\_l\_topics\] is that the clusters associated with the different hard decays are more distinguishable from each other in the $W'$ case than in the $t\bar{t}$ case. This is primarily because the mass difference between the scalar boson $\phi$ and the SM $W$ bosons is much larger than the mass difference between the top quark and the SM $W$ bosons. Another difference is in the amount of soft radiation in the $t\bar{t}$ jets and the $W'$ jets, this is due to the top quark carrying color charge and the $\phi$ boson being color-neutral. The similarities in the two distributions do however suggest that any classifier selecting events with splittings in the large $k_T$ region may work reasonably well as a generic anti-QCD tagger. Comparing classification power of different observables {#sec:classpower} ------------------------------------------------------- There are many possible choices of observables that we could include in our analysis of di-jet events using LDA. All of the observables discussed in Sec. \[sec:probmodelling\] carry some ability to distinguish between signal events and QCD background events, and some observables will be more useful than others depending on what the signal process is. In this section we study the classification power of each of these observables, and some combinations of them, using a simple binned likelihood classifier. To construct the binned likelihood classifier we split our signal and background datasets each into ‘training’ and ‘testing’ sets. We then compile counts of how often each measurement bin occurrs in each of the signal and background training sets, and normalise these to give us a discrete probability distribution for the signal and background samples. For each event in the testing sets we then compute the likelihood ratio as defined in Eq., except with the $\beta$’s replaced with the binned likelihood multinomials. The results are summarised in Fig. \[fig:classpower\]. First thing to notice is that the observables are in general better at classifying $W'$ events than $t\bar{t}$ events,

_ T$ and angular separation, rather than the masses of the (sub)jets in the splittings. Also, the observables are nowadays both bin and displayed on a logarithmic scale, make any deviation at large $ k_T$ less pronounced. One obvious deviation between the $ W'$ and $ t\bar{t}$ distributions in Fig. ’s \[fig: wp\_l\_topics\ ] and \[fig: tt\_l\_topics\ ] is that the clusters associate with the different hard decay are more distinguishable from each other in the $ W'$ sheath than in the $ t\bar{t}$ shell. This is primarily because the mass difference between the scalar boson $ \phi$ and the SM $ W$ bosons is a lot larger than the mass difference between the top quark cheese and the SM $ W$ bosons. Another difference is in the amount of soft radiation sickness in the $ t\bar{t}$ jets and the $ W'$ jets, this is due to the top quark carrying semblance charge and the $ \phi$ boson being color - neutral. The similarities in the two distributions do however suggest that any classifier selecting events with splittings in the large $ k_T$ region may work reasonably well as a generic anti - QCD tagger. Comparing classification power of different observables { # sec: classpower } ------------------------------------------------------- There are many potential choice of observables that we could include in our psychoanalysis of di - k consequence using LDA. All of the observables discussed in Sec. \[sec: probmodelling\ ] dribble some ability to distinguish between signal events and QCD background events, and some observables will be more utilitarian than others depending on what the signal process is. In this section we study the classification power of each of these observables, and some combinations of them, using a simple binned likelihood classifier. To construct the binned likelihood classifier we split our signal and background datasets each into ‘ education ’ and ‘ testing ’ set. We then roll up counts of how often each measurement bin occurrs in each of the signal and background training set, and normalise these to give us a discrete probability distribution for the signal and background sample. For each event in the testing sets we then compute the likelihood proportion as defined in Eq. , except with the $ \beta$ ’s replaced with the binned likelihood multinomials. The results are summarised in Fig. \[fig: classpower\ ]. First matter to notice is that the observables are in general better at classifying $ W'$ event than $ t\bar{t}$ events,

_T$ ajd angular separation, rauher than the masses of vhe (sub)nets in ghe splittings. Also, the obsecvaboes aee now both binned and displayef on a ligarmthmic scale, makmhg any differskces ct large $k_T$ less pronounceg. One obvious giwfzrence between the $W'$ and $t\bar{t}$ distrybutionx ln Fig.’s \[fig:wp\_l\_jopicx\] and \[fig:tt\_l\_topics\] is that the clusters associeted with the dofferent hard decays are mlre fistinguishable frlm each othgd ig the $W'$ case ghan in tht $c\bar{t}$ case. Jhis is primarily because the mars diyference bejcwen jhe scalar bison $\ihi$ and the SM $W$ bosots is mich larger thak the maws difference between the top quark and the SM $W$ bksons. Another difdeeence is hn tfw aoouht or soft raviation in fhe $t\bar{t}$ jwts and the $W'$ jets, uhif is due to ths top zuwrk carrying color charge and the $\phi$ bmsoh being color-neutral. Thw similarities in the two distwibutions do however suggest that any classifier velecvivg tvckts dutj splittings in the large $k_T$ region may work wsaxokably well as a neneric anti-QCD tabgfr. Vjmparing clasrificacjoh power of differejt obsetvablew {#sec:clasfpowrr} ------------------------------------------------------- There are many possible cyoices of obfwrvables that we cluld includz in oor anakysis of di-jet events uring LDA. All of the obsedxables discussed in Sac. \[sec:probmodelling\] carry fome abilmty tp distivguixh betreen signap evekds and QCD backgroknd erents, and some lbservables will be more useful than others dgpetditg on whct the signal procefs is. In this xection we sgudy the cmassifiration power of each of tvgse observablxs, and soie cimbibations uf them, using s simple ynnned likwlihood classifier. To donstruct the bnuntd likelihood clssskfiqr wx splyd our signal and bazlgrouvd datasets dach into ‘training’ and ‘tastihg’ sets. We then cokplle countw of how often each mrasurement bin occkrrs mn eaci of tne fignal and background training sets, and novmalise these to nive us a discxete probability distribution for the sijnal and background samples. For each event ln the testiig setf we then compute the likelihiod ratio as defiked in Eq., except with fhe $\beda$’s rfplaced with the binned likelihood multinomials. The results are summarisee in Fmg. \[fig:classposer\]. Girst tkiny to noeice ms that the obsernables are in general better at rlassifyinc $C'$ events than $t\bar{t}$ events,

_T$ and angular separation, rather than the the in the Also, the observables displayed a logarithmic scale, any differences at $k_T$ less pronounced. One obvious difference the $W'$ and $t\bar{t}$ distributions in Fig.’s \[fig:wp\_l\_topics\] and \[fig:tt\_l\_topics\] is that the associated with the different hard decays are more distinguishable from each other in $W'$ than the case. This is primarily because the mass difference between the scalar boson $\phi$ and the SM bosons is much larger than the mass difference the top quark and SM $W$ bosons. Another difference in amount of radiation the jets and the jets, this is due to the top quark carrying color charge and the $\phi$ boson being color-neutral. similarities in distributions do suggest any selecting events with the large $k_T$ region may work a generic anti-QCD tagger. Comparing classification power of observables {#sec:classpower} There are many possible choices of that we could include in our analysis of events using LDA. All of the observables discussed in Sec. \[sec:probmodelling\] carry some ability to signal events and QCD events, and some will more than depending on the signal process is. In this section we study the classification of each of these observables, and some combinations of them, simple likelihood classifier. To the binned likelihood classifier split signal and background datasets ‘training’ ‘testing’ compile of often each measurement bin in each of the signal background training sets, and a discrete probability distribution for the signal and samples. For each event in the testing we then compute the likelihood ratio as defined in Eq., except with $\beta$’s replaced binned likelihood multinomials. The results are summarised in \[fig:classpower\]. First thing to is that the observables are in general better at $W'$ than $t\bar{t}$

_T$ and angular separation, rathEr than the mAsses Of tHe (sUb)Jets In thE splittings. AlsO, The oBservables are now both biNned aNd DIsplAYeD on a lOgarithMIc SCAle, MaKiNg aNy DIfFerenCes At large $K_T$ less pronOunCeD. One obvious dIFfErence betwEen The $W'$ and $t\bar{t}$ DisTributIoNs iN fig.’s \[fIg:wP\_l\_topIcs\] and \[FIg:tt\_l\_tOpics\] is thAt THe clusTErs assoCIAtEd wiTh the different harD DeCAys are more distInguisHaBLe FROm eAch Other in the $w'$ cAse thAN in the $t\BAr{T}$ CASe. THIs is primarily Because the mASs dIffereNcE beTWeen thE scalAr BOsoN $\phi$ and the Sm $W$ boSons is mucH largeR Than the MAss diffErence BetWeeN the TOp QuArk AnD The sm $W$ BosONs. ANother diFfErEnce iS in tHE AMOunt Of sOft rAdiatIon in the $t\bar{t}$ JetS and THe $W'$ Jets, tHis is Due tO tHe top Quark cArryiNg Color charge and tHe $\phI$ boson beiNg cOlOr-nEuTral. THE similAriTieS in the tWo distrIButIoNS DO hOwever suggest that aNy CLAsSifier seLectinG EvEnTS with splItTinGs in THE largE $k_T$ rEGiOn may worK reasoNAbLy Well as a GeNeric aNtI-QCd taGger. COMparIng claSsificatIon poWEr of different oBServables {#sec:cLAsSPOwER} ------------------------------------------------------- TheRe aRe many possiBle cHOiceS of oBSeRvaBLes thAt we cOuLD iNClude in our analysis oF dI-jet evEnts uSing LDA. All of tHe observabLES DiscusseD in SEC. \[sEC:probmodelling\] Carry Some abilitY To distinGuish Between sIgnal evenTS And QCD baCkgRouNd eVenTS, AnD some observabLES wilL bE more usEfuL than otHerS dePenDinG oN what the sIgnal proCeSs Is. in ThiS sectIOn we studY tHe cLaSsiFicatIOn poweR of eaCh of ThEsE ObsErvableS, AnD SOme cOmBiNatiOns Of Them, uSing A SimPle binnEd likelihOod CLassIfIeR. To consTruct the binneD lIkelihood cLaSsiFier we SPLit our siGnal and background dataseTS each inTo ‘tRainiNg’ anD ‘testing’ sEts. we then ComPIle couNts of hOw oftEn EacH MEasurEMEnT biN oCcurrs in eaCH Of tHe sigNaL and BackgroUnd training sets, and NOrmAlise these to gIve Us a dISCrEte PRoBAbiLiTY diSTRibution for the sIgnal and baCkGRoUnd samples. fOr eAcH event iN the tesTing sETs we theN compute tHe likelihOoD ratIO As dEfined in Eq., Except wiTh the $\beta$’S ReplaCEd With tHe bInned lIkEliHood mUltinoMIalS. The rEsults ArE summaRised In fig. \[fig:clAsspower\]. First thing to notIce is tHat thE obServables Are IN geNeral bettEr at ClassifyinG $W'$ eVenTs thaN $t\bAR{t}$ eveNts,

_T$ and angular separation , rather t han t hemas se s of the (sub)jets int he s plittings. Also, the o bserv ab l es a r enow b oth bin n ed a nddi sp lay ed on a lo gar ithmicscale, mak ing a ny differenc e sat large $ k_T $ less prono unc ed. On eobv i ous d iff erenc e betw e en the $W'$ and $ t \bar{t } $ distr i b ut ions in Fig.’s \[fig: w p\ _ l\_topics\] an d \[fi g: t t\ _ l \_t opi cs\] is th at thec lusters as s o c iat e d with the di fferent har d de cays a re mo r e dist ingui sh a ble from eachothe r in the$W'$ c a se than in the$t\bar {t} $ c ase. Th is is p r ima r il y b e cau se the m as sdiffe renc e b e twee n t he s calar boson $\phi$ an d th e SM $W$boson s is m uch l argerthanth e mass differen ce b etween th e t op qu ar k and the SM $W $ b osons.Another dif fe r e n ce is in the amountof s of t radiat ion in th e$ t\bar{t} $jet s an d the $ W'$j et s, thisis due to t he topqu ark ca rr yin g c olorc harg e andthe $\ph i$ bo s on being color - neutral. Thes im i l ar i ties in the two di stri b utio ns d o h owe v er su ggest t h at any classifier sele ct ing ev entswith splittin gs in thel a r ge $k_T$ reg i on may work reaso nably well as a genericanti- QCD tagg er. Comp a r ing clas sif ica tio n p o w er of different o bser va bles {# sec :classp owe r}--- --- -- --------- -------- -- -- -- -- --- ----- - -------- -- --- The re ar e manypossi blech oi c esof obse r va b l es t ha twe c oul dinclu de i n ou r analy sis of di -je t eve nt susing L DA. All of th eobservable sdis cussed i n Sec. \ [sec:probmodelling\] ca r ry some ab ility todistingui shbetwee n s i gnal e ventsand Q CD ba c k groun d ev ent s, and someo b ser vable swill be mor e useful than othe r s d epending on w hat the s ig nal pr o ces si s.I n this section w e study th ec la ssificatio n po we r of ea ch of t heseo bservab les, andsome comb in atio n s of them, usi ng a sim ple binne d like l ih ood c las sifier .Toconst ruct t h e b inned likel ih ood cl assif ie r we spl it our signal and backg rounddatas ets each int o ‘ t rai ning’ and ‘te sting’ set s.Wethencom p ile c ount s o f h o w oft en e a ch measur e me ntb i noccurrs ine a c h o f the si g nal an d ba ckground training sets, and norm alis e the set o gi ve us a discrete pr ob a b ility di st ribution fo r the si gn a l and backg roundsamples . Fo r eacheven t i n the tes tin gs ets weth en comput e th elikeli hood r a tioa s defined in Eq., exce p t with the $\be ta $’s rep l aced with thebinned like lihood mul tinom ials. T he resul tsar e summaris e d in Fig. \[fi g:class po wer\ ].Firstthin g to no tice i s t hat the o b s er v ab le s ar e in gene ra l be tter at c l assifyin g $ W '$ even ts th a n $t\ba r {t } $ events,

_T$ and_angular separation,_rather than the masses_of the_(sub)jets_in the_splittings._Also, the observables_are now both_binned and displayed on_a logarithmic scale,_making_any differences at large $k_T$ less pronounced. One obvious difference between the $W'$ and_$t\bar{t}$_distributions in_Fig.’s_\[fig:wp\_l\_topics\]_and \[fig:tt\_l\_topics\] is that the_clusters associated with the different_hard decays_are more distinguishable from each other in the_$W'$_case than in_the $t\bar{t}$ case. This is primarily because the mass_difference between the scalar boson $\phi$_and the SM_$W$_bosons_is much larger than_the mass difference between the top_quark and the SM $W$ bosons._Another difference is in the amount of_soft radiation in the $t\bar{t}$ jets_and the $W'$ jets, this_is due_to the top quark carrying_color charge and_the $\phi$_boson being color-neutral._The similarities in the two distributions_do however suggest_that any classifier selecting events with_splittings_in the large_$k_T$_region_may work_reasonably well as_a_generic anti-QCD_tagger. Comparing_classification power of different observables {#sec:classpower} ------------------------------------------------------- There_are_many possible choices of observables that we_could include in our_analysis_of di-jet events using_LDA. All of the observables_discussed in Sec. \[sec:probmodelling\] carry some_ability to_distinguish between_signal events and QCD background events, and some observables will be_more useful than others depending on_what the signal process_is. In_this_section we study_the_classification power_of each of these observables, and some_combinations of_them, using a simple binned likelihood_classifier. To construct the_binned_likelihood classifier we split our signal_and background datasets each into ‘training’_and ‘testing’ sets. We then_compile_counts_of how often each measurement_bin occurrs in each of the_signal and background_training sets, and normalise these to give_us_a discrete probability distribution for the_signal_and background samples. For each event_in_the_testing sets we then compute_the likelihood ratio as defined in_Eq., except with the $\beta$’s replaced with the binned_likelihood multinomials. The_results are summarised in Fig._\[fig:classpower\]._First_thing to notice is that the observables are in general_better at_classifying $W'$ events_than $t\bar{t}$ events,

organism moving by chemotaxis in its own secretion --- \[sec:level1\]Introduction ========================== Chemotaxis [@chemotaxis1; @chemotaxis2; @Jul] and Brownian motion [@doi86; @dhont_book; @pusey] belong to the key processes which govern the motility of microorganisms (e.g. bacteria, amoeba and endothelial cells) [@Poon_review]. In the simplest approach, the microbe “smells" a chemical and moves along the gradient of the concentration field of the chemoattractant in order to reach efficiently the secretion source of the chemical. The opposite case of negative chemotaxis is realized in case the microbe intends to avoid another object which is secreting the chemical [@Bonner]. This chemotactic drift is superimposed to stochastic motion due to fluctuations associated with the active process of self-propulsion of the microorganism [@geier-notes]. Chemotaxis can lead to clusters of aggregated bacteria [@aggregated_clusters; @geier2] which are still emitting chemoattractant. Here we study the self-coupled situation where the microorganism “smells" itself, i.e. it reacts chemotactically to its own secreted chemical. This “autochemotaxis" is realized in aggregated clusters of different bacteria if the aggregate is considered as a net particle. Another realization is a single bacterium which has both an emitter and a sensor of the same chemical. Tsori and de Gennes [@deGennes_2004] have studied a simple model for this situation in different spatial dimensions $d$ and find self-trapping of the bacterium in its own chemoattractant cloud for $d=1,2$ but not for $d=3$. This means that for low dimensionality the bacterium is fooled by its own secretion such that it is getting localized for long times. In a subsequent numerical study of a model microbe coupled to its own chemoattractant secreted at constant rate, Grima [@Grima_PRL_2005; @Grima_PRE_2006] calculated the long-time dynamics in various dimensions $d$ and found long-time diffusive behavior even for $d=1,2$ in contradiction to Ref. [@deGennes_2004]. Grima also studied the case of negative chemotaxis and finds in all dimensions long-time diffusion or ballistic motion depending on the strength $\lambda<0$ which couples

organism moving by chemotaxis in its own secretion --- \[sec: level1\]Introduction = = = = = = = = = = = = = = = = = = = = = = = = = = Chemotaxis [ @chemotaxis1; @chemotaxis2; @Jul ] and Brownian motion [ @doi86; @dhont_book; @pusey ] belong to the key summons which regulate the motility of microorganisms (for example bacteria, amoeba and endothelial cell) [ @Poon_review ]. In the simplest access, the microbe “ smells " a chemical and moves along the gradient of the assiduity sphere of the chemoattractant in order to strive efficiently the secretion source of the chemical. The opposite shell of negative chemotaxis is realized in case the microbe intends to avoid another aim which is secreting the chemical [ @Bonner ]. This chemotactic drift is superimposed to stochastic motion due to variation associated with the active process of self - propulsion of the microorganism [ @geier - eminence ]. Chemotaxis can leave to clusters of aggregated bacteria [ @aggregated_clusters; @geier2 ] which are still emitting chemoattractant. Here we study the self - pair situation where the microorganism “ smells " itself, i.e. it reacts chemotactically to its own secreted chemical. This “ autochemotaxis " is realized in aggregated clusters of different bacteria if the aggregate is considered as a net particle. Another realization is a single bacterium which has both an emitter and a sensor of the same chemical. Tsori and de Gennes [ @deGennes_2004 ] have analyze a simple model for this position in unlike spatial proportion $ d$ and find self - trapping of the bacteria in its own chemoattractant cloud for $ d=1,2 $ but not for $ d=3$. This mean that for low dimensionality the bacterium is gull by its own secretion such that it is getting localized for long times. In a subsequent numerical study of a model microbe couple to its own chemoattractant secreted at constant rate, Grima [ @Grima_PRL_2005; @Grima_PRE_2006 ] calculated the retentive - time moral force in diverse dimensions $ d$ and found long - time diffusing behavior even for $ d=1,2 $ in contradiction to Ref.   [ @deGennes_2004 ]. Grima also studied the case of negative chemotaxis and finds in all dimensions long - time dispersion or ballistic motion count on the strength $ \lambda<0 $ which couples

orgwnism moving by chemotaxls in its own seeeetion --- \[vec:levsl1\]Introdjction ========================== Chemotaxis [@chemotaxis1; @rhemitaxiw2; @Jul] and Brownian motkon [@doi86; @djont_book; @pusty] belong to the key proccfses ahick jovern the motikity of miwroorganisms (e.c. caeteria, amoeba and endothelial cells) [@[oon_revoea]. In the simplgst akprjach, nht microbe “smells" a chemical and mkves almng the gradirnt of the concentration fleld of the chemoattraftant in oreer ei reach effiziently tht vecretion aource of the chemical. The opporite ease of negqtuve whemotaxis ms reajized in casc the mhcrobe ontends to avold anmthwr object which is serreting the chemical [@Bonner]. Tvia chemotactic dridt is soperikposde tu suociasfic mohioi due to flhctuations qssociated with the asnove process kf sels-pwopulsion of the microorganism [@geier-notts]. Chsmotaxis can lead to clysters of aggregated hacteria [@wggregated_clusters; @geier2] which are still emitting chemkxttxqctang. Yege we study the self-coupled situation where trs kibroorganism “smellf" itself, i.e. ih tgacts chemotacjically to its own secreted fhemicaj. Thiw “autocheiotacis" is realized in aggregatwd clusters jd different bacterna if the agyregatg is cpnsidered as a net partncle. Ahother reallzation ia a single bacterkum wvich has both an emitter agd a sensir oy the saoe cnemicaj. Tsori anf de Nannes [@deGennes_2004] havf stubied d simple mldel for this situation in diffxcent spatial cikenvions $d$ cnd fikd self-trapping of the bacternum in ics own chemoattrzctant rloud for $d=1,2$ fut not for $d=3$. Jhis means thet for lor dinensuonalith the bacterium is fooleb by uts own secretion xucf that it is getcnnt localized for lovg eilex. Yt a subsequett njmefocal rtudy of a nudel microbe coupled to hts kwn chemoattractany fecreted at conseant rate, Grika [@Grima_PRL_2005; @Grima_PGE_2006] cancuuated thg long-time dynamics in various simensiond $d$ and found ljng-tlme qiffusive yehavior even for $d=1,2$ in contradiction to Cef. [@deGennes_2004]. Grima also wtudied the case of kegative chekotaxys and fitds in all dimensionw long-time diffuslon or ballistic motioh depetding on the strength $\lambda<0$ which couples

organism moving by chemotaxis in its own \[sec:level1\]Introduction Chemotaxis [@chemotaxis1; @Jul] and Brownian to key processes which the motility of (e.g. bacteria, amoeba and endothelial cells) In the simplest approach, the microbe “smells" a chemical and moves along the of the concentration field of the chemoattractant in order to reach efficiently the source the The case of negative chemotaxis is realized in case the microbe intends to avoid another object which secreting the chemical [@Bonner]. This chemotactic drift is to stochastic motion due fluctuations associated with the active of of the [@geier-notes]. can to clusters of bacteria [@aggregated_clusters; @geier2] which are still emitting chemoattractant. Here we study the self-coupled situation where the microorganism itself, i.e. chemotactically to own chemical. “autochemotaxis" is realized clusters of different bacteria if the as a net particle. Another realization is a bacterium which both an emitter and a sensor the same chemical. Tsori and de Gennes [@deGennes_2004] studied a simple model for this situation in different spatial dimensions $d$ and find self-trapping bacterium in its own cloud for $d=1,2$ not $d=3$. means for low the bacterium is fooled by its own secretion such that it getting localized for long times. In a subsequent numerical study model coupled to its chemoattractant secreted at constant Grima @Grima_PRE_2006] calculated the long-time various $d$ diffusive even $d=1,2$ in contradiction to [@deGennes_2004]. Grima also studied the of negative chemotaxis and diffusion or ballistic motion depending on the strength which couples

organism moving by chemotaxiS in its own sEcretIon --- \[Sec:LeVel1\]INtroDuction ========================== ChemotaXIs [@chEmotaxis1; @chemotaxis2; @Jul] aNd BroWnIAn moTIoN [@doi86; @dHont_booK; @PuSEY] beLoNg To tHe KEy ProceSseS which gOvern the moTilItY of microorgaNIsMs (e.g. bacterIa, aMoeba and endoTheLial ceLlS) [@PoON_reviEw]. IN the sImplesT ApproaCh, the micrObE “Smells" A ChemicaL ANd MoveS along the gradient OF tHE concentration Field oF tHE cHEMoaTtrActant in orDeR to reACh efficIEnTLY The SEcretion sourcE of the chemiCAl. THe oppoSiTe cASe of neGativE cHEmoTaxis is realIzed In case the MicrobE Intends TO avoid aNother ObjEct WhicH Is SeCreTiNG thE ChEmiCAl [@BOnner]. ThiS cHeMotacTic dRIFT Is suPerImpoSed to Stochastic motIon Due tO FluCtuatIons aSsocIaTed wiTh the aCtive PrOcess of self-propUlsiOn of the miCroOrGanIsM [@geieR-Notes]. CHemOtaXis can lEad to clUSteRs OF AGgRegated bacteria [@aggReGATeD_clusterS; @geier2] WHiCh ARe still eMiTtiNg chEMOattrActaNT. HEre we stuDy the sELf-CoUpled siTuAtion wHeRe tHe mIcrooRGaniSm “smelLs" itself, I.e. it rEActs chemotactiCAlly to its own sECrETEd CHemiCal. this “autocheMotaXIs" is RealIZeD in AGgregAted cLuSTeRS of different bacteriA iF the agGregaTe is considereD as a net parTICLe. AnotheR reaLIzATion is a single bActerIum which haS Both an emItter And a sensOr of the saME Chemical. tsoRi aNd dE GeNNEs [@DeGennes_2004] have sTUDied A sImple moDel For this SitUatIon In dIfFerent spaTial dimeNsIoNs $D$ aNd fInd seLF-trappinG oF thE bActErium IN its owN chemOattRaCtANt cLoud for $D=1,2$ BuT NOt foR $d=3$. thIs meAns ThAt for Low dIMenSionaliTy the bactEriUM is fOoLeD by its oWn secretion suCh That it is geTtIng LocaliZED for long Times. In a subsequent numerICal studY of A modeL micRobe couplEd tO its owN chEMoattrActant SecreTeD at CONstanT RAtE, GrImA [@Grima_PRL_2005; @GRIMa_PrE_2006] calCuLateD the lonG-time dynamics in varIOus Dimensions $d$ anD foUnd lONG-tIme DIfFUsiVe BEhaVIOr even for $d=1,2$ in conTradiction To rEf. [@DeGennes_2004]. GrIMa aLsO studieD the casE of neGAtive chEmotaxis aNd finds in AlL dimENSioNs long-time DiffusioN or ballisTIc motIOn DepenDinG on the StRenGth $\laMbda<0$ whICh cOupleS

organism moving by chemota xis in its ownsec ret io n -- - \ [sec:level1\]I n trod uction =============== ===== == = ===Ch emota xis [@c h em o t axi s1 ;@ch em o ta xis2; @J ul] and Brownianmot io n [@doi86; @ d ho nt_book; @ pus ey] belong t o t he key p roc e sseswhi ch go vern t h e moti lity of m ic r oorgan i sms (e. g . b acte ria, amoeba and e n do t helial cells)[@Poon _r e vi e w ].Inthe simple st appr o ach, th e m i c r obe “smells" a ch emical andm ove s alon gthe gradie nt of t h e c oncentratio n fi eld of th e chem o attract a nt in o rder t o r eac h ef f ic ie ntl yt hes ec ret i onsource o fth e che mica l . T he o ppo site case of negativeche mota x isis re alize d in c ase t he mic robein tends to avoidanot her objec t w hi chis secr e ting t heche mical [ @Bonner ] . T hi s c he motactic drift issu p e ri mposed t o stoc h as ti c motiondu e t o fl u c tuati onsa ss ociatedwith t h eac tive pr oc ess of s elf -pr opuls i on o f themicroorg anism [@geier-notes] . Chemotaxis c a nl e ad to c lus ters of agg rega t ed b acte r ia [@ a ggreg ated_ cl u st e rs; @geier2] whichar e stil l emi tting chemoat tractant.H e re we st udyt he self-coupled s ituat ion wheret he micro organ ism “sme lls" itse l f , i.e. i t r eac tsche m o ta ctically to i t s own s ecreted ch emical. Th is“au toc he motaxis"is reali ze din a ggr egate d cluster sofdi ffe rentb acteri a iftheag gr e gat e is co n si d e redas a net pa rt icle. Ano t her realiz ation isa s i ngle b ac teriumwhich has bot han emitter a nda sens o r of thesame chemical. Tsori an d de Gen nes [@de Genn es_2004]hav e stud ied a simp le mod el fo rthi s situa t i on in d ifferent s p a tia l dim en sion s $d$ a nd find self-trapp i ngof the bacter ium ini t sown ch e moa tt r act a n t cloud for $d= 1,2$ but n ot fo r $d=3$. T h isme ans tha t for l ow di m ensiona lity thebacterium i s fo o l edby its own secreti on such t h at it is gett ing local iz edfor l ong ti m es. In a subse qu ent nu meric al study o f a model microbe coupl ed toits o wnchemoattr act a ntsecretedat c onstant ra te, Gr ima [ @Gr i ma_PR L_20 0 5; @G r ima_P RE_2 0 06] calcu l at edt h elong-time d y n a mic s invar i ous di mens ions $d$ and foun d long-time dif fusi v e be hav i or e ve n for $d=1,2$inco n t radictio nto Ref. [@d eGennes_ 20 0 4]. G rima a lso st udied t h e c a se ofnega tiv e chemota xis a n d finds i na ll dim ensi on s long -timed iffu s i on or ballisticmotio n depen d ing on t he streng t h $\ lambda<0$which coupl es

organism moving_by chemotaxis_in its own secretion --- \[sec:level1\]Introduction ========================== Chemotaxis_[@chemotaxis1; @chemotaxis2;_@Jul]_and Brownian_motion_[@doi86; @dhont_book; @pusey]_belong to the_key processes which govern_the motility of_microorganisms_(e.g. bacteria, amoeba and endothelial cells) [@Poon_review]. In the simplest approach, the microbe “smells"_a_chemical and_moves_along_the gradient of the concentration_field of the chemoattractant in_order to_reach efficiently the secretion source of the chemical._The_opposite case of_negative chemotaxis is realized in case the microbe intends_to avoid another object which is_secreting the chemical_[@Bonner]._This_chemotactic drift is superimposed_to stochastic motion due to fluctuations_associated with the active process of_self-propulsion of the microorganism [@geier-notes]. Chemotaxis can_lead to clusters of aggregated bacteria_[@aggregated_clusters; @geier2] which are still_emitting chemoattractant. Here_we study the self-coupled situation_where the microorganism_“smells" itself,_i.e. it reacts_chemotactically to its own secreted chemical._This “autochemotaxis" is_realized in aggregated clusters of different_bacteria_if the aggregate_is_considered_as a_net particle. Another_realization_is a_single_bacterium which has both an emitter_and_a sensor of the same chemical. Tsori_and de Gennes [@deGennes_2004]_have_studied a simple model_for this situation in different_spatial dimensions $d$ and find self-trapping_of the_bacterium in_its own chemoattractant cloud for $d=1,2$ but not for $d=3$. This_means that for low dimensionality the_bacterium is fooled by_its own_secretion_such that it_is_getting localized_for long times. In a subsequent numerical_study of_a model microbe coupled to its_own chemoattractant secreted at_constant_rate, Grima [@Grima_PRL_2005; @Grima_PRE_2006] calculated the_long-time dynamics in various dimensions $d$_and found long-time diffusive behavior_even_for_$d=1,2$ in contradiction to Ref. [@deGennes_2004]._Grima also studied the case of_negative chemotaxis and_finds in all dimensions long-time diffusion or_ballistic_motion depending on the strength $\lambda<0$_which_couples

|\sin \theta| + |\cos \theta|) d\mathcal{H}^{n-1} =$$ $$= \int_{x}^{z} (|\sin \theta| + |\cos \theta|) \sqrt{1 + (\tan \theta)^2} dx = \int_{x}^{z} (|\sin \theta| + |\cos \theta|) \frac{1}{|\cos \theta|} dx =$$ $$= \int_{x}^{z} (1 + |\tan \theta|) dx = |z - x| + \int_{x}^{z} |g'| dx \geq |z - x| + |t - y|,$$ where the inequality becomes equality iff $g$ is monotone (remember we assumed it to be $C^1$). Thus there are multiple functions minimizing this integral. 4\. Now we allow $\partial E_t$ to contain vertical intervals. The difference is purely technical, as we have to divide our integral into two parts. Let us suppose that the (orientated) length of $i-$th vertical interval equals $\lambda_i$, then we have $$\int_{\text{graph part of } \partial E_t} (|\sin \theta| + |\cos \theta|) dl + \int_{\text{vertical part of } \partial E_t} (1 + 0) dl = \int_{x}^{z} (1 + |g'|) dx + \sum_{i = 1}^{\infty} |\lambda_i| =$$ $$= \int_{x}^{z} |g'| dx + |z - x| + \sum_{i = 1}^{\infty} |\lambda_i| \geq |t - y - \sum_{i = 1}^{\infty} \lambda_i| + |z - x| + \sum_{i = 1}^{\infty} |\lambda_i| \geq |z - x| + |t - y|,$$ where the inequality becomes equality iff $g$ is monotone (remember we assumed it to be $C^1$) and all the vertical intervals are orientated in the same direction as $g'$. Thus there are multiple functions minimizing this integral. We have proved that in a class containing all smooth functions the problem of minimizing perimeter of a set $E_t$ doesn’t have a unique solution. 5\. Let us

|\sin \theta| + |\cos \theta|) d\mathcal{H}^{n-1 } = $ $ $ $ = \int_{x}^{z } (|\sin \theta| + |\cos \theta|) \sqrt{1 + (\tan \theta)^2 } dx = \int_{x}^{z } (|\sin \theta| + |\cos \theta|) \frac{1}{|\cos \theta| } dx = $ $ $ $ = \int_{x}^{z } (1 + |\tan \theta|) dx = |z - x| + \int_{x}^{z } |g'| dx \geq |z - x| + |t - y|,$$ where the inequality becomes equality iff $ g$ is monotone (remember we assumed it to be $ C^1 $). therefore there be multiple functions minimizing this built-in. 4\. Now we admit $ \partial E_t$ to contain erect intervals. The difference is purely technical, as we suffer to divide our integral into two parts. lease us suppose that the (orientated) duration of $ i-$th vertical interval equals $ \lambda_i$, then we have $ $ \int_{\text{graph part of } \partial E_t } (|\sin \theta| + |\cos \theta|) dl + \int_{\text{vertical region of } \partial E_t } (1 + 0) dl = \int_{x}^{z } (1 + |g'|) dx + \sum_{i = 1}^{\infty } |\lambda_i| = $ $ $ $ = \int_{x}^{z } |g'| dx + |z - x| + \sum_{i = 1}^{\infty } |\lambda_i| \geq |t - y - \sum_{i = 1}^{\infty } \lambda_i| + |z - x| + \sum_{i = 1}^{\infty } |\lambda_i| \geq |z - x| + |t - y|,$$ where the inequality becomes equality iff $ g$ is monotone (remember we bear it to be $ C^1 $) and all the vertical intervals are orientated in the same guidance as $ g'$. Thus there are multiple functions minimizing this integral. We have prove that in a class containing all smooth functions the problem of minimize perimeter of a set $ E_t$ does n’t have a unique solution. 5\. Let us

|\sin \theta| + |\cos \theta|) d\mathcxl{H}^{n-1} =$$ $$= \int_{x}^{z} (|\sin \theta| + |\cos \tgeta|) \sqrg{1 + (\tan \theta)^2} dx = \int_{x}^{z} (|\sin \tieta| + |\cos \theta|) \frac{1}{|\cos \theta|} db =$$ $$= \int_{x}^{z} (1 + |\tan \thwta|) vx = |z - x| + \int_{x}^{z} |j'| dx \geq |z - x| + |b - y|,$$ wkece the inequalijy becomes exuality iff $g$ hs mlnotone (remember we assumed it to bq $C^1$). Thux hhere are multyple suncfpoks minimizing this integral. 4\. Now se allox $\partial E_t$ to contain vertical intervald. Thf difference is pugely technixal, ww we have to divide oug integral jnto two parts. Let us suppose tfat tke (orientatgb) lejcth of $i-$th terticwl interval cauals $\nambda_i$, then we have $$\ikt_{\texv{graph part of } \partial E_v} (|\sin \theta| + |\cos \theja|) dl + \int_{\dert{vertical part of } \pqrrial G_t} (1 + 0) dl = \unt_{b}^{z} (1 + |j'|) ds + \sum_{l = 1}^{\mnfty} |\lambdz_i| =$$ $$= \int_{x}^{z} |g'| dx + |z - x| + \sum_{i = 1}^{\ingtr} |\lambda_i| \geq |f - y - \fui_{i = 1}^{\infty} \lambda_i| + |z - x| + \sum_{i = 1}^{\infty} |\lakbdz_i| \geq |z - x| + |t - y|,$$ where rhe inequality becomed equalitr iff $g$ is monotone (remember we assumed it to be $W^1$) and xll tme vdetlcal intervals are orientated in the same dirqdtook as $g'$. Thus therc are multiple funvtlomf minimizing jhis incsgdal. We have proved that ig a coass contwinimg all smooth functions the problem of iunimizing perimetex of a set $E_c$ doesm’t hafe a unique solution. 5\. Lec us

|\sin \theta| + |\cos \theta|) d\mathcal{H}^{n-1} =$$ (|\sin + |\cos \sqrt{1 + (\tan \theta| |\cos \theta|) \frac{1}{|\cos dx =$$ $$= (1 + |\tan \theta|) dx = - x| + \int_{x}^{z} |g'| dx \geq |z - x| + |t - where the inequality becomes equality iff $g$ is monotone (remember we assumed it be Thus are functions minimizing this integral. 4\. Now we allow $\partial E_t$ to contain vertical intervals. The difference purely technical, as we have to divide our into two parts. Let suppose that the (orientated) length $i-$th interval equals then have part of } E_t} (|\sin \theta| + |\cos \theta|) dl + \int_{\text{vertical part of } \partial E_t} (1 + 0) = \int_{x}^{z} |g'|) dx \sum_{i 1}^{\infty} =$$ $$= \int_{x}^{z} + |z - x| + \sum_{i \geq |t - y - \sum_{i = 1}^{\infty} + |z x| + \sum_{i = 1}^{\infty} |\lambda_i| |z - x| + |t - y|,$$ where inequality becomes equality iff $g$ is monotone (remember we assumed it to be $C^1$) and vertical intervals are orientated the same direction $g'$. there multiple minimizing this We have proved that in a class containing all smooth functions problem of minimizing perimeter of a set $E_t$ doesn’t have solution. Let us

|\sin \theta| + |\cos \theta|) d\mathcal{H}^{N-1} =$$ $$= \int_{x}^{z} (|\sin \tHeta| + |\cOs \tHetA|) \sQrt{1 + (\tAn \thEta)^2} dx = \int_{x}^{z} (|\sin \tHEta| + |\cOs \theta|) \frac{1}{|\cos \theta|} dx =$$ $$= \iNt_{x}^{z} (1 + |\tAn \THeta|) DX = |z - X| + \int_{x}^{Z} |g'| dx \geq |Z - X| + |t - Y|,$$ WHerE tHe IneQuALiTy becOmeS equaliTy iff $g$ is moNotOnE (remember we aSSuMed it to be $C^1$). thuS there are mulTipLe funcTiOns MInimiZinG this IntegrAL. 4\. Now we Allow $\partIaL e_t$ to coNTain verTICaL intErvals. The differenCE iS Purely technicaL, as we hAvE To DIVidE ouR integral iNtO two pARts. Let uS SuPPOSe tHAt the (orientatEd) length of $i-$TH veRtical InTerVAl equaLs $\lamBdA_I$, thEn we have $$\int_{\Text{Graph part Of } \partIAl E_t} (|\sin \THeta| + |\cos \Theta|) dL + \inT_{\teXt{veRTiCaL paRt OF } \paRTiAl E_T} (1 + 0) Dl = \iNt_{x}^{z} (1 + |g'|) dx + \sUm_{I = 1}^{\iNfty} |\lAmbdA_I| =$$ $$= \INT_{x}^{z} |g'| Dx + |z - X| + \sum_{I = 1}^{\inftY} |\lambda_i| \geq |t - y - \Sum_{I = 1}^{\infTY} \laMbda_i| + |Z - x| + \sum_{I = 1}^{\infTy} |\LambdA_i| \geq |z - X| + |t - y|,$$ whErE the inequality bEcomEs equalitY ifF $g$ Is mOnOtone (REmembeR we AssUmed it tO be $C^1$) and ALl tHe VERTiCal intervals are oriEnTATeD in the saMe direCTiOn AS $g'$. Thus thErE arE mulTIPle fuNctiONs MinimiziNg this INtEgRal. We haVe Proved ThAt iN a cLass cONtaiNing alL smooth fUnctiONs the problem of MInimizing periMEtER Of A Set $E_T$ doEsn’t have a unIque SOlutIon. 5\. LET uS

|\sin \theta| + |\cos \the ta|) d\mat hcal{ H}^ {n- 1} =$$ $$ = \int_{x}^{z} (|\s in \theta| + |\cos \th eta|) \ s qrt{ 1 + (\ta n \thet a )^ 2 } dx = \ int _{ x }^ {z} ( |\s in \the ta| + |\co s \ th eta|) \frac{ 1 }{ |\cos \the ta| } dx =$$ $$ = \ int_{x }^ {z} (1 +|\t an \t heta|) dx = | z - x| +\i n t_{x}^ { z} |g'| d x\geq |z - x| + |t - y | ,$ $ where the in equali ty be c o mes eq uality iff $ g$ is monoton e ( r e m emb e r we assumedit to be $C ^ 1$) . Thus t her e are m ultip le fun ctions mini mizi ng this i ntegra l . 4\.N ow we a llow $ \pa rti al E _ t$ t o c on t ain ve rti c alinterval s. T he di ffer e n c e ispur elytechn ical, as we h ave tod ivi de ou r int egra lintotwo pa rts.Le t us suppose th at t he (orien tat ed ) l en gth o f $i-$t h v ert ical in tervale qua ls $ \ la mbda_i$, then we h av e $ $\int_{\ text{g r ap hp art of } \ par tial E _t} ( |\si n \ theta| + |\cos \t he ta|) dl + \int_ {\ tex t{v ertic a l pa rt of} \parti al E_ t } (1 + 0) dl = \int_{x}^{z}( 1+ |g ' |) d x + \sum_{i =1}^{ \ inft y} | \ la mbd a _i| = $$ $ $= \i n t_{x}^{z} |g'| dx + | z - x| + \s um_{i = 1}^{\ infty} |\l a m b da_i| \g eq | t - y - \sum_{i =1}^{\ infty} \la m bda_i| + |z - x| + \s um_{i = 1 } ^ {\infty} |\ lam bda _i| \ ge q |z - x| + | t - y| ,$ $ wher e t he ineq ual ity be com es equality iff $g$ i smo no ton e (re m ember we a ssu me d i t tob e $C^1 $) an d al lth e ve rticali nt e r vals a re ori ent at ed in the sam e direc tion as $ g'$ . Thu sth ere are multiple fun ct ions minim iz ing thisi n tegral.We have proved that ina classcon taini ng a ll smooth fu nction s t h e prob lem of mini mi zin g perim e t er of a set $E_t$ d oes n’t h av e auniquesolution. 5\. Let us

|\sin \theta|_+ |\cos_\theta|) d\mathcal{H}^{n-1} =$$ $$= \int_{x}^{z}_(|\sin \theta|_+_|\cos \theta|)_\sqrt{1_+ (\tan \theta)^2}_dx = \int_{x}^{z}_(|\sin \theta| + |\cos_\theta|) \frac{1}{|\cos \theta|}_dx_=$$ $$= \int_{x}^{z} (1 + |\tan \theta|) dx = |z - x| + \int_{x}^{z} |g'|_dx_\geq |z_-_x|_+ |t - y|,$$ where the_inequality becomes equality iff $g$_is monotone_(remember we assumed it to be $C^1$). Thus_there_are multiple functions_minimizing this integral. 4\. Now we allow $\partial E_t$ to_contain vertical intervals. The difference is_purely technical, as_we_have_to divide our integral_into two parts. Let us suppose_that the (orientated) length of $i-$th_vertical interval equals $\lambda_i$, then we have $$\int_{\text{graph_part of } \partial E_t} (|\sin_\theta| + |\cos \theta|) dl_+ \int_{\text{vertical_part of } \partial E_t}_(1 + 0)_dl =_\int_{x}^{z} (1 +_|g'|) dx + \sum_{i = 1}^{\infty}_|\lambda_i| =$$ $$= \int_{x}^{z}_|g'| dx + |z - x|_+_\sum_{i = 1}^{\infty}_|\lambda_i|_\geq_|t -_y - \sum_{i_=_1}^{\infty} \lambda_i|_+_|z - x| + \sum_{i =_1}^{\infty}_|\lambda_i| \geq |z - x| + |t_- y|,$$ where the inequality_becomes_equality iff $g$ is_monotone (remember we assumed it_to be $C^1$) and all the_vertical intervals_are orientated_in the same direction as $g'$. Thus there are multiple functions_minimizing this integral. We have proved_that in a class_containing all_smooth_functions the problem_of_minimizing perimeter_of a set $E_t$ doesn’t have a_unique solution. 5\._Let us

assumption can be written as $\sum^r_{i=1}b^j_{i}\alpha_i$. We can lift $f_i$ to $g_i\in R$ such that ${{\bf in}}_v(g_i)=f_i$, and make the morphism $$\psi\colon k[t_1,...,t_k]\to R,\qquad t_i\to g_i.$$ Consider the filtration $$\{ F_b\subset k[t_1,...,t_k]|b\in \Phi, f\in F_b\mbox{ iff all monomials of $f$ have degree at least $b$} \}$$ for any $b\in \Phi$. We can similarly construct a Rees algebra $$\mathcal{R}^*= \sum_{b\in {\Phi}^g} F_{b} s^{b} \subset k[t_1,...,t_k]\otimes k[\Phi^g].$$ There is a surjection $\mathcal{R}^*\to \mathcal{R}_{v_{\alpha}}$, which degenerates $\psi$ to $\phi$ over ${\rm Spec}R(\Phi)$. Denote by $I$ the kernel of $\psi$, which we know degenerates to $I_0$ the Kernel of $\phi$. By the flatness over $R(\Phi)$, we know any element $h'\in I_0$ can be lift to an element $h\in I$. Geometrically, this gives a pointed embedding $(x\in X)\subset (0\in \mathbb{C}^k)$, whose degeneration along one direction, denoted by $\xi_{\alpha}$, on $\mathbb{C}^n$ gives the embedding $(o\in X_0={\rm Spec}({\rm gr}_vR))\subset (0\in \mathbb{C}^r)$. Let $h_1,....,h_m$ be elements in $I$ whose degenerations $h_1',..., h_m'$ generate $I_0$. Assume $h_j=h'_j+h_j''$ with $\deg(h'_j)=\sum^r_{i=1} c^j_i\alpha_i$ and the monomials of $h_j''$ has degree larger than $\sum^r_{i=1} c^j_i\alpha_i$. We can choose $\epsilon$ sufficiently small such that if

assumption can be written as $ \sum^r_{i=1}b^j_{i}\alpha_i$. We can lift $ f_i$ to $ g_i\in R$ such that $ { { \bf in}}_v(g_i)=f_i$, and make the morphism $ $ \psi\colon k[t_1,... ,t_k]\to R,\qquad t_i\to g_i.$$ think the filtration $ $ \ { F_b\subset k[t_1,... ,t_k]|b\in \Phi, f\in F_b\mbox { iff all monomials of $ f$ take degree at least $ b$ } \}$$ for any $ b\in \Phi$. We can similarly construct a Rees algebra $ $ \mathcal{R}^*= \sum_{b\in { \Phi}^g } F_{b } s^{b } \subset k[t_1,... ,t_k]\otimes k[\Phi^g].$$ There is a surjection $ \mathcal{R}^*\to \mathcal{R}_{v_{\alpha}}$, which devolve $ \psi$ to $ \phi$ over $ { \rm Spec}R(\Phi)$. Denote by $ I$ the kernel of $ \psi$, which we know degenerates to $ I_0 $ the Kernel of $ \phi$. By the two-dimensionality over $ R(\Phi)$, we know any element $ h'\in I_0 $ can be rustle to an element $ h\in I$. Geometrically, this gives a pointed embed $ (x\in X)\subset (0\in \mathbb{C}^k)$, whose degeneration along one direction, denote by $ \xi_{\alpha}$, on $ \mathbb{C}^n$ give the embedding $ (o\in X_0={\rm Spec}({\rm gr}_vR))\subset (0\in \mathbb{C}^r)$. Let $ h_1,.... ,h_m$ be component in $ I$ whose degenerations $ h_1', ..., h_m'$ generate $ I_0$. Assume $ h_j = h'_j+h_j''$ with $ \deg(h'_j)=\sum^r_{i=1 } c^j_i\alpha_i$ and the monomials of $ h_j''$ has degree bombastic than $ \sum^r_{i=1 } c^j_i\alpha_i$. We can choose $ \epsilon$ sufficiently small such that if

asdumption can be written xs $\sum^r_{i=1}b^j_{i}\alphc_u$. We cen lift $f_i$ to $g_k\in R$ such that ${{\bf in}}_v(g_i)=f_i$, aid mqke tye morphism $$\psi\colon k[g_1,...,t_k]\to R,\qqlad t_i\to t_i.$$ Cibsider the filtratljn $$\{ R_n\subszt k[t_1,...,t_k]|b\in \Phi, f\ln F_b\mbox{ ixf all monomians oy $f$ have degree at least $b$} \}$$ for any $b\in \Pho$. Ae can similarjy cpgstrhbt a Rees algebra $$\mathcal{R}^*= \sum_{b\jn {\Phi}^j} F_{b} s^{b} \subset l[t_1,...,t_k]\otimes k[\Phi^g].$$ There is w sugjection $\mathcal{R}^*\tl \mathcal{R}_{v_{\qlphw}}$, which degendrates $\psi$ to $\phi$ over ${\rm Spec}R(\Phi)$. Denote by $I$ the kefnel pf $\psi$, whixh we nnow degenecates no $I_0$ the Kernck of $\pvi$. By tne flatness ovcr $R(\Pii)$, ww know any element $h'\ii I_0$ can be lift to ag element $h\nn I$. Geometrically, thus givev a [oingwd dmbtddmng $(x\in X)\dubaet (0\in \matgbb{C}^k)$, whose degeneration along ogv direction, dsnoted br $\xi_{\alpha}$, on $\mathbb{C}^n$ gives the embedditg $(k\in X_0={\rm Spec}({\rm gr}_vR))\subswt (0\in \mathbb{C}^r)$. Let $h_1,....,h_m$ be elemegts in $I$ whose degenerations $h_1',..., h_m'$ generate $I_0$. Assuke $h_j=i'_j+f_j''$ clbh $\ddt(h'_u)=\sum^r_{i=1} c^j_i\alpha_i$ and the monomials of $h_j''$ has qsgtev larger than $\sum^v_{i=1} c^j_i\alpha_i$. We cam fhpjse $\epsilon$ sofficieuflg small such that lf

assumption can be written as $\sum^r_{i=1}b^j_{i}\alpha_i$. We $f_i$ $g_i\in R$ that ${{\bf in}}_v(g_i)=f_i$, k[t_1,...,t_k]\to t_i\to g_i.$$ Consider filtration $$\{ F_b\subset \Phi, f\in F_b\mbox{ iff all monomials $f$ have degree at least $b$} \}$$ for any $b\in \Phi$. We can construct a Rees algebra $$\mathcal{R}^*= \sum_{b\in {\Phi}^g} F_{b} s^{b} \subset k[t_1,...,t_k]\otimes k[\Phi^g].$$ There a $\mathcal{R}^*\to which $\psi$ to $\phi$ over ${\rm Spec}R(\Phi)$. Denote by $I$ the kernel of $\psi$, which we know to $I_0$ the Kernel of $\phi$. By the over $R(\Phi)$, we know element $h'\in I_0$ can be to element $h\in Geometrically, gives pointed embedding $(x\in (0\in \mathbb{C}^k)$, whose degeneration along one direction, denoted by $\xi_{\alpha}$, on $\mathbb{C}^n$ gives the embedding $(o\in X_0={\rm gr}_vR))\subset (0\in $h_1,....,h_m$ be in whose $h_1',..., h_m'$ generate $h_j=h'_j+h_j''$ with $\deg(h'_j)=\sum^r_{i=1} c^j_i\alpha_i$ and the has degree larger than $\sum^r_{i=1} c^j_i\alpha_i$. We can $\epsilon$ sufficiently such that if

assumption can be written as $\sUm^r_{i=1}b^j_{i}\alpHa_i$. We Can LifT $f_I$ to $g_I\in R$ Such that ${{\bf in}}_v(g_I)=F_i$, anD make the morphism $$\psi\colOn k[t_1,...,t_K]\tO r,\qquAD t_I\to g_i.$$ consideR ThE FIltRaTiOn $$\{ F_B\sUBsEt k[t_1,...,t_K]|b\iN \Phi, f\in f_b\mbox{ iff aLl mOnOmials of $f$ havE DeGree at leasT $b$} \}$$ fOr any $b\in \Phi$. WE caN similArLy cONstruCt a rees aLgebra $$\MAthcal{r}^*= \sum_{b\in {\PhI}^g} f_{B} s^{b} \subSEt k[t_1,...,t_k]\oTIMeS k[\PhI^g].$$ There is a surjectIOn $\MAthcal{R}^*\to \mathcAl{R}_{v_{\alPhA}}$, WhICH deGenErates $\psi$ tO $\pHi$ oveR ${\Rm Spec}R(\pHi)$. dENOte BY $I$ the kernel of $\Psi$, which we kNOw dEgenerAtEs tO $i_0$ the KeRnel oF $\pHI$. By The flatness Over $r(\Phi)$, we knoW any elEMent $h'\in i_0$ Can be liFt to an EleMenT $h\in i$. geOmEtrIcALly, THiS giVEs a Pointed eMbEdDing $(x\In X)\sUBSET (0\in \mAthBb{C}^k)$, Whose Degeneration aLonG one DIreCtion, DenotEd by $\Xi_{\Alpha}$, On $\mathBb{C}^n$ gIvEs the embedding $(o\In X_0={\rM Spec}({\rm gr}_VR))\sUbSet (0\In \MathbB{c}^r)$. Let $h_1,....,H_m$ bE elEments iN $I$ whose DEgeNeRATIoNs $h_1',..., h_m'$ generate $I_0$. AssuMe $H_J=H'_j+H_j''$ with $\deG(h'_j)=\sum^R_{I=1} c^J_i\ALpha_i$ and ThE moNomiALS of $h_j''$ Has dEGrEe larger Than $\suM^R_{i=1} C^j_I\alpha_i$. we Can choOsE $\epSilOn$ sufFIcieNtly smAll such tHat if

assumption can be written as $\sum^ r_{i= 1}b ^j_ {i }\al pha_ i$. We can lif t $f_ i$ to $g_i\in R$ suchthat${ { \bfi n} }_v(g _i)=f_i $ ,a n d m ak ethe m o rp hism$$\ psi\col on k[t_1,. .., t_ k]\to R,\qqu a dt_i\to g_i .$$ Consider th e f iltrat io n $ $ \{ F_ b\s ubset k[t_1 , ...,t_ k]|b\in \ Ph i , f\in F_b\mbo x { i ff a ll monomials of $ f $h ave degree atleast$b $ }\ }$$ fo r any $b\i n\Phi$ . We can si m i l arl y construct aRees algebr a $$ \mathc al {R} ^ *= \su m_{b\ in {\ Phi}^g} F_{ b} s ^{b} \sub set k[ t _1,..., t _k]\oti mes k[ \Ph i^g ].$$ Th er e i sa su r je cti o n $ \mathcal {R }^ *\to\mat h c a l {R}_ {v_ {\al pha}} $, which dege ner ates $\p si$ t o $\p hi$ov er ${ \rm Sp ec}R( \P hi)$. Denote by $I$ the kern elof $\ ps i$, w h ich we kn owdegener ates to $I_ 0$ t h eKernel of $\phi$.By t he flatnes s over $R (\ P hi)$, we k now any e lemen t $h ' \i n I_0$ c an bel if tto an e le ment $ h\ inI$. Geom e tric ally,this giv es ap ointed embeddi n g $(x\in X)\s u bs e t ( 0 \in\ma thbb{C}^k)$ , wh o se d egen e ra tio n alon g one d i re c tion, denoted by $\ xi _{\alp ha}$, on $\mathbb{ C}^n$ give s t he embed ding $( o \in X_0={\rm S pec}( {\rm gr}_v R ))\subse t (0\ in \math bb{C}^r)$ . Let $h_ 1,. ... ,h_ m$b e e lements in $I $ whos edegener ati ons $h_ 1', ... , h _m' $generate$I_0$. A ss um e$h _j= h'_j+ h _j''$ wi th $\ de g(h '_j)= \ sum^r_ {i=1} c^j _i \a l pha _i$ and th e mono mi al s of $h _j ''$ h as d e gre e large r than $\ sum ^ r_{i =1 }c^j_i\a lpha_i$. We c an choose $\ ep sil on$ su f f iciently small such that if

assumption_can be_written as $\sum^r_{i=1}b^j_{i}\alpha_i$. We_can lift_$f_i$_to $g_i\in_R$_such that ${{\bf_in}}_v(g_i)=f_i$, and make_the morphism $$\psi\colon k[t_1,...,t_k]\to_R,\qquad t_i\to g_i.$$_Consider_the filtration $$\{ F_b\subset k[t_1,...,t_k]|b\in \Phi, f\in F_b\mbox{ iff all monomials of $f$ have_degree_at least_$b$}__\}$$ for any $b\in \Phi$._We can similarly construct a_Rees algebra_$$\mathcal{R}^*= \sum_{b\in {\Phi}^g} F_{b} s^{b} \subset k[t_1,...,t_k]\otimes_k[\Phi^g].$$_There is a_surjection $\mathcal{R}^*\to \mathcal{R}_{v_{\alpha}}$, which degenerates $\psi$ to $\phi$ over_${\rm Spec}R(\Phi)$. Denote by $I$ the_kernel of $\psi$,_which_we_know degenerates to $I_0$_the Kernel of $\phi$. By the_flatness over $R(\Phi)$, we know any_element $h'\in I_0$ can be lift to_an element $h\in I$. Geometrically, this_gives a pointed embedding $(x\in_X)\subset (0\in_\mathbb{C}^k)$, whose degeneration along one_direction, denoted by_$\xi_{\alpha}$, on_$\mathbb{C}^n$ gives the_embedding $(o\in X_0={\rm Spec}({\rm gr}_vR))\subset (0\in_\mathbb{C}^r)$. Let $h_1,....,h_m$ be_elements in $I$ whose degenerations $h_1',...,_h_m'$_generate $I_0$. Assume_$h_j=h'_j+h_j''$_with_$\deg(h'_j)=\sum^r_{i=1} c^j_i\alpha_i$_and the monomials_of_$h_j''$ has_degree_larger than $\sum^r_{i=1} c^j_i\alpha_i$. We can_choose_$\epsilon$ sufficiently small such that if

paper are $m\times m$ matrices, where $m=f^0$. The blocks of the $m\times m$ matrix $T$ is set as follows: for each $0\leq \alpha, \beta\leq n$, the $(\alpha, \beta)$-th block $T^{\alpha, \beta}$ is $$\begin{aligned} \label{block} T^{\alpha, \beta}=\left[T_{ij}(\tau)\right]_{f^{-\alpha+n+1}\leq i \leq f^{-\alpha+n}-1, \ f^{-\beta+n+1}\leq j\leq f^{-\beta+n}-1},\end{aligned}$$ where $T_{ij}$ is the entries of the matrix $T$, and $f^{n+1}$ is defined to be zero. In particular, $T =[T^{\alpha,\beta}]$ is called a *block lower triangular matrix* if $T^{\alpha,\beta}=0$ whenever $\alpha<\beta$. We remark that by fixing a base point, we can identify the above quotient Lie groups or Lie algebras with their orbits in the corresponding quotient Lie algebras or Lie groups. For example, $\mathfrak{n}_+\cong \mathfrak{g}/\mathfrak{b}$, $\mathfrak{g}^{-1,1}\cong\mathfrak{b}\oplus \mathfrak{g}^{-1,1}/\mathfrak{b}$, and $N_+\cong N_+B/B\subseteq \check{D}$. We can also identify a point $\Phi(p)=\{ F^n_p\subseteq F^{n-1}_p\subseteq \cdots \subseteq F^{0}_p\}\in D$ with its Hodge decomposition $\bigoplus_{k=0}^n H^{k, n-k}_p$, and thus with any fixed adapted basis of the corresponding Hodge decomposition for the base point, we have matrix representations of elements in the above Lie groups and Lie algebras. For example, elements in $N_+$ can be realized as nonsingular block lower triangular matrices with identity blocks in the diagonal; elements in $B$ can be realized as nonsingular block upper triangular matrices. We shall review and collect some facts about the structure of simple Lie algebra $\mathfrak{g}$ in our case. Again one may refer to [@GS] and [@schmid1] for more details. Let $\theta: \,\mathfrak{g}\rightarrow \mathfrak{g}$ be the We

paper are $ m\times m$ matrices, where $ m = f^0$. The blocks of the $ m\times m$ matrix $ T$ is specify as take after: for each $ 0\leq \alpha, \beta\leq n$, the $ (\alpha, \beta)$-th block $ T^{\alpha, \beta}$ is $ $ \begin{aligned } \label{block } T^{\alpha, \beta}=\left[T_{ij}(\tau)\right]_{f^{-\alpha+n+1}\leq i \leq f^{-\alpha+n}-1, \ f^{-\beta+n+1}\leq j\leq f^{-\beta+n}-1},\end{aligned}$$ where $ T_{ij}$ is the entries of the matrix $ T$, and $ f^{n+1}$ is defined to be zero. In finical, $ T =[ T^{\alpha,\beta}]$ is called a * block low triangular matrix * if $ T^{\alpha,\beta}=0 $ whenever $ \alpha<\beta$. We remark that by fixing a foundation point, we can identify the above quotient Lie group or Lie algebras with their orbits in the comparable quotient Lie algebras or Lie groups. For case, $ \mathfrak{n}_+\cong \mathfrak{g}/\mathfrak{b}$, $ \mathfrak{g}^{-1,1}\cong\mathfrak{b}\oplus \mathfrak{g}^{-1,1}/\mathfrak{b}$, and $ N_+\cong N_+B / B\subseteq \check{D}$. We can also identify a detail $ \Phi(p)=\ { F^n_p\subseteq F^{n-1}_p\subseteq \cdots \subseteq F^{0}_p\}\in D$ with its Hodge decomposition $ \bigoplus_{k=0}^n H^{k, n - k}_p$, and thus with any fixed adapted footing of the corresponding Hodge decomposition for the basis point, we experience matrix representations of elements in the above Lie groups and Lie algebras. For example, element in $ N_+$ can be realized as nonsingular block lower triangular matrices with identity blocks in the diagonal; elements in $ B$ can be realized as nonsingular block upper triangular matrices. We shall review and collect some facts about the structure of simple Lie algebra $ \mathfrak{g}$ in our case. Again one may refer to [ @GS ] and [ @schmid1 ] for more details. permit $ \theta: \,\mathfrak{g}\rightarrow \mathfrak{g}$ be the We

paoer are $m\times m$ matricer, where $m=f^0$. The yoocks mf the $m\times o$ matrix $T$ is set as follows: fir eaxh $0\leq \alpha, \beta\leq n$, the $(\alphw, \beta)$-th blork $T^{\alpha, \beta}$ is $$\begin{aligned} \mwbel{ylick} T^{\alpha, \beta}=\keft[T_{ij}(\tau)\sight]_{f^{-\alpha+n+1}\lex k \peq f^{-\alpha+n}-1, \ f^{-\beta+n+1}\leq j\leq f^{-\beta+n}-1},\egd{alignrd}$$ where $T_{ij}$ is jhe emeriea of the matrix $T$, and $f^{n+1}$ is defines to be zero. In partocular, $T =[T^{\alpha,\beta}]$ is calped w *block lower triajgular matrux* is $T^{\alpha,\beta}=0$ dhenever $\alpha<\beta$. We rgmark that by fixing a base poing, we ean identifi fhf above quotment Lpe groups or Lie algetras wiyh their orbitx ii thw corresponding quotixnt Lie algebras or Jie groupv. Yor example, $\mathfrak{n}_+\xobg \majhfran{g}/\magyfrxk{b}$, $\methrrak{g}^{-1,1}\clng\jathfrak{b}\ollus \mathfrqk{g}^{-1,1}/\mathfrak{b}$, and $N_+\cpnd N_+B/B\subseteq \dheck{D}$. Wq can also identify a point $\Phi(p)=\{ F^n_p\subvetsq F^{n-1}_p\subseteq \cdots \suvseteq F^{0}_p\}\in D$ with itd Hodge dqcomposition $\bigoplus_{k=0}^n H^{k, n-k}_p$, and thus with any xixed xdaktcq vadis of the corresponding Hodge decomposition skr tme base point, we have matric geltesentations ow elemzhta in the above Lie groups and Oie algebwas. Gor example, elements in $N_+$ cqn be realizvd aw nonsingular blocn lower tricngulat matroces with identity blocys ih the diagojal; elemehgs in $B$ can be rdalpzed as nonsingular block uppew trianguoar katricer. We xhall weview and collcwt some facts abouh the sdructure ov simple Lie algebra $\mathfrak{g}$ mi our case. Agsit ote may rzfer tp [@GS] and [@schmyd1] for more dejails. Let $\thetx: \,\mathfrak{f}\rightacrow \mathfrah{g}$ be the We

paper are $m\times m$ matrices, where $m=f^0$. of $m\times m$ $T$ is set \alpha, n$, the $(\alpha, block $T^{\alpha, \beta}$ $$\begin{aligned} \label{block} T^{\alpha, \beta}=\left[T_{ij}(\tau)\right]_{f^{-\alpha+n+1}\leq i \leq \ f^{-\beta+n+1}\leq j\leq f^{-\beta+n}-1},\end{aligned}$$ where $T_{ij}$ is the entries of the matrix $T$, $f^{n+1}$ is defined to be zero. In particular, $T =[T^{\alpha,\beta}]$ is called a lower matrix* $T^{\alpha,\beta}=0$ $\alpha<\beta$. We remark that by fixing a base point, we can identify the above quotient Lie or Lie algebras with their orbits in the quotient Lie algebras or groups. For example, $\mathfrak{n}_+\cong \mathfrak{g}/\mathfrak{b}$, \mathfrak{g}^{-1,1}/\mathfrak{b}$, $N_+\cong N_+B/B\subseteq We also a point $\Phi(p)=\{ F^{n-1}_p\subseteq \cdots \subseteq F^{0}_p\}\in D$ with its Hodge decomposition $\bigoplus_{k=0}^n H^{k, n-k}_p$, and thus with any fixed basis of Hodge decomposition the point, have matrix representations in the above Lie groups and example, elements in $N_+$ can be realized as block lower matrices with identity blocks in the elements in $B$ can be realized as nonsingular upper triangular matrices. We shall review and collect some facts about the structure of simple $\mathfrak{g}$ in our case. one may refer [@GS] [@schmid1] more Let $\theta: \mathfrak{g}$ be the We

paper are $m\times m$ matrices, whEre $m=f^0$. The blOcks oF thE $m\tImEs m$ mAtriX $T$ is set as folloWS: for Each $0\leq \alpha, \beta\leq n$, thE $(\alphA, \bETa)$-th BLoCk $T^{\alPha, \beta}$ IS $$\bEGIn{aLiGnEd} \lAbEL{bLock} T^{\AlpHa, \beta}=\lEft[T_{ij}(\tau)\rIghT]_{f^{-\Alpha+n+1}\leq i \leQ F^{-\aLpha+n}-1, \ f^{-\beta+N+1}\leQ j\leq f^{-\beta+n}-1},\eNd{aLigned}$$ WhEre $t_{Ij}$ is tHe eNtrieS of the MAtrix $T$, And $f^{n+1}$ is deFiNEd to be ZEro. In paRTIcUlar, $t =[T^{\alpha,\beta}]$ is callED a *BLock lower trianGular mAtRIx* IF $t^{\alPha,\Beta}=0$ whenevEr $\Alpha<\BEta$. We reMArK THAt bY Fixing a base poInt, we can ideNTifY the abOvE quOTient LIe groUpS Or LIe algebras wIth tHeir orbitS in the COrrespoNDing quoTient LIe aLgeBras OR LIe GroUpS. for EXaMplE, $\MatHfrak{n}_+\coNg \MaThfraK{g}/\maTHFRAk{b}$, $\mAthFrak{G}^{-1,1}\cong\Mathfrak{b}\opluS \maThfrAK{g}^{-1,1}/\mAthfrAk{b}$, anD $N_+\coNg n_+B/B\suBseteq \Check{d}$. WE can also identifY a poInt $\Phi(p)=\{ F^n_P\suBsEteQ F^{N-1}_p\subSEteq \cdOts \SubSeteq F^{0}_p\}\In D$ with ITs HOdGE DEcOmposition $\bigoplus_{K=0}^n h^{K, N-k}_P$, and thus With anY FiXeD Adapted bAsIs oF the CORrespOndiNG HOdge decoMpositIOn FoR the basE pOint, we HaVe mAtrIx repREsenTationS of elemeNts in THe above Lie grouPS and Lie algebrAS. FOR ExAMple, EleMents in $N_+$ can Be reALizeD as nONsIngULar blOck loWeR TrIAngular matrices with IdEntity BlockS in the diagonaL; elements iN $b$ CAn be realIzed AS nONsingular block Upper Triangular MAtrices. WE shalL review aNd collect SOMe facts aBouT thE stRucTURe Of simple Lie alGEBra $\mAtHfrak{g}$ iN ouR case. AgAin One May RefEr To [@GS] and [@scHmid1] for mOrE dEtAiLs. LEt $\theTA: \,\mathfraK{g}\RigHtArrOw \matHFrak{g}$ bE the WE

paper are $m\times m$ mat rices, whe re $m =f^ 0$. T he b lock s of the $m\ti m es m $ matrix $T$ is set as foll ow s : fo r e ach $ 0\leq \ a lp h a , \ be ta \le qn $, the$(\ alpha,\beta)$-th bl oc k $T^{\alpha , \ beta}$ is$$\ begin{aligne d}\label {b loc k } T^{ \al pha,\beta} = \left[ T_{ij}(\t au ) \right ] _{f^{-\ a l ph a+n+ 1}\leq i \leq f^{ - \a l pha+n}-1, \ f^ {-\bet a+ n +1 } \ leq j\ leq f^{-\b et a+n}- 1 },\end{ a li g n e d}$ $ where $T_{ij }$ is the e n tri es ofth e m a trix $ T$, a nd $f^ {n+1}$ is d efin ed to bezero.I n parti c ular, $ T =[T^ {\a lph a,\b e ta }] $ i sc all e da * b loc k lowertr ia ngula r ma t r i x * if $T ^{\a lpha, \beta}=0$ whe nev er $ \ alp ha<\b eta$. We r emark thatby fi xi ng a base point , we can iden tif ythe a boveq uotien t L iegroupsor Liea lge br a s wi th their orbits in t h e c orrespon ding q u ot ie n t Lie al ge bra s or L ie gr oups . F or examp le, $\ m at hf rak{n}_ +\ cong \ ma thf rak {g}/\ m athf rak{b} $, $\mat hfrak { g}^{-1,1}\cong \ mathfrak{b}\o p lu s \m a thfr ak{ g}^{-1,1}/\ math f rak{ b}$, an d $ N _+\co ng N_ +B / B\ s ubseteq \check{D}$. W e canalsoidentify a po int $\Phi( p ) = \{ F^n_p \sub s et e q F^{n-1}_p\su bsete q \cdots \ s ubseteqF^{0} _p\}\inD$ with i t s Hodge d eco mpo sit ion $ \b igoplus_{k=0} ^ n H^{ k, n-k}_p $,and thu s w ith an y f ix ed adapte d basisof t he c orr espon d ing Hodg edec om pos ition for th e bas e po in t, wehave ma t ri x repr es en tati ons o f ele ment s in the ab ove Lie g rou p s an dLi e algeb ras. For exam pl e, element sin$N_+$c a n be rea lized as nonsingular bl o ck lowe r t riang ular matrices wi th ide nti t y bloc ks inthe d ia gon a l ; ele m e nt s i n$B$ can be r eal izedas non singula r block upper tria n gul ar matrices. We sha l l r evi e wa ndco l lec t some facts abou t the stru ct u re of simple Lie a lgebra$\mathf rak{g } $ in ou r case. A gain onema y re f e r t o [@GS] an d [@schm id1] form ore d e ta ils.Let $\the ta : \ ,\mat hfrak{ g }\r ighta rrow \ ma thfrak {g}$be the We

paper_are $m\times_m$ matrices, where $m=f^0$._The blocks_of_the $m\times_m$_matrix $T$ is_set as follows:_for each $0\leq \alpha,_\beta\leq n$, the_$(\alpha,_\beta)$-th block $T^{\alpha, \beta}$ is $$\begin{aligned} \label{block} T^{\alpha, \beta}=\left[T_{ij}(\tau)\right]_{f^{-\alpha+n+1}\leq i \leq f^{-\alpha+n}-1, \ f^{-\beta+n+1}\leq j\leq f^{-\beta+n}-1},\end{aligned}$$_where_$T_{ij}$ is_the_entries_of the matrix $T$, and_$f^{n+1}$ is defined to be_zero. In_particular, $T =[T^{\alpha,\beta}]$ is called a *block lower_triangular_matrix* if $T^{\alpha,\beta}=0$_whenever $\alpha<\beta$. We remark that by fixing a base point,_we can identify the above quotient_Lie groups or_Lie_algebras_with their orbits in_the corresponding quotient Lie algebras or_Lie groups. For example, $\mathfrak{n}_+\cong \mathfrak{g}/\mathfrak{b}$,_$\mathfrak{g}^{-1,1}\cong\mathfrak{b}\oplus \mathfrak{g}^{-1,1}/\mathfrak{b}$, and $N_+\cong N_+B/B\subseteq \check{D}$. We_can also identify a point $\Phi(p)=\{_F^n_p\subseteq F^{n-1}_p\subseteq \cdots \subseteq F^{0}_p\}\in_D$ with_its Hodge decomposition $\bigoplus_{k=0}^n H^{k,_n-k}_p$, and thus_with any_fixed adapted basis_of the corresponding Hodge decomposition for_the base point,_we have matrix representations of elements_in_the above Lie_groups_and_Lie algebras._For example, elements_in_$N_+$ can_be_realized as nonsingular block lower triangular_matrices_with identity blocks in the diagonal; elements_in $B$ can be_realized_as nonsingular block upper_triangular matrices. We shall review and_collect some facts about the structure_of simple_Lie algebra_$\mathfrak{g}$ in our case. Again one may refer to [@GS] and_[@schmid1] for more details. Let $\theta:_\,\mathfrak{g}\rightarrow \mathfrak{g}$ be the_We

for weak (Fig. 3a) and for strong (Fig. 3b) damping, as well as for the different degrees of nonlinearity. This plot shows that the difference $z^{(1)}(\tau)$ first increases and then saturates at some level determined by the degree of nonlinearity and dissipation. We start our analysis with the case of weak dissipation (Fig. 3a). For weak dissipation, the level of saturation of $z^{(1)}(\tau)$ increases with consequent consideration of higher nonlinearities. This result may be qualitatively understood as follows. For weak dissipation, the time interval of almost lossless behavior $\tau\lesssim\tau_1$ (\[tau\_1\]) is rather long and therefore the asymptotic behavior of $z^{(1)}(\tau)$ sufficiently depends on its behavior during the time $0\leq\tau\lesssim\tau_1$. For $\tau\lesssim\tau_1$, we can use the approximation (\[B9\]), i.e. $|z^{(1)}(\tau_1)|\simeq\tau_1^2 l^3=l/\Gamma^2$. This estimate shows that the level of saturation for $z^{(1)}(\tau)$ should increase with the growth of $l$. Turn to the case of strong dissipation (Fig. 3b). For strong dissipation, the level of saturation of $z^{(1)}(\tau)$ is sufficiently lower in comparison with the case of weak dissipation (compare Fig. 3a and Fig. 3b). Moreover, the level of saturation for $l=1$ is slightly higher than for nonlinearities with $l>1$ and the difference of the saturation levels for the different nonlinearities with $l>1$ is practically invisible. To understand such behavior we note that for strong enough dissipation the time interval for almost lossless dynamics $\tau\lesssim\tau_1$ is short and the behavior of $Q(\tau)$ for another time interval $[\tau_1, \infty]$ is more sufficient (Fig. 2b). For the nonlinearity with $l=1$, $Q(\tau)$ has slowly decreasing tail, while it decays rapidly for $l>1$ (Fig. 2b). As a result, the area under the curve $Q(\tau)$ for $l=1$ is greater than the areas under the curves corresponding to different $l$ with $l>1$. Yet, because all $Q(\tau)$ are fastly decreasing functions at

for weak (Fig.   3a) and for strong (Fig.   3b) damping, as well as for the unlike academic degree of nonlinearity. This plot shows that the difference $ z^{(1)}(\tau)$ inaugural increase and then saturates at some level specify by the academic degree of nonlinearity and dissipation. We start our analysis with the case of weak waste (Fig.   3a). For weak dissipation, the degree of saturation of $ z^{(1)}(\tau)$ increases with consequent consideration of higher nonlinearities. This result may be qualitatively understand as follows. For weak dissipation, the prison term interval of almost lossless behavior $ \tau\lesssim\tau_1 $ (\[tau\_1\ ]) is rather retentive and therefore the asymptotic behavior of $ z^{(1)}(\tau)$ sufficiently depends on its behavior during the time $ 0\leq\tau\lesssim\tau_1$. For $ \tau\lesssim\tau_1 $, we can practice the approximation (\[B9\ ]), i.e. $ |z^{(1)}(\tau_1)|\simeq\tau_1 ^ 2 l^3 = l/\Gamma^2$. This estimate shows that the level of saturation for $ z^{(1)}(\tau)$ should increase with the growth of $ l$. call on to the case of strong dissipation (Fig.   3b). For strong dissipation, the level of saturation of $ z^{(1)}(\tau)$ is sufficiently lower in comparison with the case of weak profligacy (compare Fig.   3a and Fig.   3b). Moreover, the level of impregnation for $ l=1 $ is slenderly higher than for nonlinearities with $ l>1 $ and the difference of the saturation levels for the unlike nonlinearities with $ l>1 $ is practically inconspicuous. To understand such behavior we note that for firm enough dissipation the time interval for almost lossless moral force $ \tau\lesssim\tau_1 $ is short and the behavior of $ Q(\tau)$ for another time interval $ [ \tau_1, \infty]$ is more sufficient (Fig.   2b). For the nonlinearity with $ l=1 $, $ Q(\tau)$ has slowly decreasing tail, while it disintegrate rapidly for $ l>1 $ (Fig.   2b). As a result, the area under the curve $ Q(\tau)$ for $ l=1 $ is greater than the area under the curves corresponding to different $ l$ with $ l>1$. Yet, because all $ Q(\tau)$ are fastly decreasing functions at

fog weak (Fig. 3a) and for strokg (Fig. 3b) damping, cw well as fod the diwferent degrees of nonlinearmty. Rhis kjot shows that the differenbe $z^{(1)}(\tau)$ furst uncreases ehd then saturzbes ac wome level detgrmined by tve degree of nmnuiuearity and dissipation. We start our wnalysix aith the case jf wtak disappction (Fig. 3a). For weak dissipation, the lenel of saturation of $z^{(1)}(\tau)$ increases with cojseqkent consideration of higher bonlybearities. Thks result may be qualijatively understood as follows. Fur weck dissipatuob, tjg time intertal of almost lossless behdvior $\tsu\lesssim\tau_1$ (\[tsu\_1\]) ms rqther long and therefmre the asymptotic behavior oy $z^{(1)}(\tau)$ sufficiently dwpwnds mn ids bdyavkor dnrihg the tije $0\leq\tau\lssssim\tau_1$. Fir $\tau\lesssim\tau_1$, we cwb use the appdoximaeijn (\[B9\]), i.e. $|z^{(1)}(\tau_1)|\simeq\tau_1^2 l^3=l/\Gamma^2$. This estimaue shkws that the level of sqturation for $z^{(1)}(\tau)$ shluld incrqase with the growth of $l$. Turn to the case of strotg diaripcblon (Wug. 3h). For strong dissipation, the level of saturatykn on $z^{(1)}(\tau)$ is sufficlently lower in cokpwrofon with the zase oy wsak dissipation (colpare Fyg. 3a abd Fig. 3b). Mjreofer, the level of saturation for $l=1$ is slpghtoy higher than for nonlinearicies woth $l>1$ and the difference of che safuration legels for ffe different nonuinvaridies with $l>1$ is practically invisiblx. To bnderstavd soch behwvior we nlte tmdt for strong enouhh didshpation thf time interval for almost lossless dynamics $\jau\nesvsim\tau_1$ ns shovt and the behadior of $Q(\tau)$ fpr anotker tioe intervam $[\tau_1, \iifty]$ is more sufficient (Fhh. 2b). For the nmnlinearyty qith $l=1$, $Q(\tau)$ fas slowly decteasing tcnl, while ut decays rapidly nor $l>1$ (Rig. 2b). As a resulc, uhw area under tne zurde $Q(\vau)$ fjs $l=1$ is greatar tfan yhe afeas under bhe curfes corresponding to difrerent $l$ with $l>1$. Yey, necause aol $Q(\tau)$ wre fastly devreasing functions at

for weak (Fig. 3a) and for strong damping, well as the different degrees that difference $z^{(1)}(\tau)$ first and then saturates some level determined by the degree nonlinearity and dissipation. We start our analysis with the case of weak dissipation 3a). For weak dissipation, the level of saturation of $z^{(1)}(\tau)$ increases with consequent of nonlinearities. result be qualitatively understood as follows. For weak dissipation, the time interval of almost lossless behavior $\tau\lesssim\tau_1$ is rather long and therefore the asymptotic behavior $z^{(1)}(\tau)$ sufficiently depends on behavior during the time $0\leq\tau\lesssim\tau_1$. $\tau\lesssim\tau_1$, can use approximation i.e. l^3=l/\Gamma^2$. This estimate that the level of saturation for $z^{(1)}(\tau)$ should increase with the growth of $l$. Turn to the of strong 3b). For dissipation, level saturation of $z^{(1)}(\tau)$ lower in comparison with the case (compare Fig. 3a and Fig. 3b). Moreover, the of saturation $l=1$ is slightly higher than for with $l>1$ and the difference of the saturation for the different nonlinearities with $l>1$ is practically invisible. To understand such behavior we note strong enough dissipation the interval for almost dynamics is and behavior of for another time interval $[\tau_1, \infty]$ is more sufficient (Fig. 2b). the nonlinearity with $l=1$, $Q(\tau)$ has slowly decreasing tail, while rapidly $l>1$ (Fig. 2b). a result, the area the $Q(\tau)$ for $l=1$ is the under to $l$ $l>1$. Yet, because all are fastly decreasing functions at

for weak (Fig. 3a) and for strong (FiG. 3b) damping, aS well As fOr tHe DiffErenT degrees of nonlINearIty. This plot shows that thE diffErENce $z^{(1)}(\TAu)$ First IncreasES aND TheN sAtUraTeS At Some lEveL determIned by the dEgrEe Of nonlineariTY aNd dissipatIon. we start our anAlySis witH tHe cASe of wEak DissiPation (fIg. 3a). For Weak dissiPaTIon, the LEvel of sATUrAtioN of $z^{(1)}(\tau)$ increases wITh COnsequent consiDeratiOn OF hIGHer NonLinearitieS. THis reSUlt may bE QuALITatIVely understooD as follows. FOR weAk dissIpAtiON, the tiMe intErVAl oF almost lossLess Behavior $\tAu\lessSIm\tau_1$ (\[taU\_1\]) Is ratheR long aNd tHerEforE ThE aSymPtOTic BEhAviOR of $Z^{(1)}(\tau)$ suffIcIeNtly dEpenDS ON Its bEhaVior DurinG the time $0\leq\taU\leSssiM\Tau_1$. for $\taU\lessSim\tAu_1$, We can Use the ApproXiMation (\[B9\]), i.e. $|z^{(1)}(\tau_1)|\siMeq\tAu_1^2 l^3=l/\Gamma^2$. thiS eStiMaTe shoWS that tHe lEveL of satuRation fOR $z^{(1)}(\tAu)$ SHOUlD increase with the grOwTH Of $L$. Turn to tHe case OF sTrONg dissipAtIon (fig. 3b). fOR stroNg diSSiPation, thE level OF sAtUration Of $Z^{(1)}(\tau)$ is SuFfiCieNtly lOWer iN compaRison witH the cASe of weak dissipATion (compare FiG. 3A aND fiG. 3B). MorEovEr, the level oF satURatiOn foR $L=1$ iS slIGhtly HigheR tHAn FOr nonlinearities witH $l>1$ And the DiffeRence of the satUration levELS For the diFferENt NOnlinearities wIth $l>1$ iS practicalLY invisibLe. To uNderstanD such behaVIOr we note ThaT foR stRonG ENoUgh dissipatioN THe tiMe IntervaL foR almost LosSleSs dYnaMiCs $\tau\lessSim\tau_1$ is ShOrT aNd The BehavIOr of $Q(\tau)$ FoR anOtHer Time iNTerval $[\Tau_1, \inFty]$ iS mOrE SufFicient (fIg. 2B). fOr thE nOnLineAriTy With $l=1$, $q(\tau)$ HAs sLowly deCreasing tAil, WHile It DeCays rapIdly for $l>1$ (Fig. 2b). AS a Result, the aReA unDer the CURve $Q(\tau)$ fOr $l=1$ is greater than the areaS Under thE cuRves cOrreSponding tO diFferenT $l$ wITh $l>1$. Yet, BecausE all $Q(\TaU)$ arE FAstly DECrEasInG functions AT

for weak (Fig. 3a) and fo r strong ( Fig.3b) da mp ing, aswell as for th e dif ferent degrees of nonl inear it y . Th i splotshows t h at t hedi ff ere nc e $ z^{(1 )}( \tau)$first incr eas es and then sa t ur ates at so melevel determ ine d by t he de g ree o f n onlin earity and di ssipation .We sta r t our a n a ly siswith the case ofw ea k dissipation ( Fig. 3 a) . F o r we akdissipatio n, thel evel of sa t u r ati o n of $z^{(1)} (\tau)$ inc r eas es wit hcon s equent cons id e rat ion of high er n onlineari ties.T his res u lt maybe qua lit ati vely un de rst oo d as fo llo w s.For weak d is sipat ion, t h e tim e i nter val o f almost loss les s be h avi or $\ tau\l esss im \tau_ 1$ (\[ tau\_ 1\ ]) is rather lo ng a nd theref ore t heas ympto t ic beh avi orof $z^{ (1)}(\t a u)$ s u f f ic iently depends onit s be havior d uringt he t i me $0\le q\ tau \les s s im\ta u_1$ . F or $\tau \lesss i m\ ta u_1$, w ecan us ethe ap proxi m atio n (\[B 9\]), i. e. $| z ^{(1)}(\tau_1) | \simeq\tau_1^ 2 l ^ 3 =l / \Gam ma^ 2$. This es tima t e sh owst ha t t h e lev el of s a tu r ation for $z^{(1)}( \t au)$ s hould increase wit h the grow t h of $l$. Tur n t o the case of s trong dissipati o n (Fig.3b).For stro ng dissip a t ion, the le vel of sa t u ra tion of $z^{( 1 ) }(\t au )$ is s uff icientl y l owe r i n c om parison w ith theca se o fwea k dis s ipation(c omp ar e F ig. 3 a and F ig. 3 b).Mo re o ver , the l e ve l of s at ur atio n f or $l=1 $ is sli ghtly h igher tha n f o r no nl in earitie s with $l>1$an d the diff er enc e of t h e saturat ion levels for the diff e rent no nli neari ties with $l> 1$is pra cti c ally i nvisib le. T ound e r stand s uc h b eh avior we n o t e t hat f or str ong eno ugh dissipation th e ti me interval f oralmo s t l oss l es s dy na m ics $ \tau\lesssim\ta u_1$ is sh or t a nd the beh a vio rof $Q(\ tau)$ f or an o ther ti me interv al $[\tau _1 , \i n f ty] $ is moresufficie nt (Fig.2 b). F o rthe n onl ineari ty wi th $l =1$, $ Q (\t au)$has sl ow ly dec reasi ng tail, w hile it decays rapidlyfor $l >1$ ( Fig . 2b). As ar esu lt, the a reaunder thecur ve$Q(\t au) $ for$l=1 $ i s g r eater tha n the area s u nde r th e curves co r r e spo nding to differ ent$l$ with $l>1$. Y e t, because all $Q( \ t au) $ a r e fa st ly decreasingfun ct i o ns at

for_weak (Fig. 3a)_and for strong (Fig. 3b)_damping, as_well_as for_the_different degrees of_nonlinearity. This plot_shows that the difference_$z^{(1)}(\tau)$ first increases_and_then saturates at some level determined by the degree of nonlinearity and dissipation. We start_our_analysis with_the_case_of weak dissipation (Fig. 3a). For_weak dissipation, the level of_saturation of_$z^{(1)}(\tau)$ increases with consequent consideration of higher nonlinearities._This_result may be_qualitatively understood as follows. For weak dissipation, the time_interval of almost lossless behavior $\tau\lesssim\tau_1$_(\[tau\_1\]) is rather_long_and_therefore the asymptotic behavior_of $z^{(1)}(\tau)$ sufficiently depends on its_behavior during the time $0\leq\tau\lesssim\tau_1$. For_$\tau\lesssim\tau_1$, we can use the approximation (\[B9\]),_i.e. $|z^{(1)}(\tau_1)|\simeq\tau_1^2 l^3=l/\Gamma^2$. This estimate shows that_the level of saturation for_$z^{(1)}(\tau)$ should_increase with the growth of_$l$. Turn to the_case of_strong dissipation (Fig. 3b)._For strong dissipation, the level of_saturation of $z^{(1)}(\tau)$_is sufficiently lower in comparison with_the_case of weak_dissipation_(compare_Fig. 3a and_Fig. 3b). Moreover, the_level_of saturation_for_$l=1$ is slightly higher than for_nonlinearities_with $l>1$ and the difference of the_saturation levels for the_different_nonlinearities with $l>1$ is_practically invisible. To understand such_behavior we note that for strong_enough dissipation_the time_interval for almost lossless dynamics $\tau\lesssim\tau_1$ is short and the behavior_of $Q(\tau)$ for another time interval_$[\tau_1, \infty]$ is more_sufficient (Fig. 2b)._For_the nonlinearity with_$l=1$,_$Q(\tau)$ has_slowly decreasing tail, while it decays rapidly_for $l>1$_(Fig. 2b). As a result, the area_under the curve $Q(\tau)$_for_$l=1$ is greater than the areas_under the curves corresponding to different_$l$ with $l>1$. Yet, because_all_$Q(\tau)$_are fastly decreasing functions at

variables and the number of constraints. We must try to keep each of them low. For such purposes, we employ some techniques used in [@BRS90; @BS86; @N07; @NMF07]. Let us review these techniques briefly. ### Fixing a basis {#BasisFix} A technique explained in the following was introduced in [@BS86] and was used to reduce the degrees of constraints and the number of variables in [@BS86; @N07; @NMF07]. We assume that $\chi (i_1,...,i_r)=+$ for some $i_1,...,i_r$ by taking the negative of $\chi$ if necessary. Let $M_V:=(v_1,...,v_n) \in \mathbb{R}^{d \times n}$ be the representation matrix of a vector configuration $V$. Because the combinatorial type of $V$ is invariant under any invertible linear transformations, we can assume that $v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}$ for an $r$-tuple $(b_1,...,b_r) \in \Lambda (n,r)$ such that $\chi (b_1,...,b_r) = +$. We call such an $r$-tuple of indices a [*basis*]{}. We obtain a new polynomial system as follows. $$\label{BasisFix:eq1} \begin{cases} {\rm sign}(\det(v_{i_{1}},...,v_{i_{r}})) = \chi (i_{1},...,i_{r}) \text{ for all $(i_{1},...,i_{r}) \in \Lambda(n,r),$ and} \\ v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}. \end{cases}$$ The resulting polynomial system (\[BasisFix:eq1\]) depends on the choice of bases. In the next section, we present how to find a suitable basis. Finally, note that the degree of each constraint can be computed easily by the following formula: $$deg(\det(v_{i_{

variables and the number of constraints. We must try to observe each of them depleted. For such purposes, we employ some techniques use in [ @BRS90; @BS86; @N07; @NMF07 ]. Let us review these techniques concisely. # # # Fixing a basis { # BasisFix } A technique explain in the following was introduced in   [ @BS86 ] and was use to reduce the degrees of restraint and the numeral of variables in   [ @BS86; @N07; @NMF07 ]. We assume that $ \chi (i_1,... ,i_r)=+$ for some $ i_1,... ,i_r$ by taking the negative of $ \chi$ if necessary. Let $ M_V:=(v_1,... ,v_n) \in \mathbb{R}^{d \times n}$ be the theatrical performance matrix of a vector configuration $ V$. Because the combinatorial type of $ V$ is changeless under any invertible linear transformations, we can assume that $ v_{b_{1}}=(1,0,... ,0)^{T},v_{b_{2}}=(0,1,0,... ,0)^{T},... ,v_{b_{r}}=(0,... ,0,1)^{T}$ for an $ r$-tuple $ (b_1,... ,b_r) \in \Lambda (n, r)$ such that $ \chi (b_1,... ,b_r) = + $. We call such an $ r$-tuple of indices a [ * footing * ] { }. We obtain a modern polynomial system as follows. $ $ \label{BasisFix: eq1 } \begin{cases } { \rm sign}(\det(v_{i_{1}},... ,v_{i_{r } }) ) = \chi (i_{1},... ,i_{r }) \text { for all $ (i_{1},... ,i_{r }) \in \Lambda(n, r),$ and } \\ v_{b_{1}}=(1,0,... ,0)^{T},v_{b_{2}}=(0,1,0,... ,0)^{T},... ,v_{b_{r}}=(0,... ,0,1)^{T }. \end{cases}$$ The resulting polynomial arrangement (\[BasisFix: eq1\ ]) depends on the choice of bases. In the next section, we present how to find a suitable basis. Finally, note that the degree of each constraint can be computed easily by the following formula: $ $ deg(\det(v_{i _ {

vagiables and the number on constraints. We must tcy to ksep each of them low. For such purposxs, ww empooy some techniques usdd in [@BRS90; @BS86; @N07; @NMD07]. Leu us review these techniqmzs brjcfly. ### Yieing a basis {#BaxisFix} A tewhnique explaitea nn the following was introduced in [@BS86] and wax ksed to reduce the qegrsvs of constraints and the numbed of vagiables in [@BS86; @N07; @NMG07]. We assume that $\chi (i_1,...,i_r)=+$ fog sole $i_1,...,i_r$ by taking tje negative of $\syi$ if necessxry. Let $M_V:=(n_1,...,r_n) \in \mathbg{R}^{d \times n}$ be the representatiun macrix of a vgeror wonfiguratiin $V$. Fecause the gpmbinadorial yype of $V$ is ikvarient ynder any invertible ninear transformatyons, we cdn assume that $v_{b_{1}}=(1,0,...,0)^{T},v_{v_{2}}=(0,1,0,...,0)^{T},...,c_{b_{r}}=(0,...,0,1)^{T}$ xor dn $r$-gypld $(b_1,...,g_r) \ih \Lambfa (i,r)$ such thaf $\chi (b_1,...,b_r) = +$. We call such an $r$-tipjv of indices z [*basif*]{}. Re obtain a new polynomial system as fonloss. $$\label{BasisFix:eq1} \begin{xases} {\rm sign}(\det(v_{i_{1}},...,v_{i_{r}})) = \chi (i_{1},...,i_{r}) \eext{ for all $(i_{1},...,i_{r}) \in \Lambda(n,r),$ and} \\ v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}. \end{casas}$$ Thx fesboting popynomial system (\[BasisFix:eq1\]) depends on the choyde on bases. In the ncxt section, we prexejt row to find a suitayme basis. Finally, note that tre detree of ewch vonstraint can be computed wasily by thv foolowing formula: $$dey(\det(v_{i_{

variables and the number of constraints. We to each of low. For such used [@BRS90; @BS86; @N07; Let us review techniques briefly. ### Fixing a basis A technique explained in the following was introduced in [@BS86] and was used reduce the degrees of constraints and the number of variables in [@BS86; @N07; We that (i_1,...,i_r)=+$ some $i_1,...,i_r$ by taking the negative of $\chi$ if necessary. Let $M_V:=(v_1,...,v_n) \in \mathbb{R}^{d \times n}$ the representation matrix of a vector configuration $V$. the combinatorial type of is invariant under any invertible transformations, can assume $v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}$ an $(b_1,...,b_r) \in \Lambda such that $\chi (b_1,...,b_r) = +$. We call such an $r$-tuple of indices a [*basis*]{}. We obtain new polynomial follows. $$\label{BasisFix:eq1} {\rm = (i_{1},...,i_{r}) \text{ for \in \Lambda(n,r),$ and} \\ v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}. \end{cases}$$ system (\[BasisFix:eq1\]) depends on the choice of bases. the next we present how to find a basis. Finally, note that the degree of each can be computed easily by the following formula: $$deg(\det(v_{i_{

variables and the number of coNstraints. WE must Try To kEeP eacH of tHem low. For such pURposEs, we employ some techniquEs useD iN [@bRS90; @Bs86; @n07; @NmF07]. Let Us revieW ThESE teChNiQueS bRIeFly. ### FiXinG a basis {#basisFix} A tEchNiQue explained IN tHe followinG waS introduced iN [@BS86] And was UsEd tO ReducE thE degrEes of cONstraiNts and the NuMBer of vARiables IN [@bS86; @n07; @NMF07]. we assume that $\chi (i_1,...,i_R)=+$ FoR Some $i_1,...,i_r$ by takinG the neGaTIvE OF $\chI$ if Necessary. LEt $m_V:=(v_1,...,v_n) \IN \mathbb{r}^{D \tIMES n}$ bE The representaTion matrix oF A veCtor coNfIguRAtion $V$. becauSe THe cOmbinatoriaL typE of $V$ is invAriant UNder any INvertibLe lineAr tRanSforMAtIoNs, wE cAN asSUmE thAT $v_{b_{1}}=(1,0,...,0)^{t},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}$ FoR aN $r$-tupLe $(b_1,...,b_R) \IN \lAmbdA (n,r)$ Such That $\cHi (b_1,...,b_r) = +$. We call suCh aN $r$-tuPLe oF indiCes a [*bAsis*]{}. we ObtaiN a new pOlynoMiAl system as folloWs. $$\laBel{BasisFIx:eQ1} \bEgiN{cAses} {\rM Sign}(\deT(v_{i_{1}},...,V_{i_{r}})) = \Chi (i_{1},...,i_{r}) \tExt{ for aLL $(i_{1},...,i_{R}) \iN \lAMbDa(n,r),$ and} \\ v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}. \eNd{CASeS}$$ The resuLting pOLyNoMIal systeM (\[BAsiSFix:EQ1\]) DepenDs on THe Choice of Bases. IN ThE nExt sectIoN, we preSeNt hOw tO find A SuitAble baSis. FinalLy, notE That the degree oF Each constrainT CaN BE cOMputEd eAsily by the fOlloWIng fOrmuLA: $$dEg(\dET(v_{i_{

variables and the numberof constra ints. We mu st try tokeep each of t h em l ow. For such purposes, we e mp l oy s o me tech niquesu se d in[@ BR S90 ;@ BS 86; @ N07 ; @NMF0 7]. Let us re vi ew these tec h ni ques brief ly. ### Fixing abasis{# Bas i sFix} A tech niquee xplain ed in the f o llowin g was in t r od uced in [@BS86] and w a su sed to reducethe de gr e es o f c ons traints an dthe n u mber of va r i a ble s in [@BS86; @ N07; @NMF07 ] . We ass um e t h at $\c hi (i _1 , ... ,i_r)=+$ fo r so me $i_1,. ..,i_r $ by tak i ng thenegati veof$\ch i $if ne ce s sar y .Let $M_ V:=(v_1, .. ., v_n)\in\ m a t hbb{ R}^ {d \ times n}$ be the r epr esen t ati on ma trixof a v ector confi gurat io n $V$. Becausethecombinato ria ltyp eof $V $ is in var ian t under any in v ert ib l e li near transformatio ns , we can ass ume th a t$v _ {b_{1}}= (1 ,0, ..., 0 ) ^{T}, v_{b _ {2 }}=(0,1, 0,..., 0 )^ {T },...,v _{ b_{r}} =( 0,. .., 0,1)^ { T}$for an $r$-tup le $( b _1,...,b_r) \i n \Lambda (n,r ) $s u ch that $\ chi (b_1,.. .,b_ r ) =+$.W ecal l such an $ r$ - tu p le of indices a [*b as is*]{} . Weobtain a newpolynomial s y stem asfoll o ws . $$\label{Basi sFix: eq1} \begi n {cases}{\rmsign}(\d et(v_{i_{ 1 } },...,v_ {i_ {r} })) =\ c hi (i_{1},...,i _ { r})\t ext{ fo r a ll $(i_ {1} ,.. .,i _{r }) \in \Lam bda(n,r) ,$ a nd }\\v_{b_ { 1}}=(1,0 ,. .., 0) ^{T },v_{ b _{2}}= (0,1, 0,.. ., 0) ^ {T} ,...,v_ { b_ { r }}=( 0, .. .,0, 1)^ {T }. \e nd{c a ses }$$ The resultin g p o lyno mi al system (\[BasisFix: eq 1\]) depen ds on the c h o ice of b ases. In the next secti o n, we p res ent h ow t o find asui tablebas i s. Fi nally, note t hat t he de g r ee of e ach constr a i ntcan b ecomp uted ea sily by the follow i ngformula: $$de g(\ det( v _ {i _{

variables_and the_number of constraints. We_must try_to_keep each_of_them low. For_such purposes, we_employ some techniques used_in [@BRS90; @BS86;_@N07;_@NMF07]. Let us review these techniques briefly. ### Fixing a basis {#BasisFix} A technique explained in_the_following was_introduced_in [@BS86]_and was used to reduce_the degrees of constraints and_the number_of variables in [@BS86; @N07; @NMF07]. We assume that $\chi_(i_1,...,i_r)=+$_for some $i_1,...,i_r$_by taking the negative of $\chi$ if necessary. Let_$M_V:=(v_1,...,v_n) \in \mathbb{R}^{d \times n}$ be_the representation matrix_of_a_vector configuration $V$. Because_the combinatorial type of $V$ is_invariant under any invertible linear transformations,_we can assume that $v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}$ for an_$r$-tuple $(b_1,...,b_r) \in \Lambda (n,r)$ such_that $\chi (b_1,...,b_r) = +$._We call_such an $r$-tuple of indices_a [*basis*]{}. We_obtain a_new polynomial system_as follows. $$\label{BasisFix:eq1} \begin{cases} {\rm sign}(\det(v_{i_{1}},...,v_{i_{r}})) = \chi_(i_{1},...,i_{r}) \text{ for_all $(i_{1},...,i_{r}) \in \Lambda(n,r),$ and} \\ v_{b_{1}}=(1,0,...,0)^{T},v_{b_{2}}=(0,1,0,...,0)^{T},...,v_{b_{r}}=(0,...,0,1)^{T}. \end{cases}$$_The_resulting polynomial system_(\[BasisFix:eq1\])_depends_on the_choice of bases._In_the next_section,_we present how to find a_suitable_basis. Finally, note that the degree of each_constraint can be computed_easily_by the following formula:_$$deg(\det(v_{i_{

because when the RW path from $st_{\text{s}}$ to $st$ improves later, $\text{c}_\text{c}(st_{\text{s}},st)$ decreases. Hence, $st$ may be expanded accordingly. However, if PPNP now expands $st$, the subsequently visited loco-states always satisfy ULSL and create no feasible solution. Thus, when a loco-state satisfies ULSL, its examination is *postponed* (i.e., not revisited) until all other loco-states in $\mathbf{Q}$ are examined. Similarly, a loco-state $st$ is shelved when $\text{l}_\text{l}(st_{\text{s}},st) + \text{l}_{\min}(\gamma^{\text{v}}_{st}, \gamma^{\text{v}}_{\text{t}}) > \tilde{L}$. If all remaining loco-states are postponed, the reference path cannot be improved, and those states are removed accordingly. If the current loco-state $st$ passes the above pruning criteria, for each unvisited $st' \in \text{N}(st)$, PPNP assigns $\text{l}_\text{l}(st_{\text{s}},st') = \text{l}_\text{l}(st_{\text{s}},st) + \text{l}(st, st')$, $\text{c}_\text{l}(st_{\text{s}},st') = \text{c}_\text{l}(st_{\text{s}},st) + \text{MIL}(st, st')$, $\text{l}_\text{c}(st_{\text{s}},st') = \text{l}_\text{c}(st_{\text{s}},st) + \text{l}(st, st')$, $\text{c}_\text{c}(st_{\text{s}},st') = \text{c}_\text{c}(st_{\text{s}},st) + \text{MIL}(st, st')$, and also $\text{pred}_\text{l}(st') = \text{pred}_\text{c}(st') = st$. On the other hand, for each visited $st'$, PPNP updates $\text{l}_\text{l}(st_{\text{s}},st'), \text{c}_\text{l}(st_{\text{s}},st')$,

because when the RW path from $ st_{\text{s}}$ to $ st$ improves later, $ \text{c}_\text{c}(st_{\text{s}},st)$ decreases. therefore, $ st$ may be expand accordingly. However, if PPNP now expand $ st$, the subsequently travel to loco - states always meet ULSL and create no feasible solution. Thus, when a loco - state of matter satisfies ULSL, its examination is * postponed * (i.e., not revisit) until all other loco - states in $ \mathbf{Q}$ are examined. Similarly, a loco - state $ st$ is postpone when $ \text{l}_\text{l}(st_{\text{s}},st) + \text{l}_{\min}(\gamma^{\text{v}}_{st }, \gamma^{\text{v}}_{\text{t } }) > \tilde{L}$. If all remaining loco - states are postponed, the reference path cannot be better, and those states are removed accordingly. If the current loco - department of state $ st$ passes the above pruning criteria, for each unvisited $ st' \in \text{N}(st)$, PPNP assigns $ \text{l}_\text{l}(st_{\text{s}},st') = \text{l}_\text{l}(st_{\text{s}},st) + \text{l}(st, st')$, $ \text{c}_\text{l}(st_{\text{s}},st') = \text{c}_\text{l}(st_{\text{s}},st) + \text{MIL}(st, st')$, $ \text{l}_\text{c}(st_{\text{s}},st') = \text{l}_\text{c}(st_{\text{s}},st) + \text{l}(st, st')$, $ \text{c}_\text{c}(st_{\text{s}},st') = \text{c}_\text{c}(st_{\text{s}},st) + \text{MIL}(st, st')$, and besides $ \text{pred}_\text{l}(st') = \text{pred}_\text{c}(st') = st$. On the other handwriting, for each visited $ st'$, PPNP updates $ \text{l}_\text{l}(st_{\text{s}},st'), \text{c}_\text{l}(st_{\text{s}},st')$,

befause when the RW path fvom $st_{\text{s}}$ to $sj$ umprovxs lated, $\text{c}_\tdxt{c}(st_{\text{s}},st)$ decreases. Hencx, $st$ may ve expanded accordinglh. However, if PPNP now wxpands $st$, the subsequenfpy vnsmted loco-states always sadisfy ULSL and cfecte no feasible solution. Thus, when a loco-ststf satisfies ULFL, ius qxamjnation is *postponed* (i.e., not revisifed) untpl all other loco-xtates in $\mathbf{Q}$ are examlned. Similarly, a loco-shate $st$ is wheldwd when $\text{u}_\text{l}(st_{\teqc{s}},st) + \text{l}_{\jin}(\gamma^{\text{v}}_{st}, \gamma^{\text{v}}_{\text{t}}) > \tilbe{L}$. If all tznaijhng loco-staves arv postponed, tmv referance payh cannot be ikprmvee, and those states arx removed accordingli. If the cusrznt loco-state $st$ passws the dbova prjbine cdiverja, for earh unvisites $st' \in \texr{N}(st)$, PPNP assigns $\ttxt{j}_\nrxt{l}(st_{\text{s}},sf') = \texe{l}_\eext{l}(st_{\text{s}},st) + \text{l}(st, st')$, $\text{c}_\text{l}(st_{\uext{s}},at') = \text{c}_\text{l}(st_{\text{s}},st) + \text{MIL}(st, st')$, $\text{l}_\tgxt{c}(st_{\text{f}},st') = \text{l}_\text{c}(st_{\text{s}},st) + \text{l}(st, st')$, $\text{c}_\text{c}(st_{\dext{s}},ag') = \uewb{c}_\tebr{c}(dt_{\text{s}},st) + \text{MIL}(st, st')$, and also $\text{pred}_\text{j}(at') = \next{pred}_\text{c}(st') = ft$. On the oyhfr rand, for each visitzs $at'$, PPNP updates $\tedt{l}_\text{j}(st_{\tezt{s}},st'), \texu{c}_\texy{l}(st_{\text{s}},st')$,

because when the RW path from $st_{\text{s}}$ improves $\text{c}_\text{c}(st_{\text{s}},st)$ decreases. $st$ may be now $st$, the subsequently loco-states always satisfy and create no feasible solution. Thus, a loco-state satisfies ULSL, its examination is *postponed* (i.e., not revisited) until all loco-states in $\mathbf{Q}$ are examined. Similarly, a loco-state $st$ is shelved when $\text{l}_\text{l}(st_{\text{s}},st) \text{l}_{\min}(\gamma^{\text{v}}_{st}, > If remaining loco-states are postponed, the reference path cannot be improved, and those states are removed accordingly. the current loco-state $st$ passes the above pruning for each unvisited $st' \text{N}(st)$, PPNP assigns $\text{l}_\text{l}(st_{\text{s}},st') = + st')$, $\text{c}_\text{l}(st_{\text{s}},st') \text{c}_\text{l}(st_{\text{s}},st) \text{MIL}(st, $\text{l}_\text{c}(st_{\text{s}},st') = \text{l}_\text{c}(st_{\text{s}},st) \text{l}(st, st')$, $\text{c}_\text{c}(st_{\text{s}},st') = \text{c}_\text{c}(st_{\text{s}},st) + \text{MIL}(st, st')$, and also $\text{pred}_\text{l}(st') = \text{pred}_\text{c}(st') = st$. On the hand, for $st'$, PPNP $\text{l}_\text{l}(st_{\text{s}},st'),

because when the RW path from $sT_{\text{s}}$ to $st$ ImproVes LatEr, $\Text{C}_\texT{c}(st_{\text{s}},st)$ decREaseS. Hence, $st$ may be expanded aCcordInGLy. HoWEvEr, if PpNP now eXPaNDS $st$, ThE sUbsEqUEnTly viSitEd loco-sTates alwayS saTiSfy ULSL and crEAtE no feasiblE soLution. Thus, whEn a Loco-stAtE saTIsfieS ULsL, its ExaminATion is *Postponed* (I.e., NOt reviSIted) untIL AlL othEr loco-states in $\matHBf{q}$ Are examined. SimIlarly, A lOCo-STAte $St$ iS shelved whEn $\Text{l}_\TExt{l}(st_{\tEXt{S}},ST) + \TexT{L}_{\min}(\gamma^{\text{V}}_{st}, \gamma^{\texT{V}}_{\teXt{t}}) > \tilDe{l}$. If ALl remaIning LoCO-stAtes are postPoneD, the referEnce paTH cannot BE improvEd, and tHosE stAtes ARe ReMovEd ACcoRDiNglY. if tHe currenT lOcO-statE $st$ pASSES the AboVe prUning Criteria, for eaCh uNvisITed $St' \in \tExt{N}(sT)$, PPNp aSsignS $\text{l}_\Text{l}(St_{\Text{s}},st') = \text{l}_\texT{l}(st_{\Text{s}},st) + \teXt{l}(St, St')$, $\tExT{c}_\texT{L}(st_{\texT{s}},sT') = \teXt{c}_\text{L}(st_{\text{S}},St) + \tExT{mil}(sT, st')$, $\text{l}_\text{c}(st_{\texT{s}},ST') = \TeXt{l}_\text{c}(St_{\text{S}},St) + \TeXT{l}(st, st')$, $\teXt{C}_\teXt{c}(sT_{\TExt{s}},sT') = \texT{C}_\tExt{c}(st_{\teXt{s}},st) + \tEXt{mIl}(st, st')$, anD aLso $\texT{pRed}_\TexT{l}(st') = \tEXt{prEd}_\text{C}(st') = st$. On tHe othER hand, for each viSIted $st'$, PPNP updATeS $\TExT{L}_\texT{l}(sT_{\text{s}},st'), \texT{c}_\teXT{l}(st_{\Text{S}},St')$,

because when the RW pathfrom $st_{ \text {s} }$to $st $ im proves later,$ \tex t{c}_\text{c}(st_{\tex t{s}} ,s t )$ d e cr eases . Hence , $ s t $ m ay b e e xp a nd ed ac cor dingly. However,ifPP NP now expan d s$st$, thesub sequently vi sit ed loc o- sta t es al way s sat isfy U L SL and create n of easibl e soluti o n .Thus , when a loco-sta t es atisfies ULSL, its e xa m in a t ion is *postpone d* (i.e . , not r e vi s i t ed) until all oth er loco-sta t esin $\m at hbf { Q}$ ar e exa mi n ed. Similarly, a l oco-state $st$i s shelv e d when$\text {l} _\t ext{ l }( st _{\ te x t{s } }, st) + \ text{l}_ {\ mi n}(\g amma ^ { \ t ext{ v}} _{st }, \g amma^{\text{v }}_ {\te x t{t }}) > \til de{L }$ . Ifall re maini ng loco-states ar e po stponed,the r efe re nce p a th can not be improv ed, and tho se s t at es are removed acc or d i ng ly. Ifthe cu r re nt loco-sta te $s t$ p a s ses t he a b ov e prunin g crit e ri a, for ea ch unvis it ed$st ' \in \tex t{N}(s t)$, PPN P ass i gns $\text{l}_ \ text{l}(st_{\ t ex t { s} } ,st' ) = \text{l}_\ text { l}(s t_{\ t ex t{s } },st) + \t ex t {l } (st, st')$, $\text{ c} _\text {l}(s t_{\text{s}}, st') = \te x t { c}_\text {l}( s t_ { \text{s}},st)+ \te xt{MIL}(st , st')$,$\tex t{l}_\te xt{c}(st_ { \ text{s}} ,st ')= \ tex t { l} _\text{c}(st_ { \ text {s }},st)+ \ text{l} (st , s t') $,$\ text{c}_\ text{c}( st _{ \t ex t{s }},st ' ) = \tex t{ c}_ \t ext {c}(s t _{\tex t{s}} ,st) + \ t ext {MIL}(s t ,s t ')$, a nd als o $ \t ext{p red} _ \te xt{l}(s t') = \te xt{ p red} _\ te xt{c}(s t') = st$. On t he other h an d,for ea c h visited $st'$, PPNP updates $\ t ext{l}_ \te xt{l} (st_ {\text{s} },s t'), \ tex t {c}_\t ext{l} (st_{ \t ext { s }},st ' ) $,

because_when the_RW path from $st_{\text{s}}$_to $st$_improves_later, $\text{c}_\text{c}(st_{\text{s}},st)$_decreases._Hence, $st$ may_be expanded accordingly._However, if PPNP now_expands $st$, the_subsequently_visited loco-states always satisfy ULSL and create no feasible solution. Thus, when a loco-state_satisfies_ULSL, its_examination_is_*postponed* (i.e., not revisited) until_all other loco-states in $\mathbf{Q}$_are examined._Similarly, a loco-state $st$ is shelved when $\text{l}_\text{l}(st_{\text{s}},st)_+_\text{l}_{\min}(\gamma^{\text{v}}_{st}, \gamma^{\text{v}}_{\text{t}}) >_\tilde{L}$. If all remaining loco-states are postponed, the reference_path cannot be improved, and those_states are removed_accordingly. If_the_current loco-state $st$ passes_the above pruning criteria, for each_unvisited $st' \in \text{N}(st)$, PPNP assigns_$\text{l}_\text{l}(st_{\text{s}},st') = \text{l}_\text{l}(st_{\text{s}},st) + \text{l}(st, st')$, $\text{c}_\text{l}(st_{\text{s}},st')_= \text{c}_\text{l}(st_{\text{s}},st) + \text{MIL}(st, st')$, $\text{l}_\text{c}(st_{\text{s}},st')_= \text{l}_\text{c}(st_{\text{s}},st) + \text{l}(st, st')$,_$\text{c}_\text{c}(st_{\text{s}},st') =_\text{c}_\text{c}(st_{\text{s}},st) + \text{MIL}(st, st')$, and_also $\text{pred}_\text{l}(st') =_\text{pred}_\text{c}(st') =_st$. On the_other hand, for each visited $st'$,_PPNP updates $\text{l}_\text{l}(st_{\text{s}},st'),_\text{c}_\text{l}(st_{\text{s}},st')$,

0,1,1]_{\vec f},\quad \vec v_4=[n_B,kn_B,1,1]_{\vec f},$$ where $[a_1,\cdots,a_4]_{\vec f}=\sum_ia_i\vec f_i$. We have chosen basis (\[basisf\]) so that the toric data become the standard form in which the last components of the vectors are $1$. We can draw the toric diagram using the first three components of these vectors $\vec v_i$, which coincides with that in Fig. \[c4.eps\](b). Further generalization {#more.sec} ====================== Up to now we have considered a brane system with two kinds of fivebranes. It is also possible to introduce more than two kinds of fivebranes. To represent the types of branes we used $q_I=0$ and $1$. In this section we allow $q_I$ to be an arbitrary integer. In this case, we do not need to introduce the coefficient $k$ in (\[cscouplings\]) and we set $k=1$. This means that the $I$th fivebrane is a $(q_I,1)$5-brane, and the Chern-Simons couplings are given by $$\label{kInonzero} k_I=q_{I+1}-q_I.$$ For simplicity we assume that the Chern-Simons couplings do not vanish. This implies that all the adjoint chiral multiplets $\Phi_I$ and the vector multiplets $V_I$ become massive. It is easy to show that even if some of the $k_I$ vanish we obtain the same moduli space as derived below. By integrating out $\Phi_I$, we obtain the superpotential $$W=-\sum_{I=1}^n\frac{1}{2(q_{I+1}-q_I)}(X_IY_I-Y_{I+1}X_{I+1})^2.$$ (When we obtained the superpotential (\[ellw\]) we used $q_I=0$ and $1$, although we cannot use it here.) From the assumption (\[coulomb\]), the F-term conditions for $X_I$ and $Y_I$ give $$\frac{X_{I+1}Y_{I+1}-X_IY_I}{q_{I+1}-q_

0,1,1]_{\vec f},\quad \vec v_4=[n_B, kn_B,1,1]_{\vec f},$$ where $ [ a_1,\cdots, a_4]_{\vec f}=\sum_ia_i\vec f_i$. We have chosen basis (\[basisf\ ]) so that the toric data become the standard form in which the final component of the vectors are $ 1$. We can draw the toric diagram using the beginning three components of these vectors $ \vec v_i$, which coincide with that in Fig.   \[c4.eps\](b). Further generalization { # more.sec } = = = = = = = = = = = = = = = = = = = = = = Up to now we have consider a brane system with two kinds of fivebranes. It is besides possible to introduce more than two kinds of fivebranes. To represent the type of branes we used $ q_I=0 $ and $ 1$. In this incision we allow $ q_I$ to be an arbitrary integer. In this case, we do not necessitate to introduce the coefficient $ k$ in (\[cscouplings\ ]) and we set $ k=1$. This means that the $ I$th fivebrane is a $ (q_I,1)$5 - brane, and the Chern - Simons coupling are given by $ $ \label{kInonzero } k_I = q_{I+1}-q_I.$$ For simplicity we assume that the Chern - Simons yoke do not vanish. This implies that all the adjoint chiral multiplets $ \Phi_I$ and the vector multiplets $ V_I$ become massive. It is easy to show that even if some of the $ k_I$ vanish we obtain the like moduli space as derived below. By integrating out $ \Phi_I$, we obtain the superpotential $ $ W=-\sum_{I=1}^n\frac{1}{2(q_{I+1}-q_I)}(X_IY_I - Y_{I+1}X_{I+1})^2.$$ (When we prevail the superpotential (\[ellw\ ]) we used $ q_I=0 $ and $ 1 $, although we cannot use it here .) From the assumption (\[coulomb\ ]), the F - term conditions for $ X_I$ and $ Y_I$ give $ $ \frac{X_{I+1}Y_{I+1}-X_IY_I}{q_{I+1}-q _

0,1,1]_{\vec f},\quad \vec v_4=[n_B,kn_B,1,1]_{\vec f},$$ whtre $[a_1,\cdots,a_4]_{\vec f}=\som_ua_i\vec f_i$. We have chusen basis (\[basisf\]) so that thx toeic dqta become the standara form in which tye lest components oh the vegcors zve $1$. Wz ran draw the totic diagram gsing the firsd ghxee components of these vectors $\vec d_i$, whicn foincides with thau ig Fif. \[b4.eks\](b). Further generalization {#more.sec} ====================== Hp to nmw we have comsidered a brane system wihh tao kinds of fivebrwnes. It is qlso possible to kntroduce more than twk kinds of fivebranes. To represdnt tke types of beanfv we used $q_M=0$ and $1$. In this secbpon we dllow $q_O$ to be an arbltrarb inreger. In this case, we do not need to injroduce tha eoefficient $k$ in (\[cscoypoings\]) and we rwt $y=1$. Tgix jeans hhav the $I$th fjvebrane is a $(q_I,1)$5-brane, and the Vhqgm-Simons coupmings wrq given by $$\label{kInonzero} k_I=q_{I+1}-q_I.$$ For simklicify we assume that the Cyern-Simons couplings fo not vagish. This implies that all the adjoint chiral muldiplevs $\Phn_L$ xbd the vector multiplets $V_I$ become massive. It if essj to show that evcn if some of the $l_I$ vsgish we obtaiv the sams moduli space as ferived beloq. By integwatimg out $\Phi_I$, we obtain the syperpotentiaj $$W=-\sum_{I=1}^n\frac{1}{2(q_{I+1}-q_I)}(X_IY_N-Y_{I+1}X_{I+1})^2.$$ (When wz obtaoned yhe superpotential (\[ellw\]) we hsed $q_I=0$ and $1$, althougg we cannot use ig hvre.) Xrom the assumption (\[coulomf\]), the F-tecm couditions for $X_I$ anq $Y_I$ give $$\vrac{X_{L+1}F_{I+1}-X_IY_I}{q_{I+1}-q_

0,1,1]_{\vec f},\quad \vec v_4=[n_B,kn_B,1,1]_{\vec f},$$ where $[a_1,\cdots,a_4]_{\vec We chosen basis so that the form which the last of the vectors $1$. We can draw the toric using the first three components of these vectors $\vec v_i$, which coincides with in Fig. \[c4.eps\](b). Further generalization {#more.sec} ====================== Up to now we have considered brane with kinds fivebranes. It is also possible to introduce more than two kinds of fivebranes. To represent the of branes we used $q_I=0$ and $1$. In section we allow $q_I$ be an arbitrary integer. In case, do not to the $k$ in (\[cscouplings\]) we set $k=1$. This means that the $I$th fivebrane is a $(q_I,1)$5-brane, and the Chern-Simons couplings are by $$\label{kInonzero} simplicity we that Chern-Simons do not vanish. that all the adjoint chiral multiplets vector multiplets $V_I$ become massive. It is easy show that if some of the $k_I$ vanish obtain the same moduli space as derived below. integrating out $\Phi_I$, we obtain the superpotential $$W=-\sum_{I=1}^n\frac{1}{2(q_{I+1}-q_I)}(X_IY_I-Y_{I+1}X_{I+1})^2.$$ (When we obtained the superpotential (\[ellw\]) we and $1$, although we use it here.) the (\[coulomb\]), F-term for $X_I$ $Y_I$ give $$\frac{X_{I+1}Y_{I+1}-X_IY_I}{q_{I+1}-q_

0,1,1]_{\vec f},\quad \vec v_4=[n_B,kn_B,1,1]_{\vec f},$$ wherE $[a_1,\cdots,a_4]_{\veC f}=\sum_Ia_i\Vec F_i$. we haVe chOsen basis (\[basisF\]) So thAt the toric data become thE stanDaRD forM In Which The last COmPONenTs Of The VeCToRs are $1$. we cAn draw tHe toric diaGraM uSing the first THrEe componenTs oF these vectorS $\veC v_i$, whiCh CoiNCides WitH that In Fig. \[c4.EPs\](b). FurTher generAlIZation {#MOre.sec} ====================== UP TO nOw we Have considered a brANe SYstem with two kiNds of fIvEBrANEs. IT is Also possibLe To intROduce moRE tHAN Two KInds of fivebraNes. To represENt tHe typeS oF brANes we uSed $q_I=0$ AnD $1$. in tHis section wE allOw $q_I$ to be aN arbitRAry inteGEr. In thiS case, wE do Not Need TO iNtRodUcE The COeFfiCIenT $k$ in (\[cscoUpLiNgs\]) anD we sET $K=1$. tHis mEanS thaT the $I$Th fivebrane is A $(q_I,1)$5-BranE, And The ChErn-SiMons CoUplinGs are gIven bY $$\lAbel{kInonzero} k_I=Q_{I+1}-q_I.$$ for simpliCitY wE asSuMe thaT The CheRn-SImoNs couplIngs do nOT vaNiSH. tHiS implies that all the AdJOInT chiral mUltiplETs $\phI_i$ and the vEcTor MultIPLets $V_i$ becOMe Massive. IT is easY To ShOw that eVeN if somE oF thE $k_I$ VanisH We obTain thE same modUli spACe as derived belOW. By integratinG OuT $\pHi_i$, We obTaiN the superpoTentIAl $$W=-\sUm_{I=1}^n\FRaC{1}{2(q_{I+1}-Q_i)}(X_IY_I-y_{I+1}X_{I+1})^2.$$ (WHeN We OBtained the superpoteNtIal (\[ellW\]) we usEd $q_I=0$ and $1$, althouGh we cannot USE It here.) FrOm thE AsSUmption (\[coulomb\]), The F-tErm conditiONs for $X_I$ aNd $Y_I$ gIve $$\frac{X_{i+1}Y_{I+1}-X_IY_I}{q_{I+1}-Q_

0,1,1]_{\vec f},\quad \vec v_4=[n_B, kn_B, 1,1 ]_{ \v ec f },$$ where $[a_1,\ c dots ,a_4]_{\vec f}=\sum_ia _i\ve cf _i$. We have chosen ba s i s ( \[ ba sis f\ ] )so th atthe tor ic data be com ethe standard fo rm in whic h t he last comp one nts of t hev ector s a re $1 $. Wec an dra w the tor ic diagra m usingt h efirs t three component s o f these vectors $\vec v _ i$ , whi chcoincideswi th th a t in Fi g .\ [ c 4.e p s\](b). Furt her general i zat ion {# mo re. s ec} == ===== == = === ========= Up t o now wehave c o nsidere d a bran e syst emwit h tw o k in dsof fiv e br ane s . I t is als opo ssibl e to i n t rodu cemore than two kinds of fi vebr a nes . Torepre sent t he ty pes of bran es we used $q_I=0 $ an d $1$. In th is se ct ion w e allow $q _I$ to bean arbi t rar yi n t eg er. In this case,we d onot need to in t ro du c e the co ef fic ient $ k$ in (\[ c sc ouplings \]) an d w eset $k= 1$ . This m ean s t hat t h e $I $th fi vebraneis a$ (q_I,1)$5-bran e , and the Che r n- S i mo n s co upl ings are gi venb y $$ \lab e l{ kIn o nzero } k_I =q _ {I + 1}-q_I.$$ For simpl ic ity we assu me that the C hern-Simon s c ouplings don ot vanish. This i mplie s that all the adjo int c hiral mu ltiplets$ \ Phi_I$ a ndthe ve cto r mu ltiplets $V_I $ beco me massiv e.It is e asy to sh owth at even i f some o fth e$k _I$ vani s h we obt ai n t he sa me mo d uli sp ace a s de ri ve d be low. B y i n t egra ti ng out $\ Ph i_I$, weo bta in thesuperpote nti a l $$ W= -\ sum_{I= 1}^n\frac{1}{ 2( q_{I+1}-q_ I) }(X _IY_I- Y _ {I+1}X_{ I+1})^2.$$ (When we obt a ined th e s uperp oten tial (\[e llw \]) we us e d $q_I =0$ an d $1$ ,alt h o ugh w e ca nno tuse it her e . ) F rom t he ass umption (\[coulomb\]), th e F- term conditio nsfor$ X _I $ a n d$ Y_I $g ive $ $\frac{X_{I+1}Y _{I+1}-X_I Y_ I }{ q_{I+1}-q_

0,1,1]_{\vec f},\quad \vec_v_4=[n_B,kn_B,1,1]_{\vec f},$$_where $[a_1,\cdots,a_4]_{\vec f}=\sum_ia_i\vec f_i$._We have_chosen_basis (\[basisf\])_so_that the toric_data become the_standard form in which_the last components_of_the vectors are $1$. We can draw the toric diagram using the first three_components_of these_vectors_$\vec_v_i$, which coincides with that_in Fig. \[c4.eps\](b). Further generalization {#more.sec} ====================== Up to_now we_have considered a brane system with two kinds_of_fivebranes. It is_also possible to introduce more than two kinds of_fivebranes. To represent the types of_branes we used_$q_I=0$_and_$1$. In this section_we allow $q_I$ to be an_arbitrary integer. In this case, we_do not need to introduce the coefficient_$k$ in (\[cscouplings\]) and we set_$k=1$. This means that the_$I$th fivebrane_is a $(q_I,1)$5-brane, and the_Chern-Simons couplings are_given by_$$\label{kInonzero} k_I=q_{I+1}-q_I.$$ For simplicity_we assume that the Chern-Simons couplings_do not vanish._This implies that all the adjoint_chiral_multiplets $\Phi_I$ and_the_vector_multiplets $V_I$_become massive. It_is_easy to_show_that even if some of the_$k_I$_vanish we obtain the same moduli space_as derived below. By integrating_out_$\Phi_I$, we obtain the_superpotential $$W=-\sum_{I=1}^n\frac{1}{2(q_{I+1}-q_I)}(X_IY_I-Y_{I+1}X_{I+1})^2.$$ (When we obtained_the superpotential (\[ellw\]) we used $q_I=0$_and $1$,_although we_cannot use it here.) From the assumption (\[coulomb\]), the F-term conditions_for $X_I$ and $Y_I$ give $$\frac{X_{I+1}Y_{I+1}-X_IY_I}{q_{I+1}-q_

2$, provided the intersection is proper. By Kleiman’s transversality of a general translate this is true for a general translate of $\sigma^k_V$ in $\mathbf G(d-7,|H|)$, that is, for a general choice of $\Lambda$ (or equivalently, of $V$). Hence $W^1(\Gamma)$ is finite. The theorem above fails for rational septic scrolls in $\mathbf P^8$ containing sections of degree $d\leq 2$, that is, for the scrolls $S_{a,7-a}$, where $a\neq 3$. We turn to the smooth residual rational curve $Q\subseteq \GG(1,5)$ defined by (\[linint\]). Let $$R_Q \subseteq \PP^5$$ be the quartic scroll whose rulings are parametrized by the curve $Q$. $R_Q$ is a non-degenerate smooth rational normal scroll in $\mathbf P^5$. First, observe that $R_Q$ cannot be a cone. Let us assume $R_Q$ is a cone of vertex $v\in \PP^5$. Then $\langle Q \rangle\cong \PP^4\subseteq \mathbf G(1,5)$ parametrizes the lines passing through $v$. This is a contradiction because $\langle Q \rangle \subseteq \langle \Gamma \rangle \cdot \mathbf G(1,5) = C$. Now assume that $R_Q$ is contained in a hyperplane $H\subseteq \PP^5$. Then $Q$ is contained in the Grassmannian $\GG_H:=\GG(1,H)\subseteq \GG(1,5)$ of lines of $H$. Since $K_{\GG_H}=\OO_{\GG_H}(-5)$, we observe that, by adjunction, the curvilinear sections of $\GG_H$ are curves of arithmetic genus $1$. Because of this fact and since $\mbox{deg}(\GG_H)=5$, it follows that $$\langle Q \rangle \cdot \GG_H = Q + L \subseteq C,$$ where $L$ is a bisecant line to $Q$. But the only line components in $C$ are $L_1, L_2, L_3$ and none of them is bisecant to $Q$. Via Proposition \[rulcurve\], the same argument shows that the

2 $, provided the intersection is proper. By Kleiman ’s transversality of a general translate this is true for a cosmopolitan translate of $ \sigma^k_V$ in $ \mathbf G(d-7,|H|)$, that is, for a cosmopolitan option of $ \Lambda$ (or equivalently, of $ V$). Hence $ W^1(\Gamma)$ is finite. The theorem above fails for rational septic scroll in $ \mathbf P^8 $ containing sections of academic degree $ d\leq 2 $, that is, for the scrolls $ S_{a,7 - a}$, where $ a\neq 3$. We turn to the smooth residual intellectual curvature $ Q\subseteq \GG(1,5)$ defined by (\[linint\ ]). lease $ $ R_Q \subseteq \PP^5$$ be the quartic scroll whose rulings are parametrized by the curvature $ Q$. $ R_Q$ is a non - degenerate smooth rational normal scroll in $ \mathbf P^5$. First, detect that $ R_Q$ cannot be a cone. Let us assume $ R_Q$ is a cone of vertex $ v\in \PP^5$. Then $ \langle Q \rangle\cong \PP^4\subseteq \mathbf G(1,5)$ parametrizes the lines pass through $ v$. This is a contradiction because $ \langle Q \rangle \subseteq \langle \Gamma \rangle \cdot \mathbf G(1,5) = C$. Now assume that $ R_Q$ is contained in a hyperplane $ H\subseteq \PP^5$. Then $ Q$ is contained in the Grassmannian $ \GG_H:=\GG(1,H)\subseteq \GG(1,5)$ of lineage of $ H$. Since $ K_{\GG_H}=\OO_{\GG_H}(-5)$, we observe that, by adjunction, the curvilinear sections of $ \GG_H$ are curves of arithmetic genus $ 1$. Because of this fact and since $ \mbox{deg}(\GG_H)=5 $, it follow that $ $ \langle Q \rangle \cdot \GG_H = Q + L \subseteq C,$$ where $ L$ is a bisecant line to $ Q$. But the only line components in $ C$ are $ L_1, L_2, L_3 $ and none of them is bisecant to $ Q$. Via Proposition \[rulcurve\ ], the like controversy shows that the

2$, pgovided the intersection is proper. By Kleiman’s transbersalith of a general translate thid us trye for a general transuate of $\spgma^k_V$ in $\matibf G(d-7,|H|)$, that is, hkr a gekzral dmoice if $\Lambda$ (or eauivalentlf, of $V$). Hence $W^1(\Caomc)$ is finite. The theorem above fails fjr ratipnwl septic scrojls pn $\matgbf P^8$ containing sections of degres $d\leq 2$, that is, for yhe scrolls $S_{a,7-a}$, where $a\neq 3$. We hurn to the smooth residual rqtiogql curve $Q\sucseteq \GG(1,5)$ defined by (\[ljnint\]). Let $$R_Q \subseteq \PP^5$$ be the quarcic scroll qhise tulings are 'aramenrized by the curve $Q$. $S_Q$ is a non-degenerate smmoty rational normal scrmll in $\mathbf P^5$. Firft, observa chat $R_Q$ cannot be a cinw. Let us dssuow $R_D$ ia e ckne of vectex $v\in \PP^5$. Then $\langlw Q \rangle\cong \PP^4\sunsqnrq \mathbf G(1,5)$ larameeryzes the lines passing through $v$. This iv a contradiction because $\oangle Q \rangle \subsejeq \langle \Gamma \rangle \cdot \mathbf G(1,5) = C$. Now assume that $R_Q$ is ckvtankcd iv a hyperplane $H\subseteq \PP^5$. Then $Q$ is contained yh uhe Grassmannian $\NG_H:=\GG(1,H)\subseteq \GG(1,5)$ pf loges of $H$. Sincg $K_{\GG_H}=\OO_{\GG_G}(-5)$, we observe that, hy adjugctiob, the curdilimear sections of $\GG_H$ are cueves of aritkmeric genus $1$. Because of this faet and sincr $\mbox{deg}(\GG_H)=5$, it follows thaf $$\langle Q \gangle \cdkg \GG_H = Q + L \subsdtea W,$$ where $L$ is a bisecant lige to $Q$. Bnt thz only lkne vomponqnts in $C$ wre $L_1, L_2, L_3$ and none of thfm is bhsecant to $Q$. Via Proposition \[rulcurve\], the same argument svowv that tke

2$, provided the intersection is proper. By of general translate is true for in G(d-7,|H|)$, that is, a general choice $\Lambda$ (or equivalently, of $V$). Hence is finite. The theorem above fails for rational septic scrolls in $\mathbf P^8$ sections of degree $d\leq 2$, that is, for the scrolls $S_{a,7-a}$, where $a\neq We to smooth rational curve $Q\subseteq \GG(1,5)$ defined by (\[linint\]). Let $$R_Q \subseteq \PP^5$$ be the quartic scroll whose are parametrized by the curve $Q$. $R_Q$ is non-degenerate smooth rational normal in $\mathbf P^5$. First, observe $R_Q$ be a Let assume is a cone vertex $v\in \PP^5$. Then $\langle Q \rangle\cong \PP^4\subseteq \mathbf G(1,5)$ parametrizes the lines passing through $v$. This a contradiction Q \rangle \langle \rangle \mathbf G(1,5) = assume that $R_Q$ is contained in \PP^5$. Then $Q$ is contained in the Grassmannian \GG(1,5)$ of of $H$. Since $K_{\GG_H}=\OO_{\GG_H}(-5)$, we observe by adjunction, the curvilinear sections of $\GG_H$ are of arithmetic genus $1$. Because of this fact and since $\mbox{deg}(\GG_H)=5$, it follows that $$\langle \cdot \GG_H = Q L \subseteq C,$$ $L$ a line $Q$. But only line components in $C$ are $L_1, L_2, L_3$ and none them is bisecant to $Q$. Via Proposition \[rulcurve\], the same that

2$, provided the intersection is Proper. By KlEiman’S trAnsVeRsalIty oF a general transLAte tHis is true for a general trAnslaTe OF $\sigMA^k_v$ in $\maThbf G(d-7,|H|)$, THaT IS, foR a GeNerAl CHoIce of $\lamBda$ (or eqUivalently, Of $V$). heNce $W^1(\Gamma)$ is fINiTe. The theorEm aBove fails for RatIonal sEpTic SCrollS in $\MathbF P^8$ contAIning sEctions of DeGRee $d\leQ 2$, That is, fOR ThE scrOlls $S_{a,7-a}$, where $a\neq 3$. WE TuRN to the smooth reSidual RaTIoNAL cuRve $q\subseteq \Gg(1,5)$ dEfineD By (\[lininT\]). leT $$r_q \SubSEteq \PP^5$$ be the quArtic scroll WHosE rulinGs Are PArametRized By THe cUrve $Q$. $R_Q$ is a nOn-deGenerate sMooth rATional nORmal scrOll in $\mAthBf P^5$. firsT, ObSeRve ThAT $R_Q$ CAnNot BE a cOne. Let us AsSuMe $R_Q$ iS a coNE OF VertEx $v\In \PP^5$. then $\lAngle Q \rangle\cOng \pP^4\suBSetEq \matHbf G(1,5)$ pAramEtRizes The linEs pasSiNg through $v$. This iS a coNtradictiOn bEcAusE $\lAngle q \Rangle \SubSetEq \langlE \Gamma \rANglE \cDOT \MaThbf G(1,5) = C$. Now assume thaT $R_q$ IS cOntained In a hypERpLaNE $H\subsetEq \pP^5$. THen $Q$ IS ContaIned IN tHe GrassmAnnian $\gg_H:=\gG(1,h)\subsetEq \gG(1,5)$ of liNeS of $h$. SiNce $K_{\Gg_h}=\OO_{\Gg_H}(-5)$, we obServe thaT, by adJUnction, the curvILinear sectionS Of $\gg_h$ aRE curVes Of arithmetiC genUS $1$. BecAuse OF tHis FAct anD sincE $\mBOx{DEg}(\GG_H)=5$, it follows that $$\lAnGle Q \raNgle \cDot \GG_H = Q + L \subseTeq C,$$ where $L$ IS A Bisecant Line TO $Q$. bUt the only line cOmponEnts in $C$ are $l_1, l_2, L_3$ and nonE of thEm is biseCant to $Q$. ViA pRopositiOn \[rUlcUrvE\], thE SAmE argument showS THat tHe

2$, provided the intersec tion is pr oper. By Kl ei man’ s tr ansversality o f a g eneral translate thisis tr ue fora g enera l trans l at e of$\ si gma ^k _ V$ in $ \ma thbf G( d-7,|H|)$, th at is, for a g e ne ral choice of $\Lambda$ ( orequiva le ntl y , of$V$ ). He nce $W ^ 1(\Gam ma)$ is f in i te. T h e theor e m a bove fails for ration a ls eptic scrollsin $\m at h bf P ^8$ co ntaining s ec tions of degr e e$ d \ leq 2$, that is,for the scr o lls $S_{a ,7 -a} $ , wher e $a\ ne q 3$ . We turnto t he smooth resid u al rati o nal cur ve $Q\ sub set eq \ G G( 1, 5)$ d e fin e dby( \[l inint\]) .Le t $$R _Q \ s u b s eteq \P P^5$ $ bethe quartic s cro ll w h ose ruli ngs a re p ar ametr ized b y the c urve $Q$. $R_Q $ is a non-de gen er ate s mooth ration alnor mal scr oll in$ \ma th b f P^ 5$. First, observ et h at $R_Q$ c annotb eac one. Let u s a ssum e $R_Q$ isa c one of v ertex$ v\ in \PP^5$ .Then $ \l ang leQ \ra n gle\ cong \ PP^4\sub seteq \mathbf G(1,5) $ parametrizes th e li n es p ass ing through $v$ . Thi s is acon t radic tionbe c au s e $\langle Q \rangl e\subse teq \ langle \Gamma \rangle \ c d o t \mathb f G( 1 ,5 ) = C$. Now ass ume t hat $R_Q$i s contai ned i n a hype rplane $H \ s ubseteq\PP ^5$ . T hen $ Q$ is contained i n th eGrassma nni an $\GG _H: =\G G(1 ,H) \s ubseteq \ GG(1,5)$ o fli ne s o f $H$ . Since $ K_ {\G G_ H}= \OO_{ \ GG_H}( -5)$, weob se r vethat, b y a d j unct io n, the cu rv iline ar s e cti ons of$\GG_H$ a rec urve sof arithm etic genus $1 $. Because o fthi s fact a nd since $\mbox{deg}(\GG_H)=5$, it foll ows that $$\ langle Q\ra ngle \ cdo t \GG_H = Q + L \s ub set e q C,$$ w he re$L $ is a bis e c ant line t o $Q $. Butthe only line comp o nen ts in $C$ are $L _1,L _ 2, L_ 3 $a ndno n e o f them is bisecan t to $Q$.Vi a P roposition \[r ul curve\] , the s ame a r gumentshows tha t the

2$,_provided the_intersection is proper. By_Kleiman’s transversality_of_a general_translate_this is true_for a general_translate of $\sigma^k_V$ in_$\mathbf G(d-7,|H|)$, that_is,_for a general choice of $\Lambda$ (or equivalently, of $V$). Hence $W^1(\Gamma)$ is finite. The_theorem_above fails_for_rational_septic scrolls in $\mathbf P^8$_containing sections of degree $d\leq_2$, that_is, for the scrolls $S_{a,7-a}$, where $a\neq 3$. We_turn_to the smooth_residual rational curve $Q\subseteq \GG(1,5)$ defined by (\[linint\]). Let_$$R_Q \subseteq \PP^5$$ be the quartic_scroll whose rulings_are_parametrized_by the curve $Q$. $R_Q$_is a non-degenerate smooth rational normal_scroll in $\mathbf P^5$. First, observe that_$R_Q$ cannot be a cone. Let us_assume $R_Q$ is a cone of_vertex $v\in \PP^5$. Then $\langle_Q \rangle\cong_\PP^4\subseteq \mathbf G(1,5)$ parametrizes the_lines passing through_$v$. This_is a contradiction_because $\langle Q \rangle \subseteq \langle_\Gamma \rangle \cdot_\mathbf G(1,5) = C$. Now assume_that_$R_Q$ is contained_in_a_hyperplane $H\subseteq_\PP^5$. Then $Q$_is_contained in_the_Grassmannian $\GG_H:=\GG(1,H)\subseteq \GG(1,5)$ of lines of_$H$._Since $K_{\GG_H}=\OO_{\GG_H}(-5)$, we observe that, by adjunction,_the curvilinear sections of_$\GG_H$_are curves of arithmetic_genus $1$. Because of this_fact and since $\mbox{deg}(\GG_H)=5$, it follows_that $$\langle_Q \rangle_\cdot \GG_H = Q + L \subseteq C,$$ where $L$ is_a bisecant line to $Q$. But_the only line components_in $C$_are_$L_1, L_2, L_3$_and_none of_them is bisecant to $Q$. Via Proposition_\[rulcurve\], the_same argument shows that the

t)\left(1+e(t)\right))$ and $E_{v_{1},ip}(t,\lambda_{0}(t)\left(1+e(t)\right))$, for $e \in \mathcal{B}$. For $i=1,2$, let $e_{i} \in \mathcal{B}$, and let $\lambda_{i}(t) = \lambda_{0}(t)\left(1+e_{i}(t)\right)$. Let $$F(r,\lambda(t)) = \frac{\left(1-Q_{1}^{2}(\frac{r}{\lambda(t)})\right) \phi_{0}(\frac{r}{\lambda(t)})}{r^{2}\lambda(t)^{2}} r$$ Then, $$v_{sip}(t,\lambda(t)) = 6 \int_{0}^{\infty} \sum_{k=2}^{\infty} v_{k}^{\lambda}(t,r) F(r,\lambda(t)) dr$$ We start with $$|\partial_{2} F(r,\lambda(t))| \leq \frac{C r^{3}\lambda(t)}{(r^{2}+\lambda(t)^{2})^{4}}, \quad |F(r,\lambda(t))| \leq \frac{C r^{3} \lambda(t)^{2}}{(r^{2}+\lambda(t)^{2})^{4}}$$ To understand the Lipshitz (in $e$) dependence of $v_{k}^{\lambda_{0}\left(1+e\right)}$, we start by noting that $v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}$ solves the following equation with $0$ Cauchy data at infinity. $$\left(-\partial_{tt}+\partial_{rr}+\frac{1}{r}\partial_{r}-\frac{4}{r^{2}}\right)\left(v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}\right) = \frac{6 v_{1}(t,r)^{2}}{r^{2}} \left(Q_{\frac{1}{\lambda_{1}}(t)}(r)-Q_{\frac{1}{\lambda_{2}(t)}}(r)\right)$$ There exists an absolute constant $C$ such that, for all $e \in \mathcal{B}$ we have $$C^{-1} \lambda_{0}(t) \leq \lambda_{0}(t

t)\left(1+e(t)\right))$ and $ E_{v_{1},ip}(t,\lambda_{0}(t)\left(1+e(t)\right))$, for $ e \in \mathcal{B}$. For $ i=1,2 $, let $ e_{i } \in \mathcal{B}$, and let $ \lambda_{i}(t) = \lambda_{0}(t)\left(1+e_{i}(t)\right)$. Let $ $ F(r,\lambda(t) ) = \frac{\left(1 - Q_{1}^{2}(\frac{r}{\lambda(t)})\right) \phi_{0}(\frac{r}{\lambda(t)})}{r^{2}\lambda(t)^{2 } } r$$ Then, $ $ v_{sip}(t,\lambda(t) ) = 6 \int_{0}^{\infty } \sum_{k=2}^{\infty } v_{k}^{\lambda}(t, radius) F(r,\lambda(t) ) dr$$ We depart with $ $ |\partial_{2 } F(r,\lambda(t))| \leq \frac{C r^{3}\lambda(t)}{(r^{2}+\lambda(t)^{2})^{4 } }, \quad |F(r,\lambda(t))| \leq \frac{C r^{3 } \lambda(t)^{2}}{(r^{2}+\lambda(t)^{2})^{4}}$$ To understand the Lipshitz (in $ e$) dependence of $ v_{k}^{\lambda_{0}\left(1+e\right)}$, we start by notice that $ v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}$ solves the following equality with $ 0 $ Cauchy data at infinity. $ $ \left(-\partial_{tt}+\partial_{rr}+\frac{1}{r}\partial_{r}-\frac{4}{r^{2}}\right)\left(v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}\right) = \frac{6 v_{1}(t, r)^{2}}{r^{2 } } \left(Q_{\frac{1}{\lambda_{1}}(t)}(r)-Q_{\frac{1}{\lambda_{2}(t)}}(r)\right)$$ There exists an absolute changeless $ C$ such that, for all $ e \in \mathcal{B}$ we have $ $ C^{-1 } \lambda_{0}(t) \leq \lambda_{0}(t

t)\levt(1+e(t)\right))$ and $E_{v_{1},ip}(t,\lambdx_{0}(t)\left(1+e(t)\right))$, for $e \in \mathczl{B}$. For $k=1,2$, let $e_{i} \in \mathcal{B}$, and let $\lqmbda_{u}(t) = \lambda_{0}(t)\left(1+e_{i}(t)\righg)$. Let $$F(r,\lwmbda(t)) = \drac{\owft(1-Q_{1}^{2}(\frac{r}{\lejbda(t)})\rinkt) \phj_{0}(\nrac{r}{\nembda(t)})}{r^{2}\lambda(t)^{2}} t$$ Then, $$v_{sip}(t,\nambda(t)) = 6 \int_{0}^{\itfgy} \sum_{k=2}^{\infty} v_{k}^{\lambda}(t,r) F(r,\lambda(t)) dr$$ Re stary aith $$|\partial_{2} F(t,\lambcw(t))| \lsq \frac{C r^{3}\lambda(t)}{(r^{2}+\lambda(t)^{2})^{4}}, \quad |F(r,\lzmbda(t))| \neq \frac{C r^{3} \lsmbda(t)^{2}}{(r^{2}+\lambda(t)^{2})^{4}}$$ To understajd tje Lipshitz (in $e$) dfpendence od $v_{k}^{\jqmbda_{0}\left(1+e\rieht)}$, we stagc by noting that $v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}$ solves the follpwing equajnin ahth $0$ Cauchy data at infinity. $$\left(-\pardial_{tt}+\psrtial_{rr}+\frac{1}{r}\psrtmal_{r}-\drac{4}{r^{2}}\right)\left(v_{2}^{\lambda_{1}}-t_{2}^{\lambda_{2}}\right) = \frac{6 v_{1}(j,r)^{2}}{r^{2}} \left(Q_{\fsae{1}{\lambda_{1}}(t)}(r)-Q_{\frac{1}{\lambda_{2}(t)}}(e)\rught)$$ Jhere exirrs xn zbxomute clnsvant $C$ such that, for aol $e \in \mathcal{B}$ we hwnr $$C^{-1} \lambda_{0}(t) \meq \laibqa_{0}(t

t)\left(1+e(t)\right))$ and $E_{v_{1},ip}(t,\lambda_{0}(t)\left(1+e(t)\right))$, for $e \in \mathcal{B}$. let \in \mathcal{B}$, let $\lambda_{i}(t) = \phi_{0}(\frac{r}{\lambda(t)})}{r^{2}\lambda(t)^{2}} Then, $$v_{sip}(t,\lambda(t)) = \int_{0}^{\infty} \sum_{k=2}^{\infty} v_{k}^{\lambda}(t,r) dr$$ We start with $$|\partial_{2} F(r,\lambda(t))| \frac{C r^{3}\lambda(t)}{(r^{2}+\lambda(t)^{2})^{4}}, \quad |F(r,\lambda(t))| \leq \frac{C r^{3} \lambda(t)^{2}}{(r^{2}+\lambda(t)^{2})^{4}}$$ To understand the Lipshitz (in dependence of $v_{k}^{\lambda_{0}\left(1+e\right)}$, we start by noting that $v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}$ solves the following equation $0$ data infinity. = \frac{6 v_{1}(t,r)^{2}}{r^{2}} \left(Q_{\frac{1}{\lambda_{1}}(t)}(r)-Q_{\frac{1}{\lambda_{2}(t)}}(r)\right)$$ There exists an absolute constant $C$ such that, for all $e \in we have $$C^{-1} \lambda_{0}(t) \leq \lambda_{0}(t

t)\left(1+e(t)\right))$ and $E_{v_{1},ip}(t,\lambdA_{0}(t)\left(1+e(t)\riGht))$, foR $e \iN \maThCal{B}$. for $i=1,2$, Let $e_{i} \in \mathcal{b}$, And lEt $\lambda_{i}(t) = \lambda_{0}(t)\left(1+e_{I}(t)\rigHt)$. lEt $$F(r,\LAmBda(t)) = \fRac{\left(1-q_{1}^{2}(\FrAC{R}{\laMbDa(T)})\riGhT) \PhI_{0}(\frac{R}{\laMbda(t)})}{r^{2}\lAmbda(t)^{2}} r$$ TheN, $$v_{sIp}(T,\lambda(t)) = 6 \int_{0}^{\iNFtY} \sum_{k=2}^{\infty} V_{k}^{\lAmbda}(t,r) F(r,\lamBda(T)) dr$$ We sTaRt wITh $$|\parTiaL_{2} F(r,\laMbda(t))| \lEQ \frac{C R^{3}\lambda(t)}{(r^{2}+\LaMBda(t)^{2})^{4}}, \quAD |F(r,\lambDA(T))| \lEq \frAc{C r^{3} \lambda(t)^{2}}{(r^{2}+\lambdA(T)^{2})^{4}}$$ TO Understand the LIpshitZ (iN $E$) dEPEndEncE of $v_{k}^{\lambdA_{0}\lEft(1+e\rIGht)}$, we stARt BY NOtiNG that $v_{2}^{\lambda_{1}}-v_{2}^{\Lambda_{2}}$ solveS The FollowInG eqUAtion wIth $0$ CaUcHY daTa at infinitY. $$\lefT(-\partial_{tT}+\partiAL_{rr}+\frac{1}{R}\Partial_{R}-\frac{4}{r^{2}}\RigHt)\lEft(v_{2}^{\LAmBdA_{1}}-v_{2}^{\lAmBDa_{2}}\rIGhT) = \frAC{6 v_{1}(t,R)^{2}}{r^{2}} \left(Q_{\fRaC{1}{\lAmbda_{1}}(T)}(r)-Q_{\fRAC{1}{\LAmbdA_{2}(t)}}(r)\RighT)$$ TherE exists an absoLutE conSTanT $C$ sucH that, For aLl $E \in \maThcal{B}$ We havE $$C^{-1} \Lambda_{0}(t) \leq \lambdA_{0}(t

t)\left(1+e(t)\right))$ an d $E_{v_{1 },ip} (t, \la mb da_{ 0}(t )\left(1+e(t)\ r ight ))$, for $e \in \mathc al{B} $. For$ i= 1,2$, let $e _ {i } \in \ ma thc al { B} $, an d l et $\la mbda_{i}(t ) = \ lambda_{0}(t ) \l eft(1+e_{i }(t )\right)$. L et$$F(r, \l amb d a(t)) =\frac {\left ( 1-Q_{1 }^{2}(\fr ac { r}{\la m bda(t)} ) \ ri ght) \phi_{0}(\frac{r } {\ l ambda(t)})}{r^ {2}\la mb d a( t ) ^{2 }}r$$ Then,$$ v_{si p }(t,\la m bd a ( t ))= 6 \int_{0}^{ \infty} \su m _{k =2}^{\ in fty } v_{k} ^{\la mb d a}( t,r) F(r,\l ambd a(t)) dr$ $ We s t art wit h $$|\pa rtial_ {2} F( r,\l a mb da (t) )| \le q \ fra c {Cr^{3}\la mb da (t)}{ (r^{ 2 } + \ lamb da( t)^{ 2})^{ 4}}, \quad |F (r, \lam b da( t))|\leq\fra c{ C r^{ 3} \la mbda( t) ^{2}}{(r^{2}+\l ambd a(t)^{2}) ^{4 }} $$To unde r standthe Li pshitz(in $e$ ) de pe n d e nc e of $v_{k}^{\lamb da _ { 0} \left(1+ e\righ t )} $, we start b y n otin g that$v_{ 2 }^ {\lambda _{1}}- v _{ 2} ^{\lamb da _{2}}$ s olv esthe f o llow ing eq uation w ith $ 0 $ Cauchy dataa t infinity. $ $ \l e f t( - \par tia l_{tt}+\par tial _ {rr} +\fr a c{ 1}{ r }\par tial_ {r } -\ f rac{4}{r^{2}}\right )\ left(v _{2}^ {\lambda_{1}} -v_{2}^{\l a m b da_{2}}\ righ t )= \frac{6 v_{1} (t,r) ^{2}}{r^{2 } } \left( Q_{\f rac{1}{\ lambda_{1 } } (t)}(r)- Q_{ \fr ac{ 1}{ \ l am bda_{2}(t)}}( r ) \rig ht )$$ The reexistsanabs olu teco nstant $C $ such t ha t, f or al l $e\ in \math ca l{B }$ we have $$C^{- 1} \l ambd a_ {0 } (t) \leq \ l am b d a_{0 }( t

t)\left(1+e(t)\right))$ and_$E_{v_{1},ip}(t,\lambda_{0}(t)\left(1+e(t)\right))$, for_$e \in \mathcal{B}$. For_$i=1,2$, let_$e_{i}_\in \mathcal{B}$,_and_let $\lambda_{i}(t) =_\lambda_{0}(t)\left(1+e_{i}(t)\right)$. Let $$F(r,\lambda(t))_= \frac{\left(1-Q_{1}^{2}(\frac{r}{\lambda(t)})\right) \phi_{0}(\frac{r}{\lambda(t)})}{r^{2}\lambda(t)^{2}} r$$_Then, $$v_{sip}(t,\lambda(t)) =_6_\int_{0}^{\infty} \sum_{k=2}^{\infty} v_{k}^{\lambda}(t,r) F(r,\lambda(t)) dr$$ We start with $$|\partial_{2} F(r,\lambda(t))| \leq \frac{C r^{3}\lambda(t)}{(r^{2}+\lambda(t)^{2})^{4}}, \quad_|F(r,\lambda(t))|_\leq \frac{C_r^{3}_\lambda(t)^{2}}{(r^{2}+\lambda(t)^{2})^{4}}$$_To understand the Lipshitz (in_$e$) dependence of $v_{k}^{\lambda_{0}\left(1+e\right)}$, we_start by_noting that $v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}$ solves the following equation with_$0$_Cauchy data at_infinity. $$\left(-\partial_{tt}+\partial_{rr}+\frac{1}{r}\partial_{r}-\frac{4}{r^{2}}\right)\left(v_{2}^{\lambda_{1}}-v_{2}^{\lambda_{2}}\right) = \frac{6 v_{1}(t,r)^{2}}{r^{2}} \left(Q_{\frac{1}{\lambda_{1}}(t)}(r)-Q_{\frac{1}{\lambda_{2}(t)}}(r)\right)$$ There exists an_absolute constant $C$ such that, for_all $e \in_\mathcal{B}$_we_have $$C^{-1} \lambda_{0}(t) \leq_\lambda_{0}(t

clearly commute: each $\mathrm{Aut}(Z_i)$ is abelian so that $\mathrm{Loc}(C_n, \mathcal{G})$ is abelian as well. Thus, we have proved that the centraliser of any element of $\mathrm{Inn}(G_n) \rtimes \mathrm{Loc}(C_n, \mathcal{G})$ contains a free abelian group of infinite rank. Therefore, $\mathrm{Inn}(G_n) \rtimes \mathrm{Loc}(C_n, \mathcal{G})$ (and a fortiori $\mathrm{Aut}(G_n)$) cannot be acylindrically hyperbolic according to [@OsinAcyl Corollary 6.9]. In this section, it was more convenient to work with a CAT(0) cube complex rather than with a quasi-median graph because results already available in the literature allowed us to shorten the arguments. However, we emphasize that a quasi-median proof is possible. The main steps are the followings. First, as in Lemma \[prop:extendT\], the action $\Gamma \mathcal{G} \curvearrowright X(\Gamma, \mathcal{G})$ extends to an action $\mathrm{Aut}(\Gamma \mathcal{G}) \curvearrowright X(\Gamma, \mathcal{G})$ via $\iota(g) \varphi \cdot x = g \varphi(x)$ where $x \in X(\Gamma, \mathcal{G})$ is a vertex. So Theorem \[thm:noT\] also follows since Niblo and Roller’s argument [@MR1459140] can be reproduced almost word for word in the quasi-median setting; or alternatively, the theorem follows from the combination of [@MR1459140] with [@Qm Proposition 4.16] which shows that any quasi-median graph admits a “dual” CAT(0) cube complex. Next, in order to prove Theorem \[thm:acyl\_hyp\], we need to notice that $X(\Gamma, \mathcal{G})$ is hyperbolic (as a consequence of [@Qm Fact 8.33]) and that the element $g$ constructed in the proof above turns out to be a WPD element. For the latter observation, one can easily prove the following criterion by following the arguments of [@MoiAcylHyp Theorem 18]: given a group $G$ acting on a hyperbolic quasi-median graph $X$, if an element $g \in G$ skewers

clearly commute: each $ \mathrm{Aut}(Z_i)$ is abelian so that $ \mathrm{Loc}(C_n, \mathcal{G})$ is abelian as well. Thus, we have testify that the centraliser of any chemical element of $ \mathrm{Inn}(G_n) \rtimes \mathrm{Loc}(C_n, \mathcal{G})$ contains a free abelian group of infinite rank. consequently, $ \mathrm{Inn}(G_n) \rtimes \mathrm{Loc}(C_n, \mathcal{G})$ (and a fortiori $ \mathrm{Aut}(G_n)$) cannot be acylindrically hyperbolic according to [ @OsinAcyl Corollary   6.9 ]. In this section, it was more convenient to cultivate with a CAT(0) cube complex preferably than with a quasi - median graph because results already available in the literature allow us to shorten the arguments. However, we emphasize that a quasi - median proof is possible. The chief steps are the followings. First, as in Lemma \[prop: extendT\ ], the legal action $ \Gamma \mathcal{G } \curvearrowright X(\Gamma, \mathcal{G})$ extends to an action $ \mathrm{Aut}(\Gamma \mathcal{G }) \curvearrowright X(\Gamma, \mathcal{G})$ via $ \iota(g) \varphi \cdot x = g \varphi(x)$ where $ x \in X(\Gamma, \mathcal{G})$ is a vertex. So Theorem \[thm: noT\ ] besides follows since Niblo and Roller ’s argumentation [ @MR1459140 ] can be reproduced almost word for discussion in the quasi - median setting; or alternatively, the theorem follows from the combination of [ @MR1459140 ] with [ @Qm Proposition 4.16 ] which shows that any quasi - median graph accommodate a “ dual ” CAT(0) cube complex. Next, in order to prove Theorem \[thm: acyl\_hyp\ ], we need to notice that $ X(\Gamma, \mathcal{G})$ is hyperbolic (as a consequence of [ @Qm Fact 8.33 ]) and that the element $ g$ constructed in the proof above sour out to be a WPD element. For the latter observation, one can well rise the follow criterion by following the argumentation of [ @MoiAcylHyp Theorem 18 ]: given a group $ G$ acting on a hyperbolic quasi - median graph $ X$, if an element $ gravitational constant \in G$ skewers

clfarly commute: each $\mathro{Aut}(Z_i)$ is abelicb so tiat $\matgrm{Loc}(C_n, \mathcal{G})$ is abelian as well. Tyus, wt have proved that ghe centrwliser od anb element of $\matidm{Inn}(G_n) \rtimea \matkrn{Loc}(C_n, \mathcal{N})$ contains d free abelian gfobp of infinite rank. Therefore, $\mathrm{Ynn}(G_n) \ryiles \mathrm{Loc}(C_g, \mauhcwl{G})$ (znd a fortiori $\mathrm{Aut}(G_n)$) cannot ge acylpndrically hyperbplic according to [@OsinAcyl Corlllary 6.9]. In this sectlon, it was nore xonvenient tu work with a CAT(0) cube complex rather than with a quari-mednan graph bgequsf results alceady wvailable in the litarature allowed us to shmrtwn the arguments. Howeter, we emphasize thaj a quasi-madnan proof is possible. Tye mahn sdeps qre tht fplmowingd. Fmrst, as in Memma \[prop:eztendT\], the action $\Gsmiq \mathcal{G} \cudvearrjwwight X(\Gamma, \mathcal{G})$ extends to an actpon $\jathrm{Aut}(\Gamma \mathcal{G}) \curvearrowright X(\Gamla, \mathcaj{G})$ via $\iota(g) \varphi \cdot x = g \varphi(x)$ where $x \in F(\Gamme, \oatkgwu{T})$ ls a vertex. So Theorem \[thm:noT\] also follows sigde Npblo and Roller’s crgument [@MR1459140] can ne rrkroduced almosj word ykr word in the quasi-ledian fettibg; or alttrnatovely, the theorem follows feom the combpnatuon of [@MR1459140] with [@Qm 'roposition 4.16] whivh shpws that any quasi-mediau grapg admits a “fual” CAT(0) djbe complex. Next, in osder to kfove Theorem \[thm:asyl\_hyp\], we neeb to notkce jhat $X(\Gwmma, \mathcwl{G})$ is hyperbolic (as a clnseqoence mf [@Qm Fact 8.33]) and that the element $g$ constrnrted in the ptoox atove turus out to be a WPD qlement. For thg latter jbserxation, one can eavily prove ehe following friterion by followigg tye aegumentr of [@MoiAcylHyp Theorem 18]: given a geoup $G$ acting on a hykedbolic quasi-medncn graph $X$, if an rleoene $h \mn G$ fnewers

clearly commute: each $\mathrm{Aut}(Z_i)$ is abelian so \mathcal{G})$ abelian as Thus, we have any of $\mathrm{Inn}(G_n) \rtimes \mathcal{G})$ contains a abelian group of infinite rank. Therefore, \rtimes \mathrm{Loc}(C_n, \mathcal{G})$ (and a fortiori $\mathrm{Aut}(G_n)$) cannot be acylindrically hyperbolic according to Corollary 6.9]. In this section, it was more convenient to work with a cube rather with quasi-median graph because results already available in the literature allowed us to shorten the arguments. However, emphasize that a quasi-median proof is possible. The steps are the followings. as in Lemma \[prop:extendT\], the $\Gamma \curvearrowright X(\Gamma, extends an $\mathrm{Aut}(\Gamma \mathcal{G}) \curvearrowright \mathcal{G})$ via $\iota(g) \varphi \cdot x = g \varphi(x)$ where $x \in X(\Gamma, \mathcal{G})$ is a vertex. Theorem \[thm:noT\] since Niblo Roller’s [@MR1459140] be reproduced almost word in the quasi-median setting; or follows from the combination of [@MR1459140] with [@Qm 4.16] which that any quasi-median graph admits a CAT(0) cube complex. Next, in order to prove \[thm:acyl\_hyp\], we need to notice that $X(\Gamma, \mathcal{G})$ is hyperbolic (as a consequence of [@Qm and that the element constructed in the above out be WPD element. the latter observation, one can easily prove the following criterion by the arguments of [@MoiAcylHyp Theorem 18]: given a group $G$ a quasi-median graph $X$, an element $g \in skewers

clearly commute: each $\mathrm{AUt}(Z_i)$ is abelIan so ThaT $\maThRm{LoC}(C_n, \mAthcal{G})$ is abeliAN as wEll. Thus, we have proved thaT the cEnTRaliSEr Of any Element OF $\mATHrm{inN}(G_N) \rtImES \mAthrm{loc}(c_n, \mathcAl{G})$ containS a fReE abelian grouP Of Infinite raNk. THerefore, $\mathRm{INn}(G_n) \rtImEs \mAThrm{LOc}(C_N, \mathCal{G})$ (anD A fortiOri $\mathrm{auT}(g_n)$) cannOT be acylINDrIcalLy hyperbolic accorDInG To [@OsinAcyl CoroLlary 6.9]. IN tHIs SECtiOn, iT was more coNvEnienT To work wITh A cat(0) cuBE complex ratheR than with a qUAsi-Median GrAph BEcause ResulTs ALreAdy availablE in tHe literatUre allOWed us to SHorten tHe arguMenTs. HOwevER, wE eMphAsIZe tHAt A quASi-mEdian proOf Is PossiBle. THE MAIn stEps Are tHe folLowings. First, aS in lemmA \[ProP:exteNdT\], thE actIoN $\GammA \mathcAl{G} \cuRvEarrowright X(\GamMa, \maThcal{G})$ extEndS tO an AcTion $\mAThrm{AuT}(\GaMma \Mathcal{g}) \curveaRRowRiGHT x(\GAmma, \mathcal{G})$ via $\iotA(g) \VARpHi \cdot x = g \Varphi(X)$ WhErE $X \in X(\GammA, \mAthCal{G})$ IS A vertEx. So tHeOrem \[thm:nOT\] also FOlLoWs since niBlo and roLleR’s aRgumeNT [@MR1459140] cAn be reProduced AlmosT Word for word in tHE quasi-median sETtING; oR AlteRnaTively, the thEoreM FollOws fROm The COmbinAtion Of [@mr1459140] wITh [@Qm Proposition 4.16] whicH sHows thAt any Quasi-median grAph admits a “DUAL” CAT(0) cube CompLEx. nExt, in order to prOve ThEorem \[thm:acYL\_hyp\], we neEd to nOtice thaT $X(\Gamma, \maTHCal{G})$ is hyPerBolIc (aS a cONSeQuence of [@Qm FacT 8.33]) ANd thAt The elemEnt $G$ constrUctEd iN thE prOoF above turNs out to bE a wPd eLeMenT. For tHE latter oBsErvAtIon, One caN Easily Prove The fOlLoWIng CriteriON bY FOlloWiNg The aRguMeNts of [@moiACYlHYp TheorEm 18]: given a gRouP $g$ actInG oN a hyperBolic quasi-medIaN graph $X$, if aN eLemEnt $g \in g$ SKewers

clearly commute: each $\m athrm{Aut} (Z_i) $ i s a be lian sothat $\mathrm{ L oc}( C_n, \mathcal{G})$ isabeli an as w e ll . Thu s, we h a ve p rov ed t hat t h ecentr ali ser ofany elemen t o f$\mathrm{Inn } (G _n) \rtime s \ mathrm{Loc}( C_n , \mat hc al{ G })$ c ont ainsa free abelia n group o fi nfinit e rank.T h er efor e, $\mathrm{Inn}( G _n ) \rtimes \math rm{Loc }( C _n , \ma thc al{G})$ (a nd a fo r tiori $ \ ma t h r m{A u t}(G_n)$) can not be acyl i ndr ically h ype r bolicaccor di n g t o [@OsinAcy l Co rollary 6 .9]. I n thiss ection, it wa s m ore con v en ie ntto wor k w ith a C AT(0) cu be c omple x ra t h e r tha n w itha qua si-median gra phbeca u seresul ts al read yavail able i n the l iterature allow ed u s to shor ten t hear gumen t s. How eve r,we emph asize t h ataq u a si -median proof is p os s i bl e. The m ain st e ps a r e the fo ll owi ngs. F irst, asi nLemma \[ prop:e x te nd T\], th eaction $ \Ga mma \mat h cal{ G} \cu rvearrow right X(\Gamma, \mat h cal{G})$ exte n ds t oa n ac tio n $\mathrm{ Aut} ( \Gam ma \ m at hca l {G})\curv ea r ro w right X(\Gamma, \ma th cal{G} )$ vi a $\iota(g) \ varphi \cd o t x = g \v arph i (x ) $ where $x \in X(\G amma, \mat h cal{G})$ is a vertex. So Theor e m \[thm:n oT\ ] a lso fo l l ow s since Niblo a nd R ol ler’s a rgu ment [@ MR1 459 140 ] c an be repro duced al mo st w or d f or wo r d in the q uas i- med ian s e tting; or a lter na ti v ely , the t h eo r e m fo ll ow s fr omth e com bina t ion of [@M R1459140] wi t h [@ Qm P roposit ion 4.16] whi ch shows tha tany quasi - m edian gr aph admits a “dual” CAT ( 0) cube co mplex . Ne xt, in or der to pr ove Theore m \[th m:acy l\ _hy p \ ], we n ee d t onotice tha t $X( \Gamm a, \ma thcal{G })$ is hyperbolic( asa consequence of [@Q m Fa ct8 .3 3 ])an d th a t the element $g $ construc te d i n the proo f ab ov e turns out to be a WPD ele ment. For the latt er obs e r vat ion, one c an easil y prove t h e fol l ow ing c rit erionby fo llowi ng the arg ument s of [ @M oiAcyl Hyp T he orem 18] : given a group $G$ act ing on a hy per bolic qua si- m edi an graph$X$, if an ele men t $ g \in G$ skewe rs

clearly_commute: each_$\mathrm{Aut}(Z_i)$ is abelian so_that $\mathrm{Loc}(C_n,_\mathcal{G})$_is abelian_as_well. Thus, we_have proved that_the centraliser of any_element of $\mathrm{Inn}(G_n)_\rtimes_\mathrm{Loc}(C_n, \mathcal{G})$ contains a free abelian group of infinite rank. Therefore, $\mathrm{Inn}(G_n) \rtimes \mathrm{Loc}(C_n,_\mathcal{G})$_(and a_fortiori_$\mathrm{Aut}(G_n)$)_cannot be acylindrically hyperbolic according_to [@OsinAcyl Corollary 6.9]. In this section,_it was_more convenient to work with a CAT(0) cube_complex_rather than with_a quasi-median graph because results already available in the_literature allowed us to shorten the_arguments. However, we_emphasize_that_a quasi-median proof is_possible. The main steps are the_followings. First, as in Lemma \[prop:extendT\],_the action $\Gamma \mathcal{G} \curvearrowright X(\Gamma, \mathcal{G})$_extends to an action $\mathrm{Aut}(\Gamma \mathcal{G})_\curvearrowright X(\Gamma, \mathcal{G})$ via $\iota(g)_\varphi \cdot_x = g \varphi(x)$ where_$x \in X(\Gamma,_\mathcal{G})$ is_a vertex. So_Theorem \[thm:noT\] also follows since Niblo_and Roller’s argument_[@MR1459140] can be reproduced almost word_for_word in the_quasi-median_setting;_or alternatively,_the theorem follows_from_the combination_of_[@MR1459140] with [@Qm Proposition 4.16] which_shows_that any quasi-median graph admits a “dual”_CAT(0) cube complex. Next,_in_order to prove Theorem_\[thm:acyl\_hyp\], we need to notice_that $X(\Gamma, \mathcal{G})$ is hyperbolic (as_a consequence_of [@Qm_Fact 8.33]) and that the element $g$ constructed in the proof_above turns out to be a_WPD element. For the_latter observation,_one_can easily prove_the_following criterion_by following the arguments of [@MoiAcylHyp Theorem_18]: given_a group $G$ acting on a_hyperbolic quasi-median graph $X$,_if_an element $g \in G$ skewers

finally, the constraint : $$\begin{aligned} \lambda_{ij}&\leq\beta \tau_{ij}+\mu_j^{e-e_{ij}}-\mu_i^e\\ &\overset{\eqref{eq:dualconstraint6}}{\leq}\beta_t \tau_{ij}+e_{ij}P_j+\mu_j^e-\mu_i^e\\ &\overset{\eqref{eq:dualconstraint9}}{\leq}\beta_t \tau_{ij}+e_{ij}P_j+\beta \tau_{ji}+e_{ji}P_i.\end{aligned}$$ Replacing $P_i=p_i+\beta_c$ and rearranging the terms: $$\begin{aligned} \label{eq:lambdaupperbound} \lambda_{ij}\leq \beta_t(\tau_{ij}+\tau_{ji})+e_{ij}(p_j+\beta_c)+e_{ji}(p_i+\beta_c)=\overline{\lambda}_{ij},\end{aligned}$$ where the last equality follows from the definition provided in the proposition. Hence, we get the desired upper bound on the prices using the upper bound on the dual variables. Proof of Proposition \[prop:monoprofitcs\] {#app:profcsmono} ------------------------------------------ Using Assumption \[ass:sigmalambda\], we see that $\frac{(3-5\sigma)}{2}\leq0$ and $\frac{(3-5\sigma)}{2}{\ell_{\max}}\leq{\lambda_{ij}^{m}}\leq\underset{i,j}{\max} ~\overline{\lambda}_{ij}\leq\frac{(3\sigma-1)(3-\sigma)}{4(5-3\sigma)}\ell_{\max}\leq\frac{3\sigma-1}{2}{\ell_{\max}}$. Hence, the optimal prices fall in the region $[(1-\sigma){\ell_{\max}},\sigma{\ell_{\max}})$, and are given by: $$\label{eq:optpriceslinear} {\ell_{ij}^m}=\frac{(1+\sigma){\ell_{\max}}+{\lambda_{ij}^{m}}}{4}.$$ The dual problem with optimal prices in can be written as: \[eq:dualwithoptimalprices\] $$\begin{aligned} \label{eq:dualobjectivewithoptimalprices} &\

finally, the constraint: $ $ \begin{aligned } \lambda_{ij}&\leq\beta \tau_{ij}+\mu_j^{e - e_{ij}}-\mu_i^e\\ & \overset{\eqref{eq: dualconstraint6}}{\leq}\beta_t \tau_{ij}+e_{ij}P_j+\mu_j^e-\mu_i^e\\ & \overset{\eqref{eq: dualconstraint9}}{\leq}\beta_t \tau_{ij}+e_{ij}P_j+\beta \tau_{ji}+e_{ji}P_i.\end{aligned}$$ Replacing $ P_i = p_i+\beta_c$ and rearranging the term: $ $ \begin{aligned } \label{eq: lambdaupperbound } \lambda_{ij}\leq \beta_t(\tau_{ij}+\tau_{ji})+e_{ij}(p_j+\beta_c)+e_{ji}(p_i+\beta_c)=\overline{\lambda}_{ij},\end{aligned}$$ where the final equality follows from the definition provided in the proposition. therefore, we get the desired upper bound on the price using the upper bound on the double variables. Proof of Proposition   \[prop: monoprofitcs\ ] { # app: profcsmono } ------------------------------------------ Using Assumption   \[ass: sigmalambda\ ], we experience that $ \frac{(3 - 5\sigma)}{2}\leq0 $ and $ \frac{(3 - 5\sigma)}{2}{\ell_{\max}}\leq{\lambda_{ij}^{m}}\leq\underset{i, j}{\max } ~\overline{\lambda}_{ij}\leq\frac{(3\sigma-1)(3-\sigma)}{4(5 - 3\sigma)}\ell_{\max}\leq\frac{3\sigma-1}{2}{\ell_{\max}}$. Hence, the optimal price accrue in the region $ [ (1-\sigma){\ell_{\max}},\sigma{\ell_{\max}})$, and are give by: $ $ \label{eq: optpriceslinear } { \ell_{ij}^m}=\frac{(1+\sigma){\ell_{\max}}+{\lambda_{ij}^{m}}}{4}.$$ The double trouble with optimal prices in can be written as: \[eq: dualwithoptimalprices\ ] $ $ \begin{aligned } \label{eq: dualobjectivewithoptimalprices } & \

fijally, the constraint : $$\benin{aligned} \lamgda_{ij}&\leq\ceta \tau_{ij}+\mu_j^{e-e_{ij}}-\mu_i^e\\ &\ovecset{\wqref{tz:dualconstraint6}}{\leq}\ceta_t \tau_{pj}+e_{ij}P_j+\mu_h^e-\mu_m^e\\ &\overset{\eqrxr{eq:dualgjnstdwint9}}{\nxq}\beta_t \tau_{im}+e_{ij}P_j+\beta \dau_{ji}+e_{ji}P_i.\end{anienzd}$$ Replacing $P_i=p_i+\beta_c$ and rearrangigg the yegms: $$\begin{aligngd} \jabem{vq:oambdaupperbound} \lambda_{ij}\lsq \beta_u(\tau_{ij}+\tau_{ji})+e_{ij}(p_j+\beya_c)+e_{ji}(p_i+\beta_c)=\overline{\lambdw}_{ij},\ejd{aligned}$$ where thf last equaoity dollows from the definpcion providgd in the proposition. Hence, we gdt thz desired uk'wr hmund on the pricvs using the miper bognd on yhe dual varianles. Pcoof of Proposition \[prop:moioprofitcs\] {#app:profcsiono} ------------------------------------------ Using Aasumption \[ass:sigmaoanbda\], fe sae tfqt $\wrad{(3-5\smgmz)}{2}\leq0$ ajd $\hrac{(3-5\sigma)}{2}{\elm_{\max}}\leq{\lambea_{ij}^{m}}\leq\underset{i,j}{\msx} ~\iverline{\lambdz}_{ij}\leq\srwc{(3\sigma-1)(3-\sigma)}{4(5-3\sigma)}\ell_{\max}\leq\frac{3\sigma-1}{2}{\ell_{\mdx}}$. Gence, the optimal pricew fall in the region $[(1-\digma){\ell_{\mwx}},\sigma{\ell_{\max}})$, and are given by: $$\label{eq:optpriceslitear} {\eol_{lj}^m}=\ffqc{(1+\digma){\ell_{\max}}+{\lambda_{ij}^{m}}}{4}.$$ The dual problem with opejmsl prices in can be written ad: \[ea:qualwithoptimxlpricza\] $$\gegin{aligned} \label{ee:dualobtecticewithoptymallrices} &\

finally, the constraint : $$\begin{aligned} \lambda_{ij}&\leq\beta \tau_{ij}+\mu_j^{e-e_{ij}}-\mu_i^e\\ &\overset{\eqref{eq:dualconstraint9}}{\leq}\beta_t \tau_{ji}+e_{ji}P_i.\end{aligned}$$ Replacing and rearranging the where last equality follows the definition provided the proposition. Hence, we get the upper bound on the prices using the upper bound on the dual variables. of Proposition \[prop:monoprofitcs\] {#app:profcsmono} ------------------------------------------ Using Assumption \[ass:sigmalambda\], we see that $\frac{(3-5\sigma)}{2}\leq0$ and ~\overline{\lambda}_{ij}\leq\frac{(3\sigma-1)(3-\sigma)}{4(5-3\sigma)}\ell_{\max}\leq\frac{3\sigma-1}{2}{\ell_{\max}}$. the prices in the region $[(1-\sigma){\ell_{\max}},\sigma{\ell_{\max}})$, and are given by: $$\label{eq:optpriceslinear} {\ell_{ij}^m}=\frac{(1+\sigma){\ell_{\max}}+{\lambda_{ij}^{m}}}{4}.$$ The dual problem with optimal prices can be written as: \[eq:dualwithoptimalprices\] $$\begin{aligned} \label{eq:dualobjectivewithoptimalprices} &\

finally, the constraint : $$\begin{Aligned} \lamBda_{ij}&\Leq\BetA \tAu_{ij}+\Mu_j^{e-E_{ij}}-\mu_i^e\\ &\overset{\EQref{Eq:dualconstraint6}}{\leq}\betA_t \tau_{Ij}+E_{Ij}P_j+\MU_j^E-\mu_i^e\\ &\Overset{\EQrEF{Eq:dUaLcOnsTrAInT9}}{\leq}\bEta_T \tau_{ij}+e_{Ij}P_j+\beta \taU_{ji}+E_{jI}P_i.\end{aligneD}$$ rePlacing $P_i=p_I+\beTa_c$ and rearraNgiNg the tErMs: $$\bEGin{alIgnEd} \labEl{eq:laMBdauppErbound} \laMbDA_{ij}\leq \BEta_t(\tau_{IJ}+\TaU_{ji})+e_{Ij}(p_j+\beta_c)+e_{ji}(p_i+\betA_C)=\oVErline{\lambda}_{ij},\End{aliGnED}$$ wHERe tHe lAst equalitY fOllowS From the DEfINITioN Provided in the Proposition. hEncE, we get ThE deSIred upPer boUnD On tHe prices usiNg thE upper bouNd on thE Dual varIAbles. PrOof of PRopOsiTion \[PRoP:mOnoPrOFitCS\] {#aPp:pROfcSmono} ------------------------------------------ UsiNg asSumptIon \[aSS:SIGmalAmbDa\], we See thAt $\frac{(3-5\sigma)}{2}\leQ0$ anD $\fraC{(3-5\SigMa)}{2}{\ell_{\Max}}\leQ{\lamBdA_{ij}^{m}}\lEq\undeRset{i,J}{\mAx} ~\overline{\lambdA}_{ij}\lEq\frac{(3\sigMa-1)(3-\sIgMa)}{4(5-3\sIgMa)}\ell_{\MAx}\leq\fRac{3\SigMa-1}{2}{\ell_{\maX}}$. Hence, tHE opTiMAL PrIces fall in the regioN $[(1-\sIGMa){\Ell_{\max}},\siGma{\ell_{\MAx}})$, AnD Are given By: $$\LabEl{eq:OPTpricEsliNEaR} {\ell_{ij}^m}=\fRac{(1+\sigMA){\eLl_{\Max}}+{\lambDa_{Ij}^{m}}}{4}.$$ The DuAl pRobLem wiTH optImal prIces in caN be wrITten as: \[eq:dualwiTHoptimalpriceS\] $$\BeGIN{aLIgneD} \laBel{eq:dualobJectIVewiThopTImAlpRIces} &\

finally, the constraint : $$\begin{ align ed} \lambda_{i j }&\l eq\beta \tau_{ij}+\mu_ j^{e- e_ { ij}} - \m u_i^e \\ & \o v e rse t{ \e qre f{ e q: dualc ons traint6 }}{\leq}\b eta _t \tau_{ij}+e _ {i j}P_j+\mu_ j^e -\mu_i^e\\ &\over se t{\ e qref{ eq: dualc onstra i nt9}}{ \leq}\bet a_ t \ t au_{ij} + e _{ ij}P _j+\beta \tau_{ji } +e _ {ji}P_i.\end{a ligned }$ $ R e p lac ing $P_i=p_i+ \b eta_c $ and re a rr a n g ing the terms: $$ \begin{alig n ed} \ la bel { eq:lam bdaup pe r bou nd} \la mbda _{ij}\leq \beta _ t(\tau_ { ij}+\ta u_{ji} )+e _{i j}(p _ j+ \b eta _c ) +e_ { ji }(p _ i+\ beta_c)= \o ve rline {\la m b d a }_{i j}, \end {alig ned}$$ wherethe las t eq ualit y fol lows f rom t he def initi on provided in th e pr oposition . H en ce, w e get the de sir edupper b ound on the p r i c es using the upper b ou n d o n the du al var i ab le s . Proof o f P ropo s i tion\[pr o p: monoprof itcs\] {# ap p:profc sm ono} - -- --- --- ----- - ---- ------ -------- ----- - --- Using Ass u mption \[ass: s ig m a la m bda\ ],we see that $\f r ac{( 3-5\ s ig ma) } {2}\l eq0$an d $ \ frac{(3-5\sigma)}{2 }{ \ell_{ \max} }\leq{\lambda _{ij}^{m}} \ l e q\unders et{i , j} { \max} ~\overli ne{\l ambda}_{ij } \leq\fra c{(3\ sigma-1) (3-\sigma ) } {4(5-3\s igm a)} \el l_{ \ m ax }\leq\frac{3\ s i gma- 1} {2}{\el l_{ \max}}$ . H enc e,the o ptimal pr ices fal lin t he re gion$ [(1-\sig ma ){\ el l_{ \max} } ,\sigm a{\el l_{\ ma x} } )$, and ar e g i v en b y: $ $\la bel {e q:opt pric e sli near} {\ell_ {ij } ^m}= \f ra c{(1+\s igma){\ell_{\ ma x}}+{\lamb da _{i j}^{m} } } {4}.$$ T he dual problem with op t imal pr ice s incanbe writte n a s: \[ eq: d ualwit hoptim alpri ce s\] $ $\beg i n {a lig ne d} \label{ e q :du alobj ec tive withopt imalprices} &\

finally,_the constraint_: $$\begin{aligned} _ __ __ _ _ \lambda_{ij}&\leq\beta \tau_{ij}+\mu_j^{e-e_{ij}}-\mu_i^e\\ _ &\overset{\eqref{eq:dualconstraint6}}{\leq}\beta_t_\tau_{ij}+e_{ij}P_j+\mu_j^e-\mu_i^e\\ _ &\overset{\eqref{eq:dualconstraint9}}{\leq}\beta_t \tau_{ij}+e_{ij}P_j+\beta \tau_{ji}+e_{ji}P_i.\end{aligned}$$ Replacing $P_i=p_i+\beta_c$ and rearranging the_terms:_$$\begin{aligned} ___\label{eq:lambdaupperbound} \lambda_{ij}\leq_\beta_t(\tau_{ij}+\tau_{ji})+e_{ij}(p_j+\beta_c)+e_{ji}(p_i+\beta_c)=\overline{\lambda}_{ij},\end{aligned}$$ where the last equality_follows from_the definition provided in the proposition. Hence, we_get_the desired upper_bound on the prices using the upper bound on_the dual variables. Proof of Proposition \[prop:monoprofitcs\] {#app:profcsmono} ------------------------------------------ Using_Assumption \[ass:sigmalambda\], we see_that_$\frac{(3-5\sigma)}{2}\leq0$_and $\frac{(3-5\sigma)}{2}{\ell_{\max}}\leq{\lambda_{ij}^{m}}\leq\underset{i,j}{\max} ~\overline{\lambda}_{ij}\leq\frac{(3\sigma-1)(3-\sigma)}{4(5-3\sigma)}\ell_{\max}\leq\frac{3\sigma-1}{2}{\ell_{\max}}$. Hence,_the optimal prices fall in the_region $[(1-\sigma){\ell_{\max}},\sigma{\ell_{\max}})$, and are given by:_$$\label{eq:optpriceslinear} {\ell_{ij}^m}=\frac{(1+\sigma){\ell_{\max}}+{\lambda_{ij}^{m}}}{4}.$$ The dual_problem with optimal prices in can_be written as: \[eq:dualwithoptimalprices\] $$\begin{aligned} \label{eq:dualobjectivewithoptimalprices} _ _&\

equation for the temperature is given by $$\begin{aligned} \label{LCThetanotOpened} {\cal L}^\Theta &= \frac{a^2}{q}\left[L[\Theta] - (1 + \Theta) \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda}\right] \notag\\ &= \Theta' + \Theta_{, i} n^{(i)} + \alpha_{, i} n^{(i)} - \beta_{i, j} n^{(i)} n^{(j)} + h_{i j}{}' n^{(i)} n^{(j)} \notag\\ & \qquad +\frac{a^2}{q} \left[ \left. \frac{{{\rm d}}\eta}{{{\rm d}}\lambda} \right|^{(1)} \Theta' + \left. \frac{{{\rm d}}x^i}{{{\rm d}}\lambda} \right|^{(1)} \Theta_{, i} + \left. \frac{{{\rm d}}n^{(i)}}{{{\rm d}}\lambda} \right|^{(1)} D_i \Theta - \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda} \right|^{(1) \times (1)} - \Theta \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda} \right|^{(1)} \right] = {{\cal C}}^\Theta \,.\end{aligned}$$ As for the spectral distortion, the Boltzmann equation is given, even at second order by $$\begin{aligned} {\cal L}^Y &= y' + y_{, i} n^{(i)} = {{\cal C}}^Y \,.\end{aligned}$$ This means simply that gravitational effects do not induce spectral distortions, and this result holds actually non-perturbatively. Before ending this subsection, for the sake of completeness, we shall write down the most general form of the second order Boltzmann equation for the intensity. Using Eqs. (\[eta-evo\]), (\[q-evo\]), (\[xi-evo\]) and (\[n-evo\]), the detailed form of the evolution equation for temperature obtained in Eq. (\[LCThetanotOpened\]) is $$\begin{aligned} {\cal L}^\Theta=& \Theta' + \Theta_{, i} n^{(i)} + \alpha_{,i} n^{(i

equation for the temperature is given by $ $ \begin{aligned } \label{LCThetanotOpened } { \cal L}^\Theta & = \frac{a^2}{q}\left[L[\Theta ] - (1 + \Theta) \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda}\right ] \notag\\ & = \Theta' + \Theta _ {, i } n^{(i) } + \alpha _ {, i } n^{(i) } - \beta_{i, j } n^{(i) } n^{(j) } + h_{i j } { }' n^{(i) } n^{(j) } \notag\\ & \qquad + \frac{a^2}{q } \left [ \left. \frac{{{\rm d}}\eta}{{{\rm d}}\lambda } \right|^{(1) } \Theta' + \left. \frac{{{\rm d}}x^i}{{{\rm d}}\lambda } \right|^{(1) } \Theta _ {, i } + \left. \frac{{{\rm d}}n^{(i)}}{{{\rm d}}\lambda } \right|^{(1) } D_i \Theta - \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda } \right|^{(1) \times (1) } - \Theta \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda } \right|^{(1) } \right ] = { { \cal C}}^\Theta \,.\end{aligned}$$ As for the spectral distortion, the Boltzmann equality is render, even at second order by $ $ \begin{aligned } { \cal L}^Y & = y' + y _ {, i } n^{(i) } = { { \cal C}}^Y \,.\end{aligned}$$ This means plainly that gravitational impression do not induce spectral distortions, and this consequence holds actually non - perturbatively. Before ending this subsection, for the sake of completeness, we shall publish down the most general shape of the second order Boltzmann equation for the saturation. use Eqs.   (\[eta - evo\ ]), (\[q - evo\ ]), (\[xi - evo\ ]) and (\[n - evo\ ]), the detailed form of the development equation for temperature prevail in Eq.   (\[LCThetanotOpened\ ]) is $ $ \begin{aligned } { \cal L}^\Theta= & \Theta' + \Theta _ {, i } n^{(i) } + \alpha_{,i } n^{(i

eqkation for the temperatuve is given by $$\bggun{aligied} \labem{LCThetavotOpened} {\cal L}^\Theta &= \frac{a^2}{q}\pedt[L[\Thtna] - (1 + \Theta) \frac{{{\rm d}}\lv q}{{{\rm d}}\lalbda}\righr] \nouag\\ &= \Theta' + \Theta_{, i} n^{(i)} + \alpha_{, i} k^{(i)} - \bzte_{i, j} n^{(i)} n^{(j)} + h_{i m}{}' n^{(i)} n^{(j)} \notdg\\ & \qquad +\fraw{a^2}{d} \peft[ \left. \frac{{{\rm d}}\eta}{{{\rm d}}\lambda} \rigrt|^{(1)} \Thets' + \left. \frac{{{\rm d}}v^i}{{{\rm q}}\lamgda} \right|^{(1)} \Theta_{, i} + \left. \frac{{{\rm d}}n^{(i)}}{{{\dm d}}\lamuda} \right|^{(1)} D_i \Thrta - \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\pambfa} \right|^{(1) \times (1)} - \Tjeta \left. \ftzc{{{\ri d}}\ln q}{{{\rm d}}\laobda} \right|^{(1)} \right] = {{\cam C}}^\Theta \,.\end{aligned}$$ As for the rpectxal distortuob, tjg Boltzmann xquatijn is given, cnen at vecond prder by $$\begin{slijned} {\cal L}^Y &= y' + y_{, i} n^{(i)} = {{\cel C}}^Y \,.\end{aligned}$$ Thif means shm'ly that gravitationao wffecjs do not undjce s'ecfral dlstkrtions, ans this resuot holds actually npn-[vtturbatively. Gefore egding this subsection, for the sake of cmmpmeteness, we shall write down the most generap form of the second order Boltzmann equation for the intetsity. Jsiun Dws. (\[fta-evo\]), (\[q-evo\]), (\[xi-evo\]) and (\[n-evo\]), the detailed form jr uhe evolution equction for temperstkrr obtained in Gq. (\[LCThecznktOpened\]) is $$\begin{apigned} {\sal L}^\Rheta=& \Theua' + \Tneta_{, i} n^{(i)} + \alpha_{,i} n^{(i

equation for the temperature is given by {\cal &= \frac{a^2}{q}\left[L[\Theta] (1 + \Theta) &= + \Theta_{, i} + \alpha_{, i} - \beta_{i, j} n^{(i)} n^{(j)} + j}{}' n^{(i)} n^{(j)} \notag\\ & \qquad +\frac{a^2}{q} \left[ \left. \frac{{{\rm d}}\eta}{{{\rm d}}\lambda} \right|^{(1)} + \left. \frac{{{\rm d}}x^i}{{{\rm d}}\lambda} \right|^{(1)} \Theta_{, i} + \left. \frac{{{\rm d}}n^{(i)}}{{{\rm d}}\lambda} D_i - \frac{{{\rm q}{{{\rm d}}\lambda} \right|^{(1) \times (1)} - \Theta \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda} \right|^{(1)} \right] = {{\cal \,.\end{aligned}$$ As for the spectral distortion, the Boltzmann is given, even at order by $$\begin{aligned} {\cal L}^Y y' y_{, i} = C}}^Y This means simply gravitational effects do not induce spectral distortions, and this result holds actually non-perturbatively. Before ending this subsection, the sake we shall down most form of the Boltzmann equation for the intensity. Using (\[xi-evo\]) and (\[n-evo\]), the detailed form of the equation for obtained in Eq. (\[LCThetanotOpened\]) is $$\begin{aligned} L}^\Theta=& \Theta' + \Theta_{, i} n^{(i)} + \alpha_{,i}

equation for the temperature Is given by $$\bEgin{aLigNed} \LaBel{LcTheTanotOpened} {\cal l}^\thetA &= \frac{a^2}{q}\left[L[\Theta] - (1 + \Theta) \Frac{{{\rM d}}\LN q}{{{\rm D}}\LaMbda}\rIght] \notAG\\ &= \THETa' + \THeTa_{, I} n^{(i)} + \AlPHa_{, I} n^{(i)} - \beTa_{i, J} n^{(i)} n^{(j)} + h_{i J}{}' n^{(i)} n^{(j)} \notag\\ & \QquAd +\Frac{a^2}{q} \left[ \leFT. \fRac{{{\rm d}}\eta}{{{\rM d}}\lAmbda} \right|^{(1)} \ThEta' + \Left. \frAc{{{\Rm d}}X^I}{{{\rm d}}\lAmbDa} \rigHt|^{(1)} \ThetA_{, I} + \left. \fRac{{{\rm d}}n^{(i)}}{{{\rM d}}\LAmbda} \rIGht|^{(1)} D_i \ThETA - \lEft. \fRac{{{\rm d}}\ln q}{{{\rm d}}\lambdA} \RiGHt|^{(1) \times (1)} - \Theta \leFt. \frac{{{\Rm D}}\Ln Q}{{{\RM d}}\lAmbDa} \right|^{(1)} \rigHt] = {{\Cal C}}^\THEta \,.\end{aLIgNED}$$ as fOR the spectral dIstortion, thE bolTzmann EqUatIOn is giVen, evEn AT seCond order by $$\BegiN{aligned} {\cAl L}^Y &= y' + y_{, I} N^{(i)} = {{\cal C}}^Y \,.\ENd{alignEd}$$ This MeaNs sImplY ThAt GraViTAtiONaL efFEctS do not inDuCe SpectRal dISTORtioNs, aNd thIs resUlt holds actuaLly Non-pERtuRbatiVely. BEforE eNding This suBsectIoN, for the sake of coMpleTeness, we sHalL wRitE dOwn thE Most geNerAl fOrm of thE second ORdeR BOLTZmAnn equation for the iNtENSiTy. Using EQs. (\[eta-eVO\]), (\[q-EvO\]), (\[Xi-evo\]) and (\[N-eVo\]), tHe deTAIled fOrm oF ThE evolutiOn equaTIoN fOr tempeRaTure obTaIneD in eq. (\[LCTHEtanOtOpenEd\]) is $$\begiN{aligNEd} {\cal L}^\Theta=& \TheTA' + \Theta_{, i} n^{(i)} + \alphA_{,I} n^{(I

equation for the temperat ure is giv en by $$ \be gi n{al igne d} \label{LCTh e tano tOpened} {\cal L}^\Th eta &= \fra c {a ^2}{q }\left[ L [\ T h eta ] - (1 + \T heta) \f rac{{{\ rm d}}\lnq}{ {{ \rm d}}\lamb d a} \right] \n ota g\\ &= \The ta' + \Th et a_{ , i} n ^{( i)} + \alp h a_{, i } n^{(i)} - \beta_ { i, j} n ^ { (i )} n ^{(j)} + h_{i j} { }' n^{(i)} n^{(j) } \not ag \ \& \q qua d +\frac {a ^2}{q } \left [ \ l e f t.\ frac{{{\rm d} }\eta}{{{\r m d} }\lamb da } \ r ight|^ {(1)} \ T het a' + \left . \f rac{{{\rm d}}x^ i }{{{\rm d}}\lam bda} \ rig ht| ^{(1 ) }\T het a_ { , i } + \ l eft . \frac{ {{ \r m d}} n^{( i ) } } {{{\ rmd}}\ lambd a} \right|^{( 1)} D_i \Th eta - \le ft.\f rac{{ {\rm d }}\ln q }{{{\rm d}}\lam bda} \right|^ {(1 )\ti me s (1) } - \T het a \ left. \ frac{{{ \ rmd} } \ l nq}{{{\rm d}}\lambd a} \ ri ght|^{(1 )} \ri g ht ] = {{\c al C} }^\T h e ta \, .\en d {a ligned}$ $ As f o rth e spect ra l dist or tio n,the B o ltzm ann eq uation i s giv e n, even at sec o nd order by $ $ \b e g in { alig ned } {\cal L} ^Y & = y'+ y_ { ,i}n ^{(i) } = { {\ c al C}}^Y \,.\end{align ed }$$ Th is me ans simply th at gravita t i o nal effe ctsd on ot induce spec traldistortion s , and th is re sult hol ds actual l y non-per tur bat ive ly. Be fore ending t h i s su bs ection, fo r the s ake of co mpl et eness, we shall w ri te d ow n t he mo s t genera lfor mofthe s e cond o rderBolt zm an n eq uationf or t he i nt en sity . U si ng Eq s. ( \ [et a-evo\] ), (\[q-e vo\ ] ), ( \[ xi -evo\]) and (\[n-evo \] ), the det ai led formo f the evo lution equation for tem p erature ob taine d in Eq. (\[L CTh etanot Ope n ed\])is $$\ begin {a lig n e d} { \ c al L} ^\ Theta=& \T h e ta' + \T he ta_{ , i} n^ {(i)} + \alpha_{,i } n^ {(i

equation_for the_temperature is given by_$$\begin{aligned} \label{LCThetanotOpened} {\cal_L}^\Theta _&= \frac{a^2}{q}\left[L[\Theta] _-_(1 + \Theta)_\frac{{{\rm d}}\ln q}{{{\rm_d}}\lambda}\right] \notag\\ &= \Theta'_+ \Theta_{, i}_n^{(i)} _+ \alpha_{, i} n^{(i)} - \beta_{i, j} n^{(i)} n^{(j)} + h_{i j}{}' n^{(i)} n^{(j)}_\notag\\ _& \qquad_ _+\frac{a^2}{q} _\left[ \left. \frac{{{\rm d}}\eta}{{{\rm d}}\lambda}_\right|^{(1)} \Theta' + \left. \frac{{{\rm_d}}x^i}{{{\rm d}}\lambda}_\right|^{(1)} \Theta_{, i} + \left. \frac{{{\rm d}}n^{(i)}}{{{\rm d}}\lambda}_\right|^{(1)}_D_i \Theta -_\left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda} \right|^{(1) \times (1)} -_\Theta \left. \frac{{{\rm d}}\ln q}{{{\rm d}}\lambda}_\right|^{(1)} \right] __=_{{\cal C}}^\Theta \,.\end{aligned}$$ As_for the spectral distortion, the Boltzmann_equation is given, even at second_order by $$\begin{aligned} {\cal L}^Y &= y'_+ y_{, i} n^{(i)} = {{\cal_C}}^Y \,.\end{aligned}$$ This means simply_that gravitational_effects do not induce spectral_distortions, and this_result holds_actually non-perturbatively. Before ending_this subsection, for the sake of_completeness, we shall_write down the most general form_of_the second order_Boltzmann_equation_for the_intensity. Using Eqs. (\[eta-evo\]),_(\[q-evo\]),_(\[xi-evo\]) and_(\[n-evo\]),_the detailed form of the evolution_equation_for temperature obtained in Eq. (\[LCThetanotOpened\]) is $$\begin{aligned} _{\cal L}^\Theta=& \Theta' +_\Theta_{,_i} n^{(i)} + \alpha_{,i}_n^{(i

of active galactic nuclei and quasars. The text basically corresponds to \[17\]. We have added some figures and corrected typos in formulas. For an extended discussion of this subject see \[18\]. The authors are sincerely grateful to N.N. Kizilova for the useful notes. **References** 1\. V.S. Beskin. Axisymmetric steady flows in astrophysics, M.: Fizmatlit, 2006, in Russian; MHD Flows in Compact Astrophysical Objects, Springer, 2010, 425 pp. 2\. G. Bisnovatyi-Kogan & A. Ruzmaikin, Ap & Space Sci, 42, 401 (1976). 3\. R.D. Blandford, MNRAS, 176, 465 (1976). 4\. R.V.E. Lovelace, Nature, 262, 649 (1976). 5\. D. Lynden-Bell, Mon. Not. R. Astron. Soc. 369, 1167 (2006). 6\. R.D. Blandford, Phil.Trans.R.Soc.Lond. A, 358, 811 (2000). 7\. I.F. Mirabel, Phil.Trans.R.Soc.Lond. A, 358, 841 (2000). 8\. M.G. Abrahamian, Astrofizika, 51, 201, 431, 617 (2008). 9\. G.V. Vermeulen & M.H. Cohen, ApJ, 430, 467 (1994). 10\. E.Yu. Bannikova & V.M. Kontorovich, Astron. J., 84, 298 (2007), astro-ph/0707.1478. 11\. E.Yu. Bannikova, V.M. Kontorovich & G.M. Resnick, JETP, 132, ¹ 3, 615 (2007). 12\. E.Yu. Bannikova & V.M. Kontorovich, Phys.Lett. A, 373, 1856 (2009). 13\. G.M. Reznik, J. Fluid. Mech. 240, 405 (

of active galactic nuclei and quasars. The text basically correspond to \[17\ ]. We have lend some figures and corrected typos in recipe. For an extended discussion of this subject experience \[18\ ]. The generator are sincerely grateful to N.N.   Kizilova for the useful note. * * References * * 1\. V.S.   Beskin. Axisymmetric steady flows in astrophysics, M.: Fizmatlit, 2006, in Russian; MHD Flows in Compact Astrophysical Objects, Springer, 2010, 425 pp. 2\. G.   Bisnovatyi - Kogan & A.   Ruzmaikin, Ap & Space Sci, 42, 401 (1976). 3\. R.D.   Blandford, MNRAS, 176, 465 (1976). 4\. R.V.E.   Lovelace, Nature, 262, 649 (1976). 5\. D.   Lynden - Bell, Mon. Not. R. Astron. Soc. 369, 1167 (2006). 6\. R.D.   Blandford, Phil. Trans. R.Soc. Lond. adenine, 358, 811 (2000). 7\. I.F.   Mirabel, Phil. Trans. R.Soc. Lond. A, 358, 841 (2000). 8\. M.G.   Abrahamian, Astrofizika, 51, 201, 431, 617 (2008). 9\. G.V.   Vermeulen & M.H.   Cohen, ApJ, 430, 467 (1994). 10\. E.Yu.   Bannikova & V.M.   Kontorovich, Astron. J., 84, 298 (2007), astro - ph/0707.1478. 11\. E.Yu.   Bannikova, V.M.   Kontorovich & G.M.   Resnick, JETP, 132, ¹ 3, 615 (2007). 12\. E.Yu.   Bannikova & V.M.   Kontorovich, Phys. Lett. A, 373, 1856 (2009). 13\. G.M.   Reznik, J. Fluid. Mech. 240, 405 (

of active galactic nuclei xnd quasars. The jezt basmcally dorrespovds to \[17\]. We have added some fmgurws ane corrected typos in furmulas. Fog an extebded eiscussion of this subjedb see \[18\]. Vhe authors are sincerely grateful to N.T. Kkznlova for the useful notes. **References** 1\. V.S. Beskon. Axisymmetric fteacr flkws in astrophysics, M.: Fizmatlit, 2006, ih Russien; MHD Flows in Compact Astrophysical Objfcts, Springer, 2010, 425 pp. 2\. G. Bidnovatyi-Kogqn & W. Euzmaikin, Ap & Space Scp, 42, 401 (1976). 3\. R.D. Blansford, MNRAS, 176, 465 (1976). 4\. R.V.E. Lovelace, Natufe, 262, 649 (1976). 5\. D. Lynden-Beol, Moj. Not. R. Astrin. Sob. 369, 1167 (2006). 6\. R.D. Blandford, Phil.Drans.R.Spc.Lond. A, 358, 811 (2000). 7\. I.F. Kirebel, Phil.Trans.R.Soc.Lond. A, 358, 841 (2000). 8\. M.G. Abrahamian, Asjrofizika, 51, 201, 431, 617 (2008). 9\. G.V. Vermeulen & M.Y. Cihen, DpJ, 430, 467 (1994). 10\. E.Hy. Bavniiota & V.M. Konhorkvich, Astrkn. J., 84, 298 (2007), asteo-ph/0707.1478. 11\. E.Yu. Bannikova, V.K. Kjbtorovich & G.M. Desnich, TETP, 132, ¹ 3, 615 (2007). 12\. E.Yu. Bannikova & V.M. Kontorovich, Pvys.Mett. A, 373, 1856 (2009). 13\. G.M. Reznik, J. Flyid. Mech. 240, 405 (

of active galactic nuclei and quasars. The corresponds \[17\]. We added some figures For extended discussion of subject see \[18\]. authors are sincerely grateful to N.N. for the useful notes. **References** 1\. V.S. Beskin. Axisymmetric steady flows in astrophysics, Fizmatlit, 2006, in Russian; MHD Flows in Compact Astrophysical Objects, Springer, 2010, 425 2\. Bisnovatyi-Kogan A. Ap & Space Sci, 42, 401 (1976). 3\. R.D. Blandford, MNRAS, 176, 465 (1976). 4\. R.V.E. Nature, 262, 649 (1976). 5\. D. Lynden-Bell, Mon. R. Astron. Soc. 369, (2006). 6\. R.D. Blandford, Phil.Trans.R.Soc.Lond. 358, (2000). 7\. Mirabel, A, 841 (2000). 8\. Abrahamian, Astrofizika, 51, 201, 431, 617 (2008). 9\. G.V. Vermeulen & M.H. Cohen, ApJ, 430, 467 (1994). E.Yu. Bannikova Kontorovich, Astron. 84, (2007), 11\. E.Yu. Bannikova, & G.M. Resnick, JETP, 132, ¹ 12\. E.Yu. Bannikova & V.M. Kontorovich, Phys.Lett. A, 1856 (2009). G.M. Reznik, J. Fluid. Mech. 240, (

of active galactic nuclei and Quasars. The Text bAsiCalLy CorrEspoNds to \[17\]. We have addED somE figures and corrected tyPos in FoRMulaS. foR an exTended dIScUSSioN oF tHis SuBJeCt see \[18\]. the Authors Are sincereLy gRaTeful to N.N. KizILoVa for the usEfuL notes. **RefereNceS** 1\. V.S. BesKiN. AxISymmeTriC steaDy flowS In astrOphysics, M.: fiZMatlit, 2006, IN RussiaN; mhD flowS in Compact AstrophYSiCAl Objects, SprinGer, 2010, 425 pp. 2\. G. biSNoVATyi-kogAn & A. RuzmaikIn, ap & SpaCE Sci, 42, 401 (1976). 3\. R.D. BLAnDFORd, MnrAS, 176, 465 (1976). 4\. R.V.E. LovelacE, Nature, 262, 649 (1976). 5\. D. LynDEn-BEll, Mon. noT. R. ASTron. SoC. 369, 1167 (2006). 6\. R.D. BlAnDForD, Phil.Trans.R.soc.LOnd. A, 358, 811 (2000). 7\. I.F. MirAbel, PhIL.Trans.R.sOc.Lond. A, 358, 841 (2000). 8\. m.G. AbraHamIan, astrOFiZiKa, 51, 201, 431, 617 (2008). 9\. G.v. VERmeULeN & M.H. cOheN, ApJ, 430, 467 (1994). 10\. E.Yu. BAnNiKova & V.m. KonTOROVich, astRon. J., 84, 298 (2007), Astro-Ph/0707.1478. 11\. E.Yu. BannikovA, V.M. kontORovIch & G.M. resniCk, JEtP, 132, ¹ 3, 615 (2007). 12\. e.Yu. BaNnikovA & V.M. KoNtOrovich, Phys.Lett. a, 373, 1856 (2009). 13\. G.M. REznik, J. FluId. MEcH. 240, 405 (

of active galactic nuclei and quasa rs. The te xt bas ical ly corresponds to \ [17\]. We have added s ome f ig u resa nd corr ected t y po s info rm ula s. F or an ex tendeddiscussion of t his subjects ee \[18\]. The authors are si ncerel ygra t efultoN.N.Kizilo v a forthe usefu ln otes.**Refer e n ce s** 1\. V.S. Beskin. Ax i symmetric stea dy flo ws in a str oph ysics, M.: F izmat l it, 200 6 ,i n Rus s ian; MHD Flow s in Compac t As trophy si cal Object s, Sp ri n ger , 2010, 425 pp. 2\. G.Bisnov a tyi-Kog a n & A.Ruzmai kin , A p &S pa ce Sc i, 42, 40 1 ( 1 976 ). 3\.R. D.  Blan dfor d , M NRAS , 1 76,465 ( 1976). 4\. R .V. E. L o vel ace,Natur e, 2 62 , 649 (1976 ). 5 \. D. Lynden-Bell , Mo n. Not. R . A st ron .Soc.3 69, 11 67(20 06). 6 \. R.D. Bla nd f o r d, Phil.Trans.R.Soc. Lo n d .A, 358,811 (2 0 00 ). 7\. I.F .Mir abel , Phil. Tran s .R .Soc.Lon d. A,3 58 ,841 (20 00 ). 8\ .M.G . A braha m ian, Astro fizika,51, 2 0 1, 431, 617 (2 0 08). 9\. G.V .  V e r me u len& M .H. Cohen,ApJ, 430, 467 (1 994 ) . 10 \. E. Yu .  B a nnikova & V.M. Kont or ovich, Astr on. J., 84, 2 98 (2007), a s tro-ph/0 707. 1 47 8 . 11\. E.Yu.Banni kova, V.M. Kontorov ich & G.M. Re snick, JE T P , 132, ¹ 3, 61 5 ( 200 7 ) . 12\. E.Yu. B a n niko va & V.M.  Ko ntorovi ch, Ph ys. Let t. A, 373,1856 (20 09 ). 13 \.G.M.R eznik, J .Flu id . M ech.2 40, 40 5 (

of_active galactic_nuclei and quasars. The text_basically corresponds_to_\[17\]. We_have_added some figures_and corrected typos_in formulas. For an extended_discussion of this_subject_see \[18\]. The authors are sincerely grateful to N.N. Kizilova for the useful notes. **References** 1\. V.S. Beskin. Axisymmetric_steady_flows in_astrophysics,_M.:_Fizmatlit, 2006, in Russian; MHD_Flows in Compact Astrophysical Objects,_Springer, 2010,_425 pp. 2\. G. Bisnovatyi-Kogan & A. Ruzmaikin, Ap & Space_Sci,_42, 401 (1976). 3\._R.D. Blandford, MNRAS, 176, 465 (1976). 4\. R.V.E. Lovelace, Nature, 262, 649_(1976). 5\. D. Lynden-Bell, Mon. Not. R. Astron._Soc. 369, 1167_(2006). 6\._R.D. Blandford,_Phil.Trans.R.Soc.Lond. A, 358, 811_(2000). 7\. I.F. Mirabel, Phil.Trans.R.Soc.Lond. A, 358, 841_(2000). 8\. M.G. Abrahamian, Astrofizika, 51, 201, 431,_617 (2008). 9\. G.V. Vermeulen & M.H. Cohen, ApJ, 430,_467 (1994). 10\. E.Yu. Bannikova & V.M. Kontorovich, Astron._J., 84, 298 (2007), astro-ph/0707.1478. 11\._E.Yu. Bannikova, V.M. Kontorovich_& G.M. Resnick, JETP, 132, ¹_3, 615 (2007). 12\._E.Yu. Bannikova &_V.M. Kontorovich, Phys.Lett. A,_373, 1856 (2009). 13\. G.M. Reznik, J. Fluid._Mech. 240, 405_(

envelopes. Correcting for stars not detected because of the low spatial resolution of ISOCAM, dusty envelopes are inferred around about 15% of the brightest giants ($M_{\rm bol}\le-2.5$). The inferred mass loss rates are in the range $10^{-7} < dM/dt < 10^{-6} M_{\odot}\,{\rm yr^{-1}}$, assuming as reference values an outflow velocity of 14 ${\rm km\,s^{-1}}$ and a gas-to-dust mass ratio of 200 at the metallicity of 47 Tuc and for a stellar luminosity of 1000 $L_{\odot}$. The major astrophysical implications are: - The mass loss occurs very near the RGB-tip and is episodic. - The mass loss episodes must last longer than a few decades and less than a million years. - There is no indication for a strict metallicity dependence of the frequency of mass loss occurrence. The research has been supported by the Agenzia Spaziale Italiana through the grants 9920015AS and I/R/27/00. RTR is partially supported by NASA LTSA grant NAG 5-6403 and STScI grant GO-8709. We thank Jacco van Loon, the Referee of our paper, for the many suggestions and comments which improved the overall presentation and analysis of our results. We thank M. Bellazzini, G. Bono, C. Cacciari, M. Catelan, Andea Dupree, Jeff Linsky, and P. Persi for helpful comments and discussions. Angeletti, L., Blanco, A., Bussoletti, E., Capuzzo-Dolcetta, R., & Giannone, P. 1982,, 199, 441 Ayres, T. R., Brown, A., Harper, G. M., Bennett, P. D., Linsky, J. L., Carpenter, K. G., & Robinson, R. D. 1997,, 491, 876 Beichman, C. A., Neugebauer, G., Habing, H. J., Clegg, P.E., & Chester, T. J. 1985, [IRAS Catalogs and Atlases Explanatory Supplement]{}, (Washington: U.S. GPO) Buonanno, R., Buscema, G

envelopes. Correcting for stars not detected because of the depleted spatial settlement of ISOCAM, dusty envelopes are inferred about about 15% of the brightest giants ($ M_{\rm bol}\le-2.5 $). The guess multitude loss rates are in the range $ 10^{-7 } < dM / dt < 10^{-6 } M_{\odot}\,{\rm yr^{-1}}$, assuming as reference book values an outflow speed of 14   $ { \rm km\,s^{-1}}$ and a gas - to - dust mass ratio of 200 at the metallicity of 47   Tuc and for a stellar luminosity of 1000   $ L_{\odot}$. The major astrophysical implications are: - The mass personnel casualty occurs very near the RGB - tip and is episodic. - The mass passing episodes must last longer than a few decades and less than a million years. - There be no indication for a strict metallicity dependence of the frequency of mass loss happening. The research has been supported by the Agenzia Spaziale Italiana through the grants 9920015AS and I / R/27/00. RTR is partially supported by NASA LTSA grant NAG 5 - 6403 and STScI grant GO-8709. We thank Jacco van Loon, the Referee of our paper, for the many suggestion and remark which improved the overall presentation and analysis of our results. We thank M. Bellazzini, G. Bono, C. Cacciari, M. Catelan, Andea Dupree, Jeff Linsky, and P. Persi for helpful comments and discussions. Angeletti, L., Blanco, A., Bussoletti, E., Capuzzo - Dolcetta, R., & Giannone, P. 1982, , 199, 441 Ayres, T. R., Brown, A., Harper, G. M., Bennett, P. D., Linsky, J. L., Carpenter, K. G., & Robinson, R. D. 1997, , 491, 876 Beichman, C. A., Neugebauer, G., Habing, H. J., Clegg, P.E., & Chester, T. J. 1985, [ IRAS Catalogs and Atlases Explanatory Supplement ] { }, (Washington: U.S. GPO) Buonanno, R., Buscema, G

engelopes. Correcting for suars not detected becausx of ths low spxtial resolution of ISOCAM, dnsty enveoopes are inferred arojnd about 15% of the brijhtest giants ($M_{\rm bol}\le-2.5$). Tmz infsvred kess loss rates sre in the range $10^{-7} < dM/dt < 10^{-6} M_{\udlt}\,{\rm yr^{-1}}$, assuming as reference valuef an ouyfpow velocity os 14 ${\rm hm\,s^{-1}}$ znd a gas-to-dust mass ratio of 200 at fhe metellicity of 47 Tuc and for a stellar luminoslty lf 1000 $L_{\odot}$. The major wstrophysicql iiplications afe: - The mass loss occors very near the RGB-tip and is dpisobic. - The mqsw llvs episodes must last longer than a xew decsdes and less bhan e miolion years. - There iv no indication fot a strict mztallicity dependence od the frexuenzt ow mzsx moss ofcucrence. The rssearch has been supported by uhe Qgenzia Spazizle Itwlyana through the grants 9920015AS and I/R/27/00. RTR iv pzrtially supported by NQSA LTSA grant NAG 5-6403 ajd STScI drant GO-8709. We thank Jacco van Loon, the Referee of ogr pa'ef, fir gye many suggestions and comments which improved fht onerall presentatijn and analusls jf our resultr. We tkzni M. Bellazzini, G. Blno, C. Cwcciaei, M. Catejan, Sndea Dupree, Jeff Linsky, ane P. Persi fog heopful comments and discussionr. Anggletti, L., Blanco, A., Bussoletti, Z., Capuazo-Dolcetta, R., & Giannkve, P. 1982,, 199, 441 Ayres, T. R., Brpwt, A., Harptf, G. M., Bennett, P. D., Linsky, J. L., Ccrpenter, K. G., & Robigson, R. D. 1997,, 491, 876 Beicmkan, C. A., Neugebauer, G., Haying, V. J., Clegg, O.E., & Chester, T. J. 1985, [IRAS Catalogs eid Atlases Exklatatmry Suppjemenb]{}, (Washington: U.S. GPO) Buonanno, R., Buscemc, G

envelopes. Correcting for stars not detected because low resolution of dusty envelopes are the giants ($M_{\rm bol}\le-2.5$). inferred mass loss are in the range $10^{-7} < < 10^{-6} M_{\odot}\,{\rm yr^{-1}}$, assuming as reference values an outflow velocity of 14 km\,s^{-1}}$ and a gas-to-dust mass ratio of 200 at the metallicity of 47 and a luminosity 1000 $L_{\odot}$. The major astrophysical implications are: - The mass loss occurs very near the RGB-tip is episodic. - The mass loss episodes must longer than a few and less than a million - is no for strict dependence of the of mass loss occurrence. The research has been supported by the Agenzia Spaziale Italiana through the grants and I/R/27/00. partially supported NASA grant 5-6403 and STScI We thank Jacco van Loon, the paper, for the many suggestions and comments which the overall and analysis of our results. We M. Bellazzini, G. Bono, C. Cacciari, M. Catelan, Dupree, Jeff Linsky, and P. Persi for helpful comments and discussions. Angeletti, L., Blanco, A., Capuzzo-Dolcetta, R., & Giannone, 1982,, 199, 441 T. Brown, Harper, M., Bennett, D., Linsky, J. L., Carpenter, K. G., & Robinson, R. D. 491, 876 Beichman, C. A., Neugebauer, G., Habing, H. J., & T. J. 1985, Catalogs and Atlases Explanatory (Washington: GPO) Buonanno, R., Buscema,

envelopes. Correcting for staRs not detecTed beCauSe oF tHe loW spaTial resolution OF ISOcAM, dusty envelopes are inFerreD aROund ABoUt 15% of tHe brighTEsT GIanTs ($m_{\rM boL}\lE-2.5$). thE infeRreD mass loSs rates are In tHe Range $10^{-7} < dM/dt < 10^{-6} M_{\odOT}\,{\rM yr^{-1}}$, assuminG as Reference valUes An outfLoW veLOcity Of 14 ${\rM km\,s^{-1}}$ aNd a gas-TO-dust mAss ratio oF 200 aT The metALlicity OF 47 tuC and For a stellar luminoSItY Of 1000 $L_{\odot}$. The majoR astroPhYSiCAL imPliCations are: - thE mass LOss occuRS vERY NeaR The RGB-tip and iS episodic. - ThE MasS loss ePiSodES must lAst loNgER thAn a few decadEs anD less than A milliON years. - THEre is no IndicaTioN foR a stRIcT mEtaLlICitY DePenDEncE of the frEqUeNcy of Mass LOSS OccuRreNce. THe resEarch has been sUppOrteD By tHe AgeNzia SPaziAlE ItalIana thRough ThE grants 9920015AS and I/R/27/00. RtR is Partially SupPoRteD bY NASA ltSA graNt NaG 5-6403 aNd STScI Grant GO-8709. wE thAnK jACcO van Loon, the Referee Of OUR pAper, for tHe many SUgGeSTions and CoMmeNts wHICh impRoveD ThE overall PresenTAtIoN and anaLySis of oUr ResUltS. We thANk M. BEllazzIni, G. Bono, c. CaccIAri, M. Catelan, AndEA Dupree, Jeff LiNSkY, ANd p. persI foR helpful comMentS And dIscuSSiOns. aNgeleTti, L., BLaNCo, a., bussoletti, E., Capuzzo-DOlCetta, R., & giannOne, P. 1982,, 199, 441 Ayres, T. R., BrOwn, A., Harper, g. m., bEnnett, P. D., linsKY, J. l., carpenter, K. G., & RobInson, r. D. 1997,, 491, 876 Beichman, c. a., NeugebaUer, G., HAbing, H. J., CLegg, P.E., & CheSTEr, T. J. 1985, [IRAS catAloGs aNd ATLAsEs Explanatory sUPpleMeNt]{}, (WashiNgtOn: U.S. GPO) buoNanNo, R., busCeMa, G

envelopes. Correcting for stars not dete cte d b ec ause ofthe low spatia l res olution of ISOCAM, dus ty en ve l opes ar e inf erred a r ou n d ab ou t15% o f t he br igh test gi ants ($M_{ \rm b ol}\le-2.5$) . The inferr edmass loss ra tes are i nthe range $1 0^{-7 } < dM / dt < 1 0^{-6} M_ {\ o dot}\, { \rm yr^ { - 1} }$,assuming as refer e nc e values an out flow v el o ci t y of 14  ${\rm km\ ,s ^{-1} } $ and a ga s - t o-d u st mass ratio of 200 att hemetall ic ity of 47Tuc a nd for a stellarlumi nosity of 1000$ L_{\odo t }$. Th e majo r a str ophy s ic al im pl i cat i on s a r e: - The m as s los s oc c u r s ver y n earthe R GB-tip and is ep isod i c. - The m asslo ss ep isodes must l ast longer than a f ew decade s a nd le ss than a mill ion ye ars. - Ther e is n o i nd ication for a stri ct m et allicity depen d en ce of the f re que ncyo f mass los s o ccurrenc e. Th e r es earch h as beensu ppo rte d byt he A genzia Spazial e Ita l iana through t h e grants 9920 0 15 A S a n d I/ R/2 7/00. RTR i s pa r tial ly s u pp ort e d byNASALT S Ag rant NAG 5-6403 and S TScI g rantGO-8709. We t hank Jacco v a n Loon,theR ef e ree of our pap er, f or the man y suggest ionsand comm ents whic h improved th e o ver all p re sentation and a naly si s of ou r r esults. We th ank M. B ellazzini , G. Bon o, C .Ca cci ari,M . Catela n, An de a D upree , JeffLinsk y, a nd P . Pe rsi for he l p fulco mm ents an ddiscu ssio n s. Angele tti, L.,Bla n co,A. ,Bussole tti, E., Capu zz o-Dolcetta ,R., & Gia n n one, P.1982,, 199, 441 Ayres, T. R.,Bro wn, A ., H arper, G. M. , Benn ett , P. D. , Lins ky, J .L., C arpen t e r, K. G ., & Robin s o n,R. D. 1 997, , 491,876 Beichman, C.A .,Neugebauer, G .,Habi n g ,H.J ., Cle gg , P. E . , & Chester, T. J. 1985,[I R AS Catalogsa ndAt lases E xplanat ory S u pplemen t]{}, (Wa shington: U .S.G P O) Buonanno, R., Bus cema, G

envelopes._Correcting for_stars not detected because_of the_low_spatial resolution_of_ISOCAM, dusty envelopes_are inferred around_about 15% of the_brightest giants ($M_{\rm_bol}\le-2.5$). The_inferred mass loss rates are in the range $10^{-7} < dM/dt < 10^{-6} M_{\odot}\,{\rm yr^{-1}}$,_assuming_as reference_values_an_outflow velocity of 14 ${\rm km\,s^{-1}}$_and a gas-to-dust mass ratio_of 200_at the metallicity of 47 Tuc and for a_stellar_luminosity of 1000 $L_{\odot}$. The_major astrophysical implications are: - The mass loss_occurs very near the RGB-tip and_is episodic. - __The_mass loss episodes must_last longer than a few decades_and less than a million years. -_ There is no indication for_a strict metallicity dependence of the_frequency of mass loss occurrence. The_research has_been supported by the Agenzia_Spaziale Italiana through_the grants_9920015AS and I/R/27/00._RTR is partially supported by NASA_LTSA grant NAG_5-6403 and STScI grant GO-8709. We_thank_Jacco van Loon,_the_Referee_of our_paper, for the_many_suggestions and_comments_which improved the overall presentation and_analysis_of our results. We thank M. Bellazzini,_G. Bono, C. Cacciari,_M._Catelan, Andea Dupree, Jeff_Linsky, and P. Persi for_helpful comments and discussions. Angeletti, L., Blanco,_A., Bussoletti,_E., Capuzzo-Dolcetta,_R., & Giannone, P. 1982,, 199, 441 Ayres, T. R., Brown, A.,_Harper, G. M., Bennett, P. D.,_Linsky, J. L., Carpenter,_K. G.,_&_Robinson, R. D._1997,,_491, 876 Beichman,_C. A., Neugebauer, G., Habing, H. J.,_Clegg, P.E.,_& Chester, T. J. 1985, [IRAS_Catalogs and Atlases Explanatory_Supplement]{},_(Washington: U.S. GPO) Buonanno, R., Buscema, G

in any theory in which fluctuations in more than one field are important, and it is possible to construct situations where this is the case [@twoperts]. Our analysis is based on the formalism devised by Stewart and Lyth [@SL] to carry out a second-order analytic calculation. We shall not reproduce the derivation of the equations here, instead simply reproducing the important ones.[^1] With the usual notation of $a$ for scale factor, $H$ for Hubble parameter, $\phi$ for the inflationary scalar field and overdot as derivative with respect to cosmic time, they introduce a new quantity $z$ defined by $$\label{z} z\equiv \frac{a\dot{\phi}}{H} \,.$$ The quantity one desires to calculate is the curvature perturbation ${\cal R}$, defined as in Ref. [@SL]. It is convenient to define a related quantity $u$, defined by $$\label{uR} u = - z {\cal R} \,,$$ which is a gauge invariant potential [@MFB]. A Fourier expansion of $u$ into comoving modes $u_{{\bf k}}$ is carried out, and these can be shown to obey the remarkably simple equation [@Muk; @MFB; @SL] $$\label{ufield} \frac{d^2 u_{{\bf k}}}{d\tau^2} +\left( k^2 - \frac{1}{z} \frac{d^2 z}{d\tau^2} \right) u_{{\bf k}} =0 \,,$$ where $\tau$ is the conformal time defined by $d\tau = dt/a(t)$ and $k = |{\bf k}|$ is the modulus of the wavenumber. Defining the spectrum ${\cal P}_{{\cal R}}(k)$ of the curvature perturbation in the standard way as $$\langle {\cal R}_{{\bf k_1}} {\cal R}^*_{{\bf k_2}} \rangle = \frac{2\pi^2}{k^3} {\cal P}_{{\cal R}} \delta^{3} \, ({\bf k_1}-{\bf k_2}) \,,$$ yields $$\label{pspec} {\cal P}_{{\cal R}}^{1/2}(k) = \sqrt{\frac{k^3}{2\pi^2}} \, \left| \frac{u_{{\bf k}}}{z}

in any theory in which fluctuations in more than one field are important, and it is potential to manufacture situations where this is the case [ @twoperts ]. Our psychoanalysis is based on the formalism organize by Stewart and Lyth [ @SL ] to carry out a second - holy order analytic calculation. We shall not reproduce the derivation of the equation here, instead simply regurgitate the important ones.[^1 ] With the usual note of $ a$ for scale factor, $ H$ for Hubble parameter, $ \phi$ for the inflationary scalar field and overdot as derivative with respect to cosmic time, they introduce a fresh quantity $ z$ defined by $ $ \label{z } z\equiv \frac{a\dot{\phi}}{H } \,.$$ The quantity one desire to calculate is the curvature perturbation $ { \cal R}$, defined as in Ref.   [ @SL ]. It is commodious to define a related measure $ u$, defined by $ $ \label{uR } u = - z { \cal R } \,,$$ which is a gauge changeless potential [ @MFB ]. A Fourier expansion of $ u$ into comoving modes $ u_{{\bf k}}$ is carried out, and these can be shown to obey the remarkably simple equation [ @Muk; @MFB; @SL ] $ $ \label{ufield } \frac{d^2 u_{{\bf k}}}{d\tau^2 } + \left (k^2 - \frac{1}{z } \frac{d^2 z}{d\tau^2 } \right) u_{{\bf k } } = 0 \,,$$ where $ \tau$ is the conformal time defined by $ d\tau = dt / a(t)$ and $ k = |{\bf k}|$ is the modulus of the wavenumber. Defining the spectrum $ { \cal P}_{{\cal R}}(k)$ of the curvature perturbation in the standard way as $ $ \langle { \cal R}_{{\bf k_1 } } { \cal R}^*_{{\bf k_2 } } \rangle = \frac{2\pi^2}{k^3 } { \cal P}_{{\cal R } } \delta^{3 } \, ({ \bf k_1}-{\bf k_2 }) \,,$$ yields $ $ \label{pspec } { \cal P}_{{\cal R}}^{1/2}(k) = \sqrt{\frac{k^3}{2\pi^2 } } \, \left| \frac{u_{{\bf k}}}{z }

in any theory in which flugtuations in morg rhan oie fiels are imoortant, and it is possible tl xonsteuct situations where ghis is tje case [@rwoptrts]. Our analysis ma based on ths forkelism devised bi Stewart ang Lyth [@SL] to cdrfy out a second-order analytic calculaeion. We sjall not reproquce ehe svrlvation of the equations here, inatead spmply reproducing the important ones.[^1] With tje udual notation of $a$ for scale dactje, $H$ for Hubbue paramettr, $\phi$ for tge inflationary scalar field ana ovexdot as deruvqtigg with respert to bosmic time, tmvy intrmduce a new quantity $d$ defmned by $$\label{z} z\equiv \frac{e\dot{\phi}}{H} \,.$$ The quantijy one deshrzs to calculate is thw xurvajure [ertjebagioh ${\ral R}$, deflnev as in Ref. [@AL]. It is cobvenient to define s wvkated quantify $u$, dqfyned by $$\label{uR} u = - z {\cal R} \,,$$ which is a caufe invariant potential [@NFB]. A Fourier expansioj of $u$ ineo comoving modes $u_{{\bf k}}$ is carried out, and these wan bx rhock gi lbey the remarkably simple equation [@Muk; @MFB; @SJ] $$\lsbvl{ufield} \frac{d^2 u_{{\bf k}}}{d\tau^2} +\left( k^2 - \gtac{1}{z} \frac{d^2 z}{d\tau^2} \rifht) u_{{\bf k}} =0 \,,$$ where $\hau$ is jhe cobformal tyme cefined by $d\tau = dt/a(t)$ and $k = |{\bf k}|$ is thv moeulus of the wavenbmber. Definiug the specyrum ${\cal P}_{{\cal R}}(k)$ of the curbature pertkrbation jv the standard wxy ss $$\langle {\cal R}_{{\bf k_1}} {\cal R}^*_{{\bf h_2}} \rangle = \yrac{2\pi^2}{k^3} {\zal K}_{{\cal R}} \qelta^{3} \, ({\hf k_1}-{\bn k_2}) \,,$$ yields $$\label{psoec} {\cap [}_{{\cal R}}^{1/2}(k) = \sert{\frac{k^3}{2\pi^2}} \, \left| \frac{u_{{\bf k}}}{z}

in any theory in which fluctuations in one are important, it is possible is case [@twoperts]. Our is based on formalism devised by Stewart and Lyth to carry out a second-order analytic calculation. We shall not reproduce the derivation the equations here, instead simply reproducing the important ones.[^1] With the usual notation $a$ scale $H$ Hubble parameter, $\phi$ for the inflationary scalar field and overdot as derivative with respect to cosmic they introduce a new quantity $z$ defined by z\equiv \frac{a\dot{\phi}}{H} \,.$$ The one desires to calculate is curvature ${\cal R}$, as Ref. It is convenient define a related quantity $u$, defined by $$\label{uR} u = - z {\cal R} \,,$$ which is gauge invariant A Fourier of into modes $u_{{\bf k}}$ out, and these can be shown remarkably simple equation [@Muk; @MFB; @SL] $$\label{ufield} \frac{d^2 k}}}{d\tau^2} +\left( - \frac{1}{z} \frac{d^2 z}{d\tau^2} \right) u_{{\bf =0 \,,$$ where $\tau$ is the conformal time by $d\tau = dt/a(t)$ and $k = |{\bf k}|$ is the modulus of the wavenumber. spectrum ${\cal P}_{{\cal R}}(k)$ the curvature perturbation the way $$\langle R}_{{\bf k_1}} R}^*_{{\bf k_2}} \rangle = \frac{2\pi^2}{k^3} {\cal P}_{{\cal R}} \delta^{3} \, ({\bf k_2}) \,,$$ yields $$\label{pspec} {\cal P}_{{\cal R}}^{1/2}(k) = \sqrt{\frac{k^3}{2\pi^2}} \, k}}}{z}

in any theory in which fluctuaTions in morE than One FieLd Are iMporTant, and it is posSIble To construct situations wHere tHiS Is thE CaSe [@twoPerts]. OuR AnALYsiS iS bAseD oN ThE formAliSm devisEd by StewarT anD LYth [@SL] to carry OUt A second-ordEr aNalytic calcuLatIon. We sHaLl nOT reprOduCe the DerivaTIon of tHe equatioNs HEre, insTEad simpLY ReProdUcing the important ONeS.[^1] with the usual noTation Of $A$ FoR SCalE faCtor, $H$ for HuBbLe parAMeter, $\phI$ FoR THE inFLationary scalAr field and oVErdOt as deRiVatIVe with RespeCt TO coSmic time, theY intRoduce a neW quantITy $z$ defiNEd by $$\labEl{z} z\eqUiv \FraC{a\doT{\PhI}}{H} \,.$$ the QuANtiTY oNe dESirEs to calcUlAtE is thE curVATURe peRtuRbatIon ${\caL R}$, defined as in ref. [@sL]. It IS coNveniEnt to DefiNe A relaTed quaNtity $U$, dEfined by $$\label{uR} U = - z {\caL R} \,,$$ which is A gaUgE inVaRiant POtentiAl [@MfB]. A fourier ExpansiON of $U$ iNTO CoMoving modes $u_{{\bf k}}$ is cArRIEd Out, and thEse can BE sHoWN to obey tHe RemArkaBLY simpLe eqUAtIon [@Muk; @MFb; @SL] $$\labEL{uFiEld} \frac{D^2 u_{{\Bf k}}}{d\taU^2} +\lEft( K^2 - \frAc{1}{z} \frAC{d^2 z}{d\Tau^2} \rigHt) u_{{\bf k}} =0 \,,$$ whEre $\taU$ Is the conformal TIme defined by $d\TAu = DT/A(t)$ ANd $k = |{\bF k}|$ iS the modulus Of thE WaveNumbER. DEfiNIng thE specTrUM ${\cAL P}_{{\cal R}}(k)$ of the curvatuRe PerturBatioN in the standarD way as $$\langLE {\CAl R}_{{\bf k_1}} {\caL R}^*_{{\bf K_2}} \RaNGle = \frac{2\pi^2}{k^3} {\cal P}_{{\Cal R}} \dElta^{3} \, ({\bf k_1}-{\bf k_2}) \,,$$ YIelds $$\labEl{pspEc} {\cal P}_{{\caL R}}^{1/2}(k) = \sqrt{\frAC{K^3}{2\pi^2}} \, \left| \fRac{U_{{\bf K}}}{z}

in any theory in which fl uctuations in m ore th an one fie ld are importa n t, a nd it is possible to c onstr uc t sit u at ionswhere t h is i s t he c ase [ @ tw opert s]. Our a nalysis is ba se d on the for m al ism devise d b y Stewart an d L yth [@ SL ] t o carr y o ut asecond - orderanalyticca l culati o n. We s h a ll not reproduce the de r iv a tion of the eq uation sh er e , in ste ad simplyre produ c ing the im p o r tan t ones.[^1] Wi th the usua l no tation o f $ a $ forscale f a cto r, $H$ forHubb le parame ter, $ \ phi$ fo r the in flatio nar y s cala r f ie ldan d ov e rd ota s d erivativ ewi th re spec t t o cos mic tim e, th ey introducea n ew q u ant ity $ z$ de fine dby $$ \label {z} z \e quiv \frac{a\do t{\p hi}}{H} \ ,.$ $The q uanti t y onedes ire s to ca lculate isth e c ur vature perturbatio n$ { \c al R}$, defin e das in Ref.[@ SL] . It i s con veni e nt to defi ne a r e la te d quant it y $u$, d efi ned by $ $ \lab el{uR} u = - z {\ca l R} \,,$$ whic h is a gauge i n va r i an t pot ent ial [@MFB]. AF ouri er e x pa nsi o n of$u$ i nt o c o moving modes $u_{{\ bf k}}$is ca rried out, an d these ca n b e shownto o b ey the remarkably simp le equatio n [@Muk;@MFB; @SL] $$ \label{uf i e ld} \fra c{d ^2u_{ {\b f k} }}{d\tau^2} + \ l eft( k ^2 - \f rac {1}{z} \ fra c{d ^2 z}{d\tau ^2} \rig ht )u_ {{ \bf k}}= 0 \,,$$wh ere $ \ta u$ is the co nform al t im ed efi ned by$ d\ t a u =dt /a (t)$ an d$k = |{\ b f k }|$ isthe modul uso f th ewa venumbe r. Defining t he spectrum${ \ca l P}_ { { \cal R}} (k)$ of the curvature p e rturbat ion in t he s tandard w ayas $$\ lan g le {\c al R}_ {{\bf k _1} } {\cal R }^ *_{ {\ bf k_2}} \ r a ngl e = \fr ac{2\pi ^2}{k^3} {\cal P}_ { {\c al R}} \delta ^{3 } \, ({ \ bf k_1 }- { \bf k _2}) \,,$$ yiel ds $$\labe l{ p sp ec} {\calP }_{ {\ cal R}} ^{1/2}( k) =\ sqrt{\f rac{k^3}{ 2\pi^2}}\, \le ft| \frac{ u_{{\bfk}}}{z}

in_any theory_in which fluctuations in_more than_one_field are_important,_and it is_possible to construct_situations where this is_the case [@twoperts]. Our_analysis_is based on the formalism devised by Stewart and Lyth [@SL] to carry out_a_second-order analytic_calculation._We_shall not reproduce the derivation_of the equations here, instead_simply reproducing_the important ones.[^1] With the usual notation of_$a$_for scale factor,_$H$ for Hubble parameter, $\phi$ for the inflationary scalar_field and overdot as derivative with_respect to cosmic_time,_they_introduce a new quantity_$z$ defined by $$\label{z} z\equiv \frac{a\dot{\phi}}{H} \,.$$_The quantity one desires to calculate_is the curvature perturbation ${\cal R}$, defined_as in Ref. [@SL]. It is convenient_to define a related quantity_$u$, defined_by $$\label{uR} u = - z_{\cal R} \,,$$_which is_a gauge invariant_potential [@MFB]. A Fourier expansion of $u$_into comoving modes_$u_{{\bf k}}$ is carried out, and_these_can be shown_to_obey_the remarkably_simple equation [@Muk;_@MFB;_@SL] $$\label{ufield} \frac{d^2_u_{{\bf_k}}}{d\tau^2} +\left( k^2 - \frac{1}{z} __ \frac{d^2 z}{d\tau^2} \right) u_{{\bf k}}_=0 \,,$$ where $\tau$_is_the conformal time defined_by $d\tau = dt/a(t)$ and_$k = |{\bf k}|$ is the_modulus of_the wavenumber._Defining the spectrum ${\cal P}_{{\cal R}}(k)$ of the curvature perturbation in_the standard way as $$\langle {\cal_R}_{{\bf k_1}} {\cal R}^*_{{\bf_k_2}} \rangle_= _ _\frac{2\pi^2}{k^3}_{\cal P}_{{\cal_R}} \delta^{3} \, ({\bf_k_1}-{\bf k_2})_\,,$$ yields $$\label{pspec} {\cal P}_{{\cal R}}^{1/2}(k) =_\sqrt{\frac{k^3}{2\pi^2}} \, __ \left| \frac{u_{{\bf k}}}{z}

field limit does not correspond to the limit of fast diffusion on networks if the interaction is *one-to-all* as is the case in our mafia model. Then, the dynamics of the network-structured society is actually better described within an extended mean-field theory which accounts for the fact that for each given node there is a finite set of neighborhood compositions, cf. Appendix A. In the following we discuss the standard mean-field approach in order to highlight the novel effects introduced by the network structure and the *one-to-all* interaction scheme. We non-dimensionalize by choosing the inverse death rate, $d^{-1}$, as our basic time scale and define dimensionless time and all other parameters accordingly: $\tau = td$, $\beta = b/d$, $\sigma_i = s_i/d$, $\omega_{ij} = w_{ij}/d$, with $i,j \in \{c,m\}$. Then, the mean-field equations for the time evolution of the different population fractions read: $$\begin{aligned} \dot{c} & = & \phi\,\beta + m\,\omega_{mc} - c\,\omega_{cm} - c,\label{eq: dotc}\\ \dot{m} & = & -m\,\omega_{mc} + c\,\omega_{cm} - m,\label{eq: dotm}\\ \dot{\phi} & = & -\phi\,\beta + (1-\phi ),\label{eq: dotphi}\end{aligned}$$ where a dot signifies a time derivative with respect to $\tau$. From Eq.  we immediately see that the stationary fraction of empty places is independent of the social dynamics encoded in the transition rates $\omega_{cm}$ and $\omega_{mc}$, but only depends on the dimensionless birth rate $\beta$: $$\phi=\frac{1}{1+\beta}\,.$$ In other words, the birth rates determines how densely populated the system is. Inserting this result together with the constraint $c+m+\phi=1$ into Eq.  yields an implicit equation, whose solutions are the fixed points or stationary states of the population dynamics: $$m\,\omega_{mc}(c,m,p) - \left(\frac{\beta}{1+\beta} - m\right)\,\omega_{cm}(c,m,p) + m = 0. \label{eq:master eq}$$

field limit does not correspond to the limit of firm dissemination on networks if the interaction is * one - to - all * as is the case in our mafia model. Then, the moral force of the network - structured society is actually well described within an extended average - field theory which account for the fact that for each given node there be a finite set of neighborhood compositions, cf. Appendix A. In the following we discuss the standard average - field approach in order to foreground the novel effects introduce by the net structure and the * one - to - all * interaction scheme. We non - dimensionalize by choosing the inverse death rate, $ d^{-1}$, as our basic time plate and define dimensionless time and all other parameters accordingly: $ \tau = td$, $ \beta = b / d$, $ \sigma_i = s_i / d$, $ \omega_{ij } = w_{ij}/d$, with $ i, j \in \{c, m\}$. Then, the beggarly - plain equations for the time evolution of the different population fractions read: $ $ \begin{aligned } \dot{c } & = & \phi\,\beta + m\,\omega_{mc } - c\,\omega_{cm } - c,\label{eq: dotc}\\ \dot{m } & = & -m\,\omega_{mc } + c\,\omega_{cm } - m,\label{eq: dotm}\\ \dot{\phi } & = & -\phi\,\beta + (1-\phi), \label{eq: dotphi}\end{aligned}$$ where a acid mean a time derivative with obedience to $ \tau$. From Eq.   we immediately visit that the stationary fraction of empty places is independent of the social moral force encoded in the passage rates $ \omega_{cm}$ and $ \omega_{mc}$, but only depend on the dimensionless birth rate $ \beta$: $ $ \phi=\frac{1}{1+\beta}\,.$$ In other words, the birth rates determine how densely populated the system is. Inserting this resultant role together with the restraint $ c+m+\phi=1 $ into Eq.   give an implicit equation, whose solutions are the fixed point or stationary states of the population dynamics: $ $ m\,\omega_{mc}(c, m, p) - \left(\frac{\beta}{1+\beta } - m\right)\,\omega_{cm}(c, m, p) + m = 0. \label{eq: master eq}$$

fiepd limit does not correskond to the limit of fasv diffuaion on vetworks if the interaction ms *obe-to-aol* as is the case in ojr mafia lodel. Thwn, tie dynamics of tis netwovh-strhgtureb wociety is actoally better described witviv cn extended mean-field theory which ascounts flr the fact thwt fpw eadh given node there is a finite sef of nepghborhood composotions, cf. Appendix A. In thf foplowing we discuss the standats mqqn-field appruach in order to highljght the novel effects introducdd by the netwotk shtucture and vhe *onv-to-all* interagnion scveme. We mon-dimensionallze bb chiosing the inverse deeth rate, $d^{-1}$, as our bafic time vccle and define dimensuobless tima ana alu ouhec pzrametfrs accordingmy: $\tau = td$, $\veta = b/d$, $\sigma_i = s_i/c$, $\jnega_{ij} = w_{ij}/d$, sith $i,t \yn \{c,m\}$. Then, the mean-field equations for uhe tjme evolution of the didferent population frwctions rqad: $$\begin{aligned} \dot{c} & = & \phi\,\beta + m\,\omega_{mc} - c\,\omeca_{cm} - c,\lcncl{eq: eohc}\\ \dot{m} & = & -m\,\omega_{mc} + c\,\omega_{cm} - m,\label{eq: doti}\\ \sou{\php} & = & -\phi\,\beta + (1-\pmi ),\label{eq: dotphi}\emd{wlodned}$$ where a aot siyhiries a time derivahive wijh respect to $\twu$. Ftom Eq.  we immediately see tyat the statponaey fraction of empcy places is indgpendemt of the social dynamies enckded in the transitikv rates $\omega_{cm}$ xnd $\okega_{mc}$, but only depends on the dimeisionkess bifth tate $\beea$: $$\phi=\frac{1}{1+\heta}\,.$$ Lt other words, the hirth rdtes deterlines how densely populated the system is. Insgrthng this refult bogether with tre constraint $e+m+\phi=1$ inco Eq.  hields an pmplicit xquation, whofe solutions dte the fixed 'oints or starionqry stagds of the popukation dyucmics: $$m\,\omwga_{mc}(c,m,p) - \left(\frac{\neta}{1+\bgtz} - m\right)\,\omega_{cn}(c,m,k) + m = 0. \label{eq:maxtef ez}$$

field limit does not correspond to the fast on networks the interaction is in mafia model. Then, dynamics of the society is actually better described within extended mean-field theory which accounts for the fact that for each given node is a finite set of neighborhood compositions, cf. Appendix A. In the following discuss standard approach order to highlight the novel effects introduced by the network structure and the *one-to-all* interaction scheme. non-dimensionalize by choosing the inverse death rate, $d^{-1}$, our basic time scale define dimensionless time and all parameters $\tau = $\beta b/d$, = s_i/d$, $\omega_{ij} w_{ij}/d$, with $i,j \in \{c,m\}$. Then, the mean-field equations for the time evolution of the different population read: $$\begin{aligned} = & + - - c,\label{eq: dotc}\\ = & -m\,\omega_{mc} + c\,\omega_{cm} - & = & -\phi\,\beta + (1-\phi ),\label{eq: dotphi}\end{aligned}$$ a dot a time derivative with respect to From Eq. we immediately see that the stationary of empty places is independent of the social dynamics encoded in the transition rates $\omega_{cm}$ but only depends on dimensionless birth rate $$\phi=\frac{1}{1+\beta}\,.$$ other the rates determines densely populated the system is. Inserting this result together with the $c+m+\phi=1$ into Eq. yields an implicit equation, whose solutions are points stationary states of population dynamics: $$m\,\omega_{mc}(c,m,p) - - + m = 0.

field limit does not correspoNd to the limIt of fAst DifFuSion On neTworks if the intERactIon is *one-to-all* as is the caSe in oUr MAfia MOdEl. TheN, the dynAMiCS Of tHe NeTwoRk-STrUcturEd sOciety iS actually bEttEr Described witHIn An extended MeaN-field theory WhiCh accoUnTs fOR the fAct That fOr each GIven noDe there is A fINite seT Of neighBORhOod cOmpositions, cf. AppeNDiX a. In the followinG we disCuSS tHE StaNdaRd mean-fielD aPproaCH in ordeR To HIGHliGHt the novel effEcts introduCEd bY the neTwOrk STructuRe and ThE *One-To-all* interaCtioN scheme. We Non-dimENsionalIZe by choOsing tHe iNveRse dEAtH rAte, $D^{-1}$, aS Our BAsIc tIMe sCale and dEfInE dimeNsioNLESS timE anD all Other Parameters accOrdInglY: $\Tau = Td$, $\betA = b/d$, $\siGma_i = S_i/D$, $\omegA_{ij} = w_{ij}/D$, with $I,j \In \{c,m\}$. Then, the mean-FielD equationS foR tHe tImE evolUTion of The DifFerent pOpulatiON frAcTIONs Read: $$\begin{aligned} \doT{c} & = & \PHI\,\bEta + m\,\omegA_{mc} - c\,\omEGa_{Cm} - C,\Label{eq: dOtC}\\ \doT{m} & = & -m\,\oMEGa_{mc} + c\,\OmegA_{Cm} - M,\label{eq: Dotm}\\ \doT{\PhI} & = & -\pHi\,\beta + (1-\pHi ),\Label{eQ: dOtpHi}\eNd{aliGNed}$$ wHere a dOt signifIes a tIMe derivative wiTH respect to $\tau$. fRoM eQ.  wE ImmeDiaTely see that The sTAtioNary FRaCtiON of emPty plAcES iS Independent of the socIaL dynamIcs enCoded in the traNsition ratES $\OMega_{cm}$ anD $\omeGA_{mC}$, But only depends On the DimensionlESs birth rAte $\beTa$: $$\phi=\fraC{1}{1+\beta}\,.$$ In otHER words, thE biRth RatEs dETErMines how denseLY PopuLaTed the sYstEm is. InsErtIng ThiS reSuLt togetheR with the CoNsTrAiNt $c+M+\phi=1$ iNTo Eq.  yielDs An iMpLicIt equATion, whOse soLutiOnS aRE thE fixed pOInTS Or stAtIoNary StaTeS of thE popULatIon dynaMics: $$m\,\omegA_{mc}(C,M,p) - \leFt(\FrAc{\beta}{1+\bEta} - m\right)\,\omegA_{cM}(c,m,p) + m = 0. \label{Eq:MasTer eq}$$

field limit does not corre spond to t he li mit of f astdiff usion on netwo r ks i f the interaction is * one-t o- a ll*a sis th e casei no u r m af ia mo de l .Then, th e dynam ics of the ne tw ork-structur e dsociety is ac tually bette r d escrib ed wi t hin a n e xtend ed mea n -field theory w hi c h acco u nts for t he fac t that for each g i ve n node there is a fin it e s e t of ne ighborhood c ompos i tions,c f. A p pen d ix A. In thefollowing w e di scussth e s t andard mean -f i eld approach i n or der to hi ghligh t the no v el effe cts in tro duc ed b y t he ne tw o rks tr uct u reand the*o ne -to-a ll*i n t e ract ion sch eme. We non-dimen sio nali z e b y cho osing the i nvers e deat h rat e, $d^{-1}$, as o ur b asic time sc al e a nd defi n e dime nsi onl ess tim e and a l l o th e r pa rameters according ly : $\ tau = td $, $\b e ta = b/d$, $\ si gma _i = s _i/d$ , $\ o me ga_{ij}= w_{i j }/ d$ , with$i ,j \in \ {c, m\} $. Th e n, t he mea n-fieldequat i ons for the ti m e evolution o f t h e d i ffer ent population fra c tion s re a d: $$ \ begin {alig ne d }\ dot{c} & = & \phi\ ,\ beta + m\,\ omega_{mc} - c\,\omega _ { c m} - c, \lab e l{ e q: dotc}\\ \do t{m}& = & -m\ , \omega_{ mc} + c\,\om ega_{cm}- m,\labe l{e q:dot m}\ \ \d ot{\phi} & =& -\ph i\ ,\beta + (1-\ph i ) ,\l abe l{e q: dotphi}\ end{alig ne d} $$ w her e a d o t signif ie s a t ime deri v ativewithresp ec tt o $ \tau$.F ro m Eq. w eimme dia te ly se e th a t t he stat ionary fr act i on o fem pty pla ces is indepe nd ent of the s oci al dyn a m ics enco ded in the transition r a tes $\o meg a_{cm }$ a nd $\omeg a_{ mc}$,but only d epends on t he di m e nsion l e ss bi rt h rate $\b e t a$: $$\p hi =\fr ac{1}{1 +\beta}\,.$$ In ot h erwords, the bi rth rat e s d ete r mi n esho w de n s ely populated t he systemis . I nserting t h isre sult to getherwitht he cons traint $c +m+\phi=1 $into E q. yields an implici t equatio n , who s esolut ion s areth e f ixedpoints orstati onaryst ates o f the p opulatio n dynamics: $$m\,\omega _{mc}( c,m,p ) - \left(\f rac { \be ta}{1+\be ta}- m\right) \,\ ome ga_{c m}( c ,m,p) +m = 0. \labe l{eq : master eq } $$

field limit_does not_correspond to the limit_of fast_diffusion_on networks_if_the interaction is_*one-to-all* as is_the case in our_mafia model. Then,_the_dynamics of the network-structured society is actually better described within an extended mean-field theory_which_accounts for_the_fact_that for each given node_there is a finite set_of neighborhood_compositions, cf. Appendix A. In the following we_discuss_the standard mean-field_approach in order to highlight the novel effects introduced_by the network structure and the_*one-to-all* interaction scheme. We_non-dimensionalize_by_choosing the inverse death_rate, $d^{-1}$, as our basic time_scale and define dimensionless time and_all other parameters accordingly: $\tau = td$,_$\beta = b/d$, $\sigma_i = s_i/d$,_$\omega_{ij} = w_{ij}/d$, with $i,j_\in \{c,m\}$._Then, the mean-field equations for_the time evolution_of the_different population fractions_read: $$\begin{aligned} \dot{c} & = &_\phi\,\beta + m\,\omega_{mc}_ - c\,\omega_{cm} - c,\label{eq:_dotc}\\ \dot{m}_& = &__-m\,\omega_{mc}_ +_c\,\omega_{cm} -_m,\label{eq:_dotm}\\ \dot{\phi} &_=_& -\phi\,\beta + (1-\phi_),\label{eq:_dotphi}\end{aligned}$$ where a dot signifies a time_derivative with respect to_$\tau$._From Eq.  we immediately_see that the stationary fraction_of empty places is independent of_the social_dynamics encoded_in the transition rates $\omega_{cm}$ and $\omega_{mc}$, but only depends on_the dimensionless birth rate $\beta$: $$\phi=\frac{1}{1+\beta}\,.$$_In other words, the_birth rates_determines_how densely populated_the_system is._Inserting this result together with the constraint_$c+m+\phi=1$ into_Eq.  yields an implicit equation, whose_solutions are the fixed_points_or stationary states of the population_dynamics: $$m\,\omega_{mc}(c,m,p) - \left(\frac{\beta}{1+\beta} - m\right)\,\omega_{cm}(c,m,p)_ + m = 0. \label{eq:master_eq}$$

flux method [@waf]. The accurate chemical composition of the single crystals was determined by energy dispersive X-ray spectroscopy (EDS). The standard instrument error for this method is around 10%. The single crystals of $x=0$, 0.010, 0.014, 0.017, 0.028, 0.042, 0.047, 0.061, 0.075 and 0.109 were measured, with a high quality characterized and demonstrated in our previous report [@waf]. X-ray diffraction (XRD) was performed on Smartlab-9 diffractometer (Rikagu) from 10$^{\rm o}$ to 60$^{\rm o}$ with a scanning rate of 2$^{\rm o}$/minute. Measurements of resistivity and Hall effect were conducted by using the PPMS-9T (Quantum Design). The resistivity were measured using the standard four-probe method. The contacts cover the sides of the samples to ensure in-plane transport. Hall coefficient was measured by sweeping the field from -5T to 5T at various temperatures. We firstly confirmed the linearity of Hall resistivity $\rho_{xy}$ to magnetic field in the normal state. An excellent linearity can be observed for all the samples in the normal state except that a very weak deviation from the linearity appears below $T_{\rm SDW}$ in underdoped samples. Then the temperature dependence of Hall coefficients was obtained from the substraction of the voltages measured at 5 T and -5 T, $R_{\rm H}$ = \[$V_{xy}$(5T)-$V_{xy}$(-5T)\]$\times d$/2$I$, where $d$ is the thickness of crystals and $I$ is the current. The cotangent of Hall angle, cot$\theta_{\rm H}$=$\rho/\rho_{xy}$, was calculated at 5 T, where $\rho_{xy}$ is Hall resistivity. It should be addressed that the Hall coefficient can not be well measured at the temperature above 200 K, the possible reason is that Na ions could move above 200 K. All the samples we used are the same with that in our previous work[@waf] and the $T_{\rm c}$, $T_{\rm SDW}$ and $T_{\rm s}$ are summarized in Table 1 and plotted in Fig. 6. All the superconducting samples show nearly full shielding fraction except for the parent compound with filamentary superconductivity. RESULTS AND DISCUSSION ====================== We carefully measured the resistivity on single crystals

flux method [ @waf ]. The accurate chemical composition of the single crystal was determine by energy dispersive X - ray spectroscopy (EDS). The standard instrumental role error for this method is about 10% . The unmarried crystals of $ x=0 $, 0.010, 0.014, 0.017, 0.028, 0.042, 0.047, 0.061, 0.075 and 0.109 were measured, with a high quality characterize and demonstrated in our previous composition [ @waf ]. ten - ray diffraction (XRD) was performed on Smartlab-9 diffractometer (Rikagu) from 10$^{\rm o}$ to 60$^{\rm o}$ with a scanning rate of 2$^{\rm o}$/minute. Measurements of resistivity and Hall effect were conduct by using the PPMS-9 T (Quantum Design). The resistivity were measure using the standard four - probe method. The contacts cover the side of the samples to ensure in - plane transport. Hall coefficient was measured by sweep the field from -5 T to 5 T at various temperatures. We first confirm the linearity of Hall resistivity $ \rho_{xy}$ to magnetic field in the normal state. An excellent linearity can be observed for all the samples in the normal state except that a very watery deviation from the linearity appear below $ T_{\rm SDW}$ in underdoped samples. Then the temperature dependence of Hall coefficients was obtained from the substraction of the voltage measured at 5 T and -5 T, $ R_{\rm H}$ = \[$V_{xy}$(5T)-$V_{xy}$(-5T)\]$\times d$/2$I$, where $ d$ is the thickness of crystals and $ I$ is the current. The cotangent of Hall slant, cot$\theta_{\rm H}$=$\rho/\rho_{xy}$, was calculated at 5 T, where $ \rho_{xy}$ is Hall resistivity. It should be addressed that the Hall coefficient cannot be well measured at the temperature above 200 K, the possible reason is that Na ions could move above 200 K. All the samples we used are the like with that in our previous work[@waf ] and the $ T_{\rm c}$, $ T_{\rm SDW}$ and $ T_{\rm s}$ are summarize in Table 1 and plat in Fig. 6. All the superconducting samples show nearly full shielding fraction except for the parent compound with filamentary superconductivity. result AND DISCUSSION = = = = = = = = = = = = = = = = = = = = = = We carefully measured the resistivity on individual crystals

flkx method [@waf]. The accuraue chemical composition mf the single zrystals was determined by eiergt diskvrsive X-ray spectroscupy (EDS). Tje standqrd mnstrument error for this methkf is eround 10%. The sinnle crystalv of $x=0$, 0.010, 0.014, 0.017, 0.028, 0.042, 0.047, 0.061, 0.075 avd 0.109 were measured, with a high quality characyegized and demogstrseed jn our previous report [@waf]. X-ray dirfractimn (XRD) was petformed on Smartlab-9 diffraftomfter (Rikagu) from 10$^{\rl o}$ to 60$^{\rm o}$ witr a scanning fate of 2$^{\rm o}$/minute. Meaaurements of resistivity and Haul efyect were cineuchgd by using vhe PPIS-9T (Quantum Design). Tve resixtivity were mcasurxd uwing the standard fouc-probe method. The cogtacts coeex the sides of the sanpoes tm envure un-puant tcanaport. Jalm coefficisnt was meawured by sweeping tne dield from -5T fo 5T ae darious temperatures. We firstly confirmtd ths linearity of Hall resustivity $\rho_{xy}$ to magjetic fiejd in the normal state. An excellent linearity can be ousdrvtd for xol the samples in the normal state except that w vtry weak deviatiok from the lineariyy alkears below $T_{\ro SDW}$ nh hnderdoped samples. Then tre tenperature deprndence of Hall coefficientw was obtainvd feom the substractiln of the vultabes mrasured at 5 T and -5 T, $R_{\ro H}$ = \[$V_{xy}$(5T)-$V_{xy}$(-5T)\]$\tiles d$/2$I$, whsfe $d$ is the thicynexs of crysuxls and $I$ is the surrent. Tie cocangent uf Hsll andle, cot$\theha_{\rm M}$=$\sho/\rho_{xy}$, was calcupated ad 5 T, where $\rho_{xy}$ is Hall resistivity. It sipuld be addrgssad nhat the Kall cpefficient cag not be well keasureb at tfe temperanure abovx 200 K, the posfible reason hd that Na iois could iove aboce 200 K. Auu the samples ee used age the same with that in out lrevious work[@way] qnd the $T_{\rm c}$, $T_{\tm RDW}$ ajd $T_{\wk s}$ are summdrizdd km Tabue 1 and plobtea in Fig. 6. All the supercmndudting samples show ncarly fulo shieldyng fraction rxcept for the partnt cokponnd wiyh silamentary superconductivity. RSSULTS ANF DLSCUSSION ====================== We cwrefmlly measured che resistivity on single crystals

flux method [@waf]. The accurate chemical composition single was determined energy dispersive X-ray error this method is 10%. The single of $x=0$, 0.010, 0.014, 0.017, 0.028, 0.047, 0.061, 0.075 and 0.109 were measured, with a high quality characterized and in our previous report [@waf]. X-ray diffraction (XRD) was performed on Smartlab-9 diffractometer from o}$ 60$^{\rm with a scanning rate of 2$^{\rm o}$/minute. Measurements of resistivity and Hall effect were conducted by the PPMS-9T (Quantum Design). The resistivity were measured the standard four-probe method. contacts cover the sides of samples ensure in-plane Hall was by sweeping the from -5T to 5T at various temperatures. We firstly confirmed the linearity of Hall resistivity $\rho_{xy}$ to field in state. An linearity be for all the the normal state except that a from the linearity appears below $T_{\rm SDW}$ in samples. Then temperature dependence of Hall coefficients was from the substraction of the voltages measured at T and -5 T, $R_{\rm H}$ = \[$V_{xy}$(5T)-$V_{xy}$(-5T)\]$\times d$/2$I$, where $d$ is the thickness of $I$ is the current. cotangent of Hall cot$\theta_{\rm was at T, where is Hall resistivity. It should be addressed that the Hall coefficient not be well measured at the temperature above 200 K, reason that Na ions move above 200 K. the we used are the that our the c}$, SDW}$ and $T_{\rm s}$ summarized in Table 1 and in Fig. 6. All full shielding fraction except for the parent compound filamentary superconductivity. RESULTS AND DISCUSSION ====================== We measured the resistivity on single crystals

flux method [@waf]. The accurate cHemical comPositIon Of tHe SingLe crYstals was deterMIned By energy dispersive X-ray SpectRoSCopy (edS). the stAndard iNStRUMenT eRrOr fOr THiS methOd iS around 10%. the single cRysTaLs of $x=0$, 0.010, 0.014, 0.017, 0.028, 0.042, 0.047, 0.061, 0.075 and 0.109 were MEaSured, with a HigH quality charActErized AnD deMOnstrAteD in ouR previOUs repoRt [@waf]. X-ray DiFFractiON (XRD) was PERfOrmeD on Smartlab-9 diffraCToMEter (Rikagu) from 10$^{\Rm o}$ to 60$^{\rM o}$ WItH A ScaNniNg rate of 2$^{\rm O}$/mInute. mEasuremENtS OF ResIStivity and HalL effect were COndUcted bY uSinG The PPMs-9T (QuaNtUM DeSign). The resiStivIty were meAsured USing the STandard Four-prObe MetHod. THE cOnTacTs COveR ThE siDEs oF the sampLeS tO ensuRe in-PLANE traNspOrt. HAll coEfficient was mEasUred BY swEepinG the fIeld FrOm -5T to 5t at varIous tEmPeratures. We firsTly cOnfirmed tHe lInEarItY of HaLL resisTivIty $\Rho_{xy}$ to MagnetiC FieLd IN THe Normal state. An excelLeNT LiNearity cAn be obSErVeD For all thE sAmpLes iN THe norMal sTAtE except tHat a veRY wEaK deviatIoN from tHe LinEarIty apPEars Below $T_{\Rm SDW}$ in uNderdOPed samples. Then THe temperature DEpENDeNCe of halL coefficienTs waS ObtaIned FRoM thE SubstRactiOn OF tHE voltages measured at 5 t aNd -5 T, $R_{\rm h}$ = \[$V_{xy}$(5T)-$v_{xy}$(-5T)\]$\times d$/2$I$, whEre $d$ is the tHICKness of cRystALs ANd $I$ is the currenT. The cOtangent of hAll angle, Cot$\thEta_{\rm H}$=$\rhO/\rho_{xy}$, was CALculated At 5 T, WheRe $\rHo_{xY}$ IS HAll resistivitY. iT shoUlD be addrEssEd that tHe HAll CoeFfiCiEnt can not Be well meAsUrEd At The TempeRAture aboVe 200 k, thE pOssIble rEAson is That NA ionS cOuLD moVe above 200 k. alL THe saMpLeS we uSed ArE the sAme wITh tHat in ouR previous WorK[@Waf] aNd ThE $T_{\rm c}$, $T_{\rM SDW}$ and $T_{\rm s}$ arE sUmmarized iN TAblE 1 and plOTTed in Fig. 6. all the superconducting saMPles shoW neArly fUll sHielding fRacTion exCepT For the Parent CompoUnD wiTH FilamENTaRy sUpErconductiVITy. ReSULTs AnD DIsCUSSIOn ====================== We carefully measurED thE resistivity oN siNgle CRYsTalS

flux method [@waf]. The a ccurate ch emica l c omp os itio n of the single cr y stal s was determined by en ergydi s pers i ve X-ra y spect r os c o py(E DS ).Th e s tanda rdinstrum ent errorfor t his method i s a round 10%. Th e single cry sta ls of$x =0$ , 0.01 0,0.014 , 0.01 7 , 0.02 8, 0.042, 0 . 047, 0 . 061, 0. 0 7 5and0.109 were measur e d, with a high qu alitych a ra c t eri zed and demon st rated in ourp re v i o usr eport [@waf]. X-ray diff r act ion (X RD ) w a s perf ormed o n Sm artlab-9 di ffra ctometer(Rikag u ) from1 0$^{\rm o}$ t o 6 0$^ {\rm o} $wit ha sc a nn ing rat e of 2$^ {\ rm o}$/ minu t e . Meas ure ment s ofresistivity a ndHall eff ect w ere c ondu ct ed by using thePP MS-9T (QuantumDesi gn). Theres is tiv it y wer e measu red us ing the standa r d f ou r - p ro be method. The con ta c t scover th e side s o ft he sampl es to ens u r e in- plan e t ransport . Hall co ef ficient w as mea su red by swee p ingthe fi eld from -5Tt o 5T at variou s temperatures . W e fi r stly co nfirmed the lin e arit y of Ha llr esist ivity $ \ rh o _{xy}$ to magneticfi eld in thenormal state. An excell e n t lineari ty c a nb e observed for allthe sample s in thenorma l stateexcept th a t a verywea k d evi ati o n f rom the linea r i ty a pp ears be low $T_{\r m S DW} $ i n u nd erdoped s amples.Th en t he te mpera t ure depe nd enc eofHallc oeffic ients was o bt a ine d fromt he s ubst ra ct ionofth e vol tage s me asuredat 5 T an d - 5 T,$R _{ \rm H}$ = \[$V_{xy}$ (5 T)-$V_{xy} $( -5T )\]$\t i m es d$/2$ I$, where $d$ is the th i cknessofcryst alsand $I$ i s t he cur ren t . Thecotang ent o fHal l angle , co t$\ th eta_{\rm H } $ =$\ rho/\ rh o_{x y}$, wa s calculated at 5T , w here $\rho_{x y}$ isH a ll re s is t ivi ty . It s hould be addres sed that t he Ha ll coeffic i ent c an notbe well meas u red atthe tempe rature ab ov e 20 0 K,the possib le reaso n is that Na io n scould mo ve abo ve 20 0 K.All th e sa mples we us ed are t he sa me with th at in our previous work [@waf] andthe $T_{\rmc}$ , $T _{\rm SDW }$ a nd $T_{\rm s} $ a re su mma r izedin T a bl e 1 and p lott e d in Fig. 6. Al l th e supercond u c t ing samp les show n earl y full shieldingf raction except for t hepar e nt c om pound with fil ame nt a r y superc on ductivity. RESULTS A N D DIS CUSSIO N ==== ======= = = == = ====== We ca refully m eas ur e d the r es is t ivityon s in gle cr ystals

flux_method [@waf]._The accurate chemical composition_of the_single_crystals was_determined_by energy dispersive_X-ray spectroscopy (EDS)._The standard instrument error_for this method_is_around 10%. The single crystals of $x=0$, 0.010, 0.014, 0.017, 0.028, 0.042, 0.047, 0.061,_0.075_and 0.109_were_measured,_with a high quality characterized_and demonstrated in our previous_report [@waf]._X-ray diffraction (XRD) was performed on Smartlab-9 diffractometer_(Rikagu)_from 10$^{\rm o}$_to 60$^{\rm o}$ with a scanning rate of 2$^{\rm_o}$/minute. Measurements of resistivity and Hall_effect were conducted_by_using_the PPMS-9T (Quantum Design)._The resistivity were measured using the_standard four-probe method. The contacts cover_the sides of the samples to ensure_in-plane transport. Hall coefficient was measured_by sweeping the field from_-5T to_5T at various temperatures. We_firstly confirmed the_linearity of_Hall resistivity $\rho_{xy}$_to magnetic field in the normal_state. An excellent_linearity can be observed for all_the_samples in the_normal_state_except that_a very weak_deviation_from the_linearity_appears below $T_{\rm SDW}$ in underdoped_samples._Then the temperature dependence of Hall coefficients_was obtained from the_substraction_of the voltages measured_at 5 T and -5_T, $R_{\rm H}$ = \[$V_{xy}$(5T)-$V_{xy}$(-5T)\]$\times d$/2$I$,_where $d$_is the_thickness of crystals and $I$ is the current. The cotangent of_Hall angle, cot$\theta_{\rm H}$=$\rho/\rho_{xy}$, was calculated_at 5 T, where_$\rho_{xy}$ is_Hall_resistivity. It should_be_addressed that_the Hall coefficient can not be well_measured at_the temperature above 200 K, the_possible reason is that_Na_ions could move above 200 K._All the samples we used are_the same with that in_our_previous_work[@waf] and the $T_{\rm c}$,_$T_{\rm SDW}$ and $T_{\rm s}$ are_summarized in Table_1 and plotted in Fig. 6. All_the_superconducting samples show nearly full shielding_fraction_except for the parent compound with_filamentary_superconductivity. RESULTS_AND DISCUSSION ====================== We carefully measured the_resistivity on single crystals

and finding properties of $\aleph$.]{} From assumption (A1), if a nucleotide is proposed as a candidate for joining the end of one of the strands, then its complement is automatically a candidate for joining the beginning of the other strand. Therefore, the probability of a nucleotide being selected as a candidate for joining one strand is the same as the probability of its complement being selected for the other strand, that is,\ [**($H1$)**]{} $\mu_A=\mu_T$ and $\mu_C=\mu_G$. In other words, $\mu$ takes the form $\mu(m)=(m,0.5-m,0.5-m,m)$, where $0\leq m \leq 0.5$.\ Now, since the matrix $\aleph$ is assumed to be the same for all genomes, any sequence produced by the model is a realization of a Markov chain belonging to a family of Markov chains parameterized by $m$ and $t$, where $0< m< 0.5$ and $0\leq t\leq 1$. Further, from (A2) we have that the probability of a candidate nucleotide of type $j$ being accepted to follow a nucleotide of type $i$ at the end of one strand is equal to the probability of a nucleotide of type $\alpha(i)$ preceding a nucleotide of type $\alpha(j)$ at the beginning of the other strand, that is, the matrix $\aleph$ has the form\ [**($H2$)**]{} $a_{ij}=a_{\alpha(j)\alpha(i)}$.\ We use the [*lsqcurvefit*]{} function of the optimization toolbox in [Matlab]{} to construct estimators for the matrices $\aleph(n)$, vectors $\mu(n)$ and values $t(n)$. The [*lsqcurvefit*]{} function solves (in the least-squares sense) the nonlinear problem of determining the parameters for which the right side of equation most closely approximates $P(n)$. Firstly, we carried out free estimation without the [*a priori*]{} assumption of the properties ($H1$) and ($H2$). The solution to the optimization problem is sensitive to the value of $t(n)$ chosen to initialize the optimization algorithm. Since previous tests suggest that neither strand is significantly favored over the other during the construction of the sequence, we used $t(n)=0.5

and finding properties of $ \aleph$. ] { } From assumption (A1), if a nucleotide is proposed as a campaigner for join the end of one of the strands, then its complement is mechanically a candidate for join the beginning of the other strand. consequently, the probability of a nucleotide being selected as a candidate for joining one chain is the same as the probability of its complement being selected for the early strand, that is,\ [ * * ($ H1 $) * * ] { } $ \mu_A=\mu_T$ and $ \mu_C=\mu_G$. In other words, $ \mu$ takes the form $ \mu(m)=(m,0.5 - m,0.5 - m, m)$, where $ 0\leq m \leq 0.5$.\ nowadays, since the matrix   $ \aleph$ is assumed to be the same for all genomes, any sequence produced by the exemplar is a realization of a Markov chain belonging to a family of Markov chain parameterized by   $ m$ and   $ t$, where $ 0 < m < 0.5 $ and $ 0\leq t\leq 1$. Further, from (A2) we have that the probability of a candidate nucleotide of type   $ j$ being accept to follow a nucleotide of type   $ i$ at the end of one strand is equal to the probability of a nucleotide of character $ \alpha(i)$ preceding a nucleotide of type $ \alpha(j)$ at the beginning of the other strand, that is, the matrix $ \aleph$ has the form\ [ * * ($ H2 $) * * ] { } $ a_{ij}=a_{\alpha(j)\alpha(i)}$.\ We use the [ * lsqcurvefit * ] { } function of the optimization toolbox in [ Matlab ] { } to construct calculator for the matrix $ \aleph(n)$, vectors   $ \mu(n)$ and values $ t(n)$. The [ * lsqcurvefit * ] { } function solves (in the least - squares sense) the nonlinear problem of determining the parameters for which the right slope of equation most close approximate   $ P(n)$. first, we carried out free estimation without the [ * a priori * ] { } assumption of the properties ($ H1 $) and ($ H2 $). The solution to the optimization problem is sensitive to the value of   $ t(n)$ chosen to format the optimization algorithm. Since previous tests suggest that neither strand is significantly favored over the other during the construction of the sequence, we used $ t(n)=0.5

anf finding properties of $\xleph$.]{} From assumption (E1), if a hucleotiae is proposed as a candidatx foe joibing the end of one of the strajds, then its xomplement is automaticampy a randidate for jpining the beginning of dhd lther strand. Therefore, the probabilyty of s jucleotide beigg stlested as a candidate for joining one stdand is the same as yhe probability of its comolemfnt being selected for the otyer frrand, that ir,\ [**($H1$)**]{} $\mu_A=\mu_T$ and $\mu_C=\mu_G$. Ih other words, $\mu$ takes the form $\mu(m)=(m,0.5-k,0.5-m,m)$, where $0\oew m \neq 0.5$.\ Now, sincx the iatrix $\aleph$ lx assuked to ne the same fov all gebomes, any sequence prmduced by the modej is a redlnzation of a Markov cyaun benonghng gi a fajiky of Magkot chains padameterized by $m$ and $t$, where $0< m< 0.5$ age $0\leq t\leq 1$. Fudther, srjm (A2) we have that the probability of a bandjdate nucleotide of typw $j$ being accepted to vollow a gucleotide of type $i$ at the end of one strand is exual vo tht ivobacullty of a nucleotide of type $\alpha(i)$ preceding w nickeotide of typc $\alpha(j)$ at the bebijnogg of the othgr straus, fhat is, the matrix $\aleph$ ras tye form\ [**($H2$)**]{} $w_{ij}=a_{\slpha(j)\alpha(i)}$.\ We use the [*lsqcyrvefit*]{} funcnion of the optimizatiln toolbox nn [Matkab]{} tp construct estimators yor ths matrices $\wleph(n)$, vedgors $\mu(n)$ and valuds $n(n)$. Tve [*lsqcurvefit*]{} function sojves (in tie lecst-squards sgnse) thq nonlineag pronnem of determining the karameders for wjich the right side of equation most closely sp[roqimates $P(n)$. Firsbly, we carried jut free estimction wichout ghe [*a priogi*]{} assumpvion of the [roperties ($H1$) djd ($H2$). The soluvion to tre optimuzation oroblem is senxitive to the value of $t(n)$ chosen to inltialkae the optimizacnob algorithm. Sinve oredilux easts suggest thag ndother strand is wkgnigicantly favored oves ths other during the cjnstructuon of tre sequence, wr used $t(n)=0.5

and finding properties of $\aleph$.]{} From assumption a is proposed a candidate for of strands, then its is automatically a for joining the beginning of the strand. Therefore, the probability of a nucleotide being selected as a candidate for one strand is the same as the probability of its complement being selected the strand, is,\ $\mu_A=\mu_T$ and $\mu_C=\mu_G$. In other words, $\mu$ takes the form $\mu(m)=(m,0.5-m,0.5-m,m)$, where $0\leq m \leq 0.5$.\ since the matrix $\aleph$ is assumed to be same for all genomes, sequence produced by the model a of a chain to family of Markov parameterized by $m$ and $t$, where $0< m< 0.5$ and $0\leq t\leq 1$. Further, from (A2) we that the a candidate of $j$ accepted to follow of type $i$ at the end is equal to the probability of a nucleotide type $\alpha(i)$ a nucleotide of type $\alpha(j)$ at beginning of the other strand, that is, the $\aleph$ has the form\ [**($H2$)**]{} $a_{ij}=a_{\alpha(j)\alpha(i)}$.\ We use the [*lsqcurvefit*]{} function of the optimization toolbox to construct estimators for matrices $\aleph(n)$, vectors and $t(n)$. [*lsqcurvefit*]{} solves (in least-squares sense) the nonlinear problem of determining the parameters for which right side of equation most closely approximates $P(n)$. Firstly, we free without the [*a assumption of the properties and The solution to the is to $t(n)$ to the optimization algorithm. Since tests suggest that neither strand significantly favored over the the sequence, we used $t(n)=0.5

and finding properties of $\alePh$.]{} From assuMptioN (A1), iF a nUcLeotIde iS proposed as a caNDidaTe for joining the end of onE of thE sTRandS, ThEn its ComplemENt IS AutOmAtIcaLlY A cAndidAte For joinIng the begiNniNg Of the other stRAnD. Therefore, The Probability oF a nUcleotIdE beINg selEctEd as a CandidATe for jOining one StRAnd is tHE same as THE pRobaBility of its compleMEnT Being selected fOr the oThER sTRAnd, ThaT is,\ [**($H1$)**]{} $\mu_A=\mu_T$ AnD $\mu_C=\mU_g$. In otheR WoRDS, $\Mu$ tAKes the form $\mu(m)=(M,0.5-m,0.5-m,m)$, where $0\leQ M \leQ 0.5$.\ Now, siNcE thE Matrix $\Aleph$ Is ASsuMed to be the sAme fOr all genoMes, any SEquence PRoduced By the mOdeL is A reaLIzAtIon Of A marKOv ChaIN beLonging tO a FaMily oF MarKOV CHainS paRameTerizEd by $m$ and $t$, wherE $0< m< 0.5$ aNd $0\leQ T\leQ 1$. FurtHer, frOm (A2) wE hAve thAt the pRobabIlIty of a candidate NuclEotide of tYpe $J$ bEinG aCceptED to folLow A nuCleotidE of type $I$ At tHe END Of One strand is equal to ThE PRoBability Of a nucLEoTiDE of type $\aLpHa(i)$ PrecEDIng a nUcleOTiDe of type $\Alpha(j)$ AT tHe BeginniNg Of the oThEr sTraNd, thaT Is, thE matriX $\aleph$ haS the fORm\ [**($H2$)**]{} $a_{ij}=a_{\alpha(j)\aLPha(i)}$.\ We use the [*lSQcURVeFIt*]{} fuNctIon of the optImizATion ToolBOx In [MATlab]{} tO consTrUCt EStimators for the matrIcEs $\alepH(n)$, vecTors $\mu(n)$ and valUes $t(n)$. The [*lsQCURvefit*]{} fuNctiON sOLves (in the least-SquarEs sense) the NOnlinear ProblEm of deteRmining thE PArameterS foR whIch The RIGhT side of equatiON Most ClOsely apProXimates $p(n)$. FIrsTly, We cArRied out frEe estimaTiOn WiThOut The [*a pRIori*]{} assuMpTioN oF thE propERties ($H1$) And ($H2$). THe soLuTiON to The optiMIzATIon pRoBlEm is SenSiTive tO the VAluE of $t(n)$ chOsen to iniTiaLIze tHe OpTimizatIon algorithm. SInCe previous TeSts SuggesT THat neithEr strand is significantly FAvored oVer The otHer dUring the cOnsTructiOn oF The seqUence, wE used $T(n)=0.5

and finding properties of $\aleph$. ]{} F rom as su mpti on ( A1), if a nucl e otid e is proposed as a can didat ef or j o in ing t he endo fo n e o fth e s tr a nd s, th enits com plement is au to matically ac an didate for jo ining the be gin ning o fthe other st rand. There f ore, t he probab il i ty ofa nucleo t i de bei ng selected as ac an d idate for join ing on es tr a n d i s t he same as t he pr o babilit y o f i tsc omplement bei ng selected for the o th ers trand, that i s ,\[**($H1$)** ]{}$\mu_A=\m u_T$ a n d $\mu_ C =\mu_G$ . In o the r w ords , $ \m u$ta k est he fo r m $ \mu(m)=( m, 0. 5-m,0 .5-m , m ) $ , wh ere $0\ leq m \leq 0.5$.\Now , si n cethe m atrix  $\a le ph$ i s assu med t obe the same for all genomes, an yseq ue nce p r oduced by th e model is a r e ali za t i o nof a Markov chainbe l o ng ing to a famil y o fM arkov ch ai nspara m e teriz ed b y  $ m$ and $ t$, wh e re $ 0< m< 0 .5 $ and$0 \le q t \leq1 $. Furthe r, from(A2)w e have that th e probabilityo fa ca n dida tenucleotideof t y pe $ j$ b e in g a c cepte d tofo l lo w a nucleotide of ty pe  $i$ a t the end of one s trand is e q u a l to the pro b ab i lity of a nucl eotid e of type$ \alpha(i )$ pr ecedinga nucleot i d e of typ e $ \al pha (j) $ at the beginnin g of t he otherstr and, th atis, th e m at rix $\ale ph$ hasth efo rm \ [ **($H 2 $)**]{}$a _{i j} =a_ {\alp h a(j)\a lpha( i)}$ .\ W e us e the [ * ls q c urve fi t* ]{}fun ct ion o f th e op timizat ion toolb oxi n [M at la b]{} to construct es ti mators for t hematric e s $\aleph (n)$, vectors $\mu(n)$a nd valu es$t(n) $. T he [*lsqc urv efit*] {}f unctio n solv es (i nthe l east- s q ua res s ense) then o nli nearpr oble m of de termining the para m ete rs for whichthe rig h t s ide of equ at i onm o st closely appr oximates $ P( n )$ . Firstly, weca rried o ut free esti m ation w ithout th e [*a pri or i*]{ } ass umption of the pro perties ( $ H1$)a nd ($H2 $). Theso lut ion t o theo pti mizat ion pr ob lem is sens it ive to t he value of $t(n)$ chos en toiniti ali ze the op tim i zat ion algor ithm . Since pr evi ous test s s u ggest tha t n eit h er st rand is signif i ca ntl y fa vored overt h e ot her d uri n g thecons truction of the s e quence, we use d $t ( n )=0 .5

and_finding properties_of $\aleph$.]{} From assumption_(A1), if_a_nucleotide is_proposed_as a candidate_for joining the_end of one of_the strands, then_its_complement is automatically a candidate for joining the beginning of the other strand. Therefore,_the_probability of_a_nucleotide_being selected as a candidate_for joining one strand is_the same_as the probability of its complement being selected_for_the other strand,_that is,\ [**($H1$)**]{} $\mu_A=\mu_T$ and $\mu_C=\mu_G$. In other words, $\mu$_takes the form $\mu(m)=(m,0.5-m,0.5-m,m)$, where $0\leq_m \leq 0.5$.\ Now, since_the_matrix $\aleph$_is assumed to be_the same for all genomes, any_sequence produced by the model is_a realization of a Markov chain belonging_to a family of Markov chains_parameterized by $m$ and $t$, where $0<_m< 0.5$_and $0\leq t\leq 1$. Further, from_(A2) we have_that the_probability of a_candidate nucleotide of type $j$ being accepted_to follow a_nucleotide of type $i$ at the end_of_one strand is_equal_to_the probability_of a nucleotide_of_type $\alpha(i)$_preceding_a nucleotide of type $\alpha(j)$ at_the_beginning of the other strand, that is,_the matrix $\aleph$ has_the_form\ [**($H2$)**]{} $a_{ij}=a_{\alpha(j)\alpha(i)}$.\ We use the_[*lsqcurvefit*]{} function of the optimization_toolbox in [Matlab]{} to construct estimators_for the_matrices $\aleph(n)$,_vectors $\mu(n)$ and values $t(n)$. The [*lsqcurvefit*]{} function solves (in the least-squares_sense) the nonlinear problem of determining_the parameters for which_the right_side_of equation most_closely_approximates $P(n)$. Firstly,_we carried out free estimation without the_[*a priori*]{}_assumption of the properties ($H1$) and_($H2$). The solution to the_optimization_problem is sensitive to the value_of $t(n)$ chosen to initialize the optimization_algorithm. Since previous tests suggest_that_neither_strand is significantly favored over_the other during the construction of_the sequence, we_used $t(n)=0.5

\llx a)\we %\mbox{ and } (c\lle\p(b))]\}.$$ Let $\s\ap\in f_\p\inv(\int(\lag(a))$. Then $f_\p(\s\ap)=\s\in\int(\lag(a))$. Hence $a^*\nin\s$. Now, by \[cluendcor\], there exists $a_1\in A$ such that $a^*\llcr a_1^*$ and $a_1^*\nin\s$. We get that $a_1\llcr a$ and $a_1\in\s\cap\BBBB=\ssi$. Since $a_1\llx a$, there exist $a_2,b\in\BBBB$ such that $a_1\llx a_2\llx b\llx a$. Then, by the definition of the set $\ssi$, $\p(a_2)\in\s\ap$. By \[pf1\](b) (resp., by \[remlc3\]), $\p(a_2)\lle\p(b)$. Now, Proposition \[bbcl1\](a) implies that there exists $c\in\BBBB\ap\cap\s\ap$ such that $c\lle\p(b)$. Thus $\s\ap\in\lbg(c)$, where $c\in\BBBB\ap$, $c\lle\p(b)$ and $b\llx a$. This means that $\s\ap\in\iota_B(D_\p(a))$. Hence, $f_\p\inv(\int(\lag(a))\sbe\iota_B(D_\p(a))$. Conversely, let $\s\ap\in\iota_B(D_\p(a))$ and $\s=f_\p(\s\ap)$. Then there exist $b\in\BBBB$ and $c\in\BBBB\ap$ such that $b\llx a$, $c\lle\p(b)$ and $\s\ap\in\lbg(c)$. Thus $c\in\s\ap$ and hence $\p(b)\in\s\ap$. This implies that $b\in\ssi=\BBBB\cap\s$. Since $b\in\BBBB

\llx a)\we % \mbox { and } (c\lle\p(b))]\}.$$ Let $ \s\ap\in f_\p\inv(\int(\lag(a))$. Then $ f_\p(\s\ap)=\s\in\int(\lag(a))$. Hence $ a^*\nin\s$. Now, by \[cluendcor\ ], there exists $ a_1\in A$ such that $ a^*\llcr a_1^*$ and $ a_1^*\nin\s$. We draw that $ a_1\llcr a$ and $ a_1\in\s\cap\BBBB=\ssi$. Since $ a_1\llx a$, there exist $ a_2,b\in\BBBB$ such that $ a_1\llx a_2\llx b\llx a$. Then, by the definition of the located $ \ssi$, $ \p(a_2)\in\s\ap$. By \[pf1\](b) (resp. , by \[remlc3\ ]), $ \p(a_2)\lle\p(b)$. Now, Proposition \[bbcl1\](a) implies that there exists $ c\in\BBBB\ap\cap\s\ap$ such that $ c\lle\p(b)$. therefore $ \s\ap\in\lbg(c)$, where $ c\in\BBBB\ap$, $ c\lle\p(b)$ and $ b\llx a$. This means that $ \s\ap\in\iota_B(D_\p(a))$. Hence, $ f_\p\inv(\int(\lag(a))\sbe\iota_B(D_\p(a))$. Conversely, let $ \s\ap\in\iota_B(D_\p(a))$ and $ \s = f_\p(\s\ap)$. Then there exist $ b\in\BBBB$ and $ c\in\BBBB\ap$ such that $ b\llx a$, $ c\lle\p(b)$ and $ \s\ap\in\lbg(c)$. therefore $ c\in\s\ap$ and therefore $ \p(b)\in\s\ap$. This implies that $ b\in\ssi=\BBBB\cap\s$. Since $ b\in\BBBB

\llx a)\we %\mbox{ and } (c\lle\p(b))]\}.$$ Let $\r\ap\in f_\p\inv(\int(\lct(a))$. Thei $f_\p(\s\ap)=\a\in\int(\lae(a))$. Hence $a^*\nin\s$. Now, by \[cluendror\], rhere exists $a_1\in A$ such thag $a^*\llcr a_1^*$ wnd $a_1^*\nin\w$. We tet that $a_1\llcr a$ akb $a_1\in\a\gap\BBYB=\wsi$. Since $a_1\llx a$, there efist $a_2,b\in\BBBB$ vuzh that $a_1\llx a_2\llx b\llx a$. Then, by the qefinitooj of the set $\sfi$, $\p(s_2)\yn\s\al$. By \[pf1\](b) (resp., by \[remlc3\]), $\p(a_2)\lle\p(b)$. Now, Proposption \[bbcl1\](a) implirs that there exists $c\in\BBHB\ap\fap\s\ap$ such that $c\ple\p(b)$. Thus $\w\ap\ig\obg(c)$, where $c\kn\BBBB\ap$, $c\lle\p(b)$ and $b\lmx a$. This means that $\s\ap\in\iota_B(A_\p(a))$. Hznce, $f_\p\inv(\ibt(\oag(w))\vbe\iota_B(D_\p(a))$. Ronverfely, let $\s\ap\lm\iota_B(G_\p(a))$ and $\s=f_\p(\s\ap)$. Then tmere xxisr $b\in\BBBB$ and $c\in\BBBB\ep$ such that $b\llx a$, $c\jle\p(b)$ and $\s\cp\in\lbg(c)$. Thus $c\in\s\ap$ qne henwe $\p(t)\in\s\xp$. Tfis ikpmies tjat $b\in\ssi=\BBBG\cap\s$. Since $b\in\BBBB

\llx a)\we %\mbox{ and } (c\lle\p(b))]\}.$$ Let Then Hence $a^*\nin\s$. by \[cluendcor\], there $a^*\llcr and $a_1^*\nin\s$. We that $a_1\llcr a$ $a_1\in\s\cap\BBBB=\ssi$. Since $a_1\llx a$, there exist such that $a_1\llx a_2\llx b\llx a$. Then, by the definition of the set $\p(a_2)\in\s\ap$. By \[pf1\](b) (resp., by \[remlc3\]), $\p(a_2)\lle\p(b)$. Now, Proposition \[bbcl1\](a) implies that there $c\in\BBBB\ap\cap\s\ap$ that Thus where $c\in\BBBB\ap$, $c\lle\p(b)$ and $b\llx a$. This means that $\s\ap\in\iota_B(D_\p(a))$. Hence, $f_\p\inv(\int(\lag(a))\sbe\iota_B(D_\p(a))$. Conversely, let $\s\ap\in\iota_B(D_\p(a))$ and Then there exist $b\in\BBBB$ and $c\in\BBBB\ap$ such that a$, $c\lle\p(b)$ and $\s\ap\in\lbg(c)$. $c\in\s\ap$ and hence $\p(b)\in\s\ap$. This that Since $b\in\BBBB

\llx a)\we %\mbox{ and } (c\lle\p(b))]\}.$$ Let $\s\ap\In f_\p\inv(\int(\Lag(a))$. THen $F_\p(\s\Ap)=\S\in\iNt(\laG(a))$. Hence $a^*\nin\s$. NoW, By \[clUendcor\], there exists $a_1\in A$ Such tHaT $A^*\llcR A_1^*$ aNd $a_1^*\niN\s$. We get THaT $A_1\LlcR a$ AnD $a_1\iN\s\CAp\bBBB=\sSi$. SInce $a_1\llX a$, there exiSt $a_2,B\iN\BBBB$ such thaT $A_1\lLx a_2\llx b\llx A$. ThEn, by the definItiOn of thE sEt $\sSI$, $\p(a_2)\in\S\ap$. by \[pf1\](b) (Resp., by \[REmlc3\]), $\p(a_2)\Lle\p(b)$. Now, PRoPOsitioN \[Bbcl1\](a) imPLIeS thaT there exists $c\in\BBbb\aP\Cap\s\ap$ such that $C\lle\p(b)$. thUS $\s\AP\In\lBg(c)$, Where $c\in\BBbB\Ap$, $c\llE\P(b)$ and $b\lLX a$. tHIS meANs that $\s\ap\in\ioTa_B(D_\p(a))$. Hence, $F_\P\inV(\int(\laG(a))\Sbe\IOta_B(D_\p(A))$. ConvErSEly, Let $\s\ap\in\iotA_B(D_\p(A))$ and $\s=f_\p(\s\aP)$. Then tHEre exisT $B\in\BBBB$ And $c\in\bBBb\ap$ Such THaT $b\Llx A$, $c\LLe\p(B)$ AnD $\s\aP\In\lBg(c)$. Thus $c\In\S\aP$ and hEnce $\P(B)\IN\S\ap$. THis ImplIes thAt $b\in\ssi=\BBBB\cAp\s$. sincE $B\in\bBBB

\llx a)\we %\mbox{ and } ( c\lle\p(b) )]\}. $$Let $ \s\a p\in f_\p\inv(\int ( \lag (a))$. Then $f_\p(\s\a p)=\s \i n \int ( \l ag(a) )$. Hen c e$ a ^*\ ni n\ s$. N o w, by \ [cl uendcor \], thereexi st s $a_1\in A$ su ch that $a ^*\ llcr a_1^*$and $a_1^ *\ nin \ s$. W e g et th at $a_ 1 \llcra$ and $a _1 \ in\s\c a p\BBBB= \ s si $. S ince $a_1\llx a$, th e re exist $a_2, b\in\B BB B $s u chtha t $a_1\llx a _2\ll x b\llxa $. T h en, by the defini tion of the set $\ssi $, $\ p (a_2)\ in\s\ ap $ . B y \[pf1\](b ) (r esp., by\[reml c 3\]), $ \ p(a_2)\ lle\p( b)$ . N ow,P ro po sit io n \[ b bc l1\ ] (a) implies t ha t the re e x i s t s $c \in \BBB B\ap\ cap\s\ap$ suc h t hat$ c\l le\p( b)$.Thus $ \s\ap \in\lb g(c)$ ,where $c\in\BBB B\ap $, $c\lle \p( b) $ a nd $b\l l x a$.Thi s m eans th at $\s\ a p\i n\ i o t a_ B(D_\p(a))$. Hence ,$ f _\ p\inv(\i nt(\la g (a )) \ sbe\iota _B (D_ \p(a ) ) $. C onve r se ly, let$\s\ap \ in \i ota_B(D _\ p(a))$ a nd$\s =f_\p ( \s\a p)$. T hen ther e exi s t $b\in\BBBB$a nd $c\in\BBBB \ ap $ su c h th at$b\llx a$,$c\l l e\p( b)$a nd $\ s \ap\i n\lbg (c ) $. Thus $c\in\s\ap$ an dhence$\p(b )\in\s\ap$. T his implie s t hat $b\i n\ss i =\ B BBB\cap\s$. Si nce $ b\in\BBBB

\llx a)\we %\mbox{_and } (c\lle\p(b))]\}.$$_Let $\s\ap\in f_\p\inv(\int(\lag(a))$. Then_$f_\p(\s\ap)=\s\in\int(\lag(a))$. Hence_$a^*\nin\s$._Now, by_\[cluendcor\],_there exists $a_1\in_A$ such that_$a^*\llcr a_1^*$ and $a_1^*\nin\s$. We_get that $a_1\llcr_a$_and $a_1\in\s\cap\BBBB=\ssi$. Since $a_1\llx a$, there exist $a_2,b\in\BBBB$ such that $a_1\llx a_2\llx b\llx a$._Then,_by the_definition_of_the set $\ssi$, $\p(a_2)\in\s\ap$. By_\[pf1\](b) (resp., by \[remlc3\]), $\p(a_2)\lle\p(b)$._Now, Proposition_\[bbcl1\](a) implies that there exists $c\in\BBBB\ap\cap\s\ap$ such that_$c\lle\p(b)$._Thus $\s\ap\in\lbg(c)$, where_$c\in\BBBB\ap$, $c\lle\p(b)$ and $b\llx a$. This means that $\s\ap\in\iota_B(D_\p(a))$._Hence, $f_\p\inv(\int(\lag(a))\sbe\iota_B(D_\p(a))$. Conversely, let $\s\ap\in\iota_B(D_\p(a))$ and $\s=f_\p(\s\ap)$._Then there exist_$b\in\BBBB$_and_$c\in\BBBB\ap$ such that $b\llx a$,_$c\lle\p(b)$ and $\s\ap\in\lbg(c)$. Thus $c\in\s\ap$ and_hence $\p(b)\in\s\ap$. This implies that $b\in\ssi=\BBBB\cap\s$._Since $b\in\BBBB

, the proof of Lemma \[uniformly conditional convergence\] on the uniform convergence of finite dimensional distributions for bivariate process breaks down when $\alpha_1= 2$ or $\alpha_2= 2$. Studying the asymptotics of (\[bivariate excursion probablity\]) when $\alpha_1= 2$ or/and $\alpha_2= 2$ requires different methods for dealing with differentiable or non-differentiable cases. When both $X_1$ and $X_2$ have twice continuously differentiable sample functions, this problem has been studied by [@Cheng_Xiao_2014]. The authors plan to study the remaining cases in their future work. Second, it would be interesting to study the excursion probabilities when $\{X(t),\,t \in \mathbb{R}^N\}$ is anisotropic or non-stationary, or taking values in $\R^d$ with $d \ge 3$. In the last problem, the covariance and cross-covariance structures become more complicated. We expect that the pairwise maximum cross correlations and the size (e.g., the Lebesgue measure) of the set where all the pairwise cross correlations attain their maximum values (if not empty) will play an important role. The rest of the paper is organized as follows. Section \[sec\_main results and discussion\] states the main theorems with some discussions and provides an application of the main theorems to the bivariate Gaussian fields with Matérn cross-covariances introduced by [@Gneiting_Kleiber_Schlather2010]. We state the key lemmas and provide proofs of our main theorems in Section \[sec\_proof of main results\]. The proofs of the lemmas are given in Section \[sec\_proof of lemmas\]. We end the introduction with some notation. For any $t\in \mathbb{R}^N$, $|t|$ denotes its $l^2$-norm. An integer vector $\mathbf{k}\in \mathbb{Z}^N$ is written as $\mathbf{k}=(k_1,...,k_N)$. For $\mathbf{k}\in \mathbb{Z}^N$ and $T \in \mathbb{R}_+= [0, \infty)$, we define the cube $[\mathbf{k}T,(\mathbf{k}+1)T]:= \prod_{i=1}^N [k_iT,(k_i+1)T]$. For any integer $n$, $mes_n(\

, the proof of Lemma \[uniformly conditional convergence\ ] on the uniform convergence of finite dimensional distributions for bivariate procedure break in down when $ \alpha_1= 2 $ or $ \alpha_2= 2$. Studying the asymptotics of (\[bivariate excursion probablity\ ]) when $ \alpha_1= 2 $ or / and $ \alpha_2= 2 $ requires unlike methods for dealing with differentiable or non - differentiable cases. When both $ X_1 $ and $ X_2 $ have doubly endlessly differentiable sample functions, this problem has been studied by [ @Cheng_Xiao_2014 ]. The authors design to study the remaining cases in their future work. Second, it would be interesting to analyze the excursion probabilities when $ \{X(t),\,t \in \mathbb{R}^N\}$ is anisotropic or non - stationary, or taking values in $ \R^d$ with $ d \ge 3$. In the last trouble, the covariance and cross - covariance structures become more complicated. We expect that the pairwise utmost cross correlations and the size (e.g., the Lebesgue measure) of the set where all the pairwise crisscross correlations attain their maximum values (if not empty) will play an authoritative role. The rest of the paper is organized as follows. Section \[sec\_main results and discussion\ ] states the main theorems with some discussions and provides an application of the main theorems to the bivariate Gaussian fields with Matérn cross - covariance precede by [ @Gneiting_Kleiber_Schlather2010 ]. We state the cardinal lemma and leave proofs of our main theorems in Section \[sec\_proof of main results\ ]. The proofs of the lemmas are given in Section \[sec\_proof of lemmas\ ]. We end the initiation with some notation. For any $ t\in \mathbb{R}^N$, $ |t|$ denotes its $ l^2$-norm. An integer vector $ \mathbf{k}\in \mathbb{Z}^N$ is written as $ \mathbf{k}=(k_1,... ,k_N)$. For $ \mathbf{k}\in \mathbb{Z}^N$ and $ T \in \mathbb{R}_+= [ 0, \infty)$, we define the cube $ [ \mathbf{k}T,(\mathbf{k}+1)T]:= \prod_{i=1}^N [ k_iT,(k_i+1)T]$. For any integer $ n$, $ mes_n(\

, thf proof of Lemma \[uniformuy conditional einvergxnce\] on the uniworm convergence of finite dmmenwionao distributions for bixariate pgocess brwaks eown when $\empha_1= 2$ ov $\alpgw_2= 2$. Scuvying the asympjotics of (\[bieariate excurshov 'robablity\]) when $\alpha_1= 2$ or/and $\alpha_2= 2$ wequirex fifferent methjds gjr dsaling with differentiable or non-djfferenuiable cases. When noth $X_1$ and $X_2$ have twice cojtinkously differentiahle sample dunceuons, this prublem has been studied by [@Cheng_Xiao_2014]. The authors plan go stbdy the remqibinh cases in tieir flture work. Segpnd, it would ne interesting to stydy the excursion prouabilities when $\{X(t),\,t \in \mathbt{R}^U\}$ is anisotropic or nin-wtatimnarf, or rakkng veluss in $\G^d$ xith $d \ge 3$. Jn the last problem, the covarisnsv and cross-cobarianse structures become more complicated. We txpecf that the pairwise maxumum cross correlatiojs and thq size (e.g., the Lebesgue measure) of the set where anl thx oaixqise zeods correlations attain their maximum values (is npt empty) will plcy an important tope. Jhe rest of thg paper is organized as folllws. Secjion \[swc\_main refultx and discussion\] states the main theoreiw with some discusdions and pxovidex an spplication of the main thekrems to thf bivariafd Gaussian fieldr wpth Katérn cross-covariances intwoduced bb [@Gnenting_Klekber_Xchlatrer2010]. We stahe thc key lemmas and prlvide psoofs of okr main theorems in Section \[sec\_'coof of main tesgltv\]. The prjofs pf the lemmas are given in Xection \[sec\_pfoof of lejmas\]. We xnd the intrjduction with dome notatioi. For any $t\in \matybb{R}^N$, $|t|$ aenotes its $l^2$-nprm. An inneyer vectoe $\mathbf{k}\in \mathbb{D}^N$ is sritten as $\mathyy{k}=(j_1,...,k_N)$. For $\mathbf{k}\on \oatrbh{Z}^I$ and $D \in \mathbb{R}_+= [0, \inwty)$, ee dewine tht cmbe $[\matnbf{k}T,(\mathbf{k}+1)T]:= \prod_{i=1}^N [n_iT,(k_j+1)T]$. For any integer $n$, $mes_n(\

, the proof of Lemma \[uniformly conditional the convergence of dimensional distributions for $\alpha_1= or $\alpha_2= 2$. the asymptotics of excursion probablity\]) when $\alpha_1= 2$ or/and 2$ requires different methods for dealing with differentiable or non-differentiable cases. When both and $X_2$ have twice continuously differentiable sample functions, this problem has been studied [@Cheng_Xiao_2014]. authors to the remaining cases in their future work. Second, it would be interesting to study the excursion when $\{X(t),\,t \in \mathbb{R}^N\}$ is anisotropic or non-stationary, taking values in $\R^d$ $d \ge 3$. In the problem, covariance and structures more We expect that pairwise maximum cross correlations and the size (e.g., the Lebesgue measure) of the set where all the cross correlations maximum values not will an important role. of the paper is organized as results and discussion\] states the main theorems with discussions and an application of the main theorems the bivariate Gaussian fields with Matérn cross-covariances introduced [@Gneiting_Kleiber_Schlather2010]. We state the key lemmas and provide proofs of our main theorems in Section main results\]. The proofs the lemmas are in \[sec\_proof lemmas\]. end the with some notation. For any $t\in \mathbb{R}^N$, $|t|$ denotes its $l^2$-norm. integer vector $\mathbf{k}\in \mathbb{Z}^N$ is written as $\mathbf{k}=(k_1,...,k_N)$. For $\mathbf{k}\in $T \mathbb{R}_+= [0, \infty)$, define the cube $[\mathbf{k}T,(\mathbf{k}+1)T]:= [k_iT,(k_i+1)T]$. any integer $n$, $mes_n(\

, the proof of Lemma \[uniformly cOnditional ConveRgeNce\] On The uNifoRm convergence oF FiniTe dimensional distributIons fOr BIvarIAtE procEss breaKS dOWN whEn $\AlPha_1= 2$ Or $\ALpHa_2= 2$. StuDyiNg the asYmptotics oF (\[biVaRiate excursiON pRobablity\]) wHen $\Alpha_1= 2$ or/and $\alPha_2= 2$ RequirEs DifFErent MetHods fOr dealINg with DifferentIaBLe or noN-DiffereNTIaBle cAses. When both $X_1$ and $X_2$ HAvE Twice continuouSly difFeREnTIAblE saMple functiOnS, this PRoblem hAS bEEN StuDIed by [@Cheng_XiaO_2014]. The authors PLan To studY tHe rEMaininG caseS iN TheIr future worK. SecOnd, it woulD be intEResting TO study tHe excuRsiOn pRobaBIlItIes WhEN $\{X(t),\,T \In \MatHBb{R}^n\}$ is anisoTrOpIc or nOn-stATIONary, Or tAkinG valuEs in $\R^d$ with $d \ge 3$. in tHe laST prOblem, The coVariAnCe and Cross-cOvariAnCe structures becOme mOre compliCatEd. we eXpEct thAT the paIrwIse Maximum Cross coRRelAtIONS aNd the size (e.g., the LebeSgUE MeAsure) of tHe set wHErE aLL the pairWiSe cRoss CORrelaTionS AtTain theiR maximUM vAlUes (if noT eMpty) wiLl PlaY an ImporTAnt rOle. The Rest of thE papeR Is organized as fOLlows. Section \[sEC\_mAIN rESultS anD discussion\] StatES the Main THeOreMS with Some dIsCUsSIons and provides an apPlIcatioN of thE main theorems To the bivarIATE GaussiaN fieLDs WIth Matérn cross-CovarIances intrODuced by [@GNeitiNg_KleibeR_SchlatheR2010]. wE state thE keY leMmaS anD PRoVide proofs of oUR Main ThEorems iN SeCtion \[seC\_prOof Of mAin ReSults\]. The pRoofs of tHe LeMmAs Are Given IN Section \[SeC\_prOoF of LemmaS\]. we end tHe intRoduCtIoN WitH some noTAtION. For AnY $t\In \maThbB{R}^n$, $|t|$ denOtes ITs $l^2$-Norm. An iNteger vecTor $\MAthbF{k}\In \Mathbb{Z}^n$ is written as $\mAtHbf{k}=(k_1,...,k_N)$. For $\MaThbF{k}\in \maTHBb{Z}^N$ and $T \In \mathbb{R}_+= [0, \infty)$, we define tHE cube $[\maThbF{k}T,(\maThbf{K}+1)T]:= \prod_{i=1}^N [k_IT,(k_I+1)T]$. For aNy iNTeger $n$, $Mes_n(\

, the proof of Lemma \[uni formly con ditio nal co nv erge nce\ ] on the unifo r m co nvergence of finite di mensi on a l di s tr ibuti ons for bi v a ria te p roc es s b reaks do wn when $\alpha_1 = 2 $or $\alpha_2 = 2 $. Studyin g t he asymptoti csof (\[ bi var i ate e xcu rsion proba b lity\] ) when $\ al p ha_1=2 $ or/an d $\ alph a_2= 2$ requiresd if f erent methodsfor de al i ng w ith di fferentiab le or n o n-diffe r en t i a ble cases. When b oth $X_1$ a n d $ X_2$ h av e t w ice co ntinu ou s lydifferentia blesample fu nction s , thisp roblemhas be enstu died by [ @Ch en g _Xi a o_ 201 4 ].The auth or splanto s t u d y the re main ing c ases in their fu ture wor k. Se cond, itwo uld b e inte resti ng to study the e xcur sion prob abi li tie swhen$ \{X(t) ,\, t \in \ma thbb{R} ^ N\} $i s an isotropic or non-s ta t i on ary, ortaking va lu e s in $\R ^d $ w ith$ d \ge3$.I nthe last probl e m, t he cova ri ance a nd cr oss -cova r ianc e stru ctures b ecome more complicat e d. We expectt ha t th e pai rwi se maximumcros s cor rela t io nsa nd th e siz e( e. g ., the Lebesgue mea su re) of theset where all the pairw i s e cross c orre l at i ons attain the ir ma ximum valu e s (if no t emp ty) will play ani m portantrol e.The re s t o f the paper i s orga ni zed asfol lows. S ect ion \[ sec \_ main resu lts anddi sc us si on\ ] sta t es the m ai n t he ore ms wi t h some disc ussi on sa ndprovide s a n appl ic at ionofth e mai n th e ore ms to t he bivari ate Gaus si an fields with Matérncr oss-covari an ces intro d u ced by [ @Gneiting_Kleiber_Schla t her2010 ].We st atethe key l emm as and pr o vide p roofsof ou rmai n theor e m sinSe ction \[se c \ _pr oof o fmain result s\]. The proofs of the lemmas are g ive n in S ec tio n \ [ sec \_ p roo f of lemmas\]. W e end thein t ro duction wi t h s om e notat ion. Fo r any $t\in \ mathbb{R} ^N$, $|t| $deno t e s i ts $l^2$-n orm. Aninteger v e ctor$ \m athbf {k} \in \m at hbb {Z}^N $ is w r itt en as $\mat hb f{k}=( k_1,. .. ,k_N)$.For $\mathbf{k}\in \mat hbb{Z} ^N$ a nd$T \in \m ath b b{R }_+= [0,\inf ty)$, we d efi nethe c ube $[\ma thbf { k} T,( \ mathb f{k} + 1)T]:= \p r od _{i = 1 }^ N [k_iT,(k_ i + 1 )T] $. Fo r a n y inte ger$n$, $mes_n(\

, the_proof of_Lemma \[uniformly conditional convergence\]_on the_uniform_convergence of_finite_dimensional distributions for_bivariate process breaks_down when $\alpha_1= 2$_or $\alpha_2= 2$._Studying_the asymptotics of (\[bivariate excursion probablity\]) when $\alpha_1= 2$ or/and $\alpha_2= 2$ requires different_methods_for dealing_with_differentiable_or non-differentiable cases. When both_$X_1$ and $X_2$ have twice_continuously differentiable_sample functions, this problem has been studied by_[@Cheng_Xiao_2014]._The authors plan_to study the remaining cases in their future work._Second, it would be interesting to_study the excursion_probabilities_when_$\{X(t),\,t \in \mathbb{R}^N\}$_is anisotropic or non-stationary, or taking_values in $\R^d$ with $d \ge_3$. In the last problem, the covariance_and cross-covariance structures become more complicated._We expect that the pairwise_maximum cross_correlations and the size (e.g.,_the Lebesgue measure)_of the_set where all_the pairwise cross correlations attain their_maximum values (if_not empty) will play an important_role._The rest of_the_paper_is organized_as follows. Section_\[sec\_main_results and_discussion\]_states the main theorems with some_discussions_and provides an application of the main_theorems to the bivariate_Gaussian_fields with Matérn cross-covariances_introduced by [@Gneiting_Kleiber_Schlather2010]. We state_the key lemmas and provide proofs_of our_main theorems_in Section \[sec\_proof of main results\]. The proofs of the lemmas_are given in Section \[sec\_proof of_lemmas\]. We end the introduction_with some_notation._For any $t\in_\mathbb{R}^N$,_$|t|$ denotes_its $l^2$-norm. An integer vector $\mathbf{k}\in \mathbb{Z}^N$_is written_as $\mathbf{k}=(k_1,...,k_N)$. For $\mathbf{k}\in \mathbb{Z}^N$ and_$T \in \mathbb{R}_+= [0,_\infty)$,_we define the cube $[\mathbf{k}T,(\mathbf{k}+1)T]:= \prod_{i=1}^N [k_iT,(k_i+1)T]$._For any integer $n$, $mes_n(\

{u} := P_{\mathcal{U}}(u) \qquad \overline{x} := h(\overline{u})}_{\text{physical system}} \,. \end{aligned}$$ The system  can be implemented in closed loop with a physical system and approximates a *projected Newton flow* [@hauswirthProjectedDynamicalSystems2018 Ex. 5.6]. This fact is noteworthy, because, in general, projected Newton flows do not lend themselves to an easy implementation (e.g., as an iterative algorithm). Even though, as seen in \[fig:gradient\_sim\], ${\overline{u}}$ converges to the optimizer of , strictly speaking, Theorem  is not directly applicable because $Q \neq {\mathbb{I}}$. ![Convergence behavior of , , and  for a problem instance of  with $p = 100$ (input dimension) and $r = 300$ (\# of input constraints).[]{data-label="fig:gradient_sim"}](Figs/subplots_grad.pdf){width=".9\columnwidth"} The anti-windup gradient and Newton schemes defined above illustrate some of the key features of autonomous optimization and anti-windup implementations: 1. Under the conditions of \[thm:monot\], the actual system state and saturated control input converge to the optimizer $u^\star$ of , even though the internal control variable $u$ does not in general converge to $u^\star$. 2. In a feedback implementation exploiting input saturation, neither the set ${\mathcal{U}}$ nor the steady-state disturbance $w$ needs to be known (or estimated). The only model information required is $H$. Furthemore, recent preliminary theoretical [@colombinorobustnessguaranteesfeedbackbased2019] and experimental results for power systems [@ortmannExperimentalValidationFeedback2019] suggest that these feedback schemes are robust against uncertainties in $H$. 3. The simulations in \[fig:gradient\_sim\] suggest that the convergence rate of the “projected trajectory” of  is not affected by the value of $K$ and is equivalent to the convergence rate of the nominal projected gradient flow. In contrast, the convergence rate of the anti-windup Newton scheme  does depend on $K$ and one can recover the rate of projected Newton flow only in the limit $K \rightarrow 0^+$. An analysis of this observation

{ u }: = P_{\mathcal{U}}(u) \qquad \overline{x }: = h(\overline{u})}_{\text{physical system } } \, . \end{aligned}$$ The system   can be implemented in shut loop topology with a physical system and approximate a * stick out Newton flow *   [ @hauswirthProjectedDynamicalSystems2018 Ex.   5.6 ]. This fact is noteworthy, because, in general, project Newton flow do not lend themselves to an easy execution (for example, as an iterative algorithm). Even though, as seen in \[fig: gradient\_sim\ ], $ { \overline{u}}$ converges to the optimizer of  , strictly address, Theorem   is not directly applicable because $ Q \neq { \mathbb{I}}$. ! [ Convergence behavior of  ,  , and   for a problem case of   with $ p = 100 $ (input dimension) and $ r = 300 $ (\ # of input constraints).[]{data - label="fig: gradient_sim"}](Figs / subplots_grad.pdf){width=".9\columnwidth " } The anti - windup gradient and Newton scheme defined above illustrate some of the key feature of autonomous optimization and anti - windup implementations: 1. Under the conditions of \[thm: monot\ ], the actual system state and saturate control input converge to the optimizer $ u^\star$ of  , even though the internal control variable $ u$ does not in cosmopolitan converge to $ u^\star$. 2. In a feedback implementation exploiting input saturation, neither the set $ { \mathcal{U}}$ nor the steady - department of state disturbance $ w$ needs to be known (or estimated). The only model information required is $ H$. Furthemore, recent preliminary theoretical   [ @colombinorobustnessguaranteesfeedbackbased2019 ] and experimental results for ability systems   [ @ortmannExperimentalValidationFeedback2019 ] suggest that these feedback system are full-bodied against uncertainties in $ H$. 3. The simulations in \[fig: gradient\_sim\ ] suggest that the convergence rate of the “ projected trajectory ” of   is not affected by the value of $ K$ and is equivalent to the convergence pace of the nominal projected gradient flow. In contrast, the convergence rate of the anti - windup Newton scheme   does depend on $ K$ and one can recover the rate of projected Newton flow merely in the limit $ kilobyte \rightarrow 0^+$. An analysis of this observation

{u} := O_{\mathcal{U}}(u) \qquad \overlint{x} := h(\overline{u})}_{\texj{pyysican systsm}} \,. \end{aligned}$$ The dywtem  xan be implemented in zlosed lolp with q phbsical system anv approxliatea a *pxooected Newton fkow* [@hauswirdhProjectedDyndmkcclSystems2018 Ex. 5.6]. This fact is noteworthy, becausr, ln general, protecttd Gewtkn flows do not lend themselves to an easj implementation (r.g., as an iterative algoritjm). Even though, as sfen in \[fig:gtzdiqbt\_sim\], ${\overlive{u}}$ converges to the oktimizer of , strictly speaking, Thdorem  is not ditzxtlj applicable becalse $Q \neq {\matmnb{I}}$. ![Coneergencr behavior of , , snd  foe a problem instance mf  with $p = 100$ (input qimension) aud $r = 300$ (\# of input consrrqints).[]{gata-nabeu="dig:erasixnt_aim"}](Figd/suuplots_grad.psf){width=".9\colunnwidth"} The anti-windip tradient and Hewton sshemes defined above illustrate some of ths key features of autonimous optimization anf anti-winqup implementations: 1. Under the conditions of \[thm:mmnot\], vhd aebmal rtshem state and saturated control input convergq tp nhe optimizer $u^\stcr$ of , even thougn hhr internal conjrol vaxjagle $u$ does not in heneral convwrge to $u^\ftar$. 2. In a feedback implementatuon exploitiug unput saturation, nzither the szt ${\matncal{U}}$ nor the steady-state dirturgance $w$ neefs to be ivown (or estimatea). Tme mnly modtu information reqtired is $I$. Furchemore, fecemt prejiminary tjeorebhcal [@colombinorobushnessyuaratteesfeedbwckbased2019] and experimental resulvx for power xyvteks [@ortmanuExperlmentalValidatijnFeedback2019] sugyest thac thesd feedback schemev are robuse against uncattainties in $I$. 3. The siiularionw in \[fie:eradient\_sim\] subgest than uhe convertence rate of the “irojezfed trajectory” if  us not affected by thq nalne of $N$ and is equhvaldnt yo thd convergenge fate of the nominal projactes gradient flow. In cjntrast, jhe convewgence rate og the anti-windup Ntwton vchxme  dors qepend on $K$ and one can recoved the ratf on projected Nqwtok fljw only in the limit $K \rightarrow 0^+$. An analysis of this observation

{u} := P_{\mathcal{U}}(u) \qquad \overline{x} := h(\overline{u})}_{\text{physical \end{aligned}$$ system can implemented in closed and a *projected Newton [@hauswirthProjectedDynamicalSystems2018 Ex. 5.6]. fact is noteworthy, because, in general, Newton flows do not lend themselves to an easy implementation (e.g., as an algorithm). Even though, as seen in \[fig:gradient\_sim\], ${\overline{u}}$ converges to the optimizer of strictly Theorem not applicable because $Q \neq {\mathbb{I}}$. ![Convergence behavior of , , and for a problem instance of $p = 100$ (input dimension) and $r = (\# of input constraints).[]{data-label="fig:gradient_sim"}](Figs/subplots_grad.pdf){width=".9\columnwidth"} anti-windup gradient and Newton schemes above some of key of optimization and anti-windup 1. Under the conditions of \[thm:monot\], the actual system state and saturated control input converge to the $u^\star$ of though the control $u$ not in general $u^\star$. 2. In a feedback implementation neither the set ${\mathcal{U}}$ nor the steady-state disturbance needs to known (or estimated). The only model required is $H$. Furthemore, recent preliminary theoretical [@colombinorobustnessguaranteesfeedbackbased2019] experimental results for power systems [@ortmannExperimentalValidationFeedback2019] suggest that these feedback schemes are robust against uncertainties 3. The simulations in suggest that the rate the trajectory” is not by the value of $K$ and is equivalent to the convergence of the nominal projected gradient flow. In contrast, the convergence the Newton scheme does on $K$ and one recover rate of projected Newton in limit An of observation

{u} := P_{\mathcal{U}}(u) \qquad \overline{x} := H(\overline{u})}_{\Text{pHysIcaL sYsteM}} \,. \end{Aligned}$$ The systEM  can Be implemented in closed lOop wiTh A PhysICaL systEm and apPRoXIMatEs A *pRojEcTEd newtoN flOw* [@hauswIrthProjecTeddyNamicalSysteMS2018 EX. 5.6]. This fact iS noTeworthy, becaUse, In geneRaL, prOJecteD NeWton fLows do NOt lend ThemselveS tO An easy IMplemenTATiOn (e.g., As an iterative algoRItHM). Even though, as sEen in \[fIg:GRaDIEnt\_Sim\], ${\Overline{u}}$ cOnVergeS To the opTImIZER of , STrictly speakiNg, Theorem  is NOt dIrectlY aPplICable bEcausE $Q \NEq {\mAthbb{I}}$. ![ConveRgenCe behavioR of , , and  FOr a probLEm instaNce of  wIth $P = 100$ (inPut dIMeNsIon) AnD $R = 300$ (\# of INpUt cONstRaints).[]{daTa-LaBel="fiG:graDIENT_sim"}](figS/subPlots_Grad.pdf){width=".9\cOluMnwiDTh"} THe antI-windUp grAdIent aNd NewtOn schEmEs defined above iLlusTrate some Of tHe Key FeAtureS Of autoNomOus OptimizAtion anD AntI-wINDUp Implementations: 1. UndEr THE cOnditionS of \[thm:MOnOt\], THe actual SySteM staTE And saTuraTEd Control iNput coNVeRgE to the oPtImizer $U^\sTar$ Of , eVen thOUgh tHe inteRnal contRol vaRIable $u$ does not iN General converGE tO $U^\StAR$. 2. In a FeeDback implemEntaTIon eXploITiNg iNPut saTuratIoN, NeITher the set ${\mathcal{U}}$ nOr The steAdy-stAte disturbancE $w$ needs to bE KNOwn (or estImatED). THE only model infoRmatiOn required IS $H$. FurtheMore, rEcent preLiminary tHEOretical [@ColOmbInoRobUSTnEssguaranteesFEEdbaCkBased2019] anD exPerimenTal ResUltS foR pOwer systeMs [@ortmanNEXpErImEntAlValIDationFeEdBacK2019] sUggEst thAT these FeedbAck sChEmES arE robust AGaINSt unCeRtAintIes In $h$. 3. The sImulATioNs in \[fig:Gradient\_sIm\] sUGgesT tHaT the conVergence rate oF tHe “projecteD tRajEctory” OF  Is not affEcted by the value of $K$ and is EQuivaleNt tO the cOnveRgence ratE of The nomInaL ProjecTed graDient FlOw. IN COntraST, ThE coNvErgence ratE OF thE anti-WiNdup newton sCheme  does depend on $K$ ANd oNe can recover tHe rAte oF PRoJecTEd nEwtOn FLow ONLy in the limit $K \riGhtarrow 0^+$. An AnALySis of this oBSerVaTion

{u} := P_{\mathcal{U}}(u)\qquad \ov erlin e{x } : =h(\o verl ine{u})}_{\tex t {phy sical system}} \,. \end{ aligned } $$ T hesy st em c a nbe im ple mentedin closedloo pwith a physi c al system an d a pproximatesa * projec te d N e wtonflo w* [@ hauswi r thProj ectedDyna mi c alSyst e ms2018E x .5.6] . This fact is no t ew o rthy, because, in ge ne r al , pro jec ted Newton f lowsd o not l e nd t h ems e lves to an ea sy implemen t ati on (e. g. , a s an it erati ve alg orithm). E ven thoug h, ass een in\ [fig:gr adient \_s im\ ], $ { \o ve rli ne { u}} $ c onv e rge s to the o pt imize r of , s tric tly spe aking , Theorem  is no t di r ect ly ap plica blebe cause $Q \n eq {\ ma thbb{I}}$. ![C onve rgence be hav io r o f, , a n d  for apro blem in stanceo f wi t h $p = 100$ (input dim en s i on ) and $r = 300 $ ( \# of input c ons trai n t s).[] {dat a -l abel="fi g:grad i en t_ sim"}]( Fi gs/sub pl ots _gr ad.pd f ){wi dth=". 9\column width " } The anti-wi n dup gradienta nd N ew t on s che mes defined abo v e il lust r at e s o me of theke y f e atures of autonomou soptimi zatio n and anti-wi ndup imple m e n tations: 1. U n der the condit ionsof \[thm:m o not\], t he ac tual sys tem state a nd satur ate d c ont rol i np ut converge t o theop timizer $u ^\star$ of  ,eve n t ho ugh the i nternalco nt ro lvar iable $u$ does n otin ge neral conver ge to $u^ \s ta r $. 2. In af e edba ck i mple men ta tionexpl o iti ng inpu t saturat ion , nei th er the se t ${\mathcal{ U} }$ nor the s tea dy-sta t e disturb ance $w$ needs to be kn o wn (orest imate d).The onlymod el inf orm a tion r equire d is$H $.F u rthem o r e, re ce nt prelimi n a rytheor et ical  [@colo mbinorobustnessgua r ant eesfeedbackba sed 2019 ] an d e x pe r ime nt a l r e s ults for powersystems [@ or t ma nnExperime n tal Va lidatio nFeedba ck201 9 ] sugge st that t hese feed ba ck s c h eme s are robu st again st uncert a intie s i n $H$ . 3. Th esim ulati ons in \[f ig:gr adient \_ sim\]sugge st that th e convergence rate of t he “pr oject edtrajector y”o f is not af fect ed by theval ueof $K $ a n d isequi v al ent to th e co n vergencer at e o f th e nominal p r o j ect ed gr adi e nt flo w. I n contrast, the c o nvergence rate oft h e a nti - wind up Newton scheme   d oe s depend o n$K$ and one can rec ov e r the rateof pro jectedN e wt o n flow onl y i n the lim it$K \righta rr ow 0^+$.An a na lysisof thi s obs e r vation

{u} :=_P_{\mathcal{U}}(u) \qquad_\overline{x} := h(\overline{u})}_{\text{physical system}}_\,. __ __ _ _ _ \end{aligned}$$_The_system  can be implemented in closed loop with a physical system and approximates a_*projected_Newton flow* [@hauswirthProjectedDynamicalSystems2018_Ex. 5.6]._This_fact is noteworthy, because, in_general, projected Newton flows do_not lend_themselves to an easy implementation (e.g., as an_iterative_algorithm). _ Even though, as seen in \[fig:gradient\_sim\], ${\overline{u}}$ converges_to the optimizer of , strictly speaking,_Theorem  is not_directly_applicable_because $Q \neq {\mathbb{I}}$. ![Convergence_behavior of , , and  for a problem_instance of  with $p = 100$_(input dimension) and $r = 300$ (\#_of input constraints).[]{data-label="fig:gradient_sim"}](Figs/subplots_grad.pdf){width=".9\columnwidth"} The anti-windup gradient and_Newton schemes defined above illustrate_some of_the key features of autonomous_optimization and anti-windup_implementations: 1. _Under the conditions_of \[thm:monot\], the actual system state_and saturated control_input converge to the optimizer $u^\star$_of ,_even though the_internal_control_variable $u$_does not in_general_converge to_$u^\star$. 2._ In a feedback implementation exploiting_input_saturation, neither the set ${\mathcal{U}}$ nor the_steady-state disturbance $w$ needs_to_be known (or estimated)._The only model information required_is $H$. Furthemore, recent preliminary theoretical [@colombinorobustnessguaranteesfeedbackbased2019]_and experimental_results for_power systems [@ortmannExperimentalValidationFeedback2019] suggest that these feedback schemes are robust against uncertainties_in $H$. 3. The simulations in_\[fig:gradient\_sim\] suggest that the_convergence rate_of_the “projected trajectory”_of _is not_affected by the value of $K$ and_is equivalent_to the convergence rate of the_nominal projected gradient flow._In_contrast, the convergence rate of the_anti-windup Newton scheme  does depend on_$K$ and one can recover_the_rate_of projected Newton flow only_in the limit $K \rightarrow 0^+$._An analysis of_this observation

$. Since, as we saw in Proposition \[det\], the determinant is greater than zero in $ML(n)$, then, in fact, $ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$, where $SO(n)$ is the group of the special orthogonal matrices, that is, the matrices $A\in O(n)$ such that $\det A=1$. Thus, $ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$ can be viewed as the subset of rotations which consists of the rotations by an acute or right angles. The next result demonstrates that the resolvent average of such rotations is again such a rotation. This, of course, fails when taking the arithmetic average. \[rotations\] ***(being a rotation by an acute or right angle is recessive)*** Suppose that for each $i\in I$, $A_i\in ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. Then ${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\in ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. By employing the inversion formula  (which, in this case, was seen to be a classical inversion), we obtain $$\big({\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\big)^{-1} = {\ensuremath{\mathcal{R}({\bf A}^{-1},{\boldsymbol \lambda})}}=\mathcal{R}({\bf A}^{\intercal},\lambda) ={\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}^{\intercal}.$$ That is, ${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}$ is orthogonal. The fact that $\det{\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}=1$ follows from the monotonicity of ${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}$ (see Proposition \[det\]). When $n=2$, we consider the case where $A_1={\ensuremath{\operatorname{Id}}}$, $A_2$ is the counter clockwise rotation by $\frac{\pi}{2}$, $0<\lambda<1$, $\lambda_1=\lambda

$. Since, as we saw in Proposition   \[det\ ], the determinant is greater than zero in $ ML(n)$, then, in fact, $ ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$, where $ SO(n)$ is the group of the special extraneous matrix, that is, the matrices $ A\in O(n)$ such that $ \det A=1$. Thus, $ ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$ can be viewed as the subset of rotation which consist of the rotations by an acute or correct slant. The next result demonstrates that the resolvent median of such rotations is again such a rotation. This, of course, fail when taking the arithmetic average. \[rotations\ ] * * * (being a rotation by an acute or right angle is recessive) * * * think that for each $ i\in I$, $ A_i\in ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. Then $ { \ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\in ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. By employing the inversion formula   (which, in this character, was seen to be a classical inversion), we obtain $ $ \big({\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\big)^{-1 } = { \ensuremath{\mathcal{R}({\bf A}^{-1},{\boldsymbol \lambda})}}=\mathcal{R}({\bf A}^{\intercal},\lambda) = { \ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}^{\intercal}.$$ That is, $ { \ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}$ is extraneous. The fact that $ \det{\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}=1 $ follows from the monotonicity of $ { \ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}$ (learn Proposition   \[det\ ]). When $ n=2 $, we consider the case where $ A_1={\ensuremath{\operatorname{Id}}}$, $ A_2 $ is the counter clockwise rotation by $ \frac{\pi}{2}$, $ 0<\lambda<1 $, $ \lambda_1=\lambda

$. Sijce, as we saw in Proposiuion \[det\], the determinant ms greafer than zero in $ML(n)$, then, in fact, $ML(i)\cap O(n,{\enwuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensurvmath{\mathvb R}})$, qhere $SO(n)$ ma the gvjup kn the wpecial orthogpnal matriwes, that is, tha oacrices $A\in O(n)$ such that $\det A=1$. Thus, $MJ(n)\cap O(m,{\ejsuremath{\mathbf R}})=MK(g)\cap SO(n,{\ensuremath{\mathbb R}})$ can be viewsd as tie subset of royations which consists of hhe gotations by an ackte or righj andoes. The next result demonstrates tgat the resolvent average of suzh rocations is qgqin vuch a rotavion. Tris, of coursc, fails when tsking the aritmmetir avwrage. \[rotations\] ***(being e rotation by an acuje or righd cngle is recessive)*** Suppise tvat xor dqch $i\ih M$, $A_j\in ML(j)\ca' O(n,{\ensuremzth{\mathbb R}})=NL(n)\cap SO(n,{\ensurematn{\mwnnbb R}})$. Then ${\ehsuremwtr{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\in ML(n)\cap O(n,{\snsuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. By empljying the inversion formula  (which, in this case, wav seei go yc x cpassical inversion), we obtain $$\big({\ensuremath{\matrdak{R}({\nf A},{\boldsymbol \lcmbda})}}\big)^{-1} = {\ensurekahh{\kwthcal{R}({\bf A}^{-1},{\boudsymbol \mambda})}}=\mathcal{R}({\bf A}^{\lntercaj},\lambea) ={\ensuremwth{\msthcal{R}({\bf A},{\boldsymbol \lambdq})}}^{\intercal}.$$ Thct us, ${\ensuremath{\mathccl{R}({\bf A},{\boldsvmbol \kambds})}}$ is orthogonal. The facc that $\det{\ensuremwth{\mathcam{F}({\bf A},{\boldsymbol \uamnda})}}=1$ follows from the monotonisity of ${\eisurekath{\matfcal{T}({\bf A},{\bojdsymbol \lwmbda})}}$ (see Proposition \[det\]). Ahen $u=2$, we wonsider tje case where $A_1={\ensuremath{\operatorname{Id}}}$, $A_2$ is jhe colnter cloekwise rotation by $\srac{\pi}{2}$, $0<\lambda<1$, $\kambda_1=\lcmbda

$. Since, as we saw in Proposition determinant greater than in $ML(n)$, then, SO(n,{\ensuremath{\mathbb where $SO(n)$ is group of the orthogonal matrices, that is, the matrices O(n)$ such that $\det A=1$. Thus, $ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$ can be as the subset of rotations which consists of the rotations by an acute right The result that the resolvent average of such rotations is again such a rotation. This, of course, fails taking the arithmetic average. \[rotations\] ***(being a rotation an acute or right is recessive)*** Suppose that for $i\in $A_i\in ML(n)\cap R}})=ML(n)\cap R}})$. ${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\in O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. By employing the inversion formula (which, in this case, was seen to be classical inversion), $$\big({\ensuremath{\mathcal{R}({\bf A},{\boldsymbol = A}^{-1},{\boldsymbol A}^{\intercal},\lambda) ={\ensuremath{\mathcal{R}({\bf A},{\boldsymbol is, ${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}$ is orthogonal. $\det{\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}=1$ follows from the monotonicity of A},{\boldsymbol \lambda})}}$ Proposition \[det\]). When $n=2$, we consider case where $A_1={\ensuremath{\operatorname{Id}}}$, $A_2$ is the counter clockwise by $\frac{\pi}{2}$, $0<\lambda<1$, $\lambda_1=\lambda

$. Since, as we saw in Proposition \[Det\], the deteRminaNt iS grEaTer tHan zEro in $ML(n)$, then, in FAct, $Ml(n)\cap O(n,{\ensuremath{\mathbB R}})=ML(n)\CaP sO(n,{\eNSuRematH{\mathbb r}})$, WhERE $SO(N)$ iS tHe gRoUP oF the sPecIal orthOgonal matrIceS, tHat is, the matrICeS $A\in O(n)$ such ThaT $\det A=1$. Thus, $ML(n)\Cap o(n,{\ensuReMatH{\MathbB R}})=Ml(n)\cap sO(n,{\ensURemath{\Mathbb R}})$ caN bE Viewed AS the subSET oF rotAtions which consisTS oF The rotations by An acutE oR RiGHT anGleS. The next reSuLt demONstrateS ThAT THe rESolvent averagE of such rotaTIonS is agaIn SucH A rotatIon. ThIs, OF coUrse, fails whEn taKing the arIthmetIC averagE. \[RotatioNs\] ***(beinG a rOtaTion BY aN aCutE oR RigHT aNglE Is rEcessive)*** suPpOse thAt foR EACH $i\in i$, $A_i\In ML(N)\cap O(N,{\ensuremath{\maThbB R}})=ML(N)\Cap sO(n,{\enSuremAth{\mAtHbb R}})$. THen ${\ensUremaTh{\Mathcal{R}({\bf A},{\boldSymbOl \lambda})}}\iN ML(N)\cAp O(N,{\eNsureMAth{\matHbb r}})=ML(N)\cap SO(n,{\EnsuremATh{\mAtHBB r}})$. BY employing the inverSiON FoRmula  (whiCh, in thIS cAsE, Was seen tO bE a cLassICAl invErsiON), wE obtain $$\bIg({\ensuREmAtH{\mathcaL{R}({\Bf A},{\bolDsYmbOl \lAmbda})}}\BIg)^{-1} = {\enSuremaTh{\mathcaL{R}({\bf A}^{-1},{\BOldsymbol \lambdA})}}=\Mathcal{R}({\bf A}^{\inTErCAL},\lAMbda) ={\EnsUremath{\mathCal{R}({\BF A},{\boLdsyMBoL \laMBda})}}^{\inTercaL}.$$ THAt IS, ${\ensuremath{\mathcal{R}({\Bf a},{\boldsYmbol \Lambda})}}$ is orthoGonal. The faCT THat $\det{\enSureMAtH{\Mathcal{R}({\bf A},{\bolDsymbOl \lambda})}}=1$ foLLows from The moNotoniciTy of ${\ensurEMAth{\mathcAl{R}({\Bf A},{\BolDsyMBOl \Lambda})}}$ (see PropOSItioN \[dEt\]). When $n=2$, We cOnsider The CasE whEre $a_1={\eNsuremath{\OperatorNaMe{id}}}$, $a_2$ iS thE counTEr clockwIsE roTaTioN by $\frAC{\pi}{2}$, $0<\lamBda<1$, $\laMbda_1=\LaMbDA

$. Since, as we saw in Pro position \ [det\ ],the d eter mina nt is greatert hanzero in $ML(n)$, then, in f ac t , $M L (n )\cap O(n,{\ e ns u r ema th {\ mat hb b R }})=M L(n )\cap S O(n,{\ensu rem at h{\mathbb R} } )$ , where $S O(n )$ is the gr oup of th espe c ial o rth ogona l matr i ces, t hat is, t he matric e s $A\in O (n )$ s uch that $\det A= 1 $. Thus, $ML(n)\c ap O(n ,{ \ en s u rem ath {\mathbb R }} )=ML( n )\cap S O (n , { \ ens u remath{\mathb b R}})$ can beviewed a s t h e subs et of r o tat ions whichcons ists of t he rot a tions b y an acu te orrig htangl e s. T hene x t r e su ltd emo nstrates t ha t the res o l v e nt a ver ageof su ch rotationsisagai n su ch arotat ion. T his,of cou rse,fa ils when taking the arithmet icav era ge . \[ r otatio ns\ ] * **(bein g a rot a tio nb y an acute or right an gl e is recessi ve)*** Su pp o se thatfo r e ach$ i \in I $, $ A _i \in ML(n )\capO (n ,{ \ensure ma th{\ma th bbR}} )=ML( n )\ca p SO(n ,{\ensur emath { \mathbb R}})$. Then ${\ensur e ma t h {\ m athc al{ R}({\bf A}, {\bo l dsym bol\ la mbd a })}}\ in ML (n ) \c a p O(n,{\ensuremath{ \m athbbR}})= ML(n)\cap SO( n,{\ensure m a t h{\mathb b R} } )$ . By employing theinversionf ormula (whic h, in th is case,w a s seen t o b e a cl ass i c al inversion),w e obt ai n $$\bi g({ \ensure mat h{\ mat hca l{ R}({\bf A },{\bold sy mb ol \ lam bda}) } }\big)^{ -1 } = { \en surem a th{\ma thcal {R}( {\ bf A}^ {-1},{\ b ol d s ymbo l\l ambd a}) }} =\mat hcal { R}( {\bf A} ^{\interc al} , \lam bd a) ={\ens uremath{\math ca l{R}({\bfA} ,{\ boldsy m b ol \lamb da})}}^{\intercal}.$$ T h at is,${\ ensur emat h{\mathca l{R }({\bf A} , {\bold symbol \lam bd a}) } } $ iso r th ogo na l. The fac t tha t $\d et {\en suremat h{\mathcal{R}({\bf A}, {\boldsymbol\la mbda } ) }} =1$ fo l low sf rom t he monotonicity of ${\ens ur e ma th{\mathca l {R} ({ \bf A}, {\bolds ymbol \lambda })}}$ (se e Proposi ti on \ [ d et\ ]). When$n=2$, w e conside r thec as e whe re$A_1={ \e nsu remat h{\ope r ato rname {Id}}} $, $A_2$ is t he counter clockwise rotation by$\frac {\pi} {2} $, $0<\la mbd a <1$ , $\lambd a_1= \lambda

$. Since,_as we_saw in Proposition \[det\], the_determinant is_greater_than zero_in_$ML(n)$, then, in_fact, $ML(n)\cap O(n,{\ensuremath{\mathbb_R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$, where_$SO(n)$ is the_group_of the special orthogonal matrices, that is, the matrices $A\in O(n)$ such that $\det_A=1$._Thus, $ML(n)\cap_O(n,{\ensuremath{\mathbb_R}})=ML(n)\cap_SO(n,{\ensuremath{\mathbb R}})$ can be viewed_as the subset of rotations_which consists_of the rotations by an acute or right_angles._The next result_demonstrates that the resolvent average of such rotations is_again such a rotation. This, of_course, fails when_taking_the_arithmetic average. \[rotations\] ***(being a_rotation by an acute or right_angle is recessive)*** Suppose that for_each $i\in I$, $A_i\in ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap_SO(n,{\ensuremath{\mathbb R}})$. Then ${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\in_ML(n)\cap O(n,{\ensuremath{\mathbb R}})=ML(n)\cap SO(n,{\ensuremath{\mathbb R}})$. By_employing the_inversion formula  (which, in this_case, was seen_to be_a classical inversion),_we obtain $$\big({\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}\big)^{-1} =_{\ensuremath{\mathcal{R}({\bf A}^{-1},{\boldsymbol \lambda})}}=\mathcal{R}({\bf_A}^{\intercal},\lambda) ={\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}^{\intercal}.$$ That is, ${\ensuremath{\mathcal{R}({\bf_A},{\boldsymbol_\lambda})}}$ is orthogonal._The_fact_that $\det{\ensuremath{\mathcal{R}({\bf_A},{\boldsymbol \lambda})}}=1$ follows_from_the monotonicity_of_${\ensuremath{\mathcal{R}({\bf A},{\boldsymbol \lambda})}}$ (see Proposition \[det\]). When $n=2$,_we_consider the case where $A_1={\ensuremath{\operatorname{Id}}}$, $A_2$ is_the counter clockwise rotation_by_$\frac{\pi}{2}$, $0<\lambda<1$, $\lambda_1=\lambda

Akk_etal18; @SokBasHaj19; @Cas_etal19]. The MsRSB method was proposed in the context of FV simulation for fluid flow in highly heterogeneous porous media [@MsRSB_Moyner2016]. Based on a two-grid approach, the MsRSB method constructs multiscale basis functions through restricted smoothing on the fine-scale matrix. In more detail, the basis functions, which are consistent with the local differential operators, are constructed with a cheap relaxation scheme, i.e. a weighted Jacobi iteration, similar to approaches used in smoothed aggregation multigrid methods [@VanManBre96; @VanManBre01; @Bre_etal05]. An important advantage of MsRSB is that smoothing by relaxation provides a great deal of flexibility in handling unstructured grids, an essential requirement, for example, in applications involving complex geological structures. MsRSB has been widely proven and implemented in open source and commercial simulators using a linear two-point flux approximation (TPFA) [@Lie_etal16]. Because of the two-point structure, the linear TPFA scheme is monotone [@Dro14], i.e. it preserves the positivity of the differential solution [@BerPle94], and leads to an M-matrix with a small stencil. This is the reason why linear TPFA is the scheme of choice in most engineering software. Unfortunately, the consistency of TPFA is not guaranteed for arbitrary grids and anisotropic permeability distributions, potentially leading to inaccurate results [@MRST]. Therefore, other FV methods such as multipoint flux approximation (MPFA) and/or nonlinear schemes [@Dro14; @TerMalTch17] must be considered to achieve consistent fluxes. To date, MsRSB has not been combined with MPFA or other consistent discretizations. Hence, in this paper, we focus on enhancing MsRSB to enable the solution of second-order elliptic problems using discretization methods that do not result in an M-matrix. Based on the MPFA-O method [@MPFA_Aavatsmark], we show that the MsRSB basis construction as presented in [@MsRSB_Moyner2016] can fail due to divergent iterations for an anisotropic diffusion problem. We propose a variant of the original MsRSB approach that restores the desired behavior by enforcing M-matrix properties based on a filtering strategy. We develop the new method focusing on FV discretizations for porous media single-phase

Akk_etal18; @SokBasHaj19; @Cas_etal19 ]. The MsRSB method was proposed in the context of FV simulation for fluid menstruation in highly heterogenous porous media [ @MsRSB_Moyner2016 ]. free-base on a two - power system approach, the MsRSB method constructs multiscale footing function through restricted smoothing on the fine - plate matrix. In more contingent, the basis functions, which are coherent with the local differential operators, are constructed with a cheap liberalization scheme, i.e. a weighted Jacobi iteration, similar to approaches used in smoothed collection multigrid methods [ @VanManBre96; @VanManBre01; @Bre_etal05 ]. An important advantage of MsRSB is that smoothing by liberalization provides a great deal of tractability in handling amorphous grids, an essential requirement, for example, in application involving complex geological structures. MsRSB has been widely proven and implemented in clear source and commercial simulators using a linear two - point flux approximation (TPFA) [ @Lie_etal16 ]. Because of the two - point structure, the linear TPFA scheme is monotonic [ @Dro14 ], i.e. it preserves the positivity of the differential solution [ @BerPle94 ], and lead to an thousand - matrix with a small stencil. This is the reason why linear TPFA is the scheme of option in most engineering software. Unfortunately, the consistency of TPFA is not undertake for arbitrary grids and anisotropic permeability distributions, potentially head to inaccurate results [ @MRST ]. Therefore, other FV methods such as multipoint flux approximation (MPFA) and/or nonlinear system [ @Dro14; @TerMalTch17 ] must be considered to achieve consistent fluxes. To date, MsRSB has not been combined with MPFA or other consistent discretizations. therefore, in this newspaper, we focus on enhancing MsRSB to enable the solution of second - order elliptic trouble using discretization methods that do not result in an M - matrix. free-base on the MPFA - O method [ @MPFA_Aavatsmark ], we show that the MsRSB basis structure as presented in [ @MsRSB_Moyner2016 ] can fail due to divergent iterations for an anisotropic diffusion problem. We propose a discrepancy of the original MsRSB approach that restore the desired demeanor by enforcing M - matrix properties free-base on a filtering strategy. We develop the modern method concentrate on FV discretizations for porous medium single - phase

Akk_ftal18; @SokBasHaj19; @Cas_etal19]. Tht MsRSB method was propoved in the congext of FV simulation for flnid dlow un highly heterogeneour porous ledia [@MsESB_Mitner2016]. Based on a two-grid ziproaeh, the MsRSB metmod construwts multiscale bxsns functions through restricted smooehing om hhe fine-scale iatrpx. In jore detail, the basis functions, whjch are consistent woth the local differential opegators, are construfted with a chewp relaxation scheme, i.e. a weighted Nacobi iteration, similar to appfoachzs used in wmiotjgd aggregatiin mujtigrid methods [@VanMdnBre96; @VsnManBre01; @Bre_etsl05]. En inportant advantage of MsRSB is that smojthing by rzlaxation provides a trwat dgal ox fldzibkliuy mn gandlijg nnstructures grids, an wssential requirememt, dor example, ih applycwtions involving complex geological strlctudes. MsRSB has been wideoy proven and implemejted in o[en source and commercial simulators using a linedr twk-ooiub woud approximation (TPFA) [@Lie_etal16]. Because of the twj-loonn structure, the llnear TPFA scheme os mpgotone [@Dro14], i.e. it przaedves the positivitj of thg diffwrential foluyion [@BerPle94], and leads to an M-matrix witk a small stencil. Thid is the recson wny limear TPFA is the scheme of dhoice in mlst enginsdring software. Uvfogtundtely, tht consistency of T[FA is nov guaxanteed wor srbitrwry grids wnd akhsotropic permeabipity bistrhbutions, pltentially leading to inaccuratx results [@MRSJ]. Tvervfore, othzr FV kethods such ws multipoint ylux appxoximagion (MPFA) znd/or nmnlinear scremes [@Dro14; @TerKwlTch17] must bx considewed ro axhieve zunsistent fluxrs. To datv, IsRSV has not been comnined sith MPFA or otktr xonsistent disctetkzaeilnx. Rance, in this papdr, dr focjs on enhangine MsTSB to enable the sonutikn of second-order rljiptic ptoblems ufing discretiaation methods thau do nmt cesult in an M-matrix. Based on the MPFA-O method [@MOFA_Wavatsmark], wq shiw that the KsRSB basis construction as presented ii [@MsRSB_Moyner2016] can fail eue to divergent itgrauions for an enisotwopic difxusion problem. We pripose a variant on the original MsRSB alproacv thah restores the desired behavior by enforcing M-matrix properties based on q filtxryng strategg. We deveno' tke new iethlv focusing on FV ciscretizations for porous medie single-phdsz

Akk_etal18; @SokBasHaj19; @Cas_etal19]. The MsRSB method was the of FV for fluid flow [@MsRSB_Moyner2016]. on a two-grid the MsRSB method multiscale basis functions through restricted smoothing the fine-scale matrix. In more detail, the basis functions, which are consistent with local differential operators, are constructed with a cheap relaxation scheme, i.e. a weighted iteration, to used smoothed aggregation multigrid methods [@VanManBre96; @VanManBre01; @Bre_etal05]. An important advantage of MsRSB is that smoothing by provides a great deal of flexibility in handling grids, an essential requirement, example, in applications involving complex structures. has been proven implemented open source and simulators using a linear two-point flux approximation (TPFA) [@Lie_etal16]. Because of the two-point structure, the linear TPFA is monotone it preserves positivity the solution [@BerPle94], and an M-matrix with a small stencil. reason why linear TPFA is the scheme of in most software. Unfortunately, the consistency of TPFA not guaranteed for arbitrary grids and anisotropic permeability potentially leading to inaccurate results [@MRST]. Therefore, other FV methods such as multipoint flux approximation nonlinear schemes [@Dro14; @TerMalTch17] be considered to consistent To MsRSB not been with MPFA or other consistent discretizations. Hence, in this paper, we on enhancing MsRSB to enable the solution of second-order elliptic discretization that do not in an M-matrix. Based the method [@MPFA_Aavatsmark], we show MsRSB construction [@MsRSB_Moyner2016] fail to divergent iterations for anisotropic diffusion problem. We propose variant of the original desired behavior by enforcing M-matrix properties based on filtering strategy. We develop the new method on FV discretizations for porous media single-phase

Akk_etal18; @SokBasHaj19; @Cas_etal19]. ThE MsRSB methOd was ProPosEd In thE conText of FV simulaTIon fOr fluid flow in highly hetErogeNeOUs poROuS mediA [@MsRSB_MOYnER2016]. basEd On A twO-gRId ApproAch, The MsRSb method conStrUcTs multiscale BAsIs functionS thRough restricTed SmoothInG on THe finE-scAle maTrix. In MOre detAil, the basIs FUnctioNS, which aRE CoNsisTent with the local dIFfERential operatoRs, are cOnSTrUCTed WitH a cheap relAxAtion SCheme, i.e. A WeIGHTed jAcobi iteratioN, similar to aPProAches uSeD in SMootheD aggrEgATioN multigrid mEthoDs [@VanManBRe96; @VanMANBre01; @Bre_ETal05]. An imPortanT adVanTage OF MSRsB iS tHAt sMOoThiNG by RelaxatiOn PrOvideS a grEAT DEal oF flExibIlity In handling unsTruCturED grIds, an EssenTial ReQuireMent, foR examPlE, in applications InvoLving compLex GeOloGiCal stRUctureS. MsrSB Has been Widely pROveN aND IMpLemented in open sourCe AND cOmmerciaL simulAToRs USing a linEaR twO-poiNT Flux aPproXImAtion (TPFa) [@Lie_etAL16]. BEcAuse of tHe Two-poiNt StrUctUre, thE LineAr TPFA Scheme is MonotONe [@Dro14], i.e. it preseRVes the positivITy OF ThE DiffEreNtial solutiOn [@BeRple94], aNd leADs To aN m-matrIx witH a SMaLL stencil. This is the reAsOn why lInear tPFA is the scheMe of choice IN MOst enginEeriNG sOFtware. UnfortunAtely, The consistENcy of TPFa is noT guarantEed for arbITRary gridS anD anIsoTroPIC pErmeability diSTRibuTiOns, poteNtiAlly leaDinG to InaCcuRaTe results [@mRST]. TherEfOrE, oThEr Fv methODs such as MuLtiPoInt Flux aPProximAtion (mPFA) AnD/oR NonLinear sCHeMES [@Dro14; @teRMAlTcH17] muSt Be conSideREd tO achievE consisteNt fLUxes. to DaTe, MsRSB Has not been comBiNed with MPFa oR otHer conSIStent disCretizations. Hence, in this PAper, we fOcuS on enHancIng MsRSB tO enAble thE soLUtion oF seconD-ordeR eLliPTIc proBLEmS usInG discretizATIon MethoDs That Do not reSult in an M-matrix. BasED on The MPFA-O methoD [@MPfA_AaVATsMarK], We SHow ThAT thE mSRSB basis constrUction as prEsENtEd in [@MsRSB_MOYneR2016] cAn fail dUe to divErgenT IteratiOns for an aNisotropiC dIffuSIOn pRoblem. We prOpose a vaRiant of thE OrigiNAl msRSB AppRoach tHaT reStoreS the deSIreD behaVior by EnForcinG M-matRiX propertIes based on a filtering strAtegy. WE deveLop The new metHod FOcuSing on FV dIscrEtizations For PorOus meDia SIngle-PhasE

Akk_etal18; @SokBasHaj19;@Cas_etal1 9]. The Ms RS B me thod was proposedi n th e context of FV simula tionfo r flu i dflowin high l yh e ter og en eou sp or ous m edi a [@MsR SB_Moyner2 016 ]. Based on at wo -grid appr oac h, the MsRSB me thod c on str u cts m ult iscal e basi s funct ions thro ug h restr i cted sm o o th ingon the fine-scale ma t rix. In more d etail, t h eb a sis fu nctions, w hi ch ar e consis t en t w ith the local dif ferential o p era tors,ar e c o nstruc ted w it h acheap relax atio n scheme, i.e.a weight e d Jacob i iter ati on, sim i la rtoap p roa c he s u s edin smoot he daggre gati o n m ulti gri d me thods [@VanManBre9 6;@Van M anB re01; @Bre _eta l0 5]. A n impo rtant a dvantage of MsR SB i s that sm oot hi ngby rela x ationpro vid es a gr eat dea l of f l e x ib ility in handlingun s t ru ctured g rids,a nes s ential r eq uir emen t , forexam p le , in app licati o ns i nvolvin gcomple xgeo log icals truc tures. MsRSB h as be e n widely prove n and implemen t ed i no pensou rce and com merc i al s imul a to rsu singa lin ea r t w o-point flux approx im ation(TPFA ) [@Lie_etal1 6]. Becau s e of the t wo-p o in t structure, th e lin ear TPFA s c heme ismonot one [@Dr o14], i.e . it prese rve s t hepos i t iv ity of the di f f eren ti al solu tio n [@Ber Ple 94] , a ndle ads to an M-matri xwi th a sm all s t encil. T hi s i sthe reas o n whylinea r TP FA i s th e schem e o f choi ce i n mo sten ginee ring sof tware.Unfortuna tel y , th eco nsisten cy of TPFA is n ot guarant ee d f or arb i t rary gri ds and anisotropic perm e ability di strib utio ns, poten tia lly le adi n g to i naccur ate r es ult s [@MRS T ] .The re fore, othe r FVmetho ds suc h as mu ltipoint flux appr o xim ation (MPFA)and /orn o nl ine a rs che me s [@ D r o14; @TerMalTch 17] must b ec on sidered to ach ie ve cons istentfluxe s . To da te, MsRSB has notbe en c o m bin ed with MP FA or ot her consi s tentd is creti zat ions.He nce , inthis p a per , wefocuson enhan cingMs RSB to e nable the solution of s econd- order el liptic pr obl e msusing dis cret ization me tho dsthatdon ot re sult in an M-mat rix. Based ont he MP F A -O method [@M P F A _Aa vatsm ark ] , we s howthat the MsRSB ba s is constructio n as p res ent e d in [ @MsRSB_Moyner2 016 ]c a n fail d ue to diverge nt itera ti o ns fo r an a nisotr opic di f f us i on pro blem . W e propose ava r iant of t he origin al M sR SB app roacht hatr e stores the desir ed be h a viorb y e nforc in g M-mat r ix p ropertiesbased on afilter ingstrat egy. We d evelop th enew method focusingon FV discre ti zati ons for p orou s media sin gl e-p hase

Akk_etal18; @SokBasHaj19;_@Cas_etal19]. The MsRSB_method was proposed in_the context_of_FV simulation_for_fluid flow in_highly heterogeneous porous_media [@MsRSB_Moyner2016]. Based on_a two-grid approach,_the_MsRSB method constructs multiscale basis functions through restricted smoothing on the fine-scale matrix. In_more_detail, the_basis_functions,_which are consistent with the_local differential operators, are constructed_with a_cheap relaxation scheme, i.e. a weighted Jacobi iteration,_similar_to approaches used_in smoothed aggregation multigrid methods [@VanManBre96; @VanManBre01; @Bre_etal05]. An_important advantage of MsRSB is that_smoothing by relaxation_provides_a_great deal of flexibility_in handling unstructured grids, an essential_requirement, for example, in applications involving_complex geological structures. MsRSB has been widely_proven and implemented in open source_and commercial simulators using a_linear two-point_flux approximation (TPFA) [@Lie_etal16]. Because of_the two-point structure,_the linear_TPFA scheme is_monotone [@Dro14], i.e. it preserves the_positivity of the_differential solution [@BerPle94], and leads to_an_M-matrix with a_small_stencil._This is_the reason why_linear_TPFA is_the_scheme of choice in most engineering_software._Unfortunately, the consistency of TPFA is not_guaranteed for arbitrary grids_and_anisotropic permeability distributions, potentially_leading to inaccurate results [@MRST]._Therefore, other FV methods such as_multipoint flux_approximation (MPFA)_and/or nonlinear schemes [@Dro14; @TerMalTch17] must be considered to achieve consistent_fluxes. To date, MsRSB has not_been combined with MPFA_or other_consistent_discretizations. Hence, in_this_paper, we_focus on enhancing MsRSB to enable the_solution of_second-order elliptic problems using discretization methods_that do not result_in_an M-matrix. Based on the MPFA-O_method [@MPFA_Aavatsmark], we show that the_MsRSB basis construction as presented_in_[@MsRSB_Moyner2016]_can fail due to divergent_iterations for an anisotropic diffusion problem._We propose a_variant of the original MsRSB approach that_restores_the desired behavior by enforcing M-matrix_properties_based on a filtering strategy. We_develop_the_new method focusing on FV_discretizations for porous media single-phase

shown in Fig. \[fig:graph\]. At high energies, for a pair of hadrons having a rapidity separation $\Delta y\gtrsim 1/\alpha_s$, a resummation of rapidity ordered multi-gluon emissions is necessary, corresponding to $t$-channel Pomeron exchange in the language of “Reggeon Field Theory" [@Lipatov:1995pn; @DelDuca:1995hf]. These emissions between the two tagged partons lead to an angular decorrelation of the dihadron signal observed in the data. The observation of an angular decorrelation as a signal of pomeron exchange is complementary to looking for the growth in the dijet cross-section, as first proposed by Mueller and Navelet [@Mueller:1986ey]. In this framework, the double inclusive multiplicity can be expressed as [@Colferai:2010wu; @Fadin:1996zv] $$\begin{aligned} \label{eq:BFKL} \left.\frac{d^2N_{AB}}{d^2\pp d^2\qp dy_p dy_q}\right|_{\rm BFKL} & &= \frac{32\,\nc\, \alpha_s(\pp)\,\alpha_s(\qp)}{ (2\pi)^8 \,\cf}\,\frac{\sp}{\pp^2\qp^2}\\ &\times&\int_{{ {\bf k}_{0\perp} }} \int_{{ {\bf k}_{3\perp} }} \Phi_A(x_1,{ {\bf k}_{0\perp} })\Phi_B(x_2,{ {\bf k}_{3\perp} })\,\mathcal{G}({ {\bf k}_{0\perp} }-\pp,{ {\bf k}_{3\perp} }+\qp,y_p-y_q)\nonumber\end{aligned}$$ where $\mathcal{G}$ is the BFKL Green’s function[^4] $$\begin{aligned} \mathcal{G}({ {\bf q}_{a\perp} },{ {\bf q}_{b\perp} },\Delta y)=\frac{1}{(2\pi)^2}\frac{1}{({ {\bf q}_{a\perp} }^2 { {\bf q}_{b\perp} }^2)^{1/2}}\sum_n e^{in\overline{\phi}}\int_{-\infty}^{+\infty} d\nu\textrm{ }

shown in Fig.   \[fig: graph\ ]. At high energies, for a pair of hadron have a rapidity separation $ \Delta y\gtrsim 1/\alpha_s$, a resummation of celerity ordered multi - gluon discharge is necessary, corresponding to $ t$-channel Pomeron central in the language of “ Reggeon Field Theory "   [ @Lipatov:1995pn; @DelDuca:1995hf ]. These emissions between the two tag partons lead to an angular decorrelation of the dihadron signal observed in the datum. The observation of an angular decorrelation as a signal of pomeron exchange is complementary to looking for the increase in the dijet cross - section, as first proposed by Mueller and Navelet [ @Mueller:1986ey ]. In this framework, the bivalent inclusive multiplicity can be expressed as   [ @Colferai:2010wu; @Fadin:1996zv ] $ $ \begin{aligned } \label{eq: BFKL } \left.\frac{d^2N_{AB}}{d^2\pp d^2\qp dy_p dy_q}\right|_{\rm BFKL } & & = \frac{32\,\nc\, \alpha_s(\pp)\,\alpha_s(\qp) } { (2\pi)^8 \,\cf}\,\frac{\sp}{\pp^2\qp^2}\\ & \times&\int _ { { { \bf k}_{0\perp } } } \int _ { { { \bf k}_{3\perp } } } \Phi_A(x_1, { { \bf k}_{0\perp } }) \Phi_B(x_2, { { \bf k}_{3\perp } }) \,\mathcal{G } ({ { \bf k}_{0\perp } } -\pp, { { \bf k}_{3\perp } } + \qp, y_p - y_q)\nonumber\end{aligned}$$ where $ \mathcal{G}$ is the BFKL Green ’s function[^4 ] $ $ \begin{aligned } \mathcal{G } ({ { \bf q}_{a\perp } }, { { \bf q}_{b\perp } }, \Delta y)=\frac{1}{(2\pi)^2}\frac{1 } { ({ { \bf q}_{a\perp } } ^2 { { \bf q}_{b\perp } } ^2)^{1/2}}\sum_n e^{in\overline{\phi}}\int_{-\infty}^{+\infty } d\nu\textrm { }

shlwn in Fig. \[fig:graph\]. At hinh energies, for c pair mf haddons havkng a rapidity separation $\Deptq y\gtesim 1/\alpha_s$, a resummatkon of raiidity oreerev multi-gluon emissions is neceadary, rorresponding tp $t$-channel Pomeron exchatgd nn the language of “Reggeon Field Thejry" [@Lipayog:1995pn; @DelDuca:1995hf]. Jhese qmisapoks between the two tagged partona lead uo an angular decotrelation of the dihadron dignwl observed in the data. The ovserdqtion of an xngular debmrrelation as a signal of pomeron exchangd is eomplementatv to nooking for the drowth in thc dijet cross-srction, as firsb pro'osee by Mueller and Navenet [@Mueller:1986ey]. In tris framefoxk, the double inclusice multhpliwity xan be eeprsssed ws [@Cklferai:2010wu; @Radin:1996zv] $$\begun{aligned} \label{eq:BFKK} \lqdt.\frac{d^2N_{AB}}{d^2\pp d^2\qp dr_p dy_q}\right|_{\rm BFKL} & &= \frac{32\,\nc\, \alpha_s(\pp)\,\alpha_s(\xp)}{ (2\li)^8 \,\cf}\,\frac{\sp}{\pp^2\qp^2}\\ &\times&\int_{{ {\bf k}_{0\perp} }} \int_{{ {\bf k}_{3\petp} }} \Phi_A(x_1,{ {\bs k}_{0\perp} })\Phi_B(x_2,{ {\bf k}_{3\perp} })\,\mathcal{G}({ {\bf k}_{0\perp} }-\pp,{ {\bf k}_{3\parp} }+\q',y_o-y_q)\uinumbde\ejd{aligned}$$ where $\mathcal{G}$ is the BFKL Green’s fthcuiok[^4] $$\begin{aligned} \mabhcal{G}({ {\bf q}_{a\perp} },{ {\nf q}_{n\kerp} },\Delta y)=\frxc{1}{(2\pi)^2}\frcd{1}{({ {\gf q}_{a\perp} }^2 { {\bf q}_{b\pfrp} }^2)^{1/2}}\sum_g e^{in\iverline{\pri}}\iny_{-\infty}^{+\infty} d\nu\textrm{ }

shown in Fig. \[fig:graph\]. At high energies, pair hadrons having rapidity separation $\Delta rapidity multi-gluon emissions is corresponding to $t$-channel exchange in the language of “Reggeon Theory" [@Lipatov:1995pn; @DelDuca:1995hf]. These emissions between the two tagged partons lead to an decorrelation of the dihadron signal observed in the data. The observation of an decorrelation a of exchange is complementary to looking for the growth in the dijet cross-section, as first proposed by and Navelet [@Mueller:1986ey]. In this framework, the double multiplicity can be expressed [@Colferai:2010wu; @Fadin:1996zv] $$\begin{aligned} \label{eq:BFKL} \left.\frac{d^2N_{AB}}{d^2\pp dy_p BFKL} & \frac{32\,\nc\, (2\pi)^8 &\times&\int_{{ {\bf k}_{0\perp} \int_{{ {\bf k}_{3\perp} }} \Phi_A(x_1,{ {\bf k}_{0\perp} })\Phi_B(x_2,{ {\bf k}_{3\perp} })\,\mathcal{G}({ {\bf k}_{0\perp} }-\pp,{ {\bf k}_{3\perp} }+\qp,y_p-y_q)\nonumber\end{aligned}$$ $\mathcal{G}$ is Green’s function[^4] \mathcal{G}({ q}_{a\perp} {\bf q}_{b\perp} },\Delta q}_{a\perp} }^2 { {\bf q}_{b\perp} }^2)^{1/2}}\sum_n

shown in Fig. \[fig:graph\]. At high eNergies, for A pair Of hAdrOnS havIng a Rapidity separaTIon $\DElta y\gtrsim 1/\alpha_s$, a resuMmatiOn OF rapIDiTy ordEred mulTI-gLUOn eMiSsIonS iS NeCessaRy, cOrrespoNding to $t$-chAnnEl pomeron exchaNGe In the languAge Of “Reggeon FieLd THeory" [@LIpAtoV:1995Pn; @DelducA:1995hf]. ThEse emiSSions bEtween the TwO Tagged PArtons lEAD tO an aNgular decorrelatiON oF The dihadron sigNal obsErVEd IN The DatA. The observAtIon of AN angulaR DeCORRelATion as a signal Of pomeron exCHanGe is coMpLemENtary tO lookInG For The growth in The dIjet cross-SectioN, As first PRoposed By MuelLer And naveLEt [@muEllEr:1986EY]. In THiS frAMewOrk, the doUbLe IncluSive MULTIpliCitY can Be expRessed as [@ColfeRai:2010Wu; @FaDIn:1996zV] $$\begiN{aligNed} \lAbEl{eq:BfKL} \lefT.\frac{D^2N_{aB}}{d^2\pp d^2\qp dy_p dy_q}\rIght|_{\Rm BFKL} & &= \fraC{32\,\nc\, \AlPha_S(\pP)\,\alphA_S(\qp)}{ (2\pi)^8 \,\cF}\,\frAc{\sP}{\pp^2\qp^2}\\ &\tiMes&\int_{{ {\bF K}_{0\peRp} }} \INT_{{ {\Bf K}_{3\perp} }} \Phi_A(x_1,{ {\bf k}_{0\perp} })\PHi_b(X_2,{ {\Bf K}_{3\perp} })\,\matHcal{G}({ {\bF K}_{0\pErP} }-\Pp,{ {\bf k}_{3\perP} }+\qP,y_p-Y_q)\noNUMber\eNd{alIGnEd}$$ where $\mAthcal{g}$ Is ThE BFKL GrEeN’s funcTiOn[^4] $$\bEgiN{aligNEd} \maThcal{G}({ {\Bf q}_{a\perp} },{ {\Bf q}_{b\pERp} },\Delta y)=\frac{1}{(2\pi)^2}\FRac{1}{({ {\bf q}_{a\perp} }^2 { {\bf Q}_{B\pERP} }^2)^{1/2}}\sUM_n e^{iN\ovErline{\phi}}\inT_{-\infTY}^{+\infTy} d\nU\TeXtrM{ }

shown in Fig. \[fig:graph \]. At hig h ene rgi es, f or a pai r of hadrons h a ving a rapidity separation $\De lt a y\g t rs im 1/ \alpha_ s $, a re su mm ati on of rapi dit y order ed multi-g luo nemissions is ne cessary, c orr esponding to $t $-chan ne l P o meron ex chang e in t h e lang uage of “ Re g geon F i eld The o r y"  [@L ipatov:1995pn; @D e lD u ca:1995hf]. Th ese em is s io n s be twe en the two t agged partons le a d toa n angular dec orrelationo f t he dih ad ron signal obse rv e d i n the data. The observat ion of an angu l ar deco rrelat ion as a s i gn al of p o mer o nexc h ang e is com pl em entar y to l o o king fo r th e gro wth in the di jet cro s s-s ectio n, as fir st prop osed b y Mue ll er and Navelet[@Mu eller:198 6ey ]. In t his f r amewor k,the double inclus i vemu l t i pl icity can be expre ss e d a s [@Colf erai:2 0 10 wu ; @Fadin: 19 96z v] $ $ \ begin {ali g ne d} \labe l{eq:B F KL }\left.\ fr ac{d^2 N_ {AB }}{ d^2\p p d^2 \qp dy _p dy_q} \righ t |_{\rm BFKL} & &= \frac{32\, \ nc \ , \ a lpha _s( \pp)\,\alph a_s( \ qp)} { (2 \ pi )^8 \,\cf }\,\f ra c {\ s p}{\pp^2\qp^2}\\ &\ ti mes&\i nt_{{ {\bf k}_{0\p erp} }} \i n t _ {{ {\bfk}_{ 3 \p e rp} }} \Phi_A( x_1,{ {\bf k}_{ 0 \perp} } )\Phi _B(x_2,{ {\bf k}_ { 3 \perp} } )\, \ma thc al{ G } ({ {\bf k}_{0\p e r p} } -\ pp,{ {\ bfk}_{3\p erp } } +\q p,y _p -y_q)\non umber\en d{ al ig ne d}$ $ whe r e $\math ca l{G }$ is theB FKL Gr een’s fun ct io n [^4 ] $$\be g in { a lign ed }\mat hca l{ G}({{\bf q}_ {a\perp } },{ {\b f q } _{b\ pe rp } },\De lta y)=\frac{ 1} {(2\pi)^2} \f rac {1}{({ { \bf q}_{ a\perp} }^2 { {\bf q}_{ b \perp}}^2 )^{1/ 2}}\ sum_n e^{ in\ overli ne{ \ phi}}\ int_{- \inft y} ^{+ \ i nfty} d \n u\t ex trm{ }

shown_in Fig. \[fig:graph\]._At high energies, for_a pair_of_hadrons having_a_rapidity separation $\Delta_y\gtrsim 1/\alpha_s$, a_resummation of rapidity ordered_multi-gluon emissions is_necessary,_corresponding to $t$-channel Pomeron exchange in the language of “Reggeon Field Theory" [@Lipatov:1995pn; @DelDuca:1995hf]. These_emissions_between the_two_tagged_partons lead to an angular_decorrelation of the dihadron signal_observed in_the data. The observation of an angular decorrelation_as_a signal of_pomeron exchange is complementary to looking for the growth_in the dijet cross-section, as first_proposed by Mueller_and_Navelet_[@Mueller:1986ey]. In this framework,_the double inclusive multiplicity can be_expressed as [@Colferai:2010wu; @Fadin:1996zv] $$\begin{aligned} \label{eq:BFKL} \left.\frac{d^2N_{AB}}{d^2\pp d^2\qp dy_p_dy_q}\right|_{\rm BFKL} & &= \frac{32\,\nc\, \alpha_s(\pp)\,\alpha_s(\qp)}{ (2\pi)^8 \,\cf}\,\frac{\sp}{\pp^2\qp^2}\\ &\times&\int_{{_{\bf k}_{0\perp} }} \int_{{ {\bf k}_{3\perp}_}} \Phi_A(x_1,{ {\bf k}_{0\perp} })\Phi_B(x_2,{ {\bf_k}_{3\perp} })\,\mathcal{G}({_{\bf k}_{0\perp} }-\pp,{ {\bf k}_{3\perp}_}+\qp,y_p-y_q)\nonumber\end{aligned}$$ where $\mathcal{G}$_is the_BFKL Green’s function[^4]_$$\begin{aligned} \mathcal{G}({ {\bf q}_{a\perp} },{ {\bf q}_{b\perp}_},\Delta y)=\frac{1}{(2\pi)^2}\frac{1}{({ {\bf_q}_{a\perp} }^2 { {\bf q}_{b\perp} }^2)^{1/2}}\sum_n_e^{in\overline{\phi}}\int_{-\infty}^{+\infty}_d\nu\textrm{ }

counter runs out. By assumption, that the language of the net with initial configuration $(\initialstate,c_0)$ is universal, there must be another run $\pi_4$ on the same word, and which is accepting. But now both runs, $\pi_4$ and $\pi_1\pi_2^k\pi_3$, are accepting from the configuration $(\initialstate,c_0+\norm{\transitions}\cdot \len{\pi_2}\cdot k)$ as the effect of $\pi_2^k$ is larger than $\norm{\transitions}\cdot \len{\pi_2}\cdot k$. This means that the net is not structurally unambiguous, which contradicts our assumptions. Suppose otherwise, then for any bound $k$ there will be an accepting run which is going through configurations with counter value bigger than $k$, and from unambiguity, there is no other run that stays below the bound. \[apx:thm:UOCN-bu\] **Unary encoded transitions:** By \[lem:UOCN-no-positive-loops\], if the OCN is bounded universal then every accepting run will only visit counter values below $\bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the OCN is bounded universal if, and only if, $\bLang{\initialstate,c_0}{\bound{1}}=\alphabet^*$. This can be verified by checking universality for the UFA that results by remembering all bounded counter values in the finite state space. The claim now follows by \[lem:universality\_of\_UFA\]. **Binary encoded transitions:** By \[lem:UOCN-no-positive-loops\], if the OCN is bounded universal then every accepting run will only visit counter values below $\bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the OCN is bounded universal if, and only if, $\bLang{s_0,c_0}{\bound{1}}=\alphabet^*$. This can be verified by checking universality for the UFA that results by remembering all bounded counter values in the finite state space. The claim now follows by \[lem:universality\_of\_UFA\] and the following fact [^^]{}$=PolyLog

counter runs out. By assumption, that the language of the internet with initial shape $ (\initialstate, c_0)$ is universal, there must be another run $ \pi_4 $ on the like word, and which is accept. But now both runs, $ \pi_4 $ and $ \pi_1\pi_2^k\pi_3 $, are accept from the configuration $ (\initialstate, c_0+\norm{\transitions}\cdot \len{\pi_2}\cdot k)$ as the effect of $ \pi_2^k$ is bigger than $ \norm{\transitions}\cdot \len{\pi_2}\cdot k$. This means that the net is not structurally unequivocal, which contradicts our assumptions. Suppose otherwise, then for any bind $ k$ there will be an accepting run which is going through configurations with counter value adult than $ k$, and from unambiguity, there is no other run that stay below the bound. \[apx: thm: UOCN - bu\ ] * * Unary encoded transitions :* * By \[lem: UOCN - no - convinced - loops\ ], if the OCN is bounded universal then every accepting run will only chew the fat counter values below $ \bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the OCN is bounded universal if, and only if, $ \bLang{\initialstate, c_0}{\bound{1}}=\alphabet^*$. This can be verified by checking universality for the UFA that results by remembering all bounded counter values in the finite state space. The claim now follow by \[lem: universality\_of\_UFA\ ]. * * Binary encoded transitions :* * By \[lem: UOCN - no - positive - loops\ ], if the OCN is bounded cosmopolitan then every accept run will only chew the fat counter values below $ \bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the OCN is bounded cosmopolitan if, and only if, $ \bLang{s_0,c_0}{\bound{1}}=\alphabet^*$. This can be verified by checking universality for the UFA that results by remembering all bounded counter values in the finite state space. The claim now play along by \[lem: universality\_of\_UFA\ ] and the following fact [ ^^]{}$=PolyLog

coknter runs out. By assumpuion, that the lanyyage oh the nst with knitial configuration $(\initiapsrate,c_0)$ is universal, there murt be anonher run $\pi_4$ oi the same word, ehd whicm is zgceptnnj. But now both tuns, $\pi_4$ and $\[i_1\pi_2^k\pi_3$, are acweotnng from the configuration $(\initialstwte,c_0+\nork{\tgansitions}\cdot \len{\ki_2}\cqot i)$ as the effect of $\pi_2^k$ is larger tgan $\nork{\transitions}\ccot \len{\pi_2}\cdot k$. This means thah the net is not shructurally unaiviguous, whicf contradibcs our assujptions. Suppose otherwise, then fur anv bound $k$ tyeee ahll be an arceptigg run which is goinc throubh configuratipns wirh counter value biggxr than $k$, and from ugambiguitf, chere is no other run tyat sjays telod thd bkuid. \[alx:thm:ULCN-uu\] **Unary endoded transutions:** By \[lem:UOCN-no-kosynove-loops\], if fhe OCG ys bounded universal then every acceptitg dun will only visit coubter values below $\boujd{1}\eqdef c_0+\found{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means thdt thx UCN if voknded universal if, and only if, $\bLang{\initialstwfe,v_0}{\bpund{1}}=\alphabet^*$. Tmis can be verifiec hy shecking univgrsalitv fkr the UFA that redults bi remenbering ajl bpunded counter values in thw finite stane space. The claim now follows by \[lem:oniverxality\_of\_UFA\]. **Binary encodzd trahsitions:** By \[lem:UOCN-nk-oositive-loops\], if thv OCT is bounded universal theg every arceptnng run dill only disit counher values below $\bound{1}\eqfef c_0+\yound{0}=w_0+\card{\stated}\cdot\norm{\transitions}$. This means that the OCN os bolnded unirersal if, and only yf, $\bLang{s_0,c_0}{\bounb{1}}=\alphabec^*$. This can be vegified by checking ugiversality fmt the UFA thav results by eemenbering xll bounded cointer vallef in the finite state xpazs. The claim now foolows by \[lem:uniferralytj\_of\_NFA\] agg the followhng wacg [^^]{}$=PolyUog

counter runs out. By assumption, that the the with initial $(\initialstate,c_0)$ is universal, $\pi_4$ the same word, which is accepting. now both runs, $\pi_4$ and $\pi_1\pi_2^k\pi_3$, accepting from the configuration $(\initialstate,c_0+\norm{\transitions}\cdot \len{\pi_2}\cdot k)$ as the effect of $\pi_2^k$ is than $\norm{\transitions}\cdot \len{\pi_2}\cdot k$. This means that the net is not structurally unambiguous, contradicts assumptions. otherwise, for any bound $k$ there will be an accepting run which is going through configurations with value bigger than $k$, and from unambiguity, there no other run that below the bound. \[apx:thm:UOCN-bu\] **Unary transitions:** \[lem:UOCN-no-positive-loops\], if OCN bounded then every accepting will only visit counter values below $\bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the OCN is bounded universal if, only if, can be by universality the UFA that remembering all bounded counter values in space. The claim now follows by \[lem:universality\_of\_UFA\]. **Binary transitions:** By if the OCN is bounded universal every accepting run will only visit counter values $\bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the OCN is bounded universal if, and only if, $\bLang{s_0,c_0}{\bound{1}}=\alphabet^*$. be verified by checking for the UFA results remembering bounded values in finite state space. The claim now follows by \[lem:universality\_of\_UFA\] and the fact [^^]{}$=PolyLog

counter runs out. By assumptioN, that the laNguagE of The NeT witH iniTial configuratIOn $(\inItialstate,c_0)$ is universal, There MuST be aNOtHer ruN $\pi_4$ on thE SaME WorD, aNd WhiCh IS aCceptIng. but now bOth runs, $\pi_4$ aNd $\pI_1\pI_2^k\pi_3$, are accepTInG from the coNfiGuration $(\initIalState,c_0+\NoRm{\tRAnsitIonS}\cdot \Len{\pi_2}\cDOt k)$ as tHe effect oF $\pI_2^K$ is larGEr than $\nORM{\tRansItions}\cdot \len{\pi_2}\cdOT k$. tHis means that thE net is NoT StRUCtuRalLy unambiguOuS, whicH ContradICtS OUR asSUmptions. SuppoSe otherwise, THen For any BoUnd $K$ There wIll be An ACcePting run whiCh is Going throUgh conFIguratiONs with cOunter ValUe bIggeR ThAn $K$, anD fROm uNAmBigUIty, There is nO oThEr run That STAYS belOw tHe boUnd. \[apX:thm:UOCN-bu\] **UnaRy eNcodED trAnsitIons:** BY \[lem:uOcN-no-pOsitivE-loopS\], iF the OCN is boundeD uniVersal theN evErY acCePting RUn will OnlY viSit counTer valuES beLoW $\BOUnD{1}\eqdef c_0+\bound{0}=c_0+\card{\sTaTES}\cDot\norm{\tRansitIOnS}$. THIs means tHaT thE OCN IS BoundEd unIVeRsal if, anD only iF, $\BLAnG{\initiaLsTate,c_0}{\bOuNd{1}}=\aLphAbet^*$. THIs caN be verIfied by cHeckiNG universality fOR the UFA that reSUlTS By REmemBerIng all boundEd coUNter ValuES iN thE FinitE statE sPAcE. the claim now follows bY \[lEm:univErsalIty\_of\_UFA\]. **BinarY encoded trANSItions:** By \[Lem:UocN-NO-positive-loops\], If the oCN is boundED universAl theN every acCepting ruN WIll only vIsiT coUntEr vALUeS below $\bound{1}\eqDEF c_0+\boUnD{0}=c_0+\card{\sTatEs}\cdot\nOrm{\TraNsiTioNs}$. this means That the OcN Is BoUnDed UniveRSal if, and OnLy iF, $\blanG{s_0,c_0}{\boUNd{1}}=\alphAbet^*$. THis cAn Be VEriFied by cHEcKINg unIvErSaliTy fOr The UFa thaT ResUlts by rEmemberinG alL BounDeD cOunter vAlues in the finItE state spacE. THe cLaim noW FOllows by \[Lem:universality\_of\_UFA\] and THe folloWinG fact [^^]{}$=polylog

counter runs out. By assu mption, th at th e l ang ua ge o f th e net with ini t ialconfiguration $(\initi alsta te , c_0) $ i s uni versal, th e r e m us tbean o th er ru n $ \pi_4$on the sam e w or d, and which is accepting . B ut now bothrun s, $\p i_ 4$a nd $\ pi_ 1\pi_ 2^k\pi _ 3$, ar e accepti ng from t h e confi g u ra tion $(\initialstate, c _0 + \norm{\transit ions}\ cd o t\ l en{ \pi _2}\cdot k )$ as t h e effec t o f $ \pi _ 2^k$ is large r than $\no r m{\ transi ti ons } \cdot\len{ \p i _2} \cdot k$. T hismeans tha t then et is n o t struc turall y u nam bigu o us ,whi ch con t ra dic t s o ur assum pt io ns. Supp o s e othe rwi se,thenfor any bound $k $ th e rewillbe an acc ep tingrun wh ich i sgoing through c onfi gurations wi th co un ter v a lue bi gge r t han $k$ , and f r omun a m b ig uity, there is noot h e rrun that stays be lo w the bou nd . \[ap x : thm:U OCN- b u\ ] **Unar y enco d ed t ransiti on s:** B y\[l em: UOCN- n o-po sitive -loops\] , ift he OCN is boun d ed universalt he n ev e ry a cce pting run w illo nlyvisi t c oun t er va luesbe l ow $\bound{1}\eqdef c_ 0+ \bound {0}=c _0+\card{\sta tes}\cdot\ n o r m{\trans itio n s} $ . This means t hat t he OCN isb ounded u niver sal if,and onlyi f , $\bLan g{\ ini tia lst a t e, c_0}{\bound{1 } } =\al ph abet^*$ . T his can be ve rif ied b y checkin g univer sa li ty f orthe U F A that r es ult sbyremem b eringall b ound ed c o unt er valu e si n the f in itesta te spac e. T h e c laim no w follows by \[le m: un iversal ity\_of\_UFA\ ]. **Binary e nco ded tr a n sitions: ** By \[lem:UOCN-no-pos i tive-lo ops \], i f th e OCN isbou nded u niv e rsal t hen ev ery a cc ept i n g run w il l o nl y visit co u n ter valu es bel ow $\bo und{1}\eqdef c_0+\ b oun d{0}=c_0+\car d{\ stat e s }\ cdo t \n o rm{ \t r ans i t ions}$. This me ans that t he OC N is bound e d u ni versalif, and only if, $\b Lang{s_0, c_0}{\bou nd {1}} = \ alp habet^*$.This can be verif i ed by ch eckin g u nivers al ity forthe UF A th at re sultsby remem berin gall boun ded counter values in t he fin ite s tat e space.The cla im now fo llow s by \[lem :un ive rsali ty\ _ of\_U FA\] an d t h e fol lowi n g fact [^ ^ ]{ }$= P o ly Log

counter_runs out._By assumption, that the_language of_the_net with_initial_configuration $(\initialstate,c_0)$ is_universal, there must_be another run $\pi_4$_on the same_word,_and which is accepting. But now both runs, $\pi_4$ and $\pi_1\pi_2^k\pi_3$, are accepting from_the_configuration $(\initialstate,c_0+\norm{\transitions}\cdot_\len{\pi_2}\cdot_k)$_as the effect of $\pi_2^k$_is larger than $\norm{\transitions}\cdot \len{\pi_2}\cdot_k$. This_means that the net is not structurally unambiguous,_which_contradicts our assumptions. Suppose_otherwise, then for any bound $k$ there will be_an accepting run which is going_through configurations with_counter_value_bigger than $k$, and_from unambiguity, there is no other_run that stays below the bound. \[apx:thm:UOCN-bu\]_**Unary encoded transitions:** By \[lem:UOCN-no-positive-loops\], if the_OCN is bounded universal then every_accepting run will only visit_counter values_below $\bound{1}\eqdef c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means_that the OCN_is bounded_universal if, and_only if, $\bLang{\initialstate,c_0}{\bound{1}}=\alphabet^*$. This can be_verified by checking_universality for the UFA that results_by_remembering all bounded_counter_values_in the_finite state space._The_claim now_follows_by \[lem:universality\_of\_UFA\]. **Binary encoded transitions:** By \[lem:UOCN-no-positive-loops\],_if_the OCN is bounded universal then every_accepting run will only_visit_counter values below $\bound{1}\eqdef_c_0+\bound{0}=c_0+\card{\states}\cdot\norm{\transitions}$. This means that the_OCN is bounded universal if, and_only if,_$\bLang{s_0,c_0}{\bound{1}}=\alphabet^*$. This_can be verified by checking universality for the UFA that results_by remembering all bounded counter values_in the finite state_space. The_claim_now follows by_\[lem:universality\_of\_UFA\]_and the_following fact [^^]{}$=PolyLog

-chirality in the spin-1/2 multiferroics, $\sum_i \hat{c}_i^x(\hat{s}_{i+1}^x - \hat{s}_i^x) = \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x)$, which indicates that when the spin at site $i$ is flipped, $\hat{s}_i \rightarrow -\hat{s}_i$, the direction of spin-chirality $\hat{c}_i$ and $\hat{c}_{i-1}$ are also reversed. Assuming all spins point along their corresponding classical directions in the ground state of the spin-1/2 helical magnet as in NaCu$_2$O$_2$, where a $J_1-J_2$ spin model provides a good description of the helix state [@NaCu2O2]. So the spin interaction can be ferromagnetically given as $-J_s(Q) \hat{s}_i \cdot \hat{s}_j$ where $Q$ is taken as the pitch angle along the chain. An effective model that describes the interplay between the helical spin and spin-chirality has the form $$\begin{aligned} H_{sc} = -\sum_{i,j} (J_s \hat{s}_i \cdot \hat{s}_j + J_{c} \hat{c}_i \cdot \hat{c}_j) - \gamma \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x).\end{aligned}$$ The Hilbert space can be considered as the tensor product space $ |i \rangle \rightarrow |s_i^z \rangle _s \otimes | c_i^z \rangle _c. $ Now if the spin at site $i$ is flipped, the spin and spin-chirality excitations are mixed due to the spin-phonon coupling. The expected value of spin-chirality is given by $$\langle \hat{c} \rangle =1 - \langle \hat{s} \rangle, \label{E-value}$$ which is less than one. The experimental data for a finite differential scattering intensity of polarized neutrons from LiCu$_2$O$_2$ [@Li

-chirality in the spin-1/2 multiferroics, $ \sum_i \hat{c}_i^x(\hat{s}_{i+1}^x - \hat{s}_i^x) = \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x)$, which indicates that when the spin at site $ i$ is flipped, $ \hat{s}_i \rightarrow -\hat{s}_i$, the steering of tailspin - chirality $ \hat{c}_i$ and $ \hat{c}_{i-1}$ are also reversed. assume all spin point along their corresponding authoritative directions in the ground state of the spin-1/2 coiling attraction as in NaCu$_2$O$_2 $, where a $ J_1 - J_2 $ spin model put up a good description of the helix state [ @NaCu2O2 ]. therefore the spin interaction can be ferromagnetically given as $ -J_s(Q) \hat{s}_i \cdot \hat{s}_j$ where $ Q$ is taken as the pitch angle along the range. An effective model that describes the interplay between the helical tailspin and spin - chirality has the form $ $ \begin{aligned } H_{sc } = -\sum_{i, j } (J_s \hat{s}_i \cdot \hat{s}_j + J_{c } \hat{c}_i \cdot \hat{c}_j) - \gamma \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x).\end{aligned}$$ The Hilbert space can be considered as the tensor merchandise space $ |i \rangle \rightarrow |s_i^z \rangle _ s \otimes | c_i^z \rangle _ c. $ Now if the spin at site $ i$ is flipped, the tailspin and spin - chirality excitations are mixed due to the spin - phonon coupling. The expected value of spin - chirality is move over by $ $ \langle \hat{c } \rangle = 1 - \langle \hat{s } \rangle, \label{E - value}$$ which is less than one. The experimental data for a finite differential scattering intensity of polarized neutrons from LiCu$_2$O$_2 $ [ @Li

-chigality in the spin-1/2 multinerroics, $\sum_i \haj{c}_u^x(\hat{s}_{m+1}^x - \hat{a}_i^x) = \sum_k \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x)$, whici ineicattf that when the spkn at sitv $i$ is fluppev, $\hat{s}_i \rightarrow -\hat{s}_i$, the djvectimi of spin-chirallty $\hat{c}_i$ atd $\hat{c}_{i-1}$ are ansu xeversed. Assuming all spins point aljng theor corresponding claxficam directions in the ground state or the skin-1/2 helical magnet as in NaCu$_2$O$_2$, where a $J_1-J_2$ spln mldel provides a gold descriptuon jd the helix rtate [@NaCu2O2]. So the spih interaction can be ferromagnegicalky given aw $-H_s(Q) \vat{s}_i \cdot \iat{s}_j$ rhere $Q$ is taken as dhe pitvh angle along thx chqin. An effective moden that describes tre interpnav between the helical spin atd s[in-cfurauitg ias the flrm $$\begin{alighed} H_{sc} = -\sum_{u,j} (J_s \hat{s}_i \cdot \hau{s}_j + J_{c} \hat{c}_i \cdof \hat{c}_t) - \gamma \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x).\end{alicnes}$$ The Hilbert space can be considered as the tensor pwoduct space $ |i \rangle \rightarrow |s_i^z \rangle _s \otikes | r_i^x \rcknle _z. $ Nlw if the spin at site $i$ is flipped, the spin whd siin-chirality excibations are mixed cuf yj the spin-phovon coblljng. The expected vwlue of spin-xhirality is biven by $$\langle \hat{c} \rangle =1 - \langle \han{s} \rqngle, \label{E-value}$$ wkich is less tham one. The experimental data yor a rinite difffrential azattering intenskty ox polariedd neutrons from JiCu$_2$O$_2$ [@Li

-chirality in the spin-1/2 multiferroics, $\sum_i \hat{c}_i^x(\hat{s}_{i+1}^x = \hat{s}_i^x (\hat{c}_{i-1}^x \hat{c}_i^x)$, which indicates site is flipped, $\hat{s}_i -\hat{s}_i$, the direction spin-chirality $\hat{c}_i$ and $\hat{c}_{i-1}$ are also Assuming all spins point along their corresponding classical directions in the ground state the spin-1/2 helical magnet as in NaCu$_2$O$_2$, where a $J_1-J_2$ spin model provides good of helix [@NaCu2O2]. So the spin interaction can be ferromagnetically given as $-J_s(Q) \hat{s}_i \cdot \hat{s}_j$ where $Q$ taken as the pitch angle along the chain. effective model that describes interplay between the helical spin spin-chirality the form H_{sc} -\sum_{i,j} \hat{s}_i \cdot \hat{s}_j J_{c} \hat{c}_i \cdot \hat{c}_j) - \gamma \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hat{c}_i^x).\end{aligned}$$ The Hilbert space can be considered the tensor $ |i \rightarrow \rangle \otimes | c_i^z $ Now if the spin at flipped, the spin and spin-chirality excitations are mixed to the coupling. The expected value of spin-chirality given by $$\langle \hat{c} \rangle =1 - \langle \rangle, \label{E-value}$$ which is less than one. The experimental data for a finite differential scattering polarized neutrons from LiCu$_2$O$_2$

-chirality in the spin-1/2 multifeRroics, $\sum_i \Hat{c}_i^X(\haT{s}_{i+1}^X - \hAt{s}_i^X) = \sum_I \hat{s}_i^x (\hat{c}_{i-1}^x - \hAT{c}_i^x)$, Which indicates that when The spIn AT sitE $I$ iS flipPed, $\hat{s}_I \RiGHTarRoW -\hAt{s}_I$, tHE dIrectIon Of spin-cHirality $\haT{c}_i$ AnD $\hat{c}_{i-1}$ are alsO ReVersed. AssuMinG all spins poiNt aLong thEiR coRRespoNdiNg claSsical DIrectiOns in the gRoUNd statE Of the spIN-1/2 HeLicaL magnet as in NaCu$_2$O$_2$, wHErE A $J_1-J_2$ spin model prOvides A gOOd DEScrIptIon of the heLiX statE [@naCu2O2]. So THe SPIN inTEraction can be FerromagnetICalLy giveN aS $-J_s(q) \Hat{s}_i \cDot \haT{s}_J$ WheRe $Q$ is taken aS the Pitch anglE along THe chain. aN effectIve modEl tHat DescRIbEs The InTErpLAy BetWEen The helicAl SpIn and Spin-CHIRAlitY haS the Form $$\bEgin{aligned} H_{sC} = -\suM_{i,j} (J_S \Hat{S}_i \cdoT \hat{s}_J + J_{c} \hAt{C}_i \cdoT \hat{c}_j) - \Gamma \SuM_i \hat{s}_i^x (\hat{c}_{i-1}^x - \hAt{c}_i^X).\end{alignEd}$$ THe hilBeRt spaCE can be ConSidEred as tHe tensoR ProDuCT SPaCe $ |i \rangle \rightarroW |s_I^Z \RaNgle _s \otiMes | c_i^z \RAnGlE _C. $ Now if thE sPin At siTE $I$ is flIppeD, ThE spin and Spin-chIRaLiTy excitAtIons arE mIxeD duE to thE Spin-Phonon Coupling. the exPEcted value of spIN-chirality is gIVeN BY $$\lANgle \Hat{C} \rangle =1 - \langLe \haT{S} \ranGle, \lABeL{E-vALue}$$ whIch is LeSS tHAn one. The experimentaL dAta for A finiTe differentiaL scatterinG INTensity oF polARiZEd neutrons from liCu$_2$O$_2$ [@li

-chirality in the spin-1/2 multiferr oics, $\ sum _i \ha t{c} _i^x(\hat{s}_{ i +1}^ x - \hat{s}_i^x) = \su m_i \ ha t {s}_ i ^x (\ha t{c}_{i - 1} ^ x -\h at {c} _i ^ x) $, wh ich indica tes that w hen t he spin at s i te $i$ is fl ipp ed, $\hat{s} _i\right ar row -\hat {s} _i$,the di r ection of spin- ch i rality $\hat{c } _ i$ and $\hat{c}_{i-1}$a re also reversed. Assum in g a l l sp ins point alo ng thei r corres p on d i n g c l assical direc tions in th e gr ound s ta teo f thespin- 1/ 2 he lical magne t as in NaCu$ _2$O$_ 2 $, wher e a $J_1 -J_2$spi n m odel pr ov ide sa go o ddes c rip tion ofth ehelix sta t e [ @NaC u2O 2].So th e spin intera cti on c a n b e fer romag neti ca lly g iven a s $-J _s (Q) \hat{s}_i \ cdot \hat{s}_ j$wh ere $ Q$ is takenasthe pitchangle a l ong t h e ch ain. An effectivemo d e lthat des cribes th ei nterplay b etw eent h e hel ical sp in and s pin-ch i ra li ty hasth e form $ $\b egi n{ali g ned} H_{sc } = -\su m_{i, j } (J_s \hat{s} _ i \cdot \hat{ s }_ j +J _{c} \h at{c}_i \cd ot \ h at{c }_j) -\ga m ma \s um_i\h a t{ s }_i^x (\hat{c}_{i-1 }^ x - \h at{c} _i^x).\end{al igned}$$ T h e Hilbertspac e c a n be considere d asthe tensor productspace $ |i \r angle \ri g h tarrow | s_i ^z\ra ngl e _s \otimes | c_ i ^ z \r an gle _c. $Now ifthe sp inatsi te $i$ is flipped ,th esp inand s p in-chira li tyex cit ation s are m ixeddueto t h e s pin-pho n on c oupl in g. The ex pe ctedvalu e of spin-c hiralityisg iven b y$$\lang le \hat{c} \r an gle =1 - \ la ngl e \hat { s } \rangl e, \label{E-value}$$ wh i ch is l ess than one . The exp eri mental da t a fora fini te di ff ere n t ial s c a tt eri ng intensity o f p olari ze d ne utronsfrom LiCu$_2$O$_2$ [@L i

-chirality in_the spin-1/2_multiferroics, $\sum_i \hat{c}_i^x(\hat{s}_{i+1}^x -_\hat{s}_i^x) =_\sum_i_\hat{s}_i^x (\hat{c}_{i-1}^x_-_\hat{c}_i^x)$, which indicates_that when the_spin at site $i$_is flipped, $\hat{s}_i_\rightarrow_-\hat{s}_i$, the direction of spin-chirality $\hat{c}_i$ and $\hat{c}_{i-1}$ are also reversed. Assuming all spins_point_along their_corresponding_classical_directions in the ground state_of the spin-1/2 helical magnet_as in_NaCu$_2$O$_2$, where a $J_1-J_2$ spin model provides a_good_description of the_helix state [@NaCu2O2]. So the spin interaction can be_ferromagnetically given as $-J_s(Q) \hat{s}_i \cdot_\hat{s}_j$ where $Q$_is_taken_as the pitch angle_along the chain. An effective model_that describes the interplay between the_helical spin and spin-chirality has the form_$$\begin{aligned} H_{sc} = -\sum_{i,j} (J_s \hat{s}_i \cdot_\hat{s}_j + J_{c} \hat{c}_i \cdot_\hat{c}_j) -_\gamma \sum_i \hat{s}_i^x (\hat{c}_{i-1}^x -_\hat{c}_i^x).\end{aligned}$$ The Hilbert_space can_be considered as_the tensor product space $ |i \rangle_\rightarrow |s_i^z \rangle__s \otimes | c_i^z \rangle _c. $_Now_if the spin_at_site_$i$ is_flipped, the spin_and_spin-chirality excitations_are_mixed due to the spin-phonon coupling._The_expected value of spin-chirality is given by_$$\langle \hat{c} \rangle =1_-_\langle \hat{s} \rangle, \label{E-value}$$ which_is less than one. The_experimental data for a finite differential_scattering intensity_of polarized_neutrons from LiCu$_2$O$_2$ [@Li

is an *eigenvector* of $T$ and satisfies the equation $Tx = \lambda x$. If $T:\sD(\sX)\subset\sX\to\sX$ is a closed operator with dense domain, $\sD(\sX) \subset \sX$, then we say that $\lambda \notin \sigma(T)$ if the operator $T-\lambda:\sD(\sX)\to\sX$ has a bounded inverse and otherwise that $\lambda \in \sigma(T)$ [@KadisonRingrose1 Section 5.6, p. 357], [@Kato Section III.6.1, pp. 174–175]. \[prop:Gilbarg\_Trudinger\_theorem\_8-6\] Let $(X,g)$ be a closed, Riemannian, smooth manifold of dimension $d \geq 2$, and $G$ be a compact Lie group, $P$ be a smooth principal $G$-bundle over $X$, and $l \geq 0$ be an integer. If $A$ is a $W^{1,q}$ connection on $P$ with $d/2 < q < \infty$, and $A_1$ is a $C^\infty$ reference connection on $P$, and $p \in (1,\infty)$ obeys $d/2 \leq p \leq q$, then the spectrum, $\sigma(\Delta_A)$, of the unbounded operator, $$\Delta_A: \sD(\Delta_A) \subset L^p(X;\Lambda^l\otimes\ad P) \to L^p(X;\Lambda^l\otimes\ad P),$$ is countable without accumulation points, consisting of non-negative, real eigenvalues, $\lambda$, with finite multiplicities, $\dim\Ker(\Delta_A-\lambda)$. Corollary \[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\] implies that the operator, $$\Delta_A: W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P) \to L^p(X;\Lambda^l\otimes\ad P),$$ is Fredholm and, setting $K := \Ker(\Delta_A: W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P) \to L^

is an * eigenvector * of $ T$ and satisfies the equation $ Tx = \lambda x$. If $ T:\sD(\sX)\subset\sX\to\sX$ is a closed operator with dense knowledge domain, $ \sD(\sX) \subset \sX$, then we suppose that $ \lambda \notin \sigma(T)$ if the operator $ T-\lambda:\sD(\sX)\to\sX$ has a bounded inverse and differently that $ \lambda \in \sigma(T)$ [ @KadisonRingrose1 Section 5.6, p. 357 ], [ @Kato Section III.6.1, pp. 174–175 ]. \[prop: Gilbarg\_Trudinger\_theorem\_8 - 6\ ] Let $ (X, g)$ be a shut, Riemannian, placid manifold of property $ d \geq 2 $, and $ G$ be a compact Lie group, $ P$ be a placid principal $ G$-bundle over $ X$, and $ l \geq 0 $ be an integer. If $ A$ is a $ W^{1,q}$ joining on $ P$ with $ d/2 < q < \infty$, and $ A_1 $ is a $ C^\infty$ reference book connection on $ P$, and $ phosphorus \in (1,\infty)$ obeys $ d/2 \leq p \leq q$, then the spectrum, $ \sigma(\Delta_A)$, of the unbounded hustler, $ $ \Delta_A: \sD(\Delta_A) \subset L^p(X;\Lambda^l\otimes\ad P) \to L^p(X;\Lambda^l\otimes\ad P),$$ is countable without collection points, consisting of non - negative, substantial eigenvalues, $ \lambda$, with finite multiplicities, $ \dim\Ker(\Delta_A-\lambda)$. Corollary \[cor: Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\ ] imply that the operator, $ $ \Delta_A: W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad phosphorus) \to L^p(X;\Lambda^l\otimes\ad P),$$ is Fredholm and, setting $ kilobyte: = \Ker(\Delta_A: W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P) \to L^

is an *eigenvector* of $T$ and satisfies the gqyation $Tx = \lzmbda x$. Iw $T:\sD(\sX)\subset\sX\to\sX$ is a clodee opeeator with dense domaiv, $\sD(\sX) \suhset \sX$, rhen qe say thav $\lambda \notin \digmc(T)$ if the operatpr $T-\lambda:\vD(\sX)\to\sX$ has a buuuded inverse and otherwise that $\lambqa \in \sogla(T)$ [@KadisonRindrost1 Sqctikn 5.6, p. 357], [@Kato Section III.6.1, pp. 174–175]. \[prop:Gilgarg\_Truvinger\_theorem\_8-6\] Lrt $(X,g)$ be a closed, Riemanniwn, slooth manifold of fimension $d \geq 2$, and $G$ be a zompact Lit yroup, $P$ be z smooth principal $G$-bundle over $X$, anb $l \geq 0$ be ab ijjeger. If $A$ iw a $W^{1,z}$ connection on $P$ widh $d/2 < q < \infty$, and $A_1$ ls a $R^\infry$ reference connectimn on $P$, and $p \in (1,\igfty)$ obeyv $b/2 \leq p \leq q$, then thw wpecttum, $\shgma(\Awltx_A)$, kf tge unblunved operatod, $$\Delta_A: \sD(\Eelta_A) \subset L^p(X;\Lakbqq^l\otimes\ad P) \fo L^p(X;\Jaibda^l\otimes\ad P),$$ is countable without acbumumation points, consistint of non-negative, real eigenvaltes, $\lambda$, with finite multiplicities, $\dim\Ker(\Delta_D-\lambva)$. Zoriljxey \[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sofkltv\_cpnnection\] implles that the operayog, $$\Cglta_A: W_{A_1}^{2,p}(X;\Lambaa^l\otimes\zd P) \to L^p(X;\Lambda^l\ltimes\aq P),$$ iw Fredholi anc, setting $K := \Ker(\Delta_A: W_{A_1}^{2,p}(X;\Oambda^l\otimef\qd P) \to L^

is an *eigenvector* of $T$ and satisfies $Tx \lambda x$. $T:\sD(\sX)\subset\sX\to\sX$ is a $\sD(\sX) \sX$, then we that $\lambda \notin if the operator $T-\lambda:\sD(\sX)\to\sX$ has a inverse and otherwise that $\lambda \in \sigma(T)$ [@KadisonRingrose1 Section 5.6, p. 357], [@Kato III.6.1, pp. 174–175]. \[prop:Gilbarg\_Trudinger\_theorem\_8-6\] Let $(X,g)$ be a closed, Riemannian, smooth manifold of $d 2$, $G$ a compact Lie group, $P$ be a smooth principal $G$-bundle over $X$, and $l \geq 0$ an integer. If $A$ is a $W^{1,q}$ connection $P$ with $d/2 < < \infty$, and $A_1$ is $C^\infty$ connection on and \in obeys $d/2 \leq \leq q$, then the spectrum, $\sigma(\Delta_A)$, of the unbounded operator, $$\Delta_A: \sD(\Delta_A) \subset L^p(X;\Lambda^l\otimes\ad P) \to L^p(X;\Lambda^l\otimes\ad is countable points, consisting non-negative, eigenvalues, with finite multiplicities, \[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\] implies that the operator, $$\Delta_A: L^p(X;\Lambda^l\otimes\ad P),$$ is Fredholm and, setting $K := W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P) L^

is an *eigenvector* of $T$ and satiSfies the eqUatioN $Tx = \LamBdA x$. If $t:\sD(\sx)\subset\sX\to\sX$ iS A cloSed operator with dense doMain, $\sd(\sx) \SubsET \sx$, then We say thAT $\lAMBda \NoTiN \siGmA(t)$ iF the oPerAtor $T-\laMbda:\sD(\sX)\to\SX$ hAs A bounded inveRSe And otherwiSe tHat $\lambda \in \sIgmA(T)$ [@KadiSoNRiNGrose1 secTion 5.6, p. 357], [@kato SeCTion IIi.6.1, pp. 174–175]. \[prop:GiLbARg\_TrudINger\_theOREm\_8-6\] let $(X,G)$ be a closed, RiemannIAn, SMooth manifold oF dimenSiON $d \GEQ 2$, anD $G$ bE a compact LIe Group, $p$ Be a smooTH pRINCipAL $G$-bundle over $X$, And $l \geq 0$ be an INteGer. If $A$ Is A $W^{1,q}$ COnnectIon on $p$ wITh $d/2 < Q < \infty$, and $A_1$ iS a $C^\iNfty$ referEnce coNNection ON $P$, and $p \iN (1,\infty)$ ObeYs $d/2 \Leq p \LEq Q$, tHen ThE SpeCTrUm, $\sIGma(\delta_A)$, of ThE uNbounDed oPERATor, $$\DEltA_A: \sD(\delta_a) \subset L^p(X;\LamBda^L\otiMEs\aD P) \to L^P(X;\LamBda^l\OtImes\aD P),$$ is coUntabLe Without accumulaTion Points, conSisTiNg oF nOn-negATive, reAl eIgeNvalues, $\Lambda$, wITh fInITE MuLtiplicities, $\dim\Ker(\deLTA_A-\Lambda)$. CoRollarY \[CoR:FREdholmneSs\_And\_IndeX\_lAplacE\_opeRAtOr\_on\_W2p\_SoBolev\_cONnEcTion\] impLiEs that ThE opEraTor, $$\DeLTa_A: W_{a_1}^{2,p}(X;\LamBda^l\otimEs\ad P) \TO L^p(X;\Lambda^l\otiMEs\ad P),$$ is FredhoLM aND, SeTTing $k := \KeR(\Delta_A: W_{A_1}^{2,p}(X;\lambDA^l\otImes\AD P) \To L^

is an *eigenvector* of $T $ and sati sfies th e e qu atio n $T x = \lambda x$ . If $T:\sD(\sX)\subset\sX \to\s X$ is a cl osedoperato r w i t h d en se do ma i n, $\sD (\s X) \sub set \sX$,the nwe say that$ \l ambda \not in\sigma(T)$ i f t he ope ra tor $T-\l amb da:\s D(\sX) \ to\sX$ has a bo un d ed inv e rse and o th erwi se that $\lambda\ in \sigma(T)$ [@K adison Ri n gr o s e1Sec tion 5.6,p. 357] , [@Kato Se c t i onI II.6.1, pp. 1 74–175]. \ [ pro p:Gilb ar g\_ T ruding er\_t he o rem \_8-6\] Let $(X ,g)$ be a close d , Riema n nian, s moothman ifo ld o f d im ens io n $d \g eq2 $,and $G$be a comp actL i e grou p,$P$be asmooth princi pal $G$ - bun dle o ver $ X$,an d $l\geq 0 $ bean integer. If $A $ is a $W^{1, q}$ c onn ec tiono n $P$wit h $ d/2 < q < \inf t y$, a n d $A _1$ is a $C^\infty $r e fe rence co nnecti o non $P$, and $ p \ in ( 1 , \inft y)$o be ys $d/2\leq p \l eq q$, th en the s pe ctr um, $\si g ma(\ Delta_ A)$, ofthe u n bounded operat o r, $$\Delta_A : \ s D (\ D elta _A) \subset L^ p(X; \ Lamb da^l \ ot ime s \ad P ) \to L ^ p( X ;\Lambda^l\otimes\a dP),$$is co untable witho ut accumul a t i on point s, c o ns i sting of non-n egati ve, real e i genvalue s, $\ lambda$, with fin i t e multip lic iti es, $\ d i m\ Ker(\Delta_A- \ l ambd a) $. Cor oll ary \[c or: Fre dho lmn es s\_and\_i ndex\_La pl ac e\ _o per ator\ _ on\_W2p\ _S obo le v\_ conne c tion\] impl iesth at the operat o r, $ $\De lt a_ A: W _{A _1 }^{2, p}(X ; \La mbda^l\ otimes\ad P) \toL^ p( X;\Lamb da^l\otimes\a dP),$$ is F re dho lm and , setting$K := \Ker(\Delta_A: W_ { A_1}^{2 ,p} (X;\L ambd a^l\otime s\a d P) \ toL ^

is_an *eigenvector*_of $T$ and satisfies_the equation_$Tx_= \lambda_x$. If_$T:\sD(\sX)\subset\sX\to\sX$ is a_closed operator with_dense domain, $\sD(\sX) \subset_\sX$, then we_say_that $\lambda \notin \sigma(T)$ if the operator $T-\lambda:\sD(\sX)\to\sX$ has a bounded inverse and otherwise_that_$\lambda \in_\sigma(T)$_[@KadisonRingrose1_Section 5.6, p. 357], [@Kato_Section III.6.1, pp. 174–175]. \[prop:Gilbarg\_Trudinger\_theorem\_8-6\] Let_$(X,g)$ be_a closed, Riemannian, smooth manifold of dimension $d_\geq_2$, and $G$_be a compact Lie group, $P$ be a smooth_principal $G$-bundle over $X$, and $l_\geq 0$ be_an_integer._If $A$ is a_$W^{1,q}$ connection on $P$ with $d/2_< q < \infty$, and $A_1$_is a $C^\infty$ reference connection on $P$,_and $p \in (1,\infty)$ obeys $d/2_\leq p \leq q$, then_the spectrum,_$\sigma(\Delta_A)$, of the unbounded operator,_$$\Delta_A: \sD(\Delta_A) \subset_L^p(X;\Lambda^l\otimes\ad P)_\to L^p(X;\Lambda^l\otimes\ad P),$$_is countable without accumulation points, consisting_of non-negative, real_eigenvalues, $\lambda$, with finite multiplicities, $\dim\Ker(\Delta_A-\lambda)$. Corollary_\[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\]_implies that the_operator,_$$\Delta_A:_W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P)_\to L^p(X;\Lambda^l\otimes\ad P),$$_is_Fredholm and,_setting_$K := \Ker(\Delta_A: W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P) \to_L^

$ is the same as if $\hat{\gamma}=\gamma_{0}.$ Define $m_{i}=M^{\prime}W\lambda_{0}(x_{i})\rho _{i}$ and $m_{i}^{\ast}=\tilde{\lambda}^{\Sigma^{\ast}}(x_{i},\lambda_{\beta })\Sigma^{\ast}(x_{i})^{-1}\rho_{i}.$ The asymptotic variance of the GMM estimator for orthogonal instruments $\lambda_{0}(x)$ is$$(M^{\prime}WM)^{-1}M^{\prime}WE[\lambda_{0}(x_{i})\rho_{i}\rho_{i}^{\prime }\lambda_{0}(x_{i})^{\prime}]WM(M^{\prime}WM)^{-1}=(E[m_{i}m_{i}^{\ast\prime }])^{-1}E[m_{i}m_{i}^{\prime}](E[m_{i}m_{i}^{\ast}])^{-1\prime}.$$ The fact that this matrix is minimized in the positive semidefinite sense for $m_{i}=m_{i}^{\ast}$ is well known, e.g. see Newey and McFadden (1994). *Q.E.D.* The following result is useful for the results of Section 7: <span style="font-variant:small-caps;">Lemma A1:</span> *If Assumption 4 is satisfied then* $\hat{R}_{1}\overset{p}{\longrightarrow}0.$ *If Assumption 5 is satisfied then* $\hat{R}_{2}\overset{p}{\longrightarrow}0.$ Proof: Define $\hat{\Delta}_{i\ell}=m(z_{i},\hat{\gamma}_{\ell})-m(z_{i},\gamma_{0})-\bar{m}(\hat{\gamma}_{\ell})$ for $i\in I_{\ell}$ and let $Z_{\ell}^{c}$ denote the observations $z_{i}$ for $i\notin I_{\ell}$. Note that $\hat{\gamma}_{\ell}$ depends only on $Z_{\ell}^{c}$. By construction and independence of $Z_{\ell}^{c}$ and $z_{i},i\in I_{\ell}$ we have $E[\hat{\Delta}_{i\ell}|Z_{\ell}^{c}]=0.$ Also by independence of the observations, $E[\hat{\Delta}_{i\ell}\hat{\Delta}_{j\ell}|Z_{\ell

$ is the same as if $ \hat{\gamma}=\gamma_{0}.$ Define $ m_{i}=M^{\prime}W\lambda_{0}(x_{i})\rho _ { i}$ and $ m_{i}^{\ast}=\tilde{\lambda}^{\Sigma^{\ast}}(x_{i},\lambda_{\beta }) \Sigma^{\ast}(x_{i})^{-1}\rho_{i}.$ The asymptotic variance of the GMM estimator for orthogonal instrument $ \lambda_{0}(x)$ is$$(M^{\prime}WM)^{-1}M^{\prime}WE[\lambda_{0}(x_{i})\rho_{i}\rho_{i}^{\prime } \lambda_{0}(x_{i})^{\prime}]WM(M^{\prime}WM)^{-1}=(E[m_{i}m_{i}^{\ast\prime } ]) ^{-1}E[m_{i}m_{i}^{\prime}](E[m_{i}m_{i}^{\ast}])^{-1\prime}.$$ The fact that this matrix is minimize in the positive semidefinite sense for $ m_{i}=m_{i}^{\ast}$ is well known, e.g. witness Newey and McFadden (1994). * Q.E.D. * The following result is utilitarian for the results of Section 7: < span style="font - variant: minor - caps;">Lemma A1:</span > * If Assumption 4 is satisfied then * $ \hat{R}_{1}\overset{p}{\longrightarrow}0.$ * If Assumption 5 is quenched then * $ \hat{R}_{2}\overset{p}{\longrightarrow}0.$ Proof: Define $ \hat{\Delta}_{i\ell}=m(z_{i},\hat{\gamma}_{\ell})-m(z_{i},\gamma_{0})-\bar{m}(\hat{\gamma}_{\ell})$ for $ i\in I_{\ell}$ and let $ Z_{\ell}^{c}$ denote the observation $ z_{i}$ for $ i\notin I_{\ell}$. Note that $ \hat{\gamma}_{\ell}$ depends merely on $ Z_{\ell}^{c}$. By construction and independence of $ Z_{\ell}^{c}$ and $ z_{i},i\in I_{\ell}$ we have $ E[\hat{\Delta}_{i\ell}|Z_{\ell}^{c}]=0.$ Also by independence of the observation, $ E[\hat{\Delta}_{i\ell}\hat{\Delta}_{j\ell}|Z_{\ell

$ is the same as if $\hat{\gamma}=\namma_{0}.$ Define $m_{i}=M^{\krume}W\lakbda_{0}(x_{i})\dho _{i}$ and $m_{i}^{\ast}=\tilde{\lambda}^{\Sigma^{\ast}}(x_{i},\lembdq_{\beta })\Wigma^{\ast}(x_{i})^{-1}\rho_{i}.$ The asyoptotic vwriance if tie GMM estimator for ortmjgonzp invvruments $\lambda_{0}(w)$ is$$(M^{\prime}WK)^{-1}M^{\prime}WE[\lambdd_{0}(x_{k})\rko_{i}\rho_{i}^{\prime }\lambda_{0}(x_{i})^{\prime}]WM(M^{\prime}WM)^{-1}=(E[i_{i}m_{i}^{\ast\lrlme }])^{-1}E[m_{i}m_{i}^{\prime}](E[i_{i}m_{i}^{\sft}])^{-1\prjme}.$$ The fact that this matrix is mjnimizev in the positife semidefinite sense for $l_{i}=m_{i}^{\wst}$ is well known, f.g. see Newei anq McFadden (1994). *Q.D.D.* The following result is useful for the results of Sdctiou 7: <span stylg="yint-gdriant:small-raps;">Leima A1:</span> *If Assumpthon 4 is satisfied thek* $\hat{C}_{1}\oveeset{p}{\longrightarrow}0.$ *Ih Assumption 5 is satysfied than* $\hat{R}_{2}\overset{p}{\longeithtartow}0.$ Prmof: Awfive $\gav{\Demta}_{i\elp}=m(z_{m},\hat{\gamma}_{\elm})-m(z_{i},\gamma_{0})-\bae{m}(\hat{\gamma}_{\ell})$ for $i\on P_{\rll}$ and let $A_{\ell}^{c}$ qegote the observations $z_{i}$ for $i\notin I_{\eln}$. Nkte that $\hat{\gamma}_{\ell}$ depends only on $Z_{\ell}^{c}$. Bi construceion and independence of $Z_{\ell}^{c}$ and $z_{i},i\in I_{\ell}$ we vave $X[\hxt{\Dtlbw}_{k\wlp}|Z_{\ell}^{c}]=0.$ Also by independence of the observatioga, $T[\han{\Delta}_{i\ell}\hat{\Deltc}_{j\ell}|Z_{\ell

$ is the same as if $\hat{\gamma}=\gamma_{0}.$ _{i}$ $m_{i}^{\ast}=\tilde{\lambda}^{\Sigma^{\ast}}(x_{i},\lambda_{\beta })\Sigma^{\ast}(x_{i})^{-1}\rho_{i}.$ asymptotic variance of instruments is$$(M^{\prime}WM)^{-1}M^{\prime}WE[\lambda_{0}(x_{i})\rho_{i}\rho_{i}^{\prime }\lambda_{0}(x_{i})^{\prime}]WM(M^{\prime}WM)^{-1}=(E[m_{i}m_{i}^{\ast\prime }])^{-1}E[m_{i}m_{i}^{\prime}](E[m_{i}m_{i}^{\ast}])^{-1\prime}.$$ fact that this is minimized in the positive semidefinite for $m_{i}=m_{i}^{\ast}$ is well known, e.g. see Newey and McFadden (1994). *Q.E.D.* The result is useful for the results of Section 7: <span style="font-variant:small-caps;">Lemma A1:</span> *If 4 satisfied $\hat{R}_{1}\overset{p}{\longrightarrow}0.$ Assumption 5 is satisfied then* $\hat{R}_{2}\overset{p}{\longrightarrow}0.$ Proof: Define $\hat{\Delta}_{i\ell}=m(z_{i},\hat{\gamma}_{\ell})-m(z_{i},\gamma_{0})-\bar{m}(\hat{\gamma}_{\ell})$ for $i\in I_{\ell}$ and let $Z_{\ell}^{c}$ denote observations $z_{i}$ for $i\notin I_{\ell}$. Note that $\hat{\gamma}_{\ell}$ only on $Z_{\ell}^{c}$. By and independence of $Z_{\ell}^{c}$ and I_{\ell}$ have $E[\hat{\Delta}_{i\ell}|Z_{\ell}^{c}]=0.$ by of observations, $E[\hat{\Delta}_{i\ell}\hat{\Delta}_{j\ell}|Z_{\ell

$ is the same as if $\hat{\gamma}=\gammA_{0}.$ Define $m_{i}=M^{\Prime}w\laMbdA_{0}(x_{I})\rho _{I}$ and $M_{i}^{\ast}=\tilde{\lambDA}^{\SigMa^{\ast}}(x_{i},\lambda_{\beta })\Sigma^{\Ast}(x_{i})^{-1}\RhO_{I}.$ The ASyMptotIc variaNCe OF The gMm eStiMaTOr For orThoGonal inStruments $\lAmbDa_{0}(X)$ is$$(M^{\prime}WM)^{-1}M^{\PRiMe}WE[\lambda_{0}(X_{i})\rHo_{i}\rho_{i}^{\prime }\LamBda_{0}(x_{i})^{\pRiMe}]Wm(m^{\primE}WM)^{-1}=(e[m_{i}m_{i}^{\Ast\priME }])^{-1}E[m_{i}m_{i}^{\Prime}](E[m_{i}m_{I}^{\aST}])^{-1\prime}.$$ tHe fact tHAT tHis mAtrix is minimized iN ThE Positive semideFinite SeNSe FOR $m_{i}=M_{i}^{\aSt}$ is well knOwN, e.g. seE newey anD mcfADDen (1994). *q.e.D.* The followinG result is usEFul For the ReSulTS of SecTion 7: <sPaN StyLe="font-variaNt:smAll-caps;">LeMma A1:</spAN> *If AssuMPtion 4 is SatisfIed TheN* $\hat{r}_{1}\OvErSet{P}{\lONgrIGhTarROw}0.$ *IF AssumptIoN 5 iS satiSfieD THEN* $\hat{r}_{2}\ovErseT{p}{\lonGrightarrow}0.$ PrOof: defiNE $\haT{\DeltA}_{i\ell}=M(z_{i},\hAt{\Gamma}_{\Ell})-m(z_{i},\Gamma_{0})-\BaR{m}(\hat{\gamma}_{\ell})$ foR $i\in i_{\ell}$ and leT $Z_{\eLl}^{C}$ deNoTe the OBservaTioNs $z_{I}$ for $i\noTin I_{\ell}$. nOte ThAT $\HAt{\Gamma}_{\ell}$ depends onlY oN $z_{\ElL}^{c}$. By consTructiON aNd INdependeNcE of $z_{\ell}^{C}$ ANd $z_{i},i\In I_{\eLL}$ wE have $E[\haT{\Delta}_{I\ElL}|Z_{\Ell}^{c}]=0.$ AlsO bY indepEnDenCe oF the oBServAtions, $e[\hat{\DeltA}_{i\ell}\HAt{\Delta}_{j\ell}|Z_{\elL

$ is the same as if $\hat{ \gamma}=\g amma_ {0} .$De fine $m_ {i}=M^{\prime} W \lam bda_{0}(x_{i})\rho _{i }$ an d$ m_{i } ^{ \ast} =\tilde { \l a m bda }^ {\ Sig ma ^ {\ ast}} (x_ {i},\la mbda_{\bet a } )\ Sigma^{\ast} ( x_ {i})^{-1}\ rho _{i}.$ The a sym ptotic v ari a nce o f t he GM M esti m ator f or orthog on a l inst r uments$ \ la mbda _{0}(x)$ is$$(M^{ \ pr i me}WM)^{-1}M^{ \prime }W E [\ l a mbd a_{ 0}(x_{i})\ rh o_{i} \ rho_{i} ^ {\ p r i me} \lambda_{0}(x _{i})^{\pri m e}] WM(M^{ \p rim e }WM)^{ -1}=( E[ m _{i }m_{i}^{\as t\pr ime }])^{ -1}E[m _ {i}m_{i } ^{\prim e}](E[ m_{ i}m _{i} ^ {\ as t}] )^ { -1\ p ri me} . $$The fact t ha t thi s ma t r i x ismin imiz ed in the positive se mide f ini te se nse f or $ m_ {i}=m _{i}^{ \ast} $is well known,e.g. see Newe y a nd Mc Fa dden( 1994). *Q .E. D.* Th e follo w ing r e s u lt is useful for the r e s ul ts of Se ction7 : < s pan styl e= "fo nt-v a r iant: smal l -c aps;">Le mma A1 : </ sp an> *If A ssumpt io n 4 is sati s fied then* $\hat{R }_{1} \ overset{p}{\lo n grightarrow}0 . $* I fA ssum pti on 5 is sat isfi e d th en*$ \h at{ R }_{2} \over se t {p } {\longrightarrow}0. $ Proof : Def ine $\hat{\De lta}_{i\el l } = m(z_{i}, \hat { \g a mma}_{\ell})-m (z_{i },\gamma_{ 0 })-\bar{ m}(\h at{\gamm a}_{\ell} ) $ for $i\ inI_{ \el l}$ a nd let $Z_{\ell } ^ {c}$ d enote t heobserva tio ns$z_ {i} $for $i\no tin I_{\ el l} $. N ote that $\hat{\g am ma} _{ \el l}$ d e pendsonlyon $ Z_ {\ e ll} ^{c}$.B yc o nstr uc ti on a ndin depen denc e of $Z_{\e ll}^{c}$and $z_{ i} ,i \in I_{ \ell}$ we hav e$E[\hat{\D el ta} _{i\el l } |Z_{\ell }^{c}]=0.$ Also by inde p endence of theobse rvations, $E [\hat{ \De l ta}_{i \ell}\ hat{\ De lta } _ {j\el l } |Z _{\ el l

$ is_the same_as if $\hat{\gamma}=\gamma_{0}.$ Define_$m_{i}=M^{\prime}W\lambda_{0}(x_{i})\rho _{i}$ and_$m_{i}^{\ast}=\tilde{\lambda}^{\Sigma^{\ast}}(x_{i},\lambda_{\beta })\Sigma^{\ast}(x_{i})^{-1}\rho_{i}.$_The asymptotic_variance_of the GMM_estimator for orthogonal_instruments $\lambda_{0}(x)$ is$$(M^{\prime}WM)^{-1}M^{\prime}WE[\lambda_{0}(x_{i})\rho_{i}\rho_{i}^{\prime }\lambda_{0}(x_{i})^{\prime}]WM(M^{\prime}WM)^{-1}=(E[m_{i}m_{i}^{\ast\prime }])^{-1}E[m_{i}m_{i}^{\prime}](E[m_{i}m_{i}^{\ast}])^{-1\prime}.$$ The_fact that this_matrix_is minimized in the positive semidefinite sense for $m_{i}=m_{i}^{\ast}$ is well known, e.g. see_Newey_and McFadden_(1994)._*Q.E.D.* The_following result is useful for_the results of Section 7: <span_style="font-variant:small-caps;">Lemma A1:</span>_*If Assumption 4 is satisfied then* $\hat{R}_{1}\overset{p}{\longrightarrow}0.$ *If_Assumption_5 is satisfied_then* $\hat{R}_{2}\overset{p}{\longrightarrow}0.$ Proof: Define $\hat{\Delta}_{i\ell}=m(z_{i},\hat{\gamma}_{\ell})-m(z_{i},\gamma_{0})-\bar{m}(\hat{\gamma}_{\ell})$ for $i\in I_{\ell}$ and let_$Z_{\ell}^{c}$ denote the observations $z_{i}$ for_$i\notin I_{\ell}$. Note_that_$\hat{\gamma}_{\ell}$_depends only on $Z_{\ell}^{c}$._By construction and independence of $Z_{\ell}^{c}$_and $z_{i},i\in I_{\ell}$ we have $E[\hat{\Delta}_{i\ell}|Z_{\ell}^{c}]=0.$_Also by independence of the observations, $E[\hat{\Delta}_{i\ell}\hat{\Delta}_{j\ell}|Z_{\ell

0.71 & 13.97 & 1.43& 2MASSW J0929+34 & L8 && 1.69 & 16.69 & 1.96& 1.07 & 0.89& 14.73& & 2MASSW J1523+30 &L8 & 13.00 & 1.65 & 15.95 & 1.60 & 0.90 & 0.70 & 14.35 & 1.49 & SDSS 0032+14 & L8 & & 1.67 & 16.58 & 1.59 & 0.92 &0.67 & 14.99 & & SDSS 0857+57 &L8 && 1.72& 14.80 & 1.86 & 1.00 & 0.86& 12.94 & 1.63& –0.19 SDSS 0830+48 & L9 & & 1.64 & 15.22 & 1.54 & 0.82 & 0.72 & 13.68 & 1.70& 2MASSW J0310+16 & L9 & & & 15.88 & 1.70 & 0.97& 0.73 & 14.18 & 1.64& 2MASSW J0328+23 & L9.5 & & 1.71 & 16.35 & 1.48 & 0.88 & 0.60 & 14.87 & & SDSS 0423–04 & T0 & & 1.68 & 14.30 & 1.34 & 0.79 & 0.55 & 12.96 & 1.51& SDSS 0837–00 & T0.5 & & 1.69 & 16.90 & 0.92 & 0.69 & 0.23 & 15.98 & & SDSS 0151+12 & T1 & & 1.84 & 16.25 & 1.07 & 0.71 & 0.36 & 15.18 & & SDSS 1254–01 & T2 & & 1.74 & 14.66 & 0.82 & 0.53 & 0.29 & 13.84 & 1.59 & –0.40 SDSS 1021–03 & T3 & & 1.78 & 15.88 & 0.62 & 0.47 & 0.15 & 15.26 & & SDSS 1750+17 & T3.5 & & & 16.14 & 0.12 & 0.20 & –0.08 & 16.02 & & 2

0.71 & 13.97 & 1.43 & 2MASSW J0929 + 34 & L8 & & 1.69 & 16.69 & 1.96 & 1.07 & 0.89 & 14.73 & & 2MASSW J1523 + 30 & L8 & 13.00 & 1.65 & 15.95 & 1.60 & 0.90 & 0.70 & 14.35 & 1.49 & SDSS 0032 + 14 & L8 & & 1.67 & 16.58 & 1.59 & 0.92 & 0.67 & 14.99 & & SDSS 0857 + 57 & L8 & & 1.72 & 14.80 & 1.86 & 1.00 & 0.86 & 12.94 & 1.63 & – 0.19 SDSS 0830 + 48 & L9 & & 1.64 & 15.22 & 1.54 & 0.82 & 0.72 & 13.68 & 1.70 & 2MASSW J0310 + 16 & L9 & & & 15.88 & 1.70 & 0.97 & 0.73 & 14.18 & 1.64 & 2MASSW J0328 + 23 & L9.5 & & 1.71 & 16.35 & 1.48 & 0.88 & 0.60 & 14.87 & & SDSS 0423–04 & T0 & & 1.68 & 14.30 & 1.34 & 0.79 & 0.55 & 12.96 & 1.51 & SDSS 0837–00 & T0.5 & & 1.69 & 16.90 & 0.92 & 0.69 & 0.23 & 15.98 & & SDSS 0151 + 12 & T1 & & 1.84 & 16.25 & 1.07 & 0.71 & 0.36 & 15.18 & & SDSS 1254–01 & T2 & & 1.74 & 14.66 & 0.82 & 0.53 & 0.29 & 13.84 & 1.59 & – 0.40 SDSS 1021–03 & T3 & & 1.78 & 15.88 & 0.62 & 0.47 & 0.15 & 15.26 & & SDSS 1750 + 17 & T3.5 & & & 16.14 & 0.12 & 0.20 & – 0.08 & 16.02 & & 2

0.71 & 13.97 & 1.43& 2MASSW J0929+34 & L8 && 1.69 & 16.69 & 1.96& 1.07 & 0.89& 14.73& & 2OASSW J1523+30 &L8 & 13.00 & 1.65 & 15.95 & 1.60 & 0.90 & 0.70 & 14.35 & 1.49 & SSSS 0032+14 & L8 & & 1.67 & 16.58 & 1.59 & 0.92 &0.67 & 14.99 & & SDSS 0857+57 &L8 && 1.72& 14.80 & 1.86 & 1.00 & 0.86& 12.94 & 1.63& –0.19 WDSS 0830+48 & L9 & & 1.64 & 15.22 & 1.54 & 0.82 & 0.72 & 13.68 & 1.70& 2MASSW J0310+16 & L9 & & & 15.88 & 1.70 & 0.97& 0.73 & 14.18 & 1.64& 2MASSW O0328+23 & L9.5 & & 1.71 & 16.35 & 1.48 & 0.88 & 0.60 & 14.87 & & WDSS 0423–04 & T0 & & 1.68 & 14.30 & 1.34 & 0.79 & 0.55 & 12.96 & 1.51& VDSS 0837–00 & T0.5 & & 1.69 & 16.90 & 0.92 & 0.69 & 0.23 & 15.98 & & SDSS 0151+12 & T1 & & 1.84 & 16.25 & 1.07 & 0.71 & 0.36 & 15.18 & & SDSS 1254–01 & T2 & & 1.74 & 14.66 & 0.82 & 0.53 & 0.29 & 13.84 & 1.59 & –0.40 SDFS 1021–03 & E3 & & 1.78 & 15.88 & 0.62 & 0.47 & 0.15 & 15.26 & & SDSS 1750+17 & T3.5 & & & 16.14 & 0.12 & 0.20 & –0.08 & 16.02 & & 2

0.71 & 13.97 & 1.43& 2MASSW J0929+34 && & 16.69 1.96& 1.07 & &L8 13.00 & 1.65 15.95 & 1.60 0.90 & 0.70 & 14.35 & & SDSS 0032+14 & L8 & & 1.67 & 16.58 & 1.59 & &0.67 & 14.99 & & SDSS 0857+57 &L8 && 1.72& 14.80 & 1.86 1.00 0.86& & –0.19 SDSS 0830+48 & L9 & & 1.64 & 15.22 & 1.54 & 0.82 & 0.72 13.68 & 1.70& 2MASSW J0310+16 & L9 & & 15.88 & 1.70 0.97& 0.73 & 14.18 & 2MASSW & L9.5 & & & 1.48 & & 0.60 & 14.87 & & SDSS 0423–04 & T0 & & 1.68 & 14.30 & 1.34 0.79 & 12.96 & SDSS & & & 1.69 & 0.92 & 0.69 & 0.23 & SDSS 0151+12 & T1 & & 1.84 16.25 & & 0.71 & 0.36 & 15.18 & SDSS 1254–01 & T2 & & 1.74 14.66 & 0.82 & 0.53 & 0.29 & 13.84 & 1.59 & –0.40 SDSS 1021–03 & & 1.78 & & 0.62 & & & & SDSS 1750+17 T3.5 & & & 16.14 & 0.12 & 0.20 & –0.08 16.02 & & 2

0.71 & 13.97 & 1.43& 2MASSW J0929+34 & L8 && 1.69 & 16.69 & 1.96& 1.07 & 0.89& 14.73& & 2MASSW J1523+30 &L8 & 13.00 & 1.65 & 15.95 & 1.60 & 0.90 & 0.70 & 14.35 & 1.49 & SDSS 0032+14 & L8 & & 1.67 & 16.58 & 1.59 & 0.92 &0.67 & 14.99 & & SDSS 0857+57 &L8 && 1.72& 14.80 & 1.86 & 1.00 & 0.86& 12.94 & 1.63& –0.19 SdSS 0830+48 & L9 & & 1.64 & 15.22 & 1.54 & 0.82 & 0.72 & 13.68 & 1.70& 2MASSW J0310+16 & l9 & & & 15.88 & 1.70 & 0.97& 0.73 & 14.18 & 1.64& 2MASSw J0328+23 & L9.5 & & 1.71 & 16.35 & 1.48 & 0.88 & 0.60 & 14.87 & & sDSs 0423–04 & T0 & & 1.68 & 14.30 & 1.34 & 0.79 & 0.55 & 12.96 & 1.51& sDSS 0837–00 & t0.5 & & 1.69 & 16.90 & 0.92 & 0.69 & 0.23 & 15.98 & & SDSs 0151+12 & T1 & & 1.84 & 16.25 & 1.07 & 0.71 & 0.36 & 15.18 & & SDSS 1254–01 & T2 & & 1.74 & 14.66 & 0.82 & 0.53 & 0.29 & 13.84 & 1.59 & –0.40 SDSS 1021–03 & T3 & & 1.78 & 15.88 & 0.62 & 0.47 & 0.15 & 15.26 & & SDss 1750+17 & T3.5 & & & 16.14 & 0.12 & 0.20 & –0.08 & 16.02 & & 2

0.71 & 13.97 & 1.43& 2MASS W J0929+34 & L8 && 1. 69 & 1 6.69 & 1.96& 1.07& 0.8 9& 14.73& & 2MASSW J15 23+30 & L 8 &1 3. 00 &1.65 &1 5. 9 5 &1. 60 &0. 9 0& 0.7 0 & 14.35& 1.49 & S DSS 0 032+14 & L8& & 1.67 & 16 .58 & 1.59 & 0. 92&0.67&14. 9 9 & & SD SS 08 57+57& L8 &&1.72& 14. 80 & 1.86 & 1.00& 0. 86&12.94 & 1.63& –0. 1 9S DSS 0830+48 &L9 & & 1 . 64 & 15 .22 & 1.54 &0. 82 &0 .72 & 1 3 .6 8 & 1. 7 0& 2MASSW J03 10+16 & L9& && 15.8 8& 1 . 70 & 0 .97&0. 7 3 & 14.18 & 1. 64&2MASSW J0 328+23 & L9.5& & 1.71 & 16. 35& 1 .48& 0 .8 8 & 0 . 60& 1 4.8 7 && SDSS 0 42 3– 04 &T0 & & 1 .68& 1 4.30 & 1. 34 & 0.79 & 0 .55 & 1 2 .96 & 1. 51& S DSS08 37–00 & T0. 5 & & 1 .69 & 16.90 & 0 .92& 0.69 &0.2 3& 1 5. 98 && SDSS015 1+1 2 & T1& & 1.8 4 &16 . 2 5 & 1.07 & 0.71 & 0.3 6& 15 .18 & &SDSS 1 2 54 –0 1 & T2 &&1.7 4 &1 4 .66 & 0.8 2 & 0.53 &0.29 & 13 .8 4 & 1.5 9& –0.4 0SDS S 1 021–0 3 & T 3 & &1.78 & 1 5.88& 0.62 & 0.47 & 0.15 & 15.26& & S DS S 175 0+1 7 & T3.5 && &1 6.14 & 0 . 12 &0 .20 & –0.0 8& 1 6 .02 & & 2

0.71 &_13.97 &_1.43& 2MASSW J0929+34 &_L8 &&_1.69_& 16.69_&_1.96& 1.07 &_0.89& 14.73& &_2MASSW J1523+30 &L8 &_13.00 & 1.65_&_15.95 & 1.60 & 0.90 & 0.70 & 14.35 & 1.49 & SDSS 0032+14_&_L8 &_&_1.67_& 16.58 & 1.59 &_0.92 &0.67 & 14.99 &_& SDSS_0857+57 &L8 && 1.72& 14.80 & 1.86 &_1.00_& 0.86& 12.94_& 1.63& –0.19 SDSS 0830+48 & L9 & &_1.64 & 15.22 & 1.54 &_0.82 & 0.72_&_13.68_& 1.70& 2MASSW J0310+16_& L9 & & & 15.88_& 1.70 & 0.97& 0.73 &_14.18 & 1.64& 2MASSW J0328+23 & L9.5_& & 1.71 & 16.35 &_1.48 & 0.88 & 0.60_& 14.87_& & SDSS 0423–04 &_T0 & &_1.68 &_14.30 & 1.34_& 0.79 & 0.55 & 12.96_& 1.51& SDSS_0837–00 & T0.5 & & 1.69_&_16.90 & 0.92_&_0.69_& 0.23_& 15.98 &_&_SDSS 0151+12_&_T1 & & 1.84 & 16.25_&_1.07 & 0.71 & 0.36 & 15.18_& & SDSS 1254–01_&_T2 & & 1.74_& 14.66 & 0.82 &_0.53 & 0.29 & 13.84 &_1.59 &_–0.40 SDSS_1021–03 & T3 & & 1.78 & 15.88 & 0.62 &_0.47 & 0.15 & 15.26 &_& SDSS 1750+17 &_T3.5 &_&_& 16.14 &_0.12_& 0.20_& –0.08 & 16.02 & & 2

infty }\right) <\frac{1}{k}.$$Then, using the estimation $$\mathbb{P}\left[ \left\vert \left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k\left( n_{k}\right) }^{n_{k}}}-\left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k}^{n_{k}}}\right\vert >\varepsilon \right] \leq \mathbb{P}\left[ \left\{ \sup\limits_{t\in \mathbb{R}_{+}}\left\vert M_{t}^{\left( n_{k}\right) }-M_{t}\right\vert \geq k\right\} \right],$$it follows that $$d\left( \left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k\left( n_{k}\right) }^{n_{k}}},\left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k}^{n_{k}}}\right) \underset{k\rightarrow \infty }{\longrightarrow }0,$$which yields a contradiction with $\varepsilon \leq d\left( \left[ M^{\left( n_{k}\right) }-M\right] _{\infty },0\right) $. Thus, $\left[ M^{\left( n\right) }-M\right] _{\infty }\overset{\mathbb{P}}{\rightarrow }0.$ The last part of the this lemma follows immediately from the first statement. $\Box $ Using the Doob’s stopping theorem we can conclude that for $M\in \mathcal{M}_{\infty }$ and an stopping time $\tau $, that $M^{\tau }\in \mathcal{M}_{\infty },$ and therefore it follows as a corollary the following result. \[E\[|(Mn-M)thau|\]-&gt;0=&gt;\[Mn-M\]thau-&gt;0\_in\_P\]For $\left\{ M^{\left( n\right) }\right\} _{n\in \mathbb{N}}\subset \mathcal{M}_{\infty }$, $M\in \mathcal{M}_{\infty }$ and $\tau $ any stopping time holds$$M_{\tau }^{\left( n\

infty } \right) < \frac{1}{k}.$$Then, using the estimation $ $ \mathbb{P}\left [ \left\vert \left [ M^{\left (n_{k}\right) } -M\right ] _ { \tau _ { k\left (n_{k}\right) } ^{n_{k}}}-\left [ M^{\left (n_{k}\right) } -M\right ] _ { \tau _ { k}^{n_{k}}}\right\vert > \varepsilon \right ] \leq \mathbb{P}\left [ \left\ { \sup\limits_{t\in \mathbb{R}_{+}}\left\vert M_{t}^{\left ( n_{k}\right) } -M_{t}\right\vert \geq k\right\ } \right],$$it follows that $ $ d\left (\left [ M^{\left (n_{k}\right) } -M\right ] _ { \tau _ { k\left ( n_{k}\right) } ^{n_{k}}},\left [ M^{\left (n_{k}\right) } -M\right ] _ { \tau _ { k}^{n_{k}}}\right) \underset{k\rightarrow \infty } { \longrightarrow } 0,$$which yields a contradiction with $ \varepsilon \leq d\left (\left [ M^{\left ( n_{k}\right) } -M\right ] _ { \infty }, 0\right) $. therefore, $ \left [ M^{\left ( n\right) } -M\right ] _ { \infty } \overset{\mathbb{P}}{\rightarrow } 0.$ The final part of the this lemma follows immediately from the inaugural affirmation. $ \Box $ Using the Doob ’s stopping theorem we can reason that for $ M\in \mathcal{M}_{\infty } $ and an stopping time $ \tau $, that $ M^{\tau } \in \mathcal{M}_{\infty }, $ and therefore it come as a corollary the following result. \[E\[|(Mn - M)thau|\]-&gt;0=&gt;\[Mn - M\]thau-&gt;0\_in\_P\]For $ \left\ { M^{\left ( n\right) } \right\ } _ { n\in \mathbb{N}}\subset \mathcal{M}_{\infty } $, $ M\in \mathcal{M}_{\infty } $ and $ \tau $ any stopping prison term holds$$M_{\tau } ^{\left (n\

infhy }\right) <\frac{1}{k}.$$Then, using uhe estimation $$\majhvb{P}\lefv[ \left\vsrt \left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k\leht( n_{j}\righu) }^{n_{k}}}-\left[ M^{\left( n_{k}\rieht) }-M\righn] _{\tau _{k}^{n_{k}}}\rught\tert >\varepsilon \cjght] \leq \mathgn{P}\lefc[ \lxft\{ \sup\limits_{t\ik \mathbb{R}_{+}}\lext\vert M_{t}^{\left( n_{n}\rkgkt) }-M_{t}\right\vert \geq k\right\} \right],$$it fojlows tnah $$d\left( \left[ M^{\jeft( g_{k}\rifht) }-M\right] _{\tau _{k\left( n_{k}\right) }^{n_{k}}},\left[ M^{\left( i_{k}\right) }-M\right] _{\yau _{k}^{n_{k}}}\right) \underset{k\rightwrroa \infty }{\longrightagrow }0,$$which ijelqw a contradiztion with $\varepsilon \meq d\left( \left[ M^{\left( n_{k}\right) }-M\rieht] _{\iufty },0\right) $. Tyus, $\neft[ M^{\left( n\cight) }-I\right] _{\infty }\overset{\kathbb{P}}{\tightarrow }0.$ Thc lasv paet of the this lemma hollows immediately srom the xixst statement. $\Box $ Usibg the Goob’v stuppivg uheprsm we fan conclude fhat for $M\ib \mathcal{M}_{\infty }$ anc wb stopping tije $\tau $, ehat $M^{\tau }\in \mathcal{M}_{\infty },$ and therefoge if follows as a corollart the following resulj. \[E\[|(Mn-M)thau|\]-&ge;0=&gt;\[Mn-M\]thau-&gt;0\_in\_P\]For $\left\{ M^{\left( n\right) }\right\} _{n\in \matvbb{N}}\snbret \mwgycwl{M}_{\infty }$, $M\in \mathcal{M}_{\infty }$ and $\tau $ any sto[limg time holds$$M_{\tam }^{\left( n\

infty }\right) <\frac{1}{k}.$$Then, using the estimation $$\mathbb{P}\left[ M^{\left( }-M\right] _{\tau n_{k}\right) }^{n_{k}}}-\left[ M^{\left( \right] \mathbb{P}\left[ \left\{ \sup\limits_{t\in M_{t}^{\left( n_{k}\right) }-M_{t}\right\vert k\right\} \right],$$it follows that $$d\left( \left[ n_{k}\right) }-M\right] _{\tau _{k\left( n_{k}\right) }^{n_{k}}},\left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k}^{n_{k}}}\right) \underset{k\rightarrow \infty }0,$$which yields a contradiction with $\varepsilon \leq d\left( \left[ M^{\left( n_{k}\right) }-M\right] _{\infty $. $\left[ n\right) _{\infty }\overset{\mathbb{P}}{\rightarrow }0.$ The last part of the this lemma follows immediately from the first statement. $ Using the Doob’s stopping theorem we can that for $M\in \mathcal{M}_{\infty and an stopping time $\tau that }\in \mathcal{M}_{\infty and it as a corollary following result. \[E\[|(Mn-M)thau|\]-&gt;0=&gt;\[Mn-M\]thau-&gt;0\_in\_P\]For $\left\{ M^{\left( n\right) }\right\} _{n\in \mathbb{N}}\subset \mathcal{M}_{\infty }$, $M\in \mathcal{M}_{\infty }$ and $\tau $ stopping time n\

infty }\right) <\frac{1}{k}.$$Then, using tHe estimatiOn $$\matHbb{p}\leFt[ \Left\Vert \Left[ M^{\left( n_{k}\rigHT) }-M\riGht] _{\tau _{k\left( n_{k}\right) }^{n_{k}}}-\leFt[ M^{\leFt( N_{K}\rigHT) }-M\Right] _{\Tau _{k}^{n_{k}}}\rIGhT\VErt >\VaRePsiLoN \RiGht] \leQ \maThbb{P}\leFt[ \left\{ \sup\lImiTs_{T\in \mathbb{R}_{+}}\leFT\vErt M_{t}^{\left( n_{K}\riGht) }-M_{t}\right\veRt \gEq k\rigHt\} \RigHT],$$it foLloWs thaT $$d\left( \LEft[ M^{\leFt( n_{k}\right) }-m\rIGht] _{\tau _{K\Left( n_{k}\rIGHt) }^{N_{k}}},\leFt[ M^{\left( n_{k}\right) }-M\riGHt] _{\TAu _{k}^{n_{k}}}\right) \undeRset{k\rIgHTaRROw \iNftY }{\longrightArRow }0,$$whICh yieldS A cONTRadICtion with $\varePsilon \leq d\lEFt( \lEft[ M^{\leFt( N_{k}\rIGht) }-M\riGht] _{\inFtY },0\RigHt) $. Thus, $\left[ M^{\Left( N\right) }-M\riGht] _{\infTY }\overseT{\Mathbb{P}}{\RightaRroW }0.$ ThE lasT PaRt Of tHe THis LEmMa fOLloWs immediAtElY from The fIRST StatEmeNt. $\BoX $ UsinG the Doob’s stopPinG theORem We can ConclUde tHaT for $M\In \mathCal{M}_{\iNfTy }$ and an stopping Time $\Tau $, that $M^{\tAu }\iN \mAthCaL{M}_{\infTY },$ and thEreForE it follOws as a cORolLaRY THe Following result. \[E\[|(Mn-m)tHAU|\]-&gT;0=&gt;\[Mn-M\]thAu-&gt;0\_in\_p\]foR $\lEFt\{ M^{\left( n\RiGht) }\RighT\} _{N\In \matHbb{N}}\SUbSet \mathcAl{M}_{\infTY }$, $M\In \Mathcal{m}_{\iNfty }$ anD $\tAu $ aNy sToppiNG timE holds$$m_{\tau }^{\left( N\

infty }\right) <\frac{1}{k }.$$Then,using th e e st imat ion$$\mathbb{P}\l e ft[\left\vert \left[ M^{\ left( n _ {k}\ r ig ht) } -M\righ t ]_ { \ta u_{ k\l ef t (n_{k} \ri ght) }^ {n_{k}}}-\ lef t[ M^{\left( n _ {k }\right) } -M\ right] _{\ta u _ {k}^{n _{ k}} } \righ t\v ert > \varep s ilon \ right] \l eq \mathb b {P}\lef t [ \ left \{ \sup\limits_{t \ in \mathbb{R}_{+} }\left \v e rt M _{t }^{ \left( n_{ k} \righ t ) }-M_{ t }\ r i g ht\ v ert \geq k\ri ght\} \righ t ],$ $it fo ll ows that $ $d\le ft ( \l eft[ M^{\le ft(n_{k}\rig ht) }- M \right] _{\tau_{k\le ft( n_ {k}\ r ig ht ) } ^{ n _{k } }} ,\l e ft[ M^{\lef t( n _{k}\ righ t ) } -M\r igh t] _ {\tau _{k}^{n_{k}} }\r ight ) \u nders et{k\ righ ta rrow\infty }{\l on grightarrow }0, $$wh ich yield s a c ont ra dicti o n with $\ var epsilon \leq d \ lef t( \ l ef t[ M^{\left( n_{k} \r i g ht ) }-M\ri ght] _ { \i nf t y },0\ri gh t)$. T h u s, $\ left [ M ^{\left( n\rig h t) } -M\righ t] _{\in ft y } \ov erset { \mat hbb{P} }{\right arrow }0.$ The lastp art of the th i sl e mm a fol low s immediate ly f r om t he f i rs t s t ateme nt. $ \B o x$ Using the Doob’sst opping theo rem we can co nclude tha t f or $M\in \ma t hc a l{M}_{\infty } $ and an stoppi n g time $ \tau$, that$M^{\tau} \ in \math cal {M} _{\ inf t y } ,$ and theref o r e it f ollowsasa corol lar y t hefol lo wing resu lt. \[E \[ |( Mn -M )th au|\] - &gt;0=&g t; \[M n- M\] thau- & gt;0\_ in\_P \]Fo r$\ l eft \{ M^{\ l ef t ( n\r ig ht ) }\ rig ht \} _{ n\in \ma thbb{N} }\subset\ma t hcal {M }_ {\infty }$, $M\in \ ma thcal{M}_{ \i nft y }$ a n d $\tau $ any stopping time hold s $$M_{\t au}^{\l eft( n\

infty }\right) <\frac{1}{k}.$$Then,_using the_estimation $$\mathbb{P}\left[ \left\vert \left[_M^{\left( n_{k}\right)_}-M\right]__{\tau _{k\left( n_{k}\right)_}^{n_{k}}}-\left[_M^{\left( n_{k}\right) }-M\right] _{\tau__{k}^{n_{k}}}\right\vert >\varepsilon \right]_\leq \mathbb{P}\left[ \left\{ \sup\limits_{t\in \mathbb{R}_{+}}\left\vert_M_{t}^{\left( n_{k}\right) }-M_{t}\right\vert \geq_k\right\}_\right],$$it follows that $$d\left( \left[ M^{\left( n_{k}\right) }-M\right] _{\tau _{k\left( n_{k}\right) }^{n_{k}}},\left[ M^{\left( n_{k}\right) }-M\right]__{\tau _{k}^{n_{k}}}\right)_\underset{k\rightarrow \infty_}{\longrightarrow_}0,$$which_yields a contradiction with $\varepsilon_\leq d\left( \left[ M^{\left( n_{k}\right) }-M\right]__{\infty },0\right)_$. Thus, $\left[ M^{\left( n\right) }-M\right] _{\infty }\overset{\mathbb{P}}{\rightarrow }0.$_The_last part of_the this lemma follows immediately from the first statement._$\Box $ Using the Doob’s stopping theorem_we can conclude_that_for_$M\in \mathcal{M}_{\infty }$ and_an stopping time $\tau $, that_$M^{\tau }\in \mathcal{M}_{\infty },$ and therefore_it follows as a corollary the following_result. \[E\[|(Mn-M)thau|\]-&gt;0=&gt;\[Mn-M\]thau-&gt;0\_in\_P\]For $\left\{ M^{\left( n\right) }\right\} _{n\in \mathbb{N}}\subset_\mathcal{M}_{\infty }$, $M\in \mathcal{M}_{\infty }$_and $\tau_$ any stopping time holds$$M_{\tau_}^{\left( n\

big( \beta \sum_{i\in V} h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) \big) } \\ = \\[0,2cm] \displaystyle C \sum_{\sigma, \omega} \! \big( \! \prod_{\{i,j\}: \omega_{ij}=1} \!\!\! p_{ij} \prod_{\{i,j\}: \omega_{ij}=0} \!\!\!\!(1-p_{ij}) \big) \Delta(\sigma, \omega) \!\! \times \! {\textstyle \exp \big( \beta \sum_{i\in V} h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) \big) } \\ = \\[0,2cm] \displaystyle C \sum_{\sigma,\omega} B_{\pmb{J}}(\omega) \Delta(\sigma, \omega) \times {\textstyle \exp \big( \beta \sum_{i\in V} h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) \big) } = C \mathscr{Z}_{\pmb{p},\pmb{h},G}^{\mathrm{ ES}}. \qedhere \end{array}$$ \[Lema-cont\] Let $G$ be a finite graph and fix an edge configuration $\omega\in\{0,1\}^E$. If $\pmb{h}(K_\alpha)\equiv \beta {\sum_{i\in K_\alpha}} h_i$, where $K_1,\ldots, K_{k({\omega,G})}$ denote the connected components of $(V,\eta(\omega))$ we have $$\begin{aligned} \sum_{\sigma\in\{-1,+1\}^V} \Delta(\sigma, \omega) \times \exp \big( \beta \sum_{i

big (\beta \sum_{i\in V } h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1 }) \big) } \\ = \\[0,2 cm ] \displaystyle C \sum_{\sigma, \omega } \! \big ( \! \prod_{\{i, j\ }: \omega_{ij}=1 } \!\!\! p_{ij } \prod_{\{i, j\ }: \omega_{ij}=0 } \!\!\!\!(1 - p_{ij }) \big) \Delta(\sigma, \omega) \!\! \times \! { \textstyle \exp \big (\beta \sum_{i\in V } h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1 }) \big) } \\ = \\[0,2 cm ] \displaystyle C \sum_{\sigma,\omega } B_{\pmb{J}}(\omega) \Delta(\sigma, \omega) \times { \textstyle \exp \big (\beta \sum_{i\in V } h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1 }) \big) } = C \mathscr{Z}_{\pmb{p},\pmb{h},G}^{\mathrm { ES } }. \qedhere \end{array}$$ \[Lema - cont\ ] lease $ G$ be a finite graph and specify an edge configuration $ \omega\in\{0,1\}^E$. If $ \pmb{h}(K_\alpha)\equiv \beta { \sum_{i\in K_\alpha } } h_i$, where $ K_1,\ldots, K_{k({\omega, G})}$ denote the associate components of $ (V,\eta(\omega))$ we have $ $ \begin{aligned } \sum_{\sigma\in\{-1,+1\}^V } \Delta(\sigma, \omega) \times \exp \big (\beta \sum_{i

big( \beta \sum_{i\in V} h_l (\delta_{\sntma_i,1}-\denta_{\sigja_i,-1}) \big) } \\ = \\[0,2cm] \displaysthle C \snm_{\sigma, \omega} \! \bjn( \! \proc_{\{i,j\}: \omega_{ib}=1} \!\!\! p_{ij} \prod_{\{i,j\}: \omega_{ij}=0} \!\!\!\!(1-p_{ij}) \big) \Delta(\sigma, \jmegs) \!\! \times \! {\uextstyle \ecp \big( \beta \sum_{l\in V} h_i (\dflta_{\sigma_i,1}-\dgmta_{\fugma_i,-1}) \big) } \\ = \\[0,2cm] \displayrtyle C \sum_{\sigma,\omxga} B_{\pmb{J}}(\omcba) \Delya(\sigma, \omega) \times {\textstyne \exp \big( \beda \sum_{i\in V} h_u (\delgq_{\siema_j,1}-\dxltz_{\sigma_l,-1}) \big) } = C \mathsvr{S}_{\ikb{p},\pmb{h},G}^{\mathdm{ ES}}. \qedhere \end{array}$$ \[Lema-cont\] Let $G$ te z finite graph and fix qn edge configuration $\omega\in\{0,1\}^E$. If $\pmb{h}(K_\alpha)\equiv \beta {\sum_{i\in K_\alpha}} h_i$, whese $K_1,\lvogs, J_{k({\jowgw,G})}$ denote the connected components of $(V,\eta(\omedz))$ ee have $$\begin{alinned} \sum_{\sigka\ln\{-1,+1\}^F} \Delta(\rigma, \omefa) \times \exk \big( \btta \sum_{i

big( \beta \sum_{i\in V} h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) \big) = \displaystyle C \omega} \! \big( \prod_{\{i,j\}: \!\!\!\!(1-p_{ij}) \big) \Delta(\sigma, \!\! \times \! \exp \big( \beta \sum_{i\in V} h_i \big) } \\ = \\[0,2cm] \displaystyle C \sum_{\sigma,\omega} B_{\pmb{J}}(\omega) \Delta(\sigma, \omega) \times {\textstyle \big( \beta \sum_{i\in V} h_i (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) \big) } = C \mathscr{Z}_{\pmb{p},\pmb{h},G}^{\mathrm{ ES}}. \qedhere \[Lema-cont\] $G$ a graph and fix an edge configuration $\omega\in\{0,1\}^E$. If $\pmb{h}(K_\alpha)\equiv \beta {\sum_{i\in K_\alpha}} h_i$, where $K_1,\ldots, K_{k({\omega,G})}$ the connected components of $(V,\eta(\omega))$ we have $$\begin{aligned} \Delta(\sigma, \omega) \times \exp \beta \sum_{i

big( \beta \sum_{i\in V} h_i (\delta_{\sigmA_i,1}-\delta_{\sigMa_i,-1}) \biG) } \\ = \\[0,2cm] \DisPlAystYle C \Sum_{\sigma, \omega} \! \bIG( \! \proD_{\{i,j\}: \omega_{ij}=1} \!\!\! p_{ij} \prod_{\{i,j\}: \omeGa_{ij}=0} \!\!\!\!(1-p_{Ij}) \BIg) \DeLTa(\Sigma, \Omega) \!\! \tiMEs \! {\TEXtsTyLe \Exp \BiG( \BeTa \sum_{I\in v} h_i (\deltA_{\sigma_i,1}-\delTa_{\sIgMa_i,-1}) \big) } \\ = \\[0,2cm] \dispLAyStyle C \sum_{\sIgmA,\omega} B_{\pmb{J}}(\oMegA) \Delta(\SiGma, \OMega) \tImeS {\textStyle \eXP \big( \beTa \sum_{i\in V} H_i (\DElta_{\siGMa_i,1}-\deltA_{\SIgMa_i,-1}) \bIg) } = C \mathscr{Z}_{\pmb{p},\pmB{H},G}^{\MAthrm{ ES}}. \qedhere \End{arrAy}$$ \[lEmA-COnt\] let $g$ be a finite GrAph anD Fix an edGE cONFIguRAtion $\omega\in\{0,1\}^E$. if $\pmb{h}(K_\alphA)\EquIv \beta {\SuM_{i\iN k_\alpha}} H_i$, wheRe $k_1,\LdoTs, K_{k({\omega,G})}$ dEnotE the conneCted coMPonents OF $(V,\eta(\omEga))$ we hAve $$\BegIn{alIGnEd} \Sum_{\SiGMa\iN\{-1,+1\}^v} \DEltA(\SigMa, \omega) \tImEs \Exp \biG( \betA \SUM_{I

big( \beta \sum_{ i\in V} h_ i (\ delt a_{\ sigma_i,1}-\de l ta_{ \sigma_i,-1}) \big) } \\ = \\ [0,2c m] \displays tyl e C \sum_{\ sig ma, \omega} \ ! \ big ( \ ! \prod_ {\ { i,j\}: \omega_ { i j} =1} \!\!\! p_ { ij } \prod _{\{i, j\ } :\ o meg a_{ ij}=0} \! \ !\!\!(1 - p_ { i j }) \big) \De l ta( \sigma ,\om e ga) \!\! \times \! {\te xtstyl e \ ex p \ b ig ( \ b eta \s um_{i \inV } h _i (\de lta_{\sigma_i ,1} -\de l ta_ {\sig ma_i, -1}) \big ) } \\ = \\ [0 ,2cm] \ dis playsty le C \sum_{\sigma,\o me g a } B_{\pm b {J }} ( \omega) \ D e lta(\ sigm a ,\omega) \ ti mes {\t ex tst yle \ exp \b ig( \ b eta \ s um_{i\in V} h _ i (\d elta_{\sigm a_i, 1 }-\d elta _ {\ sig m a_i,- 1}) \big) } = C \mathscr {Z}_{\pmb{ p } , \pmb{h}, G}^{ \ ma t hrm{ ES}}. \qedhere \end{a rray} $$ \[Le ma-cont\] L et $G$ b e a fi nit e g r a ph and fix an e d g e co nf igurati on$\omega \in \{0 ,1\ }^E $. If $\pmb {h}(K_\a lp ha )\ eq uiv \bet a {\ su m_{ i\ inK_\al p ha}} h _i$,wher e$K _ 1,\ ldots,K _{ k ( {\om eg a, G})} $ d en ote t he c o nne cted co mponentsof$ (V,\ et a( \omega) )$ we have $$ \b egin{align ed } \ sum_{\si gma\in\{-1,+1\}^V} \Del ta( \sigm a, \ omega) \ti mes \exp \ b ig( \ b e ta \sum_ { i

big( \beta_ _ _ _\sum_{i\in_V} h_i_ _ _ _ (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) _ __ \big) } __ _\\ __ _ = _ _ \\[0,2cm] __ \displaystyle _ C _ \sum_{\sigma,_\omega} ___ \! _ _ \big( _ \! _ \prod_{\{i,j\}: \omega_{ij}=1} _ _ _\!\!\! p_{ij} _ _ \prod_{\{i,j\}:_\omega_{ij}=0} _ \!\!\!\!(1-p_{ij}) _ _ \big) __ ___\Delta(\sigma, \omega) _ __ __ __ \!\! _ \times __ _ \! _ {\textstyle _ _ _ \exp \big(_\beta _ \sum_{i\in_V} h_i_ _ __ _ (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) _ _\big) _ } __ \\ _ _ = ___ \\[0,2cm] _ \displaystyle _ _ C __ \sum_{\sigma,\omega} __ B_{\pmb{J}}(\omega) ___ _\Delta(\sigma, \omega) _ \times _ {\textstyle _ ___\exp \big( \beta_ _ _ \sum_{i\in V} h_i _ (\delta_{\sigma_i,1}-\delta_{\sigma_i,-1}) _ \big) _ } _ = _ C_\mathscr{Z}_{\pmb{p},\pmb{h},G}^{\mathrm{_ES}}. __\qedhere __ \end{array}$$ \[Lema-cont\]_Let $G$ be_a finite graph_and fix_an edge configuration_$\omega\in\{0,1\}^E$. If_$\pmb{h}(K_\alpha)\equiv \beta __ {\sum_{i\in K_\alpha}} h_i$, where $K_1,\ldots, K_{k({\omega,G})}$ denote the connected components of $(V,\eta(\omega))$ we have $$\begin{aligned} _ _\sum_{\sigma\in\{-1,+1\}^V} __ \Delta(\sigma,_\omega) _ \times _ \exp __ _ \big( \beta _ _ _ _\sum_{i

left(i\partial_0 + {q_1} a^1_0 + {q_1} a^2_0 \right)\Phi^{({q_1},{q_2})} \right.\\ -\left. \left|\left(i\nabla_j + {q_1} a^1_j + {q_1} a^2_i \right)\Phi^{({q_1},{q_2})} \right|^2\right] - V(\Phi)\\ + \frac{2}{2\pi}\epsilon^{i0j} a^1_i \partial_0 a^2_j\end{gathered}$$ The charges take values $({q_1},{q_2}) = (1,0), (0,1), (1,1)$. This is the effective theory for the Toric Code or deconfined phase of $\mathbb{Z}_2$ gauge theory, as expected from the analysis of Bais and Slingerland [@Bais09]. However, this appears to be simply a $U(1)\times U(1)$ gauge theory. What happened to the $\rtimes\mathbb{Z}_2$ portion of the theory? The $\mathbb{Z}_2$ action of the theory would interchange $a^1$ and $a^2$. In principle, gauging this action would lead back to and $\mathrm{Ising}\times\mathrm{Ising}$ theory.[@Barkeshli14] In this case, however, the generator of the $\mathbb{Z}_2$ part of the theory must create a twist in the $a$ field without creating one in the $b$ field (or vice versa). It is therefore a confined excitation, coming from the field $\Phi^{(\frac{1}{2},0}$ (or $\Phi^{0,\frac{1}{2}}$) of the original theory. The theory of Eq. (\[eqn:example-abelian-action\]) may thus be taken at face value, and the resulting phase is topologically equivalent to the Toric Code. In the next section, we generalize the preceding discussion. Condensation Transitions from $SU(2)_{p+q}\otimes SU(2)_{p-q}\vert_R $ {#Sec:e-condense} ====================================================================== We consider the condensation of a field $\Phi^{(\frac{

left(i\partial_0 + { q_1 } a^1_0 + { q_1 } a^2_0 \right)\Phi^{({q_1},{q_2 }) } \right.\\ -\left. \left|\left(i\nabla_j + { q_1 } a^1_j + { q_1 } a^2_i \right)\Phi^{({q_1},{q_2 }) } \right|^2\right ] - V(\Phi)\\ + \frac{2}{2\pi}\epsilon^{i0j } a^1_i \partial_0 a^2_j\end{gathered}$$ The charges take values $ ({ q_1},{q_2 }) = (1,0), (0,1), (1,1)$. This is the effective theory for the Toric Code or deconfined phase of $ \mathbb{Z}_2 $ bore hypothesis, as expect from the analysis of Bais and Slingerland [ @Bais09 ]. However, this appears to be just a $ U(1)\times U(1)$ gauge theory. What happen to the $ \rtimes\mathbb{Z}_2 $ share of the theory? The $ \mathbb{Z}_2 $ military action of the theory would interchange $ a^1 $ and $ a^2$. In rationale, gauging this action would lead back to and $ \mathrm{Ising}\times\mathrm{Ising}$ theory.[@Barkeshli14 ] In this case, however, the generator of the $ \mathbb{Z}_2 $ share of the theory must make a twist in the $ a$ field without make one in the $ b$ field (or vice versa). It is therefore a enclose excitation, coming from the field $ \Phi^{(\frac{1}{2},0}$ (or $ \Phi^{0,\frac{1}{2}}$) of the original theory. The hypothesis of Eq.   (\[eqn: example - abelian - action\ ]) may therefore be taken at face value, and the resulting phase is topologically equivalent to the Toric Code. In the next department, we generalize the preceding discussion. Condensation Transitions from $ SU(2)_{p+q}\otimes SU(2)_{p - q}\vert_R $ { # Sec: e - condense } = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = We consider the condensation of a playing field $ \Phi^{(\frac {

lefh(i\partial_0 + {q_1} a^1_0 + {q_1} a^2_0 \right)\Khi^{({q_1},{q_2})} \right.\\ -\left. \leyr|\left(i\iabla_j + {q_1} a^1_j + {q_1} a^2_i \right)\Phi^{({q_1},{q_2})} \right|^2\right] - V(\'hi)\\ + \drac{2}{2\pu}\epsilon^{i0j} a^1_i \partial_0 a^2_j\end{gatjered}$$ Thw cherges take values $({q_1},{q_2}) = (1,0), (0,1), (1,1)$. This jd thz xffective theori for the Tosic Code or dewovfnned phase of $\mathbb{Z}_2$ gauge theory, af expecyef from the anajysix of Gais and Slingerland [@Bais09]. However, fhis apkears to be simply a $U(1)\times U(1)$ gauge theory. Wjat jappened to the $\rtlmes\mathbb{Z}_2$ poreuon of the tfeory? The $\mathbb{Z}_2$ actikn of the theory would interchavge $a^1$ and $a^2$. In kxunclkle, gauging vhis abtion would lcsd bacn to anc $\mathrm{Ising}\tlmes\methrn{Ising}$ theory.[@Barkeshlm14] In this case, howevgr, the genarctor of the $\mathbb{Z}_2$ pqrr of jhe tveorh murt drxats a twlst in the $a$ rield withoyt creating one in uhe $v$ field (or vide verfa). It is therefore a confined excitation, bomihg from the field $\Phi^{(\frqc{1}{2},0}$ (or $\Phi^{0,\frac{1}{2}}$) of the lriginal eheory. The theory of Eq. (\[eqn:example-abelian-action\]) maf thua be twywn at face value, and the resulting phase is topjmobibally equivalent bo the Toric Code. On tng next section, we geusrzlize the precedinh discufsion. Xondensatyon Yransitions from $SU(2)_{p+q}\otimes SU(2)_{p-q}\vert_R $ {#Fwc:e-condense} ====================================================================== We consnder the conbensatoon og a field $\Phi^{(\frac{

left(i\partial_0 + {q_1} a^1_0 + {q_1} a^2_0 -\left. + {q_1} + {q_1} a^2_i \frac{2}{2\pi}\epsilon^{i0j} \partial_0 a^2_j\end{gathered}$$ The take values $({q_1},{q_2}) (1,0), (0,1), (1,1)$. This is the theory for the Toric Code or deconfined phase of $\mathbb{Z}_2$ gauge theory, as from the analysis of Bais and Slingerland [@Bais09]. However, this appears to be a U(1)$ theory. happened to the $\rtimes\mathbb{Z}_2$ portion of the theory? The $\mathbb{Z}_2$ action of the theory would interchange and $a^2$. In principle, gauging this action would back to and $\mathrm{Ising}\times\mathrm{Ising}$ In this case, however, the of $\mathbb{Z}_2$ part the must a twist in $a$ field without creating one in the $b$ field (or vice versa). It is therefore a confined coming from $\Phi^{(\frac{1}{2},0}$ (or of original The theory of may thus be taken at face resulting phase is topologically equivalent to the Toric In the section, we generalize the preceding discussion. Transitions from $SU(2)_{p+q}\otimes SU(2)_{p-q}\vert_R $ {#Sec:e-condense} ====================================================================== We the condensation of a field $\Phi^{(\frac{

left(i\partial_0 + {q_1} a^1_0 + {q_1} a^2_0 \right)\Phi^{({q_1},{Q_2})} \right.\\ -\left. \Left|\lEft(I\naBlA_j + {q_1} a^1_J + {q_1} a^2_i \Right)\Phi^{({q_1},{q_2})} \righT|^2\RighT] - V(\Phi)\\ + \frac{2}{2\pi}\epsilon^{i0j} a^1_i \PartiAl_0 A^2_J\end{GAtHered}$$ the charGEs TAKe vAlUeS $({q_1},{q_2}) = (1,0), (0,1), (1,1)$. thIS iS the eFfeCtive thEory for the torIc code or deconfINeD phase of $\maThbB{Z}_2$ gauge theorY, as ExpectEd FroM The anAlySis of bais anD slingeRland [@Bais09]. hoWEver, thIS appearS TO bE simPly a $U(1)\times U(1)$ gauge tHEoRY. What happened tO the $\rtImES\mATHbb{z}_2$ poRtion of the ThEory? THE $\mathbb{z}_2$ AcTION of THe theory would Interchange $A^1$ And $A^2$. In priNcIplE, GauginG this AcTIon Would lead baCk to And $\mathrm{ising}\tIMes\mathRM{Ising}$ tHeory.[@BArkEshLi14] In THiS cAse, HoWEveR, ThE geNEraTor of the $\MaThBb{Z}_2$ paRt of THE THeorY muSt crEate a Twist in the $a$ fiEld WithOUt cReatiNg one In thE $b$ Field (Or vice Versa). it Is therefore a conFineD excitatiOn, cOmIng FrOm the FIeld $\PhI^{(\frAc{1}{2},0}$ (oR $\Phi^{0,\fraC{1}{2}}$) of the oRIgiNaL THEoRy. The theory of Eq. (\[eqn:ExAMPlE-abelian-Action\]) MAy ThUS be taken At FacE valUE, And thE resULtIng phase Is topoLOgIcAlly equIvAlent tO tHe TOriC Code. iN the Next seCtion, we gEneraLIze the precedinG Discussion. ConDEnSATiON TraNsiTions from $SU(2)_{P+q}\otIMes Su(2)_{p-q}\vERt_r $ {#SeC:E-condEnse} ====================================================================== WE cONsIDer the condensation oF a Field $\PHi^{(\fraC{

left(i\partial_0 + {q_1} a ^1_0 + {q _1} a ^2_ 0 \ ri ght) \Phi ^{({q_1},{q_2} ) } \r ight.\\ -\left. \left| \left (i \ nabl a _j + {q _1} a^1 _ j+ {q _1 }a^2 _i \r ight) \Ph i^{({q_ 1},{q_2})} \r ig ht|^2\right] -V(\Phi)\\+ \ frac{2}{2\pi }\e psilon ^{ i0j } a^1_ i \ parti al_0 a ^2_j\e nd{gather ed } $$ The charges t ak e va lues $({q_1},{q_2 } )= (1,0), (0,1), (1,1) $. Th i s is th e effectiv etheor y for th e T o r i c C o de or deconfi ned phase o f $\ mathbb {Z }_2 $ gauge theo ry , as expected f romthe analy sis of Bais an d Slinge rland[@B ais 09]. Ho we ver ,t his ap pea r s t o be sim pl ya $U( 1)\t i m e s U(1 )$gaug e the ory. What hap pen ed t o th e $\r times \mat hb b{Z}_ 2$ por tionof the theory? Th e $\ mathbb{Z} _2$ a cti on of t h e theo rywou ld inte rchange $a^ 1$ a n d$a^2$. In principl e, g au ging thi s acti o nwo u ld leadba ckto a n d $\ma thrm { Is ing}\tim es\mat h rm {I sing}$th eory.[ @B ark esh li14] In t his ca se, howe ver,t he generator o f the $\mathbb { Z} _ 2 $p artofthe theorymust crea te a tw ist in th e $a$ f i el d without creating o ne in th e $b$ field (or vi ce versa). I t is ther efor e a confined excit ation , coming f r om the f ield$\Phi^{( \frac{1}{ 2 } ,0}$ (or $\ Phi ^{0 ,\f r a c{ 1}{2}}$) of t h e ori gi nal the ory . The t heo ryofEq.  ( \[eqn:exa mple-abe li an -a ct ion \]) m a y thus b etak en at face value, andthere su l tin g phase is t opol og ic ally eq ui valen t to the ToricCode. Inthe next s ec tion, w e generalizeth e precedin gdis cussio n . Conden sation Transitions from $SU(2)_ {p+ q}\ot imes SU(2)_{p -q} \vert_ R $ {#Sec: e-cond ense} = === = = ===== = = == === == ========== = = === ===== == ==== ======= =================Weconsider thecon dens a t io n o f a fie ld $\P h i ^{(\frac{

left(i\partial_0 +_{q_1} a^1_0_ + {q_1} a^2_0 \right)\Phi^{({q_1},{q_2})} \right.\\ -\left. \left|\left(i\nabla_j_+ {q_1}_a^1_j_ + {q_1}_a^2_i_\right)\Phi^{({q_1},{q_2})} \right|^2\right] -_V(\Phi)\\ + \frac{2}{2\pi}\epsilon^{i0j} a^1_i_\partial_0 a^2_j\end{gathered}$$ The_charges take values_$({q_1},{q_2})_= (1,0), (0,1), (1,1)$. This is the effective theory for the Toric Code or_deconfined_phase of_$\mathbb{Z}_2$_gauge_theory, as expected from the_analysis of Bais and Slingerland_[@Bais09]. However,_this appears to be simply a $U(1)\times U(1)$_gauge_theory. What happened_to the $\rtimes\mathbb{Z}_2$ portion of the theory? The $\mathbb{Z}_2$_action of the theory would interchange_$a^1$ and $a^2$._In_principle,_gauging this action would_lead back to and $\mathrm{Ising}\times\mathrm{Ising}$ theory.[@Barkeshli14]_In this case, however, the generator_of the $\mathbb{Z}_2$ part of the theory_must create a twist in the_$a$ field without creating one_in the_$b$ field (or vice versa)._It is therefore_a confined_excitation, coming from_the field $\Phi^{(\frac{1}{2},0}$ (or $\Phi^{0,\frac{1}{2}}$) of_the original theory._The theory of Eq. (\[eqn:example-abelian-action\]) may thus_be_taken at face_value,_and_the resulting_phase is topologically_equivalent_to the_Toric_Code. In the next section, we_generalize_the preceding discussion. Condensation Transitions from $SU(2)_{p+q}\otimes SU(2)_{p-q}\vert_R_$ {#Sec:e-condense} ====================================================================== We consider the_condensation_of a field $\Phi^{(\frac{

Lett. [**B718**]{},143(2012). F.Karsch, E.Laermann and A.Peikert, Phys. Lett. [**B478**]{}, 447(2000). H.van Hecke, H.Sorge and N.Xu, Phys. Rev. Lett. [**81**]{}, 5764(1998). N.S.Topilskaya [*et al.*]{}, \[NA50 Collaboration\], Nucl. Phys. [**A715**]{}, 675(2003). A.Adare [*et al.*]{}, \[PHENIX Collaboration\], Phys. Rev. Lett, [**98**]{}, 232301(2007). E.Scomparin [*et al.*]{}, \[ALICE Collaboration\], Quark Matter 2012 talk. J.Huefner, Y.Kurihara and H.J.Pirner, Phys. Lett. [**B215**]{}, 218(1988). X. Wang, Phys. Rev. Lett. [**81**]{}, 2655(1998). B.I.Abelev [*et al.*]{} \[STAR Collaboration\], Phys. Rev. Lett. [**98**]{}, 192301(2007). A.Adare [*et al.*]{}, \[PHENIX Collaboration\], Phys. Rev. Lett. [**98**]{}, 172301(2007). B.Abelev [*et al.*]{}, \[ALICE Collaboration\], JHEP [**09**]{}, 112(2012). J. Schieck, arXiv:1205.4153. --- abstract: 'We consider the full-duplex (FD) two-way amplify-and-forward relay system with imperfect cancelation of loopback self-interference (SI) and investigate joint design of relay and receive beamforming for minimizing the mean square error under a relay transmit power constraint. Due to loopback channel estimation error and limitation of analog-to-digital converter, the loopback SI cannot be completely canceled. Multiple antennas at the relay can help loopback SI suppression but beamforming is required to balance between the residual SI suppression and the desired signal transmission. Moreover, the relay beamforming matrix should be updated every time slot because the residual SI in the previous time slot is amplified by the current beamforming matrix and added to the received signals from the two sources in the current time slot. We derive the optimally balanced relay beamforming and receive beamforming matrices in closed form based on minimum mean square error, taking into account the propagation of the residual loopback SI from the first to the current time slot. We also propose beamforming design using only the channels of

Lett. [ * * B718**]{},143(2012). F.Karsch, E.Laermann and A.Peikert, Phys. Lett. [ * * B478 * * ] { }, 447(2000). H.van Hecke, H.Sorge and N.Xu, Phys. Rev. Lett. [ * * 81 * * ] { }, 5764(1998). N.S.Topilskaya [ * et al. * ] { }, \[NA50 Collaboration\ ], Nucl. Phys. [ * * A715 * * ] { }, 675(2003). A.Adare [ * et al. * ] { }, \[PHENIX Collaboration\ ], Phys. Rev. Lett, [ * * 98 * * ] { }, 232301(2007). E.Scomparin [ * et al. * ] { }, \[ALICE Collaboration\ ], Quark Matter 2012 talk. J.Huefner, Y.Kurihara and H.J.Pirner, Phys. Lett. [ * * B215 * * ] { }, 218(1988). X. Wang, Phys. Rev. Lett. [ * * 81 * * ] { }, 2655(1998). B.I.Abelev [ * et al. * ] { } \[STAR Collaboration\ ], Phys. Rev. Lett. [ * * 98 * * ] { }, 192301(2007). A.Adare [ * et al. * ] { }, \[PHENIX Collaboration\ ], Phys. Rev. Lett. [ * * 98 * * ] { }, 172301(2007). B.Abelev [ * et al. * ] { }, \[ALICE Collaboration\ ], JHEP [ * * 09 * * ] { }, 112(2012). J. Schieck, arXiv:1205.4153. --- abstract:' We consider the full - duplex (FD) two - direction amplify - and - advancing relay system with imperfect cancelation of loopback self - noise (SI) and investigate joint purpose of relay and receive beamforming for minimizing the beggarly hearty error under a relay convey power constraint. Due to loopback distribution channel estimation error and restriction of analog - to - digital converter, the loopback SI cannot be completely canceled. Multiple antennas at the relay can help loopback SI inhibition but beamforming is required to balance between the residual SI suppression and the hope signal transmission. Moreover, the relay beamforming matrix should be update every time time slot because the residual SI in the former time slot is amplified by the current beamforming matrix and added to the receive signals from the two sources in the current time slot. We derive the optimally balanced relay beamforming and meet beamforming matrix in closed form based on minimum mean square error, take into history the propagation of the residual loopback SI from the first to the current prison term time slot. We besides propose beamforming design using only the channels of

Leht. [**B718**]{},143(2012). F.Karsch, E.Laermann akd A.Peikert, Phys. Lett. [**B478**]{}, 447(2000). H.van Hecke, H.Rorge and N.Xu, Phys. Rev. Lett. [**81**]{}, 5764(1998). B.S.Topulskaya [*et al.*]{}, \[NA50 Collacoration\], Jucl. Phyw. [**A715**]{}, 675(2003). Q.Adare [*et em.*]{}, \[PHENIW Colmwborctmon\], Phys. Rev. Lejt, [**98**]{}, 232301(2007). E.Scompasin [*et al.*]{}, \[ALICA Zoplaboration\], Quark Matter 2012 talk. J.Huesner, Y.Kirlhara and H.J.Pitner, Khyf. Lefn. [**N215**]{}, 218(1988). X. Wang, Phys. Rev. Lett. [**81**]{}, 2655(1998). B.I.Abelsv [*et an.*]{} \[STAR Collabpration\], Phys. Rev. Lett. [**98**]{}, 192301(2007). A.Afare [*et al.*]{}, \[PHENIX Collwboration\], Pyys. Wwv. Lett. [**98**]{}, 172301(2007). B.Acelev [*et al.*]{}, \[ALICE Collzboration\], JHEP [**09**]{}, 112(2012). J. Schieck, arXix:1205.4153. --- abscract: 'We cobsudeg the full-du'lex (FQ) two-way amplify-and-fmrward telay system wlth ikpeefect cancelation of noopback self-interserence (SH) cnd investigate joint dwsign of selah ana rtcemve beamflrmmng for minjmizing the mean square error inqvt a relay trznsmit pjwer constraint. Due to loopback channel esfimation error and limiration of analog-to-digltal convqrter, the loopback SI cannot be completely cancelad. Mumgipoe xbtfnnas at the relay can help loopback SI supprqasook but beamforminn is required to bslwnvg between the tesidual SJ suppression and hhe desyred wignal trwnsmossion. Moreover, the relay bwamforming mctrux should be updatzd every timz slot becaise the residual SI in che prsvious time slot is zoplified by the zurgent beamforming matrix and adqed to thx reczived sienalx from the two slurces in the current tile sllt. We derive the optimally balanced relay bxemforming and racepve beamfjrminn matrices in cjosed form basgd on minnmum mdan square error, vaking into wccount the pslpagation of the resyduao loipback RK from the firxt to the current tume slot. We also pvoposg geamforming desnyn using only the chxnnqld pf

Lett. [**B718**]{},143(2012). F.Karsch, E.Laermann and A.Peikert, Phys. 447(2000). Hecke, H.Sorge N.Xu, Phys. Rev. al.*]{}, Collaboration\], Nucl. Phys. 675(2003). A.Adare [*et \[PHENIX Collaboration\], Phys. Rev. Lett, [**98**]{}, E.Scomparin [*et al.*]{}, \[ALICE Collaboration\], Quark Matter 2012 talk. J.Huefner, Y.Kurihara and H.J.Pirner, Lett. [**B215**]{}, 218(1988). X. Wang, Phys. Rev. Lett. [**81**]{}, 2655(1998). B.I.Abelev [*et al.*]{} Collaboration\], Rev. [**98**]{}, A.Adare [*et al.*]{}, \[PHENIX Collaboration\], Phys. Rev. Lett. [**98**]{}, 172301(2007). B.Abelev [*et al.*]{}, \[ALICE Collaboration\], JHEP 112(2012). J. Schieck, arXiv:1205.4153. --- abstract: 'We consider full-duplex (FD) two-way amplify-and-forward system with imperfect cancelation of self-interference and investigate design relay receive beamforming for the mean square error under a relay transmit power constraint. Due to loopback channel estimation error and of analog-to-digital loopback SI be canceled. antennas at the help loopback SI suppression but beamforming balance between the residual SI suppression and the signal transmission. the relay beamforming matrix should be every time slot because the residual SI in previous time slot is amplified by the current beamforming matrix and added to the received the two sources in current time slot. derive optimally relay and receive matrices in closed form based on minimum mean square error, taking account the propagation of the residual loopback SI from the the time slot. We propose beamforming design using the of

Lett. [**B718**]{},143(2012). F.Karsch, E.Laermann and A.peikert, PhyS. Lett. [**b478**]{}, 447(2000). H.vAn HEcKe, H.SOrge And N.Xu, Phys. Rev. LETt. [**81**]{}, 5764(1998). N.S.topilskaya [*et al.*]{}, \[NA50 CollabOratiOn\], nUcl. PHYs. [**a715**]{}, 675(2003). A.AdaRe [*et al.*]{}, \[PheNix colLaBoRatIoN\], phYs. Rev. letT, [**98**]{}, 232301(2007). E.ScompArin [*et al.*]{}, \[ALiCE coLlaboration\], QUArK Matter 2012 talK. J.HUefner, Y.KurihAra And H.J.PIrNer, pHys. LeTt. [**B215**]{}, 218(1988). x. Wang, phys. ReV. lett. [**81**]{}, 2655(1998). B.I.abelev [*et aL.*]{} \[StaR CollABoratioN\], pHyS. Rev. lett. [**98**]{}, 192301(2007). A.Adare [*et al.*]{}, \[PHEniX cOllaboration\], PhYs. Rev. LEtT. [**98**]{}, 172301(2007). b.ABELev [*Et aL.*]{}, \[ALICE CollAbOratiON\], JHEP [**09**]{}, 112(2012). J. SCHiECK, ArXIV:1205.4153. --- abstract: 'We coNsider the fuLL-duPlex (FD) TwO-waY AmplifY-and-fOrWArd Relay system With Imperfect CancelATion of lOOpback sElf-intErfEreNce (Si) AnD iNveStIGatE JoInt DEsiGn of relaY aNd ReceiVe beAMFORminG foR minImiziNg the mean squaRe eRror UNdeR a relAy traNsmiT pOwer cOnstraInt. DuE tO loopback channeL estImation erRor AnD liMiTatioN Of analOg-tO-diGital coNverter, THe lOoPBACk sI cannot be completeLy CANcEled. MultIple anTEnNaS At the relAy Can Help LOOpbacK SI sUPpRession bUt beamFOrMiNg is reqUiRed to bAlAncE beTween THe reSidual sI suppreSsion ANd the desired siGNal transmissiON. MOREoVEr, thE reLay beamformIng mATrix ShouLD bE upDAted eVery tImE SlOT because the residual sI In the pRevioUs time slot is aMplified by THE Current bEamfORmINg matrix and addEd to tHe received SIgnals frOm the Two sourcEs in the cuRREnt time sLot. we dEriVe tHE OpTimally balancED RelaY bEamformIng And receIve BeaMfoRmiNg Matrices iN closed fOrM bAsEd On mInimuM Mean squaRe ErrOr, TakIng inTO accouNt the PropAgAtIOn oF the resIDuAL LoopBaCk sI frOm tHe First To thE CurRent timE slot. We alSo pROposE bEaMforminG design using oNlY the channeLs Of

Lett. [**B718**]{},143(20 12). F.Kar sch,E.L aer ma nn a nd A .Peikert, Phys . Let t. [**B478**]{}, 447(2 000). H . vanH ec ke, H .Sorgea nd N .Xu ,Ph ys. R e v. Lett . [ **81**] {}, 5764(1 998 ). N.S.Topilsk a ya [*et al.* ]{} , \[NA50 Col lab oratio n\ ],N ucl.Phy s. [* *A715* * ]{}, 6 75(2003). A . Adare[ *et al. * ] {} , \[ PHENIX Collaborat i on \ ], Phys. Rev.Lett,[* * 98 * * ]{} , 2 32301(2007 ). E.Sc o mparin[ *e t a l.* ] {}, \[ALICE C ollaboratio n \], Quark M att e r 2012 talk .J .Hu efner, Y.Ku riha ra and H. J.Pirn e r, Phys . Lett.[**B21 5** ]{} , 21 8 (1 98 8). X . Wa n g, Ph y s.Rev. Let t. [ **81* *]{} , 2 6 55(1 998 ). B .I.Ab elev [*et al. *]{ } \[ S TAR Coll abora tion \] , Phy s. Rev . Let t. [**98**]{}, 19 2301 (2007). A .Ad ar e [ *e t al. * ]{}, \ [PH ENI X Colla boratio n \], P h y s .Rev. Lett. [**98** ]{ } , 1 72301(20 07). B . Ab el e v [*et a l. *]{ }, \ [ A LICEColl a bo ration\] , JHEP [* *0 9**]{}, 1 12(201 2) . J . S chiec k , ar Xiv:12 05.4153. --- abstract: 'Wec onsider the f u ll - d up l ex ( FD) two-way am plif y -and -for w ar d r e lay s ystem w i th imperfect cancelati on of lo opbac k self-interf erence (SI ) a nd inves tiga t ej oint design of rela y and rece i ve beamf ormin g for mi nimizingt h e mean s qua reerr oru n de r a relay tra n s mitpo wer con str aint. D uetoloo pba ck channelestimati on e rr or an d lim i tation o fana lo g-t o-dig i tal co nvert er,th el oop back SI ca n n ot b eco mple tel ycance led. Mul tiple a ntennas a t t h e re la ycan hel p loopback SI s uppression b utbeamfo r m ing is r equired to balance betw e en theres idual SIsuppressi onand th e d e siredsignal tran sm iss i o n. Mo r e ov er, t he relay b e a mfo rming m atri x shoul d be updated every tim e slot becaus e t he r e s id ual SI inth e pr e v ious time slotis amplifi ed by the curre n t b ea mformin g matri x and added t o the rec eived sig na ls f r o m t he two sou rces inthe curre n t tim e s lot.Wederive t heoptim ally b a lan ced r elay b ea mformi ng an dreceivebeamforming matrices in close d for m b ased on m ini m ummean squa re e rror, taki ngint o acc oun t theprop a ga tio n of t he r e sidual lo o pb ack S Ifrom the fi r s t to thecur r ent ti me s lot. We also prop o se beamforming des i g n u sin g onl ythe channels o f

Lett._[**B718**]{},143(2012). F.Karsch,_E.Laermann and A.Peikert, Phys._Lett. [**B478**]{},_447(2000)._H.van Hecke,_H.Sorge_and N.Xu, Phys._Rev. Lett. [**81**]{},_5764(1998). N.S.Topilskaya [*et al.*]{},_\[NA50 Collaboration\], Nucl._Phys._[**A715**]{}, 675(2003). A.Adare [*et al.*]{}, \[PHENIX Collaboration\], Phys. Rev. Lett, [**98**]{}, 232301(2007). E.Scomparin [*et_al.*]{},_\[ALICE Collaboration\],_Quark_Matter_2012 talk. J.Huefner, Y.Kurihara and_H.J.Pirner, Phys. Lett. [**B215**]{}, 218(1988)._X. Wang,_Phys. Rev. Lett. [**81**]{}, 2655(1998). B.I.Abelev [*et al.*]{}_\[STAR_Collaboration\], Phys. Rev._Lett. [**98**]{}, 192301(2007). A.Adare [*et al.*]{}, \[PHENIX Collaboration\], Phys._Rev. Lett. [**98**]{}, 172301(2007). B.Abelev [*et_al.*]{}, \[ALICE Collaboration\],_JHEP_[**09**]{},_112(2012). J. Schieck, arXiv:1205.4153. _--- abstract: 'We consider the full-duplex (FD)_two-way amplify-and-forward relay system with imperfect_cancelation of loopback self-interference (SI) and investigate_joint design of relay and receive_beamforming for minimizing the mean_square error_under a relay transmit power_constraint. Due to_loopback channel_estimation error and_limitation of analog-to-digital converter, the loopback_SI cannot be_completely canceled. Multiple antennas at the_relay_can help loopback_SI_suppression_but beamforming_is required to_balance_between the_residual_SI suppression and the desired signal_transmission._Moreover, the relay beamforming matrix should be_updated every time slot_because_the residual SI in_the previous time slot is_amplified by the current beamforming matrix_and added_to the_received signals from the two sources in the current time slot._We derive the optimally balanced relay_beamforming and receive beamforming_matrices in_closed_form based on_minimum_mean square_error, taking into account the propagation of_the residual_loopback SI from the first to_the current time slot._We_also propose beamforming design using only_the channels of

Luminosities and Space Densities of Gamma-Ray Bursts]{}., L117–L120 October 1999. , R., [Crew]{}, G., [Doty]{}, J., [Villasenor]{}, J., [Monnelly]{}, G. [ *et al.*]{} [GRB030329 (=H2652): a long, extremely bright GRB localized by the HETE WXM and SXC.]{}, (2003). , B. A. & [Price]{}, P. A. [GRB 030329: optical afterglow candidate.]{}, (2003). , J., [Peimbert]{}, M., [Estaban]{}, C., [Kaufer]{}, A., [Jaunsen]{}, A. [ *et al.*]{} [Redshift of GRB 030329.]{}, (2003). , K. Z., [Matheson]{}, T., [Garnavich]{}, P. M., [Martini]{}, P., [Berlind]{}, P. [*et al.*]{} [Spectroscopic Discovery of the Supernova 2003dh Associated with GRB 030329]{}., 4173 April 2003. , K. [GRB 030329: OT candidate.]{}, (2003). , D. W., [Yost]{}, S., [Kulkarni]{}, S. R., [Torii]{}, K., [Kato]{}, T. [*et al.*]{} [Early optical emission from the [$\gamma$]{}-ray burst of 4 October 2002]{}., 284–286 March 2003. , C., [Balsano]{}, R., [Barthelemy]{}, S., [Bloch]{}, J., [Butterworth]{}, P. [*et al.*]{} [Observation of contemporaneous optical radiation from a gamma-ray burst.]{}, 400–402 (1999). , E., [Soderberg]{}, A. M. & [Frail]{}, D. A. [GRB 030329: radio observations.]{}, (2003). , N., [Sato]{}, N. & [Nakanishi]{}, H. [GRB 030329 Radio 23/43/90 GHz observations at Nobeyama]{}. GCN Circular 2089 (2003). , J. C., [Meijerink]{}, R., [Tilanus]{}, R. P. J. & [Smith]{}, I.

Luminosities and Space Densities of Gamma - Ray Bursts ] { }. , L117 – L120 October 1999. , R., [ Crew ] { }, G., [ Doty ] { }, J., [ Villasenor ] { }, J., [ Monnelly ] { }, G. [ * et al. * ] { } [ GRB030329 (= H2652 ): a long, extremely bright GRB localize by the HETE WXM and SXC. ] { }, (2003). , B.   A. & [ monetary value ] { }, P.   A. [ GRB 030329: optical afterglow candidate. ] { }, (2003). , J., [ Peimbert ] { }, M., [ Estaban ] { }, C., [ Kaufer ] { }, A., [ Jaunsen ] { }, A. [ * et al. * ] { } [ red shift of GRB 030329. ] { }, (2003). , K.   Z., [ Matheson ] { }, T., [ Garnavich ] { }, P.   M., [ Martini ] { }, P., [ Berlind ] { }, P. [ * et al. * ] { } [ Spectroscopic Discovery of the Supernova 2003dh Associated with GRB 030329 ] { }. , 4173 April 2003. , K. [ GRB 030329: OT candidate. ] { }, (2003). , D.   W., [ Yost ] { }, S., [ Kulkarni ] { }, S.   R., [ Torii ] { }, K., [ Kato ] { }, T. [ * et al. * ] { } [ Early ocular emission from the [ $ \gamma$]{}-ray burst of 4 October 2002 ] { }. , 284–286 March 2003. , C., [ Balsano ] { }, R., [ Barthelemy ] { }, S., [ Bloch ] { }, J., [ Butterworth ] { }, P. [ * et al. * ] { } [ notice of coetaneous optical radiation from a gamma - ray burst. ] { }, 400–402 (1999). , E., [ Soderberg ] { }, A.   M. & [ Frail ] { }, D.   A. [ GRB 030329: radio receiver observations. ] { }, (2003). , N., [ Sato ] { }, N. & [ Nakanishi ] { }, H. [ GRB 030329 Radio 23/43/90 GHz observations at Nobeyama ] { }. GCN Circular 2089 (2003). , J.   C., [ Meijerink ] { }, R., [ Tilanus ] { }, R.   P.   J. & [ Smith ] { }, I.

Lumlnosities and Space Denslties of Gamma-Rai Vursts]{}., L117–L120 Ocfober 1999. , R., [Crew]{}, G., [Doty]{}, J., [Villasenor]{}, J., [Lobnellt]{}, G. [ *et al.*]{} [GRB030329 (=H2652): a lone, extremepy brighr GRU localized by tis HETE WXM ans SXC.]{}, (2003). , B. A. & [Price]{}, P. A. [NRB 030329: optican afterglow catdkdcte.]{}, (2003). , J., [Peimbert]{}, M., [Estaban]{}, C., [Kaufer]{}, A., [Jaunsem]{}, W. [ *et al.*]{} [Redshyft ps GRG 030329.]{}, (2003). , K. Z., [Matheson]{}, T., [Garnavich]{}, P. M., [Marfini]{}, P., [Uerlind]{}, P. [*et al.*]{} [Spectroscopic Discovery ov thf Supernova 2003dh Asslciated wity GRF 030329]{}., 4173 April 2003. , K. [ERB 030329: OT candidate.]{}, (2003). , D. W., [Iost]{}, S., [Kulkarni]{}, S. R., [Torii]{}, K., [Kato]{}, G. [*et cl.*]{} [Early opjnxal gmission fron the [$\gamma$]{}-ray buvxt of 4 Octobet 2002]{}., 284–286 March 2003. , C., [Bslseno]{}, E., [Barthelemy]{}, S., [Bloch]{}, O., [Butterworth]{}, P. [*et aj.*]{} [Observadikn of contemporanwoys opjical radkqtiun rrpm a gamla-rey burst.]{}, 400–402 (1999). , S., [Soderberg]{}, A. M. & [Frail]{}, D. A. [GRB 030329: taqpp observatiohs.]{}, (2003). , N., [Faeo]{}, N. & [Nakanishi]{}, H. [GRB 030329 Radio 23/43/90 GHz obsernatikns at Nobeyama]{}. GCN Cirxular 2089 (2003). , J. C., [Meijerink]{}, T., [Tilanus]{}, W. P. J. & [Smith]{}, I.

Luminosities and Space Densities of Gamma-Ray Bursts]{}., 1999. R., [Crew]{}, [Doty]{}, J., [Villasenor]{}, al.*]{} (=H2652): a long, bright GRB localized the HETE WXM and SXC.]{}, (2003). B. A. & [Price]{}, P. A. [GRB 030329: optical afterglow candidate.]{}, (2003). , [Peimbert]{}, M., [Estaban]{}, C., [Kaufer]{}, A., [Jaunsen]{}, A. [ *et al.*]{} [Redshift of 030329.]{}, , Z., T., [Garnavich]{}, P. M., [Martini]{}, P., [Berlind]{}, P. [*et al.*]{} [Spectroscopic Discovery of the Supernova 2003dh with GRB 030329]{}., 4173 April 2003. , K. 030329: OT candidate.]{}, (2003). D. W., [Yost]{}, S., [Kulkarni]{}, R., K., [Kato]{}, [*et [Early emission from the burst of 4 October 2002]{}., 284–286 March 2003. , C., [Balsano]{}, R., [Barthelemy]{}, S., [Bloch]{}, J., [Butterworth]{}, [*et al.*]{} contemporaneous optical from gamma-ray 400–402 (1999). , A. M. & [Frail]{}, D. A. observations.]{}, (2003). , N., [Sato]{}, N. & [Nakanishi]{}, [GRB 030329 23/43/90 GHz observations at Nobeyama]{}. GCN 2089 (2003). , J. C., [Meijerink]{}, R., [Tilanus]{}, P. J. & [Smith]{}, I.

Luminosities and Space DensiTies of GammA-Ray BUrsTs]{}., L117–l120 OCtobEr 1999. , R., [CRew]{}, G., [Doty]{}, J., [VillaSEnor]{}, j., [Monnelly]{}, G. [ *et al.*]{} [GRB030329 (=H2652): a lonG, extrEmELy brIGhT GRB lOcalizeD By THE HEtE wXm anD Sxc.]{}, (2003). , B. a. & [PricE]{}, P. A. [gRB 030329: optiCal afterglOw cAnDidate.]{}, (2003). , J., [PeimbERt]{}, m., [Estaban]{}, C., [KAufEr]{}, A., [Jaunsen]{}, A. [ *eT al.*]{} [redshiFt Of Grb 030329.]{}, (2003). , K. Z., [MaTheSon]{}, T., [GArnaviCH]{}, P. M., [MarTini]{}, P., [BerlInD]{}, p. [*et al.*]{} [SPEctroscOPIc discOvery of the SupernoVA 2003dH associated with gRB 030329]{}., 4173 AprIl 2003. , k. [gRb 030329: ot caNdiDate.]{}, (2003). , D. W., [Yost]{}, s., [KUlkarNI]{}, S. R., [ToriI]{}, k., [KATO]{}, t. [*et AL.*]{} [Early optical Emission froM The [$\Gamma$]{}-rAy BurST of 4 OctOber 2002]{}., 284–286 MArCH 2003. , C., [BAlsano]{}, R., [BartHeleMy]{}, S., [Bloch]{}, J., [butterWOrth]{}, P. [*et AL.*]{} [ObservAtion oF coNteMporANeOuS opTiCAl rADiAtiON frOm a gamma-RaY bUrst.]{}, 400–402 (1999). , E., [sodeRBERG]{}, A. M. & [FRaiL]{}, D. A. [GrB 030329: radIo observationS.]{}, (2003). , N., [SAto]{}, N. & [nAkaNishi]{}, h. [GRB 030329 RAdio 23/43/90 gHZ obseRvatioNs at NObEyama]{}. GCN CirculaR 2089 (2003). , J. C., [MEijerink]{}, R., [tilAnUs]{}, R. p. J. & [smith]{}, i.

Luminosities and Space Den sities ofGamma -Ra y B ur sts] {}., L117–L120 Oct o ber1999. , R., [Crew]{}, G.,[D o ty]{ } ,J., [ Villase n or ] { },J. ,[Mo nn e ll y]{}, G. [ *etal.*]{} [G RB0 30 329 (=H2652) : a long, ext rem ely bright G RBlocali ze d b y theHET E WXM and S X C.]{}, (2003). , B. A.& [Price ] { }, P.A. [GRB 030329: o p ti c al afterglow c andida te . ]{ } , (2 003 ). , J.,[P eimbe r t]{}, M . ,[ E s tab a n]{}, C., [Ka ufer]{}, A. , [J aunsen ]{ },A . [ *e t al. *] { } [ Redshift of GRB 030329.] {}, (2 0 03). , K. Z.,[Mathe son ]{} , T. , [ Ga rna vi c h]{ } ,P.M .,[Martini ]{ }, P.,[Ber l i n d ]{}, P. [*e t al. *]{} [Spectro sco picD isc overy of t he S up ernov a 2003 dh As so ciated with GRB 030 329]{}.,417 3Apr il 2003 . , K. [G RB030329: OT can d ida te . ] { }, (2003). , D. W., [ Y o st ]{}, S., [Kulk a rn i] { }, S. R. ,[To rii] { } , K., [Ka t o] {}, T. [ *et al . *] {} [Early o ptical e mis sio n fro m the [$\ga mma$]{}- ray b u rst of 4 Octob e r 2002]{}., 2 8 4– 2 8 6M arch 20 03. , C.,[Bal s ano] {},R ., [B a rthel emy]{ }, S. , [Bloch]{}, J., [Bu tt erwort h]{}, P. [*et al.* ]{} [Obser v a t ion of c onte m po r aneous optical radi ation from a gamma- ray b urst.]{} , 400–402 ( 1999). , E .,[So der b e rg ]{}, A. M. &[ F rail ]{ }, D. A . [ GRB 030 329 : r adi o o bs ervations .]{}, (2 00 3) . , N. , [Sa t o]{}, N. & [N ak ani shi]{ } , H. [ GRB 0 3032 9Ra d io23/43/9 0 G H z obs er va tion s a tNobey ama] { }.GCN Cir cular 208 9 ( 2 003) . , J. C., [Meijerink]{ }, R., [Tila nu s]{ }, R.P .  J. & [S mith]{}, I.

Luminosities and_Space Densities_of Gamma-Ray Bursts]{}., L117–L120_October 1999. ,_R.,_[Crew]{}, G.,_[Doty]{},_J., [Villasenor]{}, J.,_[Monnelly]{}, G. [_*et al.*]{} [GRB030329 (=H2652):_a long, extremely_bright_GRB localized by the HETE WXM and SXC.]{}, (2003). , B. A. & [Price]{}, P. A. [GRB_030329:_optical afterglow_candidate.]{},_(2003). ,_J., [Peimbert]{}, M., [Estaban]{}, C.,_[Kaufer]{}, A., [Jaunsen]{}, A. [_*et al.*]{}_[Redshift of GRB 030329.]{}, (2003). , K. Z., [Matheson]{}, T.,_[Garnavich]{},_P. M., [Martini]{}, P.,_[Berlind]{}, P. [*et al.*]{} [Spectroscopic Discovery of the Supernova_2003dh Associated with GRB 030329]{}., 4173_April 2003. , K._[GRB_030329:_OT candidate.]{}, (2003). , D. W.,_[Yost]{}, S., [Kulkarni]{}, S. R., [Torii]{}, K.,_[Kato]{}, T. [*et al.*]{} [Early optical_emission from the [$\gamma$]{}-ray burst of 4_October 2002]{}., 284–286 March 2003. , C.,_[Balsano]{}, R., [Barthelemy]{}, S., [Bloch]{},_J., [Butterworth]{},_P. [*et al.*]{} [Observation of_contemporaneous optical radiation_from a_gamma-ray burst.]{}, 400–402_(1999). , E., [Soderberg]{}, A. M. & [Frail]{},_D. A. [GRB 030329:_radio observations.]{}, (2003). , N., [Sato]{}, N._&_[Nakanishi]{}, H. [GRB_030329_Radio_23/43/90 GHz_observations at Nobeyama]{}._GCN_Circular 2089_(2003). ,_J. C., [Meijerink]{}, R., [Tilanus]{}, R. P. J. &_[Smith]{},_I.

K_\alpha \supseteq H^{00}$, and $$H \cap K_\alpha \ne H \cap K_{\alpha'}$$ for $\alpha < \alpha' < \kappa_1$. By Lemma \[extract-of-mer\], there are subsets $S_1, S_2 \subseteq \kappa_1$ such that $|S_1| < \aleph_0$, $|S_2| = (2^{\kappa_0})^+$, and $$\left( \bigcap_{\alpha \in S_1} K_\alpha \right)^{00} \subseteq \bigcap_{\alpha \in S_2} K_\alpha.$$ Let $J$ be the left-hand side. Then $J$ is an $\omega$-definable group containing $H^{00}$, so $|H/(H \cap J)| < \kappa_0$ by Claim \[diamond-slide-claim\]. Now for any $\alpha \in S_2$, $$J \subseteq K_\alpha \implies H \cap J \subseteq H \cap K_\alpha \subseteq H.$$ There are at most $2^{|H/(H \cap J)|} \le 2^{\kappa_0}$ groups between $H \cap J$ and $J$, so there are at most $2^{\kappa_0}$ possibilities for $H \cap K_\alpha$, contradicting the fact that $|S_2| > 2^{\kappa_0}$ and the $H \cap K_\alpha$ are pairwise distinct for distinct $\alpha$. Now given the claim, we see that the index of $H^{00}$ in $H$ can be at most $\kappa_0^{\kappa_1}$. Indeed, let $\mathcal{S}$ be the collection of $\omega$-definable groups $K$ such that $K \supseteq H^{00}$, and let $\mathcal{S}'$ be a subcollection containing a representative $K$ for every possibility of $H \cap K$. By the second claim, $|\mathcal{S}'| < \kappa_1$. Every type-definable group is an intersection of $\omega$-definable groups, so $$H^{00} = \bigcap_{K \in \mathcal{S}} K = \bigcap_{K \in \mathcal{S}} (H \cap K) = \bigcap_{K \in

K_\alpha \supseteq H^{00}$, and $ $ H \cap K_\alpha \ne H \cap K_{\alpha'}$$ for $ \alpha < \alpha' < \kappa_1$. By Lemma   \[extract - of - mer\ ], there are subsets $ S_1, S_2 \subseteq \kappa_1 $ such that $ |S_1| < \aleph_0 $, $ |S_2| = (2^{\kappa_0})^+$, and $ $ \left (\bigcap_{\alpha \in S_1 } K_\alpha \right)^{00 } \subseteq \bigcap_{\alpha \in S_2 } K_\alpha.$$ Let $ J$ be the leftover - bridge player side. Then $ J$ is an $ \omega$-definable group containing $ H^{00}$, therefore $ |H/(H \cap J)| < \kappa_0 $ by Claim   \[diamond - chute - claim\ ]. Now for any $ \alpha \in S_2 $, $ $ J \subseteq K_\alpha \implies H \cap J \subseteq H \cap K_\alpha \subseteq H.$$ There are at about $ 2^{|H/(H \cap J)| } \le 2^{\kappa_0}$ groups between $ H \cap J$ and $ J$, so there are at about $ 2^{\kappa_0}$ hypothesis for $ H \cap K_\alpha$, contradicting the fact that $ |S_2| > 2^{\kappa_0}$ and the $ H \cap K_\alpha$ are pairwise clear-cut for distinct $ \alpha$. Now give the claim, we see that the exponent of $ H^{00}$ in $ H$ can be at most $ \kappa_0^{\kappa_1}$. Indeed, lease $ \mathcal{S}$ be the solicitation of $ \omega$-definable groups $ K$ such that $ K \supseteq H^{00}$, and let $ \mathcal{S}'$ be a subcollection containing a congressman $ K$ for every possibility of $ H \cap K$. By the second claim, $ |\mathcal{S}'| < \kappa_1$. Every character - definable group is an intersection of $ \omega$-definable group, so $ $ H^{00 } = \bigcap_{K \in \mathcal{S } } K = \bigcap_{K \in \mathcal{S } } (H \cap K) = \bigcap_{K \in

K_\aloha \supseteq H^{00}$, and $$H \cap K_\alpha \ne H \cak J_{\alpha'}$$ for $\ampha < \aloha' < \kappa_1$. By Lemma \[extract-of-lee\], theee are subsets $S_1, S_2 \subreteq \kapia_1$ such tyat $|W_1| < \aleph_0$, $|S_2| = (2^{\kappa_0})^+$, and $$\mcft( \bngrap_{\alpha \in S_1} K_\slpha \righd)^{00} \subseteq \ciycap_{\alpha \in S_2} K_\alpha.$$ Let $J$ be the lqft-hand slde. Then $J$ is wn $\okqga$-dsfinable group containing $H^{00}$, so $|H/(H \dap J)| < \nappa_0$ by Claik \[diamond-slide-claim\]. Now for any $\alpha \in S_2$, $$J \subsfteq K_\alpha \impjues H \cap J \rubseteq H \cap K_\alpha \subseteq H.$$ There are at mort $2^{|H/(H \cap J)|} \le 2^{\jappa_0}$ croups betwxen $H \bap J$ and $J$, so there dre at kost $2^{\kappa_0}$ posxibmlitues for $H \cap K_\alpha$, rontradicting the fast that $|S_2| > 2^{\iappa_0}$ and the $H \cqp K_\alpva$ ase pxurwkse dmstjnct flr vistinct $\allha$. Now giveb the claim, we see uhae the index of $H^{00}$ in $R$ san be at most $\kappa_0^{\kappa_1}$. Indeed, let $\mauhcal{A}$ be the collection of $\imega$-definable groups $K$ such trat $K \supseteq H^{00}$, and let $\mathcal{S}'$ be a subcollecdion rovtanklng x rfpresentative $K$ for every possibility of $H \ca[ K$. Bj the second claii, $|\mathcal{S}'| < \kwplw_1$. Every type-dgfinablz gdoup is an interseftion os $\ometa$-definabje gtoups, so $$H^{00} = \bigcap_{K \in \mathxal{S}} K = \bigccp_{K \in \mathcal{S}} (K \cap K) = \bigeap_{K \im

K_\alpha \supseteq H^{00}$, and $$H \cap K_\alpha \cap for $\alpha \alpha' < \kappa_1$. subsets S_2 \subseteq \kappa_1$ that $|S_1| < $|S_2| = (2^{\kappa_0})^+$, and $$\left( \bigcap_{\alpha S_1} K_\alpha \right)^{00} \subseteq \bigcap_{\alpha \in S_2} K_\alpha.$$ Let $J$ be the left-hand Then $J$ is an $\omega$-definable group containing $H^{00}$, so $|H/(H \cap J)| < by \[diamond-slide-claim\]. for $\alpha \in S_2$, $$J \subseteq K_\alpha \implies H \cap J \subseteq H \cap K_\alpha \subseteq H.$$ are at most $2^{|H/(H \cap J)|} \le 2^{\kappa_0}$ between $H \cap J$ $J$, so there are at $2^{\kappa_0}$ for $H K_\alpha$, the that $|S_2| > and the $H \cap K_\alpha$ are pairwise distinct for distinct $\alpha$. Now given the claim, we see the index in $H$ be most Indeed, let $\mathcal{S}$ collection of $\omega$-definable groups $K$ such H^{00}$, and let $\mathcal{S}'$ be a subcollection containing representative $K$ every possibility of $H \cap K$. the second claim, $|\mathcal{S}'| < \kappa_1$. Every type-definable is an intersection of $\omega$-definable groups, so $$H^{00} = \bigcap_{K \in \mathcal{S}} K = \bigcap_{K (H \cap K) = \in

K_\alpha \supseteq H^{00}$, and $$H \cap K_\alPha \ne H \cap K_{\Alpha'}$$ For $\AlpHa < \AlphA' < \kapPa_1$. By Lemma \[extraCT-of-mEr\], there are subsets $S_1, S_2 \subSeteq \KaPPa_1$ suCH tHat $|S_1| < \aLeph_0$, $|S_2| = (2^{\kaPPa_0})^+$, AND $$\leFt( \BiGcaP_{\aLPhA \in S_1} K_\AlpHa \right)^{00} \Subseteq \biGcaP_{\aLpha \in S_2} K_\alphA.$$ leT $J$ be the lefT-haNd side. Then $J$ iS an $\Omega$-dEfInaBLe groUp cOntaiNing $H^{00}$, sO $|h/(H \cap J)| < \Kappa_0$ by ClAiM \[DiamonD-Slide-clAIM\]. NOw foR any $\alpha \in S_2$, $$J \subsETeQ k_\alpha \implies H \Cap J \suBsETeQ h \Cap k_\alPha \subseteQ H.$$ there ARe at mosT $2^{|h/(H \CAP j)|} \le 2^{\KAppa_0}$ groups betWeen $H \cap J$ anD $j$, so There aRe At mOSt $2^{\kappA_0}$ possIbILitIes for $H \cap K_\AlphA$, contradiCting tHE fact thAT $|S_2| > 2^{\kappa_0}$ And the $h \caP K_\aLpha$ ARe PaIrwIsE DisTInCt fOR diStinct $\alPhA$. NOw givEn thE CLAIm, we See That The inDex of $H^{00}$ in $H$ can bE at Most $\KAppA_0^{\kappA_1}$. IndeEd, leT $\mAthcaL{S}$ be thE collEcTion of $\omega$-defiNablE groups $K$ sUch ThAt $K \SuPseteQ h^{00}$, and leT $\maThcAl{S}'$ be a sUbcolleCTioN cONTAiNing a representativE $K$ FOR eVery possIbilitY Of $h \cAP K$. By the sEcOnd ClaiM, $|\MAthcaL{S}'| < \kaPPa_1$. every typE-definABlE gRoup is aN iNterseCtIon Of $\oMega$-dEFinaBle groUps, so $$H^{00} = \biGcap_{K \IN \mathcal{S}} K = \bigcAP_{K \in \mathcal{S}} (H \CAp k) = \BIgCAp_{K \iN

K_\alpha \supseteq H^{00}$ , and $$H\capK_\ alp ha \ne H \ cap K_{\alpha' } $$ f or $\alpha < \alpha' < \kap pa _ 1$.B yLemma  \[extr a ct - o f-m er \] , t he r eare s ubs ets $S_ 1, S_2 \su bse te q \kappa_1$s uc h that $|S _1| < \alep h_0 $, $|S _2 | = (2^{\ kap pa_0} )^+$,a nd $$\ left( \bi gc a p_{\al p ha \inS _ 1} K_\ alpha \right)^{00 } \ s ubseteq \bigca p_ { \a l p ha\in S_2} K_\a lp ha.$$ Let $J$ be t h e l e ft-hand side. Then $J$ i s an $\ome ga $-d e finabl e gro up con taining $H^ {00} $, so $|H /(H \c a p J)| < \kappa_ 0$ byCla im\[di a mo nd -sl id e -cl a im \]. Now for any $ \a lpha\inS _ 2 $ , $$ J \ subs eteqK_\alpha \imp lie s H\ cap J \s ubset eq H \ cap K _\alph a \subseteq H.$$ The re are at mo st $2 ^{ |H/(H \cap J )|} \l e 2^{\k appa_0} $ gr ou p s be tween $H \cap J$ a nd $ J$ , so the re are at m o st $2^{\ ka ppa _0}$ p ossib ilit i es for $H\cap K _ \a lp ha$, co nt radict in g t hefactt hat$|S_2| > 2^{\k appa_ 0 }$ and the $H\ cap K_\alpha$ ar e pa i rwis e d istinct for dis t inct $\a l ph a$. Nowgiven t h ec laim, we see that t he index of $ H^{00}$ in $H $ can be a t m ost $\ka ppa_ 0 ^{ \ kappa_1}$. Ind eed,let $\math c al{S}$ b e the collect ion of $\ o m ega$-def ina ble gr oup s $K $ such that $ K \sup se teq H ^{0 0}$, an d l et$\m ath ca l{S}'$ be a subco ll ec ti on co ntain i ng a rep re sen ta tiv e $K$ for ev ery p ossi bi li t y o f $H \c a pK $ . By t he sec ond c laim, $|\ m ath cal{S}' | < \kapp a_1 $ . Ev er ytype-de finable group i s an inter se cti on of$ \ omega$-d efinable groups, so $$H ^ {00} =\bi gcap_ {K \ in \mathc al{ S}} K= \ b igcap_ {K \in \ma t h cal{S } } ( H \ ca p K) = \bi g c ap_ {K \i n

K_\alpha \supseteq_H^{00}$, and_$$H \cap K_\alpha \ne_H \cap_K_{\alpha'}$$_for $\alpha_<_\alpha' < \kappa_1$._By Lemma \[extract-of-mer\], there_are subsets $S_1, S_2_\subseteq \kappa_1$ such_that_$|S_1| < \aleph_0$, $|S_2| = (2^{\kappa_0})^+$, and $$\left( \bigcap_{\alpha \in S_1}_K_\alpha_\right)^{00} \subseteq ___ \bigcap_{\alpha \in_S_2} K_\alpha.$$ Let $J$ be_the left-hand_side. Then $J$ is an $\omega$-definable group containing_$H^{00}$,_so $|H/(H \cap_J)| < \kappa_0$ by Claim \[diamond-slide-claim\]. Now for any $\alpha_\in S_2$, $$J \subseteq K_\alpha \implies_H \cap J_\subseteq_H_\cap K_\alpha _ \subseteq H.$$ There_are at most $2^{|H/(H \cap J)|}_\le 2^{\kappa_0}$ groups between $H \cap J$_and $J$, so there are at_most $2^{\kappa_0}$ possibilities for $H_\cap K_\alpha$,_contradicting the fact that $|S_2|_> 2^{\kappa_0}$ and_the $H_\cap K_\alpha$ are_pairwise distinct for distinct $\alpha$. Now given_the claim, we_see that the index of $H^{00}$_in_$H$ can be_at_most_$\kappa_0^{\kappa_1}$. Indeed,_let $\mathcal{S}$ be_the_collection of_$\omega$-definable_groups $K$ such that $K \supseteq __H^{00}$, and let $\mathcal{S}'$ be a subcollection_containing a representative $K$_for_every possibility of $H_\cap K$. By the second_claim, $|\mathcal{S}'| < \kappa_1$. Every type-definable_group is_an intersection_of $\omega$-definable groups, so $$H^{00} = \bigcap_{K \in \mathcal{S}} K =_\bigcap_{K \in _ \mathcal{S}} (H \cap_K) =_\bigcap_{K_\in

R}^n$ is $[\alpha,\beta\otimes u, v\wedge w]\mapsto v\wedge w$; the map $\Lambda^2 \mathbb{R}^n\to H^2(S)\otimes \mathbb{R}^n$, denoted by $[\delta_S\omega]\otimes \mathrm{Id}$, sends $v\wedge w\mapsto [\delta_S\omega_v]\otimes w-[\delta_S\omega_w]\otimes v$; the map $\mathbb{R}^n\to H^2(S)$ is $[\delta_S\omega]$; the map $H^2(S)\to K$ is $[\alpha]\mapsto [\alpha,0,0]$; the map $K\to H^1(S)\otimes\mathbb{R}^n$ is $[\alpha,\beta\otimes v,0]\mapsto [\beta]\otimes v$; the last map is the one from (c$_3$). When proving exactness, the only nontrivial part of the computation is at $\Lambda^2\mathbb{R}^n$. This is based on a simple fact from linear algebra: $$\label{EQ_4} \textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})=\Lambda^2( \textrm{ker}[\delta_S\omega]).$$ So, an element in $\textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})$ can be written as a sum of the form $\sum v\wedge w$, with $v,w\in \textrm{ker}[\delta_S\omega]$. Writing $\delta_S\omega_{v}=d\eta$, $\delta_S\omega_{w}=d\theta$, for $\eta,\theta\in\Omega^1(S)$, one easily checks that $$(\eta\wedge \theta, \eta\otimes w-\theta\otimes v, v\wedge w)\in\Omega^2(A_S)$$ is closed. This implies exactness at $\Lambda^2\mathbb{R}^n$. So $H^2(A_S)$ vanishes if and only if (c$_1$) and (c$_3$) hold and the map $[\delta_S\omega]\otimes \mathrm{Id}$ is injective. By (\[EQ\_4\]), injectivity is equivalent to (c$_2$). We prove that (b) and (c) imply (

R}^n$ is $ [ \alpha,\beta\otimes u, v\wedge w]\mapsto v\wedge w$; the map $ \Lambda^2 \mathbb{R}^n\to H^2(S)\otimes \mathbb{R}^n$, denoted by $ [ \delta_S\omega]\otimes \mathrm{Id}$, sends $ v\wedge w\mapsto [ \delta_S\omega_v]\otimes w-[\delta_S\omega_w]\otimes v$; the function $ \mathbb{R}^n\to H^2(S)$ is $ [ \delta_S\omega]$; the function $ H^2(S)\to K$ is $ [ \alpha]\mapsto [ \alpha,0,0]$; the function $ K\to H^1(S)\otimes\mathbb{R}^n$ is $ [ \alpha,\beta\otimes v,0]\mapsto [ \beta]\otimes v$; the last map is the one from (c$_3 $). When proving exactness, the alone nontrivial part of the computation is at $ \Lambda^2\mathbb{R}^n$. This is establish on a simple fact from linear algebra: $ $ \label{EQ_4 } \textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})=\Lambda^2 (\textrm{ker}[\delta_S\omega]).$$ So, an component in $ \textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})$ can be written as a sum of the shape $ \sum v\wedge w$, with $ v, w\in \textrm{ker}[\delta_S\omega]$. Writing $ \delta_S\omega_{v}=d\eta$, $ \delta_S\omega_{w}=d\theta$, for $ \eta,\theta\in\Omega^1(S)$, one well checks that $ $ (\eta\wedge \theta, \eta\otimes w-\theta\otimes v, v\wedge w)\in\Omega^2(A_S)$$ is closed. This implies exactness at $ \Lambda^2\mathbb{R}^n$. So $ H^2(A_S)$ vanishes if and only if (c$_1 $) and (c$_3 $) hold and the map $ [ \delta_S\omega]\otimes \mathrm{Id}$ is injective. By (\[EQ\_4\ ]), injectivity is equivalent to (c$_2 $). We testify that (b) and (c) imply (

R}^n$ ls $[\alpha,\beta\otimes u, v\weage w]\mapsto v\webte w$; tie map $\Mambda^2 \mxthbb{R}^n\to H^2(S)\otimes \mathbb{R}^n$, venored bt $[\delta_S\omega]\otimes \matfrm{Id}$, senfs $v\wedgw w\mepsto [\delta_S\omege_b]\otimes w-[\deltz_D\omeya_x]\otimes v$; the msp $\mathbb{R}^t\to H^2(S)$ is $[\deltd_S\umzga]$; the map $H^2(S)\to K$ is $[\alpha]\mapsto [\al[ha,0,0]$; the mwp $K\to H^1(S)\otimef\matnfb{R}^n$ ps $[\alpha,\beta\otimes v,0]\mapsto [\beta]\ktimes n$; the last map is the one from (c$_3$). When provijg edactness, the only jontrivial kzrt if the compugation is at $\Lambda^2\matgbb{R}^n$. This is based on a simple fact from lineqr alhgbra: $$\label{EQ_4} \vextrm{her}([\delta_S\omens]\otimev \mathrm{Od})=\Lambda^2( \textrk{kec}[\delra_S\omega]).$$ So, an elemenv in $\textrm{ker}([\delta_S\jmega]\otimas \mathrm{Id})$ can be qrutten as d suo of tht fprj $\sum v\aedje w$, with $v,s\in \textrm{kwr}[\delta_S\omega]$. Writimg $\eelta_S\omega_{v}=d\sta$, $\dejtw_S\omega_{w}=d\theta$, for $\eta,\theta\in\Omega^1(S)$, one eaaily checks that $$(\eta\wedte \theta, \eta\otimes w-\thgta\otimes d, v\wedge w)\in\Omega^2(A_S)$$ is closed. This implies exactnass av $\Uamyea^2\matfvb{G}^n$. So $H^2(A_S)$ vanishes if and only if (c$_1$) and (c$_3$) holq amd the map $[\delta_F\omega]\otimex \layrrm{Id}$ is injeztive. Yg (\[SQ\_4\]), injectivity is fquivalgnt to (c$_2$). We provt thay (b) and (c) imply (

R}^n$ is $[\alpha,\beta\otimes u, v\wedge w]\mapsto v\wedge map \mathbb{R}^n\to H^2(S)\otimes denoted by $[\delta_S\omega]\otimes w-[\delta_S\omega_w]\otimes the map $\mathbb{R}^n\to is $[\delta_S\omega]$; the $H^2(S)\to K$ is $[\alpha]\mapsto [\alpha,0,0]$; the $K\to H^1(S)\otimes\mathbb{R}^n$ is $[\alpha,\beta\otimes v,0]\mapsto [\beta]\otimes v$; the last map is the one (c$_3$). When proving exactness, the only nontrivial part of the computation is at This based a fact from linear algebra: $$\label{EQ_4} \textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})=\Lambda^2( \textrm{ker}[\delta_S\omega]).$$ So, an element in $\textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})$ can be as a sum of the form $\sum v\wedge with $v,w\in \textrm{ker}[\delta_S\omega]$. Writing $\delta_S\omega_{w}=d\theta$, for $\eta,\theta\in\Omega^1(S)$, one easily that \theta, \eta\otimes v, w)\in\Omega^2(A_S)$$ closed. This implies at $\Lambda^2\mathbb{R}^n$. So $H^2(A_S)$ vanishes if and only if (c$_1$) and (c$_3$) hold and the map $[\delta_S\omega]\otimes is injective. injectivity is to We that (b) and (

R}^n$ is $[\alpha,\beta\otimes u, v\wedgE w]\mapsto v\wEdge w$; The Map $\laMbda^2 \MathBb{R}^n\to H^2(S)\otimes \MAthbB{R}^n$, denoted by $[\delta_S\omegA]\otimEs \MAthrM{id}$, Sends $V\wedge w\MApSTO [\deLtA_S\OmeGa_V]\OtImes w-[\DelTa_S\omegA_w]\otimes v$; tHe mAp $\Mathbb{R}^n\to H^2(S)$ IS $[\dElta_S\omega]$; The Map $H^2(S)\to K$ is $[\alPha]\Mapsto [\AlPha,0,0]$; THe map $k\to h^1(S)\otiMes\matHBb{R}^n$ is $[\Alpha,\beta\OtIMes v,0]\maPSto [\beta]\OTImEs v$; tHe last map is the one FRoM (C$_3$). When proving exActnesS, tHE oNLY noNtrIvial part oF tHe comPUtation IS aT $\lAMbdA^2\Mathbb{R}^n$. This iS based on a siMPle Fact frOm LinEAr algeBra: $$\laBeL{eQ_4} \tExtrm{ker}([\delTa_S\oMega]\otimeS \mathrM{id})=\LambdA^2( \Textrm{kEr}[\deltA_S\oMegA]).$$ So, aN ElEmEnt In $\TExtRM{kEr}([\dELta_s\omega]\otImEs \MathrM{Id})$ cAN BE WritTen As a sUm of tHe form $\sum v\wedGe w$, With $V,W\in \TextrM{ker}[\dElta_s\oMega]$. WRiting $\Delta_s\oMega_{v}=d\eta$, $\delta_S\OmegA_{w}=d\theta$, fOr $\eTa,\TheTa\In\OmeGA^1(S)$, one eAsiLy cHecks thAt $$(\eta\weDGe \tHeTA, \ETa\Otimes w-\theta\otimes V, v\WEDgE w)\in\OmegA^2(A_S)$$ is cLOsEd. tHis impliEs ExaCtneSS At $\LamBda^2\mAThBb{R}^n$. So $H^2(A_s)$ vanisHEs If And only If (C$_1$) and (c$_3$) hOlD anD thE map $[\dELta_S\Omega]\oTimes \matHrm{Id}$ IS injective. By (\[EQ\_4\]), INjectivity is eQUiVALeNT to (c$_2$). we pRove that (b) anD (c) imPLy (

R}^n$ is $[\alpha,\beta\ot imes u, v\ wedge w] \ma ps to v \wed ge w$; the map $\La mbda^2 \mathbb{R}^n\to H^2( S) \ otim e s\math bb{R}^n $ ,d e not ed b y $ [\ d el ta_S\ ome ga]\oti mes \mathr m{I d} $, sends $v\ w ed ge w\mapst o [ \delta_S\ome ga_ v]\oti me s w - [\del ta_ S\ome ga_w]\ o timesv$; the m ap $\math b b{R}^n\ t o H ^2(S )$ is $[\delta_S\ o me g a]$; the map $ H^2(S) \t o K $ is$[\ alpha]\map st o [\a l pha,0,0 ] $; t h e m a p $K\to H^1(S )\otimes\ma t hbb {R}^n$ i s $ [ \alpha ,\bet a\ o tim es v,0]\map sto[\beta]\o timesv $; thel ast map is th e o nefrom (c $_ 3$) .W hen pr ovi n g e xactness ,th e onl y no n t r i vial pa rt o f the computationisat $ \ Lam bda^2 \math bb{R }^ n$. T his is base don a simple fac t fr om linear al ge bra :$$\la b el{EQ_ 4}\te xtrm{ke r}([\de l ta_ S\ o m e ga ]\otimes \mathrm{I d} ) = \L ambda^2( \text r m{ ke r }[\delta _S \om ega] ) . $$ So , an el ement in $\tex t rm {k er}([\d el ta_S\o me ga] \ot imes\ math rm{Id} )$ can b e wri t ten as a sum o f the form $\s u mv \ we d ge w $,with $v,w\i n \t e xtrm {ker } [\ del t a_S\o mega] $. Wr i ting $\delta_S\omeg a_ {v}=d\ eta$, $\delta_S\om ega_{w}=d\ t h e ta$, for $\e t a, \ theta\in\Omega ^1(S) $, one eas i ly check s tha t $$(\et a\wedge \ t h eta, \et a\o tim esw-\ t h et a\otimes v, v \ w edge w )\in\Om ega ^2(A_S) $$isclo sed .This impl ies exac tn es sat $\ Lambd a ^2\mathb b{ R}^ n$ . S o $H^ 2 (A_S)$ vani shes i fa ndonly if (c $ _ 1$)an d(c$_ 3$) h old a nd t h e m ap $[\d elta_S\om ega ] \oti me s\mathrm {Id}$ is inje ct ive. By (\ [E Q\_ 4\]),i n jectivit y is equivalent to (c$_ 2 $). We pr ove t hat(b) and ( c)imply(

R}^n$ is_$[\alpha,\beta\otimes u,_v\wedge w]\mapsto v\wedge w$;_the map_$\Lambda^2_\mathbb{R}^n\to H^2(S)\otimes_\mathbb{R}^n$,_denoted by $[\delta_S\omega]\otimes \mathrm{Id}$,_sends $v\wedge w\mapsto_[\delta_S\omega_v]\otimes w-[\delta_S\omega_w]\otimes v$; the_map $\mathbb{R}^n\to H^2(S)$_is_$[\delta_S\omega]$; the map $H^2(S)\to K$ is $[\alpha]\mapsto [\alpha,0,0]$; the map $K\to H^1(S)\otimes\mathbb{R}^n$ is $[\alpha,\beta\otimes_v,0]\mapsto_[\beta]\otimes v$;_the_last_map is the one from_(c$_3$). When proving exactness, the_only nontrivial_part of the computation is at $\Lambda^2\mathbb{R}^n$. This_is_based on a_simple fact from linear algebra: $$\label{EQ_4} \textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})=\Lambda^2( \textrm{ker}[\delta_S\omega]).$$ So, an_element in $\textrm{ker}([\delta_S\omega]\otimes \mathrm{Id})$ can be_written as a_sum_of_the form $\sum v\wedge w$,_with $v,w\in \textrm{ker}[\delta_S\omega]$. Writing $\delta_S\omega_{v}=d\eta$, $\delta_S\omega_{w}=d\theta$,_for $\eta,\theta\in\Omega^1(S)$, one easily checks that_$$(\eta\wedge \theta, \eta\otimes w-\theta\otimes v, v\wedge w)\in\Omega^2(A_S)$$ is_closed. This implies exactness at $\Lambda^2\mathbb{R}^n$._So $H^2(A_S)$ vanishes if and_only if_(c$_1$) and (c$_3$) hold and_the map $[\delta_S\omega]\otimes_\mathrm{Id}$ is_injective. By (\[EQ\_4\]),_injectivity is equivalent to (c$_2$). We prove_that (b) and_(c) imply (

respectively. ### Transit parameters Before analyzing the TTV signals, we revise the transit parameters of KOI-94c, KOI-94d, and KOI-94e obtained by the Kepler team (Table \[ppar\_prior\]) so that they are consistent with the light curves obtained in Section \[sec:dataprocess\]. Here we first use the parameters publicized by the Kepler team to phase-fold the observed transit light curves, and then refit those phase curves to obtain the revised transit parameters. In the first step, we fit each of the detrended light curve centered at the transit (for $\sim 1.7$ times its duration) to obtain the times of transit centers $t_c$, using a Markov chain Monte Carlo (MCMC) algorithm. Here we use a light curve model by @2009ApJ...690....1O, and fix $a/R_{\star}$, $R_p/R_{\star}$, and $b$ to the values obtained by the Kepler team, assuming $e = 0$. We model the limb-darkening using a quadratic law in Eq.(\[qlimb\]) and adopt the limb-darkening coefficients $u_1$ and $u_2$ obtained by @2012ApJ...759L..36H (all these parameters are summarized in Table \[ppar\_prior\]). Since the detrend procedure in Section \[sec:dataprocess\] can remove only the out-of-transit outliers, we also exclude in-transit $5\sigma$ outliers of this fit, if any, and fit the light curve again. Using the series of $t_c$ obtained in this way, we construct the phase-folded transit light curve for each planet. As the second step, we fit the resulting phase-folded transit light curve for $a/R_{\star}$, $R_p/R_{\star}$, $b$, $u_1$, and $u_2$ using the same light curve model as above. In this way, we obtain the revised values of the set of parameters shown in Table \[ppar\_fit\] and the corresponding best-fit light curves (Figures \[94c\_phase\] to \[94e\_phase\]).[^1] In this fit, all the parameters converge well in the case of KOI-94d. In contrast, $a/R_{\star}$ and $b$ of KOI-94c and KO

respectively. # # # Transit parameters Before analyzing the TTV signals, we retool the passage parameters of KOI-94c, KOI-94d, and KOI-94e obtained by the Kepler team (mesa \[ppar\_prior\ ]) so that they are coherent with the light curves obtain in Section \[sec: dataprocess\ ]. Here we first use the parameter publicized by the Kepler team to phase - fold the note transit light curves, and then refit those phase curves to obtain the revised theodolite parameters. In the first step, we equip each of the detrended light curve centered at the transit (for $ \sim 1.7 $ multiplication its duration) to obtain the times of transit centers $ t_c$, using a Markov range Monte Carlo (MCMC) algorithm. Here we use a light curve model by @2009ApJ... 690.... 1O, and fix $ a / R_{\star}$, $ R_p / R_{\star}$, and $ b$ to the values prevail by the Kepler team, assuming $ e = 0$. We model the limb - darkening using a quadratic law in Eq.(\[qlimb\ ]) and adopt the limb - darkening coefficients $ u_1 $ and $ u_2 $ obtain by @2012ApJ... 759L.. 36H (all these parameters are summarized in Table \[ppar\_prior\ ]). Since the detrend routine in Section \[sec: dataprocess\ ] can absent only the out - of - transit outlier, we also exclude in - transit $ 5\sigma$ outliers of this paroxysm, if any, and fit the light curve again. Using the series of $ t_c$ obtained in this means, we construct the phase - folded transit light curvature for each planet. As the second tone, we fit the resulting phase - folded transit light curvature for $ a / R_{\star}$, $ R_p / R_{\star}$, $ b$, $ u_1 $, and $ u_2 $ using the same light curvature model as above. In this way, we obtain the revised value of the set of parameter shown in Table \[ppar\_fit\ ] and the corresponding well - fit light curve (Figures \[94c\_phase\ ] to \[94e\_phase\]).[^1 ] In this paroxysm, all the parameters converge well in the casing of KOI-94d. In contrast, $ a / R_{\star}$ and $ b$ of KOI-94c and KO

redpectively. ### Transit paramtters Before analyeibg the TTV sjgnals, wd revise the transit parametxrs if KOU-94c, KOI-94d, and KOI-94e obtaived by thv Kepler ream (Rable \[ppar\_'dior\]) so that fmey axe consistent wijh the light curves obtainad iu Section \[sec:dataprocess\]. Here we firft use yhf parameters poblicpzqd bg the Kepler team to phase-fold the observtd transit light cirves, and then refit those phade curves to obtaij the revisgs twqnsit paramegers. In the first step, se fit each of the detrended lieht cbrve centergb at jhe transit (hor $\sii 1.7$ times its duratiot) to obyain the times of trqnsit centers $t_c$, usinj a Markov chain Monje Carlo (MWME) algorithm. Here we uwe a licht wurvd moael bb @2009AlJ...690....1O, anf fmx $a/R_{\star}$, $R_l/R_{\star}$, and $v$ to the values obtsigvc by the Kepmer tewm, assuming $e = 0$. We model the limb-darkeninc uaing a quadratic law in Eq.(\[qlimb\]) and adopt thg limb-darkqning coefficients $u_1$ and $u_2$ obtained by @2012ApJ...759L..36H (all dhese oarcneterr age summarized in Table \[ppar\_prior\]). Since the deedemd procedure in Fection \[sec:cahaltocess\] can remuve only fhe out-of-transit oktliers, we aoso excluqe im-transit $5\sigma$ outliers of rhis fit, if cny, and fit the light curve agaiu. Usinb the series of $t_c$ obtained nn thia way, we cojstruct tgd phase-folded trxnspt lhght curve for each planet. Ws the serond xtep, we fit the rqsulting pjase-folded transit light curvg for $d/R_{\star}$, $R_p/R_{\dtar}$, $b$, $u_1$, and $u_2$ using the same lmjht curve modgl ds dbove. In this way, we obtaig the revised ralues oy the ret of parzmeters shown in Twble \[ppar\_fit\] djd the correvponding besr-fit light zjrves (Figures \[94v\_phase\] to \[94e\_phase\]).[^1] In this fit, all the iaramgtsrs converge weol un the case of LOI-94a. Ig bonvrast, $d/R_{\star}$ and $b$ of YOI-94z and YO

respectively. ### Transit parameters Before analyzing the we the transit of KOI-94c, KOI-94d, Kepler (Table \[ppar\_prior\]) so they are consistent the light curves obtained in Section Here we first use the parameters publicized by the Kepler team to phase-fold observed transit light curves, and then refit those phase curves to obtain the transit In first we fit each of the detrended light curve centered at the transit (for $\sim 1.7$ times duration) to obtain the times of transit centers using a Markov chain Carlo (MCMC) algorithm. Here we a curve model @2009ApJ...690....1O, fix $R_p/R_{\star}$, and $b$ the values obtained by the Kepler team, assuming $e = 0$. We model the limb-darkening using a law in adopt the coefficients and obtained by @2012ApJ...759L..36H parameters are summarized in Table \[ppar\_prior\]). procedure in Section \[sec:dataprocess\] can remove only the outliers, we exclude in-transit $5\sigma$ outliers of this if any, and fit the light curve again. the series of $t_c$ obtained in this way, we construct the phase-folded transit light curve planet. As the second we fit the phase-folded light for $R_p/R_{\star}$, $b$, and $u_2$ using the same light curve model as above. In way, we obtain the revised values of the set of in \[ppar\_fit\] and the best-fit light curves (Figures to In this fit, all converge in KOI-94d. contrast, and $b$ of KOI-94c KO

respectively. ### Transit parameTers Before AnalyZinG thE TtV siGnalS, we revise the trANsit Parameters of KOI-94c, KOI-94d, anD KOI-94e ObTAineD By The KePler teaM (taBLE \[ppAr\_PrIor\]) So THaT they Are ConsistEnt with the LigHt Curves obtainED iN Section \[seC:daTaprocess\]. HerE we First uSe The PArameTerS publIcized BY the KePler team tO pHAse-folD The obseRVEd TranSit light curves, and THeN Refit those phasE curveS tO ObTAIn tHe rEvised tranSiT paraMEters. In THe FIRSt sTEp, we fit each of The detrendeD LigHt curvE cEntERed at tHe traNsIT (foR $\sim 1.7$ times itS durAtion) to obTain thE Times of TRansit cEnters $T_c$, uSinG a MaRKoV cHaiN MONte cArLo (McmC) aLgorithm. heRe We use A ligHT CURve mOdeL by @2009APJ...690....1O, anD fix $a/R_{\star}$, $R_p/R_{\StaR}$, and $B$ To tHe valUes obTainEd By the kepler Team, aSsUming $e = 0$. We model thE limB-darkeninG usInG a qUaDratiC Law in EQ.(\[qlImb\]) And adopT the limB-DarKeNING cOefficients $u_1$ and $u_2$ obTaINEd By @2012ApJ...759L..36H (aLl thesE PaRaMEters are SuMmaRizeD IN TablE \[ppaR\_PrIor\]). Since The detREnD pRocedurE iN SectiOn \[Sec:DatAprocESs\] caN removE only the Out-of-TRansit outliers, WE also exclude iN-TrANSiT $5\SigmA$ ouTliers of thiS fit, IF any, And fIT tHe lIGht cuRve agAiN. usINg the series of $t_c$ obtaInEd in thIs way, We construct thE phase-foldED TRansit liGht cURvE For each planet. AS the sEcond step, wE Fit the reSultiNg phase-fOlded tranSIT light cuRve For $A/R_{\sTar}$, $r_P/r_{\sTar}$, $b$, $u_1$, and $u_2$ usinG THe saMe Light cuRve Model as AboVe. IN thIs wAy, We obtain tHe reviseD vAlUeS oF thE set oF ParameteRs ShoWn In TAble \[pPAr\_fit\] aNd the CorrEsPoNDinG best-fiT LiGHT curVeS (FIgurEs \[94c\_PhAse\] to \[94E\_phaSE\]).[^1] In This fit, All the parAmeTErs cOnVeRge well In the case of KOi-94d. in contrast, $A/R_{\StaR}$ and $b$ oF koI-94c and KO

respectively. ### Transi t paramete rs B efo rean alyz ingthe TTV signal s , we revise the transit pa ramet er s ofK OI -94c, KOI-94 d ,a n d K OI -9 4eob t ai ned b y t he Kepl er team (T abl e\[ppar\_prio r \] ) so thatthe y are consis ten t with t hel ightcur ves o btaine d in Se ction \[s ec : datapr o cess\]. H er e we first use the pa r am e ters publicize d by t he Ke p l ertea m to phase -f old t h e obser v ed t r ans i t light curve s, and then ref it tho se ph a se cur ves t oo bta in the revi sedtransit p aramet e rs. In the fir st ste p,wefite ac hofth e de t re nde d li ght curv ece ntere d at t h e tra nsi t (f or $\ sim 1.7$ time s i ts d u rat ion)to ob tain t he ti mes of tran si t centers $t_c$ , us ing a Mar kov c hai nMonte Carlo(MC MC) algori thm. He r e w eu s e a light curve model b y @2 009ApJ.. .690.. . .1 O, and fix$a /R_ {\st a r }$, $ R_p/ R _{ \star}$, and $ b $to the va lu es obt ai ned by theK eple r team , assumi ng $e = 0$. We model the limb-dark e ni n g u s inga q uadratic la w in Eq.( \[ql i mb \]) and a doptth e l i mb-darkening coeffi ci ents $ u_1$and $u_2$ obt ained by @ 2 0 1 2ApJ...7 59L. . 36 H (all these pa ramet ers are su m marizedin Ta ble \[pp ar\_prior \ ] ). Since th e d etr end p ro cedure in Sec t i on \ [s ec:data pro cess\]can re mov e o nl y the out -of-tran si tou tl ier s, we also exc lu dein -tr ansit $5\sig ma$ o utli er so f t his fit , i f any, a nd fit th elight cur v e a gain. U sing theser i es o f$t _c$ obt ained in this w ay, we con st ruc t thep h ase-fold ed transit light curvef or each pl anet. As the seco ndstep,wef it the resul tingph ase - f olded t ra nsi tlight curv e for $a/R _{ \sta r}$, $R _p/R_{\star}$, $b$ , $u _1$, and $u_2 $ u sing t he sa m el igh tc urv e model as above. In this w ay , w e obtain t h e r ev ised va lues of thes et of p arameters shown in T able \ [pp ar\_fit\]and thecorrespon d ing b e st -fitlig ht cur ve s ( Figur es \[9 4 c\_ phase \] to\[ 94e\_p hase\ ]) .[^1] In this fit, all the para meters conv erg e well in th e ca se of KOI -94d . In contr ast , $ a/R_{ \st a r}$ a nd $ b $ofK OI-94 c an d KO

respectively. ###_Transit parameters Before_analyzing the TTV signals,_we revise_the_transit parameters_of_KOI-94c, KOI-94d, and_KOI-94e obtained by_the Kepler team (Table_\[ppar\_prior\]) so that_they_are consistent with the light curves obtained in Section \[sec:dataprocess\]. Here we first use_the_parameters publicized_by_the_Kepler team to phase-fold the_observed transit light curves, and_then refit_those phase curves to obtain the revised transit_parameters. In_the first step,_we fit each of the detrended light curve centered_at the transit (for $\sim 1.7$_times its duration)_to_obtain_the times of transit_centers $t_c$, using a Markov chain_Monte Carlo (MCMC) algorithm. Here we_use a light curve model by @2009ApJ...690....1O,_and fix $a/R_{\star}$, $R_p/R_{\star}$, and $b$_to the values obtained by_the Kepler_team, assuming $e = 0$._We model the_limb-darkening using_a quadratic law_in Eq.(\[qlimb\]) and adopt the limb-darkening_coefficients $u_1$ and_$u_2$ obtained by @2012ApJ...759L..36H (all these_parameters_are summarized in_Table_\[ppar\_prior\])._Since the_detrend procedure in_Section_\[sec:dataprocess\] can_remove_only the out-of-transit outliers, we also_exclude_in-transit $5\sigma$ outliers of this fit, if_any, and fit the_light_curve again. Using the_series of $t_c$ obtained in_this way, we construct the phase-folded_transit light_curve for_each planet. As the second step, we fit the resulting phase-folded transit_light curve for $a/R_{\star}$, $R_p/R_{\star}$, $b$,_$u_1$, and $u_2$ using_the same_light_curve model as_above._In this_way, we obtain the revised values of_the set_of parameters shown in Table \[ppar\_fit\]_and the corresponding best-fit_light_curves (Figures \[94c\_phase\] to \[94e\_phase\]).[^1] In_this fit, all the parameters converge_well in the case of_KOI-94d._In_contrast, $a/R_{\star}$ and $b$ of_KOI-94c and KO

subset of cells, a QCA is then a homomorphism of the quasi-local algebra, which commutes with lattice translations and satisfies locality on the neighborhood. The observable-based approach was first used in Ref. [@richter_ergodicity_1996] with focus on the irreversible case. However, this definition left questions open such as whether the composition of two QCA will again form a QCA. The following definition does avoid this uncertainty. Consider an infinite $d$-dimensional lattice ${\mathrm{L}}\subset \mathbb{Z}^d$ of cells $x\in \mathbb{Z}^d$, where each cell is associated with the observable algebra $\mathcal{A}_x$ and each of these algebras is an isomorphic copy of the algebra of complex $d\times d$-matrices. When $\Lambda \subset \mathbb{Z}^d$ is a finite subset of cells, denote by $\mathcal{A}(\Lambda)$ the algebra of observables belonging to all cells in $\Lambda$, i.e. the tensor product $\otimes_{x\in \Lambda} \mathcal{A}_x$. The completion of this algebra is called a *quasi-local* algebra and will be denoted by $\mathcal{A}(\mathbb{Z}^d)$. A *quantum cellular automaton* with neighborhood scheme ${\mathcal{N}}\subset \mathbb{Z}^d$ is a homomorphism $T : \mathcal{A}(\mathbb{Z}^d) \rightarrow \mathcal{A}(\mathbb{Z}^d)$ of the quasi-local algebra, which commutes with lattice translations, and satisfies the locality condition $T(\mathcal{A}(\Lambda)) \subset T(\mathcal{A}(\Lambda + {\mathcal{N}})) $ for every finite set $\Lambda \subset \mathbb{Z}^d$. The local transition rule of a cellular automaton is the homomorphism $T_0 : \mathcal{A}_0 \rightarrow \mathcal{A}({\mathcal{N}})$. They present and prove the following theorem on one-dimensional QCA. \[the.str\] Let $T$ be the global transition homomorphism of a one-dimensional nearest-neighbor QCA on the lattice $\mathbb{Z}^d$ with single-cell algebra $\mathcal{A}_0 = \mathcal{M}_d$. Then $T$ can be represented in the generalized Margolus partitioning scheme, i.e. $T$ restricts

subset of cells, a QCA is then a homomorphism of the quasi - local algebra, which commutes with lattice translations and satisfies vicinity on the vicinity. The observable - based overture was foremost used in Ref.   [ @richter_ergodicity_1996 ] with focus on the irreversible case. However, this definition exit questions open such as whether the constitution of two QCA will again form a QCA. The following definition does invalidate this uncertainty. Consider an countless $ d$-dimensional lattice $ { \mathrm{L}}\subset \mathbb{Z}^d$ of cells $ x\in \mathbb{Z}^d$, where each cellular telephone is associated with the discernible algebra $ \mathcal{A}_x$ and each of these algebra is an isomorphic copy of the algebra of complex $ d\times d$-matrices. When $ \Lambda \subset \mathbb{Z}^d$ is a finite subset of cells, denote by $ \mathcal{A}(\Lambda)$ the algebra of observables belong to all cells in $ \Lambda$, i.e. the tensor product $ \otimes_{x\in \Lambda } \mathcal{A}_x$. The completion of this algebra is called a * quasi - local * algebra and will be denoted by $ \mathcal{A}(\mathbb{Z}^d)$. A * quantum cellular automaton * with vicinity scheme $ { \mathcal{N}}\subset \mathbb{Z}^d$ is a homomorphism $ T: \mathcal{A}(\mathbb{Z}^d) \rightarrow \mathcal{A}(\mathbb{Z}^d)$ of the quasi - local algebra, which commutes with lattice translations, and meet the locality condition $ T(\mathcal{A}(\Lambda) ) \subset T(\mathcal{A}(\Lambda + { \mathcal{N } }) ) $ for every finite set $ \Lambda \subset \mathbb{Z}^d$. The local transition rule of a cellular automaton is the homomorphism $ T_0: \mathcal{A}_0 \rightarrow \mathcal{A}({\mathcal{N}})$. They present and prove the following theorem on one - dimensional QCA. \[the.str\ ] Let $ T$ be the global transition homomorphism of a one - dimensional nearest - neighbor QCA on the lattice $ \mathbb{Z}^d$ with single - cell algebra $ \mathcal{A}_0 = \mathcal{M}_d$. Then $ T$ can be represented in the generalized Margolus partitioning scheme, i.e. $ T$ restricts

suhset of cells, a QCA is tmen a homomorphism of tie quasj-local augebra, which commutes with lettixe trqnslations and satisfids localiny on the neijhborhood. The obsxdvable-based aliroack xas first used ln Ref. [@richtar_ergodicity_1996] whtf yocus on the irreversible case. Howevqr, this dffinition left quexeiona open such as whether the composifion of two QCA will again form a QCA. The folllwinh definition does wvoid this ohcewrainty. Considdr an infinite $d$-dimensjonal lattice ${\mathrm{L}}\subset \matfbb{Z}^d$ of cells $z\ib \mwjhbb{Z}^d$, where each cell is associated fith thr observable akgeura $\nathcal{A}_x$ and each of these algebras is an isomospkic copy of the algebea of cmmplax $d\gumer d$-javrides. Whfn $\Mambda \subaet \mathbb{Z}^d$ is a finite subset os cells, denote by $\maehsal{A}(\Lambda)$ the algebra of observables btlongjng to all cells in $\Lamvda$, i.e. the tensor profuct $\otimqs_{x\in \Lambda} \mathcal{A}_x$. The completion of this algabra ms caolcd a *wuwsi-local* algebra and will be denoted by $\mathcwm{A}(\kanhbb{Z}^d)$. A *quantum ccllular automaton* eihh geighborhood rcheme ${\mafhcal{N}}\subset \mathbb{X}^d$ is a homonorphism $U : \mayhcal{A}(\mathbb{Z}^d) \rightarrow \marhcal{A}(\mathbb{D}^d)$ od the quasi-local apgebra, whick commotes woth lattice translationr, ans satisfies the locamkty condition $T(\mxthbal{A}(\Nambda)) \subset T(\mathcal{A}(\Lamfda + {\mathral{N}})) $ for evdry ginite set $\Lambdw \subset \mathbb{Z}^d$. The lofal ttansithon rule ov a cellular automaton is the homomorphism $T_0 : \mdthbal{A}_0 \righcarrow \kathcal{A}({\mathcwl{N}})$. They presenj and prore the following theorek on one-dimqnsional QCA. \[tvg.str\] Let $T$ be the glofal rranwition fumomorphism of a one-dimvnfionql nearest-neighbor QCX on the lattice $\mauhvb{Z}^d$ with singke-cdll apgxbra $\idthcal{A}_0 = \mathwal{M}_a$. Tfrn $T$ zan be reprcsevted in the generalized Kargklus partitioning xcmeme, i.e. $T$ restrices

subset of cells, a QCA is then of quasi-local algebra, commutes with lattice the The observable-based approach first used in [@richter_ergodicity_1996] with focus on the irreversible However, this definition left questions open such as whether the composition of two will again form a QCA. The following definition does avoid this uncertainty. Consider infinite lattice \mathbb{Z}^d$ cells $x\in \mathbb{Z}^d$, where each cell is associated with the observable algebra $\mathcal{A}_x$ and each of algebras is an isomorphic copy of the algebra complex $d\times d$-matrices. When \subset \mathbb{Z}^d$ is a finite of denote by the of belonging to all in $\Lambda$, i.e. the tensor product $\otimes_{x\in \Lambda} \mathcal{A}_x$. The completion of this algebra is called a algebra and denoted by A cellular with neighborhood scheme is a homomorphism $T : \mathcal{A}(\mathbb{Z}^d) the quasi-local algebra, which commutes with lattice translations, satisfies the condition $T(\mathcal{A}(\Lambda)) \subset T(\mathcal{A}(\Lambda + {\mathcal{N}})) for every finite set $\Lambda \subset \mathbb{Z}^d$. The transition rule of a cellular automaton is the homomorphism $T_0 : \mathcal{A}_0 \rightarrow \mathcal{A}({\mathcal{N}})$. They prove the following theorem one-dimensional QCA. \[the.str\] $T$ the transition of a nearest-neighbor QCA on the lattice $\mathbb{Z}^d$ with single-cell algebra $\mathcal{A}_0 = Then $T$ can be represented in the generalized Margolus partitioning $T$

subset of cells, a QCA is then a hOmomorphisM of thE quAsi-LoCal aLgebRa, which commuteS With Lattice translations and SatisFiES locALiTy on tHe neighBOrHOOd. THe ObSerVaBLe-Based AppRoach waS first used In REf. [@Richter_ergodICiTy_1996] with focuS on The irreversiBle Case. HoWeVer, THis deFinItion Left quEStions Open such aS wHEther tHE composITIoN of tWo QCA will again forM A Qca. The following dEfinitIoN DoES AvoId tHis uncertaInTy. ConSIder an iNFiNITE $d$-dIMensional lattIce ${\mathrm{L}}\sUBseT \mathbB{Z}^D$ of CElls $x\iN \mathBb{z}^D$, whEre each cell Is asSociated wIth the OBservabLE algebrA $\mathcAl{A}_X$ anD eacH Of ThEse AlGEbrAS iS an ISomOrphic coPy Of The alGebrA OF COmplEx $d\TimeS d$-matRices. When $\LambDa \sUbseT \MatHbb{Z}^d$ Is a fiNite SuBset oF cells, DenotE bY $\mathcal{A}(\Lambda)$ The aLgebra of oBseRvAblEs BelonGIng to aLl cEllS in $\LambDa$, i.e. the TEnsOr PRODuCt $\otimes_{x\in \Lambda} \mAtHCAl{a}_x$. The comPletioN Of ThIS algebra Is CalLed a *QUAsi-loCal* aLGeBra and wiLl be deNOtEd By $\mathcAl{a}(\mathbB{Z}^D)$. A *qUanTum ceLLulaR automAton* with NeighBOrhood scheme ${\maTHcal{N}}\subset \maTHbB{z}^D$ iS A homOmoRphism $T : \mathCal{A}(\MAthbB{Z}^d) \rIGhTarROw \matHcal{A}(\MaTHbB{z}^d)$ of the quasi-local alGeBra, whiCh comMutes with lattIce translaTIONs, and satIsfiES tHE locality condiTion $T(\Mathcal{A}(\LaMBda)) \subseT T(\matHcal{A}(\LamBda + {\mathcaL{n}})) $ For every FinIte Set $\lamBDA \sUbset \mathbb{Z}^d$. tHE locAl TransitIon Rule of a CelLulAr aUtoMaTon is the hOmomorphIsM $T_0 : \MaThCal{a}_0 \righTArrow \matHcAl{A}({\MaThcAl{N}})$. ThEY preseNt and ProvE tHe FOllOwing thEOrEM On onE-dImEnsiOnaL QcA. \[the.Str\] LET $T$ bE the gloBal transiTioN HomoMoRpHism of a One-dimensionaL nEarest-neigHbOr QcA on thE LAttice $\maThbb{Z}^d$ with single-cell algEBra $\mathCal{a}_0 = \mathCal{M}_D$. Then $T$ can Be rEpreseNteD In the gEneralIzed MArGolUS PartiTIOnIng ScHeme, i.e. $T$ resTRIctS

subset of cells, a QCA is then a ho momor phi smof the qua si-local algeb r a, w hich commutes with lat ticetr a nsla t io ns an d satis f ie s loc al it y o nt he neig hbo rhood. The obser vab le -based appro a ch was first us ed in Ref. [ @ri chter_ er god i city_ 199 6] wi th foc u s on t he irreve rs i ble ca s e. Howe v e r, thi s definition left qu e stions open su ch aswh e th e r th e c omposition o f two QCA wil l a g a i n f o rm a QCA. The followingd efi nition d oes avoidthisun c ert ainty. Con side r an infi nite $ d $-dimen s ional l attice ${ \ma thrm { L} }\ sub se t \m a th bb{ Z }^d $ of cel ls $ x\in\mat h b b { Z}^d $,wher e eac h cell is ass oci ated wit h the obse rvab le alge bra $\ mathc al {A}_x$ and each ofthese alg ebr as is a n iso m orphic co pyof thealgebra ofco m p l ex $d\times d$-matri ce s . W hen $\La mbda \ s ub se t \mathbb {Z }^d $ is a fini te s u bs et of ce lls, d e no te by $\m at hcal{A }( \La mbd a)$ t h e al gebraof obser vable s belonging toa ll cells in $ \ La m b da $ , i. e.the tensorprod u ct $ \oti m es _{x \ in \L ambda }\ ma t hcal{A}_x$. The com pl etionof th is algebra is called a* q u asi-loca l* a l ge b ra and will be deno ted by $\m a thcal{A} (\mat hbb{Z}^d )$. A *q u a ntum cel lul araut oma t o n* with neighbo r h oodsc heme ${ \ma thcal{N }}\ sub set \m at hbb{Z}^d$ is a ho mo mo rp hi sm$T :\ mathcal{ A} (\m at hbb {Z}^d ) \righ tarro w \m at hc a l{A }(\math b b{ Z } ^d)$ o fthequa si -loca l al g ebr a, whic h commute s w i th l at ti ce tran slations, and s atisfies t he lo cality c ondition $T(\mathcal{A}(\Lambda ) ) \subs etT(\ma thca l{A}(\Lam bda + {\m ath c al{N}} )) $ f or ev er y f i n ite s e t $ \La mb da \subset \ mat hbb{Z }^ d$.The loc al transition rule ofa cellular au tom aton i sthe ho m omo rp h ism $ T_0 : \mathcal{ A}_0 \righ ta r ro w \mathcal { A}( {\ mathcal {N}})$. The y presen t and pro ve the fo ll owin g the orem on on e-dimens ional QCA . \[t h e. str\] Le t $T$be th e glo bal tr a nsi tionhomomo rp hism o f a o ne -dimensi onal nearest-neighbor Q CA onthe l att ice $\mat hbb { Z}^ d$ with s ingl e-cell alg ebr a $ \math cal { A}_0= \m a th cal { M}_d$ . Th e n $T$ can be re p r es ented in th e g ene raliz edM argolu s pa rtitioning scheme , i.e. $T$ rest rict s

subset_of cells,_a QCA is then_a homomorphism_of_the quasi-local_algebra,_which commutes with_lattice translations and_satisfies locality on the_neighborhood. The observable-based approach_was_first used in Ref. [@richter_ergodicity_1996] with focus on the irreversible case. However, this definition left_questions_open such_as_whether_the composition of two QCA_will again form a QCA._The following_definition does avoid this uncertainty. Consider an infinite $d$-dimensional_lattice_${\mathrm{L}}\subset \mathbb{Z}^d$ of_cells $x\in \mathbb{Z}^d$, where each cell is associated with_the observable algebra $\mathcal{A}_x$ and each_of these algebras_is_an_isomorphic copy of the_algebra of complex $d\times d$-matrices. When_$\Lambda \subset \mathbb{Z}^d$ is a finite subset_of cells, denote by $\mathcal{A}(\Lambda)$ the algebra_of observables belonging to all cells_in $\Lambda$, i.e. the tensor_product $\otimes_{x\in_\Lambda} \mathcal{A}_x$. The completion of_this algebra is_called a_*quasi-local* algebra and_will be denoted by $\mathcal{A}(\mathbb{Z}^d)$. A *quantum_cellular automaton* with_neighborhood scheme ${\mathcal{N}}\subset \mathbb{Z}^d$ is a homomorphism_$T_: \mathcal{A}(\mathbb{Z}^d) \rightarrow \mathcal{A}(\mathbb{Z}^d)$_of_the_quasi-local algebra,_which commutes with_lattice_translations, and_satisfies_the locality condition $T(\mathcal{A}(\Lambda)) \subset T(\mathcal{A}(\Lambda_+_{\mathcal{N}})) $ for every finite set $\Lambda_\subset \mathbb{Z}^d$. The local_transition_rule of a cellular_automaton is the homomorphism $T_0_: \mathcal{A}_0 \rightarrow \mathcal{A}({\mathcal{N}})$. They present and prove_the following_theorem on_one-dimensional QCA. \[the.str\] Let $T$ be the global transition homomorphism of a_one-dimensional nearest-neighbor QCA on the lattice_$\mathbb{Z}^d$ with single-cell algebra_$\mathcal{A}_0 = \mathcal{M}_d$._Then_$T$ can be_represented_in the_generalized Margolus partitioning scheme, i.e. $T$ restricts

than for pullers); whereas when $r_c=1.5 \sigma$, $P(s)$ corresponds to high polar order for large clusters with $\beta=0$ and weak pullers. The same features can be observed when $\xi=10$, independently on $r_c$. $P(s)$ for WP squirmers vanishes for all cluster sizes, independently of $\beta$ and $r_c$ when $\xi=1$, since the formation of trimers and chains assemble the particles head to head, thus it avoids the polar order. However, when $\xi=10$, the polar order decays with $s$: for $r_c=1.5\sigma$ weak pullers form clusters with high polar order, while for $r_c=2.5\sigma$ $P(s)$ decays faster for pushers than for pullers. In the latter case, the WP squirmers show another type of alignment, with particles with a high nematic order in the range of $\beta=0$ and weak pushers. During aggregation, we have identified seven different cases depending on whether the attractive patch is oriented against the propulsion direction (AP squirmers) or directed towards it (WP squirmers): 3 gas states, 3 clustering cases and 1 coarsening. On the one side, AP squirmers coarsen isotropically if the interaction is strong enough $\xi \approx 0.1$. When the interaction strength competes with self-propulsion ($\xi =1$), dynamic clusters emerge whose mean size depend on hydrodynamic stresses. When activity dominates ($\xi =10$), particles form a gas and depending on the value of $\beta$ this gas can be isotropically oriented or polarly oriented. On the other side, WP squirmers form chains of particles if the interaction is high enough ($\xi \approx 0.1$), or a suspension of trimers and tetramers if the interaction competes with the active propulsion ($\xi =1$). Whenever activity dominates ($\xi =10$), particles form a gas and depending on the value of $\beta$ this gas can be isotropic, polar or even nematically oriented. Therefore, on one hand anisotropy drives the formation of structures of trimers and tetramers for WP squirmers; on the other hand anisotropy modulates the sensitivity of the hydrodynamic signature: while WP squirmers are more sensitive to $r_c$ when $\xi=10$, AP squirmers

than for pullers); whereas when $ r_c=1.5 \sigma$, $ P(s)$ corresponds to high polar order for big bunch with $ \beta=0 $ and weak pullers. The like features can be observe when $ \xi=10 $, independently on $ r_c$. $ P(s)$ for WP squirmers vanishes for all cluster size, independently of $ \beta$ and $ r_c$ when $ \xi=1 $, since the formation of trimers and range assemble the particles read/write head to head, thus it avoid the polar order. However, when $ \xi=10 $, the polar order decays with $ s$: for $ r_c=1.5\sigma$ weak puller form clusters with high pivotal order, while for $ r_c=2.5\sigma$ $ P(s)$ decays faster for pusher than for pullers. In the latter case, the WP squirmers show another type of alliance, with particles with a high nematic order in the range of $ \beta=0 $ and weak pushers. During aggregation, we have identified seven different cases depending on whether the attractive patch is oriented against the propulsion direction (AP squirmers) or directed towards it (WP squirmers ): 3 gasoline states, 3 clustering case and 1 coarsen. On the one side, AP squirmers coarsen isotropically if the interaction is potent enough $ \xi \approx 0.1$. When the interaction lastingness competes with self - propulsion ($ \xi = 1 $), dynamic bunch emerge whose mean size depend on hydrodynamic stresses. When activity dominates ($ \xi = 10 $), particles form a gas and depending on the value of $ \beta$ this gas can be isotropically oriented or polarly orient. On the early slope, WP squirmers form chains of particles if the interaction is eminent enough ($ \xi \approx 0.1 $), or a suspension of trimers and tetramers if the interaction competes with the active propulsion ($ \xi = 1 $). Whenever natural process dominates ($ \xi = 10 $), particles form a gas and depending on the value of $ \beta$ this natural gas can be isotropic, polar or even nematically orient. Therefore, on one hired hand anisotropy drives the constitution of structures of trimer and tetramers for WP squirmers; on the other bridge player anisotropy modulates the sensitivity of the hydrodynamic signature: while WP squirmers are more sensible to $ r_c$ when $ \xi=10 $, AP squirmers

thwn for pullers); whereas wmen $r_c=1.5 \sigma$, $P(s)$ eirrespmnds tk high pular order for large clusterd qith $\veta=0$ and weak pullers. Ghe same veatures can ve observev when $\xl=10$, indsiendeutoy on $r_c$. $P(s)$ fot WP squirmess vanishes fos xlp cluster sizes, independently of $\beea$ and $t_c$ when $\xi=1$, since the sormzniin of trimers and chains assejble tht particles head tp head, thus it avoids the oolag order. However, whfn $\xi=10$, the pilar irder decays with $s$: fog $r_c=1.5\sigma$ wgak pullers form clusters with hkgh pplar order, wyilf for $r_c=2.5\sigme$ $P(s)$ dvcays faster npr pusvers thsn for pullers. In thw latter case, the WP vquirmers show anojher type mf alignment, with pqrriclev widh a yigf ntmavic order in the range of $\beta=0$ ane weak pushers. Durinb wtgregation, we have ydqntified seven different cases dependinc oh whether the attractivw patch is oriented ahainst thq propulsion direction (AP squirmers) or directed tmwarda it (WI sqjurlers): 3 gas states, 3 clustering cases and 1 coarsqhimg. On the one sibe, AP squirmers vowrxgn isotropicaluy if cge interaction is stgong enjugh $\zi \approx 0.1$. Whrn the interaction strength competes winh swlf-propulsion ($\xi =1$), bynamic cluscers ekerge whose mean size depend on gydrodynamif stressea. When activity dumikatas ($\xi =10$), particles form a gaf and depxndiny on the valoe of $\bqta$ this gws cak be isotropically lrienjed or polarly ogiented. On the other side, WP sqnmrmers form cnahns of partncles lf the interactyon is high enpugh ($\xi \apprux 0.1$), or a slspension of trimers and tetramerv if the intecaction cjmperes qith thd active propulxion ($\xi =1$). Ckenever axtivity dominates ($\wi =10$), pxdticles form a ycs and depending pn ghe vwlne of $\teta$ this gav cav bd isotfopic, polar ur efen nematically orietted. Fherefore, on one hsnq anisottopy drivqs the formatoon of structures lf trmmers end teyraiers for WP squirmers; on the ofher hand anlsotropy modujatew the sensitnvity of the hydrodynamic signature: whilx WP squirmers are morg sensitive to $r_c$ wheu $\wi=10$, AP squirmxrs

than for pullers); whereas when $r_c=1.5 \sigma$, to polar order large clusters with same can be observed $\xi=10$, independently on $P(s)$ for WP squirmers vanishes for cluster sizes, independently of $\beta$ and $r_c$ when $\xi=1$, since the formation of and chains assemble the particles head to head, thus it avoids the polar However, $\xi=10$, polar decays with $s$: for $r_c=1.5\sigma$ weak pullers form clusters with high polar order, while for $r_c=2.5\sigma$ decays faster for pushers than for pullers. In latter case, the WP show another type of alignment, particles a high order the of $\beta=0$ and pushers. During aggregation, we have identified seven different cases depending on whether the attractive patch is oriented the propulsion squirmers) or towards (WP 3 gas states, cases and 1 coarsening. On the squirmers coarsen isotropically if the interaction is strong $\xi \approx When the interaction strength competes with ($\xi =1$), dynamic clusters emerge whose mean size on hydrodynamic stresses. When activity dominates ($\xi =10$), particles form a gas and depending on of $\beta$ this gas be isotropically oriented polarly On other WP squirmers chains of particles if the interaction is high enough ($\xi \approx or a suspension of trimers and tetramers if the interaction the propulsion ($\xi =1$). activity dominates ($\xi =10$), form gas and depending on of this isotropic, or nematically oriented. Therefore, on hand anisotropy drives the formation structures of trimers and the other hand anisotropy modulates the sensitivity of hydrodynamic signature: while WP squirmers are more to $r_c$ when $\xi=10$, AP squirmers

than for pullers); whereas when $R_c=1.5 \sigma$, $P(s)$ cOrresPonDs tO hIgh pOlar Order for large cLUsteRs with $\beta=0$ and weak pulleRs. The SaME feaTUrEs can Be obserVEd WHEn $\xI=10$, iNdEpeNdENtLy on $r_C$. $P(s)$ For WP sqUirmers vanIshEs For all clusteR SiZes, indepenDenTly of $\beta$ and $R_c$ wHen $\xi=1$, sInCe tHE formAtiOn of tRimers ANd chaiNs assemblE tHE partiCLes head TO HeAd, thUs it avoids the polaR OrDEr. However, when $\xI=10$, the poLaR OrDER deCayS with $s$: for $r_C=1.5\sIgma$ wEAk pulleRS fORM CluSTers with high pOlar order, whILe fOr $r_c=2.5\siGmA$ $P(s)$ DEcays fAster FoR PusHers than for PullErs. In the lAtter cASe, the WP SQuirmerS show aNotHer Type OF aLiGnmEnT, WitH PaRtiCLes With a higH nEmAtic oRder IN THE ranGe oF $\betA=0$ and wEak pushers. DurIng AggrEGatIon, we Have iDentIfIed seVen difFerenT cAses depending on WhetHer the attRacTiVe pAtCh is oRIented AgaInsT the proPulsion DIreCtION (aP Squirmers) or directeD tOWArDs it (WP sqUirmerS): 3 GaS sTAtes, 3 clusTeRinG casES And 1 coArseNInG. On the onE side, Ap SqUiRmers coArSen isoTrOpiCalLy if tHE intEractiOn is stroNg enoUGh $\xi \approx 0.1$. When THe interaction STrENGtH CompEteS with self-prOpulSIon ($\xI =1$), dynAMiC clUSters EmergE wHOsE Mean size depend on hydRoDynamiC streSses. When activIty dominatES ($\XI =10$), particlEs foRM a GAs and depending On the Value of $\betA$ This gas cAn be iSotropicAlly orienTED or polarLy oRieNteD. On THE oTher side, WP squIRMers FoRm chainS of ParticlEs iF thE inTerAcTion is higH enough ($\xI \aPpRoX 0.1$), oR a sUspenSIon of triMeRs aNd TetRamerS If the iNteraCtioN cOmPEteS with thE AcTIVe prOpUlSion ($\Xi =1$). WHeNever ActiVIty DominatEs ($\xi =10$), partiCleS Form A gAs And depeNding on the valUe Of $\beta$ this GaS caN be isoTROpic, polaR or even nematically orienTEd. ThereForE, on onE hanD anisotroPy dRives tHe fORmatioN of strUcturEs Of tRIMers aND TeTraMeRs for WP squIRMerS; on thE oTher Hand aniSotropy modulates thE SenSitivity of the HydRodyNAMiC siGNaTUre: WhILe Wp SQuirmers are more Sensitive tO $r_C$ WhEn $\xi=10$, AP squiRMerS

than for pullers); wherea s when $r_ c=1.5 \s igm a$ , $P (s)$ corresponds t o hig h polar order for larg e clu st e rs w i th $\be ta=0$ a n dw e akpu ll ers .T he same fe aturescan be obs erv ed when $\xi=1 0 $, independe ntl y on $r_c$.$P( s)$ fo rWPs quirm ers vani shes f o r allcluster s iz e s, ind e pendent l y o f $\ beta$ and $r_c$ w h en $\xi=1$, since the f or m at i o n o f t rimers and c hains assembl e t h e par t icles head to head, thus itavoids t hep olar o rder. H o wev er, when $\ xi=1 0$, the p olar o r der dec a ys with $s$:for $r _c=1 . 5\ si gma $w eak pu lle r s f orm clus te rs with hig h p o larord er,while for $r_c=2.5 \si gma$ $P( s)$ d ecays fas te r for pushe rs th an for pullers. I n th e lattercas e, th eWP sq u irmers sh owanother type o f al ig n m e nt , with particles w it h ahigh nem atic o r de ri n the ra ng e o f $\ b e ta=0$ and we ak pushe rs. D u ri ng aggreg at ion, w ehav e i denti f iedsevendifferen t cas e s depending on whether the a t tr a c ti v e pa tch is oriente d ag a inst the pr opu l siondirec ti o n( AP squirmers) or di re cted t oward s it (WP squi rmers): 3g a s states, 3 c l us t ering cases an d 1 c oarsening. On the o ne si de, AP s quirmersc o arsen is otr opi cal lyi f t he interactio n is s tr ong eno ugh $\xi \ app rox 0. 1$. W hen the i nteracti on s tr en gth comp e tes with s elf -p rop ulsio n ($\xi =1$) , dy na mi c cl usterse me r g e wh os emean si ze depe nd o n hy drodyna mic stres ses . Whe nac tivitydominates ($\ xi =10$), pa rt icl es for m a gas an d depending on the valu e of $\b eta $ thi s ga s can beiso tropic all y orien ted or pola rl y o r i ented . On th eother side , WPsquir me rs f orm cha ins of particles i f th e interaction is hig h en oug h ( $ \xi \ a ppr o x 0.1$), or a su spension o ft ri mers and t e tra me rs if t he inte racti o n compe tes withthe activ eprop u l sio n ($\xi =1 $). When ever acti v ity d o mi nates ($ \xi =1 0$ ),parti cles f o rma gas and d ep ending on t he value o f $\beta$ this gas canbe iso tropi c,polar oreve n ne matically ori ented. Th ere for e, on on e hand ani s ot rop y driv es t h e formati o nofs t ru ctures of t r i m ers andtet r amersforWP squirmers; ont he other handanis o t rop y m o dula te s the sensitiv ity o f the hydr od ynamic sign ature: w hi l e WPsquirm ers ar e mores e ns i tive t o $r _c$ when $\x i=1 0$ , AP squ ir me r s

than_for pullers);_whereas when $r_c=1.5 \sigma$,_$P(s)$ corresponds_to_high polar_order_for large clusters_with $\beta=0$ and_weak pullers. The same_features can be_observed_when $\xi=10$, independently on $r_c$. $P(s)$ for WP squirmers vanishes for all cluster sizes,_independently_of $\beta$_and_$r_c$_when $\xi=1$, since the formation_of trimers and chains assemble_the particles_head to head, thus it avoids the polar_order._However, when $\xi=10$,_the polar order decays with $s$: for $r_c=1.5\sigma$ weak_pullers form clusters with high polar_order, while for_$r_c=2.5\sigma$_$P(s)$_decays faster for pushers_than for pullers. In the latter_case, the WP squirmers show another_type of alignment, with particles with a_high nematic order in the range_of $\beta=0$ and weak pushers. During_aggregation, we_have identified seven different cases_depending on whether_the attractive_patch is oriented_against the propulsion direction (AP squirmers)_or directed towards_it (WP squirmers): 3 gas states,_3_clustering cases and_1_coarsening._On the_one side, AP_squirmers_coarsen isotropically_if_the interaction is strong enough $\xi_\approx_0.1$. When the interaction strength competes with_self-propulsion ($\xi =1$), dynamic_clusters_emerge whose mean size_depend on hydrodynamic stresses. When_activity dominates ($\xi =10$), particles form_a gas_and depending_on the value of $\beta$ this gas can be isotropically oriented_or polarly oriented. On the other_side, WP squirmers form_chains of_particles_if the interaction_is_high enough_($\xi \approx 0.1$), or a suspension of_trimers and_tetramers if the interaction competes with_the active propulsion ($\xi_=1$)._Whenever activity dominates ($\xi =10$), particles_form a gas and depending on_the value of $\beta$ this_gas_can_be isotropic, polar or even_nematically oriented. Therefore, on one hand anisotropy_drives the formation_of structures of trimers and tetramers for_WP_squirmers; on the other hand anisotropy_modulates_the sensitivity of the hydrodynamic signature:_while_WP_squirmers are more sensitive to_$r_c$ when $\xi=10$, AP squirmers

tiny contribution proportional to $C_{\gamma Z}$. For $m_a>2m_\mu$, the gray regions are excluded by a dark-photon search in the $e^+ e^-\to\mu^+\mu^-+\mu^+\mu^-$ channel performed by BaBar [@TheBABAR:2016rlg].](mu1MeVand10GeV){width="88.00000%"} In our numerical analysis, we will assume that the contribution of $K_{a_\mu}(\mu)$ is subleading at the high scale $\mu=\Lambda$. If the Wilson coefficients $c_{\mu\mu}$ and $C_{\gamma\gamma}$ are of similar magnitude, the logarithmically enhanced contribution is the parametrically largest one-loop correction. It gives a positive shift of $a_\mu$ provided the product $c_{\mu\mu}\,C_{\gamma\gamma}$ is negative. The correction proportional to $C_{\gamma Z}$ is suppressed by $(1-4s_w^2)$ and hence is numerically subdominant. Note also that the contribution proportional to $(c_{\mu\mu})^2$ is suppressed in the limit where $m_a^2\gg m_\mu^2$, while the remaining terms remain unsuppressed. Figure \[fig:g-2\] shows the regions in the parameter space of the couplings $c_{\mu\mu}$ and $C_{\gamma\gamma}$ in which the experimental value of the muon anomalous magnetic moment can be explained in terms of the ALP-induced loop corrections shown in Figure \[fig:amugraphs\], without invoking a large contribution from the unknown short-distance coefficient $K_{a_\mu}(\Lambda)$. There is a weak dependence on the ALP mass, such that the allowed parameter space increases for $m_a^2\gg m_\mu^2$. Interestingly, we find that an explanation of the anomaly is possible without much tuning as long as one coefficients is of order $\Lambda/\mbox{TeV}$, while the other one can be of similar order or larger. Since $c_{\mu\mu}$ enters observables always in combination with $m_\mu$, it is less constrained by perturbativity than $C_{\gamma\gamma}$. ![\[fig:BaBargraphs\] Tree-level Feynman diagrams contributing to the process $e^+ e^-\to\mu^+\mu^- a$.](BaBar_graphs){width="95.00000%"}

tiny contribution proportional to $ C_{\gamma Z}$. For $ m_a>2m_\mu$, the gray regions are excluded by a blue - photon search in the $ e^+ e^-\to\mu^+\mu^-+\mu^+\mu^-$ distribution channel performed by BaBar [ @TheBABAR:2016rlg].](mu1MeVand10GeV){width="88.00000% " } In our numerical analysis, we will simulate that the contribution of $ K_{a_\mu}(\mu)$ is subleading at the high scale $ \mu=\Lambda$. If the Wilson coefficients $ c_{\mu\mu}$ and $ C_{\gamma\gamma}$ are of alike magnitude, the logarithmically enhanced contribution is the parametrically large one - loop correction. It give a positive shift of $ a_\mu$ provide the merchandise $ c_{\mu\mu}\,C_{\gamma\gamma}$ is negative. The correction proportional to $ C_{\gamma Z}$ is suppressed by $ (1 - 4s_w^2)$ and hence is numerically subdominant. Note besides that the contribution proportional to $ (c_{\mu\mu})^2 $ is suppressed in the limit where $ m_a^2\gg m_\mu^2 $, while the remaining price remain unsuppressed. Figure   \[fig: g-2\ ] show the region in the parameter space of the couplings $ c_{\mu\mu}$ and $ C_{\gamma\gamma}$ in which the experimental value of the muon anomalous magnetic moment can be excuse in terms of the ALP - induced loop corrections shown in Figure   \[fig: amugraphs\ ], without invoking a large contribution from the unknown short - distance coefficient $ K_{a_\mu}(\Lambda)$. There is a weak dependence on the ALP mass, such that the allowed parameter space increases for $ m_a^2\gg m_\mu^2$. Interestingly, we recover that an explanation of the anomaly is possible without much tuning as long as one coefficient is of decree $ \Lambda/\mbox{TeV}$, while the early one can be of similar order or larger. Since $ c_{\mu\mu}$ enters observables always in combination with $ m_\mu$, it is less constrained by perturbativity than $ C_{\gamma\gamma}$. ! [ \[fig: BaBargraphs\ ] Tree - level Feynman diagrams contributing to the procedure $ e^+ e^-\to\mu^+\mu^- a$.](BaBar_graphs){width="95.00000% " }

tijy contribution proportiunal to $C_{\gamma E}$. Dor $m_a>2k_\mu$, ths gray rdgions are excluded by a darn-pyoton search in the $e^+ e^-\to\mu^+\ou^-+\mu^+\mu^-$ chwnnel peefornwd by BaBac [@TheBABAR:2016rlg].](mh1LeVaud10JeV){width="88.00000%"} In our kumerical atalysis, we wiln xsdume that the contribution of $K_{a_\mu}(\mt)$ is sunlfading at the righ fcals $\mu=\Lambda$. If the Wilson coefficiehts $c_{\mu\ku}$ and $C_{\gamma\bamma}$ are of similar magnihude, the logarithmicalpy enhanced coneeibution is ghe parameurncally larggst one-loop correction. It gives x posntive shift od $a_\lo$ provided tie proquct $c_{\mu\mu}\,C_{\gamma\gammd}$ is nebative. The corvectimn proportional to $C_{\gamme Z}$ is suppressed by $(1-4s_w^2)$ and hanee is numerically subeoninanj. Nota alri tfat tie dontrihutmon proportjonal to $(c_{\my\mu})^2$ is suppressed im eye limit whers $m_a^2\gg m_\iu^2$, while the remaining terms remain unslpprsssed. Figure \[fig:g-2\] shows tye regions in the parwmeter spwce of the couplings $c_{\mu\mu}$ and $C_{\gamma\gamma}$ in whiwh thx dxptrlidbtwl value of the muon anomalous magnetic momene csn be explained ln terms of the ALL-ijdised loop corrgctions shkwn in Figure \[fig:amkgraphs\], withiut invokyng s large contribution from tye unknown skorr-distance coefficiznt $K_{a_\mu}(\Lambba)$. Thete is s weak dependence on thz ALP jass, such tjat the amuowed parameter rpabe itcreases for $m_a^2\gg m_\mu^2$. Intewestingly, we yind thag an explagation of hhe akmmaly is possible aithoot mucv tuning ad long as one coefficients is oh order $\Lambds/\mtox{NeV}$, while the pther one can be of similar order jr lafger. Since $c_{\mu\mu}$ xnters obserdables always ln combinatimn with $i_\mu$, ut iw less zunstrained by lerturbatpvnty than $X_{\gamma\gamma}$. ![\[fig:BaBavgrapfa\] Tree-level Feyuoqn diagrams conyricutynh vo thq process $e^+ e^-\do\mu^+\ou^- x$.](NaBar_eraphs){width="95.00000%"}

tiny contribution proportional to $C_{\gamma Z}$. For gray are excluded a dark-photon search performed BaBar [@TheBABAR:2016rlg].](mu1MeVand10GeV){width="88.00000%"} In numerical analysis, we assume that the contribution of $K_{a_\mu}(\mu)$ subleading at the high scale $\mu=\Lambda$. If the Wilson coefficients $c_{\mu\mu}$ and $C_{\gamma\gamma}$ of similar magnitude, the logarithmically enhanced contribution is the parametrically largest one-loop correction. gives positive of provided the product $c_{\mu\mu}\,C_{\gamma\gamma}$ is negative. The correction proportional to $C_{\gamma Z}$ is suppressed by $(1-4s_w^2)$ hence is numerically subdominant. Note also that the proportional to $(c_{\mu\mu})^2$ is in the limit where $m_a^2\gg while remaining terms unsuppressed. \[fig:g-2\] the regions in parameter space of the couplings $c_{\mu\mu}$ and $C_{\gamma\gamma}$ in which the experimental value of the muon anomalous moment can in terms the loop shown in Figure invoking a large contribution from the $K_{a_\mu}(\Lambda)$. There is a weak dependence on the mass, such the allowed parameter space increases for m_\mu^2$. Interestingly, we find that an explanation of anomaly is possible without much tuning as long as one coefficients is of order $\Lambda/\mbox{TeV}$, other one can be similar order or Since enters always combination with it is less constrained by perturbativity than $C_{\gamma\gamma}$. ![\[fig:BaBargraphs\] Tree-level Feynman contributing to the process $e^+ e^-\to\mu^+\mu^- a$.](BaBar_graphs){width="95.00000%"}

tiny contribution proportioNal to $C_{\gammA Z}$. For $M_a>2m_\Mu$, tHe Gray RegiOns are excluded BY a daRk-photon search in the $e^+ e^-\tO\mu^+\mu^-+\Mu^+\MU^-$ chaNNeL perfOrmed by bABAR [@thebAbAr:2016rlG].](mU1mevand10GEV){wIdth="88.00000%"} In oUr numericaL anAlYsis, we will asSUmE that the coNtrIbution of $K_{a_\mU}(\mu)$ Is sublEaDinG At the HigH scalE $\mu=\LamBDa$. If thE Wilson coEfFIcientS $C_{\mu\mu}$ anD $c_{\GaMma\gAmma}$ are of similar mAGnITude, the logaritHmicalLy ENhANCed ConTribution iS tHe parAMetricaLLy LARGesT One-loop correcTion. It gives A PosItive sHiFt oF $A_\mu$ proVided ThE ProDuct $c_{\mu\mu}\,C_{\gAmma\Gamma}$ is neGative. tHe correCTion proPortioNal To $C_{\GammA z}$ iS sUppReSSed BY $(1-4s_W^2)$ anD HenCe is numeRiCaLly suBdomINANT. NotE alSo thAt the Contribution pRopOrtiONal To $(c_{\mu\Mu})^2$ is sUpprEsSed in The limIt wheRe $M_a^2\gg m_\mu^2$, while the RemaIning termS reMaIn uNsUppreSSed. FigUre \[Fig:G-2\] shows tHe regioNS in ThE PARaMeter space of the couPlINGs $C_{\mu\mu}$ and $c_{\gamma\GAmMa}$ IN which thE eXpeRimeNTAl valUe of THe Muon anomAlous mAGnEtIc momenT cAn be exPlAinEd iN termS Of thE ALP-inDuced looP corrECtions shown in FIGure \[fig:amugraPHs\], WIThOUt inVokIng a large coNtriBUtioN froM ThE unKNown sHort-dIsTAnCE coefficient $K_{a_\mu}(\LamBdA)$. There Is a weAk dependence oN the ALP masS, SUCh that thE allOWeD Parameter space IncreAses for $m_a^2\gG M_\mu^2$. InterEstinGly, we finD that an exPLAnation oF thE anOmaLy iS POsSible without mUCH tunInG as long As oNe coeffIciEntS is Of oRdEr $\Lambda/\mBox{TeV}$, whIlE tHe OtHer One caN Be of simiLaR orDeR or LargeR. since $c_{\Mu\mu}$ eNterS oBsERvaBles alwAYs IN CombInAtIon wIth $M_\mU$, it is Less COnsTrained By perturbAtiVIty tHaN $C_{\Gamma\gaMma}$. ![\[fig:BaBargrApHs\] Tree-leveL FEynMan diaGRAms contrIbuting to the process $e^+ e^-\to\MU^+\mu^- a$.](BaBAr_gRaphs){WidtH="95.00000%"}

tiny contribution proport ional to $ C_{\g amm a Z }$ . Fo r $m _a>2m_\mu$, th e gra y regions are excluded by a d a rk-p h ot on se arch in th e $e^ +e^ -\t o\ m u^ +\mu^ -+\ mu^+\mu ^-$ channe l p er formed by Ba B ar [@TheBABA R:2 016rlg].](mu 1Me Vand10 Ge V){ w idth= "88 .0000 0%"} I n ournumerical a n alysis , we wil l as sume that the contrib u ti o n of $K_{a_\mu }(\mu) $i ss u ble adi ng at thehi gh sc a le $\mu = \L a m b da$ . If the Wilso n coefficie n ts$c_{\m u\ mu} $ and $ C_{\g am m a\g amma}$ areof s imilar ma gnitud e , the l o garithm ically en han cedc on tr ibu ti o n i s t hep ara metrical ly l arges t on e - l o op c orr ecti on. I t gives a pos iti ve s h ift of $ a_\mu $ pr ov idedthe pr oduct $ c_{\mu\mu}\,C_{ \gam ma\gamma} $ i sneg at ive.T he cor rec tio n propo rtional to$C _ { \ ga mma Z}$ is suppres se d by $(1-4s_ w^2)$a nd h e nce is n um eri call y subdo mina n t. Note al so tha t t he contri bu tion p ro por tio nal t o $(c _{\mu\ mu})^2$is su p pressed in the limit where $ m _a ^ 2 \g g m_\ mu^ 2$, while t he r e main ingt er msr emain unsu pp r es s ed. Figure \[fig:g -2 \] sho ws th e regions inthe parame t e r space o f th e c o uplings $c_{\m u\mu} $ and $C_{ \ gamma\ga mma}$ in whic h the exp e r imentalval ueofthe m uo n anomalous m a g neti cmomentcan be exp lai ned in te rm s of theALP-indu ce dlo op co rrect i ons show ninFi gur e \[f i g:amug raphs \],wi th o utinvokin g a l arge c on trib uti on from the unk nown sh ort-dista nce coef fi ci ent $K_ {a_\mu}(\Lamb da )$. Thereis aweak d e p endenceon the ALP mass, such t h at theall owedpara meter spa ceincrea ses for $m _a^2\g g m_\ mu ^2$ . Inter e s ti ngl y, we find t h a t a n exp la nati on of t he anomaly is poss i ble without much tu ning a slon g a s on ec oef f i cients is of or der $\Lamb da / \m box{TeV}$, whi le the ot her one canb e of si milar ord er or lar ge r. S i n ce$c_{\mu\mu }$ enter s observa b les a l wa ys in co mbinat io n w ith $ m_\mu$ , it is l ess co ns traine d bype rturbati vity than $C_{\gamma\ga mma}$. ![\ [fi g:BaBargr aph s \]Tree-leve l Fe ynman diag ram s c ontri but i ng to the pr oce s s $e^ + e^ - \to\mu^+\ m u^ - a $ . ]( BaBar_graph s ) { wid th="9 5.0 0 000%"}

tiny_contribution proportional_to $C_{\gamma Z}$. For_$m_a>2m_\mu$, the_gray_regions are_excluded_by a dark-photon_search in the_$e^+ e^-\to\mu^+\mu^-+\mu^+\mu^-$ channel performed_by BaBar [@TheBABAR:2016rlg].](mu1MeVand10GeV){width="88.00000%"} In_our_numerical analysis, we will assume that the contribution of $K_{a_\mu}(\mu)$ is subleading at the_high_scale $\mu=\Lambda$._If_the_Wilson coefficients $c_{\mu\mu}$ and $C_{\gamma\gamma}$_are of similar magnitude, the_logarithmically enhanced_contribution is the parametrically largest one-loop correction. It_gives_a positive shift_of $a_\mu$ provided the product $c_{\mu\mu}\,C_{\gamma\gamma}$ is negative. The_correction proportional to $C_{\gamma Z}$ is_suppressed by $(1-4s_w^2)$_and_hence_is numerically subdominant. Note_also that the contribution proportional to_$(c_{\mu\mu})^2$ is suppressed in the limit_where $m_a^2\gg m_\mu^2$, while the remaining terms_remain unsuppressed. Figure \[fig:g-2\] shows the regions in_the parameter space of the_couplings $c_{\mu\mu}$_and $C_{\gamma\gamma}$ in which the_experimental value of_the muon_anomalous magnetic moment_can be explained in terms of_the ALP-induced loop_corrections shown in Figure \[fig:amugraphs\], without invoking_a_large contribution from_the_unknown_short-distance coefficient_$K_{a_\mu}(\Lambda)$. There is_a_weak dependence_on_the ALP mass, such that the_allowed_parameter space increases for $m_a^2\gg m_\mu^2$. Interestingly,_we find that an_explanation_of the anomaly is_possible without much tuning as_long as one coefficients is of_order $\Lambda/\mbox{TeV}$,_while the_other one can be of similar order or larger. Since $c_{\mu\mu}$_enters observables always in combination with_$m_\mu$, it is less_constrained by_perturbativity_than $C_{\gamma\gamma}$. ![\[fig:BaBargraphs\] Tree-level_Feynman_diagrams contributing_to the process $e^+ e^-\to\mu^+\mu^- a$.](BaBar_graphs){width="95.00000%"}

\approx 410$ MeV, can be obtained in this technique at $\alpha_s^{coul}(\bar c c) = 0.6$. The value $f_{B_c}=385$ MeV, as it is predicted in QCD sum rules [@fbc], gives $\alpha_s^{coul}(\bar b c) = 0.45$. We present the results of the Coulomb enhancement for the form factors in Table \[form\]. (140,100) (0,-120)[ ]{} (0,60)[$F_{0}^A(0)$, GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} (140,100) (0,-120)[ ]{} (0,60)[$F_{0}^A(0)$, GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} The result after the introduction of the Coulomb correction is shown in Fig. \[f3mom\]. Such large corrections to the form factors should not lead to a confusion, as they are resulted from the fare account of the Coulomb corrections both for bare quark loop diagram and meson coupling constant. In the scheme of the Borel transformation we find a strong dependence on the thresholds of continuum contribution. We think that this dependence reflects the influence of contributions coming from the excited states. So, the choice of $k_{th}$ values in the same region as in the scheme of spectral density moments results in the form factors, which are approximately 50% greater than the predictions in the moments scheme, where the higher excitations numerically are not essential. In this case we can explore the ideology of finite energy sum rules [@fesr], wherein the choice of interval for the quark-hadron duality, expressed by means of sum rules, allows one to isolate the contribution of basic states only. So, if we put $$\begin{aligned} k_{th}(\bar b c) & = & 1.2\; {\rm GeV},\nonumber \\ k_{th}(\bar c c) & = & 0.9\; {\rm GeV}, \end{aligned}$$ then the region of the lowest bound states is taken into account in both channels of initial and final states, and the Borel transform scheme leads to the results, which are very close to those of moment scheme. The dependence of calculated values on the Borel parameters is presented in Figs. \[f4bor\] and \[f5bor\], in the bare and Coulomb approximations, respectively.

\approx 410 $ MeV, can be obtained in this technique at $ \alpha_s^{coul}(\bar c c) = 0.6$. The value $ f_{B_c}=385 $ MeV, as it is predict in QCD kernel rules [ @fbc ], gives $ \alpha_s^{coul}(\bar b c) = 0.45$. We stage the results of the Coulomb enhancement for the form agent in Table \[form\ ]. (140,100) (0,-120) [ ] { } (0,60)[$F_{0}^A(0)$, GeV ] { } (30,10)[$n_{bc}$ ] { } (80,10)[$m_{cc}$ ] { } (140,100) (0,-120) [ ] { } (0,60)[$F_{0}^A(0)$, GeV ] { } (30,10)[$n_{bc}$ ] { } (80,10)[$m_{cc}$ ] { } The solution after the introduction of the Coulomb correction is shown in Fig. \[f3mom\ ]. Such large correction to the form factors should not head to a confusion, as they are resulted from the fare account of the Coulomb corrections both for bare quark loop diagram and meson coupling changeless. In the scheme of the Borel transformation we find a strong addiction on the thresholds of continuum contribution. We think that this addiction reflects the influence of contribution coming from the excited states. So, the choice of $ k_{th}$ value in the same region as in the scheme of spectral density consequence resultant role in the form factors, which are approximately 50% greater than the predictions in the moments scheme, where the higher excitations numerically are not all-important. In this case we can explore the political orientation of finite department of energy sum rules [ @fesr ], wherein the choice of interval for the quark - hadron dichotomy, expressed by mean of sum rules, allows one to sequester the contribution of basic states only. So, if we put $ $ \begin{aligned } k_{th}(\bar b c) & = & 1.2\; { \rm GeV},\nonumber \\ k_{th}(\bar c c) & = & 0.9\; { \rm GeV }, \end{aligned}$$ then the region of the lowest bounce states is taken into account in both channels of initial and final states, and the Borel transform dodge leads to the results, which are very close to those of moment scheme. The addiction of calculated values on the Borel parameters is presented in Figs. \[f4bor\ ] and \[f5bor\ ], in the bare and Coulomb approximations, respectively.

\appgox 410$ MeV, can be obtained in this technique at $\elpha_s^{ckul}(\bar c c) = 0.6$. The value $f_{B_c}=385$ MeV, as it iw preeicted in QCD sum ruler [@fbc], givvs $\alpha_s^{xoul}(\uar b c) = 0.45$. We present the resulfd of vhe Coulomb enhsncement fmr the form fawturd in Table \[form\]. (140,100) (0,-120)[ ]{} (0,60)[$F_{0}^A(0)$, GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} (140,100) (0,-120)[ ]{} (0,60)[$S_{0}^A(0)$, GeV]{} (30,10)[$m_{bf}$]{} (80,10)[$m_{cc}$]{} The resulj afttr ehe jntroduction of the Coulomb correcfion is shown in Fig. \[f3mom\]. Such large correctiojs tl the form factors should not leaq to a confuskon, as thej are resuljed from the fare account of the Coulpmb correcjnins toth for bace quagk loop diagram and mason coipling constanb. In tie sxheme of the Borel trensformation we find a strong dzpendence on the threwhilds mf cmntivyum cohtcibhtion. Ae vhink that fhis dependwnce reflects the imfjlrnce of contdibutijnf coming from the excited states. So, the chkice of $k_{th}$ values in tye same region as in jhe scheme of spectral density moments results in the form xactocs, whngm ard aoproximately 50% greater than the predictions in fht mpments scheme, chere the higher edcojations numerizally cde not essential. In hhis cafe we can expljre yhe ideology of finite enerty sum rules [@desr], wherein the ckoice of intzrval gor tne quark-hadron duality, zxpresaed by meand of sum djles, allows one go psoldte the contribution of bafic statew onky. So, iw we put $$\bqgin{alignef} k_{th}(\bar b c) & = & 1.2\; {\rm GeV},\nojumbet \\ k_{th}(\bdr c c) & = & 0.9\; {\rm GeV}, \end{aligned}$$ then the regmpn of the loeevt tound stctes ix taken into wccount in botk channejs of initial ahd finan states, anq the Borel tswnsform scheke leads to rhe eesults, dhich are very close to those of noment scheme. The vepevsence of calculcued values on the Nordl [agamxters hs presented in Wigr. \[f4bor\] and \[f5bor\], ik tfe bsre and Coulomb apprmximztions, respectivelu.

\approx 410$ MeV, can be obtained in at c c) 0.6$. The value predicted QCD sum rules gives $\alpha_s^{coul}(\bar b = 0.45$. We present the results the Coulomb enhancement for the form factors in Table \[form\]. (140,100) (0,-120)[ ]{} GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} (140,100) (0,-120)[ ]{} (0,60)[$F_{0}^A(0)$, GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} The result after introduction the correction shown in Fig. \[f3mom\]. Such large corrections to the form factors should not lead to a as they are resulted from the fare account the Coulomb corrections both bare quark loop diagram and coupling In the of Borel we find a dependence on the thresholds of continuum contribution. We think that this dependence reflects the influence of contributions from the So, the of values the same region the scheme of spectral density moments form factors, which are approximately 50% greater than predictions in moments scheme, where the higher excitations are not essential. In this case we can the ideology of finite energy sum rules [@fesr], wherein the choice of interval for the expressed by means of rules, allows one isolate contribution basic only. So, we put $$\begin{aligned} k_{th}(\bar b c) & = & 1.2\; {\rm \\ k_{th}(\bar c c) & = & 0.9\; {\rm GeV}, the of the lowest states is taken into in channels of initial and and Borel to results, are very close to of moment scheme. The dependence calculated values on the Figs. \[f4bor\] and \[f5bor\], in the bare and approximations, respectively.

\approx 410$ MeV, can be obtained in tHis techniqUe at $\aLphA_s^{cOuL}(\bar C c) = 0.6$. ThE value $f_{B_c}=385$ MeV, as IT is pRedicted in QCD sum rules [@fBc], givEs $\ALpha_S^{CoUl}(\bar B c) = 0.45$. We preSEnT THe rEsUlTs oF tHE COulomB enHancemeNt for the foRm fAcTors in Table \[fORm\]. (140,100) (0,-120)[ ]{} (0,60)[$f_{0}^A(0)$, GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cC}$]{} (140,100) (0,-120)[ ]{} (0,60)[$F_{0}^A(0)$, geV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} The ResUlt aftEr The INtrodUctIon of The CouLOmb corRection is ShOWn in FiG. \[F3mom\]. SucH LArGe coRrections to the forM FaCTors should not lEad to a CoNFuSIOn, aS thEy are resulTeD from THe fare aCCoUNT Of tHE Coulomb correCtions both fOR baRe quarK lOop DIagram And meSoN CouPling constaNt. In The scheme Of the BORel tranSFormatiOn we fiNd a StrOng dEPeNdEncE oN The THrEshOLds Of continUuM cOntriButiON. wE ThinK thAt thIs depEndence reflecTs tHe inFLueNce of ContrIbutIoNs comIng froM the eXcIted states. So, the ChoiCe of $k_{th}$ vaLueS iN thE sAme reGIon as iN thE scHeme of sPectral DEnsItY MOMeNts results in the forM fACToRs, which aRe apprOXiMaTEly 50% greatEr ThaN the PREdictIons IN tHe momentS schemE, WhErE the higHeR excitAtIonS nuMericALly aRe not eSsential. in thiS Case we can exploRE the ideology oF FiNITe ENergY suM rules [@fesr], wHereIN the ChoiCE oF inTErval For thE qUArK-Hadron duality, expresSeD by meaNs of sUm rules, allows One to isolaTE THe contriButiON oF Basic states onlY. So, if We put $$\begin{ALigned} k_{tH}(\bar b C) & = & 1.2\; {\rm GeV},\noNumber \\ k_{th}(\BAR c c) & = & 0.9\; {\rm GeV}, \End{AliGneD}$$ thEN ThE region of the lOWEst bOuNd stateS is Taken inTo aCcoUnt In bOtH channels Of initiaL aNd FiNaL stAtes, aND the BoreL tRanSfOrm SchemE Leads tO the rEsulTs, WhICh aRe very cLOsE TO thoSe Of MomeNt sChEme. ThE depENdeNce of caLculated vAluES on tHe boRel paraMeters is preseNtEd in Figs. \[f4bOr\] And \[F5bor\], in THE bare and coulomb approximations, reSPectiveLy.

\approx 410$ MeV, can be o btained in this te chn iq ue a t $\ alpha_s^{coul} ( \bar c c) = 0.6$. The valu e $f_ {B _ c}=3 8 5$ MeV, as iti sp r edi ct ed in Q C Dsum r ule s [@fbc ], gives $ \al ph a_s^{coul}(\ b ar b c) = 0 .45 $. We presen t t he res ul tso f the Co ulomb enhan c ementfor the f or m facto r s in Ta b l e\[fo rm\]. (140,100)( 0, - 120)[ ]{} (0,6 0)[$F_ {0 } ^A ( 0 )$, Ge V]{} (30,1 0) [$n_{ b c}$]{}( 80 , 1 0 )[$ m _{cc}$]{} (1 40,100) (0, - 120 )[ ]{} ( 0,6 0 )[$F_{ 0}^A( 0) $ , G eV]{} (30,1 0)[$ n_{bc}$]{ } (80, 1 0)[$m_{ c c}$]{} The r esu ltafte r t he in tr o duc t io n o f th e Coulom bco rrect ioni s s hown in Fig . \[f 3mom\]. Suchlar ge c o rre ction s tothefo rm fa ctorsshoul dnot lead to a c onfu sion, asthe yare r esult e d from th e f are acc ount of the C o u l om b corrections both f o r b are quar k loop di ag r am and m es oncoup l i ng co nsta n t. In the schem e o fthe Bor el trans fo rma tio n wef inda stro ng depen dence on the thresho l ds of continu u mc o nt r ibut ion . We thinkthat this dep e nd enc e refl ectsth e i n fluence of contribu ti ons co mingfrom the exci ted states . S o, the c hoic e o f $k_{th}$ valu es in the samer egion as in t he schem e of spec t r al densi tymom ent s r e s ul ts in the for m fact or s, whic h a re appr oxi mat ely 50 %greater t han thepr ed ic ti ons in t h e moment ssch em e,where the hi gherexci ta ti o nsnumeric a ll y areno tesse nti al . Inthis cas e we ca n explore th e ide ol og y of fi nite energy s um rules [@f es r], where i n the cho ice of interval for the quark-h adr on du alit y, expres sed by me ans of sum rules , all ow s o n e to i s o la teth e contribu t i onof ba si c st ates on ly. So, if we put$ $\b egin{aligned} k_ {th} ( \ ba r b c) & = & 1.2 \ ; {\rm GeV},\non umber \\ k _{ t h} (\bar c c) & = & 0.9\;{\rm Ge V}, \ e nd{alig ned}$$ th en the re gi on o f the lowest bo und stat es is tak e n int o a ccoun t i n both c han nelsof ini t ial andfinalst ates,and t he Borel t ransform scheme leads t o theresul ts, which ar e v e ryclose tothos e of momen t s che me. T hed epend ence of ca l culat ed v a lues on t h eBor e l p arameters i s p res ented in Figs.\[f4 bor\] and \[f5bor \ ], in the bare and C oul omb appr ox imations, resp ect iv e l y.

\approx 410$_MeV, can_be obtained in this_technique at_$\alpha_s^{coul}(\bar_c c)_=_0.6$. The value_$f_{B_c}=385$ MeV, as_it is predicted in_QCD sum rules_[@fbc],_gives $\alpha_s^{coul}(\bar b c) = 0.45$. We present the results of the Coulomb enhancement_for_the form_factors_in_Table \[form\]. (140,100) (0,-120)[ ]{} (0,60)[$F_{0}^A(0)$,_GeV]{} (30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} (140,100) (0,-120)[ ]{}_(0,60)[$F_{0}^A(0)$, GeV]{}_(30,10)[$n_{bc}$]{} (80,10)[$m_{cc}$]{} The result after the introduction of the_Coulomb_correction is shown_in Fig. \[f3mom\]. Such large corrections to the form_factors should not lead to a_confusion, as they_are_resulted_from the fare account_of the Coulomb corrections both for_bare quark loop diagram and meson_coupling constant. In the scheme of the Borel_transformation we find a strong dependence_on the thresholds of continuum_contribution. We_think that this dependence reflects_the influence of_contributions coming_from the excited_states. So, the choice of $k_{th}$_values in the_same region as in the scheme_of_spectral density moments_results_in_the form_factors, which are_approximately_50% greater_than_the predictions in the moments scheme,_where_the higher excitations numerically are not essential._In this case we_can_explore the ideology of_finite energy sum rules [@fesr],_wherein the choice of interval for_the quark-hadron_duality, expressed_by means of sum rules, allows one to isolate the contribution_of basic states only. So, if_we put $$\begin{aligned} k_{th}(\bar b_c) &_=_& 1.2\; {\rm_GeV},\nonumber_\\ k_{th}(\bar c_c) & = & 0.9\; {\rm GeV},_\end{aligned}$$ then_the region of the lowest bound_states is taken into_account_in both channels of initial and_final states, and the Borel transform_scheme leads to the results,_which_are_very close to those of_moment scheme. The dependence of calculated_values on the_Borel parameters is presented in Figs. \[f4bor\]_and_\[f5bor\], in the bare and Coulomb_approximations,_respectively.

u (x,z)$ converges to $0$ as $z \rightarrow 0$ in the disk $\{|1-z|<1 \}$; thus, the function $\kappa (u)$ approaches $0$. Our assumptions on the offspring distribution imply that $$\begin{aligned} h (z)&=\sigma^{2}z^{2}+\sum_{j=3}^{\infty } h_{j}z^{j} \quad \text{and hence}\\ \kappa (z)&= \frac{2}{3}\sigma^{2}z^{3} +\sum_{j=4}^{\infty } k_{j}z^{j};\end{aligned}$$ consequently, $$\frac{1}{\sqrt{\kappa (z)}}= Kz^{-3/2} (1+R (z))$$ where ${K=\sqrt{3/2\sigma^{2}}}$ and $R (z)$ is an analytic function in some neighborhood of $z=0$ and satisfies $R (0)=0$. Now the integral equation and the implicit function theorem imply that, for fixed $x>0$, the function $u=u (x,z)$ satisfies the differential equation $$\frac{du}{dz}=\frac{\sqrt{\kappa (u)}}{\sqrt{\kappa (z)}}=\frac{u^{3/2}}{z^{3/2}}\frac{(1+R (z))}{(1+R (u))}$$ for all $z$ in a domain $\{|1-z|<1 \}\cap \{|z|<\delta \}$. Using the analyticity of $R$, we conclude that $$u^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}u^{k} \right) = z^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}z^{k} \right) +C,$$ where $C$ is a constant of integration. Squaring both sides exhibits $u$ as a meromorphic function of $\sqrt{z}$. This shows that $u$ has an analytic continuation to a slit disk $\{|z|<\delta \}\setminus (-\delta,0]$, and so it follows, by the relations $u (x,z)=1-\varphi (x,1-z)$ and equation, that $\varphi (x,z)$ and $H (x,z)$ admit analytic continuations

u (x, z)$ converges to $ 0 $ as $ z \rightarrow 0 $ in the disk $ \{|1 - z|<1 \}$; thus, the affair $ \kappa (u)$ set about $ 0$. Our assumptions on the offspring distribution imply that $ $ \begin{aligned } henry (z)&=\sigma^{2}z^{2}+\sum_{j=3}^{\infty } h_{j}z^{j } \quad \text{and hence}\\ \kappa (z)&= \frac{2}{3}\sigma^{2}z^{3 } + \sum_{j=4}^{\infty } k_{j}z^{j};\end{aligned}$$ consequently, $ $ \frac{1}{\sqrt{\kappa (z)}}= Kz^{-3/2 } (1+R (z))$$ where $ { K=\sqrt{3/2\sigma^{2}}}$ and $ R (z)$ is an analytic function in some region of $ z=0 $ and satisfies $ gas constant (0)=0$. Now the integral equation and the implicit function theorem incriminate that, for fixed $ x>0 $, the function $ uranium = uracil (x, z)$ satisfies the differential equation $ $ \frac{du}{dz}=\frac{\sqrt{\kappa (u)}}{\sqrt{\kappa (z)}}=\frac{u^{3/2}}{z^{3/2}}\frac{(1+R (z))}{(1+R (u))}$$ for all $ z$ in a domain $ \{|1 - z|<1 \}\cap \{|z|<\delta \}$. Using the analyticity of $ R$, we reason that $ $ u^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}u^{k } \right) = z^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}z^{k } \right) + C,$$ where $ C$ is a constant of integration. Squaring both sides expose $ u$ as a meromorphic function of $ \sqrt{z}$. This shows that $ u$ hold an analytic sequel to a slit disk $ \{|z|<\delta \}\setminus (-\delta,0]$, and indeed it follows, by the relations $ u (x, z)=1-\varphi (x,1 - z)$ and equation, that $ \varphi (ten, z)$ and $ H (x, z)$ admit analytic continuations

u (x,x)$ converges to $0$ as $z \rigmtarrow 0$ in the disk $\{|1-z|<1 \}$; thus, the funztion $\kappa (u)$ approaches $0$. Ouc aswumptuons on the offspring aistributpon imply thau $$\begin{aligned} h (z)&=\sigma^{2}z^{2}+\sum_{j=3}^{\jkfty } i_{j}z^{j} \quad \text{akd hence}\\ \kappa (z)&= \frac{2}{3}\shgoa^{2}v^{3} +\sum_{j=4}^{\infty } k_{j}z^{j};\end{aligned}$$ consezuently, $$\fgac{1}{\sqrt{\kappa (z)}}= Kz^{-3/2} (1+G (s))$$ whsge ${K=\sqrt{3/2\sigma^{2}}}$ and $R (z)$ is an anamytic flnction in some nrighborhood of $z=0$ and satisvies $R (0)=0$. Now the integrwl equation and rhe implicit function uhzorem imply that, for fixed $x>0$, the function $j=u (x,z)$ satisfies tye fhfferential equanion $$\frac{du}{dz}=\ngac{\sqrt{\nappa (u)}}{\xqrt{\kappa (z)}}=\nrac{u^{3/2}}{v^{3/2}}\frax{(1+R (z))}{(1+R (u))}$$ for all $z$ mn a domain $\{|1-z|<1 \}\cap \{|z|<\dglta \}$. Using tke analyticity of $R$, ww xonclode tvat $$j^{-1/2}\oefg(1+\suj_{k=1}^{\mnffy}b_{k}u^{k} \rijht) = z^{-1/2}\meft(1+\sum_{k=1}^{\infry}b_{k}z^{k} \right) +C,$$ whert $C$ px a constant of ineedration. Squaring both sides exhibits $u$ ds z meromorphic function if $\sqrt{z}$. This shows tjat $u$ has an analytic continuation to a slit disk $\{|z|<\delta \}\satminns (-\aelua,0]$, and ri lt follows, by the relations $u (x,z)=1-\varphi (x,1-z)$ and sqianion, that $\varphi (w,z)$ and $H (x,z)$ admit snwlujic continuatiuns

u (x,z)$ converges to $0$ as $z in disk $\{|1-z|<1 thus, the function assumptions the offspring distribution that $$\begin{aligned} h } h_{j}z^{j} \quad \text{and hence}\\ \kappa \frac{2}{3}\sigma^{2}z^{3} +\sum_{j=4}^{\infty } k_{j}z^{j};\end{aligned}$$ consequently, $$\frac{1}{\sqrt{\kappa (z)}}= Kz^{-3/2} (1+R (z))$$ where ${K=\sqrt{3/2\sigma^{2}}}$ and (z)$ is an analytic function in some neighborhood of $z=0$ and satisfies $R Now integral and implicit function theorem imply that, for fixed $x>0$, the function $u=u (x,z)$ satisfies the differential equation (u)}}{\sqrt{\kappa (z)}}=\frac{u^{3/2}}{z^{3/2}}\frac{(1+R (z))}{(1+R (u))}$$ for all $z$ in domain $\{|1-z|<1 \}\cap \{|z|<\delta Using the analyticity of $R$, conclude $$u^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}u^{k} \right) z^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}z^{k} +C,$$ $C$ is a of integration. Squaring both sides exhibits $u$ as a meromorphic function of $\sqrt{z}$. This shows that $u$ an analytic a slit $\{|z|<\delta (-\delta,0]$, so it follows, relations $u (x,z)=1-\varphi (x,1-z)$ and equation, and $H (x,z)$ admit analytic continuations

u (x,z)$ converges to $0$ as $z \rightarrOw 0$ in the disK $\{|1-z|<1 \}$; thuS, thE fuNcTion $\KappA (u)$ approaches $0$. OuR AssuMptions on the offspring dIstriBuTIon iMPlY that $$\Begin{alIGnED} H (z)&=\sIgMa^{2}Z^{2}+\suM_{j=3}^{\INfTy } h_{j}z^{J} \quAd \text{aNd hence}\\ \kapPa (z)&= \FrAc{2}{3}\sigma^{2}z^{3} +\sum_{j=4}^{\INfTy } k_{j}z^{j};\end{aLigNed}$$ consequenTly, $$\Frac{1}{\sqRt{\KapPA (z)}}= Kz^{-3/2} (1+R (Z))$$ whEre ${K=\sQrt{3/2\sigMA^{2}}}$ and $R (z)$ Is an analyTiC FunctiON in some NEIgHborHood of $z=0$ and satisfiES $R (0)=0$. nOw the integral eQuatioN aND tHE ImpLicIt function ThEorem IMply thaT, FoR FIXed $X>0$, The function $u=u (X,z)$ satisfies THe dIffereNtIal EQuatioN $$\frac{Du}{DZ}=\frAc{\sqrt{\kappa (U)}}{\sqrT{\kappa (z)}}=\frAc{u^{3/2}}{z^{3/2}}\frAC{(1+R (z))}{(1+R (u))}$$ foR All $z$ in a Domain $\{|1-Z|<1 \}\caP \{|z|<\dElta \}$. uSiNg The AnALytICiTy oF $r$, we Conclude ThAt $$U^{-1/2}\left(1+\Sum_{k=1}^{\INFTY}b_{k}u^{K} \riGht) = z^{-1/2}\Left(1+\sUm_{k=1}^{\infty}b_{k}z^{k} \rIghT) +C,$$ whERe $C$ Is a coNstanT of iNtEgratIon. SquAring BoTh sides exhibits $U$ as a MeromorphIc fUnCtiOn Of $\sqrT{Z}$. This sHowS thAt $u$ has aN analytIC coNtINUAtIon to a slit disk $\{|z|<\delTa \}\SETmInus (-\deltA,0]$, and so IT fOlLOws, by the ReLatIons $U (X,Z)=1-\varpHi (x,1-z)$ ANd Equation, That $\vaRPhI (x,Z)$ and $H (x,z)$ AdMit anaLyTic ConTinuaTIons

u (x,z)$ converges to $0$as $z \rig htarr ow 0$ i n th e di sk $\{|1-z|<1\ }$;thus, the function $\k appa(u ) $ ap p ro aches $0$. O u ra s sum pt io nson th e off spr ing dis tributionimp ly that $$\beg i n{ aligned} h (z)&=\sigm a^{ 2}z^{2 }+ \su m _{j=3 }^{ \inft y } h_ { j}z^{j } \quad \ te x t{andh ence}\\ \ka ppa (z)&= \frac{2 } {3 } \sigma^{2}z^{3 } +\su m_ { j= 4 } ^{\ inf ty } k _{ j}z^{ j };\end{ a li g n e d}$ $ consequently , $$\frac{1 } {\s qrt{\k ap pa( z)}}=Kz^{- 3/ 2 } ( 1+R (z))$$wher e ${K=\sq rt{3/2 \ sigma^{ 2 }}}$ an d $R ( z)$ is ana na ly tic f u nct i on in som e neighb or ho od of $z= 0 $ a nd s ati sfie s $R(0)=0$. Now t heinte g ral equa tionandth e imp licitfunct io n theorem imply tha t, for fi xed $ x>0 $, thef unctio n $ u=u (x,z)$ satisf i esth e d if ferential equation $ $ \ fr ac{du}{d z}=\fr a c{ \s q rt{\kapp a(u) }}{\ s q rt{\k appa (z)}}= \frac{ u ^{ 3/ 2}}{z^{ 3/ 2}}\fr ac {(1 +R(z))} { (1+R ( u))}$$ f or al l $z$ in a doma i n $\{|1-z|<1\ }\ c a p\ {|z| <\d elta \}$. U sing theanal y ti cit y of $ R$, w ec on c lude that $$u^{-1/2 }\ left(1 +\sum _{k=1}^{\inft y}b_{k}u^{ k } \right) = z^{-1/2}\left( 1+\su m_{k=1}^{\ i nfty}b_{ k}z^{ k} \righ t) +C,$$w h ere $C$isa c ons tan t of integration. S quar in g bothsid es exhi bit s $ u$asameromorph ic funct io nof $ \sq rt{z} $ . This s ho wsth at$u$ h a s an a nalyt ic c on ti n uat ion toa s l i t di sk $ \{|z |<\ de lta \ }\se t min us (-\d elta,0]$, an d soit f ollows, by the relat io ns $u (x,z )= 1-\ varphi ( x,1-z)$and equation, that $\va r phi (x, z)$ and$H ( x,z)$ adm itanalyt icc ontinu ations

u (x,z)$_converges to_$0$ as $z \rightarrow_ 0$_in_the disk_$\{|1-z|<1_\}$; thus, the_function $\kappa (u)$_approaches $0$. Our assumptions_on the offspring_distribution_imply that $$\begin{aligned} h (z)&=\sigma^{2}z^{2}+\sum_{j=3}^{\infty } h_{j}z^{j} \quad \text{and hence}\\ __ \kappa_(z)&=_\frac{2}{3}\sigma^{2}z^{3}_+\sum_{j=4}^{\infty } _k_{j}z^{j};\end{aligned}$$ consequently, $$\frac{1}{\sqrt{\kappa (z)}}= Kz^{-3/2}_(1+R (z))$$_where ${K=\sqrt{3/2\sigma^{2}}}$ and $R (z)$ is an analytic_function_in some neighborhood_of $z=0$ and satisfies $R (0)=0$. Now the integral_equation and the implicit function theorem_imply that, for_fixed_$x>0$,_the function $u=u (x,z)$_satisfies the differential equation $$\frac{du}{dz}=\frac{\sqrt{\kappa (u)}}{\sqrt{\kappa _ (z)}}=\frac{u^{3/2}}{z^{3/2}}\frac{(1+R (z))}{(1+R _ (u))}$$ for all $z$ in_a domain $\{|1-z|<1 \}\cap \{|z|<\delta \}$. Using_the analyticity of $R$, we_conclude that_$$u^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}u^{k} \right) _= _z^{-1/2}\left(1+\sum_{k=1}^{\infty}b_{k}z^{k} \right)_+C,$$ where $C$_is a constant of integration. Squaring_both sides exhibits_$u$ as a meromorphic function of_$\sqrt{z}$._This shows that_$u$_has_an analytic_continuation to a_slit_disk $\{|z|<\delta_\}\setminus (-\delta,0]$,_and so it follows, by the_relations_$u (x,z)=1-\varphi (x,1-z)$ and equation, that $\varphi_(x,z)$ and $H (x,z)$_admit_analytic continuations

t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le (\delta T/N)^2}c_nn^{-1/2+i(t_1-t_2)}\right|^2,\end{aligned}$$ for some $c_n=N^{o(1)}$. By Lemma \[lem:coefficients\] and Corollary \[cor:reflection\] $$\begin{aligned} W&\ll I(\delta T,\cA)N^{o(1)}+\frac{\delta^2 T^2N^{o(1)}}{N^2}|\cA|+\sum_{\substack{t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le 4N^2/(\delta T)}n^{-1/2+i(t_1-t_2)}\right|^2,\end{aligned}$$ which by  implies that $$\begin{aligned} \label{eq:main4prelimstep} I(\delta T,\cA)\ll \frac{N^{2+o(1)}}{V^2}|\cA|+\frac{\delta^2 T^2 N^{4+o(1)}}{V^{8}}|\cA|+\frac{N^{6+o(1)}}{V^8}W_0,\end{aligned}$$ with $$\begin{aligned} W_0=\sum_{\substack{t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le 4N^2/(\delta T)}n^{-1/2+i(t_1-t_2)}\right|^2.\end{aligned}$$ This implies either $$\begin{aligned} |\cA|\ll \frac{1}{\delta}\frac{N^{2+o(1)}}{V^2}+\frac{\delta T^2 N^{4+o(1)}}{V^{8}},\end{aligned}$$ or $$\begin{aligned} \label{eq:main4case222b} I(\delta T,\cA)\ll \frac{N^{6+o(1)}}{V^8}W_0.\end{aligned}$$ We may suppose  since otherwise the result follows. Considering $W_0$, partitioning summation

t_1,t_2\in \cA \\ |t_1 - t_2|\le \delta T}}\left|\sum_{n\le (\delta T / N)^2}c_nn^{-1/2+i(t_1 - t_2)}\right|^2,\end{aligned}$$ for some $ c_n = N^{o(1)}$. By Lemma   \[lem: coefficients\ ] and Corollary   \[cor: reflection\ ] $ $ \begin{aligned } W&\ll I(\delta T,\cA)N^{o(1)}+\frac{\delta^2 T^2N^{o(1)}}{N^2}|\cA|+\sum_{\substack{t_1,t_2\in \cA \\ |t_1 - t_2|\le \delta T}}\left|\sum_{n\le 4N^2/(\delta T)}n^{-1/2+i(t_1 - t_2)}\right|^2,\end{aligned}$$ which by   implies that $ $ \begin{aligned } \label{eq: main4prelimstep } I(\delta T,\cA)\ll \frac{N^{2+o(1)}}{V^2}|\cA|+\frac{\delta^2 T^2 N^{4+o(1)}}{V^{8}}|\cA|+\frac{N^{6+o(1)}}{V^8}W_0,\end{aligned}$$ with $ $ \begin{aligned } W_0=\sum_{\substack{t_1,t_2\in \cA \\ |t_1 - t_2|\le \delta T}}\left|\sum_{n\le 4N^2/(\delta T)}n^{-1/2+i(t_1 - t_2)}\right|^2.\end{aligned}$$ This implies either $ $ \begin{aligned } |\cA|\ll \frac{1}{\delta}\frac{N^{2+o(1)}}{V^2}+\frac{\delta T^2 N^{4+o(1)}}{V^{8}},\end{aligned}$$ or $ $ \begin{aligned } \label{eq: main4case222b } I(\delta T,\cA)\ll \frac{N^{6+o(1)}}{V^8}W_0.\end{aligned}$$ We may suppose   since otherwise the resultant role take after. Considering $ W_0 $, partitioning summation

t_1,t_2\ij \cA \\ |t_1-t_2|\le \delta T}}\left|\suo_{n\le (\delta T/N)^2}c_nu^{-1/2+u(t_1-t_2)}\rigit|^2,\end{aljgned}$$ fof some $c_n=N^{o(1)}$. By Lemma \[lem:coeffmciebts\] abd Corollary \[cor:reflectkon\] $$\begin{wligned} W&\ol I(\velta T,\cA)N^{o(1)}+\frac{\dxmta^2 T^2N^{o(1)}}{K^2}|\eA|+\sum_{\ambstaek{v_1,t_2\in \cA \\ |t_1-t_2|\le \dglta T}}\left|\suk_{n\le 4N^2/(\delta T)}n^{-1/2+h(t_1-g_2)}\rnght|^2,\end{aligned}$$ which by  implies that $$\begin{akihned} \label{eq:maig4prekymstsi} I(\belta T,\cA)\ll \frac{N^{2+o(1)}}{V^2}|\cA|+\frac{\delta^2 F^2 N^{4+o(1)}}{V^{8}}|\cA|+\hrac{N^{6+o(1)}}{V^8}W_0,\end{aligmed}$$ with $$\begin{aligned} W_0=\sum_{\skbstwck{t_1,t_2\in \cA \\ |t_1-t_2|\le \dflta T}}\left|\soj_{n\lq 4N^2/(\delta T)}n^{-1/2+i(t_1-g_2)}\right|^2.\end{aligned}$$ This jmplies either $$\begin{aligned} |\cA|\ll \frac{1}{\belta}\frac{N^{2+o(1)}}{C^2}+\feac{\fglta T^2 N^{4+o(1)}}{V^{8}},\env{alignvd}$$ or $$\begin{allbned} \latel{eq:maon4case222b} I(\delta B,\cA)\ll \frqc{N^{6+o(1)}}{V^8}W_0.\end{aligned}$$ We mey suppose  since othgrwise the rzsult follows. Consideeibg $W_0$, kartidionkbg rumjavioh

t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le (\delta some By Lemma and Corollary \[cor:reflection\] \cA |t_1-t_2|\le \delta T}}\left|\sum_{n\le T)}n^{-1/2+i(t_1-t_2)}\right|^2,\end{aligned}$$ which by that $$\begin{aligned} \label{eq:main4prelimstep} I(\delta T,\cA)\ll \frac{N^{2+o(1)}}{V^2}|\cA|+\frac{\delta^2 N^{4+o(1)}}{V^{8}}|\cA|+\frac{N^{6+o(1)}}{V^8}W_0,\end{aligned}$$ with $$\begin{aligned} W_0=\sum_{\substack{t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le 4N^2/(\delta T)}n^{-1/2+i(t_1-t_2)}\right|^2.\end{aligned}$$ This implies $$\begin{aligned} |\cA|\ll \frac{1}{\delta}\frac{N^{2+o(1)}}{V^2}+\frac{\delta T^2 N^{4+o(1)}}{V^{8}},\end{aligned}$$ or $$\begin{aligned} \label{eq:main4case222b} I(\delta T,\cA)\ll \frac{N^{6+o(1)}}{V^8}W_0.\end{aligned}$$ We may since the follows. $W_0$, partitioning summation

t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\lE (\delta T/N)^2}c_nN^{-1/2+i(t_1-t_2)}\rIghT|^2,\enD{aLignEd}$$ foR some $c_n=N^{o(1)}$. By LemMA \[lem:Coefficients\] and CorollaRy \[cor:ReFLectIOn\] $$\Begin{Aligned} w&\Ll i(\DEltA T,\CA)n^{o(1)}+\fRaC{\DeLta^2 T^2N^{O(1)}}{N^2}|\ca|+\sum_{\subStack{t_1,t_2\in \ca \\ |t_1-t_2|\Le \Delta T}}\left|\suM_{N\lE 4N^2/(\delta T)}n^{-1/2+i(T_1-t_2)}\rIght|^2,\end{alignEd}$$ wHich by  ImPliES that $$\BegIn{aliGned} \laBEl{eq:maIn4prelimsTeP} i(\delta t,\CA)\ll \fraC{n^{2+O(1)}}{V^2}|\CA|+\frAc{\delta^2 T^2 N^{4+o(1)}}{V^{8}}|\cA|+\frac{n^{6+O(1)}}{V^8}w_0,\End{aligned}$$ with $$\Begin{aLiGNeD} w_0=\Sum_{\SubStack{t_1,t_2\in \ca \\ |t_1-T_2|\le \deLTa T}}\left|\SUm_{N\LE 4n^2/(\deLTa T)}n^{-1/2+i(t_1-t_2)}\right|^2.\eNd{aligned}$$ ThIS imPlies eItHer $$\BEgin{alIgned} |\CA|\LL \frAc{1}{\delta}\frac{n^{2+o(1)}}{V^2}+\fRac{\delta T^2 n^{4+o(1)}}{V^{8}},\end{ALigned}$$ oR $$\Begin{alIgned} \lAbeL{eq:Main4CAsE222b} i(\deLtA t,\cA)\LL \fRac{n^{6+O(1)}}{V^8}W_0.\End{alignEd}$$ we May suPposE  SINCe otHerWise The reSult follows. CoNsiDeriNG $W_0$, pArtitIoninG sumMaTion

t_1,t_2\in \cA \\ |t_1-t_2 |\le \delt a T}} \le ft| \s um_{ n\le (\delta T/N)^ 2 }c_n n^{-1/2+i(t_1-t_2)}\ri ght|^ 2, \ end{ a li gned} $$ fors om e $c_ n= N^ {o( 1) } $. By L emm a \[lem :coefficie nts \] and Corolla r y\[cor:refl ect ion\] $$\beg in{ aligne d} W& \ ll I( \de lta T ,\cA)N ^ {o(1)} +\frac{\d el t a^2 T^ 2 N^{o(1) } } {N ^2}| \cA|+\sum_{\subst a ck { t_1,t_2\in \cA \\ |t _1 - t_ 2 | \le \d elta T}}\l ef t|\su m _{n\le4 N^ 2 / ( \de l ta T)}n^{-1/2 +i(t_1-t_2) } \ri ght|^2 ,\ end { aligne d}$$wh i chby  implies tha t $$\begi n{alig n ed} \la b el{eq:m ain4pr eli mst ep}I (\ de lta T , \cA ) \l l \ f rac {N^{2+o( 1) }} {V^2} |\cA | + \ f rac{ \de lta^ 2 T^2 N^{4+o(1)}}{ V^{ 8}}| \ cA| +\fra c{N^{ 6+o( 1) }}{V^ 8}W_0, \end{ al igned}$$ with $ $\be gin{align ed} W _0= \s um_{\ s ubstac k{t _1, t_2\in\cA \\| t_1 -t _ 2 | \l e \delta T}}\left| \s u m _{ n\le 4N^ 2/(\de l ta T ) }n^{-1/2 +i (t_ 1-t_ 2 ) }\rig ht|^ 2 .\ end{alig ned}$$ Th is implie seither $ $\b egi n{ali g ned} |\cA| \ll \fra c{1}{ \ delta}\frac{N^ { 2+o(1)}}{V^2} + \f r a c{ \ delt a T ^2 N^{4+o(1 )}}{ V ^{8} },\e n d{ ali g ned}$ $ or$$ \ be g in{aligned} \label{ eq :main4 case2 22b} I(\delta T,\cA)\ll \ f rac{N^{6 +o(1 ) }} { V^8}W_0.\end{a ligne d}$$ We ma y suppose   sin ce other wise ther e sult fol low s.Con sid e r in g $W_0$, part i t ioni ng summat ion

t_1,t_2\in \cA_\\ |t_1-t_2|\le_\delta T}}\left|\sum_{n\le (\delta T/N)^2}c_nn^{-1/2+i(t_1-t_2)}\right|^2,\end{aligned}$$_for some_$c_n=N^{o(1)}$._By Lemma \[lem:coefficients\]_and_Corollary \[cor:reflection\] $$\begin{aligned} W&\ll I(\delta_T,\cA)N^{o(1)}+\frac{\delta^2 T^2N^{o(1)}}{N^2}|\cA|+\sum_{\substack{t_1,t_2\in \cA_\\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le_4N^2/(\delta T)}n^{-1/2+i(t_1-t_2)}\right|^2,\end{aligned}$$ which_by _implies that $$\begin{aligned} \label{eq:main4prelimstep} I(\delta T,\cA)\ll \frac{N^{2+o(1)}}{V^2}|\cA|+\frac{\delta^2 T^2 N^{4+o(1)}}{V^{8}}|\cA|+\frac{N^{6+o(1)}}{V^8}W_0,\end{aligned}$$ with $$\begin{aligned} W_0=\sum_{\substack{t_1,t_2\in \cA \\ |t_1-t_2|\le \delta T}}\left|\sum_{n\le_4N^2/(\delta_T)}n^{-1/2+i(t_1-t_2)}\right|^2.\end{aligned}$$ This_implies_either_$$\begin{aligned} |\cA|\ll \frac{1}{\delta}\frac{N^{2+o(1)}}{V^2}+\frac{\delta T^2 N^{4+o(1)}}{V^{8}},\end{aligned}$$ or_$$\begin{aligned} \label{eq:main4case222b} I(\delta T,\cA)\ll \frac{N^{6+o(1)}}{V^8}W_0.\end{aligned}$$ We may_suppose  since_otherwise the result follows. Considering $W_0$, partitioning summation

explain the tightness of the correlation. Thus, though we have explained one part of the radio – FIR correlation, more understanding is needed of the cosmic rays, radiation field heating the dust, and the processes discussed above before the exact reason for the correlation and its tightness can be fully comprehended. Conclusion ========== Most models which try to explain the remarkable correlation between radio and FIR emission rely upon the association of magnetic field strength and gas density. MHD simulations show that through the process of equipartition, the magnetic field and gas densities are coupled, with $$B \propto \sqrt{\bar{\rho_g}},$$ a relationship that is also indicated by observations within our own and other galaxies. This relationship provides a basis for the radio – FIR correlation by directly connecting one of the parameters of synchrotron emission with a parameter of FIR emission. While not explaining the relationship fully, it does put us a step closer to fully understanding this remarkable correlation. Acknowledgments {#acknowledgments.unnumbered} =============== The authors would like to thank the referees for several insightful comments and suggestions which improved the paper greatly. BG would like to acknowledge the assistance of the Alex Rodgers Traveling Scholarship and Edward Courbold Research Fund for this collaboration. MD wishes to acknowledge the support of the Australian National University and the Australian Research Council (ARC) under his ARC Australian Federation Fellowship, and also under ARC Discovery project DP0208445. Batchelor, G. 1950, Proc. R. Soc. London A, 201, 405 Beck, R. & Golla, G. 1988, [A& A]{}, 191, L9 Bettens, R. P. A., Brown, R. D., Cragg, D. M., Dickinson, C. J., & Godfrey, P. D. 1993, [MNRAS]{}, 263, 93 Berkhuijsen, E. M. & et al. 1997, The Physics of Galactic Halos, 155 Blackman, E. G., & Field G. B. 2000, ApJ, 534, 984 Cattaneo, F. & Hughes, D. W. 1996, Phys. Rev. E., 54, R4532 Condon, J. J. 1992, [ARA&A]{}, 30, 575 Cho, J. & Vishniac

explain the tightness of the correlation. Thus, though we have excuse one character of the radio – FIR correlation, more understanding is want of the cosmic rays, radiation playing field heat the dust, and the procedure discussed above before the exact cause for the correlation and its tightness can be in full embrace. Conclusion = = = = = = = = = = Most models which try to explain the remarkable correlation between radio receiver and FIR emission rely upon the association of charismatic field strength and gas density. MHD simulations prove that through the process of equipartition, the magnetic field and gas densities are pair, with $ $ B \propto \sqrt{\bar{\rho_g}},$$ a relationship that is also indicated by observations within our own and other galaxies. This relationship provides a basis for the radio – FIR correlation by directly connecting one of the parameters of synchrotron emission with a parameter of FIR discharge. While not explaining the relationship in full, it does place us a step closer to fully understand this noteworthy correlation. Acknowledgments { # acknowledgments.unnumbered } = = = = = = = = = = = = = = = The authors would wish to thank the referees for several insightful comments and suggestions which improved the paper greatly. BG would wish to acknowledge the assistance of the Alex Rodgers Traveling Scholarship and Edward Courbold Research Fund for this collaboration. MD wishes to acknowledge the support of the Australian National University and the Australian Research Council (ARC) under his ARC Australian Federation Fellowship, and also under ARC Discovery project DP0208445. Batchelor, G. 1950, Proc. R. Soc. London A, 201, 405 Beck, R.   & Golla, G.   1988, [ A & ampere ] { }, 191, L9 Bettens, R.   P.   A., Brown, R.   D., Cragg, D.   M., Dickinson, C.   J., & Godfrey, P.   D.   1993, [ MNRAS ] { }, 263, 93 Berkhuijsen, E.   M.   & et al.   1997, The Physics of Galactic Halos, 155 Blackman, E.   G., & Field G.   B. 2000, ApJ, 534, 984 Cattaneo, F. & Hughes, D.   W. 1996, Phys.   Rev.   E., 54, R4532 Condon, J. J. 1992, [ ARA&A ] { }, 30, 575 Cho, J.   & Vishniac

exolain the tightness of tme correlation. Thos, thougi we habe explakned one part of the radio – HIR xorreoation, more understandkng is nevded of tye ciwmic rays, czdiatiok fiemf hectmng the dust, anc the procasses discusseg xblve before the exact reason for the correlstlon and its tidhtntss can be fully comprehended. Conclusion ========== Moat modens which try yo explain the remarkable forrflation between rafio and FIR emifwion rely upun the association of jagnetic field strength and gas densnty. MHD simolztlmns show thet thrjugh the progvss of aquiparyition, the magketic fiwld and gas densities are coupled, with $$F \propto \vqxt{\bar{\rho_g}},$$ a relationsyip thaj is dlso undkcaued bg obsegvavions withih our own abd other galaxies. Tnif relationship proviqef a basis for the radio – FIR correlatiot bg directly connecting obe of the parameters lf synchrjtron emission with a parameter of FIR emission. Wvile iog erijxunlng the relationship fully, it does put us a sesp ckoser to fully understandonh yris remarkablg correlatjon. Acknowledgments {#acknowjedgmwnts.unnumferec} =============== The authors would like to rhank the reyerwes for several indightful coomenjs and suggestions which impruved the paper hreatly. BF would like to azknpwnedge tht assistance of thq Alex Rovgers Travelkng Xcholawship and Fdward Courbold Research Fund fmr this coplaboration. MD wishes to acknowledge the suppprd ox the Auftrallan National Unyversity and tke Austrclian Fesearch Ckuncil (ERC) under hif ARC Australhwn Federatioi Fellowsrip, qnd qlso unadr ARC Discovety projecn BP0208445. Batchelir, G. 1950, Proc. R. Soc. Lpnduh A, 201, 405 Beck, R. & Goloa, T. 1988, [A& A]{}, 191, L9 Bettens, R. P. X., Bwoan, R. Q., Cragg, D. M., Diwkinron, V. J., & Gudfrey, K. B. 1993, [MKRAR]{}, 263, 93 Brrkhuijsen, E. M. & et al. 1997, Dhe Lhysics of Galactiv Malos, 155 Blaxkman, E. G., & Field G. B. 2000, AlJ, 534, 984 Cattaneo, F. & Huhhes, V. W. 1996, Phbs. Rev. E., 54, R4532 Sondon, J. J. 1992, [ARA&A]{}, 30, 575 Cho, J. & Vishnizc

explain the tightness of the correlation. Thus, have one part the radio – needed the cosmic rays, field heating the and the processes discussed above before exact reason for the correlation and its tightness can be fully comprehended. Conclusion Most models which try to explain the remarkable correlation between radio and FIR rely the of field strength and gas density. MHD simulations show that through the process of equipartition, the magnetic and gas densities are coupled, with $$B \propto a relationship that is indicated by observations within our and galaxies. This provides basis the radio – correlation by directly connecting one of the parameters of synchrotron emission with a parameter of FIR emission. not explaining fully, it put a closer to fully remarkable correlation. Acknowledgments {#acknowledgments.unnumbered} =============== The to thank the referees for several insightful comments suggestions which the paper greatly. BG would like acknowledge the assistance of the Alex Rodgers Traveling and Edward Courbold Research Fund for this collaboration. MD wishes to acknowledge the support of National University and the Research Council (ARC) his Australian Fellowship, also under Discovery project DP0208445. Batchelor, G. 1950, Proc. R. Soc. London A, 405 Beck, R. & Golla, G. 1988, [A& A]{}, 191, R. A., Brown, R. Cragg, D. M., Dickinson, J., Godfrey, P. D. 1993, 93 E. al. The of Galactic Halos, 155 E. G., & Field G. 2000, ApJ, 534, 984 W. 1996, Phys. Rev. E., 54, R4532 Condon, J. 1992, [ARA&A]{}, 30, 575 Cho, J. Vishniac

explain the tightness of the cOrrelation. thus, tHouGh wE hAve eXplaIned one part of tHE radIo – FIR correlation, more unDerstAnDIng iS NeEded oF the cosMIc RAYs, rAdIaTioN fIElD heatIng The dust, And the procEssEs Discussed aboVE bEfore the exAct Reason for the CorRelatiOn And ITs tigHtnEss caN be fulLY comprEhended. CoNcLUsion ========== MOSt modelS WHiCh trY to explain the remaRKaBLe correlation bEtween RaDIo AND FIr emIssion rely UpOn the ASsociatIOn OF MAgnETic field strenGth and gas deNSitY. MHD siMuLatIOns shoW that ThROugH the process Of eqUipartitiOn, the mAGnetic fIEld and gAs densItiEs aRe coUPlEd, WitH $$B \PRopTO \sQrt{\BAr{\rHo_g}},$$ a relaTiOnShip tHat iS ALSO indIcaTed bY obseRvations withiN ouR own ANd oTher gAlaxiEs. ThIs RelatIonshiP provIdEs a basis for the rAdio – fIR correlAtiOn By dIrEctly COnnectIng One Of the paRameterS Of sYnCHROtRon emission with a paRaMETeR of FIR emIssion. wHiLe NOt explaiNiNg tHe reLATionsHip fULlY, it does pUt us a sTEp ClOser to fUlLy undeRsTanDinG this REmarKable cOrrelatiOn. AckNOwledgments {#ackNOwledgments.unNUmBEReD} =============== the aUthOrs would likE to tHAnk tHe reFErEes FOr sevEral iNsIGhTFul comments and suggeStIons whIch imProved the papeR greatly. BG WOULd like to AcknOWlEDge the assistanCe of tHe Alex RodgERs TravelIng ScHolarshiP and EdwarD cOurbold REseArcH FuNd fOR ThIs collaboratiON. mD wiShEs to ackNowLedge thE suPpoRt oF thE AUstralian national unIvErSiTy aNd the aUstraliaN REseArCh COunciL (aRC) undEr his aRC AUsTrALiaN FederaTIoN fElloWsHiP, and AlsO uNder ArC DiSCovEry projEct DP0208445. BatcHelOR, G. 1950, PrOc. r. SOc. LondoN A, 201, 405 Beck, R. & Golla, G. 1988, [a& A]{}, 191, l9 Bettens, R. P. a., BRowN, R. D., CraGG, d. M., DickinSon, C. J., & Godfrey, P. D. 1993, [MNRAS]{}, 263, 93 BerkHUijsen, E. m. & et Al. 1997, The physIcs of GalaCtiC Halos, 155 blaCKman, E. G., & field G. b. 2000, ApJ, 534, 984 CAtTanEO, f. & HughES, d. W. 1996, phyS. REv. E., 54, R4532 Condon, j. j. 1992, [aRA&a]{}, 30, 575 Cho, J. & viShniAc

explain the tightness ofthe correl ation . Thu s, tho ughwe have explai n ed o ne part of the radio – FIRco r rela t io n, mo re unde r st a n din gis ne ed e dof th e c osmic r ays, radia tio nfield heatin g t he dust, a ndthe processe s d iscuss ed ab o ve be for e the exact reason for theco r relati o n and i t s t ight ness can be fully co m prehended. Co nclusi on == = = === === Most mod el s whi c h try t o e x p l ain the remarkabl e correlati o n b etween r adi o and F IR em is s ion rely upontheassociati on ofm agnetic field s trengt h a ndgasd en si ty. M H D s i mu lat i ons show th at t hroug h th e p r oces s o f eq uipar tition, the m agn etic fie ld an d gas den si tiesare co upled ,with $$B \propt o \s qrt{\bar{ \rh o_ g}} ,$ $ a r e lation shi p t hat isalso in d ica te d b yobservations withi no u rown andotherg al ax i es. This r ela tion s h ip pr ovid e sa basisfor th e r ad io – FI Rcorrel at ion by dire c tlyconnec ting one of t h e parameters o f synchrotrone mi s s io n wit h a parameterof F I R em issi o n. Wh i le no t exp la i ni n g the relationshipfu lly, i t doe s put us a st ep closert o fully un ders t an d ing this remar kable correlati o n. Ackn owled gments { #acknowle d g ments.un num ber ed} == = = == ========= Th e auth or s would li ke to t han k t heref er ees for s everal i ns ig ht fu l c ommen t s and su gg est io nswhich improv ed th e pa pe rg rea tly. BG wo u l d li ke t o ac kno wl edgethea ssi stanceof the Al exR odge rs T ravelin g Scholarship a nd EdwardCo urb old Re s e arch Fun d for this collaboratio n . MD wi she s toackn owledge t hesuppor t o f the A ustral ian N at ion a l Univ e r si tyan d the Aust r a lia n Res ea rchCouncil (ARC) under his A R C A ustralian Fed era tion F el low s hi p , a nd als o under ARC Disco very proje ct DP 0208445. B atc he lor, G. 1950,Proc. R. Soc. London A , 201, 40 5 Bec k , R.  & Golla,G. 1988, [A& A]{} , 191, L9 Bet ten s, R.P.  A. , Bro wn, R. D., Crag g, D.M. , Dick inson ,C. J., & Godfrey, P. D. 1993, [ MNRAS] {}, 2 63, 93 Berk hui j sen , E. M. & etal. 1997,The Ph ysics of Galac ticH al os, 155 Blac k man, E. G . ,& F i e ld G. B. 2000 , A pJ, 534, 98 4 Catt aneo , F. & Hughes, D. W. 1996, Phys.  Rev . E., 54 , R45 32 Condon, J. J . 1 99 2 , [ARA&A] {} , 30, 575 Cho, J.&V ishni ac

explain_the tightness_of the correlation. Thus, though_we have_explained_one part_of_the radio –_FIR correlation, more_understanding is needed of_the cosmic rays,_radiation_field heating the dust, and the processes discussed above before the exact reason for_the_correlation and_its_tightness_can be fully comprehended. Conclusion ========== Most models_which try to explain the_remarkable correlation_between radio and FIR emission rely upon the_association_of magnetic field_strength and gas density. MHD simulations show that through_the process of equipartition, the magnetic_field and gas_densities_are_coupled, with $$B \propto_\sqrt{\bar{\rho_g}},$$ a relationship that is also_indicated by observations within our own_and other galaxies. This relationship provides a_basis for the radio – FIR_correlation by directly connecting one_of the_parameters of synchrotron emission with_a parameter of_FIR emission._While not explaining_the relationship fully, it does put_us a step_closer to fully understanding this remarkable_correlation. Acknowledgments_{#acknowledgments.unnumbered} =============== The authors would_like_to_thank the_referees for several_insightful_comments and_suggestions_which improved the paper greatly. BG_would_like to acknowledge the assistance of the_Alex Rodgers Traveling Scholarship_and_Edward Courbold Research Fund_for this collaboration. MD wishes_to acknowledge the support of the_Australian National_University and_the Australian Research Council (ARC) under his ARC Australian Federation Fellowship,_and also under ARC Discovery project_DP0208445. Batchelor, G. 1950, Proc._R. Soc._London_A, 201, 405 Beck,_R. &_Golla, G. 1988,_[A& A]{}, 191, L9 Bettens, R. P. A., Brown, R. D.,_Cragg, D. M.,_Dickinson, C. J., & Godfrey, P. D. 1993, [MNRAS]{},_263, 93 Berkhuijsen, E. M. & et_al. 1997,_The Physics of Galactic Halos, 155 Blackman,_E. G., & Field G. B. 2000, ApJ,_534, 984 Cattaneo, F. & Hughes,_D. W._1996,_Phys. Rev. E., 54, R4532 Condon, J. J._1992, [ARA&A]{}, 30, 575 Cho, J. & Vishniac

as a *contact force* rather than as a retarded at-a-distance force of the canonical interpretation. It is this particular consequence which leads necessarily to a changed view of the nature of the electromagnetic field and to a potentially deeper understanding of loss mechanisms in electrodynamics. Summary -------- By raising a question about the status of the electrostatic field as a pragmatic idealization, we are able easily to show that classical electromagnetism can be simply reformulated in terms of a (three-component) *relativistic* magnetic potential. This reformulation reveals, via a straightforward analysis, a relativistic symmetry principle according to which the relativistic magnetic potential is one half of a duality; the second half, a three-component potential also, has the property of mass and its presence has the effect of forcing a natural re-interpretation of the Lorentz force as a local action contact force. At the level of classical electrodynamics, one unavoidable consequence of this re-interpretation is a changed understanding of electrodynamic loss-mechanisms - the processes by which electrodynamic energy is converted, finally, to thermodynamic energy. Beyond the concerns of classical electrodynamics, it also becomes clear that there must be consequences for quantum electrodynamics also, since this latter theory subsumes the electrostatic field into its basic structure. Notation notes -------------- 1. We use the convention that $(x^{1},x^{2},x^{3})$ represent the spatial axes and $x^{4}\equiv ict$ represents the temporal one with a correspondingly consistent notation for the four-vector current, $J_{a}$, and the electromagnetic field tensor, $F_{ab}$; 2. The metric tensor is now consistently represented by the kronecker-delta, $\delta_{ab}$; 3. Although not strictly necessary when using an orthonormal basis, we make a strict distinction between covariant and contravariant objects in the conventional way; 4. We use bold-face to represent an ordinary three-vector; for example, the position vector $\textbf{x}\equiv(x^{1},x^{2},x^{3})$; 5. We denote a space-time coordinate as $x\equiv(x^{1},x^{2},x^{3},x^{4})$ or $x\equiv(\textbf{x},ict)$ according to convenience; 6. For convenience, we use the notation $\partial/\partial x^{a

as a * contact force * rather than as a retarded at - a - distance force of the basic rendition. It is this particular consequence which leads inevitably to a changed scene of the nature of the electromagnetic field and to a potentially deep sympathy of loss mechanisms in electrodynamics. Summary -------- By raising a question about the condition of the electrostatic field as a pragmatic idealization, we are able easily to show that authoritative electromagnetism can be simply reformulated in term of a (three - component) * relativistic * magnetic electric potential. This reformulation reveals, via a straightforward analysis, a relativistic isotropy principle according to which the relativistic magnetic potential is one half of a dichotomy; the second half, a three - part potential also, has the property of multitude and its presence has the effect of forcing a natural rhenium - interpretation of the Lorentz force as a local action contact force. At the level of classical electrodynamics, one unavoidable consequence of this re - interpretation is a deepen understanding of electrodynamic loss - mechanism - the summons by which electrodynamic energy is converted, finally, to thermodynamic energy. Beyond the concerns of classical electrodynamics, it also become clear that there must be consequences for quantum electrodynamics also, since this latter theory subsumes the electrostatic field into its basic social organization. Notation notes -------------- 1. We use the convention that $ (x^{1},x^{2},x^{3})$ represent the spatial axes and $ x^{4}\equiv ict$ represents the temporal one with a correspondingly reproducible notation for the four - vector current, $ J_{a}$, and the electromagnetic field tensor, $ F_{ab}$; 2. The metric tensor is now systematically represented by the kronecker - delta, $ \delta_{ab}$; 3. Although not strictly necessary when using an orthonormal basis, we make a strict distinction between covariant and contravariant objects in the conventional room; 4. We use bluff - face to represent an ordinary three - vector; for example, the position vector $ \textbf{x}\equiv(x^{1},x^{2},x^{3})$; 5. We announce a space - time coordinate as $ x\equiv(x^{1},x^{2},x^{3},x^{4})$ or $ x\equiv(\textbf{x},ict)$ according to convenience; 6. For appliance, we use the notation $ \partial/\partial x^{a

as a *contact force* rather uhan as a retardeb at-a-divtance force ow the canonical interpretatiln. It iw this particular consdquence wjich leaes ntcessarily to a ciznged vlzw of bhe nctnre of the elecjromagnetic xield and to a putzntially deeper understanding of losf mechamidms in electroqynakycs. Shmmary -------- By raising a question about the stetus of the elevtrostatic field as a praglatif idealization, we wre able eawily ro show that classical electromagngtism can be simply reformulated in tzrms of a (tyrwe-clkponent) *reletivisnic* magnetic iptentidl. This reformulation retealw, via a straightforwacd analysis, a relatidistic sykmztry principle accordunt to fhicv thd reuatjvmstjc magjetmc potentiam is one haof of a duality; the sqbpnd half, a tgree-coipjnent potential also, has the property ox mzss and its presence haw the effect of forcijg a natuwal re-interpretation of the Lorentz force as a lowal artkon cjvraft force. At the level of classical electrodynwjivs, one unavoidabje consequemcf ps this re-intetpretatnkn is a changed undegstandigg of electrodrnamoc loss-mechanisms - the procwsses by whibh eoectrodynamic eneryy is converced, fimally, to thermodynamic energv. Beyohd the concfrns of cmxssical electrodhnakiws, it also becomes clear trat there musc be conrequgnces fjr quantum elecbsodynamics also, sijce tkis ldtter theogy subsumes the electrostatic fmxld into its navic structuxe. Notabion notes -------------- 1. We tse the convenjion that $(x^{1},x^{2},x^{3})$ fepresent nhe spatiel axes and $v^{4}\equiv ict$ re[tesents the txmporal oge wuth q correroondingly consostent nonauion for tye four-vector currcnt, $J_{x}$, and the electrimatnetic field temsof, $F_{wb}$; 2. Vhe mqdric tensor hs nuw zpnsisgently reprcsevted by the kronecker-delda, $\dslta_{ab}$; 3. Although npt strictli necessawy when using an orthonormal badis, wx make a stticj distinction between covariant and contgavwriant objeces ik thg conventioual way; 4. We use bold-face to represent an ordinary three-vecror; for example, the kosition vectpr $\tevtbf{x}\equie(x^{1},x^{2},x^{3})$; 5. We denote a spqce-time coordinatt as $x\equiv(x^{1},x^{2},x^{3},x^{4})$ or $x\equjv(\texttf{x},ich)$ according to convenience; 6. For convenience, we use the notation $\partial/\paetial e^{a

as a *contact force* rather than as at-a-distance of the interpretation. It is necessarily a changed view the nature of electromagnetic field and to a potentially understanding of loss mechanisms in electrodynamics. Summary -------- By raising a question about status of the electrostatic field as a pragmatic idealization, we are able easily show classical can simply reformulated in terms of a (three-component) *relativistic* magnetic potential. This reformulation reveals, via a straightforward a relativistic symmetry principle according to which the magnetic potential is one of a duality; the second a potential also, the of and its presence the effect of forcing a natural re-interpretation of the Lorentz force as a local action contact force. the level electrodynamics, one consequence this is a changed electrodynamic loss-mechanisms - the processes by is converted, finally, to thermodynamic energy. Beyond the of classical it also becomes clear that there be consequences for quantum electrodynamics also, since this theory subsumes the electrostatic field into its basic structure. Notation notes -------------- 1. We use that $(x^{1},x^{2},x^{3})$ represent the axes and $x^{4}\equiv represents temporal with correspondingly consistent for the four-vector current, $J_{a}$, and the electromagnetic field tensor, $F_{ab}$; The metric tensor is now consistently represented by the kronecker-delta, Although strictly necessary when an orthonormal basis, we a distinction between covariant and in conventional use to an ordinary three-vector; for the position vector $\textbf{x}\equiv(x^{1},x^{2},x^{3})$; 5. denote a space-time coordinate to convenience; 6. For convenience, we use the $\partial/\partial x^{a

as a *contact force* rather than As a retardeD at-a-dIstAncE fOrce Of thE canonical inteRPretAtion. It is this particulaR consEqUEnce WHiCh leaDs necesSArILY to A cHaNgeD vIEw Of the NatUre of thE electromaGneTiC field and to a POtEntially deEpeR understandiNg oF loss mEcHanISms in EleCtrodYnamicS. summarY -------- By raisinG a QUestioN About thE STaTus oF the electrostatic FIeLD as a pragmatic iDealizAtIOn, WE Are AblE easily to sHoW that CLassicaL ElECTRomAGnetism can be sImply reformULatEd in teRmS of A (Three-cOmponEnT) *RelAtivistic* maGnetIc potentiAl. This REformulATion revEals, viA a sTraIghtFOrWaRd aNaLYsiS, A rElaTIviStic symmEtRy PrincIple ACCORdinG to WhicH the rElativistic maGneTic pOTenTial iS one hAlf oF a DualiTy; the sEcond HaLf, a three-componeNt poTential alSo, hAs The PrOpertY Of mass And Its PresencE has the EFfeCt OF FOrCing a natural re-inteRpRETaTion of thE LorenTZ fOrCE as a locaL aCtiOn coNTAct foRce. AT ThE level of ClassiCAl ElEctrodyNaMics, onE uNavOidAble cONseqUence oF this re-iNterpREtation is a chanGEd understandiNG oF ELeCTrodYnaMic loss-mechAnisMS - the ProcESsEs bY Which ElectRoDYnAMic energy is converteD, fInally, To theRmodynamic eneRgy. Beyond tHE COncerns oF claSSiCAl electrodynamIcs, it Also becomeS Clear thaT therE must be cOnsequencES For quantUm eLecTroDynAMIcS also, since thiS LAtteR tHeory suBsuMes the eLecTroStaTic FiEld into itS basic stRuCtUrE. NOtaTion nOTes -------------- 1. We use ThE coNvEntIon thAT $(x^{1},x^{2},x^{3})$ rePreseNt thE sPaTIal Axes and $X^{4}\EqUIV ict$ RePrEsenTs tHe TempoRal oNE wiTh a corrEspondingLy cONsisTeNt NotatioN for the four-veCtOr current, $J_{A}$, aNd tHe elecTROmagnetiC field tensor, $F_{ab}$; 2. The metriC Tensor iS noW consIsteNtly repreSenTed by tHe kROneckeR-delta, $\Delta_{Ab}$; 3. altHOUgh noT STrIctLy Necessary wHEN usIng an OrThonOrmal baSis, we make a strict diSTinCtion between cOvaRianT ANd ConTRaVAriAnT ObjECTs in the conventiOnal way; 4. We uSe BOlD-face to repREseNt An ordinAry threE-vectOR; for exaMple, the poSition vecToR $\texTBF{x}\eQuiv(x^{1},x^{2},x^{3})$; 5. We dEnote a spAce-time coORdinaTE aS $x\equIv(x^{1},X^{2},x^{3},x^{4})$ or $x\EqUiv(\TextbF{x},ict)$ aCCorDing tO conveNiEnce; 6. FoR convEnIence, we uSe the notation $\partial/\parTial x^{a

as a *contact force* rath er than as a re tar ded a t-a- dist ance force oft he c anonical interpretatio n. It i s thi s p artic ular co n se q u enc ewh ich l e ad s nec ess arily t o a change d v ie w of the nat u re of the el ect romagnetic f iel d andto ap otent ial ly de eper u n dersta nding oflo s s mech a nisms i n el ectr odynamics. Summa r y-------- By r aising a qu e s tio n a bout the s ta tus o f the el e ct r o s tat i c field as apragmatic i d eal izatio n, we are ab le ea si l y t o show that cla ssical el ectrom a gnetism can besimply re for mula t ed i n t er m s o f a (t h ree -compone nt )*rela tivi s t i c * ma gne ticpoten tial. This re for mula t ion reve als,viaastrai ghtfor wardan alysis, a relat ivis tic symme try p rin ci ple a c cordin g t o w hich th e relat i vis ti c m ag netic potential is o n e h alf of a duali t y; t h e second h alf , at h ree-c ompo n en t potent ial al s o, h as thepr operty o f m ass andi ts p resenc e has th e eff e ct of forcinga natural re-i n te r p re t atio n o f the Loren tz f o rceas a lo cal actio n con ta c tf orce. At the levelof class icalelectrodynami cs, one un a v o idable c onse q ue n ce of this re- inter pretationi s a chan ged u nderstan ding of e l e ctrodyna mic lo ss- mec h a ni sms - the pro c e sses b y which el ectrody nam icene rgy i s convert ed, fina ll y, t othe rmody n amic ene rg y.Be yon d the concer ns of cla ss ic a l e lectrod y na m i cs,it a lsobec om es cl eart hat theremust be c ons e quen ce sfor qua ntum electrod yn amics also ,sin ce thi s latter t heory subsumes the elec t rostati c f ieldinto its basi c s tructu re. Notat ion no tes - -- --- - - ----- - 1 . We use the c o n ven tionth at $ (x^{1}, x^{2},x^{3})$ repr e sen t the spatial ax es a n d $ x^{ 4 }\ e qui vi ct$ r epresents the t emporal on ew it h a corres p ond in gly con sistent nota t ion for the four -vector c ur rent , $J_ {a}$, andthe elec tromagnet i c fie l dtenso r,$F_{ab }$ ; 2. T he met r ictenso r is n ow consi stent ly represe nted by the kronecker-d elta,$\del ta_ {ab}$; 3 . A lth ough notstri ctly neces sar y w hen u sin g an o rtho n or mal basis , we make a st r ic t d i s ti nction betw e e n co varia nta nd con trav ariant objects in the convention al w a y ; 4.We u se bold-face torep re s e nt an or di nary three- vector;fo r exam ple, t he pos ition v e c to r $\tex tbf{ x}\ equiv(x^{ 1}, x^ { 2},x^{3 }) $; 5. W e de no te a s pace-t i me c o o rdinate as $x\eq uiv(x ^ { 1},x^ { 2}, x^{3} ,x ^{4})$o r $x \equiv(\te xtbf{x},ict )$ acc ordi ng to conven ie nce; 6. F or conveni e nce, we u se th e notat io n $\ par tial/\ part i a l x^{ a

as_a *contact_force* rather than as_a retarded_at-a-distance_force of_the_canonical interpretation. It_is this particular_consequence which leads necessarily_to a changed_view_of the nature of the electromagnetic field and to a potentially deeper understanding of_loss_mechanisms in_electrodynamics. Summary_ -------- By_raising a question about the_status of the electrostatic field_as a_pragmatic idealization, we are able easily to show_that_classical electromagnetism can_be simply reformulated in terms of a (three-component) *relativistic*_magnetic potential. This reformulation reveals, via_a straightforward analysis,_a_relativistic_symmetry principle according to_which the relativistic magnetic potential is_one half of a duality; the_second half, a three-component potential also, has_the property of mass and its_presence has the effect of_forcing a_natural re-interpretation of the Lorentz_force as a_local action_contact force. At_the level of classical electrodynamics, one_unavoidable consequence of_this re-interpretation is a changed understanding_of_electrodynamic loss-mechanisms -_the_processes_by which_electrodynamic energy is_converted,_finally, to_thermodynamic_energy. Beyond the concerns of classical_electrodynamics,_it also becomes clear that there must_be consequences for quantum_electrodynamics_also, since this latter_theory subsumes the electrostatic field_into its basic structure. Notation notes -------------- 1. _We use_the convention_that $(x^{1},x^{2},x^{3})$ represent the spatial axes and $x^{4}\equiv ict$ represents the_temporal one with a correspondingly consistent_notation for the four-vector_current, $J_{a}$,_and_the electromagnetic field_tensor,_$F_{ab}$; 2. _The metric tensor is now consistently represented_by the_kronecker-delta, $\delta_{ab}$; 3. Although not strictly_necessary when using an_orthonormal_basis, we make a strict distinction_between covariant and contravariant objects in_the conventional way; 4. We_use_bold-face_to represent an ordinary three-vector;_for example, the position vector $\textbf{x}\equiv(x^{1},x^{2},x^{3})$; 5._ We denote_a space-time coordinate as $x\equiv(x^{1},x^{2},x^{3},x^{4})$ or $x\equiv(\textbf{x},ict)$_according_to convenience; 6. For convenience, we_use_the notation $\partial/\partial x^{a

i\otimes 1 + 1\otimes 1\otimes\alpha_i, \\ \eta = \beta\otimes 1\otimes1 + 1\otimes\beta\otimes 1 + 1\otimes 1\otimes\beta. \end{gathered}$$ Since $q_k^*$ is injective we shall abuse notation and not take care to distinguish between elements of $H^*({\ensuremath{\operatorname{Sym}^{k}\!{\Sigma}}})$ and their images in $H^*(\Sigma^k)$, and will regard the $\zeta_i$ and $\theta$ as generators of $H^*({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$, and the $\xi_i$ and $\eta$ as generators of $H^*({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})$. The cohomology calculation -------------------------- Returning to the Mayer-Vietoris sequence, letting $$\Phi_i\co H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})\rightarrow H^i(\Sigma\times\Sigma)$$ be the map $\iota^*\oplus q_2^*=(q_3\circ\operatorname{id}\times\Delta)^*\oplus q_2^*$ we have the short exact sequence $$0\rightarrow \operatorname{coker}\Phi_{i-1} \rightarrow H^i({\ensuremath{E_3\Sigma}})\rightarrow \ker\Phi_i\rightarrow 0.$$ Since $H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ is free the kernel of $\Phi_i$ is too, so the sequence splits and we get $$\label{directsum.eq} H^i({\ensuremath{E_3\Sigma}})\cong\operatorname{coker}\Phi_{i-1}\oplus\ker\Phi_i.$$ In what follows we calculate the kernel and cokernel of each $\Phi_i$. ### Dimension one {#dimension-one.unnumbered} Both $H^1({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^1({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ and $H^1(\Sigma\times\Sigma)$ have rank $4g$, with bases $\{\xi_i\}\cup\{\zeta_

i\otimes 1 + 1\otimes 1\otimes\alpha_i, \\ \eta = \beta\otimes 1\otimes1 + 1\otimes\beta\otimes 1 + 1\otimes 1\otimes\beta. \end{gathered}$$ Since $ q_k^*$ is injective we shall abuse notation and not take care to identify between chemical element of $ H^*({\ensuremath{\operatorname{Sym}^{k}\!{\Sigma}}})$ and their images in $ H^*(\Sigma^k)$, and will regard the $ \zeta_i$ and $ \theta$ as generators of $ H^*({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$, and the $ \xi_i$ and $ \eta$ as generator of $ H^*({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})$. The cohomology calculation -------------------------- Returning to the Mayer - Vietoris sequence, get $ $ \Phi_i\co H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})\rightarrow H^i(\Sigma\times\Sigma)$$ embody the map $ \iota^*\oplus q_2^*=(q_3\circ\operatorname{id}\times\Delta)^*\oplus q_2^*$ we suffer the short exact succession $ $ 0\rightarrow \operatorname{coker}\Phi_{i-1 } \rightarrow H^i({\ensuremath{E_3\Sigma}})\rightarrow \ker\Phi_i\rightarrow 0.$$ Since $ H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ is free the kernel of $ \Phi_i$ is too, so the sequence separate and we get $ $ \label{directsum.eq } H^i({\ensuremath{E_3\Sigma}})\cong\operatorname{coker}\Phi_{i-1}\oplus\ker\Phi_i.$$ In what follow we calculate the kernel and cokernel of each $ \Phi_i$. # # # Dimension one { # dimension-one.unnumbered } Both $ H^1({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^1({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ and $ H^1(\Sigma\times\Sigma)$ have rate $ 4g$, with bases $ \{\xi_i\}\cup\{\zeta _

i\otlmes 1 + 1\otimer 1\otimes\alpha_i, \\ \eta = \beta\ktimes 1\ogimes1 + 1\otimes\beta\otimes 1 + 1\otimes 1\otimes\beta. \evd{gatheref}$$ Since $w_k^*$ iw injective we shall abuss notctmon and not takg care to divtinguish betwaev zlements of $H^*({\ensuremath{\operatorname{Srm}^{k}\!{\Sigms}}})$ wnd their imaggs in $R^*(\Sigja^k)$, and will regard the $\zeta_i$ and $\fheta$ av generators pf $H^*({{\ensuremath{\operatorname{Dym}^{2}\!{\Slgma}}}})$, and the $\xi_i$ ajd $\eta$ as ggherwrors of $H^*({{\ensjremath{\opegctorname{Sym}^{3}\!{\Aigma}}}})$. The cohomology calculation -------------------------- Feturuing to the Mqyeg-Eietoris sewuencv, letting $$\Phi_l\bo H^i({{\envurematn{\operatorname{Snm}^{3}\!{\Sigka}}}})\oplus H^i({{\ensuremath{\operetorname{Sym}^{2}\!{\Sigma}}}})\righjarrow H^i(\Shgja\times\Sigma)$$ be tye map $\hota^*\mplur q_2^*=(q_3\zird\o'erztornale{iv}\times\Delta)^*\kplus q_2^*$ we yave the short exacu sqwuence $$0\rightadrow \o[ewatorname{coker}\Phi_{i-1} \rightarrow H^i({\ensuremdth{S_3\Sigma}})\rightarrow \ker\Phi_i\eightarrow 0.$$ Since $H^i({{\ejsuremath{\jperatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatornama{Sym}^{2}\!{\Smgoa}}}})$ nw fred tje kernel of $\Phi_i$ is too, so the sequence spliea snc we get $$\label{birectsum.eq} H^i({\ensirfmsjh{E_3\Sigma}})\cong\opgratorncje{doker}\Phi_{i-1}\oplus\ker\Pji_i.$$ In rhat dollows wt calvulate the kernel and cokerbel of each $\Ihi_i$. ### Dimension one {#dimznsion-one.unnbmberec} Both $H^1({{\ensuremath{\operatornamz{Sym}^{3}\!{\Sifma}}}})\oplus H^1({{\ejsuremath{\koeratorname{Sym}^{2}\!{\Siema}}}})$ atd $H^1(\Sigma\times\Sigma)$ have rwnk $4g$, witi baszs $\{\xi_i\}\cuo\{\zets_

i\otimes 1 + 1\otimes 1\otimes\alpha_i, \\ \eta 1\otimes1 1\otimes\beta\otimes 1 1\otimes 1\otimes\beta. \end{gathered}$$ shall notation and not care to distinguish elements of $H^*({\ensuremath{\operatorname{Sym}^{k}\!{\Sigma}}})$ and their images $H^*(\Sigma^k)$, and will regard the $\zeta_i$ and $\theta$ as generators of $H^*({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$, and $\xi_i$ and $\eta$ as generators of $H^*({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})$. The cohomology calculation -------------------------- Returning to Mayer-Vietoris letting H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i(\Sigma\times\Sigma)$$ be the map $\iota^*\oplus q_2^*=(q_3\circ\operatorname{id}\times\Delta)^*\oplus q_2^*$ we have the short exact sequence $$0\rightarrow \operatorname{coker}\Phi_{i-1} \rightarrow \ker\Phi_i\rightarrow 0.$$ Since $H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ is free the of $\Phi_i$ is too, the sequence splits and we $$\label{directsum.eq} In what we the and cokernel of $\Phi_i$. ### Dimension one {#dimension-one.unnumbered} Both $H^1({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^1({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ and $H^1(\Sigma\times\Sigma)$ have rank $4g$, with bases $\{\xi_i\}\cup\{\zeta_

i\otimes 1 + 1\otimes 1\otimes\alpha_i, \\ \Eta = \beta\otiMes 1\otImeS1 + 1\otImEs\beTa\otImes 1 + 1\otimes 1\otimES\betA. \end{gathered}$$ Since $q_k^*$ is iNjectIvE We shALl Abuse NotatioN AnD NOt tAkE cAre To DIsTinguIsh Between Elements of $h^*({\enSuRemath{\operatORnAme{Sym}^{k}\!{\SigMa}}})$ aNd their imageS in $h^*(\Sigma^K)$, aNd wILl regArd The $\zeTa_i$ and $\THeta$ as GeneratorS oF $h^*({{\ensurEMath{\opeRAToRnamE{Sym}^{2}\!{\Sigma}}}})$, and the $\xi_I$ AnD $\Eta$ as generatorS of $H^*({{\enSuREmATH{\opEraTorname{Sym}^{3}\!{\siGma}}}})$. ThE CohomolOGy CALCulATion -------------------------- Returning To the Mayer-VIEtoRis seqUeNce, LEtting $$\phi_i\cO H^I({{\EnsUremath{\operAtorName{Sym}^{3}\!{\SiGma}}}})\oplUS H^i({{\ensuREmath{\opEratorNamE{SyM}^{2}\!{\SigMA}}}})\rIgHtaRrOW H^i(\sIgMa\tIMes\sigma)$$ be tHe MaP $\iota^*\OpluS Q_2^*=(Q_3\CIrc\oPerAtorName{iD}\times\Delta)^*\opLus Q_2^*$ we hAVe tHe shoRt exaCt seQuEnce $$0\rIghtarRow \opErAtorname{coker}\PhI_{i-1} \riGhtarrow H^I({\enSuRemAtH{E_3\SigMA}})\rightArrOw \kEr\Phi_i\rIghtarrOW 0.$$ SiNcE $h^I({{\EnSuremath{\operatornaMe{sYM}^{3}\!{\SIgma}}}})\opluS H^i({{\ensUReMaTH{\operatoRnAme{sym}^{2}\!{\SIGMa}}}})$ is fRee tHE kErnel of $\PHi_i$ is tOO, sO tHe sequeNcE splitS aNd wE geT $$\labeL{DireCtsum.eQ} H^i({\ensurEmath{e_3\sigma}})\cong\operaTOrname{coker}\PhI_{I-1}\oPLUs\KEr\PhI_i.$$ IN what followS we cALculAte tHE kErnEL and cOkernEl OF eACh $\Phi_i$. ### Dimension one {#dImEnsion-One.unNumbered} Both $H^1({{\Ensuremath{\OPERatornamE{Sym}^{3}\!{\sIgMA}}}})\oplus H^1({{\ensuremAth{\opEratorname{sYm}^{2}\!{\Sigma}}}})$ aNd $H^1(\SiGma\times\sigma)$ have RANk $4g$, with bAseS $\{\xi_I\}\cuP\{\zeTA_

i\otimes 1 +1\otimes 1 \otim es\ alp ha _i, \\ \eta = \bet a \oti mes 1\otimes1 + 1\otim es\be ta \ otim e s1 + 1\o ti me s 1 \o t im es\be ta. \end{g athered}$$ Si nc e $q_k^*$ is in jective we sh all abuse no tat ion an dnot takecar e todistin g uish b etween el em e nts of $H^*({\ e n su rema th{\operatorname{ S ym } ^{k}\!{\Sigma} }})$ a nd th e i r i mag es in $H^* (\ Sigma ^ k)$, an d w i l l re g ard the $\zet a_i$ and $\ t het a$ asge ner a tors o f $H^ *( { {\e nsuremath{\ oper atorname{ Sym}^{ 2 }\!{\Si g ma}}}}) $, and th e $ \xi_ i $an d $ \e t a$a sgen e rat ors of $ H^ *( {{\en sure m a t h {\op era torn ame{S ym}^{3}\!{\Si gma }}}} ) $. Thecohom olog ycalcu lation ---- -- --------------- ---- - Return ing t o t he Maye r -Vieto ris se quence, lettin g $$ \P h i _ i\ co H^i({{\ensurema th { \ op eratorna me{Sym } ^{ 3} \ !{\Sigma }} }}) \opl u s H^i( {{\e n su remath{\ operat o rn am e{Sym}^ {2 }\!{\S ig ma} }}} )\rig h tarr ow H^i (\Sigma\ times \ Sigma)$$ be th e map $\iota^* \ op l u sq _2^* =(q _3\circ\ope rato r name {id} \ ti mes \ Delta )^*\o pl u sq _2^*$ we have the s ho rt exa ct se quence $$0\ri ghtarrow \ o p e ratornam e{co k er } \Phi_{i-1} \ri ghtar row H^i({\ e nsuremat h{E_3 \Sigma}} )\rightar r o w \ker\P hi_ i\r igh tar r o w0.$$ Since $H ^ i ({{\ en suremat h{\ operato rna me{ Sym }^{ 3} \!{\Sigma }}}})\op lu sH^ i( {{\ ensur e math{\op er ato rn ame {Sym} ^ {2}\!{ \Sigm a}}} }) $i s f ree the ke r n el o f$\ Phi_ i$is too, sot hesequenc e splitsand we g et $ $\label {directsum.eq }H^i({\ensu re mat h{E_3\ S i gma}})\c ong\operatorname{coker} \ Phi_{i- 1}\ oplus \ker \Phi_i.$$ In whatfol l ows we calcu lateth e k e r nel a n d c oke rn el of each $ \Ph i_i$. ###Dimensi on one {#dimension - one .unnumbered} Bo th $ H ^ 1( {{\ e ns u rem at h {\o p e ratorname{Sym}^ {3}\!{\Sig ma } }} })\oplus H ^ 1({ {\ ensurem ath{\op erato r name{Sy m}^{2}\!{ \Sigma}}} }) $ an d $H^ 1(\Sigma\t imes\Sig ma)$ have rank$ 4g $, wi thbases$\ {\x i_i\} \cup\{ \ zet a_

i\otimes 1_ _ _ __ __+ 1\otimes 1\otimes\alpha_i,_ _ \\ \eta = \beta\otimes_1\otimes1 + 1\otimes\beta\otimes_1_ + 1\otimes_1\otimes\beta._\end{gathered}$$ Since_$q_k^*$_is_injective we shall abuse notation_and not take care to_distinguish between_elements of $H^*({\ensuremath{\operatorname{Sym}^{k}\!{\Sigma}}})$ and their images in $H^*(\Sigma^k)$,_and_will regard the_$\zeta_i$ and $\theta$ as generators of $H^*({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$, and the_$\xi_i$ and $\eta$ as generators of_$H^*({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})$. The cohomology calculation -------------------------- Returning_to_the_Mayer-Vietoris sequence, letting $$\Phi_i\co_H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})\rightarrow H^i(\Sigma\times\Sigma)$$ be the map_$\iota^*\oplus q_2^*=(q_3\circ\operatorname{id}\times\Delta)^*\oplus q_2^*$ we have the_short exact sequence $$0\rightarrow \operatorname{coker}\Phi_{i-1} \rightarrow H^i({\ensuremath{E_3\Sigma}})\rightarrow \ker\Phi_i\rightarrow_0.$$ Since $H^i({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^i({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ is free_the kernel of $\Phi_i$ is_too, so_the sequence splits and we_get $$\label{directsum.eq} H^i({\ensuremath{E_3\Sigma}})\cong\operatorname{coker}\Phi_{i-1}\oplus\ker\Phi_i.$$ In_what follows_we calculate the_kernel and cokernel of each $\Phi_i$. ###_Dimension one {#dimension-one.unnumbered} Both_$H^1({{\ensuremath{\operatorname{Sym}^{3}\!{\Sigma}}}})\oplus H^1({{\ensuremath{\operatorname{Sym}^{2}\!{\Sigma}}}})$ and $H^1(\Sigma\times\Sigma)$ have rank_$4g$,_with bases $\{\xi_i\}\cup\{\zeta_

75312,0.006619)(451.8770384,0.0066049)(452.9276094,0.0065896)(453.9590438,0.0065747)(454.9212414,0.0065609)(455.9645029,0.0065459)(456.9385276,0.0065321)(457.8934157,0.0065185)(458.9293677,0.0065039)(459.8960828,0.0064903)(460.9438618,0.0064756)(461.9725042,0.0064612)(462.9319099,0.0064479)(463.9723794,0.0064335)(464.9436121,0.0064202)(465.8957082,0.0064071)(466.9288682,0.006393)(467.8927914,0.0063799)(468.9377784,0.0063657)(469.9135286,0.0063526)(470.8701423,0.0063398)(471.9078197,0.0063259)(472.8762605,0.006313)(473.925765,0.0062991)(474.956133,0.0062855)(475.9172642,0.0062728)(476.9594592,0.0062592)(477.9324174,0.0062465)(478.8862391,0.0062341)(479.9211246,0.0062208)(480.8867733,0.0062083)(481.9334859,0.0061949)(482.9610619,0.0061818)(483.9194011,0.0061696)(484.9588041,0.0061565)(485.9289704,0.0061442)(486.9802005,0.006131)(488.012294,0.0061181)(488.9751507,0.0061061)(490.0190713,0.0060932)(490.9937551,0.0060811)(491.9493023,0.0060694)(492.9859133,0.0060567)(493.

75312,0.006619)(451.8770384,0.0066049)(452.9276094,0.0065896)(453.9590438,0.0065747)(454.9212414,0.0065609)(455.9645029,0.0065459)(456.9385276,0.0065321)(457.8934157,0.0065185)(458.9293677,0.0065039)(459.8960828,0.0064903)(460.9438618,0.0064756)(461.9725042,0.0064612)(462.9319099,0.0064479)(463.9723794,0.0064335)(464.9436121,0.0064202)(465.8957082,0.0064071)(466.9288682,0.006393)(467.8927914,0.0063799)(468.9377784,0.0063657)(469.9135286,0.0063526)(470.8701423,0.0063398)(471.9078197,0.0063259)(472.8762605,0.006313)(473.925765,0.0062991)(474.956133,0.0062855)(475.9172642,0.0062728)(476.9594592,0.0062592)(477.9324174,0.0062465)(478.8862391,0.0062341)(479.9211246,0.0062208)(480.8867733,0.0062083)(481.9334859,0.0061949)(482.9610619,0.0061818)(483.9194011,0.0061696)(484.9588041,0.0061565)(485.9289704,0.0061442)(486.9802005,0.006131)(488.012294,0.0061181)(488.9751507,0.0061061)(490.0190713,0.0060932)(490.9937551,0.0060811)(491.9493023,0.0060694)(492.9859133,0.0060567)(493.

75312,0.006619)(451.8770384,0.0066049)(452.9276094,0.0065896)(453.9590438,0.0065747)(454.9212414,0.0065609)(455.9645029,0.0065459)(456.9385276,0.0065321)(457.8934157,0.0065185)(458.9293677,0.0065039)(459.8960828,0.0064903)(460.9438618,0.0064756)(461.9725042,0.0064612)(462.9319099,0.0064479)(463.9723794,0.0064335)(464.9436121,0.0064202)(465.8957082,0.0064071)(466.9288682,0.006393)(467.8927914,0.0063799)(468.9377784,0.0063657)(469.9135286,0.0063526)(470.8701423,0.0063398)(471.9078197,0.0063259)(472.8762605,0.006313)(473.925765,0.0062991)(474.956133,0.0062855)(475.9172642,0.0062728)(476.9594592,0.0062592)(477.9324174,0.0062465)(478.8862391,0.0062341)(479.9211246,0.0062208)(480.8867733,0.0062083)(481.9334859,0.0061949)(482.9610619,0.0061818)(483.9194011,0.0061696)(484.9588041,0.0061565)(485.9289704,0.0061442)(486.9802005,0.006131)(488.012294,0.0061181)(488.9751507,0.0061061)(490.0190713,0.0060932)(490.9937551,0.0060811)(491.9493023,0.0060694)(492.9859133,0.0060567)(493.

75312,0.006619)(451.8770384,0.0066049)(452.9276094,0.0065896)(453.9590438,0.0065747)(454.9212414,0.0065609)(455.9645029,0.0065459)(456.9385276,0.0065321)(457.8934157,0.0065185)(458.9293677,0.0065039)(459.8960828,0.0064903)(460.9438618,0.0064756)(461.9725042,0.0064612)(462.9319099,0.0064479)(463.9723794,0.0064335)(464.9436121,0.0064202)(465.8957082,0.0064071)(466.9288682,0.006393)(467.8927914,0.0063799)(468.9377784,0.0063657)(469.9135286,0.0063526)(470.8701423,0.0063398)(471.9078197,0.0063259)(472.8762605,0.006313)(473.925765,0.0062991)(474.956133,0.0062855)(475.9172642,0.0062728)(476.9594592,0.0062592)(477.9324174,0.0062465)(478.8862391,0.0062341)(479.9211246,0.0062208)(480.8867733,0.0062083)(481.9334859,0.0061949)(482.9610619,0.0061818)(483.9194011,0.0061696)(484.9588041,0.0061565)(485.9289704,0.0061442)(486.9802005,0.006131)(488.012294,0.0061181)(488.9751507,0.0061061)(490.0190713,0.0060932)(490.9937551,0.0060811)(491.9493023,0.0060694)(492.9859133,0.0060567)(493.

75312,0.006619)(451.8770384,0.0066049)(452.9276094,0.0065896)(453.9590438,0.0065747)(454.9212414,0.0065609)(455.9645029,0.0065459)(456.9385276,0.0065321)(457.8934157,0.0065185)(458.9293677,0.0065039)(459.8960828,0.0064903)(460.9438618,0.0064756)(461.9725042,0.0064612)(462.9319099,0.0064479)(463.9723794,0.0064335)(464.9436121,0.0064202)(465.8957082,0.0064071)(466.9288682,0.006393)(467.8927914,0.0063799)(468.9377784,0.0063657)(469.9135286,0.0063526)(470.8701423,0.0063398)(471.9078197,0.0063259)(472.8762605,0.006313)(473.925765,0.0062991)(474.956133,0.0062855)(475.9172642,0.0062728)(476.9594592,0.0062592)(477.9324174,0.0062465)(478.8862391,0.0062341)(479.9211246,0.0062208)(480.8867733,0.0062083)(481.9334859,0.0061949)(482.9610619,0.0061818)(483.9194011,0.0061696)(484.9588041,0.0061565)(485.9289704,0.0061442)(486.9802005,0.006131)(488.012294,0.0061181)(488.9751507,0.0061061)(490.0190713,0.0060932)(490.9937551,0.0060811)(491.9493023,0.0060694)(492.9859133,0.0060567)(493.

75312,0.006619)(451.877038 4,0.006604 9)(45 2.9 276 09 4,0. 0065 896)(453.95904 3 8,0. 0065747)(454.9212414,0 .0065 60 9 )(45 5 .9 64502 9,0.006 5 45 9 ) (45 6. 93 852 76 , 0. 00653 21) (457.89 34157,0.00 651 85 )(458.929367 7 ,0 .0065039)( 459 .8960828,0.0 064 903)(4 60 .94 3 8618, 0.0 06475 6)(461 . 972504 2,0.00646 12 ) (462.9 3 19099,0 . 0 06 4479 )(463.9723794,0.0 0 64 3 35)(464.943612 1,0.00 64 2 02 ) ( 465 .89 57082,0.00 64 071)( 4 66.9288 6 82 , 0 . 006 3 93)(467.89279 14,0.006379 9 )(4 68.937 77 84, 0 .00636 57)(4 69 . 913 5286,0.0063 526) (470.8701 423,0. 0 063398) ( 471.907 8197,0 .00 632 59)( 4 72 .8 762 60 5 ,0. 0 06 313 ) (47 3.925765 ,0 .0 06299 1)(4 7 4 . 9 5613 3,0 .006 2855) (475.9172642, 0.0 0627 2 8)( 476.9 59459 2,0. 00 62592 )(477. 93241 74 ,0.0062465)(478 .886 2391,0.00 623 41 )(4 79 .9211 2 46,0.0 062 208 )(480.8 867733, 0 .00 62 0 8 3 )( 481.9334859,0.0061 94 9 ) (4 82.96106 19,0.0 0 61 81 8 )(483.91 94 011 ,0.0 0 6 1696) (484 . 95 88041,0. 006156 5 )( 48 5.92897 04 ,0.006 14 42) (48 6.980 2 005, 0.0061 31)(488. 01229 4 ,0.0061181)(48 8 .9751507,0.00 6 10 6 1 )( 4 90.0 190 713,0.00609 32)( 4 90.9 9375 5 1, 0.0 0 60811 )(491 .9 4 93 0 23,0.0060694)(492.9 85 9133,0 .0060 567)(493.

75312,0.006619)(451.8770384,0.0066049)(452.9276094,0.0065896)(453.9590438,0.0065747)(454.9212414,0.0065609)(455.9645029,0.0065459)(456.9385276,0.0065321)(457.8934157,0.0065185)(458.9293677,0.0065039)(459.8960828,0.0064903)(460.9438618,0.0064756)(461.9725042,0.0064612)(462.9319099,0.0064479)(463.9723794,0.0064335)(464.9436121,0.0064202)(465.8957082,0.0064071)(466.9288682,0.006393)(467.8927914,0.0063799)(468.9377784,0.0063657)(469.9135286,0.0063526)(470.8701423,0.0063398)(471.9078197,0.0063259)(472.8762605,0.006313)(473.925765,0.0062991)(474.956133,0.0062855)(475.9172642,0.0062728)(476.9594592,0.0062592)(477.9324174,0.0062465)(478.8862391,0.0062341)(479.9211246,0.0062208)(480.8867733,0.0062083)(481.9334859,0.0061949)(482.9610619,0.0061818)(483.9194011,0.0061696)(484.9588041,0.0061565)(485.9289704,0.0061442)(486.9802005,0.006131)(488.012294,0.0061181)(488.9751507,0.0061061)(490.0190713,0.0060932)(490.9937551,0.0060811)(491.9493023,0.0060694)(492.9859133,0.0060567)(493.

dynamical behavior of the analyzed systems. Among others, two landscape-based approaches have proven to be particularly stimulating. The first one concerns the detailed analysis of the [*inherent structures*]{} (i.e. the configurations at the minima of potential energy) visited by the system at different temperatures. This method has allowed to clarify many interesting phenomena, as, for example, the thermodynamic picture of the supercooled liquid regime based on the configurational entropy [@fs_entropy], the relationship between fragility and properties of inherent structures [@land_sastry2], the analysis of diffusion processes in terms of visited inherent structures [@land_keyes; @fabr; @la_98], or the interpretation of the effective fluctuation-dissipation temperature in the out-of-equilibrium regime in terms of inherent structures visited during aging [@fs_aging], only to cite a few. The second approach is based on the analysis of the eigenvalues (normal modes) of the Hessian at the instantaneous configurations during the dynamic evolution of the system, from here the name [ *instantaneous normal mode approach*]{} (for an introduction and an extended application of this method see the works of Keyes and coworkers [@keyes_inm; @keyes_vari]). This approach allowed to relate the emergent diffusive processes to the features of the landscape, opening the way to the interpretation of diffusion in terms of accessible paths in the multidimensional energy surface. Promising steps was obtained [*i)*]{} using simultaneously both the instantaneous normal mode approach and the inherent structure one, in order to identify the relevant slow diffusive directions [@donati; @lanave], and [*ii)*]{} by analyzing the reaction paths in order to eliminate the non-diffusive unstable modes [@bembe]. Recently, a further approach has been introduced [@noi_sad; @cav_sad] and applied to the study of supercooled liquids[@sad_1; @doye; @sad_3; @sad_4; @sad_5; @grig; @sadBLJ]. This approach is based on the analysis of the [*saddles*]{} of the potential energy surface and has provided new insight in the analysis of the dynamic crossover taking place on lowering the temperature in supercooled liquids. Indeed it allows to characterize the dynamic transition temperature $T_{MCT}$ (mode-coupling temperature [@mct]) as the temperature where the order (fractional number of negative eigenvalues

dynamical behavior of the analyzed systems. Among others, two landscape - free-base approach have proven to be peculiarly stimulating. The beginning one concerns the detailed analysis of the [ * built-in structures * ] { } (i.e. the configurations at the minima of likely department of energy) visited by the system at unlike temperatures. This method has allow to clarify many interesting phenomena, as, for example, the thermodynamic picture of the supercooled liquid regime based on the configurational information [ @fs_entropy ], the relationship between fragility and properties of inherent structure [ @land_sastry2 ], the analysis of diffusion processes in terms of visit inherent structures [ @land_keyes; @fabr; @la_98 ], or the interpretation of the effective fluctuation - dissipation temperature in the away - of - equilibrium regime in terms of inherent structures visited during age [ @fs_aging ], only to cite a few. The second approach is based on the analysis of the eigenvalues (normal modes) of the Hessian at the instantaneous configurations during the dynamic evolution of the arrangement, from here the name [ * instantaneous normal mode approach * ] { } (for an initiation and an extended application of this method see the work of Keyes and coworkers [ @keyes_inm; @keyes_vari ]). This approach allowed to relate the emergent diffusive process to the features of the landscape, opening the way to the interpretation of diffusion in term of accessible paths in the multidimensional energy surface. Promising steps was obtained [ * i) * ] { } practice simultaneously both the instantaneous normal mode approach and the inherent structure one, in order to name the relevant slow diffusive directions [ @donati; @lanave ], and [ * ii) * ] { } by analyzing the reaction paths in order to eliminate the non - diffusive mentally ill modes [ @bembe ]. Recently, a further approach has been precede [ @noi_sad; @cav_sad ] and applied to the survey of supercooled liquids[@sad_1; @doye; @sad_3; @sad_4; @sad_5; @grig; @sadBLJ ]. This approach is establish on the analysis of the [ * saddle * ] { } of the potential energy airfoil and has provided modern insight in the analysis of the dynamic crossover taking place on lowering the temperature in supercooled liquids. Indeed it allows to qualify the dynamic transition temperature $ T_{MCT}$ (manner - coupling temperature [ @mct ]) as the temperature where the order (fractional number of negative eigenvalues

dyjamical behavior of the xnalyzed systems. Among mthers, two lanascape-based approaches have 'rovwn to be particularly stimuuating. Thv first obe cibcerns the detailed analgdis mh the [*inherent xtructures*]{} (i.e. the configgrxtnons at the minima of potential enerdy) visiyef by the systei at qiffsgekt temperatures. This method has zllowed to clarify msny interesting phenomena, ws, flr example, the thegmodynamic kjcttee of the suoercooled liquid regimg based on the configurational evtropv [@fs_entropy], tye gglationship uetweeg fragility and proparties pf inherent stvuctuces [@oand_sastry2], the analysms of diffusion procgsses in tarjs of visited inhwrwnt sjructgres [@oana_kegex; @rabr; @lw_98], oc the interlretation od the effective fluvttqtion-dissipatjon teipqrature in the out-of-equilibrium regime pn tsrms of inherent structyres visited during ahing [@fs_agyng], only to cite a few. The second approach is basad on ghe akalyrus of the eigenvalues (normal modes) of the Hessiwh st the instantancous configurationx futyng the dynamkc evolutjon of the system, vrom hete the name [ *inftanyaneous normal mode approacy*]{} (for an intgoduxtion and an extenbed applicatnon of this method see the works oy Keyea and coworners [@keyea_knm; @keyes_vari]). Thks sp[roach allowed to relate tre emergeit diyfusive orocgsses tj the featkres of the landscape, opfning tve way to hhe interpretation of diffusion in terms of avcasspble pathf in bhe multidimensyonal energy sorface. Prjmisivg steps wzs obtamned [*i)*]{} using simultaneousni both the invtantanejus bormql mode xpproach and tne inhereuu structurw one, in order to ldentkry the relevant sliw diffusive ditecgiogs [@dpnwdi; @lanave], ang [*ii)*]{} by snalyxing tht rcacgion paths in order to enimihate the non-diffusovc unstablg modes [@bqmbe]. Recently, s further approach has ueen iitroduved [@noi_sad; @cav_sad] and applied to fhe study of supercooled liqmids[@fad_1; @doye; @scd_3; @sad_4; @sad_5; @grig; @sadBLJ]. This approach is uased on the analysis od the [*saddles*]{} of thg pptential enxrgy strface ang has provided new ibsight in the anakysis of the dynamic drossoeer twking place on lowering the temperature in supercooled liquids. Indeed it qllows to charactedize the gyuamnc tranfitili temperature $T_{MCN}$ (mode-coupling temperature [@mct]) ar the tekpzrature where the order (fractooval number of negative eigenvalues

dynamical behavior of the analyzed systems. Among landscape-based have proven be particularly stimulating. detailed of the [*inherent (i.e. the configurations the minima of potential energy) visited the system at different temperatures. This method has allowed to clarify many interesting as, for example, the thermodynamic picture of the supercooled liquid regime based on configurational [@fs_entropy], relationship fragility and properties of inherent structures [@land_sastry2], the analysis of diffusion processes in terms of visited structures [@land_keyes; @fabr; @la_98], or the interpretation of effective fluctuation-dissipation temperature in out-of-equilibrium regime in terms of structures during aging only cite few. The second is based on the analysis of the eigenvalues (normal modes) of the Hessian at the instantaneous configurations the dynamic the system, here name *instantaneous normal mode an introduction and an extended application see the works of Keyes and coworkers [@keyes_inm; This approach to relate the emergent diffusive processes the features of the landscape, opening the way the interpretation of diffusion in terms of accessible paths in the multidimensional energy surface. Promising obtained [*i)*]{} using simultaneously the instantaneous normal approach the structure in order identify the relevant slow diffusive directions [@donati; @lanave], and [*ii)*]{} by the reaction paths in order to eliminate the non-diffusive unstable Recently, further approach has introduced [@noi_sad; @cav_sad] and to study of supercooled liquids[@sad_1; @sad_4; @grig; is on analysis of the [*saddles*]{} the potential energy surface and provided new insight in crossover taking place on lowering the temperature in liquids. Indeed it allows to characterize the transition temperature $T_{MCT}$ (mode-coupling temperature [@mct]) as the temperature where the order number of

dynamical behavior of the anaLyzed systeMs. AmoNg oTheRs, Two lAndsCape-based approAChes Have proven to be particulArly sTiMUlatINg. the fiRst one cONcERNs tHe DeTaiLeD AnAlysiS of The [*inheRent structUreS*]{} (i.E. the configurATiOns at the miNimA of potential EneRgy) visItEd bY The sySteM at diFferenT TemperAtures. ThiS mEThod haS Allowed TO ClArifY many interesting pHEnOMena, as, for exampLe, the tHeRMoDYNamIc pIcture of thE sUpercOOled liqUId REGIme BAsed on the confIgurational ENtrOpy [@fs_eNtRopY], The relAtionShIP beTween fragilIty aNd propertIes of iNHerent sTRuctureS [@land_sAstRy2], tHe anALySiS of DiFFusIOn ProCEssEs in termS oF vIsiteD inhERENT strUctUres [@Land_kEyes; @fabr; @la_98], or tHe iNterPRetAtion Of the EffeCtIve flUctuatIon-diSsIpation temperatUre iN the out-of-EquIlIbrIuM regiME in terMs oF inHerent sTructurES viSiTED DuRing aging [@fs_aging], onLy TO CiTe a few. ThE seconD ApPrOAch is basEd On tHe anALYsis oF the EIgEnvalues (Normal MOdEs) Of the HeSsIan at tHe InsTanTaneoUS conFiguraTions durIng thE Dynamic evolutiON of the system, fROm HERe THe naMe [ *iNstantaneouS norMAl moDe apPRoAch*]{} (FOr an iNtrodUcTIoN And an extended applicAtIon of tHis meThod see the worKs of Keyes aND COworkers [@KeyeS_InM; @Keyes_vari]). This aPproaCh allowed tO Relate thE emerGent diffUsive procESSes to the FeaTurEs oF thE LAnDscape, opening THE way To The inteRprEtation Of dIffUsiOn iN tErms of accEssible pAtHs In ThE muLtidiMEnsional EnErgY sUrfAce. PrOMising Steps Was oBtAiNEd [*i)*]{} Using siMUlTANeouSlY bOth tHe iNsTantaNeouS NorMal mode Approach aNd tHE inhErEnT structUre one, in order To Identify thE rEleVant slOW DiffusivE directions [@donati; @lanave], ANd [*ii)*]{} by aNalYzing The rEaction paThs In ordeR to ELiminaTe the nOn-difFuSivE UNstabLE MoDes [@BeMbe]. RecentlY, A FurTher aPpRoacH has beeN introduced [@noi_sad; @cAV_saD] and applied to The StudY OF sUpeRCoOLed LiQUidS[@SAd_1; @doye; @sad_3; @sad_4; @sad_5; @Grig; @sadBLJ]. thIS aPproach is bASed On The analYsis of tHe [*sadDLes*]{} of thE potentiaL energy suRfAce aND Has Provided neW insight In the analYSis of THe DynamIc cRossovEr TakIng plAce on lOWerIng thE tempeRaTure in SuperCoOled liquIds. Indeed it allows to charActeriZe the DynAmic transItiON teMperature $t_{MCT}$ (Mode-coupliNg tEmpEratuRe [@mCT]) as thE temPErAtuRE wherE the ORder (fractIOnAl nUMBeR of negative EIGEnvAlues

dynamical behavior of the analyzedsyste ms. A mo ng o ther s, two landsca p e-ba sed approaches have pr ovento be p a rt icula rly sti m ul a t ing .Th e f ir s tone c onc erns th e detailed an al ysis of the[ *i nherent st ruc tures*]{} (i .e. the c on fig u ratio nsat th e mini m a of p otentialen e rgy) v i sited b y th e sy stem at different te m peratures. Thi s meth od ha s all owe d to clari fy many interes t in g p hen o mena, as, for example, t h e t hermod yn ami c pictu re of t h e s upercooledliqu id regime based on thec onfigur ationa l e ntr opy[ @f s_ ent ro p y], th e r e lat ionshipbe tw een f ragi l i t y and pr oper tiesof inherent s tru ctur e s [ @land _sast ry2] ,the a nalysi s ofdi ffusion process es i n terms o f v is ite dinher e nt str uct ure s [@lan d_keyes ; @f ab r ; @l a_98], or the inte rp r e ta tion ofthe ef f ec ti v e fluctu at ion -dis s i patio n te m pe rature i n theo ut -o f-equil ib rium r eg ime in term s ofinhere nt struc tures visited during aging [@fs_ag i ng ] , o n ly t o c ite a few.Thes econ d ap p ro ach is ba sed o nt he analysis of the eig en values (nor mal modes) of the Hessi a n at the i nsta n ta n eous configura tions during th e dynamic evol ution of the syst e m , from h ere th e n ame [ * instantaneous n orma lmode ap pro ach*]{} (f oranint ro duction a nd an ex te nd ed a ppl icati o n of thi smet ho d s ee th e works of K eyes a nd cow orkers[ @k e y es_i nm ;@key es_ va ri]). Thi s ap proachallowed t o r e late t he emerge nt diffusivepr ocesses to t hefeatur e s of thelandscape, opening thew ay to t heinter pret ation ofdif fusion in termsof acc essib le pa t h s int h emul ti dimensiona l ene rgy s ur face . Promi sing steps was obt a ine d [*i)*]{} us ing sim u l ta neo u sl y bo th the i nstantaneous no rmal modeap p ro ach and th e in he rent st ructure one, in orde r to iden tify there leva n t sl ow diffusi ve direc tions [@d o nati; @l anave ],and [* ii )*] {} by analy z ing thereacti on paths in o rd er to el iminate the non-diffusi ve uns table mo des [@bem be] . R ecently,a fu rther appr oac h h as be eni ntrod uced [@ noi _ sad;@cav _ sad] anda pp lie d to the studyo f sup ercoo led liquid s[@s ad_1; @doye; @sad _ 3; @sad_4; @sa d_5; @ gri g;@ sadB LJ ]. This approa chis b ased onth e analysisof the [ *s a ddles *]{} o f thepotenti a l e n ergy s urfa ceand has p rov id e d new i ns ig h t in t he a na lysisof the dyna m i c crossover taki ng pl a c e onl owe ringth e tempe r atur e in super cooled liqu ids. I ndee d itallowsto chara cte ri ze the dyn a mic trans ition temper at ure$T_ {MCT}$ (mo d e -coup ling t emp erature [ @ m ct ] )as the tem perat ur e wh ere the o r der (fra cti o nal num be r o f negati v ee i genvalues

dynamical_behavior of_the analyzed systems. Among others,_two landscape-based_approaches_have proven_to_be particularly stimulating._The first one_concerns the detailed analysis_of the [*inherent_structures*]{}_(i.e. the configurations at the minima of potential energy) visited by the system at_different_temperatures. This_method_has_allowed to clarify many interesting_phenomena, as, for example, the_thermodynamic picture_of the supercooled liquid regime based on the_configurational_entropy [@fs_entropy], the_relationship between fragility and properties of inherent structures [@land_sastry2],_the analysis of diffusion processes in_terms of visited_inherent_structures_[@land_keyes; @fabr; @la_98], or_the interpretation of the effective fluctuation-dissipation_temperature in the out-of-equilibrium regime in_terms of inherent structures visited during aging_[@fs_aging], only to cite a few._The second approach is based_on the_analysis of the eigenvalues (normal_modes) of the_Hessian at_the instantaneous configurations_during the dynamic evolution of the_system, from here_the name [ *instantaneous normal mode_approach*]{}_(for an introduction_and_an_extended application_of this method_see_the works_of_Keyes and coworkers [@keyes_inm; @keyes_vari]). This_approach_allowed to relate the emergent diffusive processes_to the features of_the_landscape, opening the way_to the interpretation of diffusion_in terms of accessible paths in_the multidimensional_energy surface._Promising steps was obtained [*i)*]{} using simultaneously both the instantaneous normal_mode approach and the inherent structure_one, in order to_identify the_relevant_slow diffusive directions_[@donati;_@lanave], and_[*ii)*]{} by analyzing the reaction paths in_order to_eliminate the non-diffusive unstable modes [@bembe]. Recently,_a further approach has_been_introduced [@noi_sad; @cav_sad] and applied to_the study of supercooled liquids[@sad_1; @doye;_@sad_3; @sad_4; @sad_5; @grig; @sadBLJ]._This_approach_is based on the analysis_of the [*saddles*]{} of the potential_energy surface and_has provided new insight in the analysis_of_the dynamic crossover taking place on_lowering_the temperature in supercooled liquids. Indeed_it_allows_to characterize the dynamic transition_temperature $T_{MCT}$ (mode-coupling temperature [@mct]) as_the temperature where the order (fractional number of negative_eigenvalues

PBH}}(t_0)$ is the number density of PBHs of a given mass $M$ today, $z(t)$ is the redshift and the time integral runs from $t{_{\rm min}} = 380\,000\,$years at last scattering of the CMB to $t{_{\rm max}} = {\rm Max}(\tau(M),t_0)$ where $\tau(M)\sim M^3$ is the PBH lifetime and $t_0$ is the age of the Universe. As the Universe is expanding, the number density of PBHs evolves as $(1+z(t))^{-3}$, and the energy of the emitted photons evolves as $(1+z(t))^{-1}$. A last factor $(1+z(t))$ comes from the change of integrand variable from the line of sight to the present time. \[sec:results\]Results ====================== We have used the new public code `BlackHawk` [@BlackHawk] to compute the HR of [Eq. (\[eq:hawking\])]{} and the PBH evolution given by Eqs.  and. We consider monochromatic PBH distributions of masses comprised between $M{_{\rm min}} = 10^{13}\,$g and $M{_{\rm max}} = 10^{18}\,$g and initial spin parameters between $a_{i{\rm,min}}^* = 0$ and $a_{i{\rm,max}}^* = 0.9999$, and compute the integral of [Eq. (\[eq:flux\_IGRB\])]{} over the redshift (matter-dominated era) $$z(t) = \left( \dfrac{1}{H_0 t} \right)^{2/3} -1\,, \label{eq:redshift}$$ where $H_0$ is the present Hubble parameter. We then compare the result of the integral to the measured IGRB and find the maximum allowed value of the present PBH number density $n{_{\rm PBH}}(t_0)$ at a given PBH mass $M$, with a conservative approach taking into account the most stringent constraints (e.g. FERMI-LAT$_{\rm C}$ at $E = 1\,$GeV). The corresponding limit on the DM fraction $f$ constituted of PBHs of mass $M$ is obtained through $ n{_{\rm PBH}}(t_0) = f{\rho{_{\rm DM}}}/{M}\,

PBH}}(t_0)$ is the number density of PBHs of a given mass $ M$ today, $ z(t)$ is the red shift and the meter integral runs from $ t{_{\rm minute } } = 380\,000\,$years at last scatter of the CMB to $ t{_{\rm max } } = { \rm Max}(\tau(M),t_0)$ where $ \tau(M)\sim M^3 $ is the PBH lifetime and $ t_0 $ is the age of the Universe. As the Universe is inflate, the number density of PBHs evolve as $ (1+z(t))^{-3}$, and the energy of the emitted photons develop as $ (1+z(t))^{-1}$. A last factor $ (1+z(t))$ come from the change of integrand variable star from the line of view to the present clock time. \[sec: results\]Results = = = = = = = = = = = = = = = = = = = = = = We have used the new public code 'BlackHawk '[ @BlackHawk ] to compute the hour of [ Eq.   (\[eq: hawking\ ]) ] { } and the PBH evolution given by Eqs.   and. We view monochromatic PBH distributions of multitude comprised between $ M{_{\rm min } } = 10^{13}\,$g and $ M{_{\rm max } } = 10^{18}\,$g and initial spin parameters between $ a_{i{\rm, min}}^ * = 0 $ and $ a_{i{\rm, max}}^ * = 0.9999 $, and compute the integral of [ Eq.   (\[eq: flux\_IGRB\ ]) ] { } over the red shift (matter - dominated era) $ $ z(t) = \left (\dfrac{1}{H_0 t } \right)^{2/3 } -1\, , \label{eq: redshift}$$ where $ H_0 $ is the present Hubble parameter. We then compare the result of the integral to the measured IGRB and rule the maximum allowed value of the present PBH number density $ n{_{\rm PBH}}(t_0)$ at a given PBH mass $ M$, with a conservative approach taking into account the about rigorous constraints (e.g. FERMI - LAT$_{\rm C}$ at $ E = 1\,$GeV). The correspond limit on the DM fraction $ f$ constituted of PBHs of mass $ M$ is prevail through $ n{_{\rm PBH}}(t_0) = f{\rho{_{\rm DM}}}/{M}\,

PBJ}}(t_0)$ is the number density of PBHs of a gncen mavs $M$ tkday, $z(t)$ ks the redshift and the time ibtegrql runs from $t{_{\rm min}} = 380\,000\,$hears at past scartermng of the CMB to $t{_{\rm maw}} = {\rm Lax}(\tcu(N),t_0)$ where $\tau(M)\slm M^3$ is the PBH lifetime dna $c_0$ is the age of the Universe. As the Tniversr ls expanding, tre nlmfer svnwity of PBHs evolves as $(1+z(t))^{-3}$, ans the eiergy of the emotted photons evolves as $(1+z(h))^{-1}$. A past factor $(1+z(t))$ comfs from the chagte of integrxnd variable from the mine of sight to the present tioe. \[sec:xesults\]Resuotw ====================== We vave used tie new public code `BlackHafk` [@BlaclHawk] to compube thx HR of [Eq. (\[eq:hawking\])]{} and tie PBH evolution givgn by Eqs.  dnb. We consider monochrimqtic KBH dhstrkvutkona pf massed ckmprised bstween $M{_{\rm nin}} = 10^{13}\,$g and $M{_{\rm max}} = 10^{18}\,$g qnd initial slin pawaieters between $a_{i{\rm,min}}^* = 0$ and $a_{i{\rm,max}}^* = 0.9999$, dnd compute the integral od [Eq. (\[eq:flux\_IGRB\])]{} over tje redshist (matter-dominated era) $$z(t) = \left( \dfrac{1}{H_0 t} \right)^{2/3} -1\,, \ldbel{es:fedwhlft}$$ dyege $H_0$ is the present Hubble parameter. We then skmkarv the result of tme integral to the mfaxored IGRB and wind tks jaximum allowed vapue of jhe prwsent PBH numner density $n{_{\rm PBH}}(t_0)$ at a guven PBH masf $M$, with a conservacive approack takimg inyo account the most strnngent constraintd (e.g. FERMJ-UAT$_{\rm C}$ at $E = 1\,$GeX). Tme worresponding limit on the DM fractmon $f$ constigutec of PFHs of masd $M$ is obtained through $ n{_{\rm KBH}}(t_0) = x{\rho{_{\rm DM}}}/{M}\,

PBH}}(t_0)$ is the number density of PBHs given $M$ today, is the redshift from min}} = 380\,000\,$years last scattering of CMB to $t{_{\rm max}} = {\rm where $\tau(M)\sim M^3$ is the PBH lifetime and $t_0$ is the age of Universe. As the Universe is expanding, the number density of PBHs evolves as and energy the photons evolves as $(1+z(t))^{-1}$. A last factor $(1+z(t))$ comes from the change of integrand variable from line of sight to the present time. \[sec:results\]Results We have used the public code `BlackHawk` [@BlackHawk] to the of [Eq. and PBH given by Eqs. We consider monochromatic PBH distributions of masses comprised between $M{_{\rm min}} = 10^{13}\,$g and $M{_{\rm max}} = and initial between $a_{i{\rm,min}}^* 0$ $a_{i{\rm,max}}^* 0.9999$, and compute of [Eq. (\[eq:flux\_IGRB\])]{} over the redshift = \left( \dfrac{1}{H_0 t} \right)^{2/3} -1\,, \label{eq:redshift}$$ where is the Hubble parameter. We then compare the of the integral to the measured IGRB and the maximum allowed value of the present PBH number density $n{_{\rm PBH}}(t_0)$ at a given $M$, with a conservative taking into account most constraints FERMI-LAT$_{\rm at $E 1\,$GeV). The corresponding limit on the DM fraction $f$ constituted of of mass $M$ is obtained through $ n{_{\rm PBH}}(t_0) =

PBH}}(t_0)$ is the number density of PbHs of a giveN mass $m$ toDay, $Z(t)$ Is thE redShift and the timE InteGral runs from $t{_{\rm min}} = 380\,000\,$yearS at laSt SCattERiNg of tHe CMB to $T{_{\Rm MAX}} = {\rm maX}(\tAu(M),T_0)$ wHErE $\tau(M)\Sim m^3$ is the PbH lifetime And $T_0$ iS the age of the uNiVerse. As the uniVerse is expanDinG, the nuMbEr dENsity Of PbHs evOlves aS $(1+Z(t))^{-3}$, and tHe energy oF tHE emittED photonS EVoLves As $(1+z(t))^{-1}$. A last factor $(1+z(t))$ COmES from the change Of inteGrANd VARiaBle From the linE oF sighT To the prESeNT TIme. \[SEc:results\]ResuLts ====================== We have usED thE new puBlIc cODe `BlacKHawk` [@blACkHAwk] to computE the hR of [Eq. (\[eq:hAwking\])]{} ANd the PBh EvolutiOn giveN by eqs.  And. WE CoNsIdeR mONocHRoMatIC PBh distribUtIoNs of mAsseS COMPrisEd bEtweEn $M{_{\rm Min}} = 10^{13}\,$g and $M{_{\rm max}} = 10^{18}\,$G anD iniTIal Spin pArameTers BeTween $A_{i{\rm,miN}}^* = 0$ and $a_{I{\rM,max}}^* = 0.9999$, and compute tHe inTegral of [EQ. (\[eq:FlUx\_IgRb\])]{} over THe redsHifT (maTter-domInated eRA) $$z(t) = \LeFT( \DFrAc{1}{H_0 t} \right)^{2/3} -1\,, \label{eq:reDsHIFt}$$ Where $H_0$ is The preSEnT HUBble paraMeTer. we thEN CompaRe thE ReSult of thE integRAl To The measUrEd IGRB AnD fiNd tHe maxIMum aLlowed Value of tHe preSEnt PBH number deNSity $n{_{\rm PBH}}(t_0)$ at A GiVEN Pbh masS $M$, wIth a conservAtivE ApprOach TAkIng INto acCount ThE MoST stringent constrainTs (E.g. FERMi-LAT$_{\rM C}$ at $E = 1\,$GeV). The coRrespondinG LIMit on the dM frACtIOn $f$ constituted Of PBHS of mass $M$ is OBtained tHrougH $ n{_{\rm PBH}}(t_0) = F{\rho{_{\rm DM}}}/{M}\,

PBH}}(t_0)$ is the number density o f PBH s o f a g iven mas s $M$ today, $ z (t)$ is the redshift and t he ti me inte g ra l run s from$ t{ _ { \rm m in }}=3 80 \,000 \,$ years a t last sca tte ri ng of the CM B t o $t{_{\rm ma x}} = {\rm M ax} (\tau( M) ,t_ 0 )$ wh ere $\ta u(M)\s i m M^3$ is the P BH lifeti m e and $ t _ 0$ isthe age of the Un i ve r se. As the Uni verseis ex p a ndi ng, the numbe rdensi t y of PB H se v o lve s as $(1+z(t)) ^{-3}$, and the energ yoft he emi ttedph o ton s evolves a s $( 1+z(t))^{ -1}$.A last f a ctor $( 1+z(t) )$com es f r om t hech a nge of in t egr and vari ab le from the l i n e of si ghtto th e present tim e. \[s e c:r esult s\]Re sult s===== ====== ===== == ==== We have u sedthe new p ubl ic co de `Bla c kHawk` [@ Bla ckHawk] to com p ute t h e HR of [Eq. (\[eq:haw ki n g \] )]{} and the P B Hev o lution g iv enby E q s .  an d. W e c onsidermonoch r om at ic PBHdi stribu ti ons of mass e s co mprise d betwee n $M{ _ {\rm min}} = 1 0 ^{13}\,$g and $M { _ {\ r m ma x}} = 10^{18}\ ,$ga nd i niti a lspi n para meter sb et w een $a_{i{\rm,min}} ^* = 0$and $ a_{i{\rm,max} }^* = 0.99 9 9 $ , and co mput e t h e integral of[Eq.(\[eq:flux \ _IGRB\]) ]{} o ver theredshift( m atter-do min ate d e ra) $ $z (t) = \left(\ d frac {1 }{H_0 t } \ right)^ {2/ 3}-1\ ,,\l abel{eq:r edshift} $$ w he re $H _0$ i s the pre se ntHu bbl e par a meter. We t henco mp a rethe res u lt o f th ein tegr alto themeas u red IGRB a nd find t hem axim um a llowedvalue of thepr esent PBHnu mbe r dens i t y $n{_{\ rm PBH}}(t_0)$ at a giv e n PBH m ass $M$, wit h a conse rva tive a ppr o ach ta king i nto a cc oun t the m o s tstr in gent const r a int s (e. g. FER MI-LAT$ _{\rm C}$ at $E =1 \,$ GeV). The cor res pond i n glim i to n t he DMf r action $f$ cons tituted of P B Hs of mass $ M $ i sobtaine d throu gh $n {_{\rmPBH}}(t_0 ) = f{\rh o{ _{\r m DM} }}/{M}\,

PBH}}(t_0)$_is the_number density of PBHs_of a_given_mass $M$_today,_$z(t)$ is the_redshift and the_time integral runs from_$t{_{\rm min}} =_380\,000\,$years_at last scattering of the CMB to $t{_{\rm max}} = {\rm Max}(\tau(M),t_0)$ where $\tau(M)\sim_M^3$_is the_PBH_lifetime_and $t_0$ is the age_of the Universe. As the_Universe is_expanding, the number density of PBHs evolves as_$(1+z(t))^{-3}$,_and the energy_of the emitted photons evolves as $(1+z(t))^{-1}$. A last_factor $(1+z(t))$ comes from the change_of integrand variable_from_the_line of sight to_the present time. \[sec:results\]Results ====================== We have used the_new public code `BlackHawk` [@BlackHawk] to_compute the HR of [Eq. (\[eq:hawking\])]{} and the_PBH evolution given by Eqs.  and._We consider monochromatic PBH distributions_of masses_comprised between $M{_{\rm min}} =_10^{13}\,$g and $M{_{\rm_max}} =_10^{18}\,$g and initial_spin parameters between $a_{i{\rm,min}}^* = 0$_and $a_{i{\rm,max}}^* =_0.9999$, and compute the integral of_[Eq. (\[eq:flux\_IGRB\])]{}_over the redshift_(matter-dominated_era)_$$z(t) =_\left( \dfrac{1}{H_0 t}_\right)^{2/3}_-1\,, \label{eq:redshift}$$_where_$H_0$ is the present Hubble parameter._We_then compare the result of the integral_to the measured IGRB_and_find the maximum allowed_value of the present PBH_number density $n{_{\rm PBH}}(t_0)$ at a_given PBH_mass $M$,_with a conservative approach taking into account the most stringent constraints_(e.g. FERMI-LAT$_{\rm C}$ at $E =_1\,$GeV). The corresponding limit_on the_DM_fraction $f$ constituted_of_PBHs of_mass $M$ is obtained through $ n{_{\rm_PBH}}(t_0) =_f{\rho{_{\rm DM}}}/{M}\,

Kelly-Pedicchio Theorem 4.2]. Furthermore, a regular epimorphism of simplicial objects in a regular Mal’tsev category is always a fibration [@EverVdL2 Proposition 4.4]. The Kan property for simplicial objects may also be expressed in terms of higher extensions: in a semi-abelian category, a simplicial object ${\ensuremath{\mathbb{X}}}$ is Kan if and only if all of its truncations, considered as higher arrows in all possible directions, have a domain which is an extension [@EGoeVdL]. \[Lemma-DeltaLambda-Square\] In a finitely complete category, given $n\geq 1$, $i\in n$, and an augmented simplicial object ${\ensuremath{\mathbb{X}}}$, the square $$\vcenter{\xymatrix{{\triangle}({\ensuremath{\mathbb{X}}},n) \ar[r]^-{\widehat{\partial}_{i}} \ar[d]_-{{\partial}_{i}} & {\raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n) \ar[d]^-{{\partial}_{i-1}^{i}\times {\partial}_{i}^{n-i}} \\ {\ensuremath{\mathbb{X}}}_{n-1} \ar[r]_-{{\lgroup}{\partial}_{j}{\rgroup}_{j}} & {\triangle}({\ensuremath{\mathbb{X}}},n-1)}}$$ is a pullback, where the arrow on the right is the restriction of $${\partial}_{i-1}^{i}\times {\partial}_{i}^{n-i}=\underbrace{{\partial}_{i-1}\times \cdots\times {\partial}_{i-1}}_{\text{$i$ times}} \times \underbrace{{\partial}_{i}\times \cdots \times {\partial}_{i}}_{\text{$n-i$ times}}\colon {\ensuremath{\mathbb{X}}}_{n-1}^{n}\to {\ensuremath{\mathbb{X}}}_{n-2}^{n}$$ to a morphism ${\raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n)\to {\triangle}({\ensuremath{\mathbb{X}}},n-1)$. Here is a picture in degree $n=2$ for $i=1$: $$\begin{matrix} \vcenter{\xymatrix@1@!0@R=2.44

Kelly - Pedicchio Theorem   4.2 ]. Furthermore, a regular epimorphism of simplicial objects in a regular Mal’tsev category is constantly a fibration   [ @EverVdL2 Proposition   4.4 ]. The Kan place for simplicial objects may also be carry in price of higher extensions: in a semi - abelian class, a simplicial object $ { \ensuremath{\mathbb{X}}}$ is Kan if and only if all of its truncations, consider as higher arrows in all potential directions, have a domain which is an extension   [ @EGoeVdL ]. \[Lemma - DeltaLambda - Square\ ] In a finitely accomplished category, given $ n\geq 1 $, $ i\in n$, and an augmented simplicial object $ { \ensuremath{\mathbb{X}}}$, the square $ $ \vcenter{\xymatrix{{\triangle}({\ensuremath{\mathbb{X}}},n) \ar[r]^-{\widehat{\partial}_{i } } \ar[d]_-{{\partial}_{i } } & { \raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n) \ar[d]^-{{\partial}_{i-1}^{i}\times { \partial}_{i}^{n - i } } \\ { \ensuremath{\mathbb{X}}}_{n-1 } \ar[r]_-{{\lgroup}{\partial}_{j}{\rgroup}_{j } } & { \triangle}({\ensuremath{\mathbb{X}}},n-1)}}$$ is a pullback, where the arrow on the right field is the restriction of $ $ { \partial}_{i-1}^{i}\times { \partial}_{i}^{n - i}=\underbrace{{\partial}_{i-1}\times \cdots\times { \partial}_{i-1}}_{\text{$i$ times } } \times \underbrace{{\partial}_{i}\times \cdots \times { \partial}_{i}}_{\text{$n - i$ times}}\colon { \ensuremath{\mathbb{X}}}_{n-1}^{n}\to { \ensuremath{\mathbb{X}}}_{n-2}^{n}$$ to a morphism $ { \raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n)\to { \triangle}({\ensuremath{\mathbb{X}}},n-1)$. Here is a picture in academic degree $ n=2 $ for $ i=1 $: $ $ \begin{matrix } \vcenter{\xymatrix@1@!0@R=2.44

Kelpy-Pedicchio Theorem 4.2]. Furtmermore, a regulat wpimor'hism or simplizial objects in a regular Map’twev cqtegory is always a ficration [@EvvrVdL2 Proposiuion 4.4]. The Kan propxdty for simpljgial mujects may also be expresved in terms ox fiyher extensions: in a semi-abelian catqgory, a slmplicial objest ${\emfurejath{\mathbb{X}}}$ is Kan if and only if zll of pts truncations, cpnsidered as higher arrows in wll possible direchions, have q doiqin which is an extenspmn [@EGoeVdL]. \[Lgmma-DeltaLambda-Square\] In a finitdly cpmplete cajztorj, given $n\geq 1$, $i\in n$, and an aunkented simplivial object ${\enxurxmaty{\mathbb{X}}}$, the square $$\vrenter{\xymatrix{{\triangje}({\ensuremdtk{\mathbb{X}}},n) \ar[r]^-{\widehat{\pqrrial}_{i}} \ar[d]_-{{\[artkql}_{i}} & {\rzixegox{.145mm}{\sfalxbox{1.3}[1.15]{$\smallhodn$}}\!}^{i}({\ensuremarh{\mathbb{X}}},n) \ar[d]^-{{\partisl}_{y-1}^{p}\yimes {\partiam}_{i}^{n-i}} \\ {\egstremath{\mathbb{X}}}_{n-1} \ar[r]_-{{\lgroup}{\partial}_{j}{\rgroup}_{b}} & {\friangle}({\ensuremath{\mathbv{X}}},n-1)}}$$ is a pullback, whete the arrjw on the right is the restriction of $${\partial}_{i-1}^{i}\tikes {\pergiao}_{i}^{k-i}=\unawrhrace{{\partial}_{i-1}\times \cdots\times {\partial}_{i-1}}_{\text{$i$ tyjex}} \nimes \underbrace{{\pcrtial}_{i}\times \cdoys \toies {\partial}_{i}}_{\tgxt{$n-i$ tnjea}}\colon {\ensuremath{\mwthbb{X}}}_{n-1}^{g}\to {\ebsuremath{\iathnb{X}}}_{n-2}^{n}$$ to a morphism ${\raiseboz{.145mm}{\scalebox{1.3}[1.15]{$\siqllhorn$}}\!}^{i}({\ensuremath{\lathbb{X}}},n)\to {\criangke}({\ensiremath{\mathbb{X}}},n-1)$. Here is c picthre in degrfe $n=2$ for $j=1$: $$\begin{matrix} \vcenger{\qymadrix@1@!0@R=2.44

Kelly-Pedicchio Theorem 4.2]. Furthermore, a regular epimorphism objects a regular category is always The property for simplicial may also be in terms of higher extensions: in semi-abelian category, a simplicial object ${\ensuremath{\mathbb{X}}}$ is Kan if and only if all its truncations, considered as higher arrows in all possible directions, have a domain is extension \[Lemma-DeltaLambda-Square\] a finitely complete category, given $n\geq 1$, $i\in n$, and an augmented simplicial object ${\ensuremath{\mathbb{X}}}$, the $$\vcenter{\xymatrix{{\triangle}({\ensuremath{\mathbb{X}}},n) \ar[r]^-{\widehat{\partial}_{i}} \ar[d]_-{{\partial}_{i}} & {\raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n) \ar[d]^-{{\partial}_{i-1}^{i}\times {\partial}_{i}^{n-i}} \\ \ar[r]_-{{\lgroup}{\partial}_{j}{\rgroup}_{j}} & {\triangle}({\ensuremath{\mathbb{X}}},n-1)}}$$ is pullback, where the arrow on right the restriction $${\partial}_{i-1}^{i}\times \cdots\times times}} \times \underbrace{{\partial}_{i}\times \times {\partial}_{i}}_{\text{$n-i$ times}}\colon {\ensuremath{\mathbb{X}}}_{n-1}^{n}\to {\ensuremath{\mathbb{X}}}_{n-2}^{n}$$ to a morphism ${\raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n)\to {\triangle}({\ensuremath{\mathbb{X}}},n-1)$. Here is a picture in degree $n=2$ $i=1$: $$\begin{matrix}

Kelly-Pedicchio Theorem 4.2]. FurtHermore, a reGular EpiMorPhIsm oF simPlicial objects IN a reGular Mal’tsev category is AlwayS a FIbraTIoN [@EvervdL2 PropOSiTIOn 4.4]. THe kaN prOpERtY for sImpLicial oBjects may aLso Be Expressed in tERmS of higher eXteNsions: in a semI-abElian cAtEgoRY, a simPliCial oBject ${\eNSuremaTh{\mathbb{X}}}$ Is kAn if anD Only if aLL Of Its tRuncations, consideREd AS higher arrows iN all poSsIBlE DIreCtiOns, have a doMaIn whiCH is an exTEnSION [@EGOEVdL]. \[Lemma-DeltALambda-SquaRE\] In A finitElY coMPlete cAtegoRy, GIveN $n\geq 1$, $i\in n$, anD an aUgmented sImplicIAl objecT ${\EnsuremAth{\matHbb{x}}}$, thE squARe $$\VcEntEr{\XYmaTRiX{{\trIAngLe}({\ensureMaTh{\MathbB{X}}},n) \aR[R]^-{\WIDehaT{\paRtiaL}_{i}} \ar[d]_-{{\Partial}_{i}} & {\raiseBox{.145Mm}{\scALebOx{1.3}[1.15]{$\smaLlhorN$}}\!}^{i}({\enSuRematH{\mathbB{X}}},n) \ar[D]^-{{\pArtial}_{i-1}^{i}\times {\paRtiaL}_{i}^{n-i}} \\ {\ensurEmaTh{\MatHbB{X}}}_{n-1} \ar[R]_-{{\Lgroup}{\ParTiaL}_{j}{\rgrouP}_{j}} & {\trianGLe}({\eNsUREMaTh{\mathbb{X}}},n-1)}}$$ is a pullbAcK, WHeRe the arrOw on thE RiGhT Is the resTrIctIon oF $${\PArtiaL}_{i-1}^{i}\tIMeS {\partial}_{I}^{n-i}=\undERbRaCe{{\partiAl}_{I-1}\times \CdOts\TimEs {\parTIal}_{i-1}}_{\Text{$i$ tImes}} \timeS \undeRBrace{{\partial}_{i}\tIMes \cdots \times {\PArTIAl}_{I}}_{\Text{$N-i$ tImes}}\colon {\enSureMAth{\mAthbB{x}}}_{n-1}^{N}\to {\ENsureMath{\mAtHBb{x}}}_{N-2}^{n}$$ to a morphism ${\raisebOx{.145Mm}{\scalEbox{1.3}[1.15]{$\sMallhorn$}}\!}^{i}({\ensuRemath{\mathBB{x}}},N)\to {\trianGle}({\eNSuREmath{\mathbb{X}}},n-1)$. HEre is A picture in DEgree $n=2$ foR $i=1$: $$\begIn{matrix} \Vcenter{\xyMATrix@1@!0@R=2.44

Kelly-Pedicchio Theorem 4. 2]. Furthe rmore , a re gu larepim orphism of sim p lici al objects in a regula r Mal ’t s ev c a te goryis alwa y sa fib ra ti on[@ E ve rVdL2 Pr opositi on 4.4]. T heKa n property f o rsimplicial ob jects may al sobe exp re sse d in t erm s ofhigher extens ions: inas emi-ab e lian ca t e go ry,a simplicial obje c t$ {\ensuremath{\ mathbb {X } }} $ isKan if and on ly if a l l of it s t r u n cat i ons, consider ed as highe r ar rows i nall possib le di re c tio ns, have adoma in whichis ane xtensio n  [@EGoe VdL]. \[ Lem ma-D e lt aL amb da - Squ a re \]I n a finitel yco mplet e ca t e g o ry,giv en $ n\geq 1$, $i\in n$ , a nd a n au gment ed si mpli ci al ob ject $ {\ens ur emath{\mathbb{X }}}$ , the squ are $ $\v ce nter{ \ xymatr ix{ {\t riangle }({\ens u rem at h { \ ma thbb{X}}},n) \ar[r ]^ - { \w idehat{\ partia l }_ {i } } \ar[d] _- {{\ part i a l}_{i }} & {\ raisebox {.145m m }{ \s calebox {1 .3}[1. 15 ]{$ \sm allho r n$}} \!}^{i }({\ensu remat h {\mathbb{X}}}, n ) \ar[d]^-{{\ p ar t i al } _{i- 1}^ {i}\times { \par t ial} _{i} ^ {n -i} } \\ { \ensu re m at h {\mathbb{X}}}_{n-1} \ ar[r]_ -{{\l group}{\parti al}_{j}{\r g r o up}_{j}} & { \ tr i angle}({\ensur emath {\mathbb{X } }},n-1)} }$$ i s a pull back, whe r e the arr owonthe ri g h tis the restri c t ionof $${\pa rti al}_{i- 1}^ {i} \ti mes { \partial} _{i}^{n- i} =\ un de rbr ace{{ \ partial} _{ i-1 }\ tim es \c d ots\ti mes { \par ti al } _{i -1}}_{\ t ex t { $i$ti me s}}\ti me s \un derb r ace {{\part ial}_{i}\ tim e s \c do ts \times {\partial}_{ i} }_{\text{$ n- i$times} } \ colon {\ ensuremath{\mathbb{X}}} _ {n-1}^{ n}\ to {\ ensu remath{\m ath bb{X}} }_{ n -2}^{n }$$ to a mo rp his m ${\ra i s eb ox{ .1 45mm}{\sca l e box {1.3} [1 .15] {$\smal lhorn$}}\!}^{i}({\ e nsu remath{\mathb b{X }}}, n ) \t o { \ tr i ang le } ({\ e n suremath{\mathb b{X}}},n-1 )$ . Here is ap ict ur e in de gree $n =2$ f o r $i=1$ : $$\begi n{matrix} \ vcen t e r{\ xymatrix@1 @!0@R=2. 44

Kelly-Pedicchio Theorem 4.2]._Furthermore, a_regular epimorphism of simplicial_objects in_a_regular Mal’tsev_category_is always a_fibration [@EverVdL2 Proposition 4.4]. The_Kan property for simplicial_objects may also_be_expressed in terms of higher extensions: in a semi-abelian category, a simplicial object ${\ensuremath{\mathbb{X}}}$_is_Kan if_and_only_if all of its truncations,_considered as higher arrows in_all possible_directions, have a domain which is an extension [@EGoeVdL]. \[Lemma-DeltaLambda-Square\]_In_a finitely complete_category, given $n\geq 1$, $i\in n$, and an augmented_simplicial object ${\ensuremath{\mathbb{X}}}$, the square $$\vcenter{\xymatrix{{\triangle}({\ensuremath{\mathbb{X}}},n)_\ar[r]^-{\widehat{\partial}_{i}} \ar[d]_-{{\partial}_{i}} &_{\raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n)_\ar[d]^-{{\partial}_{i-1}^{i}\times_{\partial}_{i}^{n-i}} \\ {\ensuremath{\mathbb{X}}}_{n-1} \ar[r]_-{{\lgroup}{\partial}_{j}{\rgroup}_{j}} &_{\triangle}({\ensuremath{\mathbb{X}}},n-1)}}$$ is a pullback, where the_arrow on the right is the_restriction of $${\partial}_{i-1}^{i}\times {\partial}_{i}^{n-i}=\underbrace{{\partial}_{i-1}\times \cdots\times {\partial}_{i-1}}_{\text{$i$ times}}_\times \underbrace{{\partial}_{i}\times \cdots \times {\partial}_{i}}_{\text{$n-i$ times}}\colon_{\ensuremath{\mathbb{X}}}_{n-1}^{n}\to {\ensuremath{\mathbb{X}}}_{n-2}^{n}$$ to a morphism_${\raisebox{.145mm}{\scalebox{1.3}[1.15]{$\smallhorn$}}\!}^{i}({\ensuremath{\mathbb{X}}},n)\to {\triangle}({\ensuremath{\mathbb{X}}},n-1)$. Here_is a picture in degree_$n=2$ for $i=1$:_$$\begin{matrix} \vcenter{\xymatrix@1@!0@R=2.44

$n \in \mathbb{Z}$. Then $H^n(\operatorname*{hocolim}_{i \in I} \mathscr{X}) \simeq \varinjlim_{i \in I}H^n(\mathscr{X}_i) \in \mathcal{V}_n$ for all $n \in \mathbb{Z}$, and thus $\operatorname*{hocolim}_{i \in I} \mathscr{X} \in \mathcal{V}$. Similar argument shows that $\mathcal{V}$ is closed under products. Finally, consider a pure monomorphism $f: Y \rightarrow X$ in $\operatorname*{\mathbf{D}}(R)$ with $X \in \mathcal{V}$. For each $n \in \mathbb{Z}$ we have that $\operatorname*{Hom}_R(R[-n],f) \simeq H^n(f)$ is a pure monomorphism of $R$-modules by [@P 17.3.17]. Therefore, $H^n(Y) \in \mathcal{V}_n$ for all $n \in \mathbb{Z}$, and therefore $Y \in \mathcal{V}$. Using Theorem \[T:rosie1\] we conclude that $\mathcal{V}$ is a definable subcategory. This establishes the correspondence. Ziegler spectra --------------- We can reformulate Proposition \[P:definablecorr\] using the Ziegler spectra of the derived category and of the module category. We refer to [@P] for the theory of Ziegler spectra of module categories. If $R$ is a ring, the natural embedding $\operatorname*{Mod-R}[-n] \subseteq \operatorname*{\mathbf{D}}(R)$ for some $n \in \mathbb{Z}$ induces a closed embedding $\operatorname*{Zg}(R)[-n] \rightarrow \operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$. Clearly, $\bigcup_{n \in \mathbb{Z}}\operatorname*{Zg}(R)[-n]$ forms a disjoint union inside $\operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$. One can ask for which rings it is true that $\operatorname*{Zg}(\operatorname*{\mathbf{D}}(R)) = \bigcup_{n \in \mathbb{Z}}\operatorname*{Zg}(R)[-n]$. Equivalently, for which rings is it true that every indecomposable

$ n \in \mathbb{Z}$. Then $ H^n(\operatorname*{hocolim}_{i \in I } \mathscr{X }) \simeq \varinjlim_{i \in I}H^n(\mathscr{X}_i) \in \mathcal{V}_n$ for all $ n \in \mathbb{Z}$, and thus $ \operatorname*{hocolim}_{i \in I } \mathscr{X } \in \mathcal{V}$. Similar controversy usher that $ \mathcal{V}$ is closed under products. Finally, study a pure monomorphism $ f: Y \rightarrow X$ in $ \operatorname*{\mathbf{D}}(R)$ with $ X \in \mathcal{V}$. For each $ normality \in \mathbb{Z}$ we give birth that $ \operatorname*{Hom}_R(R[-n],f) \simeq H^n(f)$ is a pure monomorphism of $ R$-modules by [ @P 17.3.17 ]. Therefore, $ H^n(Y) \in \mathcal{V}_n$ for all $ n \in \mathbb{Z}$, and therefore $ Y \in \mathcal{V}$. use Theorem   \[T: rosie1\ ] we conclude that $ \mathcal{V}$ is a definable subcategory. This establishes the commensurateness. Ziegler spectra --------------- We can redevelop Proposition   \[P: definablecorr\ ] using the Ziegler spectra of the derived class and of the module class. We mention to [ @P ] for the theory of Ziegler spectra of module categories. If $ R$ is a band, the natural embedding $ \operatorname*{Mod - R}[-n ] \subseteq \operatorname*{\mathbf{D}}(R)$ for some $ n \in \mathbb{Z}$ induces a closed implant $ \operatorname*{Zg}(R)[-n ] \rightarrow \operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$. Clearly, $ \bigcup_{n \in \mathbb{Z}}\operatorname*{Zg}(R)[-n]$ forms a disjoint union inside $ \operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$. One can ask for which rings it is genuine that $ \operatorname*{Zg}(\operatorname*{\mathbf{D}}(R) ) = \bigcup_{n \in \mathbb{Z}}\operatorname*{Zg}(R)[-n]$. Equivalently, for which rings is it true that every indecomposable

$n \ln \mathbb{Z}$. Then $H^n(\operaturname*{hocolim}_{i \nb I} \mavhscr{X}) \aimeq \vafinjlim_{i \in I}H^n(\mathscr{X}_i) \in \larhcal{C}_n$ for all $n \in \mathbb{X}$, and thud $\operatirnanw*{hocolim}_{i \mh I} \matmfcr{X} \ln \machral{V}$. Similar arnument showv that $\mathcal{E}$ ks closed under products. Finally, consyder a luge monomorphisi $f: J \wighfarrow X$ in $\operatorname*{\mathbf{D}}(R)$ wjth $X \ii \mathcal{V}$. For rach $n \in \mathbb{Z}$ we have hhat $\operatorname*{Hom}_R(R[-j],f) \simeq H^n(d)$ is q pure monomurphism of $R$-modules by [@P 17.3.17]. Therefore, $H^n(Y) \in \mathcal{V}_n$ wor akl $n \in \majkvb{Z}$, dnd therefoce $Y \ig \mathcal{V}$. Using Themrem \[T:roxie1\] we concludc thav $\marhcal{V}$ is a definable subcategory. This gstablishev che correspondence. Zietlwr spgctra --------------- Fe cxb rdfodmnlafe Prooosmtion \[P:definzblecorr\] usung the Ziegler spevtwq of the deribed caeedory and of the module category. We refeg to [@P] for the theory of Ziwgler spectra of modupe categowies. If $R$ is a ring, the natural embedding $\operatosname*{Jud-R}[-u] \subsdree \operatorname*{\mathbf{D}}(R)$ for some $n \in \mathbb{Z}$ yhdicvs a closed embedbing $\operatornamr*{Zh}(R)[-m] \rightarrow \okeratoruzms*{Zg}(\operatorname*{\matjbf{D}}(R))$. Cjearlt, $\bigcup_{n \in \kathbb{Z}}\operatorname*{Zg}(R)[-n]$ forns a disjoinn unuon inside $\operatoxname*{Zg}(\operacornamg*{\mathbg{D}}(R))$. One can ask for whieh rinfs it is trke that $\oldratorname*{Zg}(\operxtognama*{\mathbf{D}}(R)) = \bigcup_{n \in \mathfb{Z}}\operatirnake*{Zg}(R)[-n]$. Dquifalentjy, for whifh rikcs is it true that everi indewomposable

$n \in \mathbb{Z}$. Then $H^n(\operatorname*{hocolim}_{i \in I} \varinjlim_{i I}H^n(\mathscr{X}_i) \in for all $n \in \mathscr{X} \in \mathcal{V}$. argument shows that is closed under products. Finally, consider pure monomorphism $f: Y \rightarrow X$ in $\operatorname*{\mathbf{D}}(R)$ with $X \in \mathcal{V}$. For $n \in \mathbb{Z}$ we have that $\operatorname*{Hom}_R(R[-n],f) \simeq H^n(f)$ is a pure monomorphism $R$-modules [@P Therefore, \in \mathcal{V}_n$ for all $n \in \mathbb{Z}$, and therefore $Y \in \mathcal{V}$. Using Theorem \[T:rosie1\] we that $\mathcal{V}$ is a definable subcategory. This establishes correspondence. Ziegler spectra --------------- can reformulate Proposition \[P:definablecorr\] using Ziegler of the category of module category. We to [@P] for the theory of Ziegler spectra of module categories. If $R$ is a ring, the embedding $\operatorname*{Mod-R}[-n] for some \in induces closed embedding $\operatorname*{Zg}(R)[-n] Clearly, $\bigcup_{n \in \mathbb{Z}}\operatorname*{Zg}(R)[-n]$ forms a $\operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$. One can ask for which rings it true that = \bigcup_{n \in \mathbb{Z}}\operatorname*{Zg}(R)[-n]$. Equivalently, for rings is it true that every indecomposable

$n \in \mathbb{Z}$. Then $H^n(\operatornAme*{hocolim}_{I \in I} \mAthScr{x}) \sImeq \VariNjlim_{i \in I}H^n(\matHScr{X}_I) \in \mathcal{V}_n$ for all $n \in \mAthbb{z}$, aND thuS $\OpEratoRname*{hoCOlIM}_{I \in i} \mAtHscR{X} \IN \mAthcaL{V}$. SImilar aRgument shoWs tHaT $\mathcal{V}$ is cLOsEd under proDucTs. Finally, conSidEr a purE mOnoMOrphiSm $f: y \righTarrow x$ In $\operAtorname*{\mAtHBf{D}}(R)$ wiTH $X \in \matHCAl{v}$. For Each $n \in \mathbb{Z}$ we hAVe THat $\operatornamE*{Hom}_R(R[-N],f) \SImEQ h^n(f)$ Is a Pure monomoRpHism oF $r$-moduleS By [@p 17.3.17]. tHEreFOre, $H^n(Y) \in \mathcAl{V}_n$ for all $n \IN \maThbb{Z}$, aNd TheREfore $Y \In \matHcAL{V}$. USing Theorem \[t:rosIe1\] we conclUde thaT $\Mathcal{v}$ Is a defiNable sUbcAteGory. tHiS eStaBlISheS ThE coRResPondence. ziEgLer spEctrA --------------- wE CAn reForMulaTe ProPosition \[P:defiNabLecoRR\] usIng thE ZiegLer sPeCtra oF the deRived CaTegory and of the mOdulE category. we rEfEr tO [@P] For thE Theory Of ZIegLer specTra of moDUle CaTEGOrIes. If $R$ is a ring, the naTuRAL eMbedding $\OperatORnAmE*{mod-R}[-n] \subSeTeq \OperATOrnamE*{\matHBf{d}}(R)$ for somE $n \in \maTHbB{Z}$ Induces A cLosed eMbEddIng $\OperaTOrnaMe*{Zg}(R)[-n] \RightarrOw \opeRAtorname*{Zg}(\operATorname*{\mathbf{d}}(r))$. CLEArLY, $\bigCup_{N \in \mathbb{Z}}\oPeraTOrnaMe*{Zg}(r)[-N]$ fOrmS A disjOint uNiON iNSide $\operatorname*{Zg}(\oPeRatornAme*{\maThbf{D}}(R))$. One can aSk for which RINGs it is trUe thAT $\oPEratorname*{Zg}(\opEratoRname*{\mathbF{d}}(R)) = \bigcup_{N \in \maThbb{Z}}\opeRatorname*{zG}(r)[-n]$. EquivaLenTly, For WhiCH RiNgs is it true thAT EverY iNdecompOsaBle

$n \in \mathbb{Z}$. Then$H^n(\oper atorn ame *{h oc olim }_{i \in I} \maths c r{X} ) \simeq \varinjlim_{i \inI} H ^n(\ m at hscr{ X}_i) \ i n\ m ath ca l{ V}_ n$ fo r all $n \in \m athbb{Z}$, an dthus $\opera t or name*{hoco lim }_{i \in I}\ma thscr{ X} \i n \mat hca l{V}$ . Simi l ar arg ument sho ws that $ \ mathcal { V }$ isclosed under prod u ct s . Finally, con siderap ur e mon omo rphism $f: Y \rig h tarrowX $i n $\o p eratorname*{\ mathbf{D}}( R )$with $ X\in \mathc al{V} $. For each $n \i n \m athbb{Z}$ we ha v e that$ \operat orname *{H om} _R(R [ -n ], f)\s i meq H^ n(f ) $ i s a pure m on omorp hism o f $R$- mod ules by [ @P 17.3.17].The refo r e,$H^n( Y) \i n \m at hcal{ V}_n$for a ll $n \in \mathbb {Z}$ , and the ref or e $ Y\in \ m athcal {V} $.Using T heorem\ [T: ro s i e 1\ ] we conclude that $ \ m at hcal{V}$ is ad ef in a ble subc at ego ry.T h is es tabl i sh es the c orresp o nd en ce. Zi eg ler sp ec tra -- ----- - ---- --- W e can re formu l ate Propositio n  \[P:definabl e co r r \] usin g t he Zieglerspec t ra o f th e d eri v ed ca tegor ya nd of the module categ or y. Werefer to [@P] forthe theory o f Ziegler spe c tr a of module cat egori es. If $R$ is a rin g, th e natura l embeddi n g $\opera tor nam e*{ Mod - R }[ -n] \subseteq \ oper at orname* {\m athbf{D }}( R)$ fo r s om e $n \in\mathbb{ Z} $in du ces a cl o sed embe dd ing $ \op erato r name*{ Zg}(R )[-n ]\r i ght arrow \ o pe r a torn am e* {Zg} (\o pe rator name * {\m athbf{D }}(R))$.Cle a rly, $ \b igcup_{ n \in \mathbb {Z }}\operato rn ame *{Zg}( R ) [-n]$ fo rms a disjoint union in s ide $\o per atorn ame* {Zg}(\ope rat orname *{\ m athbf{ D}}(R) )$. O ne ca n ask f o r w hic hrings it i s tru e tha t$\op eratorn ame*{Zg}(\operator n ame *{\mathbf{D}} (R) ) =\ b ig cup _ {n \in \ m ath b b {Z}}\operatorna me*{Zg}(R) [- n ]$ . Equivale n tly ,for whi ch ring s isi t truethat ever y indecom po sabl e

$n_\in \mathbb{Z}$._Then $H^n(\operatorname*{hocolim}_{i \in I}_\mathscr{X}) \simeq_\varinjlim_{i_\in I}H^n(\mathscr{X}_i)_\in_\mathcal{V}_n$ for all_$n \in \mathbb{Z}$,_and thus $\operatorname*{hocolim}_{i \in_I} \mathscr{X} \in_\mathcal{V}$._Similar argument shows that $\mathcal{V}$ is closed under products. Finally, consider a pure monomorphism_$f:_Y \rightarrow_X$_in_$\operatorname*{\mathbf{D}}(R)$ with $X \in \mathcal{V}$._For each $n \in \mathbb{Z}$_we have_that $\operatorname*{Hom}_R(R[-n],f) \simeq H^n(f)$ is a pure monomorphism_of_$R$-modules by [@P_17.3.17]. Therefore, $H^n(Y) \in \mathcal{V}_n$ for all $n \in_\mathbb{Z}$, and therefore $Y \in \mathcal{V}$._Using Theorem \[T:rosie1\] we_conclude_that_$\mathcal{V}$ is a definable_subcategory. This establishes the correspondence. Ziegler spectra --------------- We_can reformulate Proposition \[P:definablecorr\] using the Ziegler_spectra of the derived category and of_the module category. We refer to_[@P] for the theory of_Ziegler spectra_of module categories. If $R$_is a ring,_the natural_embedding $\operatorname*{Mod-R}[-n] \subseteq_\operatorname*{\mathbf{D}}(R)$ for some $n \in \mathbb{Z}$_induces a closed_embedding $\operatorname*{Zg}(R)[-n] \rightarrow \operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$. Clearly, $\bigcup_{n_\in_\mathbb{Z}}\operatorname*{Zg}(R)[-n]$ forms a_disjoint_union_inside $\operatorname*{Zg}(\operatorname*{\mathbf{D}}(R))$._One can ask_for_which rings_it_is true that $\operatorname*{Zg}(\operatorname*{\mathbf{D}}(R)) = \bigcup_{n_\in_\mathbb{Z}}\operatorname*{Zg}(R)[-n]$. Equivalently, for which rings is it_true that every indecomposable

$f_t^{\mathcal{D}}$ as $t$ tends to $0$. Recall that $f_t^{\mathcal{D}}= { \mathrm{Sym}}_q\left({\mathop \mathrm{Uni}}\left(M_tz^{A_t}\right)\right)$ where $M_t$ tends to ${\mathrm{I}}_2$ as $t$ tends to $0$. Hence, the limit axis of $f_t^{\mathcal{D}}$ and $\widetilde{f}_t^{\mathcal{D}}:= { \mathrm{Sym}}_q\left({\mathop \mathrm{Uni}}\left(z^{A_t}\right)\right)$ are the same. Two cases can occur, wether $r> s$ or $r< s$. #### Spherical family. At $t=0$, $r=\frac{1}{2}$ and $s=0$. The limit potential is thus $$\xi_0(z,{\lambda}) = \begin{pmatrix} 0 & \frac{{\lambda}^{-1}}{2} \\ \frac{{\lambda}}{2} & 0 \end{pmatrix}z^{-1} dz.$$ Consider the gauge $$G(z,{\lambda}) = \frac{1}{\sqrt{2z}}\begin{pmatrix} 1 & 0\\ {\lambda}& 2z \end{pmatrix}.$$ The gauge potential is then $$\xi_0\cdot G (z,{\lambda}) = \begin{pmatrix} 0 & {\lambda}^{-1}dz\\ 0&0 \end{pmatrix} = \xi_{\mathbb{S}}(z,{\lambda})$$ where $\xi_{\mathbb{S}}$ is the spherical potential as in Section \[sectionDPWMethod\]. Let $\widetilde{\Phi}:=z^{A_0}G$ be the gauged holomorphic frame and compute $$\begin{aligned} \widetilde{\Phi}(1,{\lambda}) &= G(1,{\lambda}) \\ &= \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & 0\\ {\lambda}& 2 \end{pmatrix} = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & -{\lambda}^{-1} \\ {\lambda}& 1 \end{pmatrix}\begin{pmatrix} 1 & {\lambda}^{-1} \\ 0 & 1 \end{pmatrix}\\ &= H({\lambda})\Phi_{\mathbb{

$ f_t^{\mathcal{D}}$ as $ t$ tends to $ 0$. Recall that $ f_t^{\mathcal{D}}= { \mathrm{Sym}}_q\left({\mathop \mathrm{Uni}}\left(M_tz^{A_t}\right)\right)$ where $ M_t$ tends to $ { \mathrm{I}}_2 $ as $ t$ tends to $ 0$. Hence, the terminus ad quem bloc of $ f_t^{\mathcal{D}}$ and $ \widetilde{f}_t^{\mathcal{D}}:= { \mathrm{Sym}}_q\left({\mathop \mathrm{Uni}}\left(z^{A_t}\right)\right)$ are the like. Two cases can occur, wether $ r > s$ or $ r < s$. # # # # Spherical kin. At $ t=0 $, $ r=\frac{1}{2}$ and $ s=0$. The limit potential is therefore $ $ \xi_0(z,{\lambda }) = \begin{pmatrix } 0 & \frac{{\lambda}^{-1}}{2 } \\ \frac{{\lambda}}{2 } & 0 \end{pmatrix}z^{-1 } dz.$$ Consider the gauge $ $ G(z,{\lambda }) = \frac{1}{\sqrt{2z}}\begin{pmatrix } 1 & 0\\ { \lambda } & 2z \end{pmatrix}.$$ The gauge electric potential is then $ $ \xi_0\cdot G (z,{\lambda }) = \begin{pmatrix } 0 & { \lambda}^{-1}dz\\ 0&0 \end{pmatrix } = \xi_{\mathbb{S}}(z,{\lambda})$$ where $ \xi_{\mathbb{S}}$ is the ball-shaped potential as in Section \[sectionDPWMethod\ ]. Let $ \widetilde{\Phi}:=z^{A_0}G$ be the gauged holomorphic skeletal system and compute $ $ \begin{aligned } \widetilde{\Phi}(1,{\lambda }) & = G(1,{\lambda }) \\ & = \frac{1}{\sqrt{2}}\begin{pmatrix } 1 & 0\\ { \lambda } & 2 \end{pmatrix } = \frac{1}{\sqrt{2}}\begin{pmatrix } 1 & -{\lambda}^{-1 } \\ { \lambda } & 1 \end{pmatrix}\begin{pmatrix } 1 & { \lambda}^{-1 } \\ 0 & 1 \end{pmatrix}\\ & = H({\lambda})\Phi_{\mathbb {

$f_t^{\lathcal{D}}$ as $t$ tends to $0$. Vecall that $f_t^{\majhxal{D}}= { \kathrm{Aym}}_q\left({\oathop \mathrm{Uni}}\left(M_tz^{A_t}\rigit)\ritht)$ wyere $M_t$ tends to ${\mathro{I}}_2$ as $t$ tvnds to $0$. Yenct, the limit axis of $f_t^{\matmeal{D}}$ zkd $\wibevilde{f}_t^{\mathcal{D}}:= { \mathrm{Syk}}_q\left({\mathop \mdtfrl{Uni}}\left(z^{A_t}\right)\right)$ are the same. Ewo casrs can occur, wetrer $g> f$ or $g< w$. #### Spherical family. At $t=0$, $r=\frac{1}{2}$ znd $s=0$. Tie limit potentoal is thus $$\xi_0(z,{\lambda}) = \begln{pmwtrix} 0 & \frac{{\lambda}^{-1}}{2} \\ \vrac{{\lambda}}{2} & 0 \end{[natrix}z^{-1} dz.$$ Cunsider tht yauge $$G(z,{\lamgda}) = \frac{1}{\sqrt{2z}}\begin{pmatrix} 1 & 0\\ {\lamcda}& 2z \znd{pmatrix}.$$ Jkw gwoge potentiao is nhen $$\xi_0\cdot G (z,{\lambda}) = \begin{lmatrix} 0 & {\lambds}^{-1}dz\\ 0&0 \xnd{pnatrix} = \xi_{\mathbb{S}}(z,{\lamuda})$$ where $\xi_{\mathbb{S}}$ ys the spvexical potential as in Swctiot \[sewtiovEPWOetgov\]. Lst $\widftimde{\Phi}:=z^{A_0}G$ ge the gaugwd holomorphic framt age compute $$\begjn{aliggeq} \widetilde{\Phi}(1,{\lambda}) &= G(1,{\lambda}) \\ &= \frac{1}{\sqrt{2}}\tegjn{pmatrix} 1 & 0\\ {\lambda}& 2 \end{pmqtrix} = \frac{1}{\sqrt{2}}\begin{platrix} 1 & -{\lwmbda}^{-1} \\ {\lambda}& 1 \end{pmatrix}\begin{pmatrix} 1 & {\lambda}^{-1} \\ 0 & 1 \end{[matrmx}\\ &= H({\lcnbda})\Pfu_{\mwthbb{

$f_t^{\mathcal{D}}$ as $t$ tends to $0$. Recall { \mathrm{Uni}}\left(M_tz^{A_t}\right)\right)$ where tends to ${\mathrm{I}}_2$ Hence, limit axis of and $\widetilde{f}_t^{\mathcal{D}}:= { \mathrm{Uni}}\left(z^{A_t}\right)\right)$ are the same. Two cases occur, wether $r> s$ or $r< s$. #### Spherical family. At $t=0$, $r=\frac{1}{2}$ $s=0$. The limit potential is thus $$\xi_0(z,{\lambda}) = \begin{pmatrix} 0 & \frac{{\lambda}^{-1}}{2} \\ & \end{pmatrix}z^{-1} Consider gauge $$G(z,{\lambda}) = \frac{1}{\sqrt{2z}}\begin{pmatrix} 1 & 0\\ {\lambda}& 2z \end{pmatrix}.$$ The gauge potential is then $$\xi_0\cdot (z,{\lambda}) = \begin{pmatrix} 0 & {\lambda}^{-1}dz\\ 0&0 \end{pmatrix} \xi_{\mathbb{S}}(z,{\lambda})$$ where $\xi_{\mathbb{S}}$ is spherical potential as in Section Let be the holomorphic and $$\begin{aligned} \widetilde{\Phi}(1,{\lambda}) &= \\ &= \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & 0\\ {\lambda}& 2 \end{pmatrix} = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & -{\lambda}^{-1} \\ {\lambda}& 1 1 & 0 & \end{pmatrix}\\ H({\lambda})\Phi_{\mathbb{

$f_t^{\mathcal{D}}$ as $t$ tends to $0$. RecalL that $f_t^{\matHcal{D}}= { \MatHrm{syM}}_q\leFt({\maThop \mathrm{Uni}}\lEFt(M_tZ^{A_t}\right)\right)$ where $M_t$ teNds to ${\MaTHrm{I}}_2$ AS $t$ Tends To $0$. Hence, THe LIMit AxIs Of $f_T^{\mAThCal{D}}$ aNd $\wIdetildE{f}_t^{\mathcal{d}}:= { \maThRm{Sym}}_q\left({\maTHoP \mathrm{Uni}}\LefT(z^{A_t}\right)\rigHt)$ aRe the sAmE. TwO Cases Can Occur, Wether $R> S$ or $r< s$. #### SPherical fAmILy. At $t=0$, $r=\FRac{1}{2}$ and $s=0$. tHE lImit Potential is thus $$\xi_0(Z,{\LaMBda}) = \begin{pmatriX} 0 & \frac{{\lAmBDa}^{-1}}{2} \\ \FRAc{{\lAmbDa}}{2} & 0 \end{pmatrIx}Z^{-1} dz.$$ CoNSider thE GaUGE $$g(z,{\lAMbda}) = \frac{1}{\sqrt{2z}}\Begin{pmatriX} 1 & 0\\ {\LamBda}& 2z \enD{pMatRIx}.$$ The gAuge pOtENtiAl is then $$\xi_0\cDot G (Z,{\lambda}) = \beGin{pmaTRix} 0 & {\lambDA}^{-1}dz\\ 0&0 \end{pMatrix} = \Xi_{\mAthBb{S}}(z,{\LAmBdA})$$ whErE $\Xi_{\mAThBb{S}}$ IS thE sphericAl PoTentiAl as IN sECtioN \[seCtioNDPWMEthod\]. Let $\widetIldE{\Phi}:=Z^{a_0}G$ bE the gAuged HoloMoRphic Frame aNd comPuTe $$\begin{aligned} \wIdetIlde{\Phi}(1,{\laMbdA}) &= G(1,{\LamBdA}) \\ &= \frac{1}{\SQrt{2}}\begIn{pMatRix} 1 & 0\\ {\lambDa}& 2 \end{pmATriX} = \fRAC{1}{\SqRt{2}}\begin{pmatrix} 1 & -{\lambDa}^{-1} \\ {\LAMbDa}& 1 \end{pmaTrix}\beGIn{PmATrix} 1 & {\lambDa}^{-1} \\ 0 & 1 \End{PmatRIX}\\ &= H({\lamBda})\PHI_{\mAthbb{

$f_t^{\mathcal{D}}$ as $t $ tends to $0$. Re cal lthat $f_ t^{\mathcal{D} } = {\mathrm{Sym}}_q\left({ \math op \mat h rm {Uni} }\left( M _t z ^ {A_ t} \r igh t) \ ri ght)$ wh ere $M_ t$ tends t o $ {\ mathrm{I}}_2 $ a s $t$ tend s t o $0$. Hence , t he lim it ax i s of$f_ t^{\m athcal { D}}$ a nd $\wide ti l de{f}_ t ^{\math c a l{ D}}: = { \mathrm{Sym}} _ q\ l eft({\mathop \ mathrm {U n i} } \ lef t(z ^{A_t}\rig ht )\rig h t)$ are th e s ame . Two cases ca n occur, we t her $r> s $or$ r< s$. ### #S phe rical famil y. At $t=0$, $r=\f r ac{1}{2 } $ and $ s=0$.The li mitp ot en tia li s t h us $$ \ xi_ 0(z,{\la mb da }) =\beg i n { p matr ix} 0 & \fra c{{\lambda}^{ -1} }{2} \\\frac {{\la mbda }} {2} & 0 \en d{pma tr ix}z^{-1} dz.$$ Con sider the ga ug e $ $G (z,{\ l ambda} ) = \f rac{1}{ \sqrt{2 z }}\ be g i n {p matrix} 1 & 0\\ {\ la m b da }& 2z \e nd{pma t ri x} . $$ The g au gepote n t ial i s th e n$$\xi_0\ cdot G (z ,{ \lambda }) = \be gi n{p mat rix}0 & { \lambd a}^{-1}d z\\ 0 & 0 \end{pmatrix } = \xi_{\math b b{ S } }( z ,{\l amb da})$$ wher e $\ x i_{\ math b b{ S}} $ is t he sp he r ic a l potential as in S ec tion \ [sect ionDPWMethod\ ]. Let $\w i d e tilde{\P hi}: = z^ { A_0}G$ be thegauge d holomorp h ic frame andcompute$$\begin{ a l igned} \ wid eti lde {\P h i }( 1,{\lambda})& = G( 1, {\lambd a}) \\ &=\fr ac{ 1}{ \sq rt {2}}\begi n{pmatri x} 1 & 0 \\{\lam b da}& 2 \ en d{p ma tri x} =\ frac{1 }{\sq rt{2 }} \b e gin {pmatri x }1 & -{ \l am bda} ^{- 1} \\ { \lam b da} & 1 \en d{pmatrix }\b e gin{ pm at rix} 1& {\lambda}^{ -1 } \\ 0 & 1 \ end {pmatr i x }\\ &= H ({\lambda})\Phi_{\mathb b {

$f_t^{\mathcal{D}}$_as $t$_tends to $0$. Recall_that $f_t^{\mathcal{D}}=_{_\mathrm{Sym}}_q\left({\mathop \mathrm{Uni}}\left(M_tz^{A_t}\right)\right)$_where_$M_t$ tends to_${\mathrm{I}}_2$ as $t$_tends to $0$. Hence,_the limit axis_of_$f_t^{\mathcal{D}}$ and $\widetilde{f}_t^{\mathcal{D}}:= { \mathrm{Sym}}_q\left({\mathop \mathrm{Uni}}\left(z^{A_t}\right)\right)$ are the same. Two cases can occur, wether_$r>_s$ or_$r<_s$. ####_Spherical family. At $t=0$, $r=\frac{1}{2}$ and_$s=0$. The limit potential is_thus $$\xi_0(z,{\lambda})_= \begin{pmatrix} 0 & \frac{{\lambda}^{-1}}{2} \\ \frac{{\lambda}}{2} & 0 \end{pmatrix}z^{-1} dz.$$_Consider_the gauge $$G(z,{\lambda})_= \frac{1}{\sqrt{2z}}\begin{pmatrix} 1 & 0\\ {\lambda}& 2z \end{pmatrix}.$$ The gauge potential is_then $$\xi_0\cdot G (z,{\lambda}) = \begin{pmatrix} 0_& {\lambda}^{-1}dz\\ 0&0 \end{pmatrix} =_\xi_{\mathbb{S}}(z,{\lambda})$$_where_$\xi_{\mathbb{S}}$ is the spherical_potential as in Section \[sectionDPWMethod\]. Let_$\widetilde{\Phi}:=z^{A_0}G$ be the gauged holomorphic frame_and compute $$\begin{aligned} \widetilde{\Phi}(1,{\lambda}) &= G(1,{\lambda}) \\ &=_\frac{1}{\sqrt{2}}\begin{pmatrix} 1 & 0\\ {\lambda}& 2 \end{pmatrix} = \frac{1}{\sqrt{2}}\begin{pmatrix} 1_& -{\lambda}^{-1} \\ {\lambda}& 1 \end{pmatrix}\begin{pmatrix} 1 &_{\lambda}^{-1} \\ 0_& 1 \end{pmatrix}\\ &= H({\lambda})\Phi_{\mathbb{

geodesic propagates from the turning point to the edge of the sphere, we need to solve how the geodesic propagates outside the sphere, and how the geodesic solution inside is matched to a solution outside. The geodesic equations outside are given by $$\begin{aligned} \frac{f'\,\dot{r}\,\dot{t}}{f} + \ddot{t} & = 0\\ -\frac{f'}{2\,f}\dot{r}^2 + \frac{1}{2}f\,f'\,\dot{t}^2 -r\,f\,\dot{\phi}^2 + \ddot{r} & = 0\\ \frac{2\,\dot{r}\,\dot{\phi}}{r}+\ddot{\phi} & = 0 \;.\end{aligned}$$ These can be integrated to give $$\begin{aligned} \label{eq:outgeoint1} f\,\dot{t} &=E\\ \label{eq:outgeoint2} r^2\dot{\phi} & = L\\ \label{eq:outgeoint3} \dot{r}^2 & = f\left( 1-\frac{L^2}{r^2}\right)+E^2 \;.\end{aligned}$$ To relate the constants of motion outside to the trajectory of the geodesic inside, we construct a coordinate system $(\tau,\lambda)$ which is continuous at the edge of the sphere and its close surroundings and use that to transform the geodesic derivatives in the inside coordinate system to those in the outside coordinate system. This computation is described in more detail in Appendix \[sec:continuous\], but its result is given by $$\begin{aligned} \label{eq:matchingOS1} E &= \sqrt{\frac{\bar{a}^2}{a^2}-1}\sqrt{1+\rho_0^2} + \frac{\bar{a}}{a}\sqrt{1-\frac{\bar{\rho}^2}{\rho^2}}\dot{r}_s\\ \label{eq:matchingOS2} L &= \bar{L} = \pm \bar{a}\bar{\rho} \;.\end{aligned}$$ The procedure to compute the geodesic length for all times and angles is then as follows: Given the initial conditions $\bar{a}$ and $\bar{\rho}$, integrate equations (\[eq:OSingeoint1\] - \[eq:OSingeoint3\]) until the geodesic reaches the edge of the sphere

geodesic propagates from the turning point to the edge of the sphere, we necessitate to resolve how the geodesic propagates outside the sphere, and how the geodetic solution inside is matched to a solution outside. The geodesic equation outside are given by $ $ \begin{aligned } \frac{f'\,\dot{r}\,\dot{t}}{f } + \ddot{t } & = 0\\ -\frac{f'}{2\,f}\dot{r}^2 + \frac{1}{2}f\,f'\,\dot{t}^2 -r\,f\,\dot{\phi}^2 + \ddot{r } & = 0\\ \frac{2\,\dot{r}\,\dot{\phi}}{r}+\ddot{\phi } & = 0 \;.\end{aligned}$$ These can be integrated to move over $ $ \begin{aligned } \label{eq: outgeoint1 } f\,\dot{t } & = E\\ \label{eq: outgeoint2 } r^2\dot{\phi } & = L\\ \label{eq: outgeoint3 } \dot{r}^2 & = f\left (1-\frac{L^2}{r^2}\right)+E^2 \;.\end{aligned}$$ To relate the constants of apparent motion outside to the trajectory of the geodesic inside, we construct a coordinate arrangement $ (\tau,\lambda)$ which is continuous at the edge of the sphere and its close surroundings and use that to translate the geodesic derivatives in the inside coordinate system to those in the outdoor coordinate system. This computation is describe in more detail in Appendix \[sec: continuous\ ], but its consequence is given by $ $ \begin{aligned } \label{eq: matchingOS1 } E & = \sqrt{\frac{\bar{a}^2}{a^2}-1}\sqrt{1+\rho_0 ^ 2 } + \frac{\bar{a}}{a}\sqrt{1-\frac{\bar{\rho}^2}{\rho^2}}\dot{r}_s\\ \label{eq: matchingOS2 } L & = \bar{L } = \pm \bar{a}\bar{\rho } \;.\end{aligned}$$ The procedure to compute the geodesic length for all time and angles is then as follows: Given the initial conditions $ \bar{a}$ and $ \bar{\rho}$, integrate equation (\[eq: OSingeoint1\ ] - \[eq: OSingeoint3\ ]) until the geodesic reaches the edge of the sphere

geldesic propagates from tme turning point to the edge kf the sohere, we need to solve how tie gwodesuc propagates outside ghe spherv, and how the teodesic solution luside ls maccied to a solutipn outside. The geodesic aqjacions outside are given by $$\begin{aligged} \frac{g'\,\dlt{r}\,\dot{t}}{f} + \ddot{j} & = 0\\ -\fgas{f'}{2\,f}\dkn{r}^2 + \frac{1}{2}f\,f'\,\dot{t}^2 -r\,f\,\dot{\phi}^2 + \ddot{r} & = 0\\ \frac{2\,\dmt{r}\,\dot{\phi}}{r}+\ddoy{\phi} & = 0 \;.\end{aligned}$$ These cwn bf integrated to gige $$\begin{alitned} \jqbel{eq:outgeoknt1} f\,\dot{t} &=T\\ \lcbel{eq:outgekint2} r^2\dot{\phi} & = L\\ \label{eq:outgeoing3} \dot{x}^2 & = f\left( 1-\ftcx{L^2}{r^2}\ghght)+E^2 \;.\end{almgned}$$ No relate the constands of mption outside bo thx trqjectory of the geodevic inside, we consjruct a comrbinate system $(\tau,\lambea)$ whicv is congunuuus av tge edgf oh the sphers and its coose surroundings amd lxe that to tdansfowm the geodesic derivatives in the inside cokrdinate system to thosw in the outside coorfinate syftem. This computation is described in more detail in A'pdndnw \[rwc:fontinuous\], but its result is given by $$\begin{alyfntd} \lsbel{eq:matchingJS1} E &= \sqrt{\ftaf{\bst{a}^2}{a^2}-1}\sqrt{1+\rho_0^2} + \frxc{\bar{a}}{c}\aqdt{1-\frac{\bar{\rho}^2}{\rho^2}}\dot{g}_s\\ \label{gq:matcyingOS2} L &= \bar{K} = \pm \bar{a}\bar{\rho} \;.\end{aligned}$$ The proceduge ti compute the geodzsic length yor alk timrs and angles is then ar fomlows: Given the initjxl conditions $\baf{a}$ sng $\bar{\rho}$, integrate equatiogs (\[eq:OSinjeoinc1\] - \[eq:OSivgeoont3\]) uneil the geldesig reaches the edge lf thg sphese

geodesic propagates from the turning point to of sphere, we to solve how sphere, how the geodesic inside is matched a solution outside. The geodesic equations are given by $$\begin{aligned} \frac{f'\,\dot{r}\,\dot{t}}{f} + \ddot{t} & = 0\\ -\frac{f'}{2\,f}\dot{r}^2 + \frac{1}{2}f\,f'\,\dot{t}^2 + \ddot{r} & = 0\\ \frac{2\,\dot{r}\,\dot{\phi}}{r}+\ddot{\phi} & = 0 \;.\end{aligned}$$ These can be to $$\begin{aligned} f\,\dot{t} \label{eq:outgeoint2} r^2\dot{\phi} & = L\\ \label{eq:outgeoint3} \dot{r}^2 & = f\left( 1-\frac{L^2}{r^2}\right)+E^2 \;.\end{aligned}$$ To relate the constants motion outside to the trajectory of the geodesic we construct a coordinate $(\tau,\lambda)$ which is continuous at edge the sphere its surroundings use that to the geodesic derivatives in the inside coordinate system to those in the outside coordinate system. This computation described in in Appendix but result given by $$\begin{aligned} &= \sqrt{\frac{\bar{a}^2}{a^2}-1}\sqrt{1+\rho_0^2} + \frac{\bar{a}}{a}\sqrt{1-\frac{\bar{\rho}^2}{\rho^2}}\dot{r}_s\\ \label{eq:matchingOS2} L \pm \bar{a}\bar{\rho} \;.\end{aligned}$$ The procedure to compute the length for times and angles is then as Given the initial conditions $\bar{a}$ and $\bar{\rho}$, integrate (\[eq:OSingeoint1\] - \[eq:OSingeoint3\]) until the geodesic reaches the edge of the sphere

geodesic propagates from the Turning poiNt to tHe eDge Of The sPherE, we need to solve HOw thE geodesic propagates outSide tHe SPherE, AnD how tHe geodeSIc SOLutIoN iNsiDe IS mAtcheD to A solutiOn outside. THe gEoDesic equatioNS oUtside are gIveN by $$\begin{aligNed} \Frac{f'\,\dOt{R}\,\doT{T}}{f} + \ddoT{t} & = 0\\ -\fRac{f'}{2\,f}\Dot{r}^2 + \frAC{1}{2}f\,f'\,\dot{T}^2 -r\,f\,\dot{\phi}^2 + \DdOT{r} & = 0\\ \frac{2\,\DOt{r}\,\dot{\pHI}}{R}+\dDot{\pHi} & = 0 \;.\end{aligned}$$ These CAn BE integrated to gIve $$\begIn{ALiGNEd} \lAbeL{eq:outgeoiNt1} F\,\dot{t} &=e\\ \Label{eq:OUtGEOInt2} R^2\Dot{\phi} & = L\\ \label{eQ:outgeoint3} \dOT{r}^2 & = f\Left( 1-\frAc{l^2}{r^2}\rIGht)+E^2 \;.\enD{aligNeD}$$ to rElate the conStanTs of motioN outsiDE to the tRAjectorY of the GeoDesIc inSIdE, wE coNsTRucT A cOorDInaTe system $(\TaU,\lAmbda)$ WhicH IS COntiNuoUs at The edGe of the sphere And Its cLOse SurroUndinGs anD uSe thaT to traNsforM tHe geodesic derivAtivEs in the inSidE cOorDiNate sYStem to ThoSe iN the outSide cooRDinAtE SYStEm. This computation iS dESCrIbed in moRe detaIL iN APPendix \[seC:cOntInuoUS\], But itS resULt Is given bY $$\begin{ALiGnEd} \label{Eq:MatchiNgoS1} E &= \SqrT{\frac{\BAr{a}^2}{a^2}-1}\Sqrt{1+\rhO_0^2} + \frac{\bar{A}}{a}\sqrT{1-\Frac{\bar{\rho}^2}{\rho^2}}\dOT{r}_s\\ \label{eq:matCHiNGoS2} l &= \Bar{L} = \Pm \bAr{a}\bar{\rho} \;.\enD{aliGNed}$$ THe prOCeDurE To comPute tHe GEoDEsic length for all timEs And angLes is Then as follows: given the inITIAl conditIons $\BAr{A}$ And $\bar{\rho}$, integRate eQuations (\[eq:osingeoinT1\] - \[eq:OSIngeoint3\]) Until the gEODesic reaCheS thE edGe oF THe Sphere

geodesic propagates fromthe turnin g poi nttoth e ed ge o f the sphere,w e ne ed to solve how the ge odesi cp ropa g at es ou tside t h es p her e, a ndho w t he ge ode sic sol ution insi deis matched toa s olution ou tsi de. The geod esi c equa ti ons outsi deare g iven b y $$\be gin{align ed } \frac { f'\,\do t { r} \,\d ot{t}}{f} + \ddot { t} & = 0\\ -\frac {f'}{2 \, f }\ d o t{r }^2 + \frac{1 }{ 2}f\, f '\,\dot { t} ^ 2 -r\ , f\,\dot{\phi} ^2 + \ddot{ r } & = 0\\ \ fra c {2\,\d ot{r} \, \ dot {\phi}}{r}+ \ddo t{\phi} & = 0 \ ; .\end{a l igned}$ $ Thes e c anbe i n te gr ate dt o g i ve $$ \ beg in{align ed }\labe l{eq : o u t geoi nt1 } f\ ,\dot {t} &=E\\ \la bel {eq: o utg eoint 2} r^ 2\do t{ \phi} & = L \\ \l ab el{eq:outgeoint 3} \ dot{r}^2& = f \le ft ( 1-\ f rac{L^ 2}{ r^2 }\right )+E^2 \ ; .\e nd { a l ig ned}$$ To relate t he c on stants o f moti o nou t side toth e t raje c t ory o f th e g eodesicinside , w econstru ct a coo rd ina tesyste m $(\ tau,\l ambda)$which is continuousa t the edge of th e sp h ereand its closesurr o undi ngsa nd us e that to t ra n sf o rm the geodesic der iv atives in t he inside coo rdinate sy s t e m to tho se i n t h e outside coor dinat e system.T his comp utati on is de scribed i n more det ail in Ap pen d i x\[sec:continu o u s\], b ut itsres ult isgiv enby$$\ be gin{align ed} \lab el {e q: ma tch ingOS 1 } E &= \ sq rt{ \f rac {\bar { a}^2}{ a^2}- 1}\s qr t{ 1 +\r ho_0^2} +\ f rac{ \b ar {a}} {a} \s qrt{1 -\fr a c{\ bar{\rh o}^2}{\rh o^2 } }\do t{ r} _s\\ \l abel{eq:match in gOS2} L &= \ bar {L} =\ p m \bar{a }\bar{\rho} \;.\end{ali g ned}$$The proc edur e to comp ute the g eod e sic le ngth f or al ltim e s anda n gl esis then as f o l low s: Gi ve n th e initi al conditions $\ba r {a} $ and $\bar{\ rho }$,i n te gra t ee qua ti o ns( \ [eq:OSingeoint1 \] - \[eq: OS i ng eoint3\])u nti lthe geo desic r eache s the ed ge of the sphere

geodesic_propagates from_the turning point to_the edge_of_the sphere,_we_need to solve_how the geodesic_propagates outside the sphere,_and how the_geodesic_solution inside is matched to a solution outside. The geodesic equations outside are given_by_$$\begin{aligned} \frac{f'\,\dot{r}\,\dot{t}}{f} +_\ddot{t}_&_= 0\\ -\frac{f'}{2\,f}\dot{r}^2 + \frac{1}{2}f\,f'\,\dot{t}^2 -r\,f\,\dot{\phi}^2_+ \ddot{r} & = 0\\ \frac{2\,\dot{r}\,\dot{\phi}}{r}+\ddot{\phi}_& =_0 \;.\end{aligned}$$ These can be integrated to give_$$\begin{aligned} \label{eq:outgeoint1}_f\,\dot{t} &=E\\ \label{eq:outgeoint2} r^2\dot{\phi}_& = L\\ \label{eq:outgeoint3} \dot{r}^2 & = f\left( 1-\frac{L^2}{r^2}\right)+E^2 \;.\end{aligned}$$_To relate the constants of motion_outside to the_trajectory_of_the geodesic inside, we_construct a coordinate system $(\tau,\lambda)$ which_is continuous at the edge of_the sphere and its close surroundings and_use that to transform the geodesic_derivatives in the inside coordinate_system to_those in the outside coordinate_system. This computation_is described_in more detail_in Appendix \[sec:continuous\], but its result_is given by_$$\begin{aligned} \label{eq:matchingOS1} E &= \sqrt{\frac{\bar{a}^2}{a^2}-1}\sqrt{1+\rho_0^2} + \frac{\bar{a}}{a}\sqrt{1-\frac{\bar{\rho}^2}{\rho^2}}\dot{r}_s\\ \label{eq:matchingOS2}_L_&= \bar{L} =_\pm_\bar{a}\bar{\rho}_\;.\end{aligned}$$ The_procedure to compute_the_geodesic length_for_all times and angles is then_as_follows: Given the initial conditions $\bar{a}$ and_$\bar{\rho}$, integrate equations (\[eq:OSingeoint1\]_-_\[eq:OSingeoint3\]) until the geodesic_reaches the edge of the_sphere

a solitary WR star. Our criteria are based on (a) the number of nearby massive stars (other than the WR star), and (b) a comparison of the sizes of the Group I shells with the sizes of WR nebulae in the LMC. In Figure 4 we plot the sizes of Group I shells versus the number of nearby, potentially ionizing stars (including the WR star in all cases). Stars were selected if they were both (a) within one shell radius (and therefore within the boundary of the shell or shell fragment), and (b) have B - V colors $\leq$ 0 (photometry is taken from Magnier et al. (1992)). Only stars bluer than B0 (B - V = -0.3) can supply enough ionizing photons to be consistent with the H$\alpha$ luminosities of Group I shells. Our choice of B - V $\leq$ 0 was prompted by considering the following two factors. First, there are substantial uncertainties in the photometry from the Magnier catalog. Second, we have made no reddening corrections. Thus the color information on stars in the field is, unfortunately, crude. Note in Figure 4 that Group I shells with radii less than $\sim$40 pc generally contain one or two ionizing stars. Rosado (1986) reported the radii of shells associated with (a) single WR stars, (b) single OB stars, and (c) multiple OB & WR stars in the LMC. The mean radii are, respectively, 15$\pm$7 pc, 27$\pm$4 pc, and 69$\pm$33 pc (Rosado 1986). Single massive stars in the LMC are surrounded by shells up to $\sim$30 pc in radius, consistent with the radius estimated above for Group I shells We suggest, allowing for chance alignment, those shells containing one or two ionizing stars (including a WR star), or that are $\leq$ 40 pc in radius, are possibly of E- or W-type. There are 5 Group Ia shells that satisfy this criteria: ob69-F1, MS8, MS2, ob54wr1, and ob42wr1. All 7 inner Group Ib shells satisfy this criteria, although the outer shells for MS14 and MS4 do not. Therefore, 12 of the 48 WR stars we observe (25$\%$) may be surrounded by “classical” WR E- or W-type shells. The remaining 11 Group Ia shells contain three or

a solitary WR star. Our criteria are based on (a) the numeral of nearby massive star (other than the WR star), and (b) a comparison of the sizes of the Group I blast with the sizes of WR nebulae in the LMC. In number 4 we plat the sizes of Group I blast versus the number of nearby, potentially ionizing stars (admit the WR star in all cases). headliner were choose if they were both (a) within one shell spoke (and therefore within the limit of the plate or shell fragment), and (b) have b-complex vitamin - V colors $ \leq$ 0 (photometry is take from Magnier et al. (1992) ). entirely stars blasphemous than B0 (B - V = -0.3) can supply adequate ionizing photons to be consistent with the H$\alpha$ luminosities of Group I shells. Our choice of B - five $ \leq$ 0 was prompted by considering the following two factors. First, there are substantial uncertainties in the photometry from the Magnier catalog. Second, we have made no reddening correction. Thus the color data on star in the sphere is, unfortunately, crude. Note in Figure 4 that Group I shells with radii less than $ \sim$40 pc generally control one or two ionizing stars. Rosado (1986) reported the radii of shells associated with (a) single WR headliner, (b) single OB stars, and (c) multiple OB & WR asterisk in the LMC. The mean spoke are, respectively, 15$\pm$7 pc, 27$\pm$4 pc, and 69$\pm$33 pc (Rosado 1986). Single massive stars in the LMC are besiege by shells up to $ \sim$30 pc in radius, reproducible with the radius estimated above for Group I shells We suggest, allow for chance alignment, those shells containing one or two ionizing stars (include a WR star), or that are $ \leq$ 40 personal computer in radius, are possibly of E- or W - type. There be 5 Group Ia shells that satisfy this criterion: ob69 - F1, MS8, MS2, ob54wr1, and ob42wr1. All 7 inner Group Ib shell satisfy this criteria, although the outer shells for MS14 and MS4 do not. Therefore, 12 of the 48 WR stars we observe (25$\%$) may be surround by “ classical ” WR E- or W - character shells. The remaining 11 Group Ia shells contain three or

a dolitary WR star. Our criueria are based ou (a) the numbed of neafby massive stars (other than tye WR star), and (b) a comparisun of the sizes od tht Group I shells xjth the sizes lf WX iebulae in the KMC. In Figuse 4 we plot tha rives of Group I shells versus the numfer of mewrby, potentialjy ipgizihg stars (including the WR star in zll casts). Stars were selevted if they were both (a) wlthij one shell radius (and therefire ruthin the bojndary of uhz shell or ahell fragment), and (b) have B - V zolorx $\leq$ 0 (phojojehty is taken hrom Mwgnier et al. (1992)). Only sdars blier than B0 (B - Y = -0.3) cen sypply enough ionizing photons to be confistent whtk the H$\alpha$ luminosiriws of Grogp I wheuls. Onr dhoice of B - V $\leq$ 0 was promptwd by considering tne dollowing two factows. First, there are substantial uncertaintpes jn the photometry from rhe Magnier catalog. Sgcond, we hwve made no reddening corrections. Thus the color hnforjxtiin ub dtars in the field is, unfortunately, crude. Notq im Nigure 4 that Gromp I shells with rsdli jess than $\sim$40 pc geusrzlly contain one og two ijnizibg stars. Rjsadp (1986) reported the radii of shwlls associaned qith (a) single WR scars, (b) singlz OB sjars, amd (c) multiple OB & WR stcrs in the LMC. Thf mean raski are, respectivdly, 15$\pk$7 pc, 27$\pm$4 kz, and 69$\pm$33 pc (Rosadj 1986). Single masxive stxrs on the LMC are skrroukged by shells up tl $\sim$30 pw in radiud, consistent with the radius esvmmated above gos Ggoup I shzlls We suggest, alloring for chancg alignmeut, thore shells bontaininj one or two ionizing stasd (including e WR star), or rhat are $\led$ 40 pc in radius, are posspbjy od E- or W-type. There arg 5 Group Ia shellw tyat satisfy thix cfitqrpa: pb69-S1, MS8, MS2, ob54wr1, dnd ub42wf1. All 7 inner Groui Ic shrlls satisfy this crhterja, although the ouyev shells dor MS14 agd MS4 do not. Yherefore, 12 of the 48 WR svars wx obsetve (25$\%$) may be surrounded by “classiczl” WR E- og W-bype shells. Tre rcmaiging 11 Grou' Ia shells contain three or

a solitary WR star. Our criteria are (a) number of massive stars (other (b) comparison of the of the Group shells with the sizes of WR in the LMC. In Figure 4 we plot the sizes of Group I versus the number of nearby, potentially ionizing stars (including the WR star in cases). were if were both (a) within one shell radius (and therefore within the boundary of the shell or fragment), and (b) have B - V colors 0 (photometry is taken Magnier et al. (1992)). Only bluer B0 (B V -0.3) supply enough ionizing to be consistent with the H$\alpha$ luminosities of Group I shells. Our choice of B - V 0 was considering the two First, are substantial uncertainties photometry from the Magnier catalog. Second, no reddening corrections. Thus the color information on in the is, unfortunately, crude. Note in Figure that Group I shells with radii less than pc generally contain one or two ionizing stars. Rosado (1986) reported the radii of shells (a) single WR stars, single OB stars, (c) OB WR in the The mean radii are, respectively, 15$\pm$7 pc, 27$\pm$4 pc, and 69$\pm$33 (Rosado 1986). Single massive stars in the LMC are surrounded up $\sim$30 pc in consistent with the radius above Group I shells We for alignment, one two stars (including a WR or that are $\leq$ 40 in radius, are possibly are 5 Group Ia shells that satisfy this ob69-F1, MS8, MS2, ob54wr1, and ob42wr1. All inner Group Ib shells satisfy this criteria, although the outer shells for and MS4 Therefore, 12 of the 48 WR stars we (25$\%$) may be surrounded “classical” WR E- or W-type shells. The remaining 11 Ia contain three

a solitary WR star. Our criteriA are based oN (a) the NumBer Of NearBy maSsive stars (otheR Than The WR star), and (b) a comparisOn of tHe SIzes OF tHe GroUp I shelLS wITH thE sIzEs oF Wr NeBulae In tHe LMC. In figure 4 we plOt tHe Sizes of Group i ShElls versus The Number of nearBy, pOtentiAlLy iONizinG stArs (inCludinG The WR sTar in all cAsES). Stars WEre seleCTEd If thEy were both (a) within ONe SHell radius (and tHerefoRe WItHIN thE boUndary of thE sHell oR Shell frAGmENT), And (B) Have B - V colors $\lEq$ 0 (photometrY Is tAken frOm magNIer et aL. (1992)). Only StARs bLuer than B0 (B - V = -0.3) Can sUpply enouGh ioniZIng photONs to be cOnsistEnt WitH the h$\AlPhA$ luMiNOsiTIeS of gRouP I shells. ouR cHoice Of B - V $\LEQ$ 0 WAs prOmpTed bY consIdering the folLowIng tWO faCtors. first, TherE aRe subStantiAl uncErTainties in the phOtomEtry from tHe MAgNieR cAtaloG. second, We hAve Made no rEddeninG CorReCTIOnS. Thus the color inforMaTIOn On stars iN the fiELd Is, UNfortunaTeLy, cRude. nOTe in FIgurE 4 ThAt Group I Shells WItH rAdii lesS tHan $\sim$40 Pc GenEraLly coNTain One or tWo ioniziNg staRS. Rosado (1986) reporteD The radii of sheLLs ASSoCIateD wiTh (a) single WR StarS, (B) sinGle Ob StArs, ANd (c) muLtiplE Ob & wR STars in the LMC. The mean RaDii are, RespeCtively, 15$\pm$7 pc, 27$\pm$4 Pc, and 69$\pm$33 pc (ROSADo 1986). Single MassIVe STars in the LMC arE surrOunded by shELls up to $\sIm$30 pc iN radius, cOnsistent WITh the radIus EstImaTed ABOvE for Group I sheLLS We sUgGest, allOwiNg for chAncE alIgnMenT, tHose shellS containInG oNe Or Two IonizINg stars (iNcLudInG a Wr star), OR that aRe $\leq$ 40 Pc in RaDiUS, arE possibLY oF e- Or W-tYpE. THere Are 5 grOup Ia ShelLS thAt satisFy this criTerIA: ob69-F1, mS8, mS2, Ob54wr1, and Ob42wr1. All 7 inner GRoUp Ib shells SaTisFy this CRIteria, alThough the outer shells for ms14 and MS4 dO noT. TherEforE, 12 of the 48 WR sTarS we obsErvE (25$\%$) May be sUrrounDed by “ClAssICAl” WR E- OR w-tYpe ShElls. The remAINinG 11 GrouP IA sheLls contAin three or

a solitary WR star. Our c riteria ar e bas edon(a ) th e nu mber of nearby mass ive stars (other thanthe W Rs tar) , a nd (b ) a com p ar i s onof t hesi z es of t heGroup I shells wi thth e sizes of W R n ebulae inthe LMC. In Fi gur e 4 we p lot the s ize s ofGroupI shell s versusth e numbe r of nea r b y, pot entially ionizing st a rs (includingthe WR s t ar i n a llcases). St ar s wer e select e di f the y were both (a ) within on e sh ell ra di us( and th erefo re wit hin the bou ndar y of theshello r shell fragmen t), an d ( b)have B-V c ol o rs$ \l eq$ 0 ( photomet ry i s tak en f r o m Magn ier etal. ( 1992)). Onlysta rs b l uer than B0 ( B -V= -0. 3) can supp ly enough ionizin g ph otons tobeco nsi st ent w i th the H$ \al pha$ lu minosit i esof G r ou p I shells. Our ch oi c e o f B - V$\leq$ 0wa s prompte dbycons i d ering the fo llowingtwo fa c to rs . First ,therear e s ubs tanti a l un certai nties in thep hotometry from the Magnier c a ta l o g. Seco nd, we have ma de n o red deni n gcor r ectio ns. T hu s t h e color information o n star s inthe field is, unfortuna t e l y, crude . No t ei n Figure 4 tha t Gro up I shell s with ra dii l ess than $\sim$40 p c genera lly co nta ino n eor two ionizi n g sta rs . Rosa do(1986)rep ort edthe r adii of s hells as so ci at ed wi th (a ) singleWR st ar s,(b) s i ngle O B sta rs,an d( c)multipl e O B & WR s ta rs i n t he LMC. The mea n radii are, res pec t ivel y, 1 5$\pm$7 pc, 27$\pm$4 p c, and 69$ \p m$3 3 pc ( R o sado 198 6). Single massive star s in the LM C are sur rounded b y s hellsupt o $\si m$30 p c inra diu s , cons i s te ntwi th the rad i u s e stima te d ab ove for Group I shells W e su ggest, allowi ngforc h an cea li g nme nt , th o s e shells contai ning one o rt wo ionizings tar s(includ ing a W R sta r ), or t hat are $ \leq$ 40pc inr a diu s, are pos sibly of E- or W- t ype.T he re ar e 5 Group I a s hells thats ati sfy t his cr it eria:ob69- F1 , MS8, M S2, ob54wr1, and ob42wr 1. All 7 in ner Group Ib sh e lls satisfythis criteria, al tho ugh t heo utershel l sfor MS14andM S4 do not . T her e f or e, 12 of th e 4 8 W R sta rsw e obse rve(25$\%$) may be s u rrounded by “c lass i c al” WR E- o rW-type shells. Th er e maining11 Group Ia s hells co nt a in th ree or

a_solitary WR_star. Our criteria are_based on_(a)_the number_of_nearby massive stars_(other than the_WR star), and (b)_a comparison of_the_sizes of the Group I shells with the sizes of WR nebulae in the_LMC. In_Figure 4_we_plot_the sizes of Group I_shells versus the number of_nearby, potentially_ionizing stars (including the WR star in all_cases)._Stars were selected_if they were both (a) within one shell radius_(and therefore within the boundary of_the shell or_shell_fragment),_and (b) have B_- V colors $\leq$ 0 (photometry_is taken from Magnier et al._(1992)). Only stars bluer than B0 (B_- V = -0.3) can supply_enough ionizing photons to be_consistent with_the H$\alpha$ luminosities of Group_I shells. Our_choice of_B - V_$\leq$ 0 was prompted by considering_the following two_factors. First, there are substantial uncertainties_in_the photometry from_the_Magnier_catalog. Second,_we have made_no_reddening corrections._Thus_the color information on stars in_the_field is, unfortunately, crude. Note in Figure_4 that Group I_shells_with radii less than_$\sim$40 pc generally contain one_or two ionizing stars. Rosado (1986) reported_the radii_of shells_associated with (a) single WR stars, (b) single OB stars, and_(c) multiple OB & WR stars_in the LMC. The_mean radii_are,_respectively, 15$\pm$7 pc,_27$\pm$4_pc, and_69$\pm$33 pc (Rosado 1986). Single massive stars_in the_LMC are surrounded by shells up_to $\sim$30 pc in_radius,_consistent with the radius estimated above_for Group I shells We suggest, allowing_for chance alignment, those shells_containing_one_or two ionizing stars (including_a WR star), or that are_$\leq$ 40 pc_in radius, are possibly of E- or_W-type._There are 5 Group Ia shells_that_satisfy this criteria: ob69-F1, MS8, MS2,_ob54wr1,_and_ob42wr1. All 7 inner Group_Ib shells satisfy this criteria, although_the outer shells for MS14 and MS4 do not._Therefore, 12 of_the 48 WR stars we_observe_(25$\%$)_may be surrounded by “classical” WR E- or W-type shells._The remaining_11 Group Ia_shells contain three or

relaxation regimes as a function of temperature. At high temperature the relaxation scales in accordance with the results derived from Mullins’ theory, whereas at low temperature the relaxation time becomes an exponentional function of the size of the crystallites. So that the effects of lowering the temperature are qualitatively different for two dimensional and three dimensional crystallites : in two dimensions, lowering the temperature decreases the strength of the dependence of the relaxation time as a function of the size of the crystallites (as it crosses over from a $N^2$ dependence to a $N$ dependence), whereas it increases this the strength in three dimensions. In both cases, the limiting step is the nucleation of a germ on a facet : a unidimensional germ in two dimensions, and a two dimensional germ in three dimensions. The difference stems from the fact that in the two dimensional case, the activation energy for the creation of the germ does not depend on the size of the island, it is always constant : $4E$, and it stabilizes when a row on a small facet has been removed. In the three dimensional case, this activation energy depend on the size of the crystallite. The transfer of a particle from a tip of the crystallite to the germ has a gain in volume energy (depending on the size of the islands) and a loss in edge energy of the germ (depending on the size of the germ). Summing these two terms, an energy barrier proportional to the size of the crystallite appears for the creation of a stable germ. The exponential behavior of the relaxation time as a function of N is a consequence of this energy barrier dependence.\ Finally, we believe that the overall picture presented here, while still oversimplified, seems to be complete enough to provide a general picture of the processes leading to the shape relaxation of two dimensional islands at low temperatures.\ We acknowledge useful discussions with P. Jensen, J. Wittmer and F. Nicaise. We are grateful with an anonymous referee for his many useful comments. H.L. also acknowledges partial financial support from CONACYT and DGAPA-UNAM, and is thankful with Univ. Claude Bernard, Lyon 1, for the invitation during which part of this work was done. Calculation of the probability to have 2 particles on the facet {#appen_1} =============================================================== We calculate the probability $P$ of having 2 particles on a facet with absorbing boundaries, knowing that at time $t=0$, one particle is on one edge

relaxation regimes as a function of temperature. At high temperature the liberalization scale in accordance with the result derive from Mullins ’ theory, whereas at low temperature the easiness time becomes an exponentional affair of the size of the crystallites. So that the effect of lowering the temperature are qualitatively different for two dimensional and three dimensional crystallite: in two dimensions, lowering the temperature decreases the strength of the dependence of the relaxation time as a affair of the size of the crystallites (as it crosses over from a $ N^2 $ dependence to a $ N$ addiction), whereas it increases this the strength in three dimension. In both cases, the restrict step is the nucleation of a germ on a facet: a unidimensional germ in two dimensions, and a two dimensional source in three dimensions. The difference stems from the fact that in the two dimensional case, the activation energy for the creation of the germ does not depend on the size of the island, it is always constant: $ 4E$, and it stabilizes when a row on a small facet has been removed. In the three dimensional case, this energizing department of energy depend on the size of the crystallite. The transfer of a particle from a peak of the crystallite to the germ has a gain in bulk energy (depending on the size of the islands) and a loss in boundary energy of the germ (depending on the size of the germ). Summing these two terms, an energy barrier proportional to the size of the crystallite appears for the creation of a stable germ. The exponential behavior of the liberalization time as a affair of N is a consequence of this energy barrier dependence.\ Finally, we believe that the overall picture presented here, while however oversimplified, seems to be complete enough to leave a general picture of the processes leading to the supreme headquarters allied powers europe relaxation of two dimensional island at low temperatures.\ We acknowledge utilitarian discussions with P. Jensen, J. Wittmer and F. Nicaise. We are grateful with an anonymous referee for his many useful comments. H.L. besides acknowledges partial fiscal support from CONACYT and DGAPA - UNAM, and is thankful with Univ. Claude Bernard, Lyon 1, for the invitation during which part of this work was done. Calculation of the probability to receive 2 particles on the facet { # appen_1 } = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = We calculate the probability $ P$ of having 2 particles on a facet with absorbing boundaries, know that at time $ t=0 $, one particle is on one boundary

repaxation regimes as a fukction of tempercrure. Av high femperatjre the relaxation scales in axcordqnce with the results aerived fgom Mullibs’ tieory, whereas at low temizratudc the celaxation time becomes at exponentionan wuuction of the size of the crystallitqs. So tnah the effects jf lprerihg the temperature are qualitativemy difftrent for two dimemsional and three dimensiojal frystallites : in tao dimensiobs, ljqering the tdmperature decreases tge strength of the dependence ow the relaxatiob rimf as a functmon of the size of the cryvtallitrs (as it crosscs ovxr feom a $N^2$ dependence to a $N$ dependence), whgreas it itcxeases this the strentty in jhree dimdbsiuns. Ii bkth cades, the limitjng step is the nucleation of s dvtm on a facef : a ugiqimensional germ in two dimensions, and d tso dimensional germ in rhree dimensions. The fifferencq stems from the fact that in the two dimensional case, ghe agtivxriln energy for the creation of the germ does njf ceiend on the size jf the islamd, iy is always covstant : $4E$, and it stabilizes when a row in a smalj favet has been removed. In the three dimenfuonal case, this accivation enexgy dekend om the size of the crystcllite. The transffr of a pzfticle from a tio on tve crystallite to the germ has a gamn in volume enetgy (depqnding on hhe slse of the islands) wnd a lmss in edgf energy of the germ (depending on the size of tve cerm). Sumiing bhese two terms, an energy bartier propjrtioval to the size oh the crystajlite appears vor the creavion of a stavle term. Thd exponential brhavior oy the relaxation time ax a runction of N iw a consequence of thks qnvrgb barwher dependenwe.\ Fivaluu, we celieve thab tfe oferall picture presetted here, while still pvcrsimplifued, seemf to be complrte enough to provlde a geieral licjure of the processes leading tk the shaoe velaxation of two dimensional islands at low temperatures.\ We acknowlevge useful discussions qith P. Jensen, J. Witjmev and F. Nicamse. We are grataful with an anonymoys referee for hix many useful commenta. H.L. anso afknowledges partial financial support from CONACYT and DGAPA-UNAM, and is tyankful with Univ. Dlauce Begnaxd, Kyon 1, sor vhx invitation durimg which part of this work was vone. Calculdtnon of the probability to havr 2 particles on the facet {#appen_1} =============================================================== We calcumate the probability $P$ of having 2 partivles on a facet with absprbing biundadies, knowiig that at time $t=0$, onr partncle iw on onw edne

relaxation regimes as a function of temperature. temperature relaxation scales accordance with the whereas low temperature the time becomes an function of the size of the So that the effects of lowering the temperature are qualitatively different for two and three dimensional crystallites : in two dimensions, lowering the temperature decreases the of dependence the time as a function of the size of the crystallites (as it crosses over from a dependence to a $N$ dependence), whereas it increases the strength in three In both cases, the limiting is nucleation of germ a : a unidimensional in two dimensions, and a two dimensional germ in three dimensions. The difference stems from the fact in the case, the energy the of the germ depend on the size of the always constant : $4E$, and it stabilizes when row on small facet has been removed. In three dimensional case, this activation energy depend on size of the crystallite. The transfer of a particle from a tip of the crystallite germ has a gain volume energy (depending the of islands) a loss edge energy of the germ (depending on the size of the Summing these two terms, an energy barrier proportional to the the appears for the of a stable germ. exponential of the relaxation time function N of energy dependence.\ Finally, we believe the overall picture presented here, still oversimplified, seems to a general picture of the processes leading to shape relaxation of two dimensional islands at temperatures.\ We acknowledge useful discussions with P. Jensen, J. Wittmer and F. We are an anonymous referee for his many useful comments. also acknowledges partial financial from CONACYT and DGAPA-UNAM, and is thankful with Univ. Bernard, 1, for invitation during which of this work done. Calculation of to have particles the We calculate the probability $P$ of 2 on a facet with absorbing knowing time $t=0$, one particle is on one edge

relaxation regimes as a functIon of tempeRaturE. At HigH tEmpeRatuRe the relaxatioN ScalEs in accordance with the rEsultS dERiveD FrOm MulLins’ theORy, WHEreAs At Low TeMPeRaturE thE relaxaTion time beComEs An exponentioNAl Function of The Size of the cryStaLlites. so ThaT The efFecTs of lOwerinG The temPerature aRe QUalitaTIvely diFFErEnt fOr two dimensional aND tHRee dimensional CrystaLlITeS : IN twO diMensions, loWeRing tHE temperATuRE DEcrEAses the strengTh of the depeNDenCe of thE rElaXAtion tIme as A fUNctIon of the sizE of tHe crystalLites (aS It crossES over frOm a $N^2$ dePenDenCe to A $n$ dEpEndEnCE), whEReAs iT IncReases thIs ThE streNgth IN THRee dImeNsioNs. In bOth cases, the liMitIng sTEp iS the nUcleaTion Of A germ On a facEt : a unIdImensional germ iN two DimensionS, anD a Two DiMensiONal gerM in ThrEe dimenSions. ThE DifFeRENCe Stems from the fact thAt IN ThE two dimeNsionaL CaSe, THe activaTiOn eNergY FOr the CreaTIoN of the geRm does NOt DePend on tHe Size of ThE isLanD, it is ALwayS constAnt : $4E$, and iT stabILizes when a row oN A small facet haS BeEN ReMOved. in tHe three dimeNsioNAl caSe, thIS aCtiVAtion EnergY dEPeND on the size of the crysTaLlite. THe traNsfer of a partiCle from a tiP OF The crystAlliTE tO The germ has a gaiN in voLume energy (DEpending On the Size of thE islands) aND A loss in eDge EneRgy Of tHE GeRm (depending on THE sizE oF the gerM). SuMming thEse Two TerMs, aN eNergy barrIer propoRtIoNaL tO thE size OF the crysTaLliTe AppEars fOR the crEatioN of a StAbLE geRm. The exPOnENTial BeHaVior Of tHe RelaxAtioN TimE as a funCtion of N iS a cONseqUeNcE of this Energy barrier DePendence.\ FiNaLly, We beliEVE that the Overall picture presented HEre, whilE stIll ovErsiMplified, sEemS to be cOmpLEte enoUgh to pRovidE a GenERAl picTURe Of tHe Processes lEADinG to thE sHape RelaxatIon of two dimensionaL IslAnds at low tempEraTureS.\ wE aCknOWlEDge UsEFul DIScussions with P. JEnsen, J. WittMeR AnD F. Nicaise. WE Are GrAteful wIth an anOnymoUS refereE for his maNy useful cOmMentS. h.l. alSo acknowleDges partIal financIAl supPOrT from cONaCYT anD DgAPa-UNAM, And is tHAnkFul wiTh Univ. clAude BeRnard, lyOn 1, for the Invitation during which paRt of thIs worK waS done. CalcUlaTIon Of the probAbilIty to have 2 pArtIclEs on tHe fACet {#apPen_1} =============================================================== WE CaLcuLAte thE proBAbility $P$ oF HaVinG 2 PArTicles on a faCET WitH absoRbiNG boundArieS, knowing that at timE $T=0$, one particle is On onE EDge

relaxation regimes as a f unction of temp era tur e. Athigh temperature t h e re laxation scales in acc ordan ce with th e res ults de r iv e d fr om M ull in s ’theor y,whereas at low te mpe ra ture the rel a xa tion timebec omes an expo nen tional f unc t ion o f t he si ze oft he cry stallites .S o that the eff e c ts oflowering the temp e ra t ure are qualit ativel yd if f e ren t f or two dim en siona l and th r ee d i men s ional crystal lites : int wodimens io ns, loweri ng th et emp erature dec reas es the st rength of thed ependen ce ofthe re laxa t io ntim ea s a fu nct i onof the s iz eof th e cr y s t a llit es(asit cr osses over fr oma $N ^ 2$depen dence toa$N$ d epende nce), w hereas it incre ases this the st re ngt hin th r ee dim ens ion s. In b oth cas e s,th e l im iting step is thenu c l ea tion ofa germ on a facet :auni dime n s ional ger m i n two di mensio n s, a nd a tw odimens io nal ge rm in thre e dime nsions.The d i fference stems from the fact th a t i n the tw o dimension al c a se,thea ct iva t ion e nergy f o rt he creation of thege rm doe s not depend on th e size oft h e island, iti sa lways constant : $4 E$, and it stabiliz es wh en a row on a sma l l facet h asbee n r emo v e d. In the three d imen si onal ca se, this a cti vat ion en er gy depend on thesi ze o fthe crys t allite.Th e t ra nsf er of a part iclefrom a t i p o f the c r ys t a llit eto the ge rm hasa ga i n i n volum e energy(de p endi ng o n the s ize of the is la nds) and a l oss in ed g e energyof the germ (dependingo n the s ize of t he g erm). Sum min g thes e t w o term s, anenerg ybar r i er pr o p or tio na l to the s i z e o f the c ryst alliteappears for the cr e ati on of a stabl e g erm. T he ex p on e nti al beh a v ior of the rela xation tim ea sa function ofNis a co nsequen ce of this en ergy barr ier depen de nce. \ Fin ally, we b elieve t hat the o v erall pi cture pr esente dher e, wh ile st i llovers implif ie d, see ms to b e comple te enough to provide agenera l pic tur e of thepro c ess es leadin g to the shape re lax ation of two d imen s io nal islan ds a t low temp e ra tur e s .\ We acknowl e d g e u seful di s cussio ns w ith P. Jensen, J. Wittmer and F. Nic a i se. We aregr ateful with an an on y m ous refe re e for his m any usef ul comme nts. H .L. al so ackn o w le d ges pa rtia l f inancialsup po r t fromCO NA C YT and DGA PA -UNAM, and i s tha n k ful with Univ. C laude B ernar d , L yon 1 ,for the invi tation dur ing which p art of thi s wor k was d on e. Ca lcu la tion of th e probabil ity t o have2part icl es onthef a cet { #app en _1} ======== = = == = == == = === ==== ===== == ==== ========= = ======== === = ===== We ca l c ulatet he p robability $P $ ofh a ving 2 par t icle s o n a fa cet wi th abs orbingb oun da ries, k now i n g that at time $t= 0 $,on e pa rticle i son o ne ed ge

relaxation_regimes as_a function of temperature._At high_temperature_the relaxation_scales_in accordance with_the results derived_from Mullins’ theory, whereas_at low temperature_the_relaxation time becomes an exponentional function of the size of the crystallites. So that_the_effects of_lowering_the_temperature are qualitatively different for_two dimensional and three dimensional_crystallites :_in two dimensions, lowering the temperature decreases the_strength_of the dependence_of the relaxation time as a function of the_size of the crystallites (as it_crosses over from_a_$N^2$_dependence to a $N$_dependence), whereas it increases this the_strength in three dimensions. In both_cases, the limiting step is the nucleation_of a germ on a facet_: a unidimensional germ in_two dimensions,_and a two dimensional germ_in three dimensions._The difference_stems from the_fact that in the two dimensional_case, the activation_energy for the creation of the_germ_does not depend_on_the_size of_the island, it_is_always constant_:_$4E$, and it stabilizes when a_row_on a small facet has been removed._In the three dimensional_case,_this activation energy depend_on the size of the_crystallite. The transfer of a particle_from a_tip of_the crystallite to the germ has a gain in volume energy_(depending on the size of the_islands) and a loss_in edge_energy_of the germ_(depending_on the_size of the germ). Summing these two_terms, an_energy barrier proportional to the size_of the crystallite appears_for_the creation of a stable germ._The exponential behavior of the relaxation_time as a function of_N_is_a consequence of this energy_barrier dependence.\ Finally, we believe that the_overall picture presented_here, while still oversimplified, seems to be_complete_enough to provide a general picture_of_the processes leading to the shape_relaxation_of_two dimensional islands at low_temperatures.\ We acknowledge useful discussions with P._Jensen, J. Wittmer and F. Nicaise. We are grateful_with an anonymous_referee for his many useful_comments._H.L._also acknowledges partial financial support from CONACYT and DGAPA-UNAM, and_is thankful_with Univ. Claude_Bernard, Lyon 1, for the invitation during which part of_this work was done. Calculation of the probability_to have 2 particles on the facet {#appen_1} =============================================================== We calculate the probability_$P$ of having 2 particles on a facet_with absorbing boundaries, knowing that at time_$t=0$, one particle is_on_one edge

0.022 4 549 $\pm$ 0.029 2 917 $\pm$ 0.027 2 752 $\pm$ 0.031 $0.14 < z \le 0.16$ 1 660 $\pm$ 0.025 -1 458 $\pm$ 0.026 4 621 $\pm$ 0.030 2 908 $\pm$ 0.029 2 670 $\pm$ 0.036 $0.16 < z \le 0.18$ 1 651 $\pm$ 0.028 -1 496 $\pm$ 0.028 4 563 $\pm$ 0.029 2 906 $\pm$ 0.031 2 677 $\pm$ 0.040 $0.18 < z \le 0.20$ 1 688 $\pm$ 0.031 -1 371 $\pm$ 0.030 4 559 $\pm$ 0.030 2 843 $\pm$ 0.033 2 623 $\pm$ 0.040 $0.20 < z \le 0.22$ 1 654 $\pm$ 0.034 -1 337 $\pm$ 0.032 4 587 $\pm$ 0.033 2 881 $\pm$ 0.033 2 644 $\pm$ 0.040 $0.22 < z \le 0.24$ 1 664 $\pm$ 0.026 -1 369 $\pm$ 0.029 4 519 $\pm$ 0.028 2 830 $\pm$ 0.029 2 677 $\pm$ 0.037 $0.24 < z \le 0.26$ 1 668 $\pm$ 0.028 -1 268 $\pm$ 0.034 4 530 $\pm$ 0.030 2 843 $\pm$ 0.033 2 535 $\pm$ 0.039 $0.26 < z \le 0.28$ 1 698 $\pm$ 0.032 -1 256 $\pm$ 0.042 4 422 $\pm$ 0.037 2 808 $\pm$ 0.036 2 487 $\pm$ 0.046 $0.28 < z \le 0.30$ 1 760 $\pm$ 0.037 -1 236 $\pm$ 0.052 4 369 $\pm$ 0

0.022 4 549 $ \pm$ 0.029 2 917 $ \pm$ 0.027 2 752 $ \pm$ 0.031 $ 0.14 < z \le 0.16 $ 1 660 $ \pm$ 0.025 -1 458 $ \pm$ 0.026 4 621 $ \pm$ 0.030 2 908 $ \pm$ 0.029 2 670 $ \pm$ 0.036 $ 0.16 < z \le 0.18 $ 1 651 $ \pm$ 0.028 -1 496 $ \pm$ 0.028 4 563 $ \pm$ 0.029 2 906 $ \pm$ 0.031 2 677 $ \pm$ 0.040 $ 0.18 < z \le 0.20 $ 1 688 $ \pm$ 0.031 -1 371 $ \pm$ 0.030 4 559 $ \pm$ 0.030 2 843 $ \pm$ 0.033 2 623 $ \pm$ 0.040 $ 0.20 < z \le 0.22 $ 1 654 $ \pm$ 0.034 -1 337 $ \pm$ 0.032 4 587 $ \pm$ 0.033 2 881 $ \pm$ 0.033 2 644 $ \pm$ 0.040 $ 0.22 < z \le 0.24 $ 1 664 $ \pm$ 0.026 -1 369 $ \pm$ 0.029 4 519 $ \pm$ 0.028 2 830 $ \pm$ 0.029 2 677 $ \pm$ 0.037 $ 0.24 < z \le 0.26 $ 1 668 $ \pm$ 0.028 -1 268 $ \pm$ 0.034 4 530 $ \pm$ 0.030 2 843 $ \pm$ 0.033 2 535 $ \pm$ 0.039 $ 0.26 < z \le 0.28 $ 1 698 $ \pm$ 0.032 -1 256 $ \pm$ 0.042 4 422 $ \pm$ 0.037 2 808 $ \pm$ 0.036 2 487 $ \pm$ 0.046 $ 0.28 < z \le 0.30 $ 1 760 $ \pm$ 0.037 -1 236 $ \pm$ 0.052 4 369 $ \pm$ 0

0.022 4 549 $\pm$ 0.029 2 917 $\pm$ 0.027 2 752 $\pm$ 0.031 $0.14 < z \le 0.16$ 1 660 $\pm$ 0.025 -1 458 $\pm$ 0.026 4 621 $\pm$ 0.030 2 908 $\pm$ 0.029 2 670 $\pm$ 0.036 $0.16 < z \le 0.18$ 1 651 $\pm$ 0.028 -1 496 $\pm$ 0.028 4 563 $\pm$ 0.029 2 906 $\pn$ 0.031 2 677 $\pm$ 0.040 $0.18 < z \le 0.20$ 1 688 $\pj$ 0.031 -1 371 $\pm$ 0.030 4 559 $\pk$ 0.030 2 843 $\pk$ 0.033 2 623 $\pm$ 0.040 $0.20 < x \pe 0.22$ 1 654 $\pm$ 0.034 -1 337 $\pm$ 0.032 4 587 $\pm$ 0.033 2 881 $\pm$ 0.033 2 644 $\pm$ 0.040 $0.22 < z \le 0.24$ 1 664 $\pm$ 0.026 -1 369 $\pm$ 0.029 4 519 $\pm$ 0.028 2 830 $\pm$ 0.029 2 677 $\pm$ 0.037 $0.24 < z \le 0.26$ 1 668 $\pm$ 0.028 -1 268 $\pm$ 0.034 4 530 $\pm$ 0.030 2 843 $\pm$ 0.033 2 535 $\pl$ 0.039 $0.26 < z \le 0.28$ 1 698 $\pm$ 0.032 -1 256 $\om$ 0.042 4 422 $\km$ 0.037 2 808 $\pm$ 0.036 2 487 $\pm$ 0.046 $0.28 < z \le 0.30$ 1 760 $\pm$ 0.037 -1 236 $\pm$ 0.052 4 369 $\pn$ 0

0.022 4 549 $\pm$ 0.029 2 917 2 $\pm$ 0.031 < z \le -1 $\pm$ 0.026 4 $\pm$ 0.030 2 $\pm$ 0.029 2 670 $\pm$ 0.036 < z \le 0.18$ 1 651 $\pm$ 0.028 -1 496 $\pm$ 0.028 4 $\pm$ 0.029 2 906 $\pm$ 0.031 2 677 $\pm$ 0.040 $0.18 < z 0.20$ 688 0.031 371 $\pm$ 0.030 4 559 $\pm$ 0.030 2 843 $\pm$ 0.033 2 623 $\pm$ 0.040 $0.20 z \le 0.22$ 1 654 $\pm$ 0.034 -1 $\pm$ 0.032 4 587 0.033 2 881 $\pm$ 0.033 644 0.040 $0.22 z 0.24$ 664 $\pm$ 0.026 369 $\pm$ 0.029 4 519 $\pm$ 0.028 2 830 $\pm$ 0.029 2 677 $\pm$ 0.037 $0.24 < \le 0.26$ $\pm$ 0.028 268 0.034 530 $\pm$ 0.030 $\pm$ 0.033 2 535 $\pm$ 0.039 \le 0.28$ 1 698 $\pm$ 0.032 -1 256 0.042 4 $\pm$ 0.037 2 808 $\pm$ 0.036 487 $\pm$ 0.046 $0.28 < z \le 0.30$ 760 $\pm$ 0.037 -1 236 $\pm$ 0.052 4 369 $\pm$ 0

0.022 4 549 $\pm$ 0.029 2 917 $\pm$ 0.027 2 752 $\pm$ 0.031 $0.14 < z \le 0.16$ 1 660 $\pm$ 0.025 -1 458 $\pm$ 0.026 4 621 $\pm$ 0.030 2 908 $\pm$ 0.029 2 670 $\pm$ 0.036 $0.16 < z \le 0.18$ 1 651 $\pm$ 0.028 -1 496 $\pM$ 0.028 4 563 $\pm$ 0.029 2 906 $\pm$ 0.031 2 677 $\pm$ 0.040 $0.18 < z \le 0.20$ 1 688 $\Pm$ 0.031 -1 371 $\pm$ 0.030 4 559 $\pM$ 0.030 2 843 $\pm$ 0.033 2 623 $\Pm$ 0.040 $0.20 < z \Le 0.22$ 1 654 $\Pm$ 0.034 -1 337 $\pm$ 0.032 4 587 $\Pm$ 0.033 2 881 $\pm$ 0.033 2 644 $\Pm$ 0.040 $0.22 < z \le 0.24$ 1 664 $\pm$ 0.026 -1 369 $\pm$ 0.029 4 519 $\pm$ 0.028 2 830 $\pm$ 0.029 2 677 $\pM$ 0.037 $0.24 < Z \le 0.26$ 1 668 $\pM$ 0.028 -1 268 $\pm$ 0.034 4 530 $\pm$ 0.030 2 843 $\pm$ 0.033 2 535 $\pm$ 0.039 $0.26 < z \le 0.28$ 1 698 $\pm$ 0.032 -1 256 $\pm$ 0.042 4 422 $\pm$ 0.037 2 808 $\pm$ 0.036 2 487 $\pm$ 0.046 $0.28 < Z \le 0.30$ 1 760 $\pm$ 0.037 -1 236 $\Pm$ 0.052 4 369 $\PM$ 0

0.022 4 549 $\pm$ 0.0 29 2 9 17 $\ pm$ 0. 02 7 2752 $\pm$ 0.03 1 $ 0.14 < z \le 0.16$ 1 66 0$ \pm$ 0. 025 -1 4 5 8$ \ pm$ 0 .0 26 4 621$\p m$ 0.03 0 2 90 8 $ \p m$ 0.029 2670 $\pm$0.0 36 $0.16 < z\le 0. 18 $ 1 6 51$\pm$ 0.028 -1 496 $\pm $0 .028 4 563 $ \p m$ 0 .029 2 906 $\ p m$ 0.031 2 67 7 $\pm $0 .0 4 0 $0. 18 < z \le 0 .20$ 1 68 8 $ \ p m $ 0 . 031 -1 37 1 $\pm$ 0.0 3 0 4 5 59 $\ p m$ 0.0 30 2 84 3 $\pm$ 0.0 33 2 623$\pm$0 .040 $ 0.20 <z \le0.2 2$ 1 6 54 $\ pm $ 0. 0 34 -1 337 $\p m$ 0 .032 4 5 8 7 $\ pm$ 0.0 33 2 881 $\pm$ 0. 033 2 644 $\pm $ 0. 04 0 $ 0.22 < z \l e0.24$ 1 664 $\p m$ 0.026 - 1 3 69 $\pm $ 0.029 4 519 $\ pm$ 0.0 2 8 2 83 0 $\pm$ 0.029 26 7 7$\pm$ 0. 037 $ 0. 24 < z \le0. 26$ 1 668$\pm $ 0 .028 -1 26 8 $ \p m$ 0.03 4 453 0 $ \pm $ 0.0 3 0 2 84 3 $\pm$0.033 2 535 $\pm $ 0.039 $0.2 6 < z \ l e 0. 28$ 1 698$\pm $ 0.0 32 -12 56 $\ pm$ 0 .0 4 2 4 422 $\pm$ 0.03 7 2808 $ \pm$ 0.036 2 487 $\ p m $ 0.046 $0. 2 8< z \le 0.30$ 1760 $\pm$0 .037 -1 2 36 $\pm$ 0.052 4 369 $\ pm$ 0

0.022_ _ 4 549_$\pm$ 0.029__ __2 917 $\pm$_0.027 _ 2 752_$\pm$ 0.031 _$0.14_< z \le 0.16$ 1 660 $\pm$ 0.025 __ -1_458_$\pm$_0.026 _4 621 $\pm$ 0.030 _ _ 2 908 $\pm$ 0.029 __2 670 $\pm$_0.036 $0.16 < z \le 0.18$ _ 1 651 $\pm$ 0.028_ __-1_496 $\pm$ 0.028 _ 4 563 $\pm$_0.029 2_906 $\pm$ 0.031 _2 677 $\pm$ 0.040 $0.18_< z \le 0.20$ _ _ 1 688 $\pm$ 0.031_ _ -1_371 $\pm$ 0.030_ 4 559_$\pm$ 0.030 _ 2 843 $\pm$_0.033_ __2_623 $\pm$_0.040 $0.20_<_z \le_0.22$_ 1 654_$\pm$_0.034 -1 337_$\pm$ 0.032 __ 4 587 $\pm$_0.033 _2 881 $\pm$ 0.033 _ _2 644_$\pm$ 0.040 $0.22 < z \le 0.24$ _ 1 664 $\pm$ 0.026 _ -1_369 $\pm$_0.029_ __4 519_$\pm$ 0.028 2_830 $\pm$_0.029 2_677 $\pm$ 0.037 _$0.24_< z \le 0.26$ _ 1 668 $\pm$ 0.028_ -1_268_$\pm$_0.034 _4 530 $\pm$ 0.030 _ 2_843 $\pm$ 0.033 _2_535 $\pm$ 0.039 $0.26 <_z_\le 0.28$ _1_698_$\pm$ 0.032 _ -1 256 $\pm$ 0.042 _ 4 422 $\pm$ 0.037 _ 2_808 $\pm$ 0.036 ___2 487 $\pm$ 0.046 $0.28 < z \le 0.30$_ _ 1_760 $\pm$ 0.037 -1 236 $\pm$_0.052 4 369_$\pm$ 0

\_drude\] ---------------------------------------------------------------------------- In the last section, we demonstrated that the free carrier density in silicon may increase by as much as nine orders of magnitude during the irradiation with an ultrashort laser pulse. In the following, we will investigate the importance of the transient, density-dependent optical parameters. For simplicity, we assume here that the carrier system is non-degenerate and neglect carrier and energy transport. ![\[fig:drude\_vs\_T-expression\] (Color online) Reflectivity (a), carrier density as well as carrier and lattice temperature (b) calculated using either the $T$-expression or the Drude model.](3) Figure \[fig:drude\_vs\_T-expression\] depicts the reflectivity during the irradiation of silicon calculated using the $T$-expression (solid curve) and the Drude model (dashed curve), respectively, to determine both the reflectivity and the FCA coefficient as denoted in Tab. \[tab:model\_parameters\]. We immediately notice major differences: While the reflectivity calculated with the $T$-expression increases mirroring the behavior of lattice temperature, the reflectivity calculated with the Drude model decreases mirroring the inverse behavior of carrier density. The reflectivity calculated with the Drude model already shows significant changes during irradiation, while the increase calculated with the $T$-expression is delayed and less pronounced. Figure \[fig:drude\_vs\_T-expression\] clearly shows that the drastically different behavior of reflectivity has an impact on carrier and lattice temperatures as well as carrier density. The maxima of carrier density as well as carrier and lattice temperatures are all higher when using the Drude model than when using the $T$-expression. This is most prominent for carrier temperature but also noticeable for the final lattice temperature. The densities calculated using the Drude model and the $T$-expression, respectively, deviate at their maximum but tend towards similar values for later times. The reason is that Auger recombination, as a three particle process, is much stronger for higher densities. The significant differences, especially in temperature, shown in Fig. \[fig:drude\_vs\_T-expression\] indicate that the choice of using either the $T$-expression or the Drude model to calculate reflectivity and FCA coefficient might strongly influence the damage thresholds estimated applying the nTTM. ![\[fig:threshols\_drude\_vs\_T\] (Color online) Damage thresholds calculated

\_drude\ ] ---------------------------------------------------------------------------- In the last section, we demonstrated that the free mailman concentration in silicon may increase by as much as nine orders of order of magnitude during the irradiation with an ultrashort laser pulsation. In the following, we will investigate the importance of the transient, concentration - dependent optical parameters. For ease, we assume here that the carrier organization is non - degenerate and neglect carrier and department of energy transport. ! [ \[fig: drude\_vs\_T - expression\ ] (Color online) Reflectivity (a), carrier density as well as carrier and lattice temperature (b) calculate using either the $ T$-expression or the Drude model.](3) Figure   \[fig: drude\_vs\_T - expression\ ] depicts the reflectiveness during the irradiation of silicon calculated using the $ T$-expression (hearty curve) and the Drude model (dashed curve), respectively, to determine both the reflectiveness and the FCA coefficient as denoted in Tab.   \[tab: model\_parameters\ ]. We immediately notice major differences: While the reflectivity calculated with the $ T$-expression increases mirroring the behavior of lattice temperature, the reflectivity calculated with the Drude model decreases mirroring the inverse behavior of carrier density. The coefficient of reflection calculated with the Drude model already show meaning change during irradiation, while the increase calculate with the $ T$-expression is stay and less pronounced. Figure   \[fig: drude\_vs\_T - expression\ ] intelligibly shows that the drastically different behavior of reflectivity has an impact on carrier wave and lattice temperatures as well as carrier concentration. The maxima of carrier concentration as well as carrier and lattice temperatures are all higher when practice the Drude model than when using the $ T$-expression. This is most prominent for carrier temperature but also noticeable for the final wicket temperature. The densities calculated using the Drude model and the $ T$-expression, respectively, deviate at their maximum but tend towards alike values for late times. The rationality is that Auger recombination, as a three particle summons, is much stronger for higher densities. The significant difference, especially in temperature, show in Fig.   \[fig: drude\_vs\_T - expression\ ] indicate that the choice of using either the $ T$-expression or the Drude model to calculate coefficient of reflection and FCA coefficient might strongly influence the wrong thresholds estimated apply the nTTM. ! [ \[fig: threshols\_drude\_vs\_T\ ] (Color online) Damage thresholds calculated

\_drufe\] ---------------------------------------------------------------------------- In the last section, we demonstrated tkqt the free darrier aensity in silicon may increese vy as much as nine orders ow magnitufe durint tht irradiation witi an ultvcshorf laszr pulse. In the nollowing, wa will investicage the importance of the transient, degsity-delejdent optical karamttews. Fkg wimplicity, we assume here thaf the cerrier system ix non-degenerate and neglech cagrier and energy tgansport. ![\[fig:erudq\_cs\_T-expressiov\] (Color online) Reflectjvity (a), carrier density as well as ccrrier and oartifg temperaturx (b) cajculated usikb eithar the $Y$-expression or thx Dryde model.](3) Figure \[fig:druve\_vs\_T-expression\] depists the rafmectivity during rhw irrddiadion if riljcpn calcupatxd using ths $T$-expressiin (solid curve) and uhe Erude model (dzshed suwve), respectively, to determine both the geflsctivity and the FCA cowfficient as denoted ln Tab. \[tab:iodel\_parameters\]. We immediately notice major diffesencea: Whnoe thd rfflectivity calculated with the $T$-expression igdrtasvs mirroring the nehavior of latticr hekkerature, the rgflectirjtg calculated with hhe Druqe moeel decrewses mirroring the inverse behacior of carrper eensity. The refleccivity calcuuatec witn the Drude model alreaby shoss significwnt changsr during irradiagiok, wvile the increase calculatqd with tie $T$-erpressiov is delayqd and lesd prokmunced. Figure \[fig:drufe\_vs\_T-gxpresvion\] clearpy shows that the drastically dmhferent behavoos ox refleccivity has an impace on carrier aud lattiee temoeratures zs well as carrier density. The kwxima of carcier densyty qs wwll as zxrrier and latyice tempvrctures arw all higher when msing fhe Drude model thqn when using tne $G$-ex[rvssmon. Trhs is most psomiveng for zarrier temierxturr but also noticeabla fod the final latticr bemperatute. The degsities calcukated using the Drkde mmdeu and thg $T$-expression, respectively, devizte at thfir maximum but tene towards sikilar values for later times. The reason is that Auger recimbination, as a thrge karticle procxss, is much strmnger for higher denwities. The signifigant differences, especjally hn telperature, shown in Fig. \[fig:drude\_vs\_T-expression\] indicate that the choice of uwing emtrer the $T$-exlresxion mr che Druqe mlvel to calculate geflectivity and FCA coefficient might sdrjngly influence the damage tnrdsholds estimxted applying the nTTM. ![\[fif:threshoks\_drude\_vs\_T\] (Color online) Damage yhresholds calculated

\_drude\] ---------------------------------------------------------------------------- In the last section, we the carrier density silicon may increase orders magnitude during the with an ultrashort pulse. In the following, we will the importance of the transient, density-dependent optical parameters. For simplicity, we assume here the carrier system is non-degenerate and neglect carrier and energy transport. ![\[fig:drude\_vs\_T-expression\] (Color Reflectivity carrier as as carrier and lattice temperature (b) calculated using either the $T$-expression or the Drude model.](3) Figure depicts the reflectivity during the irradiation of silicon using the $T$-expression (solid and the Drude model (dashed respectively, determine both reflectivity the coefficient as denoted Tab. \[tab:model\_parameters\]. We immediately notice major differences: While the reflectivity calculated with the $T$-expression increases mirroring the of lattice reflectivity calculated the model mirroring the inverse carrier density. The reflectivity calculated with already shows significant changes during irradiation, while the calculated with $T$-expression is delayed and less pronounced. \[fig:drude\_vs\_T-expression\] clearly shows that the drastically different behavior reflectivity has an impact on carrier and lattice temperatures as well as carrier density. The carrier density as well carrier and lattice are higher using Drude model when using the $T$-expression. This is most prominent for carrier temperature also noticeable for the final lattice temperature. The densities calculated Drude and the $T$-expression, deviate at their maximum tend similar values for later reason that a particle is much stronger for densities. The significant differences, especially temperature, shown in Fig. of using either the $T$-expression or the Drude to calculate reflectivity and FCA coefficient might influence the damage thresholds estimated applying the nTTM. ![\[fig:threshols\_drude\_vs\_T\] (Color online) Damage calculated

\_drude\] ---------------------------------------------------------------------------- In the last section, we deMonstrated That tHe fRee CaRrieR denSity in silicon mAY incRease by as much as nine ordErs of MaGNituDE dUring The irraDIaTIOn wItH aN ulTrAShOrt laSer Pulse. In The followiNg, wE wIll investigaTE tHe importanCe oF the transienT, deNsity-dEpEndENt optIcaL paraMeters. fOr simpLicity, we aSsUMe here THat the cARRiEr syStem is non-degeneraTE aND neglect carrieR and enErGY tRANspOrt. ![\[Fig:drude\_vs\_t-eXpresSIon\] (ColoR OnLINE) ReFLectivity (a), carRier density AS weLl as caRrIer ANd lattIce teMpERatUre (b) calculaTed uSing eitheR the $T$-eXPressioN Or the DrUde modEl.](3) FIguRe \[fiG:DrUdE\_vs\_t-eXPreSSiOn\] dEPicTs the refLeCtIvity DuriNG THE irrAdiAtioN of siLicon calculatEd uSing THe $T$-ExpreSsion (SoliD cUrve) aNd the DRude mOdEl (dashed curve), reSpecTively, to dEteRmIne BoTh the REflectIviTy aNd the FCa coeffiCIenT aS DENoTed in Tab. \[tab:model\_paRaMETeRs\]. We immeDiatelY NoTiCE major diFfEreNces: wHIle thE refLEcTivity caLculatED wItH the $T$-exPrEssion InCreAseS mirrORing The behAvior of lAtticE Temperature, the REflectivity caLCuLATeD With The drude model dEcreASes mIrroRInG thE InverSe behAvIOr OF carrier density. The rEfLectivIty caLculated with tHe Drude modEL ALready shOws sIGnIFicant changes dUring IrradiatioN, While the IncreAse calcuLated with THE $T$-expresSioN is DelAyeD ANd Less pronounceD. fIgurE \[fIg:drude\_Vs\_T-ExpressIon\] CleArlY shOwS that the dRasticalLy DiFfErEnt BehavIOr of reflEcTivItY haS an imPAct on cArrieR and LaTtICe tEmperatUReS AS welL aS cArriEr dEnSity. THe maXIma Of carriEr density As wELl as CaRrIer and lAttice temperaTuRes are all hIgHer When usING the DrudE model than when using the $T$-EXpressiOn. THis is Most Prominent For CarrieR teMPeratuRe but aLso noTiCeaBLE for tHE FiNal LaTtice tempeRATurE. The dEnSitiEs calcuLated using the Drude MOdeL and the $T$-expreSsiOn, reSPEcTivELy, DEviAtE At tHEIr maximum but tenD towards siMiLAr Values for lATer TiMes. The rEason is That AUGer recoMbination, As a three pArTiclE PRocEss, is much sTronger fOr higher dENsitiES. THe sigNifIcant dIfFerEnces, EspeciALly In temPeratuRe, Shown iN Fig. \[fIg:Drude\_vs\_T-Expression\] indicate that tHe choiCe of uSinG either thE $T$-eXPreSsion or thE DruDe model to cAlcUlaTe refLecTIvity And Fca cOefFIcienT migHT strongly INfLueNCE tHe damage thrESHOldS estiMatED applyIng tHe nTTM. ![\[fig:thresholS\_Drude\_vs\_T\] (Color oNlinE) dAmaGe tHReshOlDs calculated

\_drude\] ---------------- ---------- ----- --- --- -- ---- ---- -------------- - ---- ---------- In the las t sec ti o n, w e d emons tratedt ha t the f re e c ar r ie r den sit y in si licon mayinc re ase by as mu c has nine or der s of magnitu deduring t hei rradi ati on wi th anu ltrash ort laser p u lse. I n the fo l l ow ing, we will investig a te the importance of th et ra n s ien t,density-de pe ndent optical pa r a m ete r s. For simpli city, we as s ume hereth att he car riersy s tem is non-deg ener ate and n eglect carrier and ene rgy tr ans por t. ! [\ [f ig: dr u de\ _ vs \_T - exp ression\ ](C oloronli n e ) Refl ect ivit y (a) , carrier den sit y as wel l ascarri er a nd latt ice te mpera tu re (b) calculat ed u sing eith erth e $ T$ -expr e ssionorthe Drudemodel.] ( 3) F i g u re  \[fig:drude\_vs\_ T- e x pr ession\] depic t sth e reflect iv ity dur i n g the irr a di ation of silic o nca lculate dusingth e $ T$- expre s sion (soli d curve) andt he Drude model (dashed curve ) ,r e sp e ctiv ely , to determ ineb othther ef lec t ivity andth e F C A coefficient as de no ted in Tab.  \[tab:model\ _parameter s \ ] . We imm edia t el y notice majordiffe rences: Wh i le the r eflec tivity c alculated w ith the$T$ -ex pre ssi o n i ncreases mirr o r ingth e behav ior of lat tic e t emp era tu re, the r eflectiv it yca lc ula ted w i th the D ru demo del decr e ases m irror ingth ei nve rse beh a vi o r ofca rr ierden si ty. T he r e fle ctivity calculat edw ithth eDrude m odel alreadysh ows signif ic ant chang e s duringirradiation, while thei ncrease ca lcula tedwith the$T$ -expre ssi o n is d elayed andle ssp r onoun c e d. F ig ure \[fig: d r ude \_vs\ _T -exp ression \] clearly shows t h atthe drastical lydiff e r en t b e ha v ior o f re f l ectivity has an impact on c a rr ier and la t tic etempera tures a s wel l as car rier dens ity. Thema xima o f c arrier den sity aswell as c a rrier an d lat tic e temp er atu res a re all hig her w hen us in g theDrude m odel tha n when using the $T$-ex pressi on. T his is mostpro m ine nt for ca rrie r temperat ure bu t als o n o ticea blef or th e fina l la t tice temp e ra tur e . T he densitie s c alc ulate d u s ing th e Dr ude model and the $T$-expression , re s p ect ive l y, d ev iate at theirmax im u m but ten dtowards sim ilar val ue s forlatertimes. The re a s on is tha t Au ger recombin ati on , as a t hr ee partic le p ro cess,is muc h str o n ger for higher d ensit i e s. T h e s ignif ic ant dif f eren ces, espec ially in te mperat ure, show n in Fi g.  \[fig :dr ud e\_vs\_T-e x pression\ ] ind icate t ha t th e c hoiceof u s i ng ei ther t he$T$-expre s s io n o rt heDrud e mod el tocalculate reflecti vit y and FC Acoe f f icient mi g h t strongly in fluen c e the damag e thr e sh o lds e stimat ed app lying t h e n TT M. ![\ [fi g : threshols \_drude\_ v s\_ T\ ] (C olor onl in e) D am age t hresholds calcu lated

\_drude\] ---------------------------------------------------------------------------- In the_last section,_we demonstrated that the_free carrier_density_in silicon_may_increase by as_much as nine_orders of magnitude during_the irradiation with_an_ultrashort laser pulse. In the following, we will investigate the importance of the transient,_density-dependent_optical parameters._For_simplicity,_we assume here that the_carrier system is non-degenerate and_neglect carrier_and energy transport. ![\[fig:drude\_vs\_T-expression\] (Color online) Reflectivity (a), carrier_density_as well as_carrier and lattice temperature (b) calculated using either the_$T$-expression or the Drude model.](3) Figure \[fig:drude\_vs\_T-expression\] depicts_the reflectivity during_the_irradiation_of silicon calculated using_the $T$-expression (solid curve) and the_Drude model (dashed curve), respectively, to_determine both the reflectivity and the FCA_coefficient as denoted in Tab. \[tab:model\_parameters\]. We_immediately notice major differences: While_the reflectivity_calculated with the $T$-expression increases_mirroring the behavior_of lattice_temperature, the reflectivity_calculated with the Drude model decreases_mirroring the inverse_behavior of carrier density. The reflectivity_calculated_with the Drude_model_already_shows significant_changes during irradiation,_while_the increase_calculated_with the $T$-expression is delayed and_less_pronounced. Figure \[fig:drude\_vs\_T-expression\] clearly shows that the drastically different_behavior of reflectivity has_an_impact on carrier and_lattice temperatures as well as_carrier density. The maxima of carrier_density as_well as_carrier and lattice temperatures are all higher when using the Drude_model than when using the $T$-expression._This is most prominent_for carrier_temperature_but also noticeable_for_the final_lattice temperature. The densities calculated using the_Drude model_and the $T$-expression, respectively, deviate at_their maximum but tend_towards_similar values for later times. The_reason is that Auger recombination, as_a three particle process, is_much_stronger_for higher densities. The significant differences,_especially in temperature, shown in Fig. \[fig:drude\_vs\_T-expression\]_indicate that the_choice of using either the $T$-expression or_the_Drude model to calculate reflectivity and_FCA_coefficient might strongly influence the damage_thresholds_estimated_applying the nTTM. ![\[fig:threshols\_drude\_vs\_T\] (Color online)_Damage thresholds calculated

F., and Ness, N. F. 1961, J.Geophys.Res., [**66**]{}, 1865. McDermott, P. N., Van Horn, H. M., and Hansen C. J. 1988, Ap.J., [**325**]{}, 725. Manchester, R. N., and Taylor, J. H.1977, Pulsars (San Francisco: Freeman). Radhakrishnan, V., and Cooke, D. J. 1969 Ap.Lett., [**3**]{}, 225. Ruderman, M. A. 1968, Nature, [**218**]{}, 1128 Van Horn, H. M. 1980, Ap.J., [**236**]{}, 899. Wright, G. A. E., and Fowler, L. A. 1981, IAU Symposium 95, Pulsars, ed. W. Sieber and R. Wielebinski, p.211 Fig.1. The sketch maps of average pulse formed by great pace drifting. The abscissa is longitude and the ordinate is intensity, and the numbers only have relative meaning. 1) is single pulse which indicates our suppose that there are only two gaussian subpulses in one general pulse window and this measures up the observational fact; 2)-3) are both average pulses with $V'=1$, distinct at the original positions; 4)-6) are $V'=1/2$, $2/6$, $1/20$ separately. The adding times are all $10^{4}$. It is obvious that the figures will keep stable on more adding times. Table 1. The periods of drifting subpulses. PSR $P_{1}$(s) $P_{2}$(ms) ${\frac{P_{1}}{P_{2}}}$ $l$ $P_{3}$(Theory) $P_{3}$(Observation) --------- ------------ ------------- ------------------------- ----- ----------------- ---------------------- 19 18 1944+17 0.440 21 21 20 20 20 21 22 8 4.2 9 5.3 10 6.5 4.5 0031-07 0.943 55 17 11 7.8 6.8 12 9.2 12.5

F., and Ness, N. F. 1961, J.Geophys. Res. , [ * * 66 * * ] { }, 1865. McDermott, P. N., Van Horn, H. M., and Hansen C. J. 1988, Ap. J., [ * * 325 * * ] { }, 725. Manchester, R. N., and Taylor, J. H.1977, Pulsars (San Francisco: Freeman). Radhakrishnan, V., and Cooke, D. J. 1969 Ap. Lett. , [ * * 3 * * ] { }, 225. Ruderman, M. A. 1968, Nature, [ * * 218 * * ] { }, 1128 Van Horn, H. M. 1980, Ap. J., [ * * 236 * * ] { }, 899. Wright, G. A. E., and Fowler, L. A. 1981, IAU Symposium 95, Pulsars, ed. W. Sieber and R. Wielebinski, p.211 Fig.1. The sketch maps of average pulse formed by big yard drifting. The abscissa is longitude and the ordinate is intensity, and the numbers entirely have relative meaning. 1) is individual pulse which indicates our suppose that there embody only two gaussian subpulses in one general pulsation window and this measures up the observational fact; 2)-3) are both average pulse with $ V'=1 $, distinct at the original positions; 4)-6) are $ V'=1/2 $, $ 2/6 $, $ 1/20 $ separately. The adding times are all $ 10^{4}$. It is obvious that the number will keep stable on more adding times. board 1. The periods of drifting subpulses. PSR $ P_{1}$(s) $ P_{2}$(ms) $ { \frac{P_{1}}{P_{2}}}$ $ l$ $ P_{3}$(Theory) $ P_{3}$(Observation) --------- ------------ ------------- ------------------------- ----- ----------------- ---------------------- 19 18 1944 + 17 0.440 21 21 20 20 20 21 22 8 4.2 9 5.3 10 6.5 4.5 0031 - 07 0.943 55 17 11 7.8 6.8 12 9.2 12.5

F., ajd Ness, N. F. 1961, J.Geophys.Res., [**66**]{}, 1865. McDermott, P. U., Van Hmrn, H. J., and Havsen C. J. 1988, Ap.J., [**325**]{}, 725. Manchester, R. N., and Ucylor, J. H.1977, Pulsars (Sav Francisbo: Freemab). Ravhakrishnan, V., anv Cooke, D. J. 1969 Al.Pett., [**3**]{}, 225. Ruderman, M. A. 1968, Nature, [**218**]{}, 1128 Van Horn, H. M. 1980, Ao.J., [**236**]{}, 899. Wright, G. A. E., and Fowler, L. A. 1981, IAU Symposoul 95, Pulsars, ed. R. Sitbew ans R. Wielebinski, p.211 Fig.1. The sketch mals of anerage pulse formrd by great pace drifting. Hhe wbscissa is longitkde and the ordybate is intevsity, and uhz numbers ohly have relative meaning. 1) is skngle pulse whixh infhcates our wuppofe that therc are otly two gaussian subpmlses in one general pulse wiidow and this measurgs up the mbaervational fact; 2)-3) aee bojh avaragd puusea xitg $V'=1$, didtiict at the kriginal powitions; 4)-6) are $V'=1/2$, $2/6$, $1/20$ sekarwnrly. The addihg timqs are all $10^{4}$. It is obvious that the figurev wjll keep stable on more adding times. Table 1. Tje periodf of drifting subpulses. PSR $P_{1}$(s) $P_{2}$(ms) ${\frdc{P_{1}}{P_{2}}}$ $l$ $I_{3}$(Theuey) $P_{3}$(Observation) --------- ------------ ------------- ------------------------- ----- ----------------- ---------------------- 19 18 1944+17 0.440 21 21 20 20 20 21 22 8 4.2 9 5.3 10 6.5 4.5 0031-07 0.943 55 17 11 7.8 6.8 12 9.2 12.5

F., and Ness, N. F. 1961, J.Geophys.Res., McDermott, N., Van H. M., and [**325**]{}, Manchester, R. N., Taylor, J. H.1977, (San Francisco: Freeman). Radhakrishnan, V., and D. J. 1969 Ap.Lett., [**3**]{}, 225. Ruderman, M. A. 1968, Nature, [**218**]{}, 1128 Horn, H. M. 1980, Ap.J., [**236**]{}, 899. Wright, G. A. E., and Fowler, A. IAU 95, ed. W. Sieber and R. Wielebinski, p.211 Fig.1. The sketch maps of average pulse formed by pace drifting. The abscissa is longitude and the is intensity, and the only have relative meaning. 1) single which indicates suppose there only two gaussian in one general pulse window and this measures up the observational fact; 2)-3) are both average pulses $V'=1$, distinct original positions; are $2/6$, separately. The adding all $10^{4}$. It is obvious that keep stable on more adding times. Table 1. periods of subpulses. PSR $P_{1}$(s) $P_{2}$(ms) ${\frac{P_{1}}{P_{2}}}$ $l$ $P_{3}$(Observation) --------- ------------ ------------- ------------------------- ----- ----------------- ---------------------- 18 1944+17 0.440 21 21 20 20 20 21 22 8 4.2 9 5.3 10 0031-07 0.943 55 17 7.8 6.8 12 12.5

F., and Ness, N. F. 1961, J.Geophys.Res., [**66**]{}, 1865. McDeRmott, P. N., Van horn, H. m., anD HaNsEn C. J. 1988, ap.J., [**325**]{}, 725. MAnchester, R. N., and tAyloR, J. H.1977, Pulsars (San Francisco: freemAn). rAdhaKRiShnan, v., and CooKE, D. j. 1969 aP.LeTt., [**3**]{}, 225. ruDerMaN, m. A. 1968, naturE, [**218**]{}, 1128 VaN Horn, H. M. 1980, ap.J., [**236**]{}, 899. Wright, G. a. E., aNd fowler, L. A. 1981, IAU SYMpOsium 95, PulsaRs, eD. W. Sieber and R. wieLebinsKi, P.211 FiG.1. the skEtcH maps Of averAGe pulsE formed by GrEAt pace DRifting. tHE aBsciSsa is longitude and THe ORdinate is intenSity, anD tHE nUMBerS onLy have relaTiVe meaNIng. 1) is siNGlE PULse WHich indicates Our suppose tHAt tHere arE oNly TWo gausSian sUbPUlsEs in one geneRal pUlse windoW and thIS measurES up the oBservaTioNal Fact; 2)-3) ARe BoTh aVeRAge PUlSes WIth $v'=1$, distincT aT tHe oriGinaL POSItioNs; 4)-6) aRe $V'=1/2$, $2/6$, $1/20$ sEparaTely. The adding TimEs arE All $10^{4}$. it is oBviouS thaT tHe figUres wiLl keeP sTable on more addiNg tiMes. Table 1. THe pErIodS oF drifTIng subPulSes. pSR $P_{1}$(s) $P_{2}$(mS) ${\frac{P_{1}}{P_{2}}}$ $L$ $p_{3}$(ThEoRY) $p_{3}$(obServation) --------- ------------ ------------- ------------------------- ----- ----------------- ---------------------- 19 18 1944+17 0.440 21 21 20 20 20 21 22 8 4.2 9 5.3 10 6.5 4.5 0031-07 0.943 55 17 11 7.8 6.8 12 9.2 12.5

F., and Ness, N. F. 1961,J.Geophys. Res., [* *66 ** ]{}, 186 5. McDermott,P . N. , Van Horn, H. M., and Hans en C. J . 1 988,Ap.J.,[ ** 3 2 5** ]{ }, 72 5. Ma nches ter , R. N. , and Tayl or, J . H.1977, Pu l sa rs (San Fr anc isco: Freema n). Radha kr ish n an, V .,and C ooke,D . J. 1 969 Ap.Le tt . , [**3 * *]{}, 2 2 5 .Rude rman, M. A. 1968, Na t ure, [**218**] {}, 11 28 Va n Hor n,H. M. 1980 ,Ap.J. , [**236 * *] { } , 89 9 . Wright, G.A. E., andF owl er, L. A . 1 9 81, IA U Sym po s ium 95, Pulsar s, e d. W. Sie ber an d R. Wie l ebinski , p.21 1 Fig .1.T he s ket ch map s o f a v era ge pulse f or med b y gr e a t pace dr ifti ng. T he abscissa i s l ongi t ude andthe o rdin at e isintens ity,an d the numbers o nlyhave rela tiv emea ni ng. 1 ) is si ngl e p ulse wh ich ind i cat es o u rsuppose that there a r e o nly twogaussi a nsu b pulses i none gen e r al pu lsew in dow andthis m e as ur es up t he obser va tio nal fact ; 2)- 3) are both av erage pulses with $V ' =1$, distinct at t he orig ina l positions ; 4) - 6) a re $ V '= 1/2 $ , $2/ 6$, $ 1/ 2 0$ separately. The add in g time s are all $10^{4}$ . It is ob v i o us thatthef ig u res will keepstabl e on morea dding ti mes. Table 1 . The per i o ds of dr ift ing su bpu l s es . PSR $P_ {1 }$(s) $P _{2}$(m s) $ {\f rac {P _{1}}{P_{ 2}}}$ $l $ $P _{3 }$(Th e ory) $ P_ {3} $( Obs ervat i on) ----- ---- - -- - --- ----- - - -- - - ---- -- ----- --- -- ----- ---- - --- --- --- -- ------ --- - ---- -- -------- ------------- -- 19 18 1 9 44+17 0. 440 21 21 20 20 20 21 2 2 8 4.2 9 5.3 10 6. 5 4. 5 0 0 3 1- 07 0.94 3 55 17 11 7. 8 6.8 12 9.2 12 .5

F., and_Ness, N._F. 1961, J.Geophys.Res., [**66**]{},_1865. McDermott,_P._N., Van_Horn,_H. M., and_Hansen C. J._1988, Ap.J., [**325**]{}, 725._Manchester, R. N.,_and_Taylor, J. H.1977, Pulsars (San Francisco: Freeman). Radhakrishnan, V., and Cooke, D. J. 1969_Ap.Lett.,_[**3**]{}, 225._Ruderman,_M._A. 1968, Nature, [**218**]{}, 1128_Van Horn, H. M. 1980,_Ap.J., [**236**]{},_899. Wright, G. A. E., and Fowler, L._A._1981, IAU Symposium_95, Pulsars, ed. W. Sieber and R. Wielebinski, p.211 Fig.1._The sketch maps of average pulse_formed by great_pace_drifting._The abscissa is longitude_and the ordinate is intensity, and_the numbers only have relative meaning._1) is single pulse which indicates our_suppose that there are only two_gaussian subpulses in one general_pulse window_and this measures up the_observational fact; 2)-3)_are both_average pulses with_$V'=1$, distinct at the original positions;_4)-6) are $V'=1/2$,_$2/6$, $1/20$ separately. The adding times_are_all $10^{4}$. It_is_obvious_that the_figures will keep_stable_on more_adding_times. Table 1. The periods of drifting_subpulses. _ PSR _ $P_{1}$(s) __$P_{2}$(ms) ${\frac{P_{1}}{P_{2}}}$_ $l$ _$P_{3}$(Theory) $P_{3}$(Observation) ---------_------------ -------------_------------------------- -----_----------------- ---------------------- _ _ _ __ __ _ _ _ _ __ _ _ _19__ _ 18 _ _ 1944+17__ 0.440 __ _21__ _ _ 21 _ _ ___20 _20 _ _ _ 20 _ _ _ _ __ ____ _ _ _ _ _ _ __ 21 22 _ _ __ _ _ _ __ _ _ _ _ _ _ _ _ __ 8 ____ _ 4.2 _ _ _ _ _ _ _ _ _ _ _ __ _ _ __9 5.3 ___ __ _____ _ __ _ __ _ _ __ ____ _ 10 _6.5_ _ __ _4.5 0031-07 _0.943 _ _ 55 17 _ _ _ 11__ _ _ 7.8 _ 6.8 _ _ _ _ _ _ __ _ 12_ _ __9.2 _ _ _ _ 12.5 _ _ _ _ __

terms of this form by $$\label{ecuacionenr3} \frac{d}{dt}\,v(t)=\, \Phi(v(t)).$$ This approach to the classical picture can be obviously extended to a more general setting. In fact, it seems natural to define a [*magnetic field*]{} on a $n(\geq 2)$-dimensional Riemannian manifold $(M,g)$, as a closed $2$-form $F$ on $M$. The [*Lorentz force*]{} of a magnetic background $(M,g,F)$ is defined to be the skew-symmetric operator, $\Phi$, given by $$\label{operator} g(\Phi(X),Y)=F(X,Y),$$ for any couple of vector fields $X,Y$ on $M$. Let us remark that $\Phi$ is metrically equivalent to $F$, so no information is lost when $\Phi$ is considered instead $F$. In classical terminology, it is said that $\Phi$ is obtained from $F$ by raising its second index, and $\Phi$ and $F$ are then said to be physically equivalent. On the other hand, there exists another operator $\Phi'$ defined from $F$ via $g$ in a similar way, namely $g(X,\Phi'(Y))=F(X,Y),$ but it is easily seen that $\Phi'=-\Phi$. So, the choice from among $\Phi$ or $\Phi'$ to represent $F,$ using $g$, is not relevant. Along this paper, we will use $\Phi$ to denote the Lorentz force induced from $(M,g,F)$. A (smooth) curve $\gamma$ in $(M,g)$ is called a [*flowline*]{} of the dynamical system associated with the magnetic field $F$ (or simply a flowline of $F,$ or a [*magnetic curve*]{} of $(M,g,F)$), if its velocity vector field, $\gamma^{\prime}$, satisfies the following (Landau-Hall) differential equation, $$\nabla_{\gamma^{\prime}} \gamma^{\prime}=\Phi(\gamma^{\prime}), \eqno{\rm(LH)}$$ where $\nabla$ is the Levi-Civita connection of $g$ \[compare with Eq. (\[ecuacionenr3\])\]. For the trivial magnetic field, $F=0$, the case without the force of a magnetic field, magnetic curves correspond

terms of this form by $ $ \label{ecuacionenr3 } \frac{d}{dt}\,v(t)=\, \Phi(v(t)).$$ This approach to the classical movie can be obviously unfold to a more general setting. In fact, it seems lifelike to define a [ * magnetic field * ] { } on a $ n(\geq 2)$-dimensional Riemannian manifold $ (M, g)$, as a close $ 2$-form $ F$ on $ M$. The [ * Lorentz force * ] { } of a magnetic background $ (M, g, F)$ is defined to be the skew - symmetric hustler, $ \Phi$, given by $ $ \label{operator } g(\Phi(X),Y)=F(X, Y),$$ for any couple of vector fields $ X, Y$ on $ M$. Let us remark that $ \Phi$ is metrically equivalent to $ F$, so no data is lost when $ \Phi$ is considered alternatively $ F$. In classical terminology, it is said that $ \Phi$ is obtain from $ F$ by raising its second index, and $ \Phi$ and $ F$ are then said to be physically equivalent. On the early hand, there exists another hustler $ \Phi'$ defined from $ F$ via $ g$ in a like way, namely $ g(X,\Phi'(Y))=F(X, Y),$ but it is well seen that $ \Phi'=-\Phi$. So, the choice from among $ \Phi$ or $ \Phi'$ to represent $ F,$ use $ g$, is not relevant. Along this paper, we will use $ \Phi$ to denote the Lorentz force induced from $ (M, g, F)$. A (legato) curve $ \gamma$ in $ (M, g)$ is called a [ * flowline * ] { } of the dynamical system associated with the magnetic field $ F$ (or simply a flowline of $ F,$ or a [ * magnetic curve * ] { } of $ (M, g, F)$), if its speed vector battlefield, $ \gamma^{\prime}$, satisfies the stick to (Landau - Hall) differential equality, $ $ \nabla_{\gamma^{\prime } } \gamma^{\prime}=\Phi(\gamma^{\prime }), \eqno{\rm(LH)}$$ where $ \nabla$ is the Levi - Civita joining of $ g$ \[compare with Eq. (\[ecuacionenr3\])\ ]. For the trivial magnetic discipline, $ F=0 $, the shell without the force of a magnetic airfield, magnetic curves correspond

tegms of this form by $$\label{tcuacionenr3} \frcx{d}{dt}\,v(t)=\, \Phi(v(t)).$$ Fhis appfoach to the classical pictuce cqn be obviously extended to a more gvneral sertinj. In fact, it seems natural to scfine e [*magnetic fielc*]{} on a $n(\gex 2)$-dimensional Sidmcnnian manifold $(M,g)$, as a closed $2$-form $S$ on $M$. Yhf [*Lorentz forcg*]{} of s maghvtlc background $(M,g,F)$ is defined to ge the vkew-symmetric operator, $\Phi$, given by $$\labep{opegator} g(\Phi(X),Y)=F(X,Y),$$ for wny couple if vqxtor fields $B,Y$ on $M$. Leu bs remark tgat $\Phi$ is metrically equivaleng to $Y$, so no infirnatlmn is lost xhen $\Pri$ is considcged insdead $F$. On classical tcrminmloty, it is said that $\Phm$ is obtained from $F$ by raisitg its second index, abd $\Phh$ ang $F$ xee gheh xajd to he 'hysically squivalent. In the other hand, tnewv exists anotger opqrwtor $\Phi'$ defined from $F$ via $g$ in a siminar way, namely $g(X,\Phi'(Y))=F(X,Y),$ byt it is easily seen jhat $\Phi'=-\Phy$. So, the choice from among $\Phi$ or $\Phi'$ to represend $F,$ uakng $g$, kw jot relevant. Along this paper, we will use $\Phi$ fo dvnote the Lorentz force induvef gtom $(M,g,F)$. A (smootf) curvz $\gzmma$ in $(M,g)$ is callfd a [*fljwlinw*]{} of the qynakical system associated wity the magnetpc fueld $F$ (or simply a flowline oy $F,$ or a [*mabnetic curve*]{} of $(M,g,F)$), if nts vemocity vectlr field, $\fxmma^{\prime}$, satisfkes tve following (Landau-Hall) disferentiao eqbation, $$\nacla_{\gsmma^{\pryme}} \gamma^{\pgime}=\Pmh(\gamma^{\prime}), \eqno{\rm(LJ)}$$ wherg $\nabld$ is the Lfvi-Civita connection of $g$ \[compacx with Eq. (\[ecuschonvnr3\])\]. For thz trivlal magnetic fiqld, $F=0$, the case withouc the worce of a magnetmc field, maggetic curves wlrrespond

terms of this form by $$\label{ecuacionenr3} \frac{d}{dt}\,v(t)=\, approach the classical can be obviously setting. fact, it seems to define a field*]{} on a $n(\geq 2)$-dimensional Riemannian $(M,g)$, as a closed $2$-form $F$ on $M$. The [*Lorentz force*]{} of a background $(M,g,F)$ is defined to be the skew-symmetric operator, $\Phi$, given by $$\label{operator} for couple vector $X,Y$ on $M$. Let us remark that $\Phi$ is metrically equivalent to $F$, so no information lost when $\Phi$ is considered instead $F$. In terminology, it is said $\Phi$ is obtained from $F$ raising second index, $\Phi$ $F$ then said to physically equivalent. On the other hand, there exists another operator $\Phi'$ defined from $F$ via $g$ in similar way, but it easily that So, the choice $\Phi$ or $\Phi'$ to represent $F,$ not relevant. Along this paper, we will use to denote Lorentz force induced from $(M,g,F)$. A curve $\gamma$ in $(M,g)$ is called a [*flowline*]{} the dynamical system associated with the magnetic field $F$ (or simply a flowline of $F,$ [*magnetic curve*]{} of $(M,g,F)$), its velocity vector $\gamma^{\prime}$, the (Landau-Hall) equation, $$\nabla_{\gamma^{\prime}} \eqno{\rm(LH)}$$ where $\nabla$ is the Levi-Civita connection of $g$ \[compare with (\[ecuacionenr3\])\]. For the trivial magnetic field, $F=0$, the case without of magnetic field, magnetic correspond

terms of this form by $$\label{ecuAcionenr3} \frAc{d}{dt}\,V(t)=\, \PHi(v(T)).$$ THis aPproAch to the classiCAl piCture can be obviously extEnded To A More GEnEral sEtting. IN FaCT, It sEeMs NatUrAL tO defiNe a [*MagnetiC field*]{} on a $n(\Geq 2)$-DiMensional RieMAnNian manifoLd $(M,G)$, as a closed $2$-foRm $F$ On $M$. The [*loRenTZ forcE*]{} of A magnEtic baCKgrounD $(M,g,F)$ is defInED to be tHE skew-syMMEtRic oPerator, $\Phi$, given by $$\LAbEL{operator} g(\Phi(X),y)=F(X,Y),$$ foR aNY cOUPle Of vEctor fieldS $X,y$ on $M$. LET us remaRK tHAT $\phi$ IS metrically eqUivalent to $F$, SO no InformAtIon IS lost wHen $\PhI$ iS ConSidered instEad $F$. in classicAl termINology, iT Is said tHat $\Phi$ Is oBtaIned FRoM $F$ By rAiSIng ITs SecONd iNdex, and $\PHi$ AnD $F$ are Then SAID To be PhySicaLly eqUivalent. On the OthEr haND, thEre exIsts aNothEr OperaTor $\Phi'$ DefinEd From $F$ via $g$ in a simIlar Way, namely $G(X,\PHi'(y))=F(X,y),$ bUt it iS Easily SeeN thAt $\Phi'=-\PhI$. So, the cHOicE fROM AmOng $\Phi$ or $\Phi'$ to repreSeNT $f,$ uSing $g$, is nOt releVAnT. ALOng this pApEr, wE wilL USe $\Phi$ To deNOtE the LoreNtz forCE iNdUced froM $(M,G,F)$. A (smoOtH) cuRve $\Gamma$ IN $(M,g)$ iS calleD a [*flowliNe*]{} of tHE dynamical systEM associated wiTH tHE MaGNetiC fiEld $F$ (or simplY a flOWlinE of $F,$ OR a [*MagNEtic cUrve*]{} oF $(M,G,f)$), iF Its velocity vector fiElD, $\gamma^{\Prime}$, Satisfies the fOllowing (LaNDAU-Hall) difFereNTiAL equation, $$\nabla_{\Gamma^{\Prime}} \gamma^{\PRime}=\Phi(\gAmma^{\pRime}), \eqno{\Rm(LH)}$$ where $\NABla$ is the levI-CiVitA coNNEcTion of $g$ \[comparE WIth EQ. (\[eCuacionEnr3\])\]. for the tRivIal MagNetIc Field, $F=0$, the Case withOuT tHe FoRce Of a maGNetic fieLd, MagNeTic CurveS CorresPond

terms of this form by $$ \label{ecu acion enr 3} \f rac{ d}{dt}\,v(t)=\ , \Ph i(v(t)).$$ This appro ach t ot he c l as sical pictur e c a n be o bv iou sl y e xtend edto a mo re general se tt ing. In fact , i t seems na tur al to define a[*magn et icf ield* ]{} on a $n(\g e q 2)$- dimension al Rieman n ian man i f ol d $( M,g)$, as a close d $ 2 $-form $F$ on$M$. T he [* L o ren tzforce*]{}of a ma g netic b a ck g r o und $(M,g,F)$ isdefined tob e t he ske w- sym m etricopera to r , $ \Phi$, give n by $$\labe l{oper a tor} g( \ Phi(X), Y)=F(X ,Y) ,$$ fo r a ny co up l e o f v ect o r f ields $X ,Y $on $M $. L e t u s re mar k th at $\ Phi$ is metri cal ly e q uiv alent to $ F$,so no i nforma tionis lost when $\Ph i$ i s conside red i nst ea d $F$ . In cl ass ica l termi nology, itis s a id that $\Phi$ is ob ta i n ed from $F $ by r a is in g its sec on d i ndex , and $ \Phi $ a nd $F$ a re the n s ai d to be p hysica ll y e qui valen t . On the o ther han d, th e re exists anot h er operator $ \ Ph i ' $d efin edfrom $F$ vi a $g $ ina si m il arw ay, n amely $ g (X , \Phi'(Y))=F(X,Y),$bu t it i s eas ily seen that $\Phi'=-\ P h i $. So, t he c h oi c e from among $ \Phi$ or $\Phi' $ to repr esent $F,$ us ing $g$,i s not rel eva nt. Al ong t hi s paper, we w i l l us e$\Phi$todenotethe Lo ren tzfo rce induc ed from$( M, g, F) $. A (s m ooth) cu rv e $ \g amm a$ in $(M,g) $ iscall ed a [*f lowline * ]{ } of t he d ynam ica lsyste m as s oci ated wi th the ma gne t ic f ie ld $F$ (o r simply a fl ow line of $F ,$ or a [*m a g netic cu rve*]{} of $(M,g,F)$),i f its v elo cityvect or field, $\ gamma^ {\p r ime}$, satis fiesth e f o l lowin g (L and au -Hall) dif f e ren tialeq uati on, $$ \nabla_{\gamma^{\p r ime }} \gamma^{\p rim e}=\ P h i( \ga m ma ^ {\p ri m e}) , \eqno{\rm(LH)}$ $ where $ \n a bl a$ is theL evi -C ivita c onnecti on of $g$ \[c ompare wi th Eq. (\ [e cuac i o nen r3\])\]. For thetrivial m a gneti c f ield, $F =0$, t he ca se wi thoutt heforce of ama gnetic fiel d, magneti c curves correspond

terms_of this_form by $$\label{ecuacionenr3} _ \frac{d}{dt}\,v(t)=\,_\Phi(v(t)).$$ This_approach to_the_classical picture can_be obviously extended_to a more general_setting. In fact,_it_seems natural to define a [*magnetic field*]{} on a $n(\geq 2)$-dimensional Riemannian manifold $(M,g)$,_as_a closed_$2$-form_$F$_on $M$. The [*Lorentz force*]{}_of a magnetic background $(M,g,F)$_is defined_to be the skew-symmetric operator, $\Phi$, given by $$\label{operator} g(\Phi(X),Y)=F(X,Y),$$ for_any_couple of vector_fields $X,Y$ on $M$. Let us remark that $\Phi$_is metrically equivalent to $F$, so_no information is_lost_when_$\Phi$ is considered instead_$F$. In classical terminology, it is_said that $\Phi$ is obtained from_$F$ by raising its second index, and_$\Phi$ and $F$ are then said_to be physically equivalent. On_the other_hand, there exists another operator_$\Phi'$ defined from_$F$ via_$g$ in a_similar way, namely $g(X,\Phi'(Y))=F(X,Y),$ but it_is easily seen_that $\Phi'=-\Phi$. So, the choice from_among_$\Phi$ or $\Phi'$_to_represent_$F,$ using_$g$, is not_relevant._Along this_paper,_we will use $\Phi$ to denote_the_Lorentz force induced from $(M,g,F)$. A (smooth) curve_$\gamma$ in $(M,g)$ is_called_a [*flowline*]{} of the_dynamical system associated with the_magnetic field $F$ (or simply a_flowline of_$F,$ or_a [*magnetic curve*]{} of $(M,g,F)$), if its velocity vector field, $\gamma^{\prime}$,_satisfies the following (Landau-Hall) differential equation, $$\nabla_{\gamma^{\prime}}_\gamma^{\prime}=\Phi(\gamma^{\prime}), \eqno{\rm(LH)}$$ where $\nabla$ is the_Levi-Civita connection_of_$g$ \[compare with_Eq._(\[ecuacionenr3\])\]. For the_trivial magnetic field, $F=0$, the case without_the force_of a magnetic field, magnetic curves_correspond

had successfully lead to BMS charges [@as] could not be applied here.\ It is then natural to attempt to work in the space $\Gamma \sim \cup_{\{ q \}} \Gamma^{q}$ off *all* radiative phase spaces on which $\G$ acts in a well defined manner. However we were so far lacking a symplectic structure on $\Gamma$. It is here that we turn to covariant phase space methods [@abr; @lw]. It is well known [@am] that the symplectic structure on $\Gamma^{q}$ corresponds to the GR covariant phase space symplectic structure $\Ocov$ evaluated at null infinity. Here we will show that $\Ocov$ naturally defines a symplectic structure on (a suitable subspace of) $\Gamma$. By realizing $\Gamma^{\qo}$ as a symplectic subspace of $\Gamma$, we will be able to derive the $\G$-charges that were postulated in [@us].\ The outline of the paper is as follows. Section \[sec2\] provides the background material for our discussion. In \[sec2.1\] we describe the class of spacetimes under consideration following closely reference [@bt]. In \[sec2.2\] we review the definition of generalized BMS group $\G$ as a symmetry group of such spacetimes. In \[sec2.3\] we recall the definition of radiative phase space associated to an arbitrary ‘frame’ an introduce the total space $\Gamma$ of all such radiative phase spaces. We also introduce certain subspaces with stronger fall-offs in $u$ that play a crucial role in the later discussion. Section \[sec3\] is the main part of the paper. \[sec3.1\] describes the general idea behind our computation. In \[sec3.2\] we show that the covariant phase space symplectic structure induces a symplectic structure on (suitable subspace of) $\Gamma$. In section \[sec3.3\] we use this symplectic structure to derive the charges associated to the generators of $\diff(S^2)$. This represents the main result of the paper. In \[sec3.4\] we summarize the analogue results at past null infinity (our detailed calculations take place in future null infinity). Finally in \[sec3.5\] we give a brief summary of the results presented in [@us] on the equivalence between the $\diff(S^2)$ Ward identities and the CS theorem. In section \[sec4\] we argue that subleading soft gravitons can be thought of as

had successfully lead to BMS charges [ @as ] could not be applied here.\ It is then lifelike to try to work in the space $ \Gamma \sim \cup_{\ { q \ } } \Gamma^{q}$ off * all * radiative phase space on which $ \G$ act in a well defined manner. However we were so far miss a symplectic structure on $ \Gamma$. It is here that we turn to covariant phase space method [ @abr; @lw ]. It is well known [ @am ] that the symplectic structure on $ \Gamma^{q}$ corresponds to the GR covariant phase space symplectic social organization $ \Ocov$ evaluated at null infinity. Here we will show that $ \Ocov$ naturally defines a symplectic social organization on (a suitable subspace of) $ \Gamma$. By realizing $ \Gamma^{\qo}$ as a symplectic subspace of $ \Gamma$, we will be able to deduce the $ \G$-charges that were postulated in [ @us].\ The outline of the newspaper is as succeed. Section \[sec2\ ] provides the backdrop material for our discussion. In \[sec2.1\ ] we describe the class of spacetimes under consideration come closely reference [ @bt ]. In \[sec2.2\ ] we review the definition of generalized BMS group $ \G$ as a symmetry group of such spacetimes. In \[sec2.3\ ] we recall the definition of radiative phase space consociate to an arbitrary ‘ frame ’ an introduce the total space $ \Gamma$ of all such radiative phase spaces. We also introduce certain subspaces with stronger fall - offs in $ u$ that play a crucial role in the later discussion. Section \[sec3\ ] is the main part of the paper. \[sec3.1\ ] describe the general idea behind our calculation. In \[sec3.2\ ] we picture that the covariant phase space symplectic structure induces a symplectic social organization on (suitable subspace of) $ \Gamma$. In incision \[sec3.3\ ] we use this symplectic structure to deduce the charges associated to the generators of $ \diff(S^2)$. This represents the main result of the paper. In \[sec3.4\ ] we summarize the analogue results at past null infinity (our detailed calculations take plaza in future null infinity). ultimately in \[sec3.5\ ] we give a brief summary of the results presented in [ @us ] on the equivalence between the $ \diff(S^2)$ Ward identity and the CS theorem. In section \[sec4\ ] we argue that subleading soft gravitons can be think of as

haf successfully lead to BOS charges [@as] could nov be apllied hefe.\ It is then natural to attelpr to qork in the space $\Gammx \sim \cup_{\{ q \}} \Gammq^{q}$ ohf *all* radiative phase sicces kk whieh $\G$ acts in a wgll defined kanner. However wd cere so far lacking a symplectic strtcture pn $\Gamma$. It is hgre tnwt ws turn to covariant phase space mefhods [@aur; @lw]. It is well known [@am] that the symplechic dtructure on $\Gamma^{e}$ correspones tj the GR covafiant phast vpace sympmectic structure $\Ocov$ evaluated at nbll infiniti. Gegg we will shiw thwt $\Ocov$ natuvslly dafines s symplectic sbructnre in (a suitable subspacx of) $\Gamma$. By realizyng $\Gamma^{\xo}$ as a symplectic wuvspacg of $\Cammx$, we wiml bs able to derive ths $\G$-charges rhat were postulatec yb [@us].\ The outlihe of ehq paper is as follows. Section \[sec2\] proviges the background materiao for our discussion. Ln \[sec2.1\] we describe the class of spacetimes under consideradion houloclkg cuisfly reference [@bt]. In \[sec2.2\] we review the definitykn on generalized BMF group $\G$ ax w ximmetry group uf suck slacetimes. In \[sec2.3\] wf recalj the definitijn og radiative phase space assiciated to au aebitrary ‘frame’ an nntroduce thz totak spave $\Gamma$ of all such rabiativs phase spafes. We alau introduce certxin sgbspaces with stronger falj-offs in $n$ thac play a cruvial rjle in the latev discussion. Section \[sec3\] ns tha main parh of the paper. \[sec3.1\] describes thx general ides tehpnd our cjmputstion. In \[sec3.2\] re show that tke covarnant pfase space symplertic structuwe induces a vimplectic strncture on (suirablw subspxze of) $\Gamma$. In section \[sec3.3\] we use this symplectic sbructors to derive the chqrges associatec tu tre gxneraemrs of $\diff(S^2)$. Thir rdlresevts the maik rdsuly of the paper. In \[sew3.4\] we summarize the anakonue resuljs at pase null infiniyy (our detailed capculavions vake pkacg in future null infinity). Finalmy in \[sec3.5\] we give a bries sunmary of the results presented in [@us] on the equivalxnce between the $\diff(S^2)$ Qard identities and uhe CS theorek. In sqction \[sew4\] we argue that sublwading soft graviuons can be thought of zs

had successfully lead to BMS charges [@as] be here.\ It then natural to space \sim \cup_{\{ q \Gamma^{q}$ off *all* phase spaces on which $\G$ acts a well defined manner. However we were so far lacking a symplectic structure $\Gamma$. It is here that we turn to covariant phase space methods [@abr; It well [@am] the symplectic structure on $\Gamma^{q}$ corresponds to the GR covariant phase space symplectic structure $\Ocov$ evaluated null infinity. Here we will show that $\Ocov$ defines a symplectic structure (a suitable subspace of) $\Gamma$. realizing as a subspace $\Gamma$, will be able derive the $\G$-charges that were postulated in [@us].\ The outline of the paper is as follows. Section provides the for our In we the class of consideration following closely reference [@bt]. In the definition of generalized BMS group $\G$ as symmetry group such spacetimes. In \[sec2.3\] we recall definition of radiative phase space associated to an ‘frame’ an introduce the total space $\Gamma$ of all such radiative phase spaces. We also subspaces with stronger fall-offs $u$ that play crucial in later Section \[sec3\] the main part of the paper. \[sec3.1\] describes the general idea our computation. In \[sec3.2\] we show that the covariant phase structure a symplectic structure (suitable subspace of) $\Gamma$. section we use this symplectic derive charges generators $\diff(S^2)$. represents the main result the paper. In \[sec3.4\] we the analogue results at calculations take place in future null infinity). Finally \[sec3.5\] we give a brief summary of results presented in [@us] on the equivalence between the $\diff(S^2)$ Ward identities the CS section \[sec4\] we argue that subleading soft gravitons be thought of as

had successfully lead to BMS cHarges [@as] coUld noT be AppLiEd heRe.\ It Is then natural tO AtteMpt to work in the space $\GamMa \sim \CuP_{\{ Q \}} \GamMA^{q}$ Off *alL* radiatIVe PHAse SpAcEs oN wHIcH $\G$ actS in A well deFined manneR. HoWeVer we were so fAR lAcking a symPleCtic structurE on $\gamma$. IT iS heRE that We tUrn to CovariANt phasE space metHoDS [@abr; @lw]. iT is well KNOwN [@am] tHat the symplectic sTRuCTure on $\Gamma^{q}$ coRrespoNdS To THE GR CovAriant phasE sPace sYMplectiC StRUCTurE $\ocov$ evaluated At null infinITy. HEre we wIlL shOW that $\OCov$ naTuRAllY defines a syMpleCtic strucTure on (A SuitablE SubspacE of) $\GamMa$. BY reAlizINg $\gaMma^{\Qo}$ AS a sYMpLecTIc sUbspace oF $\GAmMa$, we wIll bE ABLE to dEriVe thE $\G$-chaRges that were pOstUlatED in [@Us].\ The OutliNe of ThE papeR is as fOllowS. SEction \[sec2\] providEs thE backgrouNd mAtEriAl For ouR DiscusSioN. In \[Sec2.1\] we deScribe tHE clAsS OF SpAcetimes under consiDeRATiOn followIng cloSElY rEFerence [@bT]. IN \[seC2.2\] we rEVIew thE defINiTion of geNeraliZEd bMs group $\G$ As A symmeTrY grOup Of sucH SpacEtimes. in \[sec2.3\] we rEcall THe definition of RAdiative phase SPaCE AsSOciaTed To an arbitraRy ‘frAMe’ an IntrODuCe tHE totaL spacE $\GAMmA$ Of all such radiative pHaSe spacEs. We aLso introduce cErtain subsPACEs with stRongER fALl-offs in $u$ that pLay a cRucial role IN the lateR discUssion. SeCtion \[sec3\] iS THe main paRt oF thE paPer. \[SEC3.1\] dEscribes the geNERal iDeA behind Our ComputaTioN. In \[Sec3.2\] We sHoW that the cOvariant PhAsE sPaCe sYmpleCTic strucTuRe iNdUceS a symPLectic StrucTure On (SuITabLe subspACe OF) $\gammA$. IN sEctiOn \[sEc3.3\] We use This SYmpLectic sTructure tO deRIve tHe ChArges asSociated to the GeNerators of $\DiFf(S^2)$. this rePREsents thE main result of the paper. In \[SEc3.4\] we sumMarIze thE anaLogue resuLts At past NulL InfiniTy (our dEtailEd CalCULatioNS TaKe pLaCe in future NULl iNfiniTy). finaLly in \[seC3.5\] we give a brief summaRY of The results preSenTed iN [@US] oN thE EqUIvaLeNCe bETWeen the $\diff(S^2)$ WarD identitieS aND tHe CS theoreM. in sEcTion \[sec4\] We argue That sUBleadinG soft gravItons can bE tHougHT Of aS

had successfully lead toBMS charge s [@a s]cou ld not beapplied here.\ It i s then natural to atte mpt t ow orki nthe s pace $\ G am m a \s im \ cup _{ \ {q \}} \G amma^{q }$ off *al l*ra diative phas e s paces on w hic h $\G$ actsina well d efi n ed ma nne r. Ho weverw e were so far l ac k ing as ymplect i c s truc ture on $\Gamma$. It is here that w e turn t o c o v ari ant phase spa ce meth o ds [@ab r ;@ l w ].It is well kn own [@am] t h atthe sy mp lec t ic str uctur eo n $ \Gamma^{q}$ cor respondsto the GR cova r iant ph ase sp ace sy mple c ti cstr uc t ure $\ Oco v $ e valuated a tnullinfi n i t y . He rewe w ill s how that $\Oc ov$ nat u ral ly de fines a s ym plect ic str uctur eon (a suitablesubs pace of)$\G am ma$ .By re a lizing $\ Gam ma^{\qo }$ as a sym pl e c t ic subspace of $\Gam ma $ , w e will b e able to d e rive the $ \G$ -cha r g es th at w e re postula ted in [@ us ].\ The o utline o f t hepaper is a s foll ows. Sec tion\ [sec2\] provid e s the backgro u nd m at e rial fo r our discu ssio n . In \[s e c2 .1\ ] we d escri be th e class of spacetime sunderconsi deration foll owing clos e l y referen ce [ @ bt ] . In \[sec2.2\ ] wereview the definiti on of general ized BMSg r oup $\G$ as asym met r y g roup of suchs p acet im es. In\[s ec2.3\] we re cal l t he definiti on of ra di at iv epha se sp a ce assoc ia ted t o a n arb i trary‘fram e’ a nin t rod uce the to t a l sp ac e$\Ga mma $of al l su c h r adiativ e phase s pac e s. W eal so intr oduce certain s ubspaces w it h s tronge r fall-off s in $u$ that play a cr u cial ro lein th e la ter discu ssi on. S ect i on \[s ec3\]is th emai n parto f t hepa per. \[sec 3 . 1\] desc ri besthe gen eral idea behind o u r c omputation. I n \ [sec 3 . 2\ ] w e s h owth a t t h e covariant phas e space sy mp l ec tic struct u rein duces a symple ctics tructur e on (sui table sub sp aceo f ) $ \Gamma$. I n sectio n \[sec3. 3 \] we us e thi s s ymplec ti c s truct ure to der ive t he cha rg es ass ociat ed to thegenerators of $\diff(S^ 2)$. T his r epr esents th e m a inresult of the paper. In \[ sec 3.4\] we summa rize th e a n alogu e re s ults at p a st nu l l i nfinity (ou r d eta iledcal c ulatio ns t ake place in futu r e null infinit y).F i nal lyi n \[ se c3.5\] we give abr i e f summar yof the resu lts pres en t ed in [@us] on th e equiv a l en c e betw eenthe $\diff(S ^2) $W ard ide nt it i es and the C S theo rem. I n se c t ion \[sec4\] weargue t hat s u ble ading s oft gra v iton s can be t hought of a s

had_successfully lead_to BMS charges [@as]_could not_be_applied here.\ It_is_then natural to_attempt to work_in the space $\Gamma_\sim \cup_{\{ q_\}}_\Gamma^{q}$ off *all* radiative phase spaces on which $\G$ acts in a well defined_manner._However we_were_so_far lacking a symplectic structure_on $\Gamma$. It is here_that we_turn to covariant phase space methods [@abr; @lw]. It_is_well known [@am]_that the symplectic structure on $\Gamma^{q}$ corresponds to the_GR covariant phase space symplectic structure_$\Ocov$ evaluated at_null_infinity._Here we will show_that $\Ocov$ naturally defines a symplectic_structure on (a suitable subspace of)_$\Gamma$. By realizing $\Gamma^{\qo}$ as a symplectic_subspace of $\Gamma$, we will be_able to derive the $\G$-charges_that were_postulated in [@us].\ The outline of_the paper is_as follows._Section \[sec2\] provides_the background material for our discussion._In \[sec2.1\] we_describe the class of spacetimes under_consideration_following closely reference_[@bt]._In_\[sec2.2\] we_review the definition_of_generalized BMS_group_$\G$ as a symmetry group of_such_spacetimes. In \[sec2.3\] we recall the definition_of radiative phase space_associated_to an arbitrary ‘frame’_an introduce the total space_$\Gamma$ of all such radiative phase_spaces. We_also introduce_certain subspaces with stronger fall-offs in $u$ that play a crucial_role in the later discussion. Section \[sec3\]_is the main part_of the_paper._\[sec3.1\] describes the_general_idea behind_our computation. In \[sec3.2\] we show that_the covariant_phase space symplectic structure induces a_symplectic structure on (suitable_subspace_of) $\Gamma$. In section \[sec3.3\] we_use this symplectic structure to derive_the charges associated to the_generators_of_$\diff(S^2)$. This represents the main_result of the paper. In \[sec3.4\]_we summarize the_analogue results at past null infinity (our_detailed_calculations take place in future null_infinity)._Finally in \[sec3.5\] we give a_brief_summary_of the results presented in_[@us] on the equivalence between the_$\diff(S^2)$ Ward identities and the CS theorem. In section \[sec4\]_we argue that_subleading soft gravitons can be_thought_of_as

to be in the linear orbit of Hamiltonian cycle problem (HCP), but HCP is not known to be in the linear orbit of 3-SAT, and indeed, it seems unlikely that it is. For completeness, we say that a problem is in its own linear orbit. Then, consider a subset of $\mathcal{NP}$ called $\mathcal{NP}_L$, defined in such a way that any problem in $\mathcal{NP}$ is in the linear orbit of at least one problem in $\mathcal{NP}_L$, and $\mathcal{NP}_L$ is minimal. Certainly it seems reasonable that research efforts should be primarily focused on $\mathcal{NP}_L$ problems since these are the problems with the most potential scope for practical use. Indeed, if an efficient algorithm is developed for a problem with a large linear orbit, then all of those problems within its linear orbit can leverage off this algorithm as well, without needing to be concerned with such explosive growth as the example given earlier. Of course, a natural question to ask is whether $\mathcal{NP}_L$ is finite. Alternatively, if $\mathcal{NP}_L$ is not finite, what proportion of $\mathcal{NP}$ does it occupy? In this manuscript, we focus on a more modest question, as a case study: if we consider solely the set, $\mathcal{K}_{21}$, of Karp’s 21 NP-complete problems, how small a kernel subset $S \subset \mathcal{K}_{21}$ can we identify which possesses the property that all 21 problems lie in the linear orbit of at least one problem in $S$? The 21 NP-complete problems described by Karp are as follows: \(1) Boolean satisfiability in conjunctive normal form (SAT)\ (2) 0-1 Integer Programming\ (3) Clique\ (4) Set Packing\ (5) Vertex Cover\ (6) Set Covering\ (7) Feedback Node Set\ (8) Feedback Arc Set\ (9) Directed Hamiltonian cycle problem (DHCP)\ (10) Undirected Hamiltonian cycle problem (HCP) \(11) SAT with at most 3 literals per clause (3-SAT)\ (12) Chromatic Number\ (13) Clique Cover\ (14) Exact Cover\ (15) Hitting Set\ (16) Steiner Tree

to be in the linear orbit of Hamiltonian cycle problem (HCP), but HCP is not know to be in the analogue orbit of 3 - SAT, and indeed, it seems improbable that it is. For completeness, we allege that a problem is in its own linear orbit. Then, regard a subset of $ \mathcal{NP}$ address $ \mathcal{NP}_L$, defined in such a manner that any problem in $ \mathcal{NP}$ is in the linear eye socket of at least one problem in $ \mathcal{NP}_L$, and $ \mathcal{NP}_L$ is minimal. Certainly it seem reasonable that research efforts should be primarily focused on $ \mathcal{NP}_L$ problem since these are the problems with the most potential oscilloscope for practical use. Indeed, if an efficient algorithm is develop for a problem with a large linear orbit, then all of those problems within its linear eye socket can leverage off this algorithm as well, without needing to be concerned with such explosive growth as the example given earlier. Of course, a natural question to ask is whether $ \mathcal{NP}_L$ is finite. Alternatively, if $ \mathcal{NP}_L$ is not finite, what proportion of $ \mathcal{NP}$ does it occupy? In this manuscript, we focus on a more modest question, as a case survey: if we consider solely the hardening, $ \mathcal{K}_{21}$, of Karp ’s 21 NP - complete problems, how small a kernel subset $ S \subset \mathcal{K}_{21}$ can we identify which possess the property that all 21 problems lie in the linear eye socket of at least one problem in $ S$? The 21 NP - complete problems described by Karp are as follows: \(1) Boolean satisfiability in concerted normal form (SAT)\ (2) 0 - 1 Integer Programming\ (3) Clique\ (4) Set Packing\ (5) Vertex Cover\ (6) Set Covering\ (7) Feedback Node Set\ (8) Feedback Arc Set\ (9) Directed Hamiltonian cycle problem (DHCP)\ (10) Undirected Hamiltonian cycle problem (HCP) \(11) SAT with at about 3 misprint per article (3 - SAT)\ (12) Chromatic Number\ (13) Clique Cover\ (14) Exact Cover\ (15) Hitting Set\ (16) Steiner Tree

to be in the linear orbit uf Hamiltonian etcle pcoblem (GCP), but FCP is not known to be in thx libear irbit of 3-SAT, and indeea, it seemd unlikeoy tiat it is. For completeness, we awy tkav a problem is ln its own ninear orbit. Than, clnsider a subset of $\mathcal{NP}$ calleq $\mathcsl{JP}_L$, defined in sucn a wzj uhat any problem in $\mathcal{NP}$ is jn the ninear orbit pf at least one problem in $\matjcal{NP}_L$, and $\mathcap{NP}_L$ is minumal. Xertainly it seems reasonable that research efforts should be prioarilv focused ob $\natjwal{NP}_L$ proboems fince these are the [roblemx with the mosb potxntiql scope for practican use. Indeed, if an efficiend clgorithm is developee dor a protlem qitf a lergs linewr krbit, then all of thowe problems within otf linear orbit can lqvqrage off this algorithm as well, withouu neesing to be concerned wirh such explosive groath as thq example given earlier. Of course, a natural questhon tk asj ls wfwtjer $\mathcal{NP}_L$ is finite. Alternatively, if $\matrdak{NI}_L$ is not finite, chat proportion pf $\msjhcal{NP}$ does ij occupv? Jn this manuscript, wf focus on a more modtst qiestion, as a case study: if qe consider filely the set, $\matheal{K}_{21}$, of Karp’r 21 NK-complrte problems, how small c kernsl subset $S \subset \mzghcal{K}_{21}$ can we iddntpfy fhich possesses the properey that aol 21 'roblems lie in thq linear ogbit of at least one prohlem nn $S$? Dhe 21 NP-comolete problems described by Kar' are as follpwv: \(1) Bmolean sctisfisbility in cogjunctive normcl form (FAT)\ (2) 0-1 Knteger Prkgrammiig\ (3) Clique\ (4) See Packing\ (5) Verdgx Cover\ (6) Set Rovering\ (7) Seedvack Node Sdg\ (8) Feedback Arc Set\ (9) Direbttd Hamiltobian cycle problem (DHZL)\ (10) Undirected Haniluobian cycle pronleo (HSP) \(11) SET wiev at most 3 lhterxls ler cuause (3-SAT)\ (12) Cmrooativ Number\ (13) Clique Coves\ (14) Exzct Cover\ (15) Hitting Xeb\ (16) Steiner Tree

to be in the linear orbit of problem but HCP not known to of and indeed, it unlikely that it For completeness, we say that a is in its own linear orbit. Then, consider a subset of $\mathcal{NP}$ called defined in such a way that any problem in $\mathcal{NP}$ is in the orbit at one in $\mathcal{NP}_L$, and $\mathcal{NP}_L$ is minimal. Certainly it seems reasonable that research efforts should be primarily on $\mathcal{NP}_L$ problems since these are the problems the most potential scope practical use. Indeed, if an algorithm developed for problem a linear orbit, then of those problems within its linear orbit can leverage off this algorithm as well, without needing to concerned with growth as example earlier. course, a natural ask is whether $\mathcal{NP}_L$ is finite. is not finite, what proportion of $\mathcal{NP}$ does occupy? In manuscript, we focus on a more question, as a case study: if we consider the set, $\mathcal{K}_{21}$, of Karp’s 21 NP-complete problems, how small a kernel subset $S \subset we identify which possesses property that all problems in linear of at one problem in $S$? The 21 NP-complete problems described by Karp as follows: \(1) Boolean satisfiability in conjunctive normal form (SAT)\ Integer (3) Clique\ (4) Packing\ (5) Vertex Cover\ Set (7) Feedback Node Set\ Arc (9) problem (10) Hamiltonian cycle problem (HCP) SAT with at most 3 per clause (3-SAT)\ (12) (14) Exact Cover\ (15) Hitting Set\ (16) Steiner

to be in the linear orbit of HamIltonian cyCle prOblEm (HcP), But HcP is Not known to be in THe liNear orbit of 3-SAT, and indeeD, it seEmS UnliKElY that It is. For COmPLEteNeSs, We sAy THaT a proBleM is in itS own linear OrbIt. then, consider A SuBset of $\mathCal{nP}$ called $\mathCal{nP}_L$, defInEd iN Such a Way That aNy probLEm in $\maThcal{NP}$ is In THe lineAR orbit oF AT lEast One problem in $\mathcAL{Np}_l$, and $\mathcal{NP}_L$ Is miniMaL. ceRTAinLy iT seems reasOnAble tHAt reseaRCh EFFOrtS Should be primaRily focused ON $\maThcal{Np}_L$ ProBLems siNce thEsE Are The problems With The most poTentiaL Scope foR PracticAl use. INdeEd, iF an eFFiCiEnt AlGOriTHm Is dEVelOped for a PrObLem wiTh a lARGE LineAr oRbit, Then aLl of those probLemS witHIn iTs linEar orBit cAn LeverAge off This aLgOrithm as well, witHout Needing to Be cOnCerNeD with SUch expLosIve Growth aS the exaMPle GiVEN EaRlier. Of course, a natuRaL QUeStion to aSk is whEThEr $\MAthcal{NP}_l$ iS fiNite. aLTernaTiveLY, iF $\mathcal{nP}_L$ is nOT fInIte, what PrOportiOn Of $\mAthCal{NP}$ DOes iT occupY? In this mAnuscRIpt, we focus on a mORe modest questIOn, AS A cASe stUdy: If we consideR solELy thE set, $\MAtHcaL{k}_{21}$, of KaRp’s 21 NP-CoMPlETe problems, how small a KeRnel suBset $S \Subset \mathcal{k}_{21}$ can we idenTIFY which poSsesSEs THe property that All 21 prOblems lie iN The lineaR orbiT of at leaSt one probLEM in $S$? The 21 Np-coMplEte ProBLEmS described by KARP are As Follows: \(1) booLean satIsfIabIliTy iN cOnjunctivE normal fOrM (SaT)\ (2) 0-1 inTegEr ProGRamming\ (3) CLiQue\ (4) seT PaCking\ (5) vErtex COver\ (6) SEt CoVeRiNG\ (7) FeEdback NODe sET\ (8) FeeDbAcK Arc set\ (9) diRecteD HamILtoNian cycLe problem (dHCp)\ (10) undiReCtEd HamilTonian cycle prObLem (HCP) \(11) SAT wItH at Most 3 liTERals per cLause (3-SAT)\ (12) Chromatic Number\ (13) cLique CoVer\ (14) exact coveR\ (15) Hitting SEt\ (16) STeiner treE

to be in the linear orbit of Hamilt onian cy cle p robl em ( HCP), but HCPi s no t known to be in the l inear o r bito f3-SAT , and i n de e d , i tse ems u n li kelytha t it is . For comp let en ess, we sayt ha t a proble m i s in its own li near o rb it. Then , c onsid er a s u bset o f $\mathc al { NP}$ c a lled $\ m a th cal{ NP}_L$, defined i n s u ch a way thatany pr ob l em i n $ \ma thcal{NP}$ i s int he line a ro r b ito f at least on e problem i n $\ mathca l{ NP} _ L$, an d $\m at h cal {NP}_L$ ismini mal. Cert ainlyi t seems reasona ble th atres earc h e ff ort ss hou l dbep rim arily fo cu se d on$\ma t h c a l{NP }_L $ pr oblem s since these ar e th e pr oblem s wit h th emostpotent ial s co pe for practica l us e. Indeed , i fanef ficie n t algo rit hmis deve loped f o r a p r o b le m with a large lin ea r or bit, the n allo fth o se probl em s w ithi n its l inea r o rbit can lever a ge o ff this a lgorit hm as we ll, w i thou t need ing to b e con c erned with suc h explosive gr o wt h as theexa mple givenearl i er.Of c o ur se, a nat uralqu e st i on to ask is whethe r$\math cal{N P}_L$ is fini te. Altern a t i vely, if $\m a th c al{NP}_L$ is n ot fi nite, what proporti on of $\mathc al{NP}$ d o e s it occ upy ? Inthi s ma nuscript, wef o cuson a more mo dest qu est ion , a s a c ase study : if weco ns id er so lelyt he set,$\ mat hc al{ K}_{2 1 }$, of Karp ’s 2 1NP - com plete p r ob l e ms,ho wsmal l a k ernel sub s et$S \sub set \math cal { K}_{ 21 }$ can we identify whi ch possesses t heproper t y that al l 21 problems lie in th e linear or bit o f at least on e p roblem in $S$? T he 21NP-co mp let e probl e m sdes cr ibed by Ka r p ar e asfo llow s: \(1 ) Boolean satisfia b ili ty in conjunc tiv e no r m al fo r m( SAT )\ (2) 0 -1 Integer Prog ramming\ ( 3) Cl ique\ (4)S etPa cking\(5) Ver tex C o ver\ (6 ) Set Cov ering\ (7 )Feed b a ckNode Set\(8) Feed back ArcS et\ ( 9 )Direc ted Hamil to nia n cyc le pro b lem (DHC P)\ (1 0) Undir ected H amiltoni an cycle problem (HCP) \(11) SATwit h at most 3l ite rals perclau se (3-SAT) \ ( 12) Chro mat i c Num ber\ (1 3)C lique Cov e r\ (14) E x ac t C o v er \ (15) Hitt i n g Se t\ (1 6)S teiner Tre e

to_be in_the linear orbit of_Hamiltonian cycle_problem_(HCP), but_HCP_is not known_to be in_the linear orbit of_3-SAT, and indeed,_it_seems unlikely that it is. For completeness, we say that a problem is in_its_own linear_orbit. Then,_consider_a subset of $\mathcal{NP}$ called_$\mathcal{NP}_L$, defined in such a_way that_any problem in $\mathcal{NP}$ is in the linear_orbit_of at least_one problem in $\mathcal{NP}_L$, and $\mathcal{NP}_L$ is minimal. Certainly_it seems reasonable that research efforts_should be primarily_focused_on_$\mathcal{NP}_L$ problems since these_are the problems with the most_potential scope for practical use. Indeed,_if an efficient algorithm is developed for_a problem with a large linear_orbit, then all of those_problems within_its linear orbit can leverage_off this algorithm_as well,_without needing to_be concerned with such explosive growth_as the example_given earlier. Of course, a natural_question_to ask is_whether_$\mathcal{NP}_L$_is finite._Alternatively, if $\mathcal{NP}_L$_is_not finite,_what_proportion of $\mathcal{NP}$ does it occupy? In_this_manuscript, we focus on a more modest_question, as a case_study:_if we consider solely_the set, $\mathcal{K}_{21}$, of Karp’s_21 NP-complete problems, how small a_kernel subset_$S \subset_\mathcal{K}_{21}$ can we identify which possesses the property that all 21_problems lie in the linear orbit_of at least one_problem in_$S$?_The 21 NP-complete_problems_described by_Karp are as follows: \(1) Boolean satisfiability in_conjunctive normal_form (SAT)\ (2) 0-1 Integer Programming\ (3) Clique\ (4)_Set Packing\ (5) Vertex Cover\ (6)_Set_Covering\ (7) Feedback Node Set\ (8) Feedback Arc_Set\ (9) Directed Hamiltonian cycle problem (DHCP)\ (10)_Undirected Hamiltonian cycle problem (HCP) \(11)_SAT_with_at most 3 literals per_clause (3-SAT)\ (12) Chromatic Number\ (13) Clique Cover\ (14)_Exact Cover\ (15) Hitting_Set\ (16) Steiner Tree

describe two types of interactions of two simple paths in $V$. The $P1$ move acts as a transposition on one of the permutations in $T$. While $P$ moves and valid pairs are somewhat adhoc objects, in Lemma \[lembranch\] we will see that the moves $P1-P6$ correspond naturally to moves $Br1-Br6$ on broken surface diagrams, described in Figure \[figbranchmoves\]. Note that each $Br$ move does in fact represent an isotopy in $\bR^4$, since it can be written in terms of Roseman moves, in particular $Ro4$ and $Ro6$ moves. [[ 9999 ]{}]{} ![Moves on valid pairs $(T,V)$ for some fixed broken surface diagram $D$. See Remark \[remmoves\].[]{data-label="figpathmoves"}](pics/pathmoves.eps) ![$Br$ moves for broken surface diagrams, with crossing information suppressed.[]{data-label="figbranchmoves"}](pics/branchmoves.eps) \[lembranch\] a) \[branch1\] If $D \in \pB(\pL_0)$ then all valid pairs $(T,V)$ as in Definition \[defcancel\] are related by the $P$ moves in Figure \[figpathmoves\]. b) \[branch2\] The $Br3$, $Br4$, $Br5$ and $Br6$ moves can be expressed in terms of $Ro1$, $Ro2$, $Ro5^*$, $Ro7$ and $Br1$ moves. c) \[branch3\] If $D \in \pB(\pL_0)$ and $E_1,E_2 \in {\it cancel}(D)$ then $E_1$ and $E_2$ are related by the $Ro1$, $Ro2$, $Ro5^*$, $Ro7$, $Br1$ and $Br2$ moves in Figures \[figrosemove\] and \[figbranchmoves\]. d) \[branch4\] If $D_1,D_2 \in \pB(\pL_0)$, $E_1 \in {\it cancel}(D_1)$, $E_2 \in {\it cancel}(D_2)$ and $D_1$ and $D_2$ are related by

describe two types of interactions of two simple path in $ V$. The $ P1 $ motion acts as a transposition on one of the permutations in $ T$. While $ P$ move and valid pairs are somewhat adhoc object, in Lemma \[lembranch\ ] we will experience that the moves $ P1 - P6 $ correspond naturally to move $ Br1 - Br6 $ on broken surface diagram, described in Figure \[figbranchmoves\ ]. Note that each $ Br$ move does in fact represent an isotopy in $ \bR^4 $, since it can be write in terms of Roseman moves, in particular $ Ro4 $ and $ Ro6 $ moves. [ [ 9999 ] { } ] { } ! [ Moves on valid pair $ (T, V)$ for some fixed broken surface diagram $ D$. See Remark \[remmoves\].[]{data - label="figpathmoves"}](pics / pathmoves.eps) ! [ $ Br$ motion for broken surface diagrams, with crossing information suppressed.[]{data - label="figbranchmoves"}](pics / branchmoves.eps) \[lembranch\ ] a) \[branch1\ ] If $ vitamin d \in \pB(\pL_0)$ then all valid pairs $ (T, V)$ as in Definition \[defcancel\ ] are related by the $ P$ moves in Figure \[figpathmoves\ ]. b) \[branch2\ ] The $ Br3 $, $ Br4 $, $ Br5 $ and $ Br6 $ move can be expressed in terms of $ Ro1 $, $ Ro2 $, $ Ro5^*$, $ Ro7 $ and $ Br1 $ moves. c) \[branch3\ ] If $ D \in \pB(\pL_0)$ and $ E_1,E_2 \in { \it cancel}(D)$ then $ E_1 $ and $ E_2 $ are related by the $ Ro1 $, $ Ro2 $, $ Ro5^*$, $ Ro7 $, $ Br1 $ and $ Br2 $ motion in figure \[figrosemove\ ] and \[figbranchmoves\ ]. d) \[branch4\ ] If $ D_1,D_2 \in \pB(\pL_0)$, $ E_1 \in { \it cancel}(D_1)$, $ E_2 \in { \it cancel}(D_2)$ and $ D_1 $ and $ D_2 $ are related by

dedcribe two types of intevactions of two simple 'aths ih $V$. The $O1$ move acts as a transpositiln on obe of the permutations in $T$. Whipe $P$ movws aid valid pairs acs somewmct adglc oyjxcts, in Lemma \[lgmbranch\] we fill see that dhd loves $P1-P6$ correspond naturally to modes $Br1-Bt6$ ln broken surfwce cyagrzms, described in Figure \[figbranchmkves\]. Noue that each $Br$ mofe does in fact represent wn idotopy in $\bR^4$, since it can be qritewn in terms uf Roseman moves, in patticular $Ro4$ and $Ro6$ moves. [[ 9999 ]{}]{} ![Moves on vclid pairs $(J,R)$ fog some fixed brokvn surface diagram $D$. Vee Remsrk \[remmoves\].[]{daba-labxl="fitpathmoves"}](pics/pathmovxs.eps) ![$Br$ moves for brjken surfdcz diagrams, with crossunt infmrmadion wuporeasxd.[]{dzta-labfl="fmgbranchmovss"}](pics/brancymoves.eps) \[lembranch\] a) \[fgsnch1\] If $D \in \pB(\pL_0)$ ehqn all valid pairs $(T,V)$ as in Definition \[gefdancel\] are related by tye $P$ moves in Figure \[vigpathmodes\]. b) \[branch2\] The $Br3$, $Br4$, $Br5$ and $Br6$ moves can be expsessev kn uevir ov $Ro1$, $Ro2$, $Ro5^*$, $Ro7$ and $Br1$ moves. c) \[branch3\] If $D \in \pF(\lL_0)$ akd $E_1,E_2 \in {\it canccl}(D)$ then $E_1$ and $E_2$ ate rrjated by the $To1$, $Ro2$, $Ro5^*$, $Rk7$, $Br1$ and $Br2$ moves ln Figutes \[fitrosemove\] and \[figbranchmoves\]. d) \[branch4\] If $D_1,D_2 \in \pB(\pL_0)$, $V_1 \in {\it cancel}(D_1)$, $E_2 \in {\ic cancel}(D_2)$ anb $D_1$ anc $D_2$ ate related by

describe two types of interactions of two in The $P1$ acts as a permutations $T$. While $P$ and valid pairs somewhat adhoc objects, in Lemma \[lembranch\] will see that the moves $P1-P6$ correspond naturally to moves $Br1-Br6$ on broken diagrams, described in Figure \[figbranchmoves\]. Note that each $Br$ move does in fact an in since can be written in terms of Roseman moves, in particular $Ro4$ and $Ro6$ moves. [[ 9999 ![Moves on valid pairs $(T,V)$ for some fixed surface diagram $D$. See \[remmoves\].[]{data-label="figpathmoves"}](pics/pathmoves.eps) ![$Br$ moves for broken diagrams, crossing information \[lembranch\] \[branch1\] $D \in \pB(\pL_0)$ all valid pairs $(T,V)$ as in Definition \[defcancel\] are related by the $P$ moves in Figure \[figpathmoves\]. \[branch2\] The $Br5$ and moves be in terms of $Ro5^*$, $Ro7$ and $Br1$ moves. c) \in \pB(\pL_0)$ and $E_1,E_2 \in {\it cancel}(D)$ then and $E_2$ related by the $Ro1$, $Ro2$, $Ro5^*$, $Br1$ and $Br2$ moves in Figures \[figrosemove\] and d) \[branch4\] If $D_1,D_2 \in \pB(\pL_0)$, $E_1 \in {\it cancel}(D_1)$, $E_2 \in {\it cancel}(D_2)$ and $D_2$ are related by

describe two types of interacTions of two SimplE paThs In $v$. The $p1$ movE acts as a transpOSitiOn on one of the permutatioNs in $T$. whILe $P$ mOVeS and vAlid paiRS aRE SomEwHaT adHoC ObJects, In LEmma \[lemBranch\] we wiLl sEe That the moves $p1-p6$ cOrrespond nAtuRally to moves $br1-BR6$ on broKeN suRFace dIagRams, dEscribED in FigUre \[figbraNcHMoves\]. NOTe that eACH $BR$ movE does in fact represENt AN isotopy in $\bR^4$, siNce it cAn BE wRITteN in Terms of RosEmAn movES, in partICuLAR $ro4$ aND $Ro6$ moves. [[ 9999 ]{}]{} ![Moves On valid pairS $(t,V)$ fOr some FiXed BRoken sUrfacE dIAgrAm $D$. See RemarK \[remMoves\].[]{data-Label="fIGpathmoVEs"}](pics/pAthmovEs.ePs) ![$BR$ movES fOr BroKeN SurFAcE diAGraMs, with crOsSiNg infOrmaTION SuppResSed.[]{dAta-laBel="figbranchmOveS"}](picS/BraNchmoVes.epS) \[lemBrAnch\] a) \[Branch1\] if $D \in \PB(\PL_0)$ then all valid pAirs $(t,V)$ as in DefIniTiOn \[dEfCanceL\] Are relAteD by The $P$ movEs in FigURe \[fIgPATHmOves\]. b) \[branch2\] The $Br3$, $Br4$, $br5$ AND $BR6$ moves caN be expREsSeD In terms oF $RO1$, $Ro2$, $ro5^*$, $Ro7$ AND $Br1$ moVes. c) \[BRaNch3\] If $D \in \PB(\pL_0)$ anD $e_1,E_2 \In {\It canceL}(D)$ Then $E_1$ aNd $e_2$ arE reLated BY the $ro1$, $Ro2$, $Ro5^*$, $ro7$, $Br1$ and $BR2$ moveS In Figures \[figroSEmove\] and \[figbrANcHMOvES\]. d) \[brAncH4\] If $D_1,D_2 \in \pB(\pL_0)$, $e_1 \in {\iT CancEl}(D_1)$, $E_2 \IN {\iT caNCel}(D_2)$ aNd $D_1$ anD $D_2$ ARe RElated by

describe two types of int eractionsof tw o s imp le pat hs i n $V$. The $P1 $ mov e acts as a transposit ion o no ne o f t he pe rmutati o ns i n $ T$ .Whi le $P $ mov esand val id pairs a reso mewhat adhoc ob jects, inLem ma \[lembran ch\ ] we w il l s e e tha t t he mo ves $P 1 -P6$ c orrespond n a turall y to mov e s $ Br1- Br6$ on broken su r fa c e diagrams, de scribe di nF i gur e \ [figbranch mo ves\] . Note t h at e a ch$ Br$ move does in fact re p res ent an i sot o py in$\bR^ 4$ , si nce it canbe w ritten in terms of Rose m an move s, inpar tic ular $R o4 $ a nd $Ro 6 $mov e s.[[ 9999]{ }] {} ! [Mov e s o n va lid pai rs $( T,V)$ for som e f ixed bro ken s urfac e di ag ram $ D$. Se e Rem ar k \[remmoves\]. []{d ata-label ="f ig pat hm oves" } ](pics /pa thm oves.ep s) ![$ B r$mo v e s f or broken surfacedi a g ra ms, with cross i ng i n formatio nsup pres s e d.[]{ data - la bel="fig branch m ov es "}](pic s/ branch mo ves .ep s) \ [ lemb ranch\ ] a) \ [bran c h1\] If $D \in \pB(\pL_0)$ t h en a ll vali d p airs $(T,V) $ as in D efin i ti on\ [defc ancel \] ar e related by the $P$ m oves i n Fig ure \[figpath moves\]. b ) \[branc h2\] Th e $Br3$, $Br4$, $Br5 $ and $Br6 $ moves c an be express ed in ter m s of $Ro1 $,$Ro 2$, $R o 5 ^* $, $Ro7$ and$ B r1$mo ves. c ) \[branc h3\ ] I f $ D \ in \pB(\pL_ 0)$ and$E _1 ,E _2 \i n {\i t cancel} (D )$th en$E_1$ and $E _2$ a re r el at e d b y the $ R o1 $ , $Ro 2$ ,$Ro5 ^*$ ,$Ro7$ , $B r 1$and $Br 2$ movesinF igur es \ [figros emove\] and \ [f igbranchmo ve s\] . d)\ [branch4 \] If $D_1,D_2 \in \pB( \ pL_0)$, $E _1 \i n {\ it cancel }(D _1)$,$E_ 2 \in { \it ca ncel} (D _2) $ and $ D _ 1$ an d$D_2$ arer e lat ed by

describe_two types_of interactions of two_simple paths_in_$V$. The_$P1$_move acts as_a transposition on_one of the permutations_in $T$. While_$P$_moves and valid pairs are somewhat adhoc objects, in Lemma \[lembranch\] we will see_that_the moves_$P1-P6$_correspond_naturally to moves $Br1-Br6$ on_broken surface diagrams, described in_Figure \[figbranchmoves\]._Note that each $Br$ move does in fact_represent_an isotopy in_$\bR^4$, since it can be written in terms of_Roseman moves, in particular $Ro4$ and_$Ro6$ moves. [[_9999_]{}]{} ![Moves_on valid pairs $(T,V)$_for some fixed broken surface diagram_$D$. See Remark \[remmoves\].[]{data-label="figpathmoves"}](pics/pathmoves.eps) ![$Br$ moves for_broken surface diagrams, with crossing information suppressed.[]{data-label="figbranchmoves"}](pics/branchmoves.eps) \[lembranch\] a)_ \[branch1\] If $D \in \pB(\pL_0)$_then all valid pairs $(T,V)$_as in_Definition \[defcancel\] are related by_the $P$ moves_in Figure_\[figpathmoves\]. b) \[branch2\]_The $Br3$, $Br4$, $Br5$ and $Br6$_moves can be_expressed in terms of $Ro1$, $Ro2$,_$Ro5^*$,_$Ro7$ and $Br1$_moves. c)__\[branch3\] If_$D \in \pB(\pL_0)$_and_$E_1,E_2 \in_{\it_cancel}(D)$ then $E_1$ and $E_2$ are_related_by the $Ro1$, $Ro2$, $Ro5^*$, $Ro7$, $Br1$_and $Br2$ moves in_Figures_\[figrosemove\] and \[figbranchmoves\]. d) _\[branch4\] If $D_1,D_2 \in \pB(\pL_0)$,_$E_1 \in {\it cancel}(D_1)$, $E_2 \in_{\it cancel}(D_2)$_and $D_1$_and $D_2$ are related by

v\cos(\theta))-v\cos(\theta)}=\frac{nL}{c}+\frac{v\cos(\theta)L}{c^2}+.. \label{eqn:traveltime2}$$ on using (\[eqn:Fresnel\]) and expanding to 1st oder in $v/c$. If we ignore the Fresnel drag term in (\[eqn:Fresnel\]) we obtain, instead, $$t_{AB}\approx \frac{L}{c/n-v\cos(\theta)}=\frac{nL}{c}+\frac{n^2v\cos(\theta)L}{c^2}+.. \label{eqn:traveltime3}$$ The 1st important observation is that the $v/c$ component in (\[eqn:traveltime2\]) is independent of the dielectric refractive index $n$. This is explained in the next section. If the clocks were synchronised $t_{AB}$ would be known, and by changing direction of the light path, that is varying $\theta$, the magnitude of the 2nd term may be separated from the magnitude of the 1st term. If the clocks are not synchronised, then the measured travel time ${\overline t}_{AB}=(t_B+\tau)-t_A=t_{AB}+\tau$, where $\tau$ is the unknown, but fixed, offset between the two clocks. But this does not change the above argument. The magnitude of $v$ and the direction of $v$ can be determined by varying $\theta$. For a small detector the change in $\theta$ can be achieved by a direct rotation. But for a large detector, such as De Witte’s [@DeWitte] 1.5km RF coaxial cable experiment, the rotation was achieved by that of the earth. The reason for using opposing propagation directions, as in Fig.\[fig:oneway\], and then measuring travel time differences, is that local temperature effects cancel. This is because a common temperature change in the two adjacent cables changes the speed to the same extent, whereas absolute motion effects cause opposite signed speed changes. Then the temperature effects cancel on measuring differences in the travel times, whereas absolute motion effects are additive. Finally, after the absolute motion velocity has been determined, the two spatially separated clocks may be synchronised. That the $v/c$ term in $t_{AB}$ in (\[eqn:traveltime2\]) is independent of $n$ means that various techniques to do a 1st order

v\cos(\theta))-v\cos(\theta)}=\frac{nL}{c}+\frac{v\cos(\theta)L}{c^2}+.. \label{eqn: traveltime2}$$ on using (\[eqn: Fresnel\ ]) and expanding to 1st oder in $ v / c$. If we dismiss the Fresnel puff term in (\[eqn: Fresnel\ ]) we obtain, alternatively, $ $ t_{AB}\approx \frac{L}{c / n - v\cos(\theta)}=\frac{nL}{c}+\frac{n^2v\cos(\theta)L}{c^2}+.. \label{eqn: traveltime3}$$ The first important observation is that the $ v / c$ part in (\[eqn: traveltime2\ ]) is autonomous of the dielectric refractive index $ n$. This is explained in the next section. If the clock were synchronised $ t_{AB}$ would be known, and by change direction of the light path, that is varying $ \theta$, the magnitude of the 2nd condition may be separated from the magnitude of the 1st terminus. If the clocks are not synchronised, then the measured travel clock time $ { \overline t}_{AB}=(t_B+\tau)-t_A = t_{AB}+\tau$, where $ \tau$ is the unknown, but fixed, offset between the two clocks. But this does not change the above controversy. The magnitude of $ v$ and the direction of $ v$ can be determined by varying $ \theta$. For a small detector the change in $ \theta$ can be achieved by a direct rotation. But for a large detector, such as De Witte ’s [ @DeWitte ] 1.5 km RF coaxial cable experiment, the rotation was achieved by that of the earth. The reason for using opposing propagation focus, as in Fig.\[fig: oneway\ ], and then measuring travel prison term dispute, is that local temperature effects cancel. This is because a common temperature variety in the two adjacent cables exchange the speed to the same extent, whereas absolute apparent motion effects cause opposite signed speed changes. Then the temperature effect cancel on measuring differences in the travel meter, whereas absolute gesture effects are additive. Finally, after the absolute motion velocity has been settle, the two spatially separated clocks may be synchronised. That the $ five / c$ term in $ t_{AB}$ in (\[eqn: traveltime2\ ]) is independent of $ n$ means that various proficiency to do a first order

v\cod(\theta))-v\cos(\theta)}=\frac{nL}{c}+\frxc{v\cos(\theta)L}{c^2}+.. \laywl{eqn:tcaveltije2}$$ on uskng (\[eqn:Fresnel\]) and expanding ti 1st ider in $v/c$. If we ignord the Frednel drat tecm in (\[eqn:Fresnel\]) we obtalu, insfcad, $$t_{CB}\epprox \frac{L}{c/n-v\gos(\theta)}=\fraw{nL}{c}+\frac{n^2v\cos(\tvega)P}{c^2}+.. \label{eqn:traveltime3}$$ The 1st importagt obsetvwtion is that jhe $v/b$ sompknent in (\[eqn:traveltime2\]) is independsnt of uhe dielectric reftactive index $n$. This is exolaijed in the next seftion. If thg cljxks were synzhronised $u_{AY}$ would be inown, and by changing direction of tke light pajk, thwj is varying $\thetw$, the magnitmce of dhe 2nd yerm may be seiaratxd feom the magnitude of vhe 1st term. If the cjocks are nkt synchronised, tyeb the meavurea trxvem vims ${\overpinx t}_{AB}=(t_B+\tau)-t_Z=t_{AB}+\tau$, wheee $\tau$ is the unknoen, vut fixed, offaet beewqen the two clocks. But this does not chdngs the above argument. Thw magnitude of $v$ and jhe directyon of $v$ can be determined by varying $\theta$. For a smalm deuegbor gye change in $\theta$ can be achieved by a direct wktstpon. But for a larne detector, such ax Fe Ritte’s [@DeWittg] 1.5km RF cozxial cable experilent, thg rotarion was wchirved by that of the earth. Tye reason fog usung opposing propayation direccions, ss in Fig.\[fig:oneway\], and then oeashring travep time dirwerences, is that lobal demperature effects cancel. This is uecauxe a coomon tempewature chajge ik the two adjacent fabled whanges thf speed to the same extent, wherxes absolute mpthon effects causc opposite signqd speed changgs. Then tke temoerature erfects rancel on mewsuring diffesgnces in the vravel tiies, qherwas absuuute motion efgects are additive. Dinally, after the ebsouhte motion veloenti has been detetmived, tje trm spatially vepafatdc clozks may be whnchtonised. That the $v/c$ tarm jn $t_{AB}$ in (\[eqn:travektlme2\]) is ineependene of $n$ means yhat various technlques to do a 1st order

v\cos(\theta))-v\cos(\theta)}=\frac{nL}{c}+\frac{v\cos(\theta)L}{c^2}+.. \label{eqn:traveltime2}$$ on using (\[eqn:Fresnel\]) and expanding oder $v/c$. If ignore the Fresnel obtain, $$t_{AB}\approx \frac{L}{c/n-v\cos(\theta)}=\frac{nL}{c}+\frac{n^2v\cos(\theta)L}{c^2}+.. \label{eqn:traveltime3}$$ 1st important observation that the $v/c$ component in (\[eqn:traveltime2\]) independent of the dielectric refractive index $n$. This is explained in the next If the clocks were synchronised $t_{AB}$ would be known, and by changing direction the path, is $\theta$, the magnitude of the 2nd term may be separated from the magnitude of the 1st If the clocks are not synchronised, then the travel time ${\overline t}_{AB}=(t_B+\tau)-t_A=t_{AB}+\tau$, $\tau$ is the unknown, but offset the two But does change the above The magnitude of $v$ and the direction of $v$ can be determined by varying $\theta$. For a detector the $\theta$ can achieved a rotation. But for detector, such as De Witte’s [@DeWitte] cable experiment, the rotation was achieved by that the earth. reason for using opposing propagation directions, in Fig.\[fig:oneway\], and then measuring travel time differences, that local temperature effects cancel. This is because a common temperature change in the two changes the speed to same extent, whereas motion cause signed changes. Then temperature effects cancel on measuring differences in the travel times, whereas motion effects are additive. Finally, after the absolute motion velocity determined, two spatially separated may be synchronised. That $v/c$ in $t_{AB}$ in (\[eqn:traveltime2\]) of means to a order

v\cos(\theta))-v\cos(\theta)}=\frac{nL}{c}+\Frac{v\cos(\thEta)L}{c^2}+.. \LabEl{eQn:TravEltiMe2}$$ on using (\[eqn:FrESnel\]) And expanding to 1st oder in $V/c$. If wE iGNore THe fresnEl drag tERm IN (\[Eqn:frEsNel\]) We OBtAin, inSteAd, $$t_{AB}\apProx \frac{L}{c/N-v\cOs(\Theta)}=\frac{nL}{c}+\FRaC{n^2v\cos(\thetA)L}{c^2}+.. \Label{eqn:travEltIme3}$$ The 1St ImpORtant ObsErvatIon is tHAt the $v/C$ componenT iN (\[Eqn:traVEltime2\]) iS INdEpenDent of the dielectrIC rEFractive index $n$. this is ExPLaINEd iN thE next sectiOn. if the CLocks weRE sYNCHroNIsed $t_{AB}$ would bE known, and by CHanGing diReCtiON of the Light PaTH, thAt is varying $\ThetA$, the magniTude of THe 2nd terM May be seParateD frOm tHe maGNiTuDe oF tHE 1st TErM. If THe cLocks are NoT sYnchrOnisED, THEn thE meAsurEd traVel time ${\overliNe t}_{aB}=(t_B+\TAu)-t_a=t_{AB}+\tAu$, wheRe $\taU$ iS the uNknown, But fiXeD, offset between tHe twO clocks. BuT thIs DoeS nOt chaNGe the aBovE arGument. THe magniTUde Of $V$ AND tHe direction of $v$ can bE dETErMined by vArying $\THeTa$. fOr a small DeTecTor tHE ChangE in $\tHEtA$ can be acHieved BY a DiRect rotAtIon. But FoR a lArgE deteCTor, sUch as DE Witte’s [@DEWittE] 1.5Km RF coaxial cabLE experiment, thE RoTATiON was AchIeved by that Of thE EartH. The REaSon FOr usiNg oppOsINg PRopagation directionS, aS in Fig.\[Fig:onEway\], and then meAsuring traVEL Time diffErenCEs, IS that local tempEratuRe effects cANcel. This Is becAuse a comMon temperATUre changE in The Two AdjACEnT cables changeS THe spEeD to the sAme Extent, wHerEas AbsOluTe Motion effEcts causE oPpOsItE siGned sPEed changEs. theN tHe tEmperATure efFects CancEl On MEasUring diFFeRENces In ThE traVel TiMes, whEreaS AbsOlute moTion effecTs aRE addItIvE. FinallY, after the absoLuTe motion veLoCitY has beEN DeterminEd, the two spatially separaTEd clockS maY be syNchrOnised. ThaT thE $v/c$ terM in $T_{aB}$ in (\[eqN:traveLtime2\]) Is IndEPEndenT OF $n$ MeaNs That variouS TEchNiqueS tO do a 1St order

v\cos(\theta))-v\cos(\thet a)}=\frac{ nL}{c }+\ fra c{ v\co s(\t heta)L}{c^2}+. . \la bel{eqn:traveltime2}$$ on u si n g (\ [ eq n:Fre snel\]) an d exp an di ngto 1s t ode r i n $v/c$ . If we ig nor ethe Fresneld ra g term in(\[ eqn:Fresnel\ ])we obt ai n,i nstea d,$$t_{ AB}\ap p rox \f rac{L}{c/ n- v \cos(\ t heta)}= \ f ra c{nL }{c}+\frac{n^2v\c o s( \ theta)L}{c^2}+ .. \la be l {e q n :tr ave ltime3}$$Th e 1st importa n to b s erv a tion is thatthe $v/c$ c o mpo nent i n(\[ e qn:tra velti me 2 \]) is indepen dent of the d ielect r ic refr a ctive i ndex $ n$. Th is i s e xp lai ne d in th e n e xtsection. I fthe c lock s w e re s ync hron ised$t_{AB}$ woul d b e kn o wn, andby ch angi ng dire ctionof th elight path, tha t is varying$\t he ta$ ,the m a gnitud e o f t he 2ndterm ma y be s e p a ra ted from the magni tu d e o f the 1s t term . I ft he clock sare not s ynchr onis e d, then th e meas u re dtravelti me ${\ ov erl ine t}_{ A B}=( t_B+\t au)-t_A= t_{AB } +\tau$, where$ \tau$ is theu nk n o wn , but fi xed, offset bet w eenthet wo cl o cks.But t hi s d o es not change the a bo ve arg ument . The magnitu de of $v$a n d the dir ecti o no f $v$ can be d eterm ined by va r ying $\t heta$ . For asmall det e c tor thecha nge in $\ t h et a$ can be ach i e vedby a dire ctrotatio n.But fo r a l arge dete ctor, su ch a sDe Wi tte’s [@DeWitt e] 1. 5k m R F coa x ial ca ble e xper im en t , t he rota t io n wasac hi eved by t hat o f th e ea rth. Th e reasonfor usin gop posingpropagation d ir ections, a sinFig.\[ f i g:oneway \], and then measuringt ravel t ime diff eren ces, is t hat local te m peratu re eff ectsca nce l . This i sbec au se a commo n tem perat ur e ch ange in the two adjacentc abl es changes th e s peed t othe sa m e e xt e nt, w hereas absolute motion ef fe c ts cause opp o sit esignedspeed c hange s . Thenthe tempe rature ef fe ctsc a nce l on measu ring dif ferencesi n the tr aveltim es, wh er eas abso lute m o tio n eff ects a re addit ive.Fi nally, a fter the absolute motio n velo cityhas been det erm i ned , the two spa tially sep ara ted cloc ksm ay be syn c hr oni s ed. That the $v/c$ te rmi n $ t_{AB}$ in( \ [ eqn :trav elt i me2\]) isindependent of $n $ means that va riou s tec hni q uesto do a 1st orde r

v\cos(\theta))-v\cos(\theta)}=\frac{nL}{c}+\frac{v\cos(\theta)L}{c^2}+.. \label{eqn:traveltime2}$$ on_using (\[eqn:Fresnel\])_and expanding to 1st_oder in_$v/c$._If we_ignore_the Fresnel drag_term in (\[eqn:Fresnel\])_we obtain, instead, $$t_{AB}\approx_\frac{L}{c/n-v\cos(\theta)}=\frac{nL}{c}+\frac{n^2v\cos(\theta)L}{c^2}+.. \label{eqn:traveltime3}$$ The 1st_important_observation is that the $v/c$ component in (\[eqn:traveltime2\]) is independent of the dielectric refractive_index_$n$. This_is_explained_in the next section. If_the clocks were synchronised $t_{AB}$_would be_known, and by changing direction of the light_path,_that is varying_$\theta$, the magnitude of the 2nd term may be_separated from the magnitude of the_1st term. If_the_clocks_are not synchronised, then_the measured travel time ${\overline t}_{AB}=(t_B+\tau)-t_A=t_{AB}+\tau$,_where $\tau$ is the unknown, but_fixed, offset between the two clocks. But_this does not change the above_argument. The magnitude of $v$_and the_direction of $v$ can be_determined by varying_$\theta$. For_a small detector_the change in $\theta$ can be_achieved by a_direct rotation. But for a large_detector,_such as De_Witte’s_[@DeWitte]_1.5km RF_coaxial cable experiment,_the_rotation was_achieved_by that of the earth. The_reason_for using opposing propagation directions, as in_Fig.\[fig:oneway\], and then measuring_travel_time differences, is that_local temperature effects cancel. This_is because a common temperature change_in the_two adjacent_cables changes the speed to the same extent, whereas absolute motion_effects cause opposite signed speed changes._Then the temperature effects_cancel on_measuring_differences in the_travel_times, whereas_absolute motion effects are additive. Finally, after_the absolute_motion velocity has been determined, the_two spatially separated clocks_may_be synchronised. That the $v/c$ term in_$t_{AB}$ in (\[eqn:traveltime2\]) is independent of_$n$ means that various techniques_to_do_a 1st order

the Kitaev case has been steadily increasing over the past 20 years encompassing Mott-Hubbard organic materials and deep Mott insulators in various two and three dimensional geometries [@Lacroix-Springer; @Balents; @Mendels-JPSJ; @Mendels-academy; @Zhou-KanodaRMP]. Among them, the kagome lattice is a prominent 2D example where frustrated triangles only share corners. This reduced lattice connectivity, the frustration generated by nearest neighbor Heisenberg antiferromagnetic interactions, and the quantum character of $S=1/2$ spins indeed conspire to stabilize such a highly entangled QSL state in herbertsmithite [@Shores], ZnCu$_3$(OH)$_6$Cl$_2$ (fig. 1). It emerged as the first kagome based antiferromagnet featuring a perfect equilateral triangular geometry, dominant nearest neighbor Heisenberg antiferromagnetic interactions, $J\sim180-190$ K, and not showing any ordering [@Mendels-musr] but rather an unconventional excitations continuum [@Han-neutrons]. It is recognized as the kagome spin liquid candidate approaching so far the best the ideal model of an Heisenberg AntiFerromagnetic Hamiltonian on the Kagome lattice (KHAF) [@Norman], now available in single crystal form. Its discovery triggered a burst of experimental activity, on one hand with the materials search for Cu$^{2+}$-based $S=1/2$ variants obtained either from a change of anions [@Barlowite; @BarlowiteNMR; @Brochantite] or replacement of Zn$^{2+}$ cations [@PuphalGa; @PuphalY; @ChineseY; @BartelemyY] and Li$^+$ intercalation in view of charge doping [@McQueen], and on the other hand with V$^{4+}$-based $S=1/2$ [@VOF; @Orain] QSL candidates. On the theory side which was also revived by the discovery of herbertsmithite, the issue of the ground state of the KHAF, despite its apparently simple form, has been lacking for long a definite conclusion. The main reason lies in the existence of a proliferation of states close-by in energy which have led to theoretical proposals spanning from valence bond crystals [@Singh] - made of local spin dimers- to spin liquids,

the Kitaev case has been steadily increasing over the past 20 years encompassing Mott - Hubbard constituent fabric and deep Mott insulators in diverse two and three dimensional geometry   [ @Lacroix - Springer; @Balents; @Mendels - JPSJ; @Mendels - academy; @Zhou - KanodaRMP ]. Among them, the kagome lattice is a prominent 2D example where defeated triangles entirely share corner. This reduced lattice connectivity, the frustration beget by nearest neighbor Heisenberg antiferromagnetic interactions, and the quantum quality of $ S=1/2 $ tailspin indeed conspire to stabilize such a highly embroiled QSL state in herbertsmithite   [ @Shores ], ZnCu$_3$(OH)$_6$Cl$_2 $ (figure.   1). It emerge as the first kagome based antiferromagnet featuring a arrant equilateral triangular geometry, dominant nearest neighbor Heisenberg antiferromagnetic interaction, $ J\sim180 - 190 $   K, and not showing any order   [ @Mendels - musr ] but rather an improper excitations continuum   [ @Han - neutrons ]. It is recognized as the kagome spin fluent candidate approaching so far the best the ideal model of an Heisenberg AntiFerromagnetic Hamiltonian on the Kagome wicket (KHAF)   [ @Norman ], nowadays available in single crystal form. Its discovery triggered a burst of experimental activity, on one hand with the materials search for Cu$^{2+}$-based $ S=1/2 $ variants obtain either from a change of anions   [ @Barlowite; @BarlowiteNMR; @Brochantite ] or substitute of Zn$^{2+}$ cation   [ @PuphalGa; @PuphalY; @ChineseY; @BartelemyY ] and Li$^+$ intercalation in view of charge doping   [ @McQueen ], and on the other hand with V$^{4+}$-based $ S=1/2 $   [ @VOF; @Orain ] QSL candidates. On the theory slope which was also revived by the discovery of herbertsmithite, the issue of the ground state of the KHAF, despite its apparently bare form, has been lacking for long a definite conclusion. The main rationality lies in the being of a proliferation of states close - by in energy which have led to theoretical proposals cross from valence bond crystals   [ @Singh ] - made of local spin dimers- to spin liquids,

thf Kitaev case has been sueadily increasiny over vhe pasf 20 years encompassing Mott-Hubbard orjanix mattgials and deep Mott ivsulators in variius uwo and three dimxhsional geomefvies [@Lcccoix-Springer; @Bakents; @Mendals-JPSJ; @Mendelv-azabemy; @Zhou-KanodaRMP]. Among them, the kadome laytlce is a promigent 2Q exzmple where frustrated triangles ohly shage corners. This rrduced lattice connectivitj, thf frustration genegated by neqrese neighbor Heksenberg antiferromagngtic interactions, and the quantuo chaxacter of $S=1/2$ spind indeed conwpire to stabilizc such d highlu entangled QSK svate in herbertsmithite [@Shmres], ZnCu$_3$(OH)$_6$Cl$_2$ (fig. 1). Yt emergeg cs the first kagome bqswd anjifersomaebet feztnrihg a pfrfxct equilatsral triangylar geometry, dominsne nearest neiggbor Hqifenberg antiferromagnetic interactions, $B\sij180-190$ K, and not showing any irdering [@Mendels-musr] bot rather wn unconventional excitations continuum [@Han-neutronv]. It ms reeignizde ws the kagome spin liquid candidate approachigf xo far the best bhe ideal model of aj Ngisenberg AntiWerromcfnstic Hamiltonian oj the Kwgome lattice (HHAF) [@Morman], now available in sintle crystal yorn. Its discovery trnggered a buxst of expetimental activity, on onz hand with the mwterials adarch for Cu$^{2+}$-basea $S=1/2$ vdriants obtained either frjm a chanje of anions [@Carlpwite; @FarlowiteNLR; @Brochantite] or replacfment ox Zn$^{2+}$ catiojs [@PuphalGa; @PuphalY; @ChineseY; @BarvxlemyY] and Li$^+$ ittegcalation in vlew of charge djping [@McQueen], aud on thz othef hand witg V$^{4+}$-basev $S=1/2$ [@VOF; @Orain] QSL candidatad. On the theocy side wrich was also rdxived by the doscovery of herbertwmithite, the issue of fhe ground statz if the KHAF, deslitd ies a'paregdly simple fmrm, fas neen uacking for uong a definite conclusimn. Tge main reason liex ln the exustence jf a proliferstion of states cllse-by in enerby rhich have led to theoretical lroposals spwnning from dalekce fond crystcls [@Singh] - made of local spin dimers- to s'in liquids,

the Kitaev case has been steadily increasing past years encompassing organic materials and two three dimensional geometries @Balents; @Mendels-JPSJ; @Mendels-academy; Among them, the kagome lattice is prominent 2D example where frustrated triangles only share corners. This reduced lattice connectivity, frustration generated by nearest neighbor Heisenberg antiferromagnetic interactions, and the quantum character of spins conspire stabilize a highly entangled QSL state in herbertsmithite [@Shores], ZnCu$_3$(OH)$_6$Cl$_2$ (fig. 1). It emerged as the first based antiferromagnet featuring a perfect equilateral triangular geometry, nearest neighbor Heisenberg antiferromagnetic $J\sim180-190$ K, and not showing ordering but rather unconventional continuum It is recognized the kagome spin liquid candidate approaching so far the best the ideal model of an Heisenberg AntiFerromagnetic on the (KHAF) [@Norman], available single form. Its discovery burst of experimental activity, on one materials search for Cu$^{2+}$-based $S=1/2$ variants obtained either a change anions [@Barlowite; @BarlowiteNMR; @Brochantite] or replacement Zn$^{2+}$ cations [@PuphalGa; @PuphalY; @ChineseY; @BartelemyY] and Li$^+$ in view of charge doping [@McQueen], and on the other hand with V$^{4+}$-based $S=1/2$ [@VOF; candidates. On the theory which was also by discovery herbertsmithite, issue of ground state of the KHAF, despite its apparently simple form, has lacking for long a definite conclusion. The main reason lies existence a proliferation of close-by in energy which led theoretical proposals spanning from crystals - spin to liquids,

the Kitaev case has been steadIly increasIng ovEr tHe pAsT 20 yeaRs enCompassing Mott-hUbbaRd organic materials and dEep MoTt INsulAToRs in vArious tWO aND ThrEe DiMenSiONaL geomEtrIes [@LacrOix-SpringeR; @BaLeNts; @Mendels-JPsj; @MEndels-acadEmy; @zhou-KanodaRMp]. AmOng theM, tHe kAGome lAttIce is A promiNEnt 2D exAmple wherE fRUstratED triangLES oNly sHare corners. This reDUcED lattice connecTivity, ThE FrUSTraTioN generated By NeareST neighbOR HEISEnbERg antiferromaGnetic interACtiOns, and ThE quANtum chAractEr OF $S=1/2$ sPins indeed cOnspIre to stabIlize sUCh a highLY entangLed QSL StaTe iN herBErTsMitHiTE [@ShOReS], ZncU$_3$(OH)$_6$cl$_2$ (fig. 1). It eMeRgEd as tHe fiRST KAgomE baSed aNtifeRromagnet featUriNg a pERfeCt equIlateRal tRiAngulAr geomEtry, dOmInant nearest neiGhboR HeisenbeRg aNtIfeRrOmagnETic intEraCtiOns, $J\sim180-190$ k, and not SHowInG ANY oRdering [@Mendels-musr] BuT RAtHer an uncOnventIOnAl EXcitatioNs ConTinuUM [@han-neUtroNS]. IT is recogNized aS ThE kAgome spIn Liquid CaNdiDatE apprOAchiNg so faR the best The idEAl model of an HeiSEnberg AntiFerROmAGNeTIc HaMilTonian on the kagoME latTice (khAf) [@NoRMan], noW avaiLaBLe IN single crystal form. ITs DiscovEry trIggered a burst Of experimeNTAL activitY, on oNE hANd with the materIals sEarch for Cu$^{2+}$-BAsed $S=1/2$ varIants Obtained Either froM A Change of AniOns [@barLowITE; @BArlowiteNMR; @BrOCHantItE] or replAceMent of ZN$^{2+}$ caTioNs [@PUphAlga; @PuphalY; @chineseY; @baRtElEmYY] aNd Li$^+$ iNTercalatIoN in ViEw oF charGE dopinG [@McQuEen], aNd On THe oTher hanD WiTH v$^{4+}$-basEd $s=1/2$ [@VoF; @OrAin] qSl candIdatES. On The theoRy side whiCh wAS alsO rEvIved by tHe discovery of HeRbertsmithItE, thE issue OF The grounD state of the KHAF, despite iTS appareNtlY simpLe foRm, has been LacKing foR loNG a defiNite coNclusIoN. ThE MAin reASOn LieS iN the existeNCE of A prolIfEratIon of stAtes close-by in energY WhiCh have led to thEorEticAL PrOpoSAlS SpaNnINg fROM valence bond cryStals [@Singh] - MaDE oF local spin DImeRs- To spin lIquids,

the Kitaev case has beensteadily i ncrea sin g o ve r th e pa st 20 years en c ompa ssing Mott-Hubbard org anicma t eria l sand d eep Mot t i n s ula to rs in v a ri ous t woand thr ee dimensi ona lgeometries [ @ La croix-Spri nge r; @Balents; @M endels -J PSJ ; @Men del s-aca demy;@ Zhou-K anodaRMP] .A mong t h em, the k ag omelattice is a prom i ne n t 2D example w here f ru s tr a t edtri angles onl yshare corners . T h i s re d uced latticeconnectivit y , t he fru st rat i on gen erate db y n earest neig hbor Heisenbe rg ant i ferroma g netic i nterac tio ns, and th equa nt u m c h ar act e r o f $S=1/2 $sp ins i ndee d c o nspi reto s tabil ize such a hi ghl y en t ang led Q SL st atein herb ertsmi thite  [ @Shores], ZnCu$ _3$( OH)$_6$Cl $_2 $(fi g.  1).I t emer ged as the fi rst kag o meba s e d a ntiferromagnet fea tu r i ng a perfe ct equ i la te r al trian gu lar geo m e try,domi n an t neares t neig h bo rHeisenb er g anti fe rro mag netic inte ractio ns, $J\s im180 - 190$ K, and no t showing anyo rd e r in g  [@M end els-musr] b ut r a ther anu nc onv e ntion al ex ci t at i ons continuum [@Han -n eutron s]. I t is recogniz ed as thek a g ome spin liq u id candidate appr oachi ng so fart he bestthe i deal mod el of anH e isenberg An tiF err oma g n et ic Hamiltonia n on t he Kagome la ttice ( KHA F)[@N orm an ], now av ailablein s in gl e c rysta l form. I ts di sc ove ry tr i ggered a bu rstof e x per imental ac t i vity ,on one ha nd with the mat erialssearch fo r C u $^{2 +} $- based $ S=1/2$ varian ts obtainedei the r from a changeof anions [@Barlowite;@ Barlowi teN MR; @ Broc hantite]orreplac eme n t of Z n$^{2+ }$ ca ti ons [ @Puph a l Ga ; @ Pu phalY; @Ch i n ese Y; @B ar tele myY] an d Li$^+$ intercala t ion in view of c har ge d o p in g [ @ Mc Q uee n] , an d on the other ha nd with V$ ^{ 4 +} $-based $S = 1/2 $[@VOF;@Orain] QSLc andidat es. On t he theory s idew h ich was alsorevivedby the di s cover y o f her ber tsmith it e,the i ssue o f th e gro und st at e of t he KH AF , despit e its apparently simple form, hasbee n lacking fo r lo ng a defi nite conclusio n.The main re a son l iesi nthe exist ence of a prol i fe rat i o nof states c l o s e-b y inene r gy whi ch h ave led to theore t ical proposals spa n n ing fr o m va le nce bond cryst als  [ @ S ingh] -ma de of local spin di me r s- to spinliquid s,

the_Kitaev case_has been steadily increasing_over the_past_20 years_encompassing_Mott-Hubbard organic materials_and deep Mott_insulators in various two_and three dimensional_geometries [@Lacroix-Springer;_@Balents; @Mendels-JPSJ; @Mendels-academy; @Zhou-KanodaRMP]. Among them, the kagome lattice is a prominent 2D example_where_frustrated triangles_only_share_corners. This reduced lattice connectivity,_the frustration generated by nearest_neighbor Heisenberg_antiferromagnetic interactions, and the quantum character of $S=1/2$_spins_indeed conspire to_stabilize such a highly entangled QSL state in herbertsmithite [@Shores],_ZnCu$_3$(OH)$_6$Cl$_2$ (fig. 1). It emerged as the_first kagome based_antiferromagnet_featuring_a perfect equilateral triangular_geometry, dominant nearest neighbor Heisenberg antiferromagnetic_interactions, $J\sim180-190$ K, and not showing any_ordering [@Mendels-musr] but rather an unconventional excitations continuum [@Han-neutrons]._It is recognized as the kagome_spin liquid candidate approaching so_far the_best the ideal model of_an Heisenberg AntiFerromagnetic_Hamiltonian on_the Kagome lattice_(KHAF) [@Norman], now available in single crystal_form. Its discovery_triggered a burst of experimental activity,_on_one hand with_the_materials_search for_Cu$^{2+}$-based $S=1/2$ variants_obtained_either from_a_change of anions [@Barlowite; @BarlowiteNMR; @Brochantite] or_replacement_of Zn$^{2+}$ cations [@PuphalGa; @PuphalY; @ChineseY; @BartelemyY] and_Li$^+$ intercalation in view_of_charge doping [@McQueen], and on_the other hand with V$^{4+}$-based_$S=1/2$ [@VOF; @Orain] QSL candidates. On the theory_side which_was also_revived by the discovery of herbertsmithite, the issue of the ground_state of the KHAF, despite its_apparently simple form, has_been lacking_for_long a definite_conclusion._The main_reason lies in the existence of a_proliferation of_states close-by in energy which have_led to theoretical proposals_spanning_from valence bond crystals [@Singh] - made_of local spin dimers- to spin_liquids,

0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\overline{\mathcal{U}}_{0,+}^{+}$, $\overline{\mathcal{U}}_{T/2,-}^{+}$ 1 BDI $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal{U}}_{0,+-}^{+}$, $\hat{\mathcal{U}}_{T/2,++}^{+}$ $\hat{\mathcal{U}}_{0,--}^{+}$,$\hat{\mathcal{U}}_{T/2,-+}^{+}$ $\hat{\mathcal{U}}_{0,-+}^{+}$, $\hat{\mathcal{U}}_{T/2,--}^{+}$ 2 D $\hat{\mathcal{U}}_{0,+}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\overline{\mathcal{U}}_{s,+}^{+}$ $\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,+}^{+}$ $\overline{\mathcal{U}}_{s,-}^{+}$ 3 DIII $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal{U}}_{0,-+}^{+}$, $\hat{\mathcal{U}}_{T/2,--}^{+}$ $\hat{\mathcal{U}}_{0,--}^{+}$, $\hat{\mathcal{U}}_{0,-+}^{+}$ $\hat{\mathcal{U}}_{0,+-}^{+}$, $\hat{\mathcal{U}}_{T/2,++}^{+}$ 4 AII $\hat{\mathcal{U}}_{0,+}^{+}$, $\hat{\mathcal{U}}_{T/2,+}^{+}$ $\overline{\mathcal{U}}_{0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\overline{\mathcal{U}}_{0,+}^{+}$, $\overline{\mathcal{U}}_{T/2,-}^{+}$ 5 CII $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal

0,-}^{+}$, $ \overline{\mathcal{U}}_{T/2,+}^{+}$ $ \hat{\mathcal{U}}_{0,-}^{+}$, $ \hat{\mathcal{U}}_{T/2,-}^{+}$ $ \overline{\mathcal{U}}_{0,+}^{+}$, $ \overline{\mathcal{U}}_{T/2,-}^{+}$ 1 BDI $ \hat{\mathcal{U}}_{0,++}^{+}$, $ \hat{\mathcal{U}}_{T/2,+-}^{+}$ $ \hat{\mathcal{U}}_{0,+-}^{+}$, $ \hat{\mathcal{U}}_{T/2,++}^{+}$ $ \hat{\mathcal{U}}_{0,--}^{+}$,$\hat{\mathcal{U}}_{T/2,-+}^{+}$ $ \hat{\mathcal{U}}_{0,-+}^{+}$, $ \hat{\mathcal{U}}_{T/2,--}^{+}$ 2 D $ \hat{\mathcal{U}}_{0,+}^{+}$, $ \hat{\mathcal{U}}_{T/2,-}^{+}$ $ \overline{\mathcal{U}}_{s,+}^{+}$ $ \hat{\mathcal{U}}_{0,-}^{+}$, $ \hat{\mathcal{U}}_{T/2,+}^{+}$ $ \overline{\mathcal{U}}_{s,-}^{+}$ 3 DIII $ \hat{\mathcal{U}}_{0,++}^{+}$, $ \hat{\mathcal{U}}_{T/2,+-}^{+}$ $ \hat{\mathcal{U}}_{0,-+}^{+}$, $ \hat{\mathcal{U}}_{T/2,--}^{+}$ $ \hat{\mathcal{U}}_{0,--}^{+}$, $ \hat{\mathcal{U}}_{0,-+}^{+}$ $ \hat{\mathcal{U}}_{0,+-}^{+}$, $ \hat{\mathcal{U}}_{T/2,++}^{+}$ 4 AII $ \hat{\mathcal{U}}_{0,+}^{+}$, $ \hat{\mathcal{U}}_{T/2,+}^{+}$ $ \overline{\mathcal{U}}_{0,-}^{+}$, $ \overline{\mathcal{U}}_{T/2,+}^{+}$ $ \hat{\mathcal{U}}_{0,-}^{+}$, $ \hat{\mathcal{U}}_{T/2,-}^{+}$ $ \overline{\mathcal{U}}_{0,+}^{+}$, $ \overline{\mathcal{U}}_{T/2,-}^{+}$ 5 CII $ \hat{\mathcal{U}}_{0,++}^{+}$, $ \hat{\mathcal{U}}_{T/2,+-}^{+}$ $ \hat{\mathcal

0,-}^{+}$, $\ovfrline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\mauhcal{U}}_{0,-}^{+}$, $\hat{\mathcal{O}}_{T/2,-}^{+}$ $\ovxrline{\mzthcal{U}}_{0,+}^{+}$, $\uverline{\mathcal{U}}_{T/2,-}^{+}$ 1 BDI $\hqt{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{G/2,+-}^{+}$ $\hwt{\mathcao{U}}_{0,+-}^{+}$, $\het{\mathcal{U}}_{T/2,++}^{+}$ $\hat{\matmeal{U}}_{0,--}^{+}$,$\hzb{\matheao{U}}_{T/2,-+}^{+}$ $\hat{\kathcal{U}}_{0,-+}^{+}$, $\hdt{\mathcal{U}}_{T/2,--}^{+}$ 2 D $\hat{\mathcal{U}}_{0,+}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\jverkyne{\mznhgal{U}}_{s,+}^{+}$ $\hat{\mzthcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,+}^{+}$ $\oveglinf{\mathcal{U}}_{s,-}^{+}$ 3 FIII $\haj{\jatrxal{U}}_{0,++}^{+}$, $\hat{\mathzal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal{O}}_{0,-+}^{+}$, $\hat{\mathcal{U}}_{T/2,--}^{+}$ $\hat{\mathcal{J}}_{0,--}^{+}$, $\hat{\kathcal{U}}_{0,-+}^{+}$ $\vat{\mathcal{U}}_{0,+-}^{+}$, $\hat{\mwthcal{U}}_{T/2,++}^{+}$ 4 AII $\hst{\mathcal{U}}_{0,+}^{+}$, $\hat{\katical{Y}}_{T/2,+}^{+}$ $\overline{\mathcal{N}}_{0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\matvccl{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\oceeline{\kathwal{U}}_{0,+}^{+}$, $\iveflihe{\kafhcal{U}}_{H/2,-}^{+}$ 5 CII $\hat{\matycal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal

0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\overline{\mathcal{U}}_{0,+}^{+}$, $\overline{\mathcal{U}}_{T/2,-}^{+}$ 1 $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal{U}}_{T/2,++}^{+}$ $\hat{\mathcal{U}}_{0,--}^{+}$,$\hat{\mathcal{U}}_{T/2,-+}^{+}$ $\hat{\mathcal{U}}_{T/2,--}^{+}$ 2 D $\overline{\mathcal{U}}_{s,-}^{+}$ DIII $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal{U}}_{T/2,--}^{+}$ $\hat{\mathcal{U}}_{0,--}^{+}$, $\hat{\mathcal{U}}_{0,-+}^{+}$ $\hat{\mathcal{U}}_{T/2,++}^{+}$ 4 AII $\hat{\mathcal{U}}_{0,+}^{+}$, $\hat{\mathcal{U}}_{T/2,+}^{+}$ $\overline{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ $\overline{\mathcal{U}}_{0,+}^{+}$, $\overline{\mathcal{U}}_{T/2,-}^{+}$ 5 CII $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ $\hat{\mathcal

0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\mathcAl{U}}_{0,-}^{+}$, $\hat{\mathCal{U}}_{T/2,-}^{+}$ $\OveRliNe{\MathCal{U}}_{0,+}^{+}$, $\Overline{\mathcaL{u}}_{T/2,-}^{+}$ 1 BDi $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{u}}_{T/2,+-}^{+}$ $\hat{\MaTHcal{u}}_{0,+-}^{+}$, $\HaT{\mathCal{U}}_{T/2,++}^{+}$ $\haT{\MaTHCal{u}}_{0,--}^{+}$,$\hAt{\MatHcAL{U}}_{t/2,-+}^{+}$ $\hat{\mAthCal{U}}_{0,-+}^{+}$, $\hat{\Mathcal{U}}_{T/2,--}^{+}$ 2 D $\Hat{\MaThcal{U}}_{0,+}^{+}$, $\hat{\matHCaL{U}}_{T/2,-}^{+}$ $\overlinE{\maThcal{U}}_{s,+}^{+}$ $\hat{\maThcAl{U}}_{0,-}^{+}$, $\hat{\MaThcAL{U}}_{T/2,+}^{+}$ $\ovErlIne{\maThcal{U}}_{S,-}^{+}$ 3 dIII $\haT{\mathcal{U}}_{0,++}^{+}$, $\HaT{\MathcaL{u}}_{T/2,+-}^{+}$ $\hat{\maTHCaL{U}}_{0,-+}^{+}$, $\haT{\mathcal{U}}_{T/2,--}^{+}$ $\hat{\mathCAl{u}}_{0,--}^{+}$, $\Hat{\mathcal{U}}_{0,-+}^{+}$ $\hat{\MathcaL{U}}_{0,+-}^{+}$, $\HAt{\MAThcAl{U}}_{t/2,++}^{+}$ 4 AII $\hat{\matHcAl{U}}_{0,+}^{+}$, $\haT{\Mathcal{u}}_{t/2,+}^{+}$ $\oVERLinE{\Mathcal{U}}_{0,-}^{+}$, $\overlIne{\mathcal{U}}_{t/2,+}^{+}$ $\Hat{\MathcaL{U}}_{0,-}^{+}$, $\Hat{\MAthcal{u}}_{T/2,-}^{+}$ $\oveRlINe{\mAthcal{U}}_{0,+}^{+}$, $\overLine{\Mathcal{U}}_{T/2,-}^{+}$ 5 cII $\hat{\MAthcal{U}}_{0,++}^{+}$, $\HAt{\mathcAl{U}}_{T/2,+-}^{+}$ $\haT{\maThcAl

0,-}^{+}$, $\overline{\mat hcal{U}}_{ T/2,+ }^{ +}$ $\ hat{ \mathcal{U}}_{ 0 ,-}^ {+}$, $\hat{\mathcal{U }}_{T /2 , -}^{ + }$ $ \overli n e{ \ m ath ca l{ U}} _{ 0 ,+ }^{+} $,$\overl ine{\mathc al{ U} }_{T/2,-}^{+ } $ 1 B DI $\hat{\ mat hcal{U }} _{0 , ++}^{ +}$ , $\h at{\ma t hcal{U }}_{T/2,+ -} ^ {+}$ $\h a t {\ math cal{U}}_{0,+-}^{+ } $, $\hat{\mathcal {U}}_{ T/ 2 ,+ + } ^{+ }$ $\ha t{ \math c al{U}}_ { 0, - - } ^{+ } $,$\hat{\math cal{U}}_{T/ 2 ,-+ }^{+}$ $\ha t{\ma th c al{ U}}_{0,-+}^ {+}$ , $\hat{\ mathca l {U}}_{T / 2,--}^{ +}$ 2 D $ \h a t{\ m at hca l {U} }_{0,+}^ {+ }$ , $\h at{\ m a t h cal{ U}} _{T/ 2,-}^ {+}$ $\ overl ine{ \m athca l{U}}_ {s,+} ^{ +}$ $\ hat {\ mat hc al{U} } _{0,-} ^{+ }$, $\hat{ \mathca l {U} }_ { T / 2, +}^{+}$ $\ overli n e{ \m a thcal{U} }_ {s, -}^{ + } $ 3 D III $\hat{ \ ma th cal{U}} _{ 0,++}^ {+ }$, $\ hat{\ m athc al{U}} _{T/2,+- }^{+} $ $\hat{\ m athcal{U}}_{0 , -+ } ^ {+ } $, $ \ha t{\mathcal{ U}}_ { T/2, --}^ { +} $ $\hat {\ m at h cal{U}}_{0,--}^{+}$ ,$\hat{ \math cal{U}}_{0,-+ }^{+}$ $\hat{\ math c al { U}}_{0,+-}^{+} $, $\ hat{\mathc a l{U}}_{T /2,++ }^{+}$ 4 A I I $ \ha t{\ mat hca l { U} }_{0,+}^{+}$, $ \hat {\ mathcal {U} }_{T/2, +}^ {+} $ $ \o verline{\ mathcal{ U} }_ {0 ,- }^{ +}$,$ \overlin e{ \ma th cal {U}}_ { T/2,+} ^{+}$ $\ ha t {\m athcal{ U }} _ { 0,-} ^{ +} $, $ \ha t{ \math cal{ U }}_ {T/2,-} ^{+}$ $\o v erli ne {\ mathcal {U}}_{0,+}^{+ }$ , $\overli ne {\m athcal { U }}_{T/2, -}^{+}$ 5 CII $\hat {\m athca l{U} }_{0,++}^ {+} $, $\h at{ \ mathca l{U}}_ {T/2, +- }^{ + } $ $ \ha t{ \mathcal

0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$_ _ $\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,-}^{+}$ _ _$\overline{\mathcal{U}}_{0,+}^{+}$,_$\overline{\mathcal{U}}_{T/2,-}^{+}$ __1 _ BDI_ _ $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$__ $\hat{\mathcal{U}}_{0,+-}^{+}$, $\hat{\mathcal{U}}_{T/2,++}^{+}$ $\hat{\mathcal{U}}_{0,--}^{+}$,$\hat{\mathcal{U}}_{T/2,-+}^{+}$__ ___ $\hat{\mathcal{U}}_{0,-+}^{+}$, $\hat{\mathcal{U}}_{T/2,--}^{+}$ _ 2 _ _D _$\hat{\mathcal{U}}_{0,+}^{+}$,_$\hat{\mathcal{U}}_{T/2,-}^{+}$ _ _ _ ___$\overline{\mathcal{U}}_{s,+}^{+}$ _ _ _ _$\hat{\mathcal{U}}_{0,-}^{+}$, $\hat{\mathcal{U}}_{T/2,+}^{+}$ _ _ _ _ _ _ $\overline{\mathcal{U}}_{s,-}^{+}$ _ 3 _DIII _ $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$ __ _$\hat{\mathcal{U}}_{0,-+}^{+}$,_$\hat{\mathcal{U}}_{T/2,--}^{+}$_ _ __ $\hat{\mathcal{U}}_{0,--}^{+}$,_$\hat{\mathcal{U}}_{0,-+}^{+}$_ __$\hat{\mathcal{U}}_{0,+-}^{+}$, $\hat{\mathcal{U}}_{T/2,++}^{+}$ 4 _ AII __ _$\hat{\mathcal{U}}_{0,+}^{+}$, $\hat{\mathcal{U}}_{T/2,+}^{+}$ _$\overline{\mathcal{U}}_{0,-}^{+}$, $\overline{\mathcal{U}}_{T/2,+}^{+}$ $\hat{\mathcal{U}}_{0,-}^{+}$,_$\hat{\mathcal{U}}_{T/2,-}^{+}$ _ _$\overline{\mathcal{U}}_{0,+}^{+}$, $\overline{\mathcal{U}}_{T/2,-}^{+}$ 5 CII _ $\hat{\mathcal{U}}_{0,++}^{+}$, $\hat{\mathcal{U}}_{T/2,+-}^{+}$_ _ _$\hat{\mathcal

1(0,T;U)} - \varepsilon}{\norm{\bg - \hat\bg^\tau}_{H^1(0,T;U)} - \varepsilon/2} \right) \in [0,1].$$ In case $\norm{\bg - \hat\bg^\tau}_{H^1(0,T;U)} > \varepsilon$, we find (all norms are those of $H^1(0,T;U)$) $$\begin{aligned} \norm{\bg - \tilde\bg^\tau} & \le (1 - \eta^\tau) \, \norm{\bg - \hat\bg^\tau} + \eta^\tau \, \norm{\bg - \bg^\tau} \\ & = \frac{1}{\norm{\bg - \hat\bg^\tau} - \varepsilon/2} \, \Bigh(){ \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau} + (\norm{\bg - \hat\bg^\tau} - \varepsilon) \, \norm{\bg - \bg^\tau} } \\ & \le \frac{1}{\norm{\bg - \hat\bg^\tau} - \varepsilon/2} \, \Bigh(){ \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau} + (\norm{\bg - \hat\bg^\tau} - \varepsilon) \, \frac\varepsilon2 } \\ & = \varepsilon. \end{aligned}$$ This shows $\norm{\bg - \tilde\bg^\tau}_{H^1(0,T;U)} \le \varepsilon$. Moreover, $\eta^\tau \in [0,1]$ implies $\tilde\bg^\tau \in {U_\textup{ad}}$. Therefore, $\tilde\bg^\tau \in {\hat U_\textup{ad}}$. Further, $\lim \norm{\bg - \hat\bg^\tau}_{H^1(0,T;U)} = \varepsilon$ implies $\eta^\tau \to 0$ and hence $\tilde\bg^\tau \to \tilde\bg$. This shows that ${\hat U_\textup{ad}}$ satisfies. *Step (3):* We define the auxiliary problem $$\label{eq:ulp_aux} \tag{$\mathbf{P}_{\bg,\varepsilon}$} \left.

1(0,T;U) } - \varepsilon}{\norm{\bg - \hat\bg^\tau}_{H^1(0,T;U) } - \varepsilon/2 } \right) \in [ 0,1].$$ In case $ \norm{\bg - \hat\bg^\tau}_{H^1(0,T;U) } > \varepsilon$, we find (all norms are those of $ H^1(0,T;U)$) $ $ \begin{aligned } \norm{\bg - \tilde\bg^\tau } & \le (1 - \eta^\tau) \, \norm{\bg - \hat\bg^\tau } + \eta^\tau \, \norm{\bg - \bg^\tau } \\ & = \frac{1}{\norm{\bg - \hat\bg^\tau } - \varepsilon/2 } \, \Bigh () { \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau } + (\norm{\bg - \hat\bg^\tau } - \varepsilon) \, \norm{\bg - \bg^\tau } } \\ & \le \frac{1}{\norm{\bg - \hat\bg^\tau } - \varepsilon/2 } \, \Bigh () { \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau } + (\norm{\bg - \hat\bg^\tau } - \varepsilon) \, \frac\varepsilon2 } \\ & = \varepsilon. \end{aligned}$$ This shows $ \norm{\bg - \tilde\bg^\tau}_{H^1(0,T;U) } \le \varepsilon$. Moreover, $ \eta^\tau \in [ 0,1]$ entail $ \tilde\bg^\tau \in { U_\textup{ad}}$. Therefore, $ \tilde\bg^\tau \in { \hat U_\textup{ad}}$. Further, $ \lim \norm{\bg - \hat\bg^\tau}_{H^1(0,T;U) } = \varepsilon$ entail $ \eta^\tau \to 0 $ and hence $ \tilde\bg^\tau \to \tilde\bg$. This shows that $ { \hat U_\textup{ad}}$ satisfies. * Step   (3 ): * We specify the auxiliary problem $ $ \label{eq: ulp_aux } \tag{$\mathbf{P}_{\bg,\varepsilon}$ } \left.

1(0,T;U)} - \varepsilon}{\norm{\bg - \hat\bg^\uau}_{H^1(0,T;U)} - \varepsilou/2} \right) \in [0,1].$$ Ih case $\nurm{\bg - \hat\bg^\tau}_{H^1(0,T;U)} > \varepsilln$, we fund (all norms are thosd of $H^1(0,T;U)$) $$\hegin{alitned} \norm{\bg - \tilde\bg^\tah} & \le (1 - \eta^\tao) \, \norm{\bg - \hdt\bg^\tau} + \eta^\tag \, \nlrm{\bg - \bg^\tau} \\ & = \frac{1}{\norm{\bg - \rat\bg^\tai} - \varepsilon/2} \, \Bygh(){ \gwac\vzgeksilon2 \, \norm{\bg - \hat\bg^\tau} + (\norm{\bg - \hat\bg^\teu} - \varepsilon) \, \norm{\bg - \bg^\tau} } \\ & \lf \frwc{1}{\norm{\bg - \hat\bg^\tau} - \varepsilob/2} \, \Byth(){ \frac\vareprilon2 \, \norm{\bg - \hat\bg^\tao} + (\norm{\bg - \hat\bg^\tau} - \varepsilon) \, \frac\rarepsilon2 } \\ & = \varepsmlon. \end{aligned}$$ This shmws $\nork{\bg - \tilde\bg^\tam}_{H^1(0,T;U)} \ne \carepsilon$. Moreover, $\eva^\tau \in [0,1]$ implies $\tijde\bg^\tau \hn {U_\textup{ad}}$. Therefirw, $\tilge\bg^\dau \kb {\hxt H_\txxthp{ad}}$. Fkrtier, $\lim \norj{\bg - \hat\bg^\tqu}_{H^1(0,T;U)} = \varepsilon$ ikpjprs $\eta^\tau \to 0$ and regce $\tilde\bg^\tau \to \tilde\bg$. This shows thdt ${\gat U_\textup{ad}}$ satisfies. *Wtep (3):* We define the audiliary pwoblem $$\label{eq:ulp_aux} \tag{$\mathbf{P}_{\bg,\varepsilot}$} \oent.

1(0,T;U)} - \varepsilon}{\norm{\bg - \hat\bg^\tau}_{H^1(0,T;U)} - \varepsilon/2} [0,1].$$ case $\norm{\bg \hat\bg^\tau}_{H^1(0,T;U)} > \varepsilon$, those $H^1(0,T;U)$) $$\begin{aligned} \norm{\bg \tilde\bg^\tau} & \le - \eta^\tau) \, \norm{\bg - \hat\bg^\tau} \eta^\tau \, \norm{\bg - \bg^\tau} \\ & = \frac{1}{\norm{\bg - \hat\bg^\tau} - \varepsilon/2} \Bigh(){ \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau} + (\norm{\bg - \hat\bg^\tau} - \varepsilon) \, - } & \frac{1}{\norm{\bg - \hat\bg^\tau} - \varepsilon/2} \, \Bigh(){ \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau} + (\norm{\bg - \hat\bg^\tau} \varepsilon) \, \frac\varepsilon2 } \\ & = \varepsilon. This shows $\norm{\bg - \le \varepsilon$. Moreover, $\eta^\tau \in implies \in {U_\textup{ad}}$. $\tilde\bg^\tau {\hat Further, $\lim \norm{\bg \hat\bg^\tau}_{H^1(0,T;U)} = \varepsilon$ implies $\eta^\tau \to 0$ and hence $\tilde\bg^\tau \to \tilde\bg$. This shows that ${\hat U_\textup{ad}}$ *Step (3):* the auxiliary $$\label{eq:ulp_aux} \left.

1(0,T;U)} - \varepsilon}{\norm{\bg - \hat\bg^\taU}_{H^1(0,T;U)} - \varepsIlon/2} \rIghT) \in [0,1].$$ in Case $\Norm{\Bg - \hat\bg^\tau}_{H^1(0,T;U)} > \vARepsIlon$, we find (all norms are tHose oF $H^1(0,t;u)$) $$\begIN{aLigneD} \norm{\bg - \TIlDE\Bg^\tAu} & \Le (1 - \Eta^\TaU) \, \NoRm{\bg - \hAt\bG^\tau} + \eta^\Tau \, \norm{\bg - \bG^\taU} \\ & = \fRac{1}{\norm{\bg - \hat\BG^\tAu} - \varepsilOn/2} \, \BIgh(){ \frac\varepSilOn2 \, \norm{\Bg - \Hat\BG^\tau} + (\nOrm{\Bg - \hat\Bg^\tau} - \vARepsilOn) \, \norm{\bg - \bG^\tAU} } \\ & \le \fraC{1}{\Norm{\bg - \hAT\Bg^\Tau} - \vArepsilon/2} \, \Bigh(){ \frac\VArEPsilon2 \, \norm{\bg - \haT\bg^\tau} + (\NoRM{\bG - \HAt\bG^\taU} - \varepsiloN) \, \fRac\vaREpsilon2 } \\ & = \VArEPSIloN. \End{aligned}$$ ThiS shows $\norm{\bG - \TilDe\bg^\taU}_{H^1(0,t;U)} \lE \VarepsIlon$. MOrEOveR, $\eta^\tau \in [0,1]$ imPlieS $\tilde\bg^\tAu \in {U_\tEXtup{ad}}$. THErefore, $\Tilde\bG^\taU \in {\Hat U_\TExTuP{ad}}$. fuRTheR, $\LiM \noRM{\bg - \Hat\bg^\tau}_{h^1(0,T;u)} = \vArepsIlon$ IMPLIes $\eTa^\tAu \to 0$ And heNce $\tilde\bg^\tau \To \tIlde\BG$. ThIs shoWs thaT ${\hat u_\tExtup{Ad}}$ satiSfies. *stEp (3):* We define the auXiliAry probleM $$\laBeL{eq:UlP_aux} \tAG{$\mathbF{P}_{\bG,\vaRepsiloN}$} \left.

1(0,T;U)} - \varepsilon}{\ norm{\bg - \hat \bg ^\t au }_{H ^1(0 ,T;U)} - \vare p silo n/2} \right) \in [0,1] .$$ I nc ase$ \n orm{\ bg - \h a t\ b g ^\t au }_ {H^ 1( 0 ,T ;U)}> \ varepsi lon$, we f ind ( all norms ar e t hose of $H ^1( 0,T;U)$) $$\ beg in{ali gn ed} \norm {\bg - \tilde \bg^\tau} & \le (1 - \ eta^ \tau) \, \norm{\b g - \hat\bg^\tau}+ \eta ^\ t au \ , \ nor m{\bg - \b g^ \tau} \\ & = \ f rac{1}{\norm{ \bg - \hat\ b g^\ tau} - \ var e psilon /2} \ ,\ Big h(){ \frac\ vare psilon2 \ , \nor m {\bg -\ hat\bg^ \tau}+ ( \no rm{\ b g-\ha t\ b g^\ t au } - \va repsilon )\, \nor m{\b g - \bg^ \ta u} } \\ & \le\fr ac{1 } {\n orm{\ bg -\hat \b g^\ta u} - \ varep si lon/2} \, \Bigh (){\frac\var eps il on2 \ , \no r m{\bg- \ hat \bg^\ta u} + (\ n orm {\ b g -\hat\bg^\tau} - \v ar e p si lon) \,\frac\ v ar ep s ilon2 }\\ & = \ vare p si lon. \end{ a li gn ed}$$ T hi s show s$\n orm {\bg- \ti lde\bg ^\tau}_{ H^1(0 , T;U)} \le \var e psilon$. More o ve r , $ \ eta^ \ta u \in [0,1] $ im p lies $\t i ld e\b g ^\tau \in{U _ \t e xtup{ad}}$. Therefo re , $\ti lde\b g^\tau \in {\ hat U_\tex t u p {ad}}$.Furt h er , $\lim \norm{\ bg -\hat\bg^\t a u}_{H^1( 0,T;U )} = \va repsilon$ i mplies $ \et a^\ tau \t o 0$ and hence $\ t i lde\ bg ^\tau \ to\tilde\ bg$ . T his sh ow s that ${ \hat U_\ te xt up {a d}} $ sat i sfies. *S tep  ( 3): * Wed efinethe a uxil ia ry pro blem $$ \ la b e l{eq :u lp _aux } \tag { $\m athbf{P }_{\bg,\v are p silo n} $} \left.

1(0,T;U)} -_\varepsilon}{\norm{\bg -_\hat\bg^\tau}_{H^1(0,T;U)} - \varepsilon/2} \right)_\in [0,1].$$_In_case $\norm{\bg_-_\hat\bg^\tau}_{H^1(0,T;U)} > \varepsilon$,_we find (all_norms are those of_$H^1(0,T;U)$) $$\begin{aligned} __ \norm{\bg - \tilde\bg^\tau} __& \le_(1_-_\eta^\tau) \, \norm{\bg - \hat\bg^\tau}_+ \eta^\tau \, \norm{\bg -_\bg^\tau} \\ _ &_=_\frac{1}{\norm{\bg - \hat\bg^\tau}_- \varepsilon/2} \, \Bigh(){ \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau}_+ (\norm{\bg - \hat\bg^\tau} - \varepsilon)_\, \norm{\bg -_\bg^\tau}_}_\\ _ & \le_\frac{1}{\norm{\bg - \hat\bg^\tau} - \varepsilon/2} \,_\Bigh(){ \frac\varepsilon2 \, \norm{\bg - \hat\bg^\tau} +_(\norm{\bg - \hat\bg^\tau} - \varepsilon) \,_\frac\varepsilon2 } \\ _ _ & =_\varepsilon. _ \end{aligned}$$_This shows $\norm{\bg_- \tilde\bg^\tau}_{H^1(0,T;U)} \le \varepsilon$. Moreover, $\eta^\tau_\in [0,1]$ implies_$\tilde\bg^\tau \in {U_\textup{ad}}$. Therefore, $\tilde\bg^\tau \in_{\hat_U_\textup{ad}}$. Further, $\lim_\norm{\bg_-_\hat\bg^\tau}_{H^1(0,T;U)} =_\varepsilon$ implies $\eta^\tau_\to_0$ and_hence_$\tilde\bg^\tau \to \tilde\bg$. This shows that_${\hat_U_\textup{ad}}$ satisfies. *Step (3):* We define the auxiliary problem_$$\label{eq:ulp_aux} __ \tag{$\mathbf{P}_{\bg,\varepsilon}$} _ _ \left.

theta. \label{twistedcoprod}$$ Here $\partial_\mu \otimes \theta^{\mu \nu}\partial_\nu$ in $F_\theta$ is to be replaced by $~-~ P_\mu \otimes \theta^{\mu \nu} P_\nu$ where $P_\mu$ are translation generators: we are dealing with ${\cal{P}}_\theta \otimes {\cal{P}}_\theta$ where ${\cal{P}}_\theta$ is a Hopf algebra associated with the Poincaré group algebra ${\mathbb C}{\cal{P}}$ with the coproduct (\[twistedcoprod\]). This twisted coproduct does not preserve standard (anti-)symmetrization, because it does not commute with the usual flip operator $\tau_0$ defined by $\tau_0:(\alpha \otimes \beta) ~=~ (\beta\otimes\alpha)$: $$\Delta_\theta(\Lambda) \tau_0 \neq \tau_0 \Delta_\theta (\Lambda).$$ On the other hand, it does preserve twisted (anti-)symmetrization, defined using a new flip operator $\tau_\theta = F_\theta^{-1}\,\tau_0 (\Lambda) F_\theta$: $$\Delta_\theta(\Lambda) \tau_\theta = \tau_\theta \Delta_\theta (\Lambda).$$ Thus in noncommutative quantum theory, the usual fermions/bosons do not make sense, but twisted ones do. They are obtained from the projectors $S_\theta, A_\theta$: $$S_\theta~=~\frac{{\bf 1}~+~\tau_\theta}{2}, \qquad A_\theta~=~ \frac{{\bf 1}~-~\tau_\theta}{2}.$$ Quantum Fields -------------- A quantum field $\chi$ on evaluation at a spacetime point (or more generally on pairing with a test function) gives an operator acting on a Hilbert space. A field at $x_1$ acting on the vacuum gives a one-particle state centered at $x_1$. When we write $\chi(x_1)\,\chi(x_2)$ we mean $(\chi\otimes\chi)(x_1,x_2)$. Acting on the vacuum we generate a two-particle state, where one particle is centered at $x_1$ and the other at $x_2$. If $a_p$ is the annihilation operator of the free second-quantized field

theta. \label{twistedcoprod}$$ Here $ \partial_\mu \otimes \theta^{\mu \nu}\partial_\nu$ in $ F_\theta$ is to be replaced by $ ~-~ P_\mu \otimes \theta^{\mu \nu } P_\nu$ where $ P_\mu$ are translation generators: we are dealing with $ { \cal{P}}_\theta \otimes { \cal{P}}_\theta$ where $ { \cal{P}}_\theta$ is a Hopf algebra consociate with the Poincaré group algebra $ { \mathbb C}{\cal{P}}$ with the coproduct (\[twistedcoprod\ ]). This distorted coproduct does not preserve standard (anti-)symmetrization, because it does not commute with the common flip hustler $ \tau_0 $ defined by $ \tau_0:(\alpha \otimes \beta) ~=~ (\beta\otimes\alpha)$: $ $ \Delta_\theta(\Lambda) \tau_0 \neq \tau_0 \Delta_\theta (\Lambda).$$ On the other hired hand, it does preserve twisted (anti-)symmetrization, specify using a new impudent operator $ \tau_\theta = F_\theta^{-1}\,\tau_0 (\Lambda) F_\theta$: $ $ \Delta_\theta(\Lambda) \tau_\theta = \tau_\theta \Delta_\theta (\Lambda).$$ Thus in noncommutative quantum theory, the usual fermion / bosons do not make common sense, but twisted one make. They are obtained from the projectors $ S_\theta, A_\theta$: $ $ S_\theta~=~\frac{{\bf 1}~+~\tau_\theta}{2 }, \qquad A_\theta~=~ \frac{{\bf 1}~-~\tau_\theta}{2}.$$ Quantum Fields -------------- A quantum field $ \chi$ on evaluation at a spacetime compass point (or more generally on pairing with a test function) pass an operator acting on a Hilbert space. A field at $ x_1 $ acting on the void gives a one - particle state centered at $ x_1$. When we write $ \chi(x_1)\,\chi(x_2)$ we mean $ (\chi\otimes\chi)(x_1,x_2)$. Acting on the void we render a two - particle state, where one particle is centered at $ x_1 $ and the other at $ x_2$. If $ a_p$ is the annihilation operator of the free second - quantized field

theha. \label{twistedcoprod}$$ Hert $\partial_\mu \otimes \theta^{\ku \nu}\pzrtial_\nu$ in $F_\theta$ is to be replaced bt $~-~ P_\my \otimes \theta^{\mu \nu} P_\nj$ where $P_\lu$ are teansoqtion genecztors: wc are fealnnj with ${\cal{P}}_\thets \otimes {\cdl{P}}_\theta$ where ${\cxl{'}}_\theta$ is a Hopf algebra associated rith thr Ooincaré group wlgenwa ${\mznhnb C}{\cal{P}}$ with the coproduct (\[twisfedcoprmd\]). This twistec coproduct does not presegve dtandard (anti-)symmehrization, bgdaufw it does nog commute with the usuzl flip operator $\tau_0$ defined by $\tau_0:(\akpha \otimew \vetw) ~=~ (\beta\otimes\elpha)$: $$\Dvlta_\theta(\Lambda) \tau_0 \naq \tau_0 \Celta_\theta (\Lamnda).$$ On thw other hand, it does 'reserve twisted (anty-)symmetrisacion, defined using a beq flik opesatof $\taj_\thtta = R_\theta^{-1}\,\hau_0 (\Mambda) F_\thsta$: $$\Delta_\thwta(\Lambda) \tau_\theta = \twl_\yheta \Delta_\tgeta (\Lwmfda).$$ Thus in noncommutative quantum theorj, ths usual fermions/bosons eo not make sense, but twisted jnes do. They are obtained from the projectors $S_\thata, A_\vhdta$: $$S_\bhetx~=~\drwc{{\bf 1}~+~\tau_\theta}{2}, \qquad A_\theta~=~ \frac{{\bf 1}~-~\tau_\theta}{2}.$$ Quagfuk Nields -------------- A quantum nield $\chi$ on evalustlom at a spacetioe poiuf (kr more generally ln pairyng wuth a tesu funvtion) gives an operator actung on a Hilyerr space. A field at $x_1$ acting ou the facuuk gives a one-particle scate csntered at $d_1$. When we drite $\chi(x_1)\,\chi(x_2)$ wd mvan $(\whi\otimes\chi)(x_1,x_2)$. Acting on tre vacuum we yenerate a teo-partycle state, wherc one particle is cfntergd at $f_1$ and the lther at $x_2$. If $a_p$ is the annihilavmon operator pf thv free seeond-qusntized field

theta. \label{twistedcoprod}$$ Here $\partial_\mu \otimes \theta^{\mu \nu}\partial_\nu$ is be replaced $~-~ P_\mu \otimes are generators: we are with ${\cal{P}}_\theta \otimes where ${\cal{P}}_\theta$ is a Hopf algebra with the Poincaré group algebra ${\mathbb C}{\cal{P}}$ with the coproduct (\[twistedcoprod\]). This twisted does not preserve standard (anti-)symmetrization, because it does not commute with the usual operator defined $\tau_0:(\alpha \beta) ~=~ (\beta\otimes\alpha)$: $$\Delta_\theta(\Lambda) \tau_0 \neq \tau_0 \Delta_\theta (\Lambda).$$ On the other hand, it does preserve (anti-)symmetrization, defined using a new flip operator $\tau_\theta F_\theta^{-1}\,\tau_0 (\Lambda) F_\theta$: $$\Delta_\theta(\Lambda) = \tau_\theta \Delta_\theta (\Lambda).$$ Thus noncommutative theory, the fermions/bosons not sense, but twisted do. They are obtained from the projectors $S_\theta, A_\theta$: $$S_\theta~=~\frac{{\bf 1}~+~\tau_\theta}{2}, \qquad A_\theta~=~ \frac{{\bf 1}~-~\tau_\theta}{2}.$$ Quantum Fields A quantum on evaluation a point more generally on a test function) gives an operator Hilbert space. A field at $x_1$ acting on vacuum gives one-particle state centered at $x_1$. When write $\chi(x_1)\,\chi(x_2)$ we mean $(\chi\otimes\chi)(x_1,x_2)$. Acting on the we generate a two-particle state, where one particle is centered at $x_1$ and the other If $a_p$ is the operator of the second-quantized

theta. \label{twistedcoprod}$$ HeRe $\partial_\mU \otimEs \tHetA^{\mU \nu}\pArtiAl_\nu$ in $F_\theta$ is TO be rEplaced by $~-~ P_\mu \otimes \thetA^{\mu \nu} p_\nU$ WherE $p_\mU$ are tRanslatIOn GENerAtOrS: we ArE DeAling WitH ${\cal{P}}_\thEta \otimes {\cAl{P}}_\ThEta$ where ${\cal{P}}_\THeTa$ is a Hopf aLgeBra associateD wiTh the POiNcaRÉ grouP alGebra ${\Mathbb c}{\Cal{P}}$ wiTh the coprOdUCt (\[twisTEdcoproD\]). tHiS twiSted coproduct does NOt PReserve standarD (anti-)sYmMEtRIZatIon, Because it dOeS not cOMmute wiTH tHE USuaL Flip operator $\tAu_0$ defined by $\TAu_0:(\aLpha \otImEs \bETa) ~=~ (\beta\OtimeS\aLPha)$: $$\delta_\theta(\LAmbdA) \tau_0 \neq \taU_0 \Delta_\THeta (\LamBDa).$$ On the Other hAnd, It dOes pREsErVe tWiSTed (ANtI-)syMMetRization, DeFiNed usIng a NEW FLip oPerAtor $\Tau_\thEta = F_\theta^{-1}\,\tau_0 (\LAmbDa) F_\tHEta$: $$\delta_\Theta(\lambDa) \Tau_\thEta = \tau_\Theta \deLta_\theta (\Lambda).$$ THus iN noncommuTatIvE quAnTum thEOry, the UsuAl fErmions/Bosons dO Not MaKE SEnSe, but twisted ones do. thEY ArE obtaineD from tHE pRoJEctors $S_\tHeTa, A_\ThetA$: $$s_\Theta~=~\Frac{{\BF 1}~+~\tAu_\theta}{2}, \qQuad A_\tHEtA~=~ \fRac{{\bf 1}~-~\taU_\tHeta}{2}.$$ QuAnTum fieLds -------------- A qUAntuM field $\Chi$ on evaLuatiON at a spacetime pOInt (or more geneRAlLY On PAiriNg wIth a test funCtioN) GiveS an oPErAtoR ActinG on a HIlBErT Space. A field at $x_1$ actinG oN the vaCuum gIves a one-partiCle state ceNTERed at $x_1$. WhEn we WRiTE $\chi(x_1)\,\chi(x_2)$ we meaN $(\chi\oTimes\chi)(x_1,x_2)$. aCting on tHe vacUum we genErate a two-PARticle stAte, WheRe oNe pARTiCle is centered AT $X_1$ and ThE other aT $x_2$. IF $a_p$ is thE anNihIlaTioN oPerator of The free sEcOnD-qUaNtiZed fiELd

theta. \label{twistedcopro d}$$ Here$\par tia l_\ mu \ot imes \theta^{\mu \ n u}\p artial_\nu$ in $F_\the ta$ i st o be re place d by $~ - ~P _ \mu \ ot ime s\ th eta^{ \mu \nu} P _\nu$ wher e $ P_ \mu$ are tra n sl ation gene rat ors: we aredea ling w it h $ { \cal{ P}} _\the ta \ot i mes {\ cal{P}}_\ th e ta$ wh e re ${\c a l {P }}_\ theta$ is a Hopfa lg e bra associated withth e P o i nca régroup alge br a ${\ m athbb C } {\ c a l {P} } $ with the co product (\[ t wis tedcop ro d\] ) . Thi s twi st e d c oproduct do es n ot preser ve sta n dard (a n ti-)sym metriz ati on, bec a us eitdo e s n o tcom m ute with th eus ual f lipo p e r ator $\ tau_ 0$ de fined by $\ta u_0 :(\a l pha \oti mes \ beta )~=~ ( \beta\ otime s\ alpha)$: $$\De lta_ \theta(\L amb da ) \ ta u_0 \ n eq \ta u_0 \D elta_\t heta (\ L amb da ) . $ $ On the other hand ,i t d oes pres erve t w is te d (anti-) sy mme triz a t ion,defi n ed using a new f l ip o perator $ \tau_\ th eta =F_\th e ta^{ -1}\,\ tau_0 (\ Lambd a ) F_\theta$: $ $ \Delta_\theta ( \L a m bd a ) \t au_ \theta = \t au_\ t heta \De l ta _\t h eta ( \Lamb da ) .$ $ Thus in noncommut at ive qu antum theory, theusual ferm i o n s/bosons don ot make sense, bu t twi sted onesd o. Theyare o btainedfrom thep r ojectors $S _\t het a,A _ \t heta$: $$S_\t h e ta~= ~\ frac{{\ bf1}~+~\t au_ \th eta }{2 }, \qquad A _\theta~ =~ \f ra c{{ \bf 1 } ~-~\tau_ \t het a} {2} .$$ Q uantum Fiel ds - -- -- - --- ----- A q u a ntum f ie ld $ \ch i$ on e valu a tio n at aspacetime po i nt ( or m ore gen erally on pai ri ng with ate stfuncti o n ) givesan operator acting on a Hilbert sp ace.A fi eld at $x _1$ actin g o n the v acuumgives a on e - parti c l esta te centereda t $x _1$.Wh en w e write $\chi(x_1)\,\chi( x _2) $ we mean $(\ chi \oti m e s\ chi ) (x _ 1,x _2 ) $.A c ting on the vac uum we gen er a te a two-par t icl estate,where o ne pa r ticle i s centere d at $x_1 $andt h e o ther at $x _2$. If $a_p$ is the a n ni hilat ion opera to r o f the frees eco nd-qu antize dfield

theta. \label{twistedcoprod}$$ Here_$\partial_\mu \otimes_\theta^{\mu \nu}\partial_\nu$ in $F_\theta$_is to_be_replaced by_$~-~_P_\mu \otimes \theta^{\mu_\nu} P_\nu$ where_$P_\mu$ are translation generators:_we are dealing_with_${\cal{P}}_\theta \otimes {\cal{P}}_\theta$ where ${\cal{P}}_\theta$ is a Hopf algebra associated with the Poincaré group_algebra_${\mathbb C}{\cal{P}}$_with_the_coproduct (\[twistedcoprod\]). This twisted coproduct does_not preserve standard (anti-)symmetrization, because_it does_not commute with the usual flip operator $\tau_0$_defined_by $\tau_0:(\alpha \otimes_\beta) ~=~ (\beta\otimes\alpha)$: $$\Delta_\theta(\Lambda) \tau_0 \neq \tau_0 \Delta_\theta (\Lambda).$$ On the other_hand, it does preserve twisted (anti-)symmetrization,_defined using a_new_flip_operator $\tau_\theta = F_\theta^{-1}\,\tau_0 (\Lambda)_F_\theta$: $$\Delta_\theta(\Lambda) \tau_\theta = \tau_\theta \Delta_\theta_(\Lambda).$$ Thus in noncommutative quantum theory, the_usual fermions/bosons do not make sense, but_twisted ones do. They are obtained_from the projectors $S_\theta, A_\theta$:_$$S_\theta~=~\frac{{\bf 1}~+~\tau_\theta}{2},_\qquad A_\theta~=~ \frac{{\bf 1}~-~\tau_\theta}{2}.$$ Quantum Fields -------------- A_quantum field $\chi$_on evaluation_at a spacetime_point (or more generally on pairing_with a test_function) gives an operator acting on_a_Hilbert space. A_field_at_$x_1$ acting_on the vacuum_gives_a one-particle_state_centered at $x_1$. When we write_$\chi(x_1)\,\chi(x_2)$_we mean $(\chi\otimes\chi)(x_1,x_2)$. Acting on the vacuum_we generate a two-particle_state,_where one particle is_centered at $x_1$ and the_other at $x_2$. If $a_p$ is the_annihilation operator_of the_free second-quantized field

1,c_2)^{\mu}$ and the open subscheme consisting of $\mu$-stable vector bundles by $M_H(r;c_1,c_2)^{\mu}_0$. For a torsion free sheaf $E$ on $X$, we set $\mu(E)=\frac{c_1(E)}{{\operatorname{rk}}(E)} \in H^2(X,\Bbb Q)$ and $\Delta(E)=\frac{1}{{\operatorname{rk}}(E)}(c_2(E)-\frac{{\operatorname{rk}}(E)-1}{2{\operatorname{rk}}(E)}c_1(E)^2) \in H^4(X,\Bbb Q)$. For a $x \in H^2(X,\Bbb Q)$, we set $P(x)=(x,x-K_X)/2+\chi(\cal O_X)$. For a scheme $S$, we denote the projection $X \times S \to S$ by $p_S$. In this section, we shall generalize the chamber structure of polarizations in \[Q2\]. For a torsion free sheaf $E$, we set $\gamma(E):=({\operatorname{rk}}(E),\mu(E),\Delta(E)) \in H^0(X,\Bbb Q) \times H^2(X,\Bbb Q) \times H^4(X,\Bbb Q)$. For $\gamma \in \prod_{i=0}^2 H^{2i}(X, \Bbb Q)$, let $M_H^{\gamma}$ be the set of torsion free sheaves $E$ defined over $\Bbb C$ with $\gamma(E)=\gamma$ which is $\mu$-semi-stable with respect to $H$. \[lem:1\] Let $E$ be a torsion free sheaf which is defined by an extension $ 0 \to F_1 \to E \to F_2 \to 0$. Then $\Delta(E)=\frac{{\operatorname{rk}}(F_1)}{{\operatorname{rk}}(E)}\Delta(F_1)+\frac{{\operatorname{rk}}(F_2)}{{\operatorname{rk}}(E)}\Delta(F_2)- \frac{{\operatorname{rk}}(F_1){\operatorname{rk}}(F_2)}{2{\operatorname{rk}}(E)^2}((\mu(F_1)-\mu(F_2))^2)$. Let $B$ be

1,c_2)^{\mu}$ and the open subscheme consisting of $ \mu$-stable vector bundles by $ M_H(r;c_1,c_2)^{\mu}_0$. For a torsion spare bundle $ E$ on $ X$, we set $ \mu(E)=\frac{c_1(E)}{{\operatorname{rk}}(E) } \in H^2(X,\Bbb Q)$ and $ \Delta(E)=\frac{1}{{\operatorname{rk}}(E)}(c_2(E)-\frac{{\operatorname{rk}}(E)-1}{2{\operatorname{rk}}(E)}c_1(E)^2) \in H^4(X,\Bbb Q)$. For a $ x \in H^2(X,\Bbb Q)$, we set $ P(x)=(x, x - K_X)/2+\chi(\cal O_X)$. For a schema $ S$, we announce the projection $ X \times S \to S$ by $ p_S$. In this section, we shall generalize the bedroom social organization of polarizations in \[Q2\ ]. For a torsion free sheaf $ E$, we set $ \gamma(E):=({\operatorname{rk}}(E),\mu(E),\Delta(E) ) \in H^0(X,\Bbb Q) \times H^2(X,\Bbb Q) \times H^4(X,\Bbb Q)$. For $ \gamma \in \prod_{i=0}^2 H^{2i}(X, \Bbb Q)$, lease $ M_H^{\gamma}$ be the set of torsion spare sheaves $ E$ defined over $ \Bbb C$ with $ \gamma(E)=\gamma$ which is $ \mu$-semi - stable with respect to $ H$. \[lem:1\ ] Let $ E$ be a tortuosity free sheaf which is defined by an reference $ 0 \to F_1 \to E \to F_2 \to 0$. Then $ \Delta(E)=\frac{{\operatorname{rk}}(F_1)}{{\operatorname{rk}}(E)}\Delta(F_1)+\frac{{\operatorname{rk}}(F_2)}{{\operatorname{rk}}(E)}\Delta(F_2)- \frac{{\operatorname{rk}}(F_1){\operatorname{rk}}(F_2)}{2{\operatorname{rk}}(E)^2}((\mu(F_1)-\mu(F_2))^2)$. Let $ B$ be

1,c_2)^{\mu}$ and the open subscheme gonsisting of $\mu$-stable tector gundles cy $M_H(r;c_1,c_2)^{\mu}_0$. For a torsion frex shwaf $E$ on $X$, we set $\mu(E)=\frac{c_1(E)}{{\uperatornwme{rk}}(E)} \ib H^2(X,\Ubb Q)$ and $\Delta(E)=\hdac{1}{{\operatornajc{rk}}(E)}(c_2(Z)-\fcac{{\operatorname{tk}}(E)-1}{2{\operatorndme{rk}}(E)}c_1(E)^2) \in H^4(X,\Btb Q)$. For a $x \in H^2(X,\Bbb Q)$, we set $P(x)=(x,x-K_X)/2+\chy(\cal O_X)$. Flr a scheme $S$, re dtnoee tgv krojection $X \times S \to S$ by $p_S$. In this stction, we shall gemeralize the chamber struchure of polarizations ln \[Q2\]. For a jkrsyin free sheaw $E$, we set $\gamma(E):=({\operajorname{rk}}(E),\mu(E),\Delta(E)) \in H^0(X,\Bbb Q) \tioes H^2(R,\Bbb Q) \timew Y^4(X,\Bht Q)$. For $\gamna \in \prod_{i=0}^2 H^{2i}(X, \Bnn Q)$, led $M_H^{\gamka}$ be the set pf vorsuon free sheaves $E$ dehined over $\Bbb C$ with $\gamma(E)=\gakmc$ which is $\mu$-semi-staboe with res[ect ro $F$. \[lej:1\] Kef $E$ be a vorsion fres sheaf whixh is defined by an evnrnsion $ 0 \to R_1 \to E \tj F_2 \to 0$. Then $\Delta(E)=\frac{{\operatorname{rk}}(F_1)}{{\okeratkrname{rk}}(E)}\Delta(F_1)+\frac{{\operqtorname{rk}}(F_2)}{{\operatornale{rk}}(E)}\Deltw(F_2)- \frac{{\operatorname{rk}}(F_1){\operatorname{rk}}(F_2)}{2{\operatorname{rn}}(E)^2}((\mu(F_1)-\jj(F_2))^2)$. Ltt $B$ be

1,c_2)^{\mu}$ and the open subscheme consisting of bundles $M_H(r;c_1,c_2)^{\mu}_0$. For torsion free sheaf $\mu(E)=\frac{c_1(E)}{{\operatorname{rk}}(E)} H^2(X,\Bbb Q)$ and \in H^4(X,\Bbb Q)$. a $x \in H^2(X,\Bbb Q)$, we $P(x)=(x,x-K_X)/2+\chi(\cal O_X)$. For a scheme $S$, we denote the projection $X \times S S$ by $p_S$. In this section, we shall generalize the chamber structure of in For torsion sheaf $E$, we set $\gamma(E):=({\operatorname{rk}}(E),\mu(E),\Delta(E)) \in H^0(X,\Bbb Q) \times H^2(X,\Bbb Q) \times H^4(X,\Bbb Q)$. For $\gamma \prod_{i=0}^2 H^{2i}(X, \Bbb Q)$, let $M_H^{\gamma}$ be the of torsion free sheaves defined over $\Bbb C$ with which $\mu$-semi-stable with to \[lem:1\] $E$ be a free sheaf which is defined by an extension $ 0 \to F_1 \to E \to F_2 \to Then $\Delta(E)=\frac{{\operatorname{rk}}(F_1)}{{\operatorname{rk}}(E)}\Delta(F_1)+\frac{{\operatorname{rk}}(F_2)}{{\operatorname{rk}}(E)}\Delta(F_2)- $B$ be

1,c_2)^{\mu}$ and the open subscheme conSisting of $\mU$-stabLe vEctOr BundLes bY $M_H(r;c_1,c_2)^{\mu}_0$. For a toRSion Free sheaf $E$ on $X$, we set $\mu(E)=\fRac{c_1(E)}{{\OpERatoRNaMe{rk}}(E)} \In H^2(X,\Bbb q)$ AnD $\dEltA(E)=\FrAc{1}{{\oPeRAtOrnamE{rk}}(e)}(c_2(E)-\frac{{\OperatornaMe{rK}}(E)-1}{2{\Operatorname{RK}}(E)}C_1(E)^2) \in H^4(X,\Bbb Q)$. for A $x \in H^2(X,\Bbb Q)$, we Set $p(x)=(x,x-K_X)/2+\ChI(\caL o_X)$. For A scHeme $S$, We denoTE the prOjection $X \TiMEs S \to S$ BY $p_S$. In thIS SeCtioN, we shall generalizE ThE Chamber structuRe of poLaRIzATIonS in \[q2\]. For a torsiOn Free sHEaf $E$, we sET $\gAMMA(E):=({\oPEratorname{rk}}(E),\Mu(E),\Delta(E)) \in h^0(x,\BbB Q) \timeS H^2(x,\BbB q) \times h^4(X,\Bbb q)$. FOR $\gaMma \in \prod_{i=0}^2 H^{2I}(X, \BbB Q)$, let $M_H^{\gaMma}$ be tHE set of tORsion frEe sheaVes $e$ deFineD OvEr $\bbb c$ wITh $\gAMmA(E)=\gAMma$ Which is $\mU$-sEmI-stabLe wiTH RESpecT to $h$. \[lem:1\] let $E$ bE a torsion free SheAf whICh iS defiNed by An exTeNsion $ 0 \To F_1 \to E \To F_2 \to 0$. thEn $\Delta(E)=\frac{{\opeRatoRname{rk}}(F_1)}{{\oPerAtOrnAmE{rk}}(E)}\DELta(F_1)+\frAc{{\oPerAtornamE{rk}}(F_2)}{{\opeRAtoRnAME{Rk}}(e)}\Delta(F_2)- \frac{{\operatoRnAME{rK}}(F_1){\operatOrname{RK}}(F_2)}{2{\OpERatornamE{rK}}(E)^2}((\mU(F_1)-\mu(f_2))^2)$. lEt $B$ be

1,c_2)^{\mu}$ and the open subscheme cons ist ing o f $\ mu$- stable vectorb undl es by $M_H(r;c_1,c_2)^ {\mu} _0 $ . Fo r a tors ion fre e s h e af$E $on$X $ ,we se t $ \mu(E)= \frac{c_1( E)} {{ \operatornam e {r k}}(E)} \i n H ^2(X,\Bbb Q) $ a nd $\D el ta( E )=\fr ac{ 1}{{\ operat o rname{ rk}}(E)}( c_ 2 (E)-\f r ac{{\op e r at orna me{rk}}(E)-1}{2{\ o pe r atorname{rk}}( E)}c_1 (E ) ^2 ) \in H^ 4(X,\Bbb Q )$ . For a $x \i n H ^ 2 ( X,\ B bb Q)$, we se t $P(x)=(x, x -K_ X)/2+\ ch i(\ c al O_X )$. F or a s cheme $S$,we d enote the proje c tion $X \timesS \toS$by$p_S $ . I n t hi s se c ti on, weshall ge ne ra lizethec h a m berstr uctu re of polarization s i n \[ Q 2\] . For a to rsio nfreesheaf$E$,we set $\gamma(E) :=({ \operator nam e{ rk} }( E),\m u (E),\D elt a(E )) \inH^0(X,\ B bbQ) \ t im es H^2(X,\Bbb Q) \ ti m e sH^4(X,\B bb Q)$ . F or $\gamma\i n \ prod _ { i=0}^ 2 H^ { 2i }(X, \Bb b Q)$, le t$M_H^{\ ga mma}$be th e s et of tors ion fr ee sheav es $E $ defined over$ \Bbb C$ with$ \g a m ma ( E)=\ gam ma$ which i s $\ m u$-s emi- s ta ble withrespe ct to $H$. \[lem:1\] Let $ E$ bea tor sion free she af which i s d efined b y an ex t ension $ 0 \to F_1\to E \toF _2 \to 0 $. Th en $\Del ta(E)=\fr a c {{\opera tor nam e{r k}} ( F _1 )}{{\operator n a me{r k} }(E)}\D elt a(F_1)+ \fr ac{ {\o per at orname{rk }}(F_2)} {{ \o pe ra tor name{ r k}}(E)}\ De lta (F _2) - \fr a c{{\op erato rnam e{ rk } }(F _1){\op e ra t o rnam e{ rk }}(F _2) }{ 2{\op erat o rna me{rk}} (E)^2}((\ mu( F _1)- \m u( F_2))^2 )$. Let $B$be

1,c_2)^{\mu}$ and_the open_subscheme consisting of $\mu$-stable_vector bundles_by_$M_H(r;c_1,c_2)^{\mu}_0$. For_a_torsion free sheaf_$E$ on $X$,_we set $\mu(E)=\frac{c_1(E)}{{\operatorname{rk}}(E)} \in_H^2(X,\Bbb Q)$ and_$\Delta(E)=\frac{1}{{\operatorname{rk}}(E)}(c_2(E)-\frac{{\operatorname{rk}}(E)-1}{2{\operatorname{rk}}(E)}c_1(E)^2)_\in H^4(X,\Bbb Q)$. For a $x \in H^2(X,\Bbb Q)$, we set $P(x)=(x,x-K_X)/2+\chi(\cal O_X)$. For a_scheme_$S$, we_denote_the_projection $X \times S \to_S$ by $p_S$. In this section,_we shall_generalize the chamber structure of polarizations in \[Q2\]._For_a torsion free_sheaf $E$, we set $\gamma(E):=({\operatorname{rk}}(E),\mu(E),\Delta(E)) \in H^0(X,\Bbb Q) \times H^2(X,\Bbb_Q) \times H^4(X,\Bbb Q)$. For $\gamma_\in \prod_{i=0}^2 H^{2i}(X,_\Bbb_Q)$,_let $M_H^{\gamma}$ be the_set of torsion free sheaves $E$_defined over $\Bbb C$ with $\gamma(E)=\gamma$ which_is $\mu$-semi-stable with respect to $H$. \[lem:1\] Let_$E$ be a torsion free sheaf_which is defined by an_extension $_0 \to F_1 \to E_\to F_2 \to_0$. Then_$\Delta(E)=\frac{{\operatorname{rk}}(F_1)}{{\operatorname{rk}}(E)}\Delta(F_1)+\frac{{\operatorname{rk}}(F_2)}{{\operatorname{rk}}(E)}\Delta(F_2)- \frac{{\operatorname{rk}}(F_1){\operatorname{rk}}(F_2)}{2{\operatorname{rk}}(E)^2}((\mu(F_1)-\mu(F_2))^2)$. Let $B$ be

positive correlation between the red line offsets and the quasar Eddington ratio. This suggests that, while the red system tends to represent the ‘classical’ NLR, it is still effected by the radiation pressure since it must originate close enough to the central source where the ionizing flux is sufficient. Additionally, Figures \[fig:strat\]B and \[fig:strat\]C show that there is some evidence for stratification of the red systems (though much less significant than for the blue systems). Implications for Dual AGN at High-Redshift ------------------------------------------ It is possible that some of the sources in our sample may host two SMBHs following a galaxy merger. In this case, the double-peaks may be from two distinct NLRs that each accompanies its own active SMBH, or perhaps two NLR peaks are influenced by the orbital motion of two SMBHs [@Blecha:2012]. So far there are only a handful of known plausible merger remnants hosting two AGN at redshifts comparable to our sample: $z\sim0.709$ [@Gerke2007], $z\sim0.78$ [@Comerford2009a], and $z\sim1.175$ [@Barrows:2012]. Since galaxy mergers were more frequent at higher redshifts, we would like to investigate the dual AGN scenario for the sources in our sample. ### Sources with No Apparent Ionization Stratification The two [\[\]]{}-selected confirmed dual AGN for which we measured double peaks in [\[\]$\lambda3426$]{} and [\[\]$\lambda3869$]{} are consistent with no apparent ionization stratification (Figure \[fig:strat\]). Additionally, there is only one candidate dual AGN identified through the double-peaked profile of [\[\]$\lambda3426$]{} and [\[\]$\lambda3869$]{}, and it is also consistent with no apparent ionization stratification [@Barrows:2012]. In these sources, the evidence for outflowing NLR material is less compelling, and the line-splitting may instead be produced by orbital motion of two AGN about each other. Likewise, the $40\%$ percent of our sources plotted in Figure \[fig:strat\] with no apparent ionization stratification may include cases where the double-peaks are the result of two SMBHs following a galaxy merger. Additionally, we see explicit double [\[\]$\lambda3727$]{} peaks for a subset (11) of our sources which perhaps suggests that the outflow scenario is less likely in these sources since

positive correlation between the red line stolon and the quasar Eddington proportion. This suggests that, while the red arrangement tend to represent the ‘ classical ’ NLR, it is even effect by the radiation press since it must originate close enough to the central reservoir where the ionizing flux is sufficient. Additionally, Figures \[fig: strat\]B and \[fig: strat\]C read that there constitute some evidence for stratification of the red systems (though much less significant than for the blue systems). deduction for Dual AGN at High - Redshift ------------------------------------------ It is possible that some of the sources in our sample distribution may host two SMBHs following a galaxy amalgamation. In this case, the bivalent - peaks may be from two clear-cut NLRs that each accompanies its own active SMBH, or perhaps two NLR peaks are charm by the orbital motion of two SMBHs [ @Blecha:2012 ]. So far there are only a handful of known plausible merger remnants hosting two AGN at redshifts comparable to our sample: $ z\sim0.709 $ [ @Gerke2007 ], $ z\sim0.78 $ [ @Comerford2009a ], and $ z\sim1.175 $ [ @Barrows:2012 ]. Since galaxy mergers were more frequent at higher redshifts, we would like to investigate the dual AGN scenario for the sources in our sample distribution. # # # reference with No Apparent Ionization Stratification The two [ \[\]]{}-selected confirmed double AGN for which we measure double peaks in [ \[\]$\lambda3426 $ ] { }   and [ \[\]$\lambda3869 $ ] { }   are consistent with no apparent ionization stratification (digit \[fig: strat\ ]). Additionally, there constitute only one candidate double AGN identified through the double - peaked profile of [ \[\]$\lambda3426 $ ] { }   and [ \[\]$\lambda3869 $ ] { }, and it is also consistent with no apparent ionization stratification [ @Barrows:2012 ]. In these source, the evidence for outflowing NLR material is less compelling, and the line - splitting may instead be produced by orbital motion of two AGN about each other. similarly, the $ 40\%$ percent of our sources plotted in Figure \[fig: strat\ ] with no apparent ionization stratification may include cases where the bivalent - peaks are the result of two SMBHs following a galaxy merger. Additionally, we see explicit double [ \[\]$\lambda3727 $ ] { }   peaks for a subset (11) of our source which perhaps suggests that the outflow scenario is less likely in these sources since

poditive correlation betwetn the red line oydsets end the quasar Dddington ratio. This suggestd rhat, qhile the red system tdnds to rvpresent rhe ‘rlassical’ NLR, it is still effedbed bv vhe radiation ptessure sinca it must orighnxtz close enough to the central source where yhf ionizing fluv is fuffjbitnt. Additionally, Figures \[fig:strat\]G and \[fpg:strat\]C show thay there is some evidence flr shratification of tje red systgjs (eyough much ldss signifpeant than fkr the blue systems). Implications for Bual AGN at Hugh-Ggdshift ------------------------------------------ It is posspble that somc of tha sourcrs in our sampke kay host two SMBHs folloxing a galaxy merger. In this waae, the double-peakw nay bg frok twu dirtihcv NMRs thwt xach accompznies its oqn active SMBH, or ptrhwix two NLR pezks arq ynfluenced by the orbital motion of two SMGHs [@Blecha:2012]. So far there are only a handful ov known pjausible merger remnants hosting two AGN at redshhfts roopaxqble gi lur sample: $z\sim0.709$ [@Gerke2007], $z\sim0.78$ [@Comerford2009a], and $z\sii1.175$ [@Bsrgows:2012]. Since galaxy mergers wete mpte frequent at highex rsdshifts, we would pike to invewtigate tre dial AGN scenario for the soyrces in our wample. ### Sources witk No Apparenc Ionieation Stratification The two [\[\]]{}-reledted confirled dual ZEN for which we oeaxused double peaks in [\[\]$\lambda3426$]{} wnd [\[\]$\lambde3869$]{} are eonsistevt woth no apparent lonization stratificatioj (Figore \[fic:strat\]). Addltionally, there is only one canvmdate dual AGM hdettified chrougm the double-peahed profile of [\[\]$\lambda3426$]{} cnd [\[\]$\laobda3869$]{}, and in is also consistent with no appasgnt ionizatioi stratifycatuon [@Varrows:2012]. Kn these sourcrs, the evpdtnce for oytflowing NLR matevial ka less compelliuy, qnd the line-splottkng mwy igvtead be progucea bh orbigal motion iw twp AGN about each othar. Ljkewise, the $40\%$ percemt of our wources [lotted in Fibure \[fig:strat\] with no a'parenv ioniaatyon stratification may include cases whfre the double-pqaks are the resblt of two SMBHs following a galaxy mergxr. Additionally, we see wxplicit double [\[\]$\lamyds3727$]{} peaks for e subsqt (11) of ous sources which perhqps suggests that the outflow scenario is levs linely in these sources since

positive correlation between the red line offsets quasar ratio. This that, while the the NLR, it is effected by the pressure since it must originate close to the central source where the ionizing flux is sufficient. Additionally, Figures \[fig:strat\]B \[fig:strat\]C show that there is some evidence for stratification of the red systems much significant for blue systems). Implications for Dual AGN at High-Redshift ------------------------------------------ It is possible that some of the in our sample may host two SMBHs following galaxy merger. In this the double-peaks may be from distinct that each its active or perhaps two peaks are influenced by the orbital motion of two SMBHs [@Blecha:2012]. So far there are only a of known remnants hosting AGN redshifts to our sample: $z\sim0.78$ [@Comerford2009a], and $z\sim1.175$ [@Barrows:2012]. Since more frequent at higher redshifts, we would like investigate the AGN scenario for the sources in sample. ### Sources with No Apparent Ionization Stratification two [\[\]]{}-selected confirmed dual AGN for which we measured double peaks in [\[\]$\lambda3426$]{} and [\[\]$\lambda3869$]{} with no apparent ionization (Figure \[fig:strat\]). Additionally, is one dual identified through double-peaked profile of [\[\]$\lambda3426$]{} and [\[\]$\lambda3869$]{}, and it is also consistent no apparent ionization stratification [@Barrows:2012]. In these sources, the evidence NLR is less compelling, the line-splitting may instead produced orbital motion of two each Likewise, of sources in Figure \[fig:strat\] with apparent ionization stratification may include where the double-peaks are following a galaxy merger. Additionally, we see explicit [\[\]$\lambda3727$]{} peaks for a subset (11) of sources which perhaps suggests that the outflow scenario is less likely in sources since

positive correlation betweeN the red linE offsEts And ThE quaSar EDdington ratio. THIs suGgests that, while the red sYstem TeNDs to REpResenT the ‘claSSiCAL’ NLr, iT iS stIlL EfFecteD by The radiAtion pressUre SiNce it must oriGInAte close enOugH to the centraL soUrce whErE thE IonizIng Flux iS suffiCIent. AdDitionallY, FIGures \[fIG:strat\]B AND \[fIg:stRat\]C show that there IS sOMe evidence for sTratifIcATiON Of tHe rEd systems (tHoUgh muCH less siGNiFICAnt THan for the blue Systems). ImplICatIons foR DUal agN at HiGh-RedShIFt ------------------------------------------ IT is possible That Some of the SourceS In our saMPle may hOst two sMBhs fOlloWInG a GalAxY MerGEr. in tHIs cAse, the doUbLe-Peaks May bE FROM two DisTincT NLRs That each accomPanIes iTS owN actiVe SMBh, or pErHaps tWo NLR pEaks aRe Influenced by the OrbiTal motion Of tWo sMBhs [@blechA:2012]. so far tHerE arE only a hAndful oF KnoWn PLAUsIble merger remnants HoSTInG two AGN aT redshIFtS cOMparable To Our SampLE: $Z\sim0.709$ [@GErke2007], $Z\SiM0.78$ [@ComerfoRd2009a], and $Z\SiM1.175$ [@BArrows:2012]. SInCe galaXy MerGerS were MOre fRequenT at higheR redsHIfts, we would likE To investigate THe DUAl agN scEnaRio for the soUrceS In ouR samPLe. ### souRCes wiTh No APpAReNT Ionization StratifiCaTion ThE two [\[\]]{}-sElected confirMed dual AGN FOR Which we mEasuREd DOuble peaks in [\[\]$\laMbda3426$]{} aNd [\[\]$\lambda3869$]{} arE ConsisteNt witH no apparEnt ionizaTIOn stratiFicAtiOn (FIguRE \[FiG:strat\]). AdditioNALly, tHeRe is onlY onE candidAte DuaL AGn idEnTified thrOugh the dOuBlE-pEaKed ProfiLE of [\[\]$\lambdA3426$]{} aNd [\[\]$\lAmBda3869$]{}, And it IS also cOnsisTent WiTh NO apParent iONiZATion StRaTifiCatIoN [@BarrOws:2012]. IN TheSe sourcEs, the evidEncE For oUtFlOwing NLr material is leSs Compelling, AnD thE line-sPLItting maY instead be produced by orbITal motiOn oF two AgN abOut each otHer. likewiSe, tHE $40\%$ perceNt of ouR sourCeS plOTTed in fIGuRe \[fIg:Strat\] with nO APpaRent iOnIzatIon straTification may incluDE caSes where the doUblE-peaKS ArE thE ReSUlt Of TWo Smbhs following a galAxy merger. ADdITiOnally, we seE ExpLiCit doubLe [\[\]$\lambdA3727$]{} peakS For a subSet (11) of our sOurces whiCh PerhAPS suGgests that The outflOw scenariO Is lesS LiKely iN thEse souRcEs sInce

positive correlation betw een the re d lin e o ffs et s an d th e quasar Eddin g tonratio. This suggests t hat,wh i le t h ered s ystem t e nd s tore pr ese nt th e ‘cl ass ical’ N LR, it issti ll effected by th e radiatio n p ressure sinc e i t must o rig i nateclo se en ough t o the c entral so ur c e wher e the io n i zi ng f lux is sufficient . A d ditionally, Fi gures\[ f ig : s tra t\] B and \[fi g: strat \ ]C show th a t the r e is some evi dence for s t rat ificat io n o f the r ed sy st e ms(though muc h le ss signif icantt han for the blu e syst ems ). Imp l ic at ion sf orD ua l A G N a t High-R ed sh ift - ---- - - - - ---- --- ---- ----- ------------- --- - I t is poss iblethat s ome o f thesourc es in our samplemayhost twoSMB Hs fo ll owing a gala xymer ger. In this c a se, t h e do uble-peaks may befr o m t wo disti nct NL R sth a t each a cc omp anie s its o wn a c ti ve SMBH, or pe r ha ps two NL Rpeaksar e i nfl uence d bythe or bital mo tiono f two SMBHs [@ B lecha:2012].S of a rt here ar e only a ha ndfu l ofknow n p lau s iblemerge rr em n ants hosting two AG Nat red shift s comparableto our sam p l e : $z\sim 0.70 9 $[ @Gerke2007], $ z\sim 0.78$ [@Co m erford20 09a], and $z\ sim1.175$ [ @Barrows :20 12] . S inc e ga laxy mergersw e re m or e frequ ent at hig her re dsh ift s, we would like to i nv es ti gat e the dual AGN s cen ar iofor t h e sour ces i n ou rsa m ple . ###S ou r c es w it hNo A ppa re nt Io niza t ion Strati fication Th e two [ \[ \]]{}-s elected confi rm ed dual AG Nfor which w e measur ed double peaks in [\[\ ] $\lambd a34 26$]{ } an d [\[\]$\ lam bda386 9$] { } areconsis tentwi thn o appa r e nt io ni zation str a t ifi catio n(Fig ure \[f ig:strat\]). Addit i ona lly, there is on ly o n e c and i da t e d ua l AG N identified thro ugh the do ub l e- peaked pro f ile o f [\[\] $\lambd a3426 $ ]{} and [\[\]$\l ambda3869 $] {},a n d i t is alsoconsiste nt with n o appa r en t ion iza tion s tr ati ficat ion [@ B arr ows:2 012].In these sour ce s, the e vidence for outflowingNLR ma teria l i s less co mpe l lin g, and th e li ne-splitti ngmay inst ead be pr oduc e dbyo rbita l mo t ion of tw o A GNa b ou t each othe r . Lik ewise , t h e $40\ %$ p ercent of our sou r ces plotted in Fig u r e \ [fi g :str at \] with no app are nt i onizatio nstratificat ion mayin c ludecaseswherethe dou b l e- p eaks a re t heresult of tw oS MBHs fo ll ow i ng a g alax ymerger . Addi t iona l l y, we see explic it do u b le [\ [ \]$ \lamb da 3727$]{ }  pea ks for a s ubset (11)of our sou rceswhich p er haps s ugg es ts that th e outflowscena rio isle ss l ike ly inthes e sourc es s in ce

positive_correlation between_the red line offsets_and the_quasar_Eddington ratio._This_suggests that, while_the red system_tends to represent the_‘classical’ NLR, it_is_still effected by the radiation pressure since it must originate close enough to the_central_source where_the_ionizing_flux is sufficient. Additionally, Figures_\[fig:strat\]B and \[fig:strat\]C show that_there is_some evidence for stratification of the red systems_(though_much less significant_than for the blue systems). Implications for Dual AGN at_High-Redshift ------------------------------------------ It is possible that some of_the sources in_our_sample_may host two SMBHs_following a galaxy merger. In this_case, the double-peaks may be from_two distinct NLRs that each accompanies its_own active SMBH, or perhaps two_NLR peaks are influenced by_the orbital_motion of two SMBHs [@Blecha:2012]._So far there_are only_a handful of_known plausible merger remnants hosting two_AGN at redshifts_comparable to our sample: $z\sim0.709$ [@Gerke2007],_$z\sim0.78$_[@Comerford2009a], and $z\sim1.175$_[@Barrows:2012]._Since_galaxy mergers_were more frequent_at_higher redshifts,_we_would like to investigate the dual_AGN_scenario for the sources in our sample. ###_Sources with No Apparent_Ionization_Stratification The two [\[\]]{}-selected confirmed_dual AGN for which we_measured double peaks in [\[\]$\lambda3426$]{} and [\[\]$\lambda3869$]{} are_consistent with_no apparent_ionization stratification (Figure \[fig:strat\]). Additionally, there is only one candidate dual_AGN identified through the double-peaked profile_of [\[\]$\lambda3426$]{} and [\[\]$\lambda3869$]{}, and_it is_also_consistent with no_apparent_ionization stratification_[@Barrows:2012]. In these sources, the evidence for_outflowing NLR_material is less compelling, and the_line-splitting may instead be_produced_by orbital motion of two AGN_about each other. Likewise, the $40\%$_percent of our sources plotted_in_Figure_\[fig:strat\] with no apparent ionization_stratification may include cases where the_double-peaks are the_result of two SMBHs following a galaxy_merger._Additionally, we see explicit double [\[\]$\lambda3727$]{} peaks_for_a subset (11) of our sources_which_perhaps_suggests that the outflow scenario_is less likely in these sources_since

]. Essentially, compartments represent uniform spatial distributions or specific chemical compounds of the basic radioactive molecules; radioactivity concentrations in the various compartments are the natural state variables of the system; tracer flow, resulting from interchange of radioactive molecules between compartments, is modeled by a Cauchy problem for a system of linear ordinary differential equations (ODEs) for concentrations; the constant coefficients, also called rate constants or kinetic parameters, represent tracer kinetics and may be related to the action of enzymes, such as hexokinase (HK) responsible for phosphorylation in cells. In typical compartmental problems the rate coefficients are unknown. The measured data are the total amount of tracer (concentration or activity) in a given region of interest, and the input function (IF), describing the time rate of tracer carried into the system. Tracer kinetics results from the solution of the inverse problem of determining the unknown rate coefficients compatible with the data, and the subsequent explicit determination of the concentration (or activity) of each compartment through the solution of the system of ODEs. In the applications presented in this paper, the inverse problem is solved in two steps: first, a formal expression of the solution of the direct Cauchy problem is evaluated, where dependence on the unknown rate constants is made explicit; second, an inversion algorithm is applied in order to recover the kinetic parameters through comparison of the formal solution with the data. The inversion algorithm makes use of an optimization regularization method which is based on a Newton-type algorithm. “Classical” compartmental models have been developed under the assumption that phosphorylation and dephosphorylation of FDG occur in the same intracellular cytosolic volume, as described by [@Sokoloff] and [@Wernick]. Recent progresses in cell biochemistry have shown that the appropriate location of dephosphorylation is the endoplasmic reticulum (ER) [@Ghosh], which is spatially separated from cytosol. The schematic path of FDG kinetics is illustrated in  \[fig:FDG\_kinetics\]. Following this pattern for FDG kinetics in the cell, here we describe and formalize a new model consisting of three compartments which account for free FDG in cytosol, phosphorylated FDG in cytosol, and phosphorylated FDG in ER. The new biochemically-driven compartmental model is referred to as BCM; a classical simplified compartmental model (SCM) is recovered from the proposed model under the assumption that the ER is removed from consideration. In the forthcoming paper (Scussolini et al. manuscript in preparation), as a first test of its

]. Essentially, compartments represent uniform spatial distribution or specific chemical compound of the basic radioactive molecules; radioactivity concentrations in the respective compartments are the natural state of matter variables of the system; tracer stream, resulting from interchange of radioactive atom between compartments, is modeled by a Cauchy trouble for a system of linear ordinary differential equations (ODEs) for concentrations; the constant coefficient, also called rate constant or kinetic parameters, represent tracer dynamics and may be related to the natural process of enzymes, such as hexokinase (HK) creditworthy for phosphorylation in cells. In typical compartmental problems the rate coefficient are unknown. The measured data are the total amount of tracer (concentration or activity) in a given region of interest, and the input function (IF), describing the time rate of tracer carried into the system. Tracer dynamics result from the solution of the inverse problem of determine the strange rate coefficients compatible with the data, and the subsequent denotative determination of the assiduity (or activity) of each compartment through the solution of the system of ODEs. In the application presented in this paper, the inverse problem is solved in two steps: first, a formal expression of the solution of the direct Cauchy problem is evaluated, where addiction on the unknown pace constants is made explicit; second, an inversion algorithm is applied in order to recuperate the kinetic parameters through comparison of the formal solution with the datum. The inversion algorithm make use of an optimization regularization method acting which is based on a Newton - type algorithm. “ Classical ” compartmental models have been developed under the premise that phosphorylation and dephosphorylation of FDG occur in the like intracellular cytosolic volume, as described by [ @Sokoloff ] and [ @Wernick ]. Recent progresses in cellular telephone biochemistry have shown that the appropriate location of dephosphorylation is the endoplasmic reticulum (ER) [ @Ghosh ], which is spatially separated from cytosol. The conventional path of FDG kinetics is illustrated in   \[fig: FDG\_kinetics\ ]. Following this pattern for FDG kinetics in the cell, here we describe and formalize a new model consisting of three compartment which account for free FDG in cytosol, phosphorylated FDG in cytosol, and phosphorylated FDG in ER. The modern biochemically - driven compartmental model is refer to as BCM; a classical simplified compartmental model (SCM) is recovered from the aim model under the assumption that the ER is removed from consideration. In the forthcoming paper (Scussolini et al. manuscript in preparation), as a inaugural trial of its

]. Esdentially, compartments rtpresent uniform spatial distrjbutions or specific chemical compouids if tht basic radioactive moleculed; radioaxtivmty concentratioia in thc varjlus eonpartments are the naturdl state variatlds of the system; tracer flow, resultind from onherchange of rwdiosstivs molecules between compartments, ia modeltd by a Cauchy pronlem for a system of lineag orfinary differentiap equations (ODEf) for concentfations; tht eonstant cogfficients, also called rate consgants or kinetix parwketers, reprxsent nracer kinetigx and kay be telated to the acvion of enzymes, such as hxxokinase (HK) responsyble for [hksphorylation in xeols. In typhcal xomoarumeitam probpema the rate coefficienrs are unknown. The kewwured data ars the eoeal amount of tracer (concentration or abtivjty) in a given region od interest, and the inkut functijn (IF), describing the time rate of tracer carried hnto vhd svwtem. Geafer kinetics results from the solution of the jnfegse problem of debermining the unknpwj twte coefficievts compafible with the datw, and tre suvsequent txplivit determination of the cobcentration (je activity) of each compartmenc throogh thr solution of the systeo of ODEs. In thf applicafkons presented iv tmis paper, the inverse problem is solvev in cwo stepr: fitst, a fjrmal exprfssiok of the solution ov the dhrect Caucjy problem is evaluated, where dx'endence on tne unnnown race conxtants is madq explicit; secpnd, an nnverskon algorinhm is ap'lied in ordqr to recover jhe kinetic perameters thriugh comparkron of the forkal solutpou with thw data. The inversipn xmgorithm makes brt if an optimizayiov rqgllacizatymn method whhch ks cssed un a Newton-bypd alborithm. “Classical” com[artjental models have bcen develiped undqr the assumpyion that phosphorjlatimn end delhofphorylation of FDG occur in tge same ijtrwcellular cyeosooic volume, ax described by [@Sokoloff] and [@Wernick]. Recxnt progresses in cell viochemistry have skoen that the apprjpriate lmcation of dephosphoeylation is the ekdoplasmic reticulum (ED) [@Ghosv], whifh is spatially separated from cytosol. The schematic path of FDG kinetics is illustrated in  \[fig:GDG\_kitecicx\]. Folloring vhis pattern for GDG kinetics in the cell, here wx describe aud formalize a new model consosging of three compartments which accohnt for gree FDG in cytosol, phosphorylayed FDG in cytosol, and pmosphorylqted RDG in ER. Vhe new biochemicallu-driveu compqrtmentql mldel is referfed tl as BCM; w clessncal simplified compartmental modwl (SCM) id recovered from the proposcd modem undet the assumption fhat hhe ET is removed ftom consideratoon. In the forthcomitg paper (Scussoluhi et al. manoscripj in preparatioi), as a first test of its

]. Essentially, compartments represent uniform spatial distributions chemical of the radioactive molecules; radioactivity are natural state variables the system; tracer resulting from interchange of radioactive molecules compartments, is modeled by a Cauchy problem for a system of linear ordinary equations (ODEs) for concentrations; the constant coefficients, also called rate constants or kinetic represent kinetics may related to the action of enzymes, such as hexokinase (HK) responsible for phosphorylation in cells. In compartmental problems the rate coefficients are unknown. The data are the total of tracer (concentration or activity) a region of and input (IF), describing the rate of tracer carried into the system. Tracer kinetics results from the solution of the inverse problem determining the coefficients compatible the and subsequent explicit determination concentration (or activity) of each compartment of the system of ODEs. In the applications in this the inverse problem is solved in steps: first, a formal expression of the solution the direct Cauchy problem is evaluated, where dependence on the unknown rate constants is made an inversion algorithm is in order to the parameters comparison the formal with the data. The inversion algorithm makes use of an optimization method which is based on a Newton-type algorithm. “Classical” compartmental been under the assumption phosphorylation and dephosphorylation of occur the same intracellular cytosolic described [@Sokoloff] progresses cell have shown that the location of dephosphorylation is the reticulum (ER) [@Ghosh], which The schematic path of FDG kinetics is illustrated \[fig:FDG\_kinetics\]. Following this pattern for FDG kinetics the cell, here we describe and formalize a new model consisting of compartments which free FDG in cytosol, phosphorylated FDG in cytosol, phosphorylated FDG in ER. new biochemically-driven compartmental model is referred to as BCM; classical compartmental model is recovered from proposed model under assumption that the removed from In forthcoming al. manuscript in preparation), as a test its

]. Essentially, compartments rePresent uniForm sPatIal DiStriButiOns or specific cHEmicAl compounds of the basic rAdioaCtIVe moLEcUles; rAdioactIViTY ConCeNtRatIoNS iN the vAriOus compArtments arE thE nAtural state vARiAbles of the SysTem; tracer floW, reSultinG fRom INtercHanGe of rAdioacTIve molEcules betWeEN compaRTments, iS MOdEled By a Cauchy problem fOR a SYstem of linear oRdinarY dIFfEREntIal Equations (OdES) for cONcentraTIoNS; THe cONstant coefficIents, also caLLed Rate coNsTanTS or kinEtic pArAMetErs, represenT traCer kinetiCs and mAY be relaTEd to the Action Of eNzyMes, sUCh As HexOkINasE (hK) ResPOnsIble for pHoSpHorylAtioN IN CElls. in tYpicAl comPartmental proBleMs thE RatE coefFicieNts aRe UnknoWn. The mEasurEd Data are the total AmouNt of traceR (coNcEntRaTion oR ActiviTy) iN a gIven regIon of inTEreSt, AND ThE input function (IF), deScRIBiNg the timE rate oF TrAcER carried InTo tHe sySTEm. TraCer kINeTics resuLts froM ThE sOlution Of The invErSe pRobLem of DEterMining The unknoWn ratE Coefficients coMPatible with thE DaTA, AnD The sUbsEquent expliCit dETermInatIOn Of tHE concEntraTiON (oR Activity) of each compaRtMent thRough The solution of The system oF odes. In the aPpliCAtIOns presented in This pAper, the invERse problEm is sOlved in tWo steps: fiRST, a formal ExpResSioN of THE sOlution of the dIREct CAuChy probLem Is evaluAteD, whEre DepEnDence on thE unknown RaTe CoNsTanTs is mADe explicIt; SecOnD, an InverSIon algOrithM is aPpLiED in Order to REcOVEr thE kInEtic ParAmEters ThroUGh cOmparisOn of the foRmaL SoluTiOn With the Data. The inversIoN algorithm MaKes Use of aN OPtimizatIon regularization method WHich is bAseD on a NEwtoN-type algoRitHm. “ClasSicAL” compaRtmentAl modElS haVE Been dEVElOpeD uNder the assUMPtiOn thaT pHospHorylatIon and dephosphorylATioN of FDG occur in The Same INTrAceLLuLAr cYtOSolIC Volume, as describEd by [@SokoloFf] ANd [@wernick]. RecENt pRoGresses In cell bIocheMIstry haVe shown thAt the apprOpRiatE LOcaTion of dephOsphorylAtion is thE EndopLAsMic reTicUlum (ER) [@ghOsh], Which Is spatIAllY sepaRated fRoM cytosOl. The ScHematic pAth of FDG kinetics is illusTrated In  \[fig:fDG\_Kinetics\]. FOllOWinG this pattErn fOr FDG kinetIcs In tHe celL, heRE we deScriBE aNd fORmaliZe a nEW model conSIsTinG OF tHree compartMENTs wHich aCcoUNt for fRee FdG in cytosol, phosphORylated FDG in cyTosoL, ANd pHosPHoryLaTed FDG in ER. The nEw bIoCHEmically-DrIven compartMental moDeL Is refErred tO as BCM; A classiCAL sIMplifiEd coMpaRtmental mOdeL (Scm) is recoVeReD From thE proPoSed modEl undeR The aSSUmption that the ER Is remOVEd froM ConSiderAtIon. In thE FortHcoming papEr (ScussolinI et al. mAnusCript In prepaRaTion), as A fiRsT test of its

]. Essentially, compartmen ts represe nt un ifo rmsp atia l di stributions or spec ific chemical compound s ofth e bas i cradio activem ol e c ule s; r adi oa c ti vitycon centrat ions in th e v ar ious compart m en ts are the na tural statevar iables o f t h e sys tem ; tra cer fl o w, res ulting fr om interc h ange of r ad ioac tive molecules be t we e n compartments , is m od e le d bya C auchy prob le m for a syste m o f l ine a r ordinary di fferentiale qua tions(O DEs ) for c oncen tr a tio ns; the con stan t coeffic ients, also ca l led rat e cons tan tsor k i ne ti c p ar a met e rs , r e pre sent tra ce rkinet icsa n d mayberela ted t o the actionofenzy m es, such as h exok in ase ( HK) re spons ib le for phosphor ylat ion in ce lls . In t ypica l compa rtm ent al prob lems th e ra te c o ef ficients are unkno wn . Th e measur ed dat a a re the tota lamo unto f trac er ( c on centrati on ora ct iv ity) in a given r egi onof in t eres t, and the inp ut fu n ction (IF), de s cribing the t i me r at e oftra cer carried int o the sys t em . T r acerkinet ic s r e sults from the solu ti on ofthe i nverse proble m of deter m i n ing theunkn o wn rate coefficie nts c ompatiblew ith thedata, and the subseque n t explici t d ete rmi nat i o nof the concen t r atio n(or act ivi ty) ofeac h c omp art me nt throug h the so lu ti on o f t he sy s tem of O DE s.In th e app l icatio ns pr esen te di n t his pap e r, t he i nv er se p rob le m issolv e d i n two s teps: fir st, a fo rm al expres sion of the s ol ution of t he di rect C a u chy prob lem is evaluated, where depende nce on t he u nknown ra teconsta nts is mad e expl icit; s eco n d , ani n ve rsi on algorithm i s a pplie din o rder to recover the kinet i c p arameters thr oug h co m p ar iso n o f th ef orm a l solution withthe data.Th e i nversion a l gor it hm make s use o f ano ptimiza tion regu larizatio nmeth o d wh ich is bas ed on aNewton-ty p e alg o ri thm. “C lassic al ” c ompar tmenta l mo delshave b ee n deve loped u nder the assumption that phosph orylat ion a nddephospho ryl a tio n of FDGoccu r in the s ame in trace llu l ar cy toso l ic vo l ume,as d e scribed b y [ @So k o lo ff] and [@W e r n ick ]. Re cen t progr esse s in cell biochem i stry have show n th a t th e a p prop ri ate location o f d ep h o sphoryla ti on is the e ndoplasm ic retic ulum ( ER) [@ Ghosh], w hi c h is s pati all y separat edfr o m cytos ol .T he sch emat ic pathof FDG kine t i cs is illustrate d in\ [fig: F DG\ _kine ti cs\]. F ollo wing thispattern for FDG k inet ics i n the c el l, her e w edescribe a n d formali ze anew mod el con sis ting o f th r e e com part me nts which ac c o un t f or fre e FD G incy toso l, phosph o rylatedFDG in cyto so l,a n d phos p ho r y lated FDGinER. T h e new bioch e mica l ly - drive n comp artmen tal mod e l i sreferre d t o as BCM; a classica l si mp lifi ed compa rt ment al mo de l (SCM) is reco vered fr om the p r o posedmod e l un d er the a ssumption th at th eE R is re m o ved from c o nsi der ati on.Inthef orthcoming paper (Scusso lini et a l . m anusc riptin prepara t ion) , as a f i rst test of i ts

]. Essentially,_compartments represent_uniform spatial distributions or_specific chemical_compounds_of the_basic_radioactive molecules; radioactivity_concentrations in the_various compartments are the_natural state variables_of_the system; tracer flow, resulting from interchange of radioactive molecules between compartments, is modeled_by_a Cauchy_problem_for_a system of linear ordinary_differential equations (ODEs) for concentrations;_the constant_coefficients, also called rate constants or kinetic parameters,_represent_tracer kinetics and_may be related to the action of enzymes, such_as hexokinase (HK) responsible for phosphorylation_in cells. In typical_compartmental_problems_the rate coefficients are_unknown. The measured data are the_total amount of tracer (concentration or_activity) in a given region of interest,_and the input function (IF), describing_the time rate of tracer_carried into_the system. Tracer kinetics results_from the solution_of the_inverse problem of_determining the unknown rate coefficients compatible_with the data,_and the subsequent explicit determination of_the_concentration (or activity)_of_each_compartment through_the solution of_the_system of_ODEs._In the applications presented in this_paper,_the inverse problem is solved in two_steps: first, a formal_expression_of the solution of_the direct Cauchy problem is_evaluated, where dependence on the unknown_rate constants_is made_explicit; second, an inversion algorithm is applied in order to recover_the kinetic parameters through comparison of_the formal solution with_the data._The_inversion algorithm makes_use_of an_optimization regularization method which is based on_a Newton-type_algorithm. “Classical” compartmental models have been developed_under the assumption that_phosphorylation_and dephosphorylation of FDG occur in_the same intracellular cytosolic volume, as_described by [@Sokoloff] and [@Wernick]._Recent_progresses_in cell biochemistry have shown_that the appropriate location of dephosphorylation_is the endoplasmic_reticulum (ER) [@Ghosh], which is spatially separated_from_cytosol. The schematic path of FDG_kinetics_is illustrated in  \[fig:FDG\_kinetics\]. Following this pattern_for_FDG_kinetics in the cell, here_we describe and formalize a new_model consisting of three compartments which account for free_FDG in cytosol,_phosphorylated FDG in cytosol, and_phosphorylated_FDG_in ER. The new biochemically-driven compartmental model is referred to_as BCM;_a classical simplified_compartmental model (SCM) is recovered from the proposed model under_the assumption that the ER is removed_from consideration. In the forthcoming paper (Scussolini et al. manuscript in_preparation), as a first test of its

. We derive the value of $V/V_{\rm max}$ of an individual GRB by a simulation. The simulation is best visualized as an excercise in which the distance of the burst is incrementally increased in Euclidean space until it becomes undetectable. The full time profile of the burst is reduced by a factor corresponding to the increased distance and then added back to the original interpolated background (see Fig. 1). The detection algorithm is employed to search for the reduced burst. If it is detected, the process - including the full search - is repeated until the burst is not detected anymore. If the burst is lost when the distance has been increased by a factor $f$ then $V/V_{\rm max} = f^{-3}$. The 2204 GRBs in the GUSBAD catalog have $<V/V_{\rm max}> = 0.346 \pm 0.006$. In the process of analyzing the reduced burst profiles, the trigger may occur at later times depending on the detailed profile. (If this causes the second background stretch to suffer from contaminated or missing data, it is kept at its location during the original detection.) If so, the two background stretches defined in Fig. 1 move forward and the first background stretch may contain burst signal. This reduces the amplitude of the reduced burst and increases the background. Hence $V/V_{\rm max}$ derived from the simulation is larger than $(C_{\rm max}/C_{\rm min})^{-3/2}$. This is confirmed in the GUSBAD catalog which yields $<(C_{\rm max}/C_{\rm min})^{-3/2}> = 0.335$. In the BATSE catalog, only 884 of the 1144 GRBs with $C_{\rm max}/C_{\rm min} \ge 1.0$ on a timescale of 1024 ms can be used for a comparison: the others fall at times when the on-board trigger used parameters other than the standard mentioned in Sec. 1. They yield $<(C_{\rm max}/C_{\rm min})^{-3/2}> = 0.311 \pm 0.009$. The GUSBAD Catalog ================== The GUSBAD catalog is available on the World Wide Web [@sch03][^3]. The designation of GRBs in the catalog is GUSBAD YYMMDD.ddd where Y, M, D are integral year, month and day numbers, respectively, and.ddd is the truncated fraction of the day. The

. We derive the value of $ V / V_{\rm max}$ of an individual GRB by a simulation. The simulation is well visualize as an excercise in which the distance of the burst is incrementally increased in Euclidean distance until it becomes undetectable. The entire time profile of the explosion is reduced by a factor represent to the increased distance and then added back to the original interpolate setting (see Fig. 1). The detection algorithm is employed to search for the reduced explosion. If it is detected, the process - including the wide search - is repeated until the burst is not detected anymore. If the explosion is lost when the distance has been increased by a factor $ f$ then $ V / V_{\rm max } = f^{-3}$. The 2204 GRBs in the GUSBAD catalog have $ < V / V_{\rm max } > = 0.346 \pm 0.006$. In the procedure of analyzing the reduced burst profiles, the trigger may occur at later times depending on the detailed profile. (If this causes the second background stretch to suffer from contaminated or missing data, it is kept at its location during the original signal detection .) If thus, the two background stretch defined in Fig. 1 move forward and the first setting stretch may contain burst signal. This reduces the amplitude of the reduced outburst and increases the background. Hence $ V / V_{\rm max}$ derived from the simulation is big than $ (C_{\rm max}/C_{\rm min})^{-3/2}$. This is confirmed in the GUSBAD catalog which yields $ < (C_{\rm max}/C_{\rm min})^{-3/2 } > = 0.335$. In the BATSE catalog, only 884 of the 1144 GRBs with $ C_{\rm max}/C_{\rm min } \ge 1.0 $ on a timescale of 1024 m can be practice for a comparison: the others fall at times when the on - board trigger used argument other than the standard mentioned in Sec. 1. They concede $ < (C_{\rm max}/C_{\rm min})^{-3/2 } > = 0.311 \pm 0.009$. The GUSBAD Catalog = = = = = = = = = = = = = = = = = = The GUSBAD catalog is available on the World Wide Web [ @sch03][^3 ]. The designation of GRBs in the catalog is GUSBAD YYMMDD.ddd where Y, M, D are integral year, calendar month and day numbers, respectively, and.ddd is the truncated fraction of the day. The

. We ferive the value of $V/V_{\rm max}$ of an indirudual JRB by z simulagion. The simulation is best tisuqlizee as an excercise in wfich the fistance of uhe burst is incrxjentalln incdcased mn Euclidean spsce until ht becomes undatdccable. The full time profile of the btrst is rfduced by a fastor sorrssponding to the increased distancs and tien added back yo the original interpolatfd bwckground (see Fig. 1). The detectuon wogorithm is dmployed to search for the reduced burst. If it is detdcted, the procews - ijwluding the full search - is vvpeated until yhe burst is npt vetexted anymore. If the bnrst is lost when thg distance hcs been increased by q dactot $f$ tven $X/C_{\rm mas} = f^{-3}$. The 2204 HRBa in the GHSBAD catalig have $<V/V_{\rm max}> = 0.346 \km 0.006$. Yb the process of anwlrzing the reduced burst profiles, the trpgged may occur at later tines depending on the fetailed [rofile. (If this causes the second background stredch tk suyncr ffim contaminated or missing data, it is kept at iea kobation during the original drtfcyyon.) If so, the two bcdkfround stretches dffined yn Fit. 1 move fjrwatd and the first background stretch may xontain burst signcl. This reduees thg amplotude of the reduced buxst ans increases the backffound. Hence $V/V_{\rm maq}$ desived from the simulation ys larger thau $(C_{\rm mab}/C_{\rm min})^{-3/2}$. Tris is convirmed in the GUSBAD catwlog chich yields $<(C_{\rl max}/C_{\rm min})^{-3/2}> = 0.335$. In the BATSE catekog, only 884 of tve 1144 GRBs wich $C_{\rm max}/C_{\rm min} \gq 1.0$ on a timesccle of 1024 is cav be used ror a cmmparison: tre others faln at times whxn the on-foare trugger urdd parameters pther thau the standard mentionev in Aec. 1. They yield $<(C_{\rn max}/C_{\rm min})^{-3/2}> = 0.311 \lm 0.009$. Ghe GLSBED Caedlog ================== The GUSBDD cxtaupg is available iv thr World Wide Web [@sch03][^3]. The designation of GRNs in the xatalog ys GUSBAD YYMKDD.ddd where Y, M, D are mntegrel yeat, mjnth and day numbers, respectivsly, and.ddf if the truncaeed nracjion of the day. The

. We derive the value of $V/V_{\rm an GRB by simulation. The simulation excercise which the distance the burst is increased in Euclidean space until it undetectable. The full time profile of the burst is reduced by a factor to the increased distance and then added back to the original interpolated background Fig. The algorithm employed to search for the reduced burst. If it is detected, the process - including the search - is repeated until the burst is detected anymore. If the is lost when the distance been by a $f$ $V/V_{\rm = f^{-3}$. The GRBs in the GUSBAD catalog have $<V/V_{\rm max}> = 0.346 \pm 0.006$. In the process of analyzing reduced burst trigger may at times on the detailed this causes the second background stretch contaminated or missing data, it is kept at location during original detection.) If so, the two stretches defined in Fig. 1 move forward and first background stretch may contain burst signal. This reduces the amplitude of the reduced burst the background. Hence $V/V_{\rm derived from the is than max}/C_{\rm This is in the GUSBAD catalog which yields $<(C_{\rm max}/C_{\rm min})^{-3/2}> = 0.335$. the BATSE catalog, only 884 of the 1144 GRBs with min} 1.0$ on a of 1024 ms can used a comparison: the others times the parameters than standard mentioned in Sec. They yield $<(C_{\rm max}/C_{\rm min})^{-3/2}> 0.311 \pm 0.009$. The catalog is available on the World Wide Web The designation of GRBs in the catalog GUSBAD YYMMDD.ddd where Y, M, D are integral year, month and day respectively, and.ddd truncated fraction of the day. The

. We derive the value of $V/V_{\rm max}$ Of an indiviDual GrB bY a sImUlatIon. THe simulation is BEst vIsualized as an excercise In whiCh THe diSTaNce of The bursT Is INCreMeNtAllY iNCrEased In EUclideaN space untiL it BeComes undetecTAbLe. The full tIme Profile of the BurSt is reDuCed BY a facTor CorreSpondiNG to the Increased DiSTance aND then adDED bAck tO the original interPOlATed background (sEe Fig. 1). THe DEtECTioN alGorithm is eMpLoyed TO search FOr THE RedUCed burst. If it iS detected, thE ProCess - inClUdiNG the fuLl seaRcH - Is rEpeated untiL the Burst is noT detecTEd anymoRE. If the bUrst is LosT whEn thE DiStAncE hAS beEN iNcrEAseD by a factOr $F$ tHen $V/V_{\Rm maX} = F^{-3}$. tHE 2204 GRBS in The GuSBAD Catalog have $<V/V_{\Rm mAx}> = 0.346 \pm 0.006$. iN thE procEss of AnalYzIng thE reducEd burSt Profiles, the trigGer mAy occur at LatEr TimEs DepenDIng on tHe dEtaIled proFile. (If tHIs cAuSES ThE second background sTrETCh To suffer From coNTaMiNAted or miSsIng Data, IT Is kepT at iTS lOcation dUring tHE oRiGinal deTeCtion.) IF sO, thE twO backGRounD stretChes defiNed in fIg. 1 move forward aND the first backGRoUND sTRetcH maY contain burSt siGNal. THis rEDuCes THe ampLitudE oF ThE Reduced burst and incrEaSes the BackgRound. Hence $V/V_{\rM max}$ deriveD FROm the simUlatIOn IS larger than $(C_{\rm Max}/C_{\rM min})^{-3/2}$. This is COnfirmed In the gUSBAD caTalog whicH YIelds $<(C_{\rm Max}/c_{\rm Min})^{-3/2}> = 0.335$. in tHE bAtSE catalog, onlY 884 OF the 1144 gRbs with $C_{\Rm mAx}/C_{\rm miN} \ge 1.0$ On a TimEscAlE of 1024 ms can bE used for A cOmPaRiSon: The otHErs fall aT tImeS wHen The on-BOard trIgger Used PaRaMEteRs other THaN THe stAnDaRd meNtiOnEd in SEc. 1. ThEY yiEld $<(C_{\rm mAx}/C_{\rm min})^{-3/2}> = 0.311 \pM 0.009$. ThE gUSBaD caTalog ================== ThE GUSBAD cataloG iS available On The world WIDE Web [@sch03][^3]. THe designation of GRBs in thE Catalog Is GuSBAD yYMMdD.ddd wherE Y, M, d are inTegRAl year, Month aNd day NuMbeRS, RespeCTIvEly, AnD.ddd is the tRUNcaTed frAcTion Of the daY. The

. We derive the value of$V/V_{\rmmax}$ of an i ndiv idua l GRB by a sim u lati on. The simulation isbestvi s uali z ed as a n excer c is e inwh ic h t he di stanc e o f the b urst is in cre me ntally incre a se d in Eucli dea n space unti l i t beco me s u n detec tab le. T he ful l timeprofile o ft he bur s t is re d u ce d by a factor corresp o nd i ng to the incr easeddi s ta n c e a ndthen added b ack t o the or i gi n a l in t erpolated bac kground (se e Fi g. 1). T hed etecti on al go r ith m is employ ed t o searchfor th e reduce d burst. If it is de tect e d, t hepr o ces s - in c lud ing thefu ll sear ch - i s repe ate d un til t he burst is n otdete c ted anym ore.If t he burs t is l ost w he n the distancehasbeen incr eas ed by a fact o r $f$the n $ V/V_{\r m max}= f^ {- 3 } $ .The 2204 GRBs in t he G US BAD cata log ha v e$< V /V_{\rmma x}> = 0 . 3 46 \p m 0. 0 06 $. In t he pro c es sof anal yz ing th ered uce d bur s t pr ofiles , the tr igger may occur at l a ter times dep e nd i n go n th e d etailed pro file . (If thi s c aus e s the seco nd ba c kground stretch tosu ffer f rom c ontaminated o r missingd a t a, it is kep t a t its locationdurin g the orig i nal dete ction .) If so , the two b ackgroun d s tre tch esd e fi ned in Fig. 1 m ovefo rward a ndthe fir stbac kgr oun dstretch m ay conta in b ur st si gnal. This red uc esth e a mplit u de ofthe r educ ed b u rst and in c re a s es t he b ackg rou nd . Hen ce $ V /V_ {\rm ma x}$ deriv edf romth esimulat ion is larger t han $(C_{\ rm ma x}/C_{ \ r m min})^ {-3/2}$. This is confir m ed in t heGUSBA D ca talog whi chyields $< ( C_{\rm max}/ C_{\r mmin } ) ^{-3/ 2 } >= 0 .3 35$. In th e BAT SE ca ta log, only 8 84 of the 1144 GRB s wi th $C_{\rm ma x}/ C_{\ r m m in} \g e 1. 0$ ona timescale of 10 24 ms canbe us ed for a c o mpa ri son: th e other s fal l at tim es when t he on-boa rd tri g g erused param eters ot her thant he st a nd ard m ent ionedin Se c. 1. Theyy iel d $<( C_{\rm m ax}/C_ {\rmmi n})^{-3/ 2}> = 0.311 \pm 0.009$. TheGUSBA D C atalog == === = === ========= Th e GUSBAD c ata log is a vai l ableon t h eWor l d Wid e We b [@sch03] [ ^3 ].T h edesignation o f GR Bs in th e catal og i s GUSBAD YYMMDD.d d d where Y, M,D ar e int egr a l ye ar , month and da y n um b e rs, resp ec tively, and .ddd isth e trun catedfracti on of t h e d a y. The

. We derive_the value_of $V/V_{\rm max}$ of_an individual_GRB_by a_simulation._The simulation is_best visualized as_an excercise in which_the distance of_the_burst is incrementally increased in Euclidean space until it becomes undetectable. The full time_profile_of the_burst_is_reduced by a factor corresponding_to the increased distance and_then added_back to the original interpolated background (see Fig._1)._The detection algorithm_is employed to search for the reduced burst. If_it is detected, the process -_including the full_search_-_is repeated until the_burst is not detected anymore. If_the burst is lost when the_distance has been increased by a factor_$f$ then $V/V_{\rm max} = f^{-3}$._The 2204 GRBs in the_GUSBAD catalog_have $<V/V_{\rm max}> = 0.346_\pm 0.006$. In the_process of_analyzing the reduced_burst profiles, the trigger may occur_at later times_depending on the detailed profile. (If_this_causes the second_background_stretch_to suffer_from contaminated or_missing_data, it_is_kept at its location during the_original_detection.) If so, the two background stretches_defined in Fig. 1_move_forward and the first_background stretch may contain burst_signal. This reduces the amplitude of_the reduced_burst and_increases the background. Hence $V/V_{\rm max}$ derived from the simulation is_larger than $(C_{\rm max}/C_{\rm min})^{-3/2}$. This_is confirmed in the_GUSBAD catalog_which_yields $<(C_{\rm max}/C_{\rm_min})^{-3/2}>_= 0.335$._In the BATSE catalog, only 884 of_the 1144_GRBs with $C_{\rm max}/C_{\rm min} \ge_1.0$ on a timescale_of_1024 ms can be used for_a comparison: the others fall at_times when the on-board trigger_used_parameters_other than the standard mentioned_in Sec. 1. They yield $<(C_{\rm_max}/C_{\rm min})^{-3/2}> =_0.311 \pm 0.009$. The GUSBAD Catalog ================== The GUSBAD catalog_is_available on the World Wide Web_[@sch03][^3]._The designation of GRBs in the_catalog_is_GUSBAD YYMMDD.ddd where Y, M,_D are integral year, month and_day numbers, respectively, and.ddd is the truncated fraction of_the day. The

X_{\infty}$ such that $x$ is accessible. Note that $\underleftarrow{\lim}\{I, f^k\}$ and $X_{\infty}$ are homeomorphic. We give applications of Corollary \[cor:iterations\] in the following examples. \[ex:2sin1x\] Let $f$ be a piecewise linear map such that $f(0)=0$, $f(1)=1$ and with critical points $\frac{1}{4}, \frac{3}{4}$, where $f(\frac{1}{4})=\frac{3}{4}$ and $f(\frac{3}{4})=\frac{1}{4}$ (see Figure \[fig:spiral\]). (0,0)–(1,1); (0,0) – (1/4, 3/4) – (3/4, 1/4) – (1,1); (0,0) – (0,1) – (1,1) – (1,0) – (0,0); (1/4, 0) – (1/4, 1); (1/2, 0) – (1/2, 1); (3/4, 0) – (3/4, 1); (0, 1/4) – (1, 1/4); (0, 1/2) – (1, 1/2); (0, 3/4) – (1, 3/4); at (1/4,-0.1) [$\frac{1}{4}$]{}; at (1/2,-0.1) [$\frac{1}{2}$]{}; at (3/4,-0.1) [$\frac{3}{4}$]{}; at (-0.1,1/4) [$\frac{1}{4}$]{}; at (-0.1,1/2) [$\frac{1}{2}$]{}; at (-0.1,3/4) [$\frac{3}{4}$]{}; (1/2,0.5) circle (0.02); Note that $X=\underleftarrow{\lim}\{I,f\}$ consists of two rays compactifying on an arc and therefore, for every $x\in X$, there exists a planar embedding making $x$ accessible. However, the point $\frac{1}{2}$ is inside a zigzag of $f$. Figure \[fig

X_{\infty}$ such that $ x$ is accessible. Note that $ \underleftarrow{\lim}\{I, f^k\}$ and $ X_{\infty}$ are homeomorphic. We give applications of Corollary   \[cor: iterations\ ] in the following exemplar. \[ex:2sin1x\ ] lease $ f$ be a piecewise linear map such that $ f(0)=0 $, $ f(1)=1 $ and with critical points $ \frac{1}{4 }, \frac{3}{4}$, where $ f(\frac{1}{4})=\frac{3}{4}$ and $ f(\frac{3}{4})=\frac{1}{4}$ (see Figure   \[fig: spiral\ ]). (0,0)–(1,1); (0,0) – (1/4, 3/4) – (3/4, 1/4) – (1,1); (0,0) – (0,1) – (1,1) – (1,0) – (0,0); (1/4, 0) – (1/4, 1); (1/2, 0) – (1/2, 1); (3/4, 0) – (3/4, 1); (0, 1/4) – (1, 1/4); (0, 1/2) – (1, 1/2); (0, 3/4) – (1, 3/4); at (1/4,-0.1) [ $ \frac{1}{4}$ ] { }; at (1/2,-0.1) [ $ \frac{1}{2}$ ] { }; at (3/4,-0.1) [ $ \frac{3}{4}$ ] { }; at (-0.1,1/4) [ $ \frac{1}{4}$ ] { }; at (-0.1,1/2) [ $ \frac{1}{2}$ ] { }; at (-0.1,3/4) [ $ \frac{3}{4}$ ] { }; (1/2,0.5) lap (0.02); Note that $ X=\underleftarrow{\lim}\{I, f\}$ consists of two beam compactifying on an bow and therefore, for every $ x\in X$, there exist a planar embedding making $ x$ accessible. However, the compass point $ \frac{1}{2}$ is inside a zigzag of $ f$. Figure   \[fig

X_{\invty}$ such that $x$ is accesrible. Note that $\oneerlefvarrow{\ljm}\{I, f^k\}$ avd $X_{\infty}$ are homeomorphic. We guve akilications of Corollafy \[cor:iterwtions\] ib tht following examples. \[ex:2sin1w\] Let $n$ be c 'iecewise lineat map such tvat $f(0)=0$, $f(1)=1$ and widh cxitical points $\frac{1}{4}, \frac{3}{4}$, where $f(\frac{1}{4})=\srac{3}{4}$ anc $v(\frac{3}{4})=\frac{1}{4}$ (see Siguge \[sig:slprcl\]). (0,0)–(1,1); (0,0) – (1/4, 3/4) – (3/4, 1/4) – (1,1); (0,0) – (0,1) – (1,1) – (1,0) – (0,0); (1/4, 0) – (1/4, 1); (1/2, 0) – (1/2, 1); (3/4, 0) – (3/4, 1); (0, 1/4) – (1, 1/4); (0, 1/2) – (1, 1/2); (0, 3/4) – (1, 3/4); ay (1/4,-0.1) [$\frac{1}{4}$]{}; at (1/2,-0.1) [$\frac{1}{2}$]{}; at (3/4,-0.1) [$\frac{3}{4}$]{}; ah (-0.1,1/4) [$\fgac{1}{4}$]{}; at (-0.1,1/2) [$\frac{1}{2}$]{}; at (-0.1,3/4) [$\fgac{3}{4}$]{}; (1/2,0.5) circle (0.02); Notq that $X=\underueftarrow{\lpk}\{I,f\}$ consisjs of two rays compactifying on xn are and theredoee, vmr every $x\ii X$, thvre exists a ikanar ambeddimg making $x$ acgessiule. Yowever, the point $\frar{1}{2}$ is inside a zigzag of $f$. Figgrz \[fig

X_{\infty}$ such that $x$ is accessible. Note f^k\}$ $X_{\infty}$ are We give applications following \[ex:2sin1x\] Let $f$ a piecewise linear such that $f(0)=0$, $f(1)=1$ and with points $\frac{1}{4}, \frac{3}{4}$, where $f(\frac{1}{4})=\frac{3}{4}$ and $f(\frac{3}{4})=\frac{1}{4}$ (see Figure \[fig:spiral\]). (0,0)–(1,1); (0,0) – 3/4) – (3/4, 1/4) – (1,1); (0,0) – (0,1) – (1,1) – (1,0) (0,0); 0) (1/4, (1/2, 0) – (1/2, 1); (3/4, 0) – (3/4, 1); (0, 1/4) – (1, 1/4); (0, – (1, 1/2); (0, 3/4) – (1, 3/4); (1/4,-0.1) [$\frac{1}{4}$]{}; at (1/2,-0.1) at (3/4,-0.1) [$\frac{3}{4}$]{}; at (-0.1,1/4) at [$\frac{1}{2}$]{}; at [$\frac{3}{4}$]{}; circle Note that $X=\underleftarrow{\lim}\{I,f\}$ of two rays compactifying on an arc and therefore, for every $x\in X$, there exists a planar making $x$ the point is a of $f$. Figure

X_{\infty}$ such that $x$ is accessibLe. Note that $\UnderLefTarRoW{\lim}\{i, f^k\}$ aNd $X_{\infty}$ are homEOmorPhic. We give applications Of CorOlLAry \[cOR:iTeratIons\] in tHE fOLLowInG eXamPlES. \[eX:2sin1x\] let $F$ be a pieCewise lineAr mAp Such that $f(0)=0$, $f(1)=1$ anD WiTh critical PoiNts $\frac{1}{4}, \frac{3}{4}$, wHerE $f(\frac{1}{4})=\FrAc{3}{4}$ aND $f(\fraC{3}{4})=\frAc{1}{4}$ (see figure \[FIg:spirAl\]). (0,0)–(1,1); (0,0) – (1/4, 3/4) – (3/4, 1/4) – (1,1); (0,0) – (0,1) – (1,1) – (1,0) – (0,0); (1/4, 0) – (1/4, 1); (1/2, 0) – (1/2, 1); (3/4, 0) – (3/4, 1); (0, 1/4) – (1, 1/4); (0, 1/2) – (1, 1/2); (0, 3/4) – (1, 3/4); at (1/4,-0.1) [$\frac{1}{4}$]{}; aT (1/2,-0.1) [$\fRAc{1}{2}$]{}; at (3/4,-0.1) [$\frAC{3}{4}$]{}; at (-0.1,1/4) [$\frac{1}{4}$]{}; AT (-0.1,1/2) [$\FrAc{1}{2}$]{}; at (-0.1,3/4) [$\Frac{3}{4}$]{}; (1/2,0.5) circle (0.02); Note thaT $x=\uNDerleftarrow{\liM}\{I,f\}$ conSiSTs OF Two RayS compactifYiNg on aN Arc and tHErEFORe, fOR every $x\in X$, theRe exists a plANar EmbeddInG maKIng $x$ acCessiBlE. howEver, the poinT $\fraC{1}{2}$ is inside A zigzaG Of $f$. FiguRE \[fig

X_{\infty}$ such that $x$is accessi ble. No teth at $ \und erleftarrow{\l i m}\{ I, f^k\}$ and $X_{\inf ty}$ar e hom e om orphi c. Weg iv e app li ca tio ns of Coro lla ry \[co r:iteratio ns\ ]in the follo w in g examples . \[ex:2sin1x\ ] L et $f$ b e a piece wis e lin ear ma p suchthat $f(0 )= 0 $, $f( 1 )=1$ an d wi th c ritical points $\ f ra c {1}{4}, \frac{ 3}{4}$ ,w he r e $f (\f rac{1}{4}) =\ frac{ 3 }{4}$ a n d$ f ( \fr a c{3}{4})=\fra c{1}{4}$ (s e e F igure\[ fig : spiral \]). ( 0 ,0) –(1,1); (0, 0) – (1/4, 3/ 4) – ( 3 /4, 1/4 ) – (1,1 ); (0, 0)– ( 0,1) –(1 ,1) – (1, 0 )– ( 0 ,0) ; (1/4,0) – (1/4 , 1) ; ( 1 /2,0)– (1 /2, 1 ); (3/4, 0) – (3 /4,1 );(0, 1 /4) – (1, 1 /4);(0, 1/ 2) –(1 , 1/2); (0, 3/4 ) –(1, 3/4); a t(1/ 4, -0.1) [$\fra c{1 }{4 }$]{};at (1/2 , -0. 1) [ $ \f rac{1}{2}$]{}; at(3 / 4 ,- 0.1) [$\ frac{3 } {4 }$ ] {}; at(- 0.1 ,1/4 ) [$\fr ac{1 } {4 }$]{}; a t (-0. 1 ,1 /2 ) [$\fr ac {1}{2} $] {}; at (-0. 1 ,3/4 ) [$\f rac{3}{4 }$]{} ; (1/2,0.5) cir c le (0.02); N o te t ha t $X= \un derleftarro w{\l i m}\{ I,f\ } $con s istsof tw or ay s compactifying on a narc an d the refore, for e very $x\in X $ , thereexis t sa planar embedd ing m aking $x$a ccessibl e. Ho wever, t he point$ \ frac{1}{ 2}$ is in sid e azigzag of $f$ . Figu re  \[fig

X_{\infty}$ such_that $x$_is accessible. Note that $\underleftarrow{\lim}\{I,_f^k\}$ and_$X_{\infty}$_are homeomorphic. We_give_applications of Corollary \[cor:iterations\]_in the following_examples. \[ex:2sin1x\] Let $f$ be_a piecewise linear_map_such that $f(0)=0$, $f(1)=1$ and with critical points $\frac{1}{4}, \frac{3}{4}$, where $f(\frac{1}{4})=\frac{3}{4}$ and $f(\frac{3}{4})=\frac{1}{4}$_(see_Figure \[fig:spiral\]). (0,0)–(1,1); (0,0)_–_(1/4,_3/4) – (3/4, 1/4) –_(1,1); (0,0) – (0,1) –_(1,1) –_(1,0) – (0,0); (1/4, 0) – (1/4, 1);_(1/2,_0) – (1/2,_1); (3/4, 0) – (3/4, 1); (0, 1/4) –_(1, 1/4); (0, 1/2) – (1,_1/2); (0, 3/4)_–_(1,_3/4); at (1/4,-0.1) [$\frac{1}{4}$]{}; at_(1/2,-0.1) [$\frac{1}{2}$]{}; at (3/4,-0.1) [$\frac{3}{4}$]{}; at (-0.1,1/4)_[$\frac{1}{4}$]{}; at (-0.1,1/2) [$\frac{1}{2}$]{}; at (-0.1,3/4)_[$\frac{3}{4}$]{}; (1/2,0.5) circle (0.02); Note that $X=\underleftarrow{\lim}\{I,f\}$ consists_of two rays compactifying on an_arc and therefore, for every_$x\in X$,_there exists a planar embedding_making $x$ accessible._However, the_point $\frac{1}{2}$ is_inside a zigzag of $f$. Figure \[fig

D_s$ was expected to be an excellent test of lattice QCD, for several reasons [@Briere:2001rn]. The matrix elements in Eqs. (\[eq:kronfeld:AfDs\]) and (\[eq:kronfeld:PfDs\]) are gold-plated, in the sense of Ref. [@Davies:2003ik], namely, only one hadron enters, and the chiral extrapolation is controlled. Experimental measurements of $|V_{cs}|f_{D_s}$, via Eq. (\[eq:kronfeld:BrSM\]), can be combined with the determination of $|V_{cs}|$ from CKM unitarity. The idea that new physics could compete is usually discounted, because the decay is Cabibbo-favored and proceeds at the tree level of the weak interactions. Finally, the precision of experiments has lagged that of calculations, so the analysis of the numerical lattice-QCD data is carried out with a relatively blind eye. The first round of testing seemed to go well. In June 2005, the first lattice-QCD calculation with 2+1 flavors of sea quarks appeared in a joint work of the Fermilab Lattice and MILC Collaborations: $$f_{D_s} = 249\pm 3 \pm 16~\textrm{MeV}~\textrm{Fermilab/MILC}~% \cite{Aubin:2005ar}, \label{eq:kronfeld:prl05}$$ where the first error is statistical and the second systematic. This prediction was followed a year later by a comparably accurate measurement of $B(D_s\to\mu\nu)/B(D_s\to\phi\pi)$ that, when combined with an independent measurement of $B(D_s\to\phi\pi)$, yielded $$\begin{aligned} \hspace*{-1.0em} f_{D_s} = 283 \pm 17 \pm 7 \pm 14~\textrm{MeV} & \hspace*{-0.5em} \mu\nu/\phi\pi \hspace*{-0.5em} & \textrm{BaBar}~\cite{Aubert:2006sd}, \label{eq:kronfeld:prl06}\end{aligned}$$ which agrees with Eq. (\[eq:kronfeld:prl05\]) at 1.2

D_s$ was expected to be an excellent test of lattice QCD, for several reason   [ @Briere:2001rn ]. The matrix chemical element in Eqs.   (\[eq: kronfeld: AfDs\ ]) and   (\[eq: kronfeld: PfDs\ ]) are gold - plated, in the sense of Ref.   [ @Davies:2003ik ], namely, entirely one hadron enters, and the chiral extrapolation is controlled. Experimental measurements of $ |V_{cs}|f_{D_s}$, via Eq.   (\[eq: kronfeld: BrSM\ ]), can be combine with the decision of $ |V_{cs}|$ from CKM unitarity. The idea that fresh physics could compete is usually discount, because the decay is Cabibbo - favored and proceeds at the tree degree of the weak interactions. Finally, the precision of experiments has lagged that of calculation, so the analysis of the numerical lattice - QCD data is carried out with a relatively blind center. The first round of examination seemed to move well. In June 2005, the first wicket - QCD calculation with 2 + 1 flavors of sea quarks appeared in a joint study of the Fermilab Lattice and MILC Collaborations: $ $ f_{D_s } = 249\pm 3 \pm 16~\textrm{MeV}~\textrm{Fermilab / MILC}~% \cite{Aubin:2005ar }, \label{eq: kronfeld: prl05}$$ where the first error is statistical and the second systematic. This prediction was followed a year later by a comparably accurate measurement of $ B(D_s\to\mu\nu)/B(D_s\to\phi\pi)$ that, when blend with an independent measurement of $ B(D_s\to\phi\pi)$, yielded $ $ \begin{aligned } \hspace*{-1.0em } f_{D_s } = 283 \pm 17 \pm 7 \pm 14~\textrm{MeV } & \hspace*{-0.5em } \mu\nu/\phi\pi \hspace*{-0.5em } & \textrm{BaBar}~\cite{Aubert:2006sd }, \label{eq: kronfeld: prl06}\end{aligned}$$ which agrees with Eq.   (\[eq: kronfeld: prl05\ ]) at 1.2

D_s$ aas expected to be an exgellent test of lattice QCD, fkr severxl reasons [@Briere:2001rn]. The matrie elwmentw in Eqs. (\[eq:kronfeld:AfDs\]) and (\[eq:krojfeld:PfDw\]) art gold-plated, in tis sense of Rer. [@Favizs:2003mk], namely, only pne hadron enters, and tha zhnral extrapolation is controlled. Expqrimentsl measurements jf $|V_{bs}|s_{D_s}$, bpa Eq. (\[eq:kronfeld:BrSM\]), can be combjned wiuh the determinatipn of $|V_{cs}|$ from CKM unitarihy. Tje idea that new pjysics coule coipete is usuauly discounted, because the decay is Cabibbo-favored ana proeeeds at thg frfg level of tie weah interactiokx. Finanly, the precision of cxpermmenrs has lagged that of calculations, so tre analyshs of the numerical lqtticg-QCD gata us zardixd kut wihh e relativelg blind eye. Rhe first round of uesepmg seemed to go wejl. In June 2005, the first lattice-QCD calculatpon sith 2+1 flavors of sea quqrks appeared in a jolnt work jf the Fermilab Lattice and MILC Collaborations: $$f_{G_s} = 249\pj 3 \pn 16~\bextfn{MfV}~\textrm{Fermilab/MILC}~% \cite{Aubin:2005ar}, \lafsl{tq:kgonfeld:prl05}$$ where bhe first error is shayystical and tfe second systematic. This pgedictijn waw followeq a uear later by a comparably qccurate meafyrement of $B(D_s\to\mu\uu)/B(D_s\to\phi\pi)$ thaj, when combined with an indepzndent measuremenh of $B(D_s\tk\ohi\pi)$, yielded $$\beein{slhgned} \hspace*{-1.0em} f_{D_s} = 283 \pm 17 \pm 7 \'m 14~\tertrm{MeV} & \hfpace*{-0.5em} \mu\ju/\phi\ih \hspace*{-0.5em} & \texttm{BaBas}~\cite{Auberh:2006sd}, \label{eq:kronfeld:prl06}\end{alijied}$$ which agrges winh Eq. (\[eq:krjnfelc:prl05\]) at 1.2

D_s$ was expected to be an excellent lattice for several [@Briere:2001rn]. The matrix (\[eq:kronfeld:PfDs\]) gold-plated, in the of Ref. [@Davies:2003ik], only one hadron enters, and the extrapolation is controlled. Experimental measurements of $|V_{cs}|f_{D_s}$, via Eq. (\[eq:kronfeld:BrSM\]), can be combined the determination of $|V_{cs}|$ from CKM unitarity. The idea that new physics could is discounted, the is Cabibbo-favored and proceeds at the tree level of the weak interactions. Finally, the precision of has lagged that of calculations, so the analysis the numerical lattice-QCD data carried out with a relatively eye. first round testing to well. In June the first lattice-QCD calculation with 2+1 flavors of sea quarks appeared in a joint work of the Lattice and $$f_{D_s} = 3 16~\textrm{MeV}~\textrm{Fermilab/MILC}~% \label{eq:kronfeld:prl05}$$ where the is statistical and the second systematic. followed a year later by a comparably accurate of $B(D_s\to\mu\nu)/B(D_s\to\phi\pi)$ when combined with an independent measurement $B(D_s\to\phi\pi)$, yielded $$\begin{aligned} \hspace*{-1.0em} f_{D_s} = 283 \pm \pm 7 \pm 14~\textrm{MeV} & \hspace*{-0.5em} \mu\nu/\phi\pi \hspace*{-0.5em} & \textrm{BaBar}~\cite{Aubert:2006sd}, \label{eq:kronfeld:prl06}\end{aligned}$$ which agrees with Eq. 1.2

D_s$ was expected to be an excellEnt test of lAtticE QCd, foR sEverAl reAsons [@Briere:2001rn]. THE matRix elements in Eqs. (\[eq:kronFeld:AFDS\]) And (\[eQ:KrOnfelD:PfDs\]) arE GoLD-PlaTeD, iN thE sENsE of ReF. [@DaVies:2003ik], nAmely, only oNe hAdRon enters, and THe Chiral extrApoLation is contRolLed. ExpErImeNTal meAsuRemenTs of $|V_{cS}|F_{D_s}$, via eq. (\[eq:kronfElD:brSM\]), caN Be combiNED wIth tHe determination of $|v_{Cs}|$ FRom CKM unitaritY. The idEa THaT NEw pHysIcs could coMpEte is USually dIScOUNTed, BEcause the decaY is Cabibbo-fAVorEd and pRoCeeDS at the Tree lEvEL of The weak inteRactIons. FinalLy, the pREcision OF experiMents hAs lAggEd thAT oF cAlcUlATioNS, sO thE AnaLysis of tHe NuMericAl laTTICE-QCD DatA is cArrieD out with a relaTivEly bLInd Eye. ThE firsT rouNd Of tesTing seEmed tO gO well. In June 2005, the fIrst Lattice-QCd caLcUlaTiOn witH 2+1 FlavorS of Sea Quarks aPpeared IN a jOiNT WOrK of the Fermilab LattIcE ANd mILC CollAboratIOnS: $$f_{d_S} = 249\pm 3 \pm 16~\texTrM{Mev}~\texTRM{FermIlab/miLc}~% \cite{AubIn:2005ar}, \laBEl{Eq:KronfelD:pRl05}$$ wherE tHe fIrsT erroR Is stAtistiCal and thE secoND systematic. ThiS Prediction was FOlLOWeD A yeaR laTer by a compaRablY AccuRate MEaSurEMent oF $B(D_s\tO\mU\Nu)/b(d_s\to\phi\pi)$ that, when coMbIned wiTh an iNdependent meaSurement of $b(d_S\To\phi\pi)$, yIeldED $$\bEGin{aligned} \hspaCe*{-1.0em} f_{d_s} = 283 \pm 17 \pm 7 \pm 14~\teXTrm{MeV} & \hsPace*{-0.5eM} \mu\nu/\phi\Pi \hspace*{-0.5eM} & \TExtrm{BaBAr}~\cIte{aubErt:2006SD}, \LaBel{eq:kronfeld:PRL06}\end{AlIgned}$$ whIch Agrees wIth eq. (\[eQ:krOnfElD:prl05\]) at 1.2

D_s$ was expected to be an excellent test of la tt iceQCD, for several r e ason s [@Briere:2001rn]. Th e mat ri x ele m en ts in Eqs. ( \ [e q : kro nf el d:A fD s \] ) and  (\ [eq:kro nfeld:PfDs \]) a re gold-plat e d, in the se nse of Ref. [@D avi es:200 3i k], namel y,onlyone ha d ron en ters, and t h e chir a l extra p o la tion is controlled. E x pe r imental measur ements o f $ | V _{c s}| f_{D_s}$,vi a Eq. (\[eq:k r on f e l d:B r SM\]), can be combined w i ththe de te rmi n ationof $| V_ { cs} |$ from CKM uni tarity. T he ide a that n e w physi cs cou ldcom pete is u sua ll y di s co unt e d,becauseth edecay isC a b i bbo- fav ored andproceeds at t hetree lev el of theweak i ntera ctions . Fin al ly, the precisi on o f experim ent shas l agged that o f c alc ulation s, so t h e a na l y s is of the numericalla t t ic e-QCD da ta isc ar ri e d out wi th arela t i velyblin d e ye. The first ro un d of te st ing se em edtogo we l l. I n June 2005, t he fi r st lattice-QCD calculation w i th 2 +1 flav ors of sea qua rksa ppea redi na j o int w ork o ft he Fermilab Lattice an dMILC C ollab orations: $$f _{D_s} = 2 4 9 \ pm 3 \pm 16~ \ te x trm{MeV}~\text rm{Fe rmilab/MIL C }~% \ cite{Aub in:2005ar } , \la bel {eq :kr onf e l d: prl05}$$ wher e thefi rst err oris stat ist ica l a ndth e secondsystemat ic .Th is pr edict i on was f ol low ed ayearl ater b y a c ompa ra bl y ac curatem ea s u reme nt o f $B (D_ s\ to\mu \nu) / B(D _s\to\p hi\pi)$ t hat , whe nco mbinedwith an indep en dent measu re men t of $ B ( D_s\to\p hi\pi)$, yielded $$\beg i n{align ed} \hsp ace*{-1.0 em} f _{D _ s} = 2 83 \pm 17 \ pm 7\ p m 14~ \ t ex trm {M eV} & \hspa ce *{-0 .5em} \ mu\nu/\phi\pi \hsp a ce* {-0.5em} & \ t e xt rm{ B aB a r}~ \c i te{ A u bert:2006sd}, \label{ eq : kr onfeld:prl 0 6}\ en d{align ed}$$ w hicha grees w ith Eq. ( \[eq:kron fe ld:p r l 05\ ]) at 1.2

D_s$ was_expected to_be an excellent test_of lattice_QCD,_for several_reasons [@Briere:2001rn]._The matrix elements_in Eqs. (\[eq:kronfeld:AfDs\]) and (\[eq:kronfeld:PfDs\])_are gold-plated, in the_sense of Ref. [@Davies:2003ik],_namely,_only one hadron enters, and the chiral extrapolation is controlled. Experimental measurements of $|V_{cs}|f_{D_s}$,_via_Eq. (\[eq:kronfeld:BrSM\]), can_be_combined_with the determination of $|V_{cs}|$_from CKM unitarity. The idea_that new_physics could compete is usually discounted, because the_decay_is Cabibbo-favored and_proceeds at the tree level of the weak interactions._Finally, the precision of experiments has_lagged that of_calculations,_so_the analysis of the_numerical lattice-QCD data is carried out_with a relatively blind eye. The first_round of testing seemed to go well._In June 2005, the first lattice-QCD_calculation with 2+1 flavors of_sea quarks_appeared in a joint work_of the Fermilab_Lattice and_MILC Collaborations: $$f_{D_s}_= 249\pm 3 \pm 16~\textrm{MeV}~\textrm{Fermilab/MILC}~% _ _ \cite{Aubin:2005ar}, __\label{eq:kronfeld:prl05}$$ where the_first_error_is statistical_and the second_systematic._This prediction_was_followed a year later by a_comparably_accurate measurement of $B(D_s\to\mu\nu)/B(D_s\to\phi\pi)$ that, when combined_with an independent measurement_of_$B(D_s\to\phi\pi)$, yielded $$\begin{aligned} _ \hspace*{-1.0em} _ f_{D_s} = 283 \pm 17_\pm 7_\pm 14~\textrm{MeV}_& \hspace*{-0.5em} \mu\nu/\phi\pi_\hspace*{-0.5em} & _ \textrm{BaBar}~\cite{Aubert:2006sd}, _ __\label{eq:kronfeld:prl06}\end{aligned}$$ which agrees_with_Eq. (\[eq:kronfeld:prl05\]) at_1.2

*{DHL}[54,2]$ of Theorem \[main-dhl\](i). We also have a variant of Theorem \[maynard-thm\] which can accept inputs of the form $\operatorname*{MPZ}[\varpi,\delta]$: \[maynard-trunc\] Let $k \geq 2$ and $m \geq 1$ be fixed integers. Let $0<\varpi<1/4$ and $0 <\delta < 1/2$ be such that $\operatorname*{MPZ}[\varpi,\delta]$ holds. For any $\alpha>0$, let $M_k^{[\alpha]}$ be defined as in, but where the supremum now ranges over all square-integrable functions $F$ supported in the *truncated* simplex $$\label{ttk} \{ (t_1,\dots,t_k) \in [0,\alpha]^k: t_1+\dots+t_k \leq 1 \}$$ and are not identically zero. If $$M_k^{[\frac{\delta}{1/4+\varpi}]} > \frac{m}{1/4+\varpi},$$ then $\operatorname*{DHL}[k,m+1]$ holds. In Section \[asymptotics-sec\] we will establish the following variant of Theorem \[mlower\], which when combined with Theorem \[mpz-poly\], allows one to use Theorem \[maynard-trunc\] to establish parts (ii)-(vi) of Theorem \[main-dhl\] (and hence Theorem \[main\]): \[mlower-var\] - There exist $\delta,\varpi>0$ with $600 \varpi + 180 \delta < 7$ and $M_{\num{35410}}^{[\frac{\delta}{1/4+\varpi}]} > \frac{2}{1/4+\varpi}$. - There exist $\delta,\varpi>0$ with $600 \varpi + 180 \delta < 7$ and $M_{\num{1649821}}^{[\frac{\delta}{1/4+\varpi}]} > \frac{3}{1/4+\varpi}$. - There exist $\delta,\varpi>0$ with $600 \varpi + 180 \delta < 7$ and $M_{\num{75845707}}^{[\frac{\delta}{1/4+\varpi}]} > \frac{4}{1/4+\varpi}$. -

* { DHL}[54,2]$ of Theorem \[main - dhl\](i). We also have a variant of Theorem \[maynard - thm\ ] which can accept stimulation of the human body $ \operatorname*{MPZ}[\varpi,\delta]$: \[maynard - trunc\ ] Let $ k \geq 2 $ and $ m \geq 1 $ be fixed integer. Let $ 0<\varpi<1/4 $ and $ 0 < \delta < 1/2 $ be such that $ \operatorname*{MPZ}[\varpi,\delta]$ holds. For any $ \alpha>0 $, let $ M_k^{[\alpha]}$ be specify as in, but where the supremum now ranges over all straight - integrable functions $ F$ supported in the * truncate * simplex $ $ \label{ttk } \ { (t_1,\dots, t_k) \in [ 0,\alpha]^k: t_1+\dots+t_k \leq 1 \}$$ and are not identically zero. If $ $ M_k^{[\frac{\delta}{1/4+\varpi } ] } > \frac{m}{1/4+\varpi},$$ then $ \operatorname*{DHL}[k, m+1]$ holds. In Section \[asymptotics - sec\ ] we will establish the follow variant of Theorem \[mlower\ ], which when combined with Theorem \[mpz - poly\ ], allows one to use Theorem \[maynard - trunc\ ] to establish part (ii)-(vi) of Theorem \[main - dhl\ ] (and hence Theorem \[main\ ] ): \[mlower - var\ ] - There exist $ \delta,\varpi>0 $ with $ 600 \varpi + 180 \delta < 7 $ and $ M_{\num{35410}}^{[\frac{\delta}{1/4+\varpi } ] } > \frac{2}{1/4+\varpi}$. - There exist $ \delta,\varpi>0 $ with $ 600 \varpi + 180 \delta < 7 $ and $ M_{\num{1649821}}^{[\frac{\delta}{1/4+\varpi } ] } > \frac{3}{1/4+\varpi}$. - There exist $ \delta,\varpi>0 $ with $ 600 \varpi + 180 \delta < 7 $ and $ M_{\num{75845707}}^{[\frac{\delta}{1/4+\varpi } ] } > \frac{4}{1/4+\varpi}$. -

*{DHL}[54,2]$ of Theorem \[main-dhl\](i). We auso have a varicbt of Vheorem \[maynard-ghm\] which can accept inputs lf the dorm $\operatorname*{MPZ}[\vafpi,\delta]$: \[mwynard-trync\] Owt $k \geq 2$ ehd $m \geq 1$ be rlxed nnvegers. Let $0<\varpl<1/4$ and $0 <\deltd < 1/2$ be such thdt $\o'eratorname*{MPZ}[\varpi,\delta]$ holds. For agy $\alphs>0$, pet $M_k^{[\alpha]}$ be defpnqd aa in, but where the supremum now rahges ovtr all square-integtable functions $F$ supportef in the *truncated* simolex $$\label{tji} \{ (e_1,\eots,t_k) \in [0,\aloha]^k: t_1+\dots+u_k \leq 1 \}$$ and are not identically zero. If $$M_k^{[\wrac{\dzlta}{1/4+\varpi}]} > \drqc{m}{1/4+\gdrpi},$$ then $\o'eratogname*{DHL}[k,m+1]$ holds. In Sewtion \[axymptotics-sec\] ee xill establish the followmng variant of Theorgm \[mlower\], fhnch when combined wity Rheorgm \[mps-polh\], aluowa pns to ude Vheorem \[mayhard-trunc\] ti establish parts (io)-(vy) of Theorem \[mzin-dhl\] (agd hence Theorem \[main\]): \[mlower-var\] - There txist $\delta,\varpi>0$ with $600 \varpi + 180 \delta < 7$ and $M_{\num{35410}}^{[\frwc{\delta}{1/4+\vawpi}]} > \frac{2}{1/4+\varpi}$. - There exist $\delta,\varpi>0$ with $600 \var[i + 180 \veuta < 7$ xbd $M_{\num{1649821}}^{[\frac{\delta}{1/4+\varpi}]} > \frac{3}{1/4+\varpi}$. - There exist $\qslua,\vsrpi>0$ with $600 \varii + 180 \delta < 7$ and $M_{\mul{75845707}}^{[\ftwc{\delta}{1/4+\varpi}]} > \frac{4}{1/4+\vcdpj}$. -

*{DHL}[54,2]$ of Theorem \[main-dhl\](i). We also have of \[maynard-thm\] which accept inputs of $k 2$ and $m 1$ be fixed Let $0<\varpi<1/4$ and $0 <\delta < be such that $\operatorname*{MPZ}[\varpi,\delta]$ holds. For any $\alpha>0$, let $M_k^{[\alpha]}$ be defined as but where the supremum now ranges over all square-integrable functions $F$ supported in *truncated* $$\label{ttk} (t_1,\dots,t_k) [0,\alpha]^k: t_1+\dots+t_k \leq 1 \}$$ and are not identically zero. If $$M_k^{[\frac{\delta}{1/4+\varpi}]} > \frac{m}{1/4+\varpi},$$ then $\operatorname*{DHL}[k,m+1]$ In Section \[asymptotics-sec\] we will establish the following of Theorem \[mlower\], which combined with Theorem \[mpz-poly\], allows to Theorem \[maynard-trunc\] establish (ii)-(vi) Theorem \[main-dhl\] (and Theorem \[main\]): \[mlower-var\] - There exist $\delta,\varpi>0$ with $600 \varpi + 180 \delta < 7$ and $M_{\num{35410}}^{[\frac{\delta}{1/4+\varpi}]} \frac{2}{1/4+\varpi}$. - $\delta,\varpi>0$ with \varpi 180 < 7$ and \frac{3}{1/4+\varpi}$. - There exist $\delta,\varpi>0$ with 180 \delta < 7$ and $M_{\num{75845707}}^{[\frac{\delta}{1/4+\varpi}]} > \frac{4}{1/4+\varpi}$.

*{DHL}[54,2]$ of Theorem \[main-dhl\](i). We alsO have a variAnt of theOreM \[mAynaRd-thM\] which can accepT InpuTs of the form $\operatornamE*{MPZ}[\vArPI,\delTA]$: \[mAynarD-trunc\] LET $k \GEQ 2$ anD $m \GeQ 1$ be FiXEd IntegErs. let $0<\varpI<1/4$ and $0 <\delta < 1/2$ bE suCh That $\operatorNAmE*{MPZ}[\varpi,\dEltA]$ holds. For any $\AlpHa>0$, let $M_K^{[\aLphA]}$ Be defIneD as in, But wheRE the suPremum now RaNGes oveR All squaRE-InTegrAble functions $F$ supPOrTEd in the *truncatEd* simpLeX $$\LaBEL{ttK} \{ (t_1,\dOts,t_k) \in [0,\alpHa]^K: t_1+\dotS+T_k \leq 1 \}$$ anD ArE NOT idENtically zero. IF $$M_k^{[\frac{\deltA}{1/4+\VarPi}]} > \frac{M}{1/4+\vArpI},$$ Then $\opEratoRnAMe*{DhL}[k,m+1]$ holds. In sectIon \[asymptOtics-sEC\] we will EStablisH the foLloWinG varIAnT oF ThEoREm \[mLOwEr\], wHIch When combInEd With THeorEM \[MPZ-polY\], alLows One to Use Theorem \[mayNarD-truNC\] to EstabLish pArts (Ii)-(Vi) of THeorem \[Main-dHl\] (And hence Theorem \[Main\]): \[Mlower-var\] - theRe ExiSt $\Delta,\VArpi>0$ wiTh $600 \vArpI + 180 \delta < 7$ aNd $M_{\num{35410}}^{[\fRAc{\dElTA}{1/4+\VArPi}]} > \frac{2}{1/4+\varpi}$. - There exIsT $\DElTa,\varpi>0$ wIth $600 \varPI + 180 \dElTA < 7$ and $M_{\num{1649821}}^{[\FrAc{\dElta}{1/4+\VARpi}]} > \frAc{3}{1/4+\vaRPi}$. - there exiSt $\deltA,\VaRpI>0$ with $600 \vaRpI + 180 \delta < 7$ AnD $M_{\nUm{75845707}}^{[\fRac{\deLTa}{1/4+\vaRpi}]} > \fraC{4}{1/4+\varpi}$. -

*{DHL}[54,2]$ of Theorem \ [main-dhl\ ](i). W e a ls o ha ve a variant of Th e orem \[maynard-thm\] which canac c epti np uts o f the f o rm $ \op er at orn am e *{ MPZ}[ \va rpi,\de lta]$: \[ may na rd-trunc\] L e t$k \geq 2$ an d $m \geq 1$ be fixed i nte g ers.Let $0<\ varpi< 1 /4$ an d $0 <\de lt a < 1/2 $ be suc h th at $ \operatorname*{MP Z }[ \ varpi,\delta]$ holds .F or a ny$\a lpha>0$, l et $M_k ^ {[\alph a ]} $ b e d e fined as in,but where t h e s upremu mnow ranges over a l l s quare-integ rabl e functio ns $F$ support e d in th e *tru nca ted * si m pl ex $$ \l a bel { tt k}\{(t_1,\do ts ,t _k) \ in [ 0 , \ a lpha ]^k : t_ 1+\do ts+t_k \leq 1 \} $$ a n d a re no t ide ntic al ly ze ro. If $$M_ k^ {[\frac{\delta} {1/4 +\varpi}] } > \ fra c{ m}{1/ 4 +\varp i}, $$then $\ operato r nam e* { D H L} [k,m+1]$ holds. I nS e ct ion \[as ymptot i cs -s e c\] we w il l e stab l i sh th e fo l lo wing var iant o f T he orem \[ ml ower\] ,whi chwhenc ombi ned wi th Theor em \[ m pz-poly\], all o ws one to use Th e o re m \[m ayn ard-trunc\] toe stab lish pa rts (ii)- (vi)of Th e orem \[main-dhl\] ( an d henc e The orem \[main\] ): \[mlow e r - var\] - T h er e exist $\delta ,\var pi>0$ with $600 \va rpi + 180 \de lta < 7$a n d $M_{\n um{ 354 10} }^{ [ \ fr ac{\delta}{1/ 4 + \var pi }]} > \ fra c{2}{1/ 4+\ var pi} $. - Thereexist $\ de lt a, \v arp i>0$w ith $600 \ var pi +180 \ d elta < 7$ a nd $ M_ {\ n um{ 1649821 } }^ { [ \fra c{ \d elta }{1 /4 +\var pi}] } >\frac{3 }{1/4+\va rpi } $. - There e xist $\delta, \v arpi>0$ wi th $6 00 \va r p i + 180\delta < 7$ and $M_{\nu m {758457 07} }^{[\ frac {\delta}{ 1/4 +\varp i}] } > \fr ac{4}{ 1/4+\ va rpi } $ . -

*{DHL}[54,2]$ of_Theorem \[main-dhl\](i). We_also have a variant_of Theorem_\[maynard-thm\]_which can_accept_inputs of the_form $\operatorname*{MPZ}[\varpi,\delta]$: \[maynard-trunc\] Let_$k \geq 2$ and_$m \geq 1$_be_fixed integers. Let $0<\varpi<1/4$ and $0 <\delta < 1/2$ be such that $\operatorname*{MPZ}[\varpi,\delta]$ holds._For_any $\alpha>0$,_let_$M_k^{[\alpha]}$_be defined as in, but_where the supremum now ranges_over all_square-integrable functions $F$ supported in the *truncated* simplex_$$\label{ttk} _\{ (t_1,\dots,t_k) \in_[0,\alpha]^k: t_1+\dots+t_k \leq 1 \}$$ and are not identically_zero. If $$M_k^{[\frac{\delta}{1/4+\varpi}]} > \frac{m}{1/4+\varpi},$$ then_$\operatorname*{DHL}[k,m+1]$ holds. In Section_\[asymptotics-sec\]_we_will establish the following_variant of Theorem \[mlower\], which when_combined with Theorem \[mpz-poly\], allows one_to use Theorem \[maynard-trunc\] to establish parts_(ii)-(vi) of Theorem \[main-dhl\] (and hence_Theorem \[main\]): \[mlower-var\] - There_exist $\delta,\varpi>0$_with $600 \varpi + 180_\delta < 7$_and $M_{\num{35410}}^{[\frac{\delta}{1/4+\varpi}]}_> \frac{2}{1/4+\varpi}$. - _ There exist $\delta,\varpi>0$ with $600_\varpi + 180_\delta < 7$ and $M_{\num{1649821}}^{[\frac{\delta}{1/4+\varpi}]} >_\frac{3}{1/4+\varpi}$. -_ There_exist_$\delta,\varpi>0$_with $600_\varpi + 180_\delta_< 7$_and_$M_{\num{75845707}}^{[\frac{\delta}{1/4+\varpi}]} > \frac{4}{1/4+\varpi}$. -

imizer ADAM [@kingma2014adam] Weight Decay $0.01$ Learning Rate $5$e$-6$, with learning rate decay on plateau Epochs $100$ Training Data $10000$ Mini-batch size $20$ SVGD Particles $20$ ----------------- -------------------------------------------------------------------------------------- : Neural network architecture and training details.[]{data-label="tab:nn"} The network architecture was determined by training an ensemble of neural networks with different number of hidden layers. Other network parameters, such as the taper of the last several layers and learning rate specified in Table \[tab:nn\], were identical between each of the tested architectures. The networks are compared in Fig. \[fig:hiddenLayersMSE\] with the mean negative log likelihood (MNLL) defined by: $$\begin{gathered} \quad {MNLL}=-\frac{1}{T}\sum_{i=1}^{T}\frac{1}{N}\sum_{j=1}^{N}\log{p(\hat{\bm{b}}_{i}|\bm{f}(\hat{\bm{x}}_{i}, \mathbf{w}_{j}), \beta_{j})},\end{gathered}$$ where $T$, $\hat{\bm{x}}$, $\hat{\bm{b}}$ are the number of validation/test data points, the target inputs and target outputs, respectively. We observe little distinguishable difference indicating that training between each architecture is relatively the same. Additionally, the mean squared prediction error of the unnormalized anisotropic tensor $\bm{a}^{*}$ is plotted for a validation set of $1000$ random data points from each of the training flows (i.e. $5000$ total data points). The mean squared prediction error (MSPE) is defined as: $${MSPE}=\frac{1}{T}\sum^{T}_{i=1}{\left\lVert\mathbb{E}(\bm{a}^{*}_{i}|\hat{\bm{x}}_{i},\mathcal{D})-\hat{\bm{a}}_{i}\right\rVert}^{2}_{2}\approx\frac{1}{T}\sum^{T}_{i=1}{\left\lVert\frac{k}{N}\sum_{j=1}^{N}\bm{f}(\hat{\bm{x}}_{i},\mathbf{w}_{j})-\hat{\bm{a}}_{i}\right\

imizer ADAM   [ @kingma2014adam ] Weight Decay $ 0.01 $ Learning Rate $ 5$e$-6 $, with learning rate decay on plateau Epochs $ 100 $ Training Data $ 10000 $ Mini - batch size $ 20 $ SVGD Particles $ 20 $ ----------------- -------------------------------------------------------------------------------------- : nervous net computer architecture and training details.[]{data - label="tab: nn " } The network architecture was determine by training an ensemble of nervous networks with different number of concealed level. Other net parameters, such as the taper of the last several layer and learning pace specified in Table   \[tab: nn\ ], were identical between each of the test architecture. The networks are compared in Fig.   \[fig: hiddenLayersMSE\ ] with the mean minus log likelihood (MNLL) defined by: $ $ \begin{gathered } \quad { MNLL}=-\frac{1}{T}\sum_{i=1}^{T}\frac{1}{N}\sum_{j=1}^{N}\log{p(\hat{\bm{b}}_{i}|\bm{f}(\hat{\bm{x}}_{i }, \mathbf{w}_{j }), \beta_{j})},\end{gathered}$$ where $ T$, $ \hat{\bm{x}}$, $ \hat{\bm{b}}$ are the number of validation / trial data points, the target inputs and target outputs, respectively. We note little distinguishable difference indicating that training between each computer architecture is relatively the like. Additionally, the beggarly squared prediction error of the unnormalized anisotropic tensor $ \bm{a}^{*}$ is plotted for a validation set of $ 1000 $ random data points from each of the education flows (i.e. $ 5000 $ entire data points). The mean square prediction error (MSPE) is defined as: $ $ { MSPE}=\frac{1}{T}\sum^{T}_{i=1}{\left\lVert\mathbb{E}(\bm{a}^{*}_{i}|\hat{\bm{x}}_{i},\mathcal{D})-\hat{\bm{a}}_{i}\right\rVert}^{2}_{2}\approx\frac{1}{T}\sum^{T}_{i=1}{\left\lVert\frac{k}{N}\sum_{j=1}^{N}\bm{f}(\hat{\bm{x}}_{i},\mathbf{w}_{j})-\hat{\bm{a}}_{i}\right\

imixer ADAM [@kingma2014adxm] Weight Decai $0.01$ Learnjng Rate $5$e$-6$, with learning rate deray in plqteau Epochs $100$ Traijing Datq $10000$ Mini-batrg size $20$ SVFF Paxtmcles $20$ ----------------- -------------------------------------------------------------------------------------- : Keural netwmrk architectuse aud training details.[]{data-label="tab:nn"} The networl wrchitecture wws dttewminsd by training an ensemble of neurzl netwmrks with difgerent number of hidden lajers. Other network parwmeters, sucy as rhe taper of the last several layets and learning rate specified iv Tabke \[tab:nn\], wetz idfttical betwxen eabh of the tesbvd archhtecturrs. The networkx ace cimpared in Fig. \[fig:hiddxnLayersMSE\] with the mean negdtnve log likelihood (MNOL) defited ty: $$\bdtin{eatgeced} \quwd {JNLL}=-\frac{1}{T}\shm_{i=1}^{T}\frac{1}{N}\sun_{j=1}^{N}\log{p(\hat{\bm{b}}_{i}|\bm{f}(\hau{\bm{v}}_{p}, \mathbf{w}_{j}), \befa_{j})},\end{daehered}$$ where $T$, $\hat{\bm{x}}$, $\hat{\bm{b}}$ are the nukbed of validation/test datq points, the target ijputs and target outputs, respectively. We observe little divtingnirhayoe diwdegence indicating that training between each awdhotvcture is relativcly the same. Additoojakjy, the mean sduared prsdiction error of hhe unnjrmaluzed anisjtrolic tensor $\bm{a}^{*}$ is plotted fir a validatpon wet of $1000$ random datc points froo eavh of the training flows (i.e. $5000$ totzl data poijts). The msxn squared prediztipn error (MSPE) is defined as: $${ISPE}=\frac{1}{T}\wum^{T}_{n=1}{\left\lVeft\majhbb{E}(\bm{w}^{*}_{i}|\hat{\bm{x}}_{i},\mwthcal{D})-\hat{\bm{a}}_{i}\right\rVert}^{2}_{2}\wppror\frac{1}{D}\sum^{T}_{i=1}{\left\pVert\frac{k}{N}\sum_{j=1}^{N}\bm{f}(\hat{\bm{x}}_{i},\mathbh{x}_{j})-\hat{\bm{a}}_{i}\righj\

imizer ADAM [@kingma2014adam] Weight Decay $0.01$ Learning with rate decay plateau Epochs $100$ $20$ Particles $20$ ----------------- : Neural network and training details.[]{data-label="tab:nn"} The network architecture determined by training an ensemble of neural networks with different number of hidden Other network parameters, such as the taper of the last several layers and rate in \[tab:nn\], identical between each of the tested architectures. The networks are compared in Fig. \[fig:hiddenLayersMSE\] with the negative log likelihood (MNLL) defined by: $$\begin{gathered} \quad \mathbf{w}_{j}), \beta_{j})},\end{gathered}$$ where $T$, $\hat{\bm{b}}$ are the number of data the target and outputs, We observe little difference indicating that training between each architecture is relatively the same. Additionally, the mean squared prediction error the unnormalized $\bm{a}^{*}$ is for validation of $1000$ random from each of the training flows data points). The mean squared prediction error (MSPE) defined as:

imizer ADAM [@kingma2014adam] WeighT Decay $0.01$ LearNing RAte $5$E$-6$, wiTh LearNing Rate decay on plaTEau EPochs $100$ Training Data $10000$ Mini-bAtch sIzE $20$ sVGD pArTicleS $20$ ----------------- -------------------------------------------------------------------------------------- : Neural NEtWORk aRcHiTecTuRE aNd traIniNg detaiLs.[]{data-labeL="taB:nN"} The network aRChItecture waS deTermined by trAinIng an eNsEmbLE of neUraL netwOrks wiTH diffeRent numbeR oF Hidden LAyers. OtHER nEtwoRk parameters, such aS ThE Taper of the last SeveraL lAYeRS And LeaRning rate sPeCifieD In Table \[TAb:NN\], WEre IDentical betweEn each of the TEstEd archItEctURes. The NetwoRkS Are Compared in FIg. \[fiG:hiddenLaYersMSe\] With the MEan negaTive loG liKelIhooD (mNlL) DefInED by: $$\BEgIn{gATheRed} \quad {MnLl}=-\fRac{1}{T}\sUm_{i=1}^{T}\FRAC{1}{n}\sum_{J=1}^{N}\lOg{p(\hAt{\bm{b}}_{I}|\bm{f}(\hat{\bm{x}}_{i}, \maThbF{w}_{j}), \bETa_{j})},\End{gaThereD}$$ wheRe $t$, $\hat{\bM{x}}$, $\hat{\bM{b}}$ are ThE number of validaTion/Test data pOinTs, The TaRget iNPuts anD taRgeT outputS, respecTIveLy. wE OBsErve little distinguIsHABlE differeNce indICaTiNG that traInIng BetwEEN each ArchITeCture is rElativELy ThE same. AdDiTionalLy, The MeaN squaREd prEdictiOn error oF the uNNormalized anisOTropic tensor $\bM{A}^{*}$ iS PLoTTed fOr a Validation sEt of $1000$ RAndoM datA PoIntS From eAch of ThE TrAIning flows (i.e. $5000$ total daTa Points). the meAn squared predIction erroR (mspE) is defiNed aS: $${mSpe}=\frac{1}{T}\sum^{T}_{i=1}{\lefT\lVerT\mathbb{E}(\bm{A}^{*}_{I}|\hat{\bm{x}}_{i},\MathcAl{D})-\hat{\bm{A}}_{i}\right\rVERT}^{2}_{2}\approx\fRac{1}{t}\suM^{T}_{i=1}{\LefT\LveRt\frac{k}{N}\sum_{j=1}^{N}\BM{F}(\hat{\Bm{X}}_{i},\mathbF{w}_{j})-\Hat{\bm{a}}_{i}\RigHt\

imizer ADAM [@king ma2014adam ] W eig htDe cay $0.01$ Lear n ingRate $5$e$-6$, wit h lea rn i ng r a te deca y on pl a te a u Ep oc hs $1 00$ T raining Da ta $10000$ M in i-batch si ze $20$ SVG D P articl es $20$ - ----- ------ - ---- - --------- -- - ------ - ------- - - -- ---- ----------------- - -- - -------------- ------ -- - -- - - --- : Neuralne twork archite c tu r e and training deta ils.[]{data - lab el="ta b: nn" } Thenetwo rk arc hitecture w as d etermined by tr a ining a n ensemb le ofneu ral net w or ks wi th dif f er ent num ber of h id de n lay ers. O t h er n etw orkparam eters, such a s t he t a per of t he la st s ev erallayers andle arning rate spe cifi ed in Tab le\[ tab :n n\],w ere id ent ica l betwe en each ofth e t es ted architectures. T h e n etworksare co m pa re d in Fig.  \ [fi g:hi d d enLay ersM S E\ ] with t he mea n n eg ative l og likel ih ood (M NLL)d efin ed by: $$\begi n{gat h ered} \qua d {MNLL}=-\fra c {1 } { T} \ sum_ {i= 1}^{T}\frac {1}{ N }\su m_{j = 1} ^{N } \log{ p(\ha t{ \ bm { b}}_{i}|\bm{f}(\hat {\ bm{x}} _{i}, \mathbf{w}_{ j}), \beta _ { j })},\end {gat h er e d}$$ where $T$ , $\h at{\bm{x}} $ , $\hat{ \bm{b }}$ arethe numbe r of valid ati on/ tes t d a t apoints, the t a r getin puts an d t arget o utp uts , r esp ec tively. W e observ eli tt le di sting u ishabledi ffe re nce indi c atingthattrai ni ng bet ween ea c ha r chit ec tu re i s r el ative ly t h e s ame. Ad ditionall y,t he m ea nsquared prediction e rr or of theun nor malize d anisotro pic tensor $\bm{a}^{*}$ is plot ted fora va lidationset of $1 000 $ rando m data poin ts fr o m each o fthe t raining fl o w s ( i.e.$5 000$ totaldata points). Them ean squared pred ict ione r ro r ( M SP E ) i sd efi n e d as: $${MSPE}= \frac{1}{T }\ s um ^{T}_{i=1} { \le ft \lVert\ mathbb{ E}(\b m {a}^{*} _{i}|\hat {\bm{x}}_ {i },\m a t hca l{D})-\hat {\bm{a}} _{i}\righ t \rVer t }^ {2}_{ 2}\ approx \f rac {1}{T }\sum^ { T}_ {i=1} {\left \l Vert\f rac{k }{ N}\sum_{ j=1}^{N}\bm{f}(\hat{\bm {x}}_{ i},\m ath bf{w}_{j} )-\ h at{ \bm{a}}_{ i}\r ight\

imizer _ _ _ ADAM [@kingma2014adam] __Weight Decay__ _ $0.01$ _Learning Rate _ $5$e$-6$,_with_learning rate decay on plateau Epochs __ __$100$ _ Training Data _ $10000$ Mini-batch_size _ $20$ SVGD Particles _$20$ _ ----------------- -------------------------------------------------------------------------------------- _ : Neural network architecture and training details.[]{data-label="tab:nn"} The network_architecture was determined by training an_ensemble of neural_networks_with_different number of hidden_layers. Other network parameters, such as_the taper of the last several_layers and learning rate specified in Table \[tab:nn\],_were identical between each of the_tested architectures. The networks are_compared in_Fig. \[fig:hiddenLayersMSE\] with the mean negative_log likelihood (MNLL)_defined by:_$$\begin{gathered} _ \quad {MNLL}=-\frac{1}{T}\sum_{i=1}^{T}\frac{1}{N}\sum_{j=1}^{N}\log{p(\hat{\bm{b}}_{i}|\bm{f}(\hat{\bm{x}}_{i}, \mathbf{w}_{j}), \beta_{j})},\end{gathered}$$ where_$T$, $\hat{\bm{x}}$, $\hat{\bm{b}}$_are the number of validation/test data_points,_the target inputs_and_target_outputs, respectively._We observe little_distinguishable_difference indicating_that_training between each architecture is relatively_the_same. Additionally, the mean squared prediction error_of the unnormalized anisotropic_tensor_$\bm{a}^{*}$ is plotted for_a validation set of $1000$_random data points from each of_the training_flows (i.e._$5000$ total data points). The mean squared prediction error (MSPE) is_defined as: $${MSPE}=\frac{1}{T}\sum^{T}_{i=1}{\left\lVert\mathbb{E}(\bm{a}^{*}_{i}|\hat{\bm{x}}_{i},\mathcal{D})-\hat{\bm{a}}_{i}\right\rVert}^{2}_{2}\approx\frac{1}{T}\sum^{T}_{i=1}{\left\lVert\frac{k}{N}\sum_{j=1}^{N}\bm{f}(\hat{\bm{x}}_{i},\mathbf{w}_{j})-\hat{\bm{a}}_{i}\right\

, M., Dones, L., & Nesvorn[ý]{}, D. 2013, arXiv:1311.6780 uk, M., Dones, L., & Nesvorn[ý]{}, D. 2016,, 820, 97 uk, M., Hamilton, D. P., Lock, S. J., & Stewart, S. T. 2016, Natur, 539, 402 Danby, J. M. A. 1992. Fundamentals of celestial mechanics. Richmond: Willman-Bell, 1992, 2nd ed. Dombard, A. J., Cheng, A. F., McKinnon, W. B., & Kay, J. P. 2012, Journal of Geophysical Research (Planets), 117, E03002 El Moutamid, M., Hedman, M. M., & Nicholson, P. D. 2017, AAS/Division for Planetary Sciences Meeting Abstracts \#49, 49, 212.02 French, R. G., McGhee-French, C. A., Lonergan, K., et al. 2017, Icar, 290, 14 Fuller, J., Luan, J., & Quataert, E. 2016,, 458, 3867 Hamilton, D. P., & Ward, W. R. 2004,, 128, 2510 Helled, R., Schubert, G., & Anderson, J. D. 2009, Icar, 199, 368 Innanen, K. A., Zheng, J. Q., Mikkola, S., & Valtonen, M. J. 1997,, 113, 1915 Kozai, Y. 1962,, 67, 446 Lainey, V., Karatekin, [Ö]{}., Desmars, J., et al. 2012,, 752, 14 Lainey, V., Jacobson, R. A., Tajeddine, R., et al. 2015, Icar, 281, 286 Laskar, J., Fienga, A., Gastineau, M., & Manche, H. 2011,, 532, A89 Lee, M. H., & Peale, S. J. 2003,, 592, 1201

, M., Dones, L., & Nesvorn[ý ] { }, D.   2013, arXiv:1311.6780 uk, M., Dones, L., & Nesvorn[ý ] { }, D.   2016, , 820, 97 uk, M., Hamilton, D.   P., Lock, S.   J., & Stewart, S.   T.   2016, Natur, 539, 402 Danby, J.   M.   A.   1992.   Fundamentals of celestial mechanics.   Richmond: Willman - Bell, 1992, 2nd ed. Dombard, A.   J., Cheng, A.   F., McKinnon, W.   B., & Kay, J.   P.   2012, Journal of Geophysical Research (Planets), 117, E03002 El Moutamid, M., Hedman, M.   M., & Nicholson, P.   D.   2017, AAS / Division for Planetary Sciences Meeting Abstracts \#49, 49, 212.02 French, R.   G., McGhee - French, C.   A., Lonergan, K., et al.   2017, Icar, 290, 14 Fuller, J., Luan, J., & Quataert, E.   2016, , 458, 3867 Hamilton, D.   P., & Ward, W.   R.   2004, , 128, 2510 Helled, R., Schubert, G., & Anderson, J.   D.   2009, Icar, 199, 368 Innanen, K.   A., Zheng, J.   Q., Mikkola, S., & Valtonen, M.   J.   1997, , 113, 1915 Kozai, Y.   1962, , 67, 446 Lainey, V., Karatekin, [ Ö ] { }. , Desmars, J., et al.   2012, , 752, 14 Lainey, V., Jacobson, R.   A., Tajeddine, R., et al.   2015, Icar, 281, 286 Laskar, J., Fienga, A., Gastineau, M., & Manche, H.   2011, , 532, A89 Lee, M.   H., & Peale, S.   J.   2003, , 592, 1201

, M., Fones, L., & Nesvorn[ý]{}, D. 2013, arXiv:1311.6780 uk, M., Dones, L., & Uwsvorn[ý]{}, D. 2016,, 820, 97 ui, M., Hamiuton, D. P., Lock, S. J., & Stewart, S. T. 2016, Iatue, 539, 402 Dqnby, J. M. A. 1992. Fundamentals ow celestiwl mechabics. Cichmond: Willman-Usll, 1992, 2nd ed. Dojnard, C. J., Cheng, A. F., McKiknon, W. B., & Kaf, J. P. 2012, Journal ox Eelphysical Research (Planets), 117, E03002 El Mottamid, K., Jedman, M. M., & Nicrolspg, P. D. 2017, AAS/Division for Planetary Sciencea Meetiig Abstracts \#49, 49, 212.02 Grench, R. G., McGhee-French, C. A., Poneggan, K., et al. 2017, Icar, 290, 14 Fuller, J., Ouan, H., & Quataert, D. 2016,, 458, 3867 Hamilton, D. P., & Ward, S. R. 2004,, 128, 2510 Helled, R., Schubert, G., & Anderron, J. B. 2009, Icar, 199, 368 Inbaben, N. A., Zheng, J. Q., Mikkjla, S., & Valtokvn, M. J. 1997,, 113, 1915 Nozai, Y. 1962,, 67, 446 Lainey, V., Kavatekmn, [Ö]{}., Eesmars, J., et al. 2012,, 752, 14 Laiiey, V., Jacobson, R. A., Tateddine, R., ec al. 2015, Icar, 281, 286 Laskar, J., Diwnga, D., Gavtindqu, O., & Jaichs, H. 2011,, 532, A89 Lex, M. H., & Peale, S. J. 2003,, 592, 1201

, M., Dones, L., & Nesvorn[ý]{}, D. uk, Dones, L., Nesvorn[ý]{}, D. 2016,, D. Lock, S. J., Stewart, S. T. Natur, 539, 402 Danby, J. M. 1992. Fundamentals of celestial mechanics. Richmond: Willman-Bell, 1992, 2nd ed. Dombard, A. J., A. F., McKinnon, W. B., & Kay, J. P. 2012, Journal of Geophysical (Planets), E03002 Moutamid, Hedman, M. M., & Nicholson, P. D. 2017, AAS/Division for Planetary Sciences Meeting Abstracts \#49, 49, French, R. G., McGhee-French, C. A., Lonergan, K., al. 2017, Icar, 290, Fuller, J., Luan, J., & E. 458, 3867 D. & W. R. 2004,, 2510 Helled, R., Schubert, G., & Anderson, J. D. 2009, Icar, 199, 368 Innanen, K. A., Zheng, Q., Mikkola, Valtonen, M. 1997,, 1915 Y. 1962,, 67, V., Karatekin, [Ö]{}., Desmars, J., et 14 Lainey, V., Jacobson, R. A., Tajeddine, R., al. 2015, 281, 286 Laskar, J., Fienga, A., M., & Manche, H. 2011,, 532, A89 Lee, H., & Peale, S. J. 2003,, 592, 1201

, M., Dones, L., & Nesvorn[ý]{}, D. 2013, arXiv:1311.6780 uk, M., DOnes, L., & NesvoRn[ý]{}, D. 2016,, 820, 97 uK, M., HAmiLtOn, D. P., lock, s. J., & Stewart, S. T. 2016, NatUR, 539, 402 DanBy, J. M. A. 1992. Fundamentals of celEstiaL mEChanICs. richmOnd: WillMAn-bELl, 1992, 2nD eD. DOmbArD, a. J., cheng, a. F., MCKinnon, w. B., & Kay, J. P. 2012, JouRnaL oF Geophysical rEsEarch (PlaneTs), 117, E03002 el Moutamid, M., HEdmAn, M. M., & NiChOlsON, P. D. 2017, AAs/DiVisioN for PlANetary sciences MEeTIng AbsTRacts \#49, 49, 212.02 FrENCh, r. G., Mcghee-French, C. A., LonerGAn, k., Et al. 2017, Icar, 290, 14 Fuller, j., Luan, J., & quATaERT, E. 2016,, 458, 3867 HAmiLton, D. P., & Ward, w. R. 2004,, 128, 2510 helleD, r., SchubeRT, G., & aNDErsON, J. D. 2009, Icar, 199, 368 InnaneN, K. A., Zheng, J. Q., MIKkoLa, S., & ValToNen, m. j. 1997,, 113, 1915 Kozai, y. 1962,, 67, 446 LainEy, v., karAtekin, [Ö]{}., DesmArs, J., Et al. 2012,, 752, 14 LaineY, V., JacoBSon, R. A., TaJEddine, R., Et al. 2015, IcAr, 281, 286 LAskAr, J., FIEnGa, a., GaStINeaU, m., & MAncHE, H. 2011,, 532, A89 lee, M. H., & PeaLe, s. J. 2003,, 592, 1201

, M., Dones, L., & Nesvorn [ý]{}, D.2013, ar Xiv :1 311. 6780 uk, M., Dones , L., & Nesvorn[ý]{}, D. 20 16,,82 0 , 97 uk , M., Hamilt o n, D . P ., L ock ,S .J., & St ewart,S. T. 2016 , N at ur, 539, 402 Da nby, J. M.  A.  1992. Funda men tals o fcel e stial me chani cs. Ri c hmond: Willman- Be l l, 199 2 , 2nd e d . D omba rd, A. J., Cheng, A. F., McKinnon,W. B., & Ka y , J.  P.  2012, Jou rn al of Geophys i ca l R ese a rch (Planets) , 117, E030 0 2 E l Mout am id, M., He dman, M .  M. , & Nichols on,P. D. 201 7, AAS / Divisio n for Pl anetar y S cie nces Me et ing A b str a ct s \ # 49, 49, 212 .0 2 Fren ch,R . G ., M cGh ee-F rench , C. A., Lone rga n, K . , e t al.  2017 , Ic ar , 290 , 14 F uller ,J., Luan, J., & Qua taert, E.  20 16 ,,45 8, 38 6 7 Hami lto n,D. P.,& Ward, W.R. 2 0 04 ,, 128, 2510 Helle d, R ., Schuber t, G., &An d erson, J .D.2009 , Icar, 199 , 3 68 Innan en, K. A. ,Zheng,J.  Q., M ik kol a,S., & Valt onen,M. J. 19 97,,1 13, 1915 Koza i , Y. 1962,, 6 7 ,4 4 6L aine y,V., Karatek in,[ Ö]{} ., D e sm ars , J.,et al .2 01 2 ,, 752, 14 Lainey,V. , Jaco bson, R. A., Tajed dine, R.,e t al. 2015 , Ic a r, 281, 286 Laska r, J. , Fienga,A ., Gasti neau, M., & M anche, H. 2 011,, 53 2,A89 Le e,M .  H ., & Peale, S . J. 2 00 3,, 592 , 1 201

, M.,_Dones, L.,_& Nesvorn[ý]{}, D. 2013, arXiv:1311.6780_uk, M.,_Dones,_L., &_Nesvorn[ý]{},_D. 2016,, 820, 97_uk, M., Hamilton,_D. P., Lock, S. J., &_Stewart, S. T. 2016, Natur,_539,_402 Danby, J. M. A. 1992. Fundamentals of celestial mechanics. Richmond: Willman-Bell, 1992, 2nd ed. Dombard, A. J., Cheng, A. F.,_McKinnon,_W. B., &_Kay,_J. P. 2012,_Journal of Geophysical Research (Planets),_117, E03002 El Moutamid, M.,_Hedman, M. M.,_& Nicholson, P. D. 2017, AAS/Division for Planetary Sciences Meeting_Abstracts_\#49, 49, 212.02 French,_R. G., McGhee-French, C. A., Lonergan, K., et al. 2017, Icar, 290,_14 Fuller, J., Luan, J., &_Quataert, E. 2016,, 458,_3867_Hamilton,_D. P., & Ward, W. R. 2004,,_128, 2510 Helled, R., Schubert, G.,_& Anderson, J. D. 2009, Icar, 199, 368_Innanen, K. A., Zheng, J. Q., Mikkola, S., &_Valtonen, M. J. 1997,, 113, 1915 Kozai, Y. 1962,, 67,_446 Lainey, V., Karatekin, [Ö]{}.,_Desmars, J.,_et al. 2012,, 752, 14 Lainey,_V., Jacobson, R. A.,_Tajeddine, R.,_et al. 2015, Icar,_281, 286 Laskar, J., Fienga, A.,_Gastineau, M., &_Manche, H. 2011,, 532, A89 Lee, M. H.,_&_Peale, S. J. 2003,, 592,_1201

. Our way of obtaining the Kerr nonlinearity numerically agrees well with those seen in experiments [@13_Observation_Schoelkopf]. The coherence times mentioned in ref. [@07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino] are for a bare cavity. But due to the hybridization of the cavity-SQ modes, the lifetimes of both the cavity and the SQ change. Since a SQ has a shorter coherence time than our desired kind of cavities, the SQ-cavity interaction will shorten the lifetime of the cavity. We therefore need to take this effect into account. To the lowest order in $g/\Delta$, the ‘inverse-Purcell’ enhanced decay rate of the cavity $\kappa \approx (1-(g/\Delta)^2)\kappa_c+(g/\Delta)^2 \gamma$ [@07_Generating_Schoelkopf; @16_Quantum_Schoelkopf; @09_Dispersive_Blais]. Here, $g$ is the qubit-cavity coupling strength, $\Delta=\omega_q-\omega_c$ is the detuning between the cavity and the qubit, and $\kappa_c$ and $\gamma$ are the decay rates of the bare cavity and the SQ respectively. Taking the above $\kappa$, the effective coherence time of a cat state in the cavity in the presence of the two photon drive, $\kappa_{eff}$, is numerically found by fitting the evolution of coherence as a decaying exponential. It is approximately equal to $2 \kappa {\left\lvert\alpha\right\rvert}^2$, where ${\left\lvert\alpha\right\rvert}$ is the size of the coherent state. For experimentally conceivable values of $\kappa_c$ [@07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino], $\gamma$ [@19_SQs_Oliver; @12_SQ_Steffen], $g$ [@07_cQED_Schoelkopf; @13_Observation_Schoelkopf; @17_Microwave_Nori], $\Delta$ [@13_Observation_Schoelkopf; @16_Quantum_Schoelkopf], and $K_q$ [@12_

. Our way of obtaining the Kerr nonlinearity numerically agrees well with those seen in experiment [ @13_Observation_Schoelkopf ]. The coherence prison term mentioned in ref.   [ @07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino ] are for a bare pit. But due to the hybridization of the cavity - SQ modes, the lifetimes of both the pit and the SQ variety. Since a SQ has a shorter coherence time than our desired kind of cavities, the SQ - pit interaction will shorten the lifetime of the cavity. We therefore necessitate to take this effect into account. To the lowest order in $ g/\Delta$, the ‘ inverse - Purcell ’ enhance decay rate of the cavity $ \kappa \approx (1-(g/\Delta)^2)\kappa_c+(g/\Delta)^2 \gamma$ [ @07_Generating_Schoelkopf; @16_Quantum_Schoelkopf; @09_Dispersive_Blais ]. Here, $ g$ is the qubit - cavity yoke strength, $ \Delta=\omega_q-\omega_c$ is the detuning between the cavity and the qubit, and $ \kappa_c$ and $ \gamma$ are the decay rate of the bare pit and the SQ respectively. Taking the above $ \kappa$, the effective coherence prison term of a cat state in the cavity in the bearing of the two photon drive, $ \kappa_{eff}$, is numerically found by fitting the evolution of coherence as a decaying exponential. It is approximately equal to $ 2 \kappa { \left\lvert\alpha\right\rvert}^2 $, where $ { \left\lvert\alpha\right\rvert}$ is the size of the coherent department of state. For experimentally conceivable values of $ \kappa_c$ [ @07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino ], $ \gamma$ [ @19_SQs_Oliver; @12_SQ_Steffen ], $ g$ [ @07_cQED_Schoelkopf; @13_Observation_Schoelkopf; @17_Microwave_Nori ], $ \Delta$ [ @13_Observation_Schoelkopf; @16_Quantum_Schoelkopf ], and $ K_q$ [ @12 _

. Oug way of obtaining the Ktrr nonlinearity uymericelly agdees welu with those seen in experimxnts [@13_Obseevation_Schoelkopf]. The cuherence nimes menriontd in ref. [@07_Ultrahigi_Bisentik; @13_Readming_Sehielkopf; @18_3D_Grassgllino] are fmr a bare cavidy. Bbt due to the hybridization of the cwvity-SQ mldes, the lifetymes jf bknh the cavity and the SQ change. Since e SQ has a shoryer coherence time than oug dedired kind of cavihies, the SQ-xaviet interactiov will shogcen the lifgtime of the cavity. We therefore need to take tyiw evxect into arcount. No the lowest order it $g/\Delts$, the ‘inverse-Pmrceln’ ebhanced decay rate of the cavity $\kappa \wpprox (1-(g/\Dalca)^2)\kappa_c+(g/\Delta)^2 \gamma$ [@07_Teberathng_Swhoeujopw; @16_Qhaituj_Schoepko'f; @09_Dispersibe_Blais]. Herw, $g$ is the qubit-cavotr coupling strsngth, $\Qejta=\omega_q-\omega_c$ is the detuning between ths cavity and the qubit, qnd $\kappa_c$ and $\gamma$ wre the dqcay rates of the bare cavity and the SQ respectieely. Vayiny the xvoge $\kappa$, the effective coherence time of a cae suatv in the cavity ik the presence of yhf yro photon drixe, $\kap'z_{erf}$, is numerically vound bi fittung the edoluyion of coherence as a decating exponennial. It is approximatepy equal to $2 \kakpa {\legt\lvert\alpha\right\rvert}^2$, chere ${\meft\lvert\aloha\right\rbdrt}$ is the size uf nhe woherent state. For experimegtally coiceivcble valjes pf $\kap[a_c$ [@07_Ultrahlgh_Visentin; @13_Reaching_Schoflkopy; @18_3D_Grdssellino], $\hamma$ [@19_SQs_Oliver; @12_SQ_Steffen], $g$ [@07_cQEV_Xchoelkopf; @13_Onsarvdtion_Schjelkoif; @17_Microwave_Nory], $\Delta$ [@13_Observction_Schjelkoof; @16_Quantum_Achoelkmpf], and $K_q$ [@12_

. Our way of obtaining the Kerr agrees with those in experiments [@13_Observation_Schoelkopf]. ref. @13_Reaching_Schoelkopf; @18_3D_Grassellino] are a bare cavity. due to the hybridization of the modes, the lifetimes of both the cavity and the SQ change. Since a has a shorter coherence time than our desired kind of cavities, the SQ-cavity will the of cavity. We therefore need to take this effect into account. To the lowest order in $g/\Delta$, ‘inverse-Purcell’ enhanced decay rate of the cavity $\kappa (1-(g/\Delta)^2)\kappa_c+(g/\Delta)^2 \gamma$ [@07_Generating_Schoelkopf; @16_Quantum_Schoelkopf; Here, $g$ is the qubit-cavity strength, is the between cavity the qubit, and and $\gamma$ are the decay rates of the bare cavity and the SQ respectively. Taking the above the effective of a state the in the presence two photon drive, $\kappa_{eff}$, is numerically the evolution of coherence as a decaying exponential. is approximately to $2 \kappa {\left\lvert\alpha\right\rvert}^2$, where ${\left\lvert\alpha\right\rvert}$ the size of the coherent state. For experimentally values of $\kappa_c$ [@07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino], $\gamma$ [@19_SQs_Oliver; @12_SQ_Steffen], $g$ [@07_cQED_Schoelkopf; @13_Observation_Schoelkopf; @17_Microwave_Nori], $\Delta$ [@13_Observation_Schoelkopf; $K_q$ [@12_

. Our way of obtaining the Kerr nOnlinearitY numeRicAllY aGreeS welL with those seen IN expEriments [@13_Observation_SchOelkoPf]. tHe coHErEnce tImes menTIoNED in ReF. [@07_ULtrAhIGh_visenTin; @13_reachinG_SchoelkopF; @18_3D_GRaSsellino] are fOR a Bare cavity. but Due to the hybrIdiZation Of The CAvity-sQ mOdes, tHe lifeTImes of Both the caViTY and thE sQ changE. sInCe a Sq has a shorter coherENcE Time than our desIred kiNd OF cAVItiEs, tHe SQ-cavity InTeracTIon will SHoRTEN thE Lifetime of the Cavity. We theREfoRe need To TakE This efFect iNtO AccOunt. To the loWest Order in $g/\DElta$, thE ‘Inverse-pUrcell’ eNhanceD deCay Rate OF tHe CavItY $\KapPA \aPprOX (1-(g/\DElta)^2)\kappA_c+(G/\DElta)^2 \gAmma$ [@07_gENERatiNg_SChoeLkopf; @16_quantum_SchoelKopF; @09_DisPErsIve_BlAis]. HeRe, $g$ iS tHe qubIt-caviTy couPlIng strength, $\DeltA=\omeGa_q-\omega_c$ Is tHe DetUnIng beTWeen thE caVitY and the Qubit, anD $\KapPa_C$ AND $\gAmma$ are the decay ratEs OF ThE bare cavIty and THe sQ REspectivElY. TaKing THE abovE $\kapPA$, tHe effectIve cohEReNcE time of A cAt statE iN thE caVity iN The pResencE of the twO photON drive, $\kappa_{eff}$, IS numerically fOUnD BY fITtinG thE evolution oF cohERencE as a DEcAyiNG expoNentiAl. iT iS Approximately equal tO $2 \kAppa {\leFt\lveRt\alpha\right\rVert}^2$, where ${\lEFT\Lvert\alpHa\riGHt\RVert}$ is the size oF the cOherent staTE. For expeRimenTally conCeivable vALUes of $\kapPa_c$ [@07_ultRahIgh_vISeNtin; @13_Reaching_SCHOelkOpF; @18_3D_GrassEllIno], $\gammA$ [@19_SQS_OlIveR; @12_SQ_stEffen], $g$ [@07_cQEd_SchoelkOpF; @13_OBsErVatIon_ScHOelkopf; @17_MIcRowAvE_NoRi], $\DelTA$ [@13_ObserVatioN_SchOeLkOPf; @16_QUantum_SCHoELKopf], AnD $K_Q$ [@12_

. Our way of obtaining the Kerr nonl inear ity nu me rica llyagrees well wi t h th ose seen in experiment s [@1 3_ O bser v at ion_S choelko p f] . Th eco her en c etimes me ntioned in ref. [ @07 _U ltrahigh_Vis e nt in; @13_Re ach ing_Schoelko pf; @18_3 D_ Gra s selli no] arefor ab are ca vity. But d u e to t h e hybri d i za tion of the cavity-SQ mo d es, the lifeti mes of b o th t hecav ity and th eSQ ch a nge. Si n ce a SQh as a shortercoherence t i methan o ur de s ired k ind o fc avi ties, the S Q-ca vity inte ractio n will s h orten t he lif eti meof t h eca vit y. Wet he ref o reneed tota ke this eff e c t into ac coun t. T o the lowestord er i n $g /\Del ta$,the‘i nvers e-Purc ell’en hanced decay ra te o f the cav ity $ \ka pp a \ap p rox (1 -(g /\D elta)^2 )\kappa _ c+( g/ \ D e lt a)^2 \gamma$ [@07_ Ge n e ra ting_Sch oelkop f ;@1 6 _Quantum _S cho elko p f ; @09 _Dis p er sive_Bla is]. H e re ,$g$ isth e qubi t- cav ity coup l ingstreng th, $\De lta=\ o mega_q-\omega_ c $ is the detu n in g be t ween th e cavity an d th e qub it,a nd $\ k appa_ c$ an d$ \g a mma$ are the decayra tes of thebare cavity a nd the SQr e s pectivel y. T a ki n g the above $\ kappa $, the eff e ctive co heren ce timeof a cats t ate in t hecav ity in t he presence oft h e tw ophotondri ve, $\k app a_{ eff }$, i s numeric ally fou nd b yfi tti ng th e evoluti on of c ohe rence as a d ecayi ng e xp on e nti al. Iti sa p prox im at elyequ al to $ 2 \k a ppa {\left \lvert\al pha \ righ t\ rv ert}^2$ , where ${\le ft \lvert\alp ha \ri ght\rv e r t}$ is t he size of the coherent state. Fo r exp erim entally c onc eivabl e v a lues o f $\ka ppa_c $[@0 7 _ Ultra h i gh _Vi se ntin; @13_ R e ach ing_S ch oelk opf; @1 8_3D_Grassellino], $\g amma$ [@19_SQ s_O live r ; @ 12_ S Q_ S tef fe n ],$ g $ [@07_cQED_Sch oelkopf; @ 13 _ Ob servation_ S cho el kopf; @ 17_Micr owave _ Nori],$\Delta$[@13_Obse rv atio n _ Sch oelkopf; @ 16_Quant um_Schoel k opf], an d $K_ q$[@12_

. Our_way of_obtaining the Kerr nonlinearity_numerically agrees_well_with those_seen_in experiments [@13_Observation_Schoelkopf]. The_coherence times mentioned_in ref. [@07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino]_are for a_bare_cavity. But due to the hybridization of the cavity-SQ modes, the lifetimes of both_the_cavity and_the_SQ_change. Since a SQ has_a shorter coherence time than_our desired_kind of cavities, the SQ-cavity interaction will shorten_the_lifetime of the_cavity. We therefore need to take this effect into_account. To the lowest order in $g/\Delta$,_the ‘inverse-Purcell’ enhanced_decay_rate_of the cavity $\kappa_\approx (1-(g/\Delta)^2)\kappa_c+(g/\Delta)^2 \gamma$ [@07_Generating_Schoelkopf; @16_Quantum_Schoelkopf; @09_Dispersive_Blais]._Here, $g$ is the qubit-cavity coupling_strength, $\Delta=\omega_q-\omega_c$ is the detuning between the_cavity and the qubit, and $\kappa_c$_and $\gamma$ are the decay_rates of_the bare cavity and the_SQ respectively. Taking_the above_$\kappa$, the effective_coherence time of a cat state_in the cavity_in the presence of the two_photon_drive, $\kappa_{eff}$, is_numerically_found_by fitting_the evolution of_coherence_as a_decaying_exponential. It is approximately equal to_$2_\kappa {\left\lvert\alpha\right\rvert}^2$, where ${\left\lvert\alpha\right\rvert}$ is the size_of the coherent state. For_experimentally_conceivable values of $\kappa_c$_[@07_Ultrahigh_Visentin; @13_Reaching_Schoelkopf; @18_3D_Grassellino], $\gamma$ [@19_SQs_Oliver;_@12_SQ_Steffen], $g$ [@07_cQED_Schoelkopf; @13_Observation_Schoelkopf; @17_Microwave_Nori], $\Delta$_[@13_Observation_Schoelkopf; @16_Quantum_Schoelkopf],_and $K_q$_[@12_

_{\a,\b} c_{\a,\b} z^{\a} (\log(z))^{\b} h_{\a,\b}(z)$$ where $\a \in \BQ$, $\b \in \BN$, and $h_{\a,\b}(z) \in \BC\{z\}_0$, where $\BC\{z\}_0$ is the ring of power series convergent at $z=0$ (identified with the ring of germs of functions analytic at $z=0$). See for example, [@Ml]. Series of the form are often known in the literature as series of the [*Nilsson class*]{}; see [@Ni1; @Ni2]. The rationality of the exponents $\{\a\}$ above is an important feature that always appears in algebraic geometry and arithmetic and rarely appears in analysis. For a further discussion; see [@Ga5]. We say that $G(z)=\sum_{n=0}^\infty a_n z^n$ is an [*resurgent series*]{} (and write $G(z) \in \RES$) if - $G(z)$ is convergent at $z=0$. - $G(z)$ has analytic continuation as a multivalued function in $\BC\setminus \La$, where $\La \subset \BC$ is a discrete subset of $\BC$. - The local monodromy is quasi-unipotent. In what follows, we will make little distinction between a germ, its analytic continuation, and the corresponding function. So, we will speak about the algebra of resurgent functions. Gevrey series of mixed type --------------------------- The following definition is motivated by the properties of some power series that are associated to knotted 3-dimensional objects. We say that series $G(z)=\sum_{n=0}^\infty a_n z^n$ is a [*Gevrey series of mixed type*]{} $(r,s)$ if - $r,s \in \BQ$ and the coefficients $a_n$ lie in a number field $K$, and - there exists a constant $C>0$ so that for every $n \in \BN$ the absolute value of every Galois conjugate of $a_n$ is less than or equal to $C^n n!^r$, and - the common denominator

_ { \a,\b } c_{\a,\b } z^{\a } (\log(z))^{\b } h_{\a,\b}(z)$$ where $ \a \in \BQ$, $ \b \in \BN$, and $ h_{\a,\b}(z) \in \BC\{z\}_0 $, where $ \BC\{z\}_0 $ is the ring of power series convergent at $ z=0 $ (identified with the gang of microbe of functions analytic at $ z=0 $). See for example, [ @Ml ]. Series of the shape are often known in the literature as serial of the [ * Nilsson class * ] { }; see [ @Ni1; @Ni2 ]. The rationality of the exponents $ \{\a\}$ above is an significant feature that always appear in algebraic geometry and arithmetic and rarely appears in psychoanalysis. For a further discussion; see [ @Ga5 ]. We say that $ G(z)=\sum_{n=0}^\infty a_n z^n$ is an [ * resurgent series * ] { } (and write $ G(z) \in \RES$) if - $ G(z)$ is convergent at $ z=0$. - $ G(z)$ has analytic continuation as a multivalued function in $ \BC\setminus \La$, where $ \La \subset \BC$ is a discrete subset of $ \BC$. - The local monodromy is quasi - unipotent. In what succeed, we will make little eminence between a germ, its analytic continuation, and the corresponding function. thus, we will speak about the algebra of resurgent functions. Gevrey series of mixed type --------------------------- The follow definition is motivated by the properties of some power series that are associated to ravel 3 - dimensional objects. We say that series $ G(z)=\sum_{n=0}^\infty a_n z^n$ is a [ * Gevrey series of mixed type * ] { } $ (r, s)$ if - $ r, s \in \BQ$ and the coefficients $ a_n$ lie in a number battlefield $ K$, and - there exists a constant $ C>0 $ so that for every $ normality \in \BN$ the absolute value of every Galois conjugate solution of $ a_n$ is less than or adequate to $ C^n n!^r$, and - the common denominator

_{\a,\b} f_{\a,\b} z^{\a} (\log(z))^{\b} h_{\a,\b}(z)$$ where $\x \in \BQ$, $\b \in \BN$, and $h_{\a,\u}(z) \in \BD\{z\}_0$, where $\BC\{z\}_0$ is the ring of power secies convtggent at $z=0$ (identified with the ring of gernw of functmkns analytic zb $z=0$). Sze for example, [@Mk]. Series ox the form are owtzn known in the literature as series of the [*Nllsson class*]{}; sge [@Ni1; @Gi2]. Ths rationality of the exponents $\{\a\}$ agove is an important feature that always appeags ij algebraic geometgy and aritymetyx and rarely appears in analysis. Fkr a further discussion; see [@Ga5]. Wd say that $G(z)=\sun_{n=0}^\unfhi a_n z^n$ is ai [*resuggent series*]{} (and writa $G(z) \in \RES$) if - $G(z)$ ix cmnvwrgent at $z=0$. - $G(z)$ has enalytic continuatiog as a muntnvalued function in $\BX\swtminos \La$, dyerd $\Lz \xugset \BF$ ia a discrefe subset od $\BC$. - The local momoqgpmy is quasi-hnipotqne. In what follows, we will make little divtihction between a germ, irs analytic continuatlon, and tre corresponding function. So, we will speak about dhe ameebxq of fwskrgent functions. Gevrey series of mixed type --------------------------- Thq fplkowing definitlon is motivated bu hhr properties ow some poser series that arf assocyated to knotttd 3-dikensional objects. We say thar series $G(z)=\slm_{n=0}^\ibfty a_n z^n$ is a [*Gerrey series uf moxed yype*]{} $(r,s)$ if - $r,s \in \BQ$ aud the coefficienhs $a_n$ lie kn a number fiela $K$, atd - there exists a constagt $C>0$ so tiat fpr everh $n \on \BN$ ehe absoluhe value of every Galois conjogate mf $a_n$ is lfss than or equal to $C^n n!^r$, and - the common dgnokindtor

_{\a,\b} c_{\a,\b} z^{\a} (\log(z))^{\b} h_{\a,\b}(z)$$ where $\a $\b \BN$, and \in \BC\{z\}_0$, where power convergent at $z=0$ with the ring germs of functions analytic at $z=0$). for example, [@Ml]. Series of the form are often known in the literature series of the [*Nilsson class*]{}; see [@Ni1; @Ni2]. The rationality of the exponents above an feature always appears in algebraic geometry and arithmetic and rarely appears in analysis. For a further discussion; [@Ga5]. We say that $G(z)=\sum_{n=0}^\infty a_n z^n$ is [*resurgent series*]{} (and write \in \RES$) if - $G(z)$ convergent $z=0$. - has continuation a multivalued function $\BC\setminus \La$, where $\La \subset \BC$ is a discrete subset of $\BC$. - The local monodromy is In what will make distinction a its analytic continuation, corresponding function. So, we will speak of resurgent functions. Gevrey series of mixed type The following is motivated by the properties of power series that are associated to knotted 3-dimensional We say that series $G(z)=\sum_{n=0}^\infty a_n z^n$ is a [*Gevrey series of mixed type*]{} $(r,s)$ $r,s \in \BQ$ and coefficients $a_n$ lie a field and there exists constant $C>0$ so that for every $n \in \BN$ the absolute of every Galois conjugate of $a_n$ is less than or $C^n and - the denominator

_{\a,\b} c_{\a,\b} z^{\a} (\log(z))^{\b} h_{\a,\b}(z)$$ where $\a \in \Bq$, $\b \in \BN$, and $h_{\A,\b}(z) \in \bC\{z\}_0$, WheRe $\bC\{z\}_0$ iS the Ring of power serIEs coNvergent at $z=0$ (identified wIth thE rINg of GErMs of fUnctionS AnALYtiC aT $z=0$). see FoR ExAmple, [@ml]. SEries of The form are OftEn Known in the liTErAture as serIes Of the [*Nilsson ClaSs*]{}; see [@NI1; @NI2]. ThE RatioNalIty of The expONents $\{\a\}$ Above is an ImPOrtant FEature tHAT aLwayS appears in algebraIC gEOmetry and arithMetic aNd RArELY apPeaRs in analysIs. for a fURther diSCuSSIOn; sEE [@Ga5]. We say that $G(Z)=\sum_{n=0}^\infty a_N Z^n$ iS an [*resUrGenT Series*]{} (And wrItE $g(z) \iN \RES$) if - $G(z)$ is cOnveRgent at $z=0$. - $G(Z)$ has anALytic coNTinuatiOn as a mUltIvaLued FUnCtIon In $\bc\seTMiNus \lA$, whEre $\La \subSeT \Bc$ is a dIscrETE SUbseT of $\bC$. - ThE locaL monodromy is qUasI-uniPOteNt. In wHat foLlowS, wE will Make liTtle dIsTinction between A gerM, its analyTic CoNtiNuAtion, ANd the cOrrEspOnding fUnction. sO, we WiLL SPeAk about the algebra oF rESUrGent funcTions. GEVrEy SEries of mIxEd tYpe --------------------------- THE FolloWing DEfInition iS motivATeD bY the proPeRties oF sOme PowEr serIEs thAt are aSsociateD to knOTted 3-dimensionaL Objects. We say tHAt SERiES $G(z)=\sUm_{n=0}^\Infty a_n z^n$ is A [*GevREy seRies OF mIxeD Type*]{} $(r,S)$ if - $r,s \In \bq$ aND the coefficients $a_n$ lIe In a numBer fiEld $K$, and - there eXists a consTANT $C>0$ so that For eVErY $N \in \BN$ the absoluTe valUe of every GALois conjUgate Of $a_n$ is leSs than or eQUAl to $C^n n!^r$, And - The ComMon DENoMinator

_{\a,\b} c_{\a,\b} z^{\a}(\log(z))^ {\b}h_{ \a, \b }(z) $$ w here $\a \in \ B Q$,$\b \in \BN$, and $h_{ \a,\b }( z ) \i n \ BC\{z \}_0$,w he r e $\ BC \{ z\} _0 $ i s the ri ng of p ower serie s c on vergent at $ z =0 $ (identif ied with the ri ngof ger ms of funct ion s ana lytica t $z=0 $). See f or exampl e , [@Ml] . Se ries of the form areo ft e n known in the liter at u re a s s eri es of the[* Nilss o n class * ]{ } ; see [@Ni1; @Ni2]. The ratio n ali ty ofth e e x ponent s $\{ \a \ }$above is an imp ortant fe aturet hat alw a ys appe ars in al geb raic ge om etr ya nda ri thm e tic and rar el yappea rs i n a n alys is. For a fu rther discuss ion ; se e [@ Ga5]. Wesayth at $G (z)=\s um_{n =0 }^\infty a_n z^ n$ i s an [*re sur ge ntse ries* ] {} (an d w rit e $G(z) \in \R E S$) i f - $G(z)$ is conver ge n t a t $z=0$. - $ G( z) $ has ana ly tic con t i nuati on a s a multiva lued f u nc ti on in $ \B C\setm in us \La$ , whe re $\L a \subse t \BC $ is a discrete subset of $\B C $. - The lo cal monodro my i s qua si-u n ip ote n t. I n wha tf ol l ows, we will make l it tle di stinc tion betweena germ, it s a nalyticcont i nu a tion, and thecorre sponding f u nction.So, w e will s peak abou t the alge bra of re sur g e nt functions. G e vrey s eries o f m ixed ty pe--- --- --- -- --------- ------- T he f ol low ing d e finition i s m ot iva ted b y the p roper ties o fs ome powers er i e s th at a re a sso ci atedto k n ott ed 3-di mensional ob j ects . W e say t hat series $G (z )=\sum_{n= 0} ^\i nfty a _ n z^n$ is a [*Gevrey series of m i xed typ e*] {} $( r,s) $ if - $r ,s \in \B Q $ andthe co effic ie nts $ a_n$l i einanumber fie l d $K $, an d - thereexists a constant$ C>0 $ so that for ev ery$ n \ in\ BN $ th ea bso l u te value of eve ry Galoisco n ju gate of $a _ n$is less t han orequal to $C^n n!^r$, a nd - t he com m o n d enominator

_{\a,\b} c_{\a,\b}_z^{\a} (\log(z))^{\b}_h_{\a,\b}(z)$$ where $\a \in_\BQ$, $\b_\in_\BN$, and_$h_{\a,\b}(z)_\in \BC\{z\}_0$, where_$\BC\{z\}_0$ is the_ring of power series_convergent at $z=0$_(identified_with the ring of germs of functions analytic at $z=0$). See for example, [@Ml]._Series_of the_form_are_often known in the literature_as series of the [*Nilsson_class*]{}; see_[@Ni1; @Ni2]. The rationality of the exponents $\{\a\}$ above_is_an important feature_that always appears in algebraic geometry and arithmetic and_rarely appears in analysis. For a_further discussion; see_[@Ga5]. We_say_that $G(z)=\sum_{n=0}^\infty a_n z^n$_is an [*resurgent series*]{} (and write_$G(z) \in \RES$) if - _$G(z)$ is convergent at $z=0$. - _$G(z)$ has analytic continuation as a_multivalued function in $\BC\setminus _ _\La$, where $\La \subset \BC$_is a discrete_subset of_$\BC$. - _The local monodromy is quasi-unipotent. In what_follows, we will_make little distinction between a germ,_its_analytic continuation, and_the_corresponding_function. So,_we will speak_about_the algebra_of_resurgent functions. Gevrey series of mixed type --------------------------- The_following_definition is motivated by the properties of_some power series that_are_associated to knotted 3-dimensional_objects. We say that series $G(z)=\sum_{n=0}^\infty_a_n z^n$ is a [*Gevrey series_of mixed_type*]{} $(r,s)$_if - $r,s \in \BQ$ and the coefficients $a_n$ lie_in a number field $K$, and -_ there exists_a constant_$C>0$_so that for_every_$n \in_\BN$ the absolute value of every Galois_conjugate of_$a_n$ is less than or equal_to $C^n n!^r$, and -__ the common denominator

z_1,z_2,\ldots) &=& %\tilde{\cal{L}}_{G}(\lambda;y_1,\ldots,y_{N(t)}) &=& \tilde{\cal{L}}_{RSL}(\lambda,\theta;y_1,\ldots,y_{N(t)}) &=& \lambda^{N} e^{-\lambda \sum z_i} %\prod_{i=1}^{N(t)} g(y_i;\lambda). %\prod_{i=1}^{N(t)} f(y_i;\lambda). %\prod_{i=1}^{N(t)} f_G(y_i;\lambda). %\prod_{i=1}^{N(t)} f_{RSL}(y_i;\lambda). \prod_{i=1}^{N(t)} f_{RSL}(y_i;\lambda,\theta).\end{aligned}$$ [@hall1] shows that the Laplace transform $\gamma$ of $Y$ is $$\gamma(s) = 1+ \frac{s}{\lambda}- %\left(\lambda\int_0^\infty \exp\{-st - \lambda \int_0^t\{1-G(x)\}dx\}dt\right)^{-1}.$$ (\_0\^{-st - \_0\^t{1-G(x;)}dx}dt)\^[-1]{}.$$ [@stadje] obtains the clump-length distribution function $F_{RSL}(y)$ by inversion of $\gamma$, but it is an infinite sum of self-convolutions of a function that may involve an integral with no analytic solution, making inference based on $\tilde{\cal{L}}_{RSL}$ difficult. M-estimation ------------ An important issue in estimation of $\lambda$ is robustness under model misspecification. Inspection of the clump-lengths from the experimental data, such as the run depicted in Figure 1, reveals that the number of clumps with lengths slightly in excess of $t_0$ is greater than expected, so that the distribution between $t_0$ and $2t_0$ is not uniform. This can be caused by variability in diameter or velocity or by errors of measurement. A desirable property for any estimator is robustness to this departure from model assumptions. For mean segment length $\mu$, the mean clump-length is given by $$E(Y;\lambda) = \frac

z_1,z_2,\ldots) & = & % \tilde{\cal{L}}_{G}(\lambda;y_1,\ldots, y_{N(t) }) & = & \tilde{\cal{L}}_{RSL}(\lambda,\theta;y_1,\ldots, y_{N(t) }) & = & \lambda^{N } e^{-\lambda \sum z_i } % \prod_{i=1}^{N(t) } g(y_i;\lambda). % \prod_{i=1}^{N(t) } f(y_i;\lambda). % \prod_{i=1}^{N(t) } f_G(y_i;\lambda). % \prod_{i=1}^{N(t) } f_{RSL}(y_i;\lambda). \prod_{i=1}^{N(t) } f_{RSL}(y_i;\lambda,\theta).\end{aligned}$$ [ @hall1 ] shows that the Laplace transform $ \gamma$ of $ Y$ is $ $ \gamma(s) = 1 + \frac{s}{\lambda}- % \left(\lambda\int_0^\infty \exp\{-st - \lambda \int_0^t\{1 - G(x)\}dx\}dt\right)^{-1}.$$ (\_0\^{-st - \_0\^t{1 - G(x;)}dx}dt)\^[-1]{}.$$ [ @stadje ] obtains the clump - duration distribution affair $ F_{RSL}(y)$ by inversion of $ \gamma$, but it is an infinite sum of self - gyrus of a function that may involve an built-in with no analytic solution, making inference based on $ \tilde{\cal{L}}_{RSL}$ difficult. M - appraisal ------------ An important issue in estimate of $ \lambda$ is robustness under model misspecification. Inspection of the clump - length from the experimental data, such as the run depicted in Figure 1, reveals that the number of clumps with length slightly in excess of $ t_0 $ is greater than expect, so that the distribution between $ t_0 $ and $ 2t_0 $ is not uniform. This can be caused by variability in diameter or velocity or by mistake of measurement. A desirable property for any estimator is robustness to this departure from model assumptions. For hateful segment length $ \mu$, the mean clump - length is move over by $ $ E(Y;\lambda) = \frac

z_1,z_2,\lfots) &=& %\tilde{\cal{L}}_{G}(\lambda;y_1,\lduts,y_{N(t)}) &=& \tilde{\cal{L}}_{RSL}(\lamuda,\thetz;y_1,\ldots,y_{V(t)}) &=& \lambda^{N} e^{-\lambda \sum z_i} %\prov_{i=1}^{N(t)} g(y_i;\lqmbda). %\prod_{i=1}^{N(t)} f(y_i;\lambda). %\orod_{i=1}^{N(t)} f_H(y_i;\lambdq). %\prov_{i=1}^{N(t)} f_{RSL}(y_i;\lambde). \lrod_{i=1}^{N(t)} f_{RSL}(y_j;\pambba,\vheta).\end{aligned}$$ [@hall1] showv that the Lapnaze transform $\gamma$ of $Y$ is $$\gamma(s) = 1+ \fwac{s}{\lamndw}- %\left(\lambda\int_0^\ynftj \qxp\{-sf - \lambda \int_0^t\{1-G(x)\}dx\}dt\right)^{-1}.$$ (\_0\^{-st - \_0\^t{1-G(x;)}ds}dt)\^[-1]{}.$$ [@stavje] obtains the clump-length distribution vunchion $F_{RSL}(y)$ by invegsion of $\ganma$, fyt it is an knfinite slk of self-cknvolutions of a function that oay iuvolve an ibtwgrwn with no aialytib solution, making infarence nased on $\tilde{\gal{L}}_{RVL}$ eifficult. M-estimation ------------ Ai important issue in estimatimn of $\lambda$ is robysrness undar mueel mias'ecjficatlon. Inspectioh of the clymp-lengths from the evirrimental dafa, sucr ws the run depicted in Figure 1, reveals uhat fhe number of clumps wirh lengths slightly ij excess jf $t_0$ is greater than expected, so that the distribgtion cetcccn $t_0$ qnf $2t_0$ is not uniform. This can be caused by variwgikiny in diameter or velocity ot hy grrors of measorement. A sesirable property for ani estinator is wobuxtness to this departure frim model asslmptuons. For mean segmznt length $\mb$, the kean vlump-length is given by $$E(Y;\lzmbda) = \frac

z_1,z_2,\ldots) &=& %\tilde{\cal{L}}_{G}(\lambda;y_1,\ldots,y_{N(t)}) &=& \tilde{\cal{L}}_{RSL}(\lambda,\theta;y_1,\ldots,y_{N(t)}) &=& \lambda^{N} z_i} g(y_i;\lambda). %\prod_{i=1}^{N(t)} %\prod_{i=1}^{N(t)} f_G(y_i;\lambda). %\prod_{i=1}^{N(t)} that Laplace transform $\gamma$ $Y$ is $$\gamma(s) 1+ \frac{s}{\lambda}- %\left(\lambda\int_0^\infty \exp\{-st - \lambda (\_0\^{-st - \_0\^t{1-G(x;)}dx}dt)\^[-1]{}.$$ [@stadje] obtains the clump-length distribution function $F_{RSL}(y)$ by inversion of but it is an infinite sum of self-convolutions of a function that may an with analytic making inference based on $\tilde{\cal{L}}_{RSL}$ difficult. M-estimation ------------ An important issue in estimation of $\lambda$ is under model misspecification. Inspection of the clump-lengths from experimental data, such as run depicted in Figure 1, that number of with slightly excess of $t_0$ greater than expected, so that the distribution between $t_0$ and $2t_0$ is not uniform. This can be by variability or velocity by of A desirable property estimator is robustness to this departure For mean segment length $\mu$, the mean clump-length given by = \frac

z_1,z_2,\ldots) &=& %\tilde{\cal{L}}_{G}(\lambda;y_1,\lDots,y_{N(t)}) &=& \tilDe{\cal{l}}_{RSl}(\laMbDa,\thEta;y_1,\Ldots,y_{N(t)}) &=& \lambda^{n} E^{-\lamBda \sum z_i} %\prod_{i=1}^{N(t)} g(y_i;\lambDa). %\proD_{i=1}^{n(T)} f(y_i;\LAmBda). %\prOd_{i=1}^{N(t)} f_G(Y_I;\lAMBda). %\PrOd_{I=1}^{N(t)} F_{Rsl}(y_I;\lambDa). \pRod_{i=1}^{N(t)} f_{rSL}(y_i;\lambdA,\thEtA).\end{aligned}$$ [@hALl1] Shows that tHe LAplace transfOrm $\Gamma$ oF $Y$ Is $$\gAMma(s) = 1+ \fRac{S}{\lambDa}- %\left(\LAmbda\iNt_0^\infty \exP\{-sT - \Lambda \INt_0^t\{1-G(x)\}dx\}DT\RiGht)^{-1}.$$ (\_0\^{-sT - \_0\^t{1-G(x;)}dx}dt)\^[-1]{}.$$ [@stadje] obtAInS The clump-length DistriBuTIoN FUncTioN $F_{RSL}(y)$ by inVeRsion OF $\gamma$, bUT iT IS An iNFinite sum of seLf-convolutiONs oF a funcTiOn tHAt may iNvolvE aN IntEgral with no AnalYtic solutIon, makINg inferENce baseD on $\tilDe{\cAl{L}}_{rSL}$ dIFfIcUlt. m-eSTimATiOn ------------ AN ImpOrtant isSuE iN estiMatiON OF $\LambDa$ iS robUstneSs under model mIssPeciFIcaTion. INspecTion Of The clUmp-lenGths fRoM the experimentaL datA, such as thE ruN dEpiCtEd in FIGure 1, reVeaLs tHat the nUmber of CLumPs WITH lEngths slightly in exCeSS Of $T_0$ is greatEr than EXpEcTEd, so that ThE diStriBUTion bEtweEN $t_0$ And $2t_0$ is noT unifoRM. THiS can be cAuSed by vArIabIliTy in dIAmetEr or veLocity or By errORs of measuremenT. a desirable proPErTY FoR Any eStiMator is robuStneSS to tHis dEPaRtuRE from Model AsSUmPTions. For mean segment LeNgth $\mu$, The meAn clump-length Is given by $$E(y;\LAMbda) = \frac

z_1,z_2,\ldots) &=& %\tild e{\cal{L}} _{G}( \la mbd a; y_1, \ldo ts,y_{N(t)}) & = & \t ilde{\cal{L}}_{RSL}(\l ambda ,\ t heta ; y_ 1,\ld ots,y_{ N (t ) } ) & =& \ lam bd a ^{ N} e^ {-\ lambda\sum z_i}%\p ro d_{i=1}^{N(t ) }g(y_i;\lam bda ). %\prod_{i =1} ^{N(t) }f(y _ i;\la mbd a). % \prod_ { i=1}^{ N(t)} f_G (y _ i;\lam b da). %\ p r od _{i= 1}^{N(t)} f_{RSL} ( y_ i ;\lambda). \pr od_{i= 1} ^ {N ( t )}f_{ RSL}(y_i;\ la mbda, \ theta). \ en d { a lig n ed}$$ [@hall1 ] shows tha t th e Lapl ac e t r ansfor m $\g am m a$of $Y$ is $ $\ga mma(s) =1+ \fr a c{s}{\l a mbda}-%\left (\l amb da\i n t_ 0^ \in ft y \e x p\ {-s t -\lambda\i nt _0^t\ {1-G ( x ) \ }dx\ }dt \rig ht)^{ -1}.$$ (\_0\^ {-s t -\ _0\ ^t{1- G(x;) }dx} dt )\^[- 1]{}.$ $ [@s ta dje] obtains th e cl ump-lengt h d is tri bu tionf unctio n $ F_{ RSL}(y) $ by in v ers io n o f$\gamma$, but it i sa n i nfinitesum of se lf - convolut io nsof a f uncti on t h at may inv olve a n i nt egral w it h no a na lyt icsolut i on,making inferen ce ba s ed on $\tilde{ \ cal{L}}_{RSL} $ d i f fi c ult. M -estimation --- - ---- ---- A n i m porta nt is su e i n estimation of $\la mb da$ is robu stness undermodel miss p e c ificatio n. I n sp e ction of the c lump- lengths fr o m the ex perim ental da ta, sucha s the run de pic ted in F ig ure 1, reveal s that t he numb erof clum pswit h l eng th s slightl y in exc es sof $ t_0 $ isg reater t ha n e xp ect ed, s o thatthe d istr ib ut i onbetween $t _ 0 $ an d$2 t_0$ is n ot un ifor m . T his can be cause d b y var ia bi lity in diameter orve locity orby er rors o f measurem ent. A desirable proper t y for a nyestim ator is robus tne ss tothi s depar ture f rom m od ela s sumpt i o ns . F or mean segm e n t l ength $ \mu$ , the m ean clump-length i s gi ven by $$E(Y; \la mbda ) =\fr a c

z_1,z_2,\ldots) &=& %\tilde{\cal{L}}_{G}(\lambda;y_1,\ldots,y_{N(t)})_&=& \tilde{\cal{L}}_{RSL}(\lambda,\theta;y_1,\ldots,y_{N(t)}) &=& \lambda^{N}_e^{-\lambda \sum z_i} %\prod_{i=1}^{N(t)} g(y_i;\lambda). %\prod_{i=1}^{N(t)}_f(y_i;\lambda). %\prod_{i=1}^{N(t)} f_G(y_i;\lambda). %\prod_{i=1}^{N(t)}_f_{RSL}(y_i;\lambda). \prod_{i=1}^{N(t)}_f_{RSL}(y_i;\lambda,\theta).\end{aligned}$$ [@hall1]_shows_that the Laplace_transform $\gamma$ of_$Y$ is $$\gamma(s) =_1+ \frac{s}{\lambda}- %\left(\lambda\int_0^\infty \exp\{-st_-_\lambda \int_0^t\{1-G(x)\}dx\}dt\right)^{-1}.$$ (\_0\^{-st - \_0\^t{1-G(x;)}dx}dt)\^[-1]{}.$$ [@stadje] obtains the clump-length distribution function $F_{RSL}(y)$ by inversion_of_$\gamma$, but_it_is_an infinite sum of self-convolutions_of a function that may_involve an_integral with no analytic solution, making inference based_on_$\tilde{\cal{L}}_{RSL}$ difficult. M-estimation ------------ An important_issue in estimation of $\lambda$ is robustness under model_misspecification. Inspection of the clump-lengths from_the experimental data,_such_as_the run depicted in_Figure 1, reveals that the number_of clumps with lengths slightly in_excess of $t_0$ is greater than expected,_so that the distribution between $t_0$_and $2t_0$ is not uniform._This can_be caused by variability in_diameter or velocity_or by_errors of measurement._A desirable property for any estimator_is robustness to_this departure from model assumptions. For_mean_segment length $\mu$,_the_mean_clump-length is_given by $$E(Y;\lambda)_=_\frac

inequality, is due to Lerman [@Lerman]. Practically speaking, it allows one to control the distance from the $\mu$ center of mass of a window $Q$ to a straight line $\ell$ in terms of the quantity $\beta_2(\mu,Q,\ell)$. \[l:lerman\] Let $\mu$ be a Radon measure on ${\mathbb{R}}^n$, let $Q\subset{\mathbb{R}}^n$ be a bounded Borel set of positive diameter such that $\mu(Q)>0$, and let $$z_Q:=\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\in {\mathbb{R}}^n$$ denote the center of mass of $Q$ with respect to $\mu$. For every straight line $\ell$ in ${\mathbb{R}}^n$, $${\mathop\mathrm{dist}\nolimits}(z_Q,\ell)\leq \beta_2(\mu,Q,\ell){\mathop\mathrm{diam}\nolimits}Q.$$ For every affine subspace $\ell$ in ${\mathbb{R}}^n$, the function ${\mathop\mathrm{dist}\nolimits}(\cdot,\ell)^2$ is convex. Thus, $${\mathop\mathrm{dist}\nolimits}(z_Q,\ell)^2 = {\mathop\mathrm{dist}\nolimits}\left(\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\right)^2\leq \int_Q {\mathop\mathrm{dist}\nolimits}(z,\ell)^2 \frac{d\mu(z)}{\mu(Q)} = \beta_2(\mu,Q,\ell)^2({\mathop\mathrm{diam}\nolimits}Q)^2$$ by Jensen’s inequality. Given two windows $R$ and $Q$ with $R\subset Q$, the approximation number $\beta_2(\mu,Q)$ controls the approximation number $\beta_2(\mu,R)$ if one has control on ${\mathop\mathrm{diam}\nolimits}Q / {\mathop\mathrm{diam}\nolimits}R$ **and** $\mu(Q) / \mu(R)$. \[l:double\] Let $\mu$ be a Radon measure on ${\mathbb{R}}^n$, let $R,Q\subset{\mathbb{R}}^n$ be a bounded Borel set of positive diameter such that $R\subset Q$ and $\mu(R

inequality, is due to Lerman [ @Lerman ]. Practically speaking, it allows one to see the distance from the $ \mu$ plaza of mass of a window $ Q$ to a straight cable $ \ell$ in terms of the quantity $ \beta_2(\mu, Q,\ell)$. \[l: lerman\ ] permit $ \mu$ be a Radon measure on $ { \mathbb{R}}^n$, let $ Q\subset{\mathbb{R}}^n$ be a jump Borel set of positive diameter such that $ \mu(Q)>0 $, and lease $ $ z_Q:=\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\in { \mathbb{R}}^n$$ denote the center of bulk of $ Q$ with respect to $ \mu$. For every straight line $ \ell$ in $ { \mathbb{R}}^n$, $ $ { \mathop\mathrm{dist}\nolimits}(z_Q,\ell)\leq \beta_2(\mu, Q,\ell){\mathop\mathrm{diam}\nolimits}Q.$$ For every affine subspace $ \ell$ in $ { \mathbb{R}}^n$, the function $ { \mathop\mathrm{dist}\nolimits}(\cdot,\ell)^2 $ is convex. Thus, $ $ { \mathop\mathrm{dist}\nolimits}(z_Q,\ell)^2 = { \mathop\mathrm{dist}\nolimits}\left(\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\right)^2\leq \int_Q { \mathop\mathrm{dist}\nolimits}(z,\ell)^2 \frac{d\mu(z)}{\mu(Q) } = \beta_2(\mu, Q,\ell)^2({\mathop\mathrm{diam}\nolimits}Q)^2$$ by Jensen ’s inequality. contribute two windows $ R$ and $ Q$ with $ R\subset Q$, the approximation number $ \beta_2(\mu, Q)$ manipulate the approximation number $ \beta_2(\mu, R)$ if one has control on $ { \mathop\mathrm{diam}\nolimits}Q / { \mathop\mathrm{diam}\nolimits}R$ * * and * * $ \mu(Q) / \mu(R)$. \[l: double\ ] permit $ \mu$ be a Radon measure on $ { \mathbb{R}}^n$, let $ R, Q\subset{\mathbb{R}}^n$ be a bounded Borel set of plus diameter such that $ R\subset Q$ and $ \mu(R

infquality, is due to Lermak [@Lerman]. Practiccoly spxaking, jt allowr one to control the distancx frim tht $\mu$ center of mass of a winfow $Q$ to a suraight line $\ell$ mh terms of ths quautmty $\beta_2(\mu,Q,\ell)$. \[l:kerman\] Let $\mu$ be a Radon mdadure on ${\mathbb{R}}^n$, let $Q\subset{\mathbb{R}}^g$ be a noknded Borel sej of kosytivs diameter such that $\mu(Q)>0$, and let $$z_S:=\int_Q z\,\hrac{d\mu(z)}{\mu(Q)}\in {\msthbb{R}}^n$$ denote the center lf mwss of $Q$ with respfct to $\mu$. Fir edwry straight line $\ell$ pu ${\mathbb{R}}^n$, $${\jathop\mathrm{dist}\nolimits}(z_Q,\ell)\led \betc_2(\mu,Q,\ell){\mathip\natjtm{diam}\nolimivs}Q.$$ For every affinc subspdce $\ell$ in ${\mathbb{R}}^n$, tme fuictiin ${\mathop\mathrm{dist}\nonimits}(\cdot,\ell)^2$ is cjnvex. Thuv, $${\jathop\mathrm{dist}\nilumits}(e_Q,\ell)^2 = {\magyop\oatgrk{djst}\nollmivs}\left(\int_Q a\,\frac{d\mu(z)}{\mu(W)}\right)^2\leq \int_Q {\mathpp\iqthrm{dist}\nolijits}(z,\ejl)^2 \frac{d\mu(z)}{\mu(Q)} = \beta_2(\mu,Q,\ell)^2({\mathop\mathrm{diak}\nomimits}Q)^2$$ by Jensen’s ineqyality. Given two windoas $R$ and $Z$ with $R\subset Q$, the approximation number $\beta_2(\mu,Q)$ contcous uhc xppgoximation number $\beta_2(\mu,R)$ if one has control jh ${\kanhop\mathrm{diam}\nollmits}Q / {\mathop\mathtm{fisi}\nolimits}R$ **ana** $\mu(Q) / \mu(D)$. \[l:double\] Let $\mu$ be a Radog meawure on ${\mwthbn{R}}^n$, let $R,Q\subset{\mathbb{R}}^n$ be a bounded Bjeel set of positivz diameter sbch thst $R\sibset Q$ and $\mu(R

inequality, is due to Lerman [@Lerman]. Practically allows to control distance from the a $Q$ to a line $\ell$ in of the quantity $\beta_2(\mu,Q,\ell)$. \[l:lerman\] Let be a Radon measure on ${\mathbb{R}}^n$, let $Q\subset{\mathbb{R}}^n$ be a bounded Borel set positive diameter such that $\mu(Q)>0$, and let $$z_Q:=\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\in {\mathbb{R}}^n$$ denote the center mass $Q$ respect $\mu$. For every straight line $\ell$ in ${\mathbb{R}}^n$, $${\mathop\mathrm{dist}\nolimits}(z_Q,\ell)\leq \beta_2(\mu,Q,\ell){\mathop\mathrm{diam}\nolimits}Q.$$ For every affine subspace $\ell$ in the function ${\mathop\mathrm{dist}\nolimits}(\cdot,\ell)^2$ is convex. Thus, $${\mathop\mathrm{dist}\nolimits}(z_Q,\ell)^2 = z\,\frac{d\mu(z)}{\mu(Q)}\right)^2\leq \int_Q {\mathop\mathrm{dist}\nolimits}(z,\ell)^2 \frac{d\mu(z)}{\mu(Q)} \beta_2(\mu,Q,\ell)^2({\mathop\mathrm{diam}\nolimits}Q)^2$$ by Jensen’s inequality. Given windows and $Q$ $R\subset the number $\beta_2(\mu,Q)$ controls approximation number $\beta_2(\mu,R)$ if one has control on ${\mathop\mathrm{diam}\nolimits}Q / {\mathop\mathrm{diam}\nolimits}R$ **and** $\mu(Q) / \mu(R)$. \[l:double\] Let be a on ${\mathbb{R}}^n$, $R,Q\subset{\mathbb{R}}^n$ a Borel set of such that $R\subset Q$ and $\mu(R

inequality, is due to Lerman [@LeRman]. PractiCally SpeAkiNg, It alLows One to control thE DistAnce from the $\mu$ center of mAss of A wINdow $q$ To A straIght linE $\ElL$ IN teRmS oF thE qUAnTity $\bEta_2(\Mu,Q,\ell)$. \[l:Lerman\] Let $\mU$ be A RAdon measure oN ${\MaThbb{R}}^n$, let $Q\SubSet{\mathbb{R}}^n$ bE a bOunded boRel SEt of pOsiTive dIameteR Such thAt $\mu(Q)>0$, and lEt $$Z_q:=\int_Q z\,\FRac{d\mu(z)}{\MU(q)}\iN {\matHbb{R}}^n$$ denote the cenTEr OF mass of $Q$ with reSpect tO $\mU$. foR EVerY stRaight line $\ElL$ in ${\maTHbb{R}}^n$, $${\maTHoP\MAThrM{Dist}\nolimits}(z_q,\ell)\leq \beta_2(\MU,Q,\eLl){\mathOp\MatHRm{diam}\NolimItS}q.$$ FoR every affinE subSpace $\ell$ iN ${\mathbB{r}}^n$, the fuNCtion ${\maThop\maThrM{diSt}\noLImItS}(\cdOt,\ELl)^2$ iS CoNveX. thuS, $${\mathop\mAtHrM{dist}\NoliMITS}(Z_Q,\elL)^2 = {\maThop\MathrM{dist}\nolimits}\LefT(\int_q Z\,\frAc{d\mu(Z)}{\mu(Q)}\rIght)^2\LeQ \int_Q {\Mathop\MathrM{dIst}\nolimits}(z,\ell)^2 \Frac{D\mu(z)}{\mu(Q)} = \beTa_2(\mU,Q,\Ell)^2({\MaThop\mAThrm{diAm}\nOliMits}Q)^2$$ by jensen’s INeqUaLITY. GIven two windows $R$ and $q$ wITH $R\Subset Q$, tHe apprOXiMaTIon numbeR $\bEta_2(\Mu,Q)$ cONTrols The aPPrOximatioN numbeR $\BeTa_2(\Mu,R)$ if onE hAs contRoL on ${\MatHop\maTHrm{dIam}\nolImits}Q / {\maThop\mAThrm{diam}\nolimiTS}R$ **and** $\mu(Q) / \mu(R)$. \[l:dOUbLE\] leT $\Mu$ be A RaDon measure oN ${\matHBb{R}}^n$, Let $R,q\SuBseT{\MathbB{R}}^n$ be A bOUnDEd Borel set of positivE dIameteR such That $R\subset Q$ aNd $\mu(R

inequality, is due to Ler man [@Lerm an].Pra cti ca llyspea king, it allow s one to control the distan ce fr om the$ \m u$ ce nter of ma s s of a w ind ow $Q $ toa s traight line $\el l$in terms of th e q uantity $\ bet a_2(\mu,Q,\e ll) $. \[ l: ler m an\]Let $\mu $ be a Radonmeasure o n$ {\math b b{R}}^n $ , l et $ Q\subset{\mathbb{ R }} ^ n$ be a bounde d Bore ls et o f p osi tive diame te r suc h that $ \ mu ( Q ) >0$ , and let $$z_ Q:=\int_Q z \ ,\f rac{d\ mu (z) } {\mu(Q )}\in { \ mat hbb{R}}^n$$ den ote the c entero f masso f $Q$ w ith re spe ctto $ \ mu $. Fo re ver y s tra i ght line $\ el l$ in $ {\ma t h b b {R}} ^n$ , $$ {\mat hop\mathrm{di st} \nol i mit s}(z_ Q,\el l)\l eq \bet a_2(\m u,Q,\ el l){\mathop\math rm{d iam}\noli mit s} Q.$ $ Fore very a ffi nesubspac e $\ell $ in $ { \ m at hbb{R}}^n$, the fu nc t i on ${\math op\mat h rm {d i st}\noli mi ts} (\cd o t ,\ell )^2$ is convex. Thus, $$ {\ mathop\ ma thrm{d is t}\ nol imits } (z_Q ,\ell) ^2 = {\m athop \ mathrm{dist}\n o limits}\left( \ in t _ Qz \,\f rac {d\mu(z)}{\ mu(Q ) }\ri ght) ^ 2\ leq \int_ Q {\m at h op \ mathrm{dist}\nolimi ts }(z,\e ll)^2 \frac{d\mu(z )}{\mu(Q)} = \beta_2( \mu, Q ,\ e ll)^2({\mathop \math rm{diam}\n o limits}Q )^2$$ by Jens en’s ineq u a lity. G ive n t wowin d o ws $R$ and $Q$w i th $ R\ subsetQ$, the ap pro xim ati onnu mber $\be ta_2(\mu ,Q )$ c on tro ls th e approxi ma tio nnum ber $ \ beta_2 (\mu, R)$if o n e h as cont r ol o n ${ \m at hop\ mat hr m{dia m}\n o lim its}Q / {\mathop \ma t hrm{ di am }\nolim its}R$ **and* *$\mu(Q) /\m u(R )$. \ [ l :double\ ] Let $\mu$ be a Radonm easureon${\ma thbb {R}}^n$,let $R,Q\ sub s et{\ma thbb{R }}^n$ b e a b ounde d Bo rel s et of posi t i vediame te r su ch that $R\subset Q$ and$ \mu (R

inequality,_is due_to Lerman [@Lerman]. Practically_speaking, it_allows_one to_control_the distance from_the $\mu$ center_of mass of a_window $Q$ to_a_straight line $\ell$ in terms of the quantity $\beta_2(\mu,Q,\ell)$. \[l:lerman\] Let $\mu$ be a Radon_measure_on ${\mathbb{R}}^n$,_let_$Q\subset{\mathbb{R}}^n$_be a bounded Borel set_of positive diameter such that_$\mu(Q)>0$, and_let $$z_Q:=\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\in {\mathbb{R}}^n$$ denote the center of_mass_of $Q$ with_respect to $\mu$. For every straight line $\ell$ in_${\mathbb{R}}^n$, $${\mathop\mathrm{dist}\nolimits}(z_Q,\ell)\leq \beta_2(\mu,Q,\ell){\mathop\mathrm{diam}\nolimits}Q.$$ For every affine subspace_$\ell$ in ${\mathbb{R}}^n$,_the_function_${\mathop\mathrm{dist}\nolimits}(\cdot,\ell)^2$ is convex. Thus,_$${\mathop\mathrm{dist}\nolimits}(z_Q,\ell)^2 = {\mathop\mathrm{dist}\nolimits}\left(\int_Q z\,\frac{d\mu(z)}{\mu(Q)}\right)^2\leq \int_Q {\mathop\mathrm{dist}\nolimits}(z,\ell)^2_\frac{d\mu(z)}{\mu(Q)} = \beta_2(\mu,Q,\ell)^2({\mathop\mathrm{diam}\nolimits}Q)^2$$ by Jensen’s inequality. Given_two windows $R$ and $Q$ with $R\subset_Q$, the approximation number $\beta_2(\mu,Q)$ controls_the approximation number $\beta_2(\mu,R)$ if_one has_control on ${\mathop\mathrm{diam}\nolimits}Q / {\mathop\mathrm{diam}\nolimits}R$_**and** $\mu(Q) /_\mu(R)$. \[l:double\] Let_$\mu$ be a_Radon measure on ${\mathbb{R}}^n$, let $R,Q\subset{\mathbb{R}}^n$_be a bounded_Borel set of positive diameter such_that_$R\subset Q$ and_$\mu(R

\]), Eq. (\[eqn:Psi\_x\_naturalnumberAssumption\]) becomes $$\begin{aligned} &\Psi_x^{\mrm{Laplace}}(H,\gamma)\nn &=\xi^{n(n-1)}e^{nNHM}\sum_{\mcal{S}_x}\exp\Big[ H\sum_{i=1}^n\sum_{a = 1}^{\tau_x}S_i^{\{a\}} \nn \aleq - \sum_{i<j}\ln \Big\{ \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\Big], \label{eqn:Psi_x_Laplace}\end{aligned}$$ where $\xi:=\sqrt{2n/\gamma}$. Here, we assume $$\begin{aligned} \xi > \max_{i<j}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big). \label{eqn:assumption_Laplace}\end{aligned}$$ By using the perturbative approximation, $$\begin{aligned} &\ln \Big\{ \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn &= \ln \xi^2 + \ln \Big\{1 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn &\approx \ln \xi^2 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2,\end{aligned}$$ we obtain the approximation of Eq. (\[eqn:Psi\_x\_Laplace\]) as $$\begin{aligned} \Psi_x^{\mrm{Laplace}}(H,\gamma) &\approx e^{nNHM}\sum_{\mcal{S}_x}\exp\Big[ H\sum_{i=1}^n\sum_{a = 1}^{\

\ ]), Eq.   (\[eqn: Psi\_x\_naturalnumberAssumption\ ]) becomes $ $ \begin{aligned } & \Psi_x^{\mrm{Laplace}}(H,\gamma)\nn & = \xi^{n(n-1)}e^{nNHM}\sum_{\mcal{S}_x}\exp\Big [ H\sum_{i=1}^n\sum_{a = 1}^{\tau_x}S_i^{\{a\ } } \nn \aleq - \sum_{i < j}\ln \Big\ { \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\Big ], \label{eqn: Psi_x_Laplace}\end{aligned}$$ where $ \xi:=\sqrt{2n/\gamma}$. Here, we assume $ $ \begin{aligned } \xi > \max_{i < j}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big). \label{eqn: assumption_Laplace}\end{aligned}$$ By using the perturbative approximation, $ $ \begin{aligned } & \ln \Big\ { \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn & = \ln \xi^2 + \ln \Big\{1 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn & \approx \ln \xi^2 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2,\end{aligned}$$ we receive the estimate of Eq.   (\[eqn: Psi\_x\_Laplace\ ]) as $ $ \begin{aligned } \Psi_x^{\mrm{Laplace}}(H,\gamma) & \approx e^{nNHM}\sum_{\mcal{S}_x}\exp\Big [ H\sum_{i=1}^n\sum_{a = 1}^{\

\]), Eq. (\[fqn:Psi\_x\_naturalnumberAssuoption\]) becomes $$\ywgin{almgned} &\Psj_x^{\mrm{Lapuace}}(H,\gamma)\nn &=\xi^{n(n-1)}e^{nNHM}\sum_{\mcal{D}_x}\wxp\Bit[ H\sum_{i=1}^n\sum_{a = 1}^{\tau_x}S_i^{\{a\}} \vn \aleq - \sul_{i<j}\ln \Bit\{ \xi^2 -\Vig(\sum_{a=1}^{\tau_e}A_i^{\{a\}}S_j^{\{a\}}+ K d_{ij}\Glg)^2\Big\}\Yij], \label{eqn:Psi_x_Laklace}\end{aligted}$$ where $\xi:=\sqst{2v/\gcmma}$. Here, we assume $$\begin{aligned} \xi > \iax_{i<j}\Bib(\skm_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ G d_{ik}\Fig). \lzbel{eqn:assumption_Laplace}\end{aligned}$$ By usiig the perturbayive approximation, $$\begin{allgnef} &\ln \Big\{ \xi^2 -\Big(\sum_{a=1}^{\hau_x}S_i^{\{a\}}S_j^{\{a\}}+ B d_{it}\Vig)^2\Big\}\nn &= \ln \xi^2 + \ln \Big\{1 -\xi^{-2}\Big(\sum_{a=1}^{\tzu_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn &\approx \lv \xi^2 -\ri^{-2}\Big(\sum_{a=1}^{\tao_r}W_i^{\{a\}}D_b^{\{a\}}+ N d_{ij}\Big)^2,\xnd{alidned}$$ we obtalm the dpproxikation of Eq. (\[eqk:Psi\_x\_Napoace\]) as $$\begin{aligned} \Pvi_x^{\mrm{Laplace}}(H,\gammw) &\approx e^{tNKM}\sum_{\mcal{S}_x}\exp\Big[ H\sun_{i=1}^b\sum_{a = 1}^{\

\]), Eq. (\[eqn:Psi\_x\_naturalnumberAssumption\]) becomes $$\begin{aligned} &\Psi_x^{\mrm{Laplace}}(H,\gamma)\nn &=\xi^{n(n-1)}e^{nNHM}\sum_{\mcal{S}_x}\exp\Big[ 1}^{\tau_x}S_i^{\{a\}} \aleq - \Big\{ \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ Here, assume $$\begin{aligned} \xi \max_{i<j}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big). By using the perturbative approximation, $$\begin{aligned} \Big\{ \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn &= \ln \xi^2 + \ln \Big\{1 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N &\approx \ln \xi^2 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2,\end{aligned}$$ we obtain the approximation of Eq. (\[eqn:Psi\_x\_Laplace\]) $$\begin{aligned} &\approx H\sum_{i=1}^n\sum_{a 1}^{\

\]), Eq. (\[eqn:Psi\_x\_naturalnumberAssUmption\]) becOmes $$\bEgiN{alIgNed} &\PSi_x^{\mRm{Laplace}}(H,\gammA)\Nn &=\xi^{N(n-1)}e^{nNHM}\sum_{\mcal{S}_x}\exp\Big[ h\sum_{i=1}^N\sUM_{a = 1}^{\taU_X}S_I^{\{a\}} \nn \aLeq - \sum_{i<J}\Ln \bIG\{ \xi^2 -\biG(\sUm_{a=1}^{\TaU_X}S_I^{\{a\}}S_j^{\{a\}}+ n d_{iJ}\Big)^2\Big\}\big], \label{eqN:PsI_x_laplace}\end{alIGnEd}$$ where $\xi:=\sQrt{2N/\gamma}$. Here, we AssUme $$\begIn{AliGNed} \xi > \Max_{I<j}\Big(\Sum_{a=1}^{\taU_X}S_i^{\{a\}}S_j^{\{A\}}+ N d_{ij}\Big). \lAbEL{eqn:asSUmption_lAPlAce}\eNd{aligned}$$ By using tHE pERturbative apprOximatIoN, $$\BeGIN{alIgnEd} &\ln \Big\{ \xi^2 -\BIg(\Sum_{a=1}^{\tAU_x}S_i^{\{a\}}S_j^{\{A\}}+ n d_{IJ}\bIg)^2\BIG\}\nn &= \ln \xi^2 + \ln \Big\{1 -\xI^{-2}\Big(\sum_{a=1}^{\tau_X}s_i^{\{a\}}s_j^{\{a\}}+ N d_{iJ}\BIg)^2\BIG\}\nn &\appRox \ln \Xi^2 -\XI^{-2}\BiG(\sum_{a=1}^{\tau_x}S_i^{\{A\}}S_j^{\{a\}}+ n d_{ij}\Big)^2,\enD{alignED}$$ we obtaIN the appRoximaTioN of eq. (\[eqN:psI\_x\_lapLaCE\]) as $$\BEgIn{aLIgnEd} \Psi_x^{\mrM{LApLace}}(H,\GammA) &\APPRox e^{NNHm}\sum_{\Mcal{S}_X}\exp\Big[ H\sum_{i=1}^n\Sum_{A = 1}^{\

\]), Eq. (\[eqn:Psi\_x\_na turalnumbe rAssu mpt ion \] ) be come s $$\begin{ali g ned} &\Psi_x^{\mrm{Laplace }}(H, \g a mma) \ nn &=\x i^{n(n- 1 )} e ^ {nN HM }\ sum _{ \ mc al{S} _x} \exp\Bi g[ H\sum_{ i=1 }^ n\sum_{a = 1 } ^{ \tau_x}S_i ^{\ {a\}} \nn \a leq - \su m_ {i< j }\ln\Bi g\{ \ xi^2 - \ Big(\s um_{a=1}^ {\ t au_x}S _ i^{\{a\ } } S_ j^{\ {a\}}+ N d_{ij}\B i g) ^ 2\Big\}\Big],\label {e q n: P s i_x _La place}\end {a ligne d }$$ whe r e$ \ x i:= \ sqrt{2n/\gamm a}$. Here,w e a ssume$$ \be g in{ali gned} \ x i > \max_{i<j} \Big (\sum_{a= 1}^{\t a u_x}S_i ^ {\{a\}} S_j^{\ {a\ }}+ N d _ {i j} \Bi g) . \l a be l{e q n:a ssumptio n_ La place }\en d { a l igne d}$ $ By usin g the perturb ati ve a p pro ximat ion,$$\b eg in{al igned} &\ln \ Big\{ \xi^2 -\B ig(\ sum_{a=1} ^{\ ta u_x }S _i^{\ { a\}}S_ j^{ \{a \}}+ Nd_{ij}\ B ig) ^2 \ B i g\ }\nn &= \ln \xi^2 + \ ln \Big\{1 -\xi^ { -2 }\ B ig(\sum_ {a =1} ^{\t a u _x}S_ i^{\ { a\ }}S_j^{\ {a\}}+ Nd_ {ij}\Bi g) ^2\Big \} \nn &\ appro x \l n \xi^ 2 -\xi^{ -2}\B i g(\sum_{a=1}^{ \ tau_x}S_i^{\{ a \} } S _j ^ {\{a \}} + N d_{ij}\ Big) ^ 2,\e nd{a l ig ned } $$ we obta in th e approximation of E q.  (\[eq n:Psi \_x\_Laplace\ ]) as $$\b e g i n{aligne d} \ P si _ x^{\mrm{Laplac e}}(H ,\gamma) & \ approx e ^{nNH M}\sum_{ \mcal{S}_ x } \exp\Big [ H \su m_{ i=1 } ^ n\ sum_{a = 1}^{ \

\]), Eq. (\[eqn:Psi\_x\_naturalnumberAssumption\])_becomes $$\begin{aligned} &\Psi_x^{\mrm{Laplace}}(H,\gamma)\nn &=\xi^{n(n-1)}e^{nNHM}\sum_{\mcal{S}_x}\exp\Big[_H\sum_{i=1}^n\sum_{a = 1}^{\tau_x}S_i^{\{a\}} \nn \aleq -_\sum_{i<j}\ln \Big\{_\xi^2_-\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N_d_{ij}\Big)^2\Big\}\Big], \label{eqn:Psi_x_Laplace}\end{aligned}$$_where $\xi:=\sqrt{2n/\gamma}$. Here,_we assume $$\begin{aligned} \xi_> \max_{i<j}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big). \label{eqn:assumption_Laplace}\end{aligned}$$_By using the_perturbative_approximation, $$\begin{aligned} &\ln \Big\{ \xi^2 -\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+ N d_{ij}\Big)^2\Big\}\nn &= \ln \xi^2 + \ln \Big\{1 -\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+_N_d_{ij}\Big)^2\Big\}\nn &\approx \ln_\xi^2_-\xi^{-2}\Big(\sum_{a=1}^{\tau_x}S_i^{\{a\}}S_j^{\{a\}}+_N d_{ij}\Big)^2,\end{aligned}$$ we obtain the_approximation of Eq. (\[eqn:Psi\_x\_Laplace\]) as $$\begin{aligned} \Psi_x^{\mrm{Laplace}}(H,\gamma) &\approx_e^{nNHM}\sum_{\mcal{S}_x}\exp\Big[ H\sum_{i=1}^n\sum_{a_= 1}^{\

. \[comp cor\] Let $\varphi$ solve on $Q_T$. Then the following holds $$\label{phi bound} \inf_{M^n} \varphi_0 \leq \varphi - \frac{t}{n} \leq \sup_{M^n} \varphi_0.$$ Define $\tilde{\varphi} = t/n$. Then $\tilde{\varphi}$ solves. Apply Theorem \[comp princ\]. \[u exp bound\] By exponentiation, this is equivalent to $$\label{u bound} \inf_{M^n} u_0 \leq u \operatorname{e}^{-t/n} \leq \sup_{M^n} u_0,$$ for the embedding function $u$, which solves. The solution $\tilde{\varphi}$ of in the proof of Corollary \[comp cor\] corresponds to the solution of whereby the initial embedding is given by $$x_0:p\mapsto (1,p).$$ As we shall be seeing shortly in §\[convergence\], the estimate on $u$ in Remark \[u exp bound\] alludes to the fact that if we were to introduce a factor of $\operatorname{e}^{-t/n}$, then the function $\tilde{u} = u\operatorname{e}^{-t/n}$ remains uniformly bounded from above and below. Thus we will be studying the convergence properties of for the corresponding rescaled embedding $\tilde{x}$ in §\[rescaled hypersurfaces\]. short time existence and first order estimates {#first order estimates} ---------------------------------------------- With the help of the parabolic theory [@ger2; @lady; @evans] we show that a solution to equation exists for a short time. Following this, as the heading suggests, we prove some first order estimates on $\varphi$. Let $F = Huv^{-1}$ as defined in. Then $$a^{ij} \equiv -\frac{\partial F}{\partial \varphi_{ij}}$$ is positive definite. Since $$F = v^{-2}(n - \sigma^{ij}\varphi_{ij}+ \varphi^i \varphi^j\varphi_{ij} v^{-2} ) = v^{-2}(n-u^2 g^{ij}\varphi_{ij}),$$ we simply differentiate $F$ with respect to $\varphi_{ij}$ to find $$-a^{ij} = \frac{\partial F}{\partial \varphi_{ij}} =

. \[comp cor\ ] Let $ \varphi$ solve on $ Q_T$. Then the following holds $ $ \label{phi bandaged } \inf_{M^n } \varphi_0 \leq \varphi - \frac{t}{n } \leq \sup_{M^n } \varphi_0.$$ specify $ \tilde{\varphi } = t / n$. Then $ \tilde{\varphi}$ solves. Apply Theorem \[comp princ\ ]. \[u exp bound\ ] By exponentiation, this is equivalent to $ $ \label{u bound } \inf_{M^n } u_0 \leq u \operatorname{e}^{-t / n } \leq \sup_{M^n } u_0,$$ for the embedding routine $ u$, which solves. The solution $ \tilde{\varphi}$ of in the proof of Corollary \[comp cor\ ] represent to the solution of whereby the initial embedding is impart by $ $ x_0: p\mapsto (1,p).$$ As we shall be seeing soon in § \[convergence\ ], the estimate on $ u$ in Remark \[u exp bound\ ] alludes to the fact that if we were to precede a factor of $ \operatorname{e}^{-t / n}$, then the function $ \tilde{u } = u\operatorname{e}^{-t / n}$ stay uniformly restrict from above and below. Thus we will be studying the convergence properties of for the comparable rescaled embedding $ \tilde{x}$ in § \[rescaled hypersurfaces\ ]. short fourth dimension existence and first holy order estimates { # inaugural order estimates } ---------------------------------------------- With the help of the parabolic theory [ @ger2; @lady; @evans ] we show that a solution to equality exists for a short time. Following this, as the heading suggests, we rise some first order estimates on $ \varphi$. Let $ F = Huv^{-1}$ as defined in. Then $ $ a^{ij } \equiv -\frac{\partial F}{\partial \varphi_{ij}}$$ is positive definite. Since $ $ F = v^{-2}(n - \sigma^{ij}\varphi_{ij}+ \varphi^i \varphi^j\varphi_{ij } v^{-2 }) = v^{-2}(n - u^2 g^{ij}\varphi_{ij}),$$ we simply differentiate $ F$ with respect to $ \varphi_{ij}$ to recover $ $ -a^{ij } = \frac{\partial F}{\partial \varphi_{ij } } =

. \[como cor\] Let $\varphi$ solve ok $Q_T$. Then the following holds $$\label{phk bound} \inf_{M^n} \varphi_0 \pew \varkki - \frac{t}{n} \leq \sup_{M^n} \varphi_0.$$ Devine $\tilee{\vacphi} = t/n$. Then $\tilde{\varphl}$ solbcs. Ap'lb Theorem \[comp krinc\]. \[u exp bmund\] By exponettkacion, this is equivalent to $$\label{u botnd} \onv_{M^n} u_0 \leq u \opgratognwme{e}^{-f/n} \leq \sup_{M^n} u_0,$$ for the embedding fhnction $u$, which solvrs. The solution $\tilde{\varphi}$ of ln the proof of Cogollary \[comk cow\] corresponds to the solution of whgreby the initial embedding is gkven yy $$x_0:p\mapsto (1,p).$$ Qs ag shall be sxeing fhortly in §\[convergenwe\], the rstimate on $u$ ln Rekarj \[u exp bound\] alludes to the fact that yf we wera co introduce a factor od $\opetatortame{d}^{-r/n}$, gheh vhe functlon $\tilde{u} = u\kperatornamw{e}^{-t/n}$ remains uniforklr bounded from above agd below. Thus we will be studying the cmnvsrgence properties of fir the corresponding tescaled eibedding $\tilde{x}$ in §\[rescaled hypersurfaces\]. short tike exmsgenec xbd first order estimates {#first order estimates} ---------------------------------------------- Wyfh tme help of the pcrabolic theory [@beg2; @kwdy; @evans] we rhow tkzt a solution to equwtion evists for a shjrt yime. Following this, as the yeading suggvsts, we prove some firdt order escimatex on $\farphi$. Let $F = Huv^{-1}$ as defnned ih. Then $$a^{ij} \fquiv -\frad{\oartial F}{\partial \vagphi_{hj}}$$ is positive definite. Sinse $$F = v^{-2}(n - \sigka^{ij}\varohi_{ik}+ \varpri^i \varphi^u\varpmh_{ij} v^{-2} ) = v^{-2}(n-u^2 g^{ij}\varohi_{ij}),$$ wa simply dlfferentiate $F$ with respect to $\terphi_{ij}$ to fimd $$-a^{ib} = \frac{\pcrtial F}{\partial \var[hi_{ij}} =

. \[comp cor\] Let $\varphi$ solve on the holds $$\label{phi \inf_{M^n} \varphi_0 \leq \varphi_0.$$ $\tilde{\varphi} = t/n$. $\tilde{\varphi}$ solves. Apply \[comp princ\]. \[u exp bound\] By this is equivalent to $$\label{u bound} \inf_{M^n} u_0 \leq u \operatorname{e}^{-t/n} \leq \sup_{M^n} for the embedding function $u$, which solves. The solution $\tilde{\varphi}$ of in the of \[comp corresponds the solution of whereby the initial embedding is given by $$x_0:p\mapsto (1,p).$$ As we shall be shortly in §\[convergence\], the estimate on $u$ in \[u exp bound\] alludes the fact that if we to a factor $\operatorname{e}^{-t/n}$, the $\tilde{u} = u\operatorname{e}^{-t/n}$ uniformly bounded from above and below. Thus we will be studying the convergence properties of for the rescaled embedding §\[rescaled hypersurfaces\]. time and order estimates {#first ---------------------------------------------- With the help of the @lady; @evans] we show that a solution to exists for short time. Following this, as the suggests, we prove some first order estimates on Let $F = Huv^{-1}$ as defined in. Then $$a^{ij} \equiv -\frac{\partial F}{\partial \varphi_{ij}}$$ is positive $$F = v^{-2}(n - \varphi^i \varphi^j\varphi_{ij} v^{-2} = g^{ij}\varphi_{ij}),$$ simply $F$ with to $\varphi_{ij}$ to find $$-a^{ij} = \frac{\partial F}{\partial \varphi_{ij}} =

. \[comp cor\] Let $\varphi$ solve on $Q_T$. then the folLowinG hoLds $$\LaBel{pHi boUnd} \inf_{M^n} \varphi_0 \LEq \vaRphi - \frac{t}{n} \leq \sup_{M^n} \varpHi_0.$$ DefInE $\TildE{\VaRphi} = t/N$. Then $\tiLDe{\VARphI}$ sOlVes. apPLy theorEm \[cOmp prinC\]. \[u exp bound\] by eXpOnentiation, tHIs Is equivaleNt tO $$\label{u bound} \Inf_{m^n} u_0 \leq U \oPerATornaMe{e}^{-T/n} \leq \Sup_{M^n} u_0,$$ FOr the eMbedding fUnCTion $u$, wHIch solvES. thE solUtion $\tilde{\varphi}$ oF In THe proof of CorolLary \[coMp COr\] CORreSpoNds to the soLuTion oF Whereby THe INITiaL Embedding is giVen by $$x_0:p\mapsTO (1,p).$$ AS we shaLl Be sEEing shOrtly In §\[COnvErgence\], the eStimAte on $u$ in REmark \[u EXp bound\] ALludes tO the faCt tHat If we WErE tO inTrODucE A fActOR of $\OperatorNaMe{E}^{-t/n}$, thEn thE FUNCtioN $\tiLde{u} = U\operAtorname{e}^{-t/n}$ reMaiNs unIForMly boUnded From AbOve anD below. thus wE wIll be studying thE conVergence pRopErTieS oF for tHE correSpoNdiNg rescaLed embeDDinG $\tILDE{x}$ In §\[rescaled hypersurFaCES\]. sHort time ExisteNCe AnD First ordEr EstImatES {#First OrdeR EsTimates} ---------------------------------------------- WIth the HElP oF the parAbOlic thEoRy [@gEr2; @lAdy; @evANs] we Show thAt a solutIon to EQuation exists fOR a short time. FoLLoWINg THis, aS thE heading sugGestS, We prOve sOMe FirST ordeR estiMaTEs ON $\varphi$. Let $F = Huv^{-1}$ as defInEd in. ThEn $$a^{ij} \Equiv -\frac{\partIal F}{\partiaL \VARphi_{ij}}$$ is PosiTIvE Definite. Since $$F = V^{-2}(n - \sigMa^{ij}\varphi_{IJ}+ \varphi^i \VarphI^j\varphi_{Ij} v^{-2} ) = v^{-2}(n-u^2 g^{ij}\VARphi_{ij}),$$ we SimPly DifFerENTiAte $F$ with respeCT To $\vaRpHi_{ij}$ to fInd $$-A^{ij} = \frac{\ParTiaL F}{\pArtIaL \varphi_{ij}} =

. \[comp cor\] Let $\varp hi$ solveon $Q _T$ . T he n th e fo llowing holds$ $\la bel{phi bound} \inf _{ M ^n}\ va rphi_ 0 \leq\ va r p hi-\f rac {t } {n } \le q \ sup_{M^ n} \varphi _0. $$ Define $\t i ld e{\varphi} =t/n$. Then $ \ti lde{\v ar phi } $ sol ves . App ly The o rem \[ comp prin c\ ] . \[u exp bou n d \] Byexponentiation, t h is is equivalentto $$\ la b el { u bo und } \inf _{ M^n}u _0 \leq u\ o p era t orname{e}^{-t /n} \leq \s u p_{ M^n} u _0 ,$$ for th e emb ed d ing function $ u$,which sol ves. T he solu t ion $\t ilde{\ var phi }$ o f i nthe p r oof of Co r oll ary \[co mp c or\]corr e s p o ndstothesolut ion of whereb y t he i n iti al em beddi ng i sgiven by $$ x_0:p \m apsto (1,p).$$ Aswe shallbese ein gshort l y in § \[c onv ergence \], the est im a t e o n $u$ in Remark \[ ue x pbound\]allude s t ot he factth atif w e wereto i n tr oduce afactor of $ \operat or name{e }^ {-t /n} $, th e n th e func tion $\t ilde{ u } = u\operator n ame{e}^{-t/n} $ r e m ai n s un ifo rmly bounde d fr o m ab ovea nd be l ow. T hus w ew il l be studying the co nv ergenc e pro perties of fo r the corr e s p onding r esca l ed embedding $\ti lde{x }$ in §\[r e scaled h ypers urfaces\ ]. short t ime exis ten ceand fi r s torder estimat e s {#f ir st orde r e stimate s}--- --- --- -- --------- -------- -- -- -- -- --- ----- - - Withth e h el p o f the parabo lic t heor y[@ g er2 ; @lady ; @ e v ans] w eshow th at a so luti o n t o equat ion exist s f o r ash or t time. Following th is , as the h ea din g sugg e s ts, we p rove some first order e s timates on $\va rphi $. Let $ F = Huv^{ -1} $ as de finedin. T he n $ $ a ^{ij} \ eq uiv - \frac{\par t i alF}{\p ar tial \varph i_{ij}}$$ is posit i vedefinite. Si nce $$F = v ^{- 2 }( n -\s i gma ^ { ij}\varphi_{ij} + \varphi^ i\ va rphi^j\var p hi_ {i j} v^{- 2} ) =v^{-2 } (n-u^2g^{ij}\va rphi_{ij} ), $$ w e sim ply differ entiate$F$ withr espec t t o $\v arp hi_{ij }$ to find $$-a^ { ij} = \f rac{\p ar tial F }{\pa rt ial \var phi_{ij}} =

. \[comp cor\]_Let $\varphi$_solve on $Q_T$. Then_the following_holds_$$\label{phi bound} __ _ _\inf_{M^n} \varphi_0 \leq \varphi_- \frac{t}{n} \leq_\sup_{M^n}_\varphi_0.$$ Define $\tilde{\varphi} = t/n$. Then $\tilde{\varphi}$ solves. Apply Theorem \[comp princ\]. \[u exp bound\] By_exponentiation,_this is_equivalent_to_$$\label{u bound} _\inf_{M^n} u_0 \leq u \operatorname{e}^{-t/n}_\leq \sup_{M^n}_u_0,$$ for the embedding function $u$, which solves. The_solution_$\tilde{\varphi}$ of in_the proof of Corollary \[comp cor\] corresponds to the_solution of whereby the initial embedding_is given by_$$x_0:p\mapsto_(1,p).$$ As_we shall be seeing_shortly in §\[convergence\], the estimate on_$u$ in Remark \[u exp bound\]_alludes to the fact that if we_were to introduce a factor of_$\operatorname{e}^{-t/n}$, then the function $\tilde{u}_= u\operatorname{e}^{-t/n}$_remains uniformly bounded from above_and below. Thus_we will_be studying the_convergence properties of for the corresponding_rescaled embedding $\tilde{x}$_in §\[rescaled hypersurfaces\]. short time existence and_first_order estimates {#first_order_estimates} ---------------------------------------------- With_the help_of the parabolic_theory_[@ger2; @lady;_@evans]_we show that a solution to_equation_exists for a short time. Following this,_as the heading suggests,_we_prove some first order_estimates on $\varphi$. Let $F =_Huv^{-1}$ as defined in. Then $$a^{ij}_\equiv -\frac{\partial_F}{\partial \varphi_{ij}}$$_is positive definite. Since $$F = v^{-2}(n - \sigma^{ij}\varphi_{ij}+ \varphi^i \varphi^j\varphi_{ij} v^{-2}_) = v^{-2}(n-u^2 g^{ij}\varphi_{ij}),$$ we simply_differentiate $F$ with respect_to $\varphi_{ij}$_to_find $$-a^{ij} =_\frac{\partial_F}{\partial \varphi_{ij}}_=

table $D$. With table $D$ a prefix sum computation is started in row major layout such that $D$ now contains the relative output destination of each row within each meta-run. In a CRCW model, with the same assignment of elements as before, tiles can now be written to their destination to form the output using table $D$. For CREW, when first creating $D$, one can ceil the tile sizes to full blocks. The resulting layout will obviously contain blocks that are not entirely filled, but contain empty memory cells. However, using the contraction described above, one can extract these empty cells. Hence, the number of I/Os to copy tiles independently in parallel is $\O{\frac{H}{PB}}$. In order to write the transposed matrix directly, the destination for each tile has to be determined. This can be done using a prefix sum computation, requiring $\O{\log P}$ I/Os. The whole step to finalise the shuffle step has I/O-complexity $\O{\frac{H}{PB} + \log P}$. #### Parallel (associative) reduce function Assuming a parallelisable reduce, each processor shall perform multiple reduce functions simultaneously on a subset of elements with intermediate key in a certain range. In a final step, the results of these partial reduce executions are then collected and reduced to the final result. To this end, the range of intermediate keys is partitioned into $\ceil{N_RPw/H}$ ranges of up to $\ceil{H/(Pw)}$ keys. @fullpaper Using the range-bounded load-balancing algorithm, elements (still ordered in meta-runs) are assigned to processors such that each processor gets elements from at most two pieces of row indices. This can be achieved by using the tuple $(\text{\it meta-run index}, \text{\it row index})$ as key. Then, elements are assigned to processors such that each processor gets elements from at most two ranges of row indices. If a processor got assigned elements that belong to the same reduce function, elements can be reduced immediately by the processor. Afterwards, for each key range, elements can be gathered to form the final result of the reduce function. This is possible with $\O{\frac{H}{PB} + \log P}$ I/Os. ### Complete sorting / merging For some choices of parameters, especially for small instances, it can be optimal to simply apply

table $ D$. With table $ D$ a prefix sum computation is depart in rowing major layout such that $ D$ now contains the relative output signal destination of each row within each meta - rivulet. In a CRCW mannequin, with the same grant of elements as before, tile can now be written to their finish to form the output using table $ D$. For CREW, when first make $ D$, one can ceil the tile sizes to full blocks. The result layout will obviously contain blocks that are not entirely fill, but contain empty memory cells. However, using the contraction described above, one can press out these empty cells. Hence, the number of I / Os to copy tiles independently in parallel is $ \O{\frac{H}{PB}}$. In order to write the transposed matrix directly, the destination for each tile has to be determined. This can be done using a prefix sum computation, necessitate $ \O{\log P}$ I / Os. The whole dance step to finalise the shuffle step has I / O - complexity $ \O{\frac{H}{PB } + \log P}$. # # # # Parallel (associative) deoxidize affair Assuming a parallelisable reduce, each processor shall do multiple reduce functions simultaneously on a subset of elements with intermediate key in a certain range. In a final step, the results of these partial reduce executions are then collected and reduce to the final consequence. To this end, the range of intermediate keys is partitioned into $ \ceil{N_RPw / H}$ range of up to $ \ceil{H/(Pw)}$ keys. @fullpaper Using the range - restrict load - balancing algorithm, elements (still regulate in meta - runs) are assigned to processors such that each processor gets component from at most two pieces of quarrel indices. This can be achieve by using the tuple $ (\text{\it meta - run index }, \text{\it rowing index})$ as samara. Then, elements are assign to processors such that each processor gets elements from at most two scope of row indices. If a central processing unit got assigned elements that belong to the same reduce function, elements can be reduced immediately by the central processing unit. Afterwards, for each key range, elements can be gathered to form the final result of the reduce affair. This is potential with $ \O{\frac{H}{PB } + \log P}$ I / Os. # # # arrant sorting / merging For some choices of parameters, especially for minor instances, it can be optimal to simply give

tahle $D$. With table $D$ a prenix sum computatnin is vtartes in row major layout such that $D$ nox cobtainw the relative output aestinatiln of eaxh riq within eedh meta-vbn. In w CREW model, with thg same assigtment of elemettr cs before, tiles can now be written tj their dfstination to sorm ehe kltkut using table $D$. For CREW, when fidst creeting $D$, one can ceil the tile sizes to fupl bpocks. The resultinh layout wiol ofciously contxin blocks that are noj entirely filled, but contain emoty mzmory cells. Hiweggr, using the contgaction descrlned abmve, one can extract tmese xmptt cells. Hence, the numbxr of I/Os to copy tijes indepanbently in parallel is $\O{\drac{H}{KB}}$. In ordde tu wdive fhe trwns'osed matris directly, rhe destination for ewbn tile has tk be dqtqrmined. This can be done using a prefix suj computation, requiring $\O{\log P}$ I/Os. The whole step to sinalise the shuffle step has I/O-complexity $\O{\frac{H}{[B} + \lke P}$. #### Pwfqlpel (associative) reduce function Assuming a parwmltlixable reduce, ecch processor shslp lgrform multiplg reducz fhnctions simultanelusly og a sybset of tlememts with intermediate key ib a certain gangw. In a final step, che results uf tnese lartial reduce executious are then collefted and ddduced to the fival rasult. To ufis end, the range of internedicte keys is kartitijned into $\feil{N_V[w/H}$ ranges of up tl $\ceip{H/([w)}$ keys. @fuplpaper Using the range-bounded load-balancing slcorpthm, elemznts (sbill ordered in meta-runs) are cssigned to pfocessors auch thet each procqssor gets elalents from av most twj piwces of row kndices. This csn be achpered by usung the tuple $(\text{\lt meja-dun index}, \text{\ic tiw index})$ as keu. Tfen, epekegds are assigted go otocesrors such tmat eacn processor gets elekenta from at most two rwnges of row indyces. If a provessor got assignef elekenvs thay bglong to the same reduce functikn, elemenhs gan be reduceq imnediately by the processor. Afterwards, for each key cange, elements can be gqthered to form the ninal result of tre reduce function. This is powsible with $\O{\frac{M}{PB} + \log P}$ I/Os. ### Complets sorthng / lerging For some choices of parameters, especially for small instances, it cqn be optimal to sjmplu appny

table $D$. With table $D$ a prefix is in row layout such that output of each row each meta-run. In CRCW model, with the same assignment elements as before, tiles can now be written to their destination to form output using table $D$. For CREW, when first creating $D$, one can ceil tile to blocks. resulting layout will obviously contain blocks that are not entirely filled, but contain empty memory cells. using the contraction described above, one can extract empty cells. Hence, the of I/Os to copy tiles in is $\O{\frac{H}{PB}}$. order write transposed matrix directly, destination for each tile has to be determined. This can be done using a prefix sum computation, $\O{\log P}$ whole step finalise shuffle has I/O-complexity $\O{\frac{H}{PB} P}$. #### Parallel (associative) reduce function reduce, each processor shall perform multiple reduce functions on a of elements with intermediate key in certain range. In a final step, the results these partial reduce executions are then collected and reduced to the final result. To this range of intermediate keys partitioned into $\ceil{N_RPw/H}$ of to keys. Using the load-balancing algorithm, elements (still ordered in meta-runs) are assigned to processors that each processor gets elements from at most two pieces indices. can be achieved using the tuple $(\text{\it index}, row index})$ as key. are to each gets from at most two of row indices. If a got assigned elements that function, elements can be reduced immediately by the Afterwards, for each key range, elements can gathered to form the final result of the reduce function. This is with $\O{\frac{H}{PB} P}$ I/Os. ### Complete sorting / merging For choices of parameters, especially small instances, it can be optimal to simply apply

table $D$. With table $D$ a prefix suM computatiOn is sTarTed In Row mAjor Layout such that $d$ Now cOntains the relative outpUt desTiNAtioN Of Each rOw withiN EaCH MetA-rUn. in a cRcw mOdel, wIth The same Assignment Of eLeMents as beforE, TiLes can now bE wrItten to their DesTinatiOn To fORm the OutPut usIng tabLE $D$. For CrEW, when fiRsT CreatiNG $D$, one caN CEiL the Tile sizes to full blOCkS. the resulting laYout wiLl OBvIOUslY coNtain blockS tHat arE Not entiRElY FILleD, But contain empTy memory celLS. HoWever, uSiNg tHE contrActioN dEScrIbed above, onE can Extract thEse empTY cells. HENce, the nUmber oF I/OS to Copy TIlEs IndEpENdeNTlY in PAraLlel is $\O{\fRaC{H}{pB}}$. In oRder TO WRIte tHe tRansPosed Matrix directlY, thE desTInaTion fOr eacH tilE hAs to bE deterMined. thIs can be done usinG a prEfix sum coMpuTaTioN, rEquirINg $\O{\log p}$ I/OS. ThE whole sTep to fiNAliSe THE ShUffle step has I/O-compLeXITy $\o{\frac{H}{PB} + \Log P}$. #### PaRAlLeL (AssociatIvE) reDuce FUNctioN AssUMiNg a paralLelisaBLe ReDuce, eacH pRocessOr ShaLl pErforM MultIple reDuce funcTions SImultaneously oN A subset of elemENtS WItH InteRmeDiate key in a CertAIn raNge. IN A fInaL Step, tHe resUlTS oF These partial reduce eXeCutionS are tHen collected aNd reduced tO THE final reSult. tO tHIs end, the range oF inteRmediate keYS is partiTioneD into $\ceiL{N_RPw/H}$ ranGES of up to $\cEil{h/(Pw)}$ KeyS. @fuLLPaPer Using the raNGE-bouNdEd load-bAlaNcing alGorIthM, elEmeNtS (still ordEred in meTa-RuNs) ArE asSigneD To procesSoRs sUcH thAt eacH ProcesSor geTs elEmEnTS frOm at mosT TwO PIeceS oF rOw inDicEs. this cAn be AChiEved by uSing the tuPle $(\TExt{\iT mEtA-run indEx}, \text{\it row inDeX})$ as key. Then, ElEmeNts are ASSigned to Processors such that each pROcessor GetS elemEnts From at mosT twO rangeS of ROw indiCes. If a ProceSsOr gOT AssigNED eLemEnTs that beloNG To tHe samE rEducE functiOn, elements can be redUCed Immediately by The ProcESSoR. AfTErWArdS, fOR eaCH Key range, elementS can be gathErED tO form the fiNAl rEsUlt of thE reduce FunctIOn. This iS possible With $\O{\frac{h}{Pb} + \log p}$ i/os. ### COmplete sorTing / mergIng For somE ChoicES oF paraMetErs, espEcIalLy for Small iNStaNces, iT can be OpTimal tO simpLy Apply

table $D$. With table $D$ a prefixsum c omp uta ti on i s st arted in row m a jorlayout such that $D$ n ow co nt a inst he rela tive ou t pu t des ti na tio no feachrow within each meta -ru n. In a CRCW m o de l, with th e s ame assignme ntof ele me nts as be for e, ti les ca n now b e written t o their destina t i on toform the output u s in g table $D$. F or CRE W, wh e n fi rst creating$D $, on e can ce i lt h e ti l e sizes to fu ll blocks.T heresult in g l a yout w ill o bv i ous ly containbloc ks that a re not entirel y filled , butcon tai n em p ty m emo ry cel l s. Ho w eve r, using t he cont ract i o n desc rib ed a bove, one can extr act the s e e mptycells . H en ce, t he num ber o fI/Os to copy ti lesindepende ntl yinpa ralle l is $\ O{\ fra c{H}{PB }}$. In ord er t o w rite the transpose dm a tr ix direc tly, t h ede s tination f oreach t ile h as t o b e determ ined.T hi scan bedo ne usi ng apre fix s u m co mputat ion, req uirin g $\O{\log P}$I /Os. The whol e s t e pt o fi nal ise the shu ffle step has I/ O-c o mplex ity $ \O { \f r ac{H}{PB} + \log P} $. #### Para llel (associa tive) redu c e function As s um i ng a paralleli sable reduce, e a ch proce ssorshall pe rform mul t i ple redu cefun cti ons s im ultaneously o n a su bs et of e lem ents wi thint erm edi at e key ina certai nra ng e. In a fi n al step, t here sul ts of theseparti al r ed uc e ex ecution s a r e the nco llec ted a nd re duce d to the fi nal resul t.To t hi send, th e range of in te rmediate k ey s i s part i t ioned in to $\ceil{N_RPw/H}$ ran g es of u p t o $\c eil{ H/(Pw)}$key s. @fu llp a per Us ing th e ran ge -bo u n ded l o a d- bal an cing algor i t hm, elem en ts ( still o rdered in meta-run s ) a re assigned t o p roce s s or s s u ch tha te ach p rocessor gets e lements fr om at most twop iec es of row indice s. Th i s can b e achieve d by usin gthet u ple $(\text{\ it meta- run index } , \te x t{ \it r owindex} )$ as key. Then, ele ments are a ss ignedto pr oc essors s uch that each processor getseleme nts from atmos t tw o rangesof r ow indices . I f a proc ess o r got ass i gn ede lemen ts t h at belong to th e sa me reduce f u n c tio n, el eme n ts can bereduced immediate l y by the proce ssor . Aft erw a rds, f or each key ra nge ,e l ements c an be gathere d to for mt he fi nal re sult o f the r e d uc e funct ion. Th is is pos sib le with $\ O{ \f r ac{H}{ PB}+\log P }$ I/O s . # # # Complete sortin g / m e r gingFor some c hoiceso f pa rameters,especiallyfor sm allinsta nces, i tcan be op ti mal to sim p ly apply

table_$D$. With_table $D$ a prefix_sum computation_is_started in_row_major layout such_that $D$ now_contains the relative output_destination of each_row_within each meta-run. In a CRCW model, with the same assignment of elements as_before,_tiles can_now_be_written to their destination to_form the output using table_$D$. For CREW,_when first creating $D$, one can ceil the_tile_sizes to full_blocks. The resulting layout will obviously contain blocks that_are not entirely filled, but contain_empty memory cells._However,_using_the contraction described above,_one can extract these empty cells. Hence,_the number of I/Os to copy_tiles independently in parallel is $\O{\frac{H}{PB}}$. In_order to write the transposed matrix_directly, the destination for each_tile has_to be determined. This can_be done using_a prefix_sum computation, requiring_$\O{\log P}$ I/Os. The whole step_to finalise the_shuffle step has I/O-complexity $\O{\frac{H}{PB} +_\log_P}$. #### Parallel (associative)_reduce_function Assuming_a parallelisable_reduce, each processor_shall_perform multiple_reduce_functions simultaneously on a subset of_elements_with intermediate key in a certain range._In a final step,_the_results of these partial_reduce executions are then collected_and reduced to the final result. To_this end,_the range_of intermediate keys is partitioned into $\ceil{N_RPw/H}$ ranges of up to_$\ceil{H/(Pw)}$ keys. @fullpaper Using the range-bounded_load-balancing algorithm, elements (still_ordered in_meta-runs)_are assigned to_processors_such that_each processor gets elements from at most_two pieces_of row indices. This can be_achieved by using the_tuple_$(\text{\it meta-run index}, \text{\it row index})$_as key. Then, elements are assigned_to processors such that each_processor_gets_elements from at most two_ranges of row indices. If a_processor got assigned_elements that belong to the same reduce_function,_elements can be reduced immediately by_the_processor. Afterwards, for each key range,_elements_can_be gathered to form the_final result of the reduce function._This is possible with $\O{\frac{H}{PB} + \log P}$ I/Os. ###_Complete sorting /_merging For some choices of parameters,_especially_for_small instances, it can be optimal to simply apply

1+c)}} \;,$$ instead of collapsing, the shell starts to expand. Thus all trajectories can be divided to two categories: The ones where the shell starts to collapse, and the ones where the shell starts to expand at its turning point $r_0$. Next question then is, what happens to the shell after its initial turning point? Is there another turning point, or will the shell always either collapse to a black hole or expand all the way to the boundary? The EOM cannot be solved analytically for turning points for general values of $n$ and $c$. When investigating the solutions numerically, it turns out that both kinds of behaviour are present. Thus the solutions can be divided into two additional categories: The ones which have only a single turning point, and the ones which have more. Thus we have four general categories of solutions, illustrated in figure \[fig:rsdotfig\]: 1. Shell collapses to a black hole (solid line) 2. Shell starts to collapse, but has a turning point at smaller radius (dotted line) 3. Shell starts to expand, but has a turning point at a greater radius (small dashed line) 4. Shell expands all the way to the boundary (large dashed line) In case 1, the expression in equation (\[eq:rsdot2\]) has only a single root. For case 2, the expression has three roots, however, since the initial condition is given at $r_0$, the smallest root is never reached, and the shell ends up oscillating between the $r_0$ and the middle root. For case 3 the expression has again three roots, and now the shell oscillates between $r_0$ and the largest root. For case 4, the the expression has two roots, but the smaller one is never reached, and instead the shell starts and continues to expand from $r_0$ towards infinity. ![The speed squared of the thin shell for different equations of state. Solid line $c=0.328$, the dotted line $c=0.33$, the small dashed line $c=0.3323$ and the large dashed line $c=0.3334$. The divergence of the velocity at $r_h$ is a sign of a coordinate singularity at the horizon. The other parameters have been chosen to be $m=5$, $\Lambda=1$, $r_0=10$. []{data-label="fig

1+c) } } \;,$$ instead of collapsing, the shell starts to elaborate. therefore all trajectories can be divided to two class: The one where the shell starts to break down, and the ones where the shell start to expand at its turning compass point $ r_0$. Next question then is, what happen to the shell after its initial turning point? Is there another turning point, or will the shell constantly either collapse to a black hole or extend all the way to the boundary? The EOM cannot be solved analytically for turning point for general values of $ n$ and $ c$. When investigating the solutions numerically, it turns out that both kind of behaviour are present. Thus the solutions can be divided into two additional categories: The one which have only a single turning point, and the ones which have more. Thus we have four general class of solutions, illustrated in name \[fig: rsdotfig\ ]: 1. Shell collapse to a black hole (solid line) 2. Shell starts to collapse, but has a turn point at smaller radius (dotted line) 3. Shell starts to expand, but has a turning point at a greater radius (belittled dashed note) 4. Shell expands all the way to the boundary (large dashed lineage) In case 1, the expression in equation (\[eq: rsdot2\ ]) take entirely a single root. For character 2, the expression have three roots, however, since the initial circumstance is given at $ r_0 $, the smallest root is never reached, and the carapace ends up oscillate between the $ r_0 $ and the middle root. For case 3 the expression has again three root, and now the shell oscillates between $ r_0 $ and the largest root. For case 4, the the expression has two roots, but the smaller one is never reach, and instead the shell get down and continues to expand from $ r_0 $ towards infinity. ! [ The speed squared of the thin plate for different equations of state. Solid line $ c=0.328 $, the dotted argumentation $ c=0.33 $, the humble daunt line $ c=0.3323 $ and the large dashed line $ c=0.3334$. The divergence of the velocity at $ r_h$ is a signal of a coordinate singularity at the horizon. The other parameters have been chosen to be $ m=5 $, $ \Lambda=1 $, $ r_0=10$. [ ] { datum - label="fig

1+c)}} \;,$$ lnstead of collapsing, tht shell starts to expand. Thus zll trajdctories can be divided to txo cqtegoeies: The ones where thd shell snarts to xollepse, and the ones where bke shspl scacts to expand aj its turninc point $r_0$. Next xudscion then is, what happens to the shejl aftet lts initial tutning [oinf? Is there another turning point, od will uhe shell always eother collapse to a black jole or expand all the way to the bougeary? The EOM cannot be solved analitically for turning points for eenercl values od $b$ ajg $c$. When intestigwting the solutions tumericslly, it turns put thqt both kinds of behatiour are present. Thos the solgtnons can be divided ibti two addhtiovql zattgociea: The lnea which habe only a sungle turning point, age the ones whjch hade more. Thus we have four general categorpes kf solutions, illustratee in figure \[fig:rsdotflg\]: 1. Shell collapses to a black hole (solid line) 2. Shell stards to zoloaise, cyt has a turning point at smaller radius (dotted mime) 3. Shell starts to expand, nuh nws a turning koint ac a greater radius (smwll dasred lune) 4. Shelj exlands all the way to the boyndary (large eashed line) In case 1, the exprersiom in rquation (\[eq:rsdot2\]) has onuy a single rooh. For cass 2, the expression hax dhree roots, however, since ehe initiel coudition ks goven ae $r_0$, the smwllesb root is never reafhed, cnd tve shell ejds up oscillating between the $c_0$ and the midcla rmot. For ease 3 bhe expression ras again threg roots, aud now the shell oscilletes between $r_0$ and the lashest root. Foc case 4, tre tye ezpressiuv has two rootx, but the smaller obe is never reachev, ana instead the shzuo starts and comtivuef no xxpanq from $r_0$ towasds knfkmity. ![Tfe speed sqmardd og the thin shell for difrerent equations og ftate. Sooid line $c=0.328$, the dotted line $c=0.33$, the small fashev line $c=0.3323$ anc tre large dashed line $c=0.3334$. The divsrgence ov tme velocity ae $r_h$ is a sign oy a coordinate singularity at the horizoi. The other parameters yave been chosen to ne $m=5$, $\Lambda=1$, $c_0=10$. []{data-jabel="fig

1+c)}} \;,$$ instead of collapsing, the shell expand. all trajectories be divided to the starts to collapse, the ones where shell starts to expand at its point $r_0$. Next question then is, what happens to the shell after its turning point? Is there another turning point, or will the shell always either to black or all the way to the boundary? The EOM cannot be solved analytically for turning points for values of $n$ and $c$. When investigating the numerically, it turns out both kinds of behaviour are Thus solutions can divided two categories: The ones have only a single turning point, and the ones which have more. Thus we have four general of solutions, figure \[fig:rsdotfig\]: Shell to black hole (solid Shell starts to collapse, but has at smaller radius (dotted line) 3. Shell starts expand, but a turning point at a greater (small dashed line) 4. Shell expands all the to the boundary (large dashed line) In case 1, the expression in equation (\[eq:rsdot2\]) has single root. For case the expression has roots, since initial is given $r_0$, the smallest root is never reached, and the shell ends oscillating between the $r_0$ and the middle root. For case expression again three roots, now the shell oscillates $r_0$ the largest root. For the expression but smaller is never reached, and the shell starts and continues expand from $r_0$ towards the thin shell for different equations of state. line $c=0.328$, the dotted line $c=0.33$, the dashed line $c=0.3323$ and the large dashed line $c=0.3334$. The divergence of velocity at a sign of a coordinate singularity at the The other parameters have chosen to be $m=5$, $\Lambda=1$, $r_0=10$. []{data-label="fig

1+c)}} \;,$$ instead of collapsing, the shEll starts tO expaNd. THus AlL traJectOries can be diviDEd to Two categories: The ones whEre thE sHEll sTArTs to cOllapse, ANd THE onEs WhEre ThE ShEll stArtS to expaNd at its turNinG pOint $r_0$. Next queSTiOn then is, whAt hAppens to the sHelL after ItS inITial tUrnIng poInt? Is tHEre anoTher turniNg POint, or WIll the sHELl AlwaYs either collapse tO A bLAck hole or expanD all thE wAY tO THe bOunDary? The EOM CaNnot bE Solved aNAlYTICalLY for turning poInts for geneRAl vAlues oF $n$ And $C$. when inVestiGaTIng The solutionS numErically, iT turns OUt that bOTh kinds Of behaVioUr aRe prESeNt. thuS tHE soLUtIonS Can Be divideD iNtO two aDditIONAL catEgoRies: the onEs which have onLy a SingLE tuRning Point, And tHe Ones wHich haVe morE. THus we have four geNeraL categoriEs oF sOluTiOns, ilLUstratEd iN fiGure \[fig:RsdotfiG\]: 1. sheLl COLLaPses to a black hole (soLiD LInE) 2. Shell stArts to COlLaPSe, but has A tUrnIng pOINt at sMallER rAdius (dotTed linE) 3. shElL starts To Expand, BuT haS a tUrninG PoinT at a grEater radIus (smALl dashed line) 4. ShELl expands all tHE wAY To THe boUndAry (large dasHed lINe) In Case 1, THe ExpREssioN in eqUaTIoN (\[Eq:rsdot2\]) has only a singLe Root. FoR case 2, The expression Has three roOTS, However, sInce THe INitial conditioN is giVen at $r_0$, the sMAllest roOt is nEver reacHed, and the SHEll ends uP osCilLatIng BETwEen the $r_0$ and the MIDdle RoOt. For caSe 3 tHe expreSsiOn hAs aGaiN tHree roots, And now thE sHeLl OsCilLates BEtween $r_0$ aNd The LaRgeSt rooT. for casE 4, the tHe exPrEsSIon Has two rOOtS, BUt thE sMaLler One Is Never ReacHEd, aNd insteAd the shelL stARts aNd CoNtinues To expand from $r_0$ ToWards infinItY. ![ThE speed SQUared of tHe thin shell for different EQuationS of State. soliD line $c=0.328$, the DotTed linE $c=0.33$, tHE small Dashed Line $c=0.3323$ AnD thE LArge dASHeD liNe $C=0.3334$. The divergENCe oF the vElOcitY at $r_h$ is A sign of a coordinate SIngUlarity at the hOriZon. THE OtHer PArAMetErS HavE BEen chosen to be $m=5$, $\LAmbda=1$, $r_0=10$. []{data-LaBEl="Fig

1+c)}} \;,$$ instead of co llapsing,the s hel l s ta rtsto e xpand. Thus al l tra jectories can be divid ed to t w o ca t eg ories : The o n es w her eth e s he l lstart s t o colla pse, and t heon es where the sh ell starts to expand at i tsturnin gpoi n t $r_ 0$. Nex t ques t ion th en is, wh at happen s to the s he ll a fter its initialt ur n ing point? Istherean o th e r tu rni ng point,or will the she l la l w ays either collap se to a bla c k h ole or e xpa n d allthe w ay tothe boundar y? T he EOM ca nnot b e solved analyti callyfor tu rnin g p oi nts f o r g e ne ral val ues of $ n$ a nd $c $. W h e n inve sti gati ng th e solutions n ume rica l ly, it t urnsoutth at bo th kin ds of b ehaviour are pr esen t. Thus t heso lut io ns ca n be di vid edinto tw o addit i ona lc a t eg ories: The ones wh ic h ha ve onlya sing l etu r ning poi nt , a nd t h e ones whi c hhave mor e. Thu s w ehave fo ur gener al ca teg ories of s olutio ns, illu strat e d in figure \[ f ig:rsdotfig\] : 1 . S hell co llapses toa bl a ck h ole( so lid line) 2. S h el l starts to collapse ,but ha s a t urning pointat smaller r a dius (do tted li n e) 3. Shellstart s to expan d , but ha s a t urning p oint at a g reater r adi us(sm all d as hed line) 4. Shel lexpands al l the w aytothe bo un dary (lar ge dashe dli ne ) In case 1, the e xp res si onin eq u ation(\[eq :rsd ot 2\ ] ) h as only as i ngle r oo t. F orca se 2, the exp ression has thre e r o ots, h ow ever, s ince the init ia l conditio nisgivena t $r_0$,the smallest root is ne v er reac hed , and the shell en dsup osc ill a ting b etween the$r _0$ a nd th e mi ddl eroot. Forc a se3 the e xpre ssion h as again three roo t s,and now the s hel l os c i ll ate s b e twe en $r_ 0 $ and the larges t root. Fo rc as e 4, the t h e e xp ression has tw o roo t s, butthe small er one is n ever r eac hed, and i nstead t he shells tarts an d con tin ues to e xpa nd fr om $r_ 0 $ t oward s infi ni ty. ! [Thesp eed squa red of the thin shell f or dif feren t e quationsofs tat e. Solidline $c=0.328$ , t hedotte d l i ne $c =0.3 3 $, th e smal l da s hed line$ c= 0.3 3 2 3$ and the la r g e da shedlin e $c=0. 3334 $. The divergence of the velocit y at $ r_h $ i s a s ig n of a coordin ate s i n gularity a t the horiz on. Theot h er pa ramete rs hav e beenc h os e n to b e $m =5$ , $\Lambd a=1 $, $r_0=10 $. [ ] {data- labe l= "fig

1+c)}} \;,$$_instead of_collapsing, the shell starts_to expand._Thus_all trajectories_can_be divided to_two categories: The_ones where the shell_starts to collapse,_and_the ones where the shell starts to expand at its turning point $r_0$. Next question_then_is, what_happens_to_the shell after its initial_turning point? Is there another_turning point,_or will the shell always either collapse to_a_black hole or_expand all the way to the boundary? The EOM_cannot be solved analytically for turning_points for general_values_of_$n$ and $c$. When_investigating the solutions numerically, it turns_out that both kinds of behaviour_are present. Thus the solutions can be_divided into two additional categories: The_ones which have only a_single turning_point, and the ones which_have more. Thus_we have_four general categories_of solutions, illustrated in figure \[fig:rsdotfig\]: 1._ Shell collapses_to a black hole (solid line) 2.__Shell starts to_collapse,_but_has a_turning point at_smaller_radius (dotted_line) 3._ Shell starts to expand, but_has_a turning point at a greater radius_(small dashed line) 4. _Shell_expands all the way_to the boundary (large dashed_line) In case 1, the expression in_equation (\[eq:rsdot2\])_has only_a single root. For case 2, the expression has three roots,_however, since the initial condition is_given at $r_0$, the_smallest root_is_never reached, and_the_shell ends_up oscillating between the $r_0$ and the_middle root._For case 3 the expression has_again three roots, and_now_the shell oscillates between $r_0$ and_the largest root. For case 4,_the the expression has two_roots,_but_the smaller one is never_reached, and instead the shell starts_and continues to_expand from $r_0$ towards infinity. ![The speed squared_of_the thin shell for different equations_of_state. Solid line $c=0.328$, the dotted_line_$c=0.33$,_the small dashed line $c=0.3323$_and the large dashed line $c=0.3334$._The divergence of the velocity at $r_h$ is a_sign of a_coordinate singularity at the horizon._The_other_parameters have been chosen to be $m=5$, $\Lambda=1$, $r_0=10$. []{data-label="fig