Reports@SCM, 10 (1)

Page 1


https://revistes.iec.cat/index.php/reports ISSNelectronicedition:2385-4227

EditorialTeam

Editor-in-chief

SimoneMarchesi,UniversitatdeBarcelona

AssociateEditors

MontserratAlsina,UniversitatPolit`ecnicadeCatalunya XavierBardina,UniversitatAut`onomadeBarcelona EnricCosme,UniversitatdeVal`encia N´uriaFagella,UniversitatdeBarcelona XavierMassaneda,UniversitatdeBarcelona ArturNicolau,UniversitatAut`onomadeBarcelona LauraPrat,UniversitatAut`onomadeBarcelona JoaquimRo´e,UniversitatAut`onomadeBarcelona OriolSerra,UniversitatPolit`ecnicadeCatalunya PabloSevilla,UniversitatPolit`ecnicadeVal`encia EnricVentura,UniversitatPolit`ecnicadeCatalunya

EditorialAssistant

RosaRodr´ıguez

FocusandScope

Reports@SCM isanon-profitelectronicresearchjournalonMathematicspublishedbytheSocietat CatalanadeMatem`atiques(SCM)whichoriginatedfromthedesiretohelpstudentsandyoung researchersintheirfirststepsintotheworldofresearchpublication.

Reports@SCM publishesshortpapersinallareasofpuremathematics,appliedmathematics,and mathematicalstatistics,includingalsomathematicalphysics,theoreticalcomputerscience,andany applicationtoscienceortechnologywheremathematicsplaysacentralrole.Tobeconsideredfor publicationin Reports@SCM anarticlemustbewritteninEnglish(withanabstractinCatalan), bemathematicallycorrect,andcontainsomeoriginalinterestingcontribution.Allsubmissionswill followapeerreviewprocessbeforebeingacceptedforpublication.

Researchannouncementscontainingpreliminaryresultsofalargerprojectarealsowelcome.Inthis case,authorsarefreetopublishinthefutureanyextendedversionofthepaperelsewhere,withthe onlyconditionofmakinganappropriatecitationto Reports@SCM

Weespeciallywelcomecontributionsfromresearchersattheinitialperiodoftheiracademiccareers, suchasMasterorPhDstudents.Wewishtogivespecialattentiontotheauthorsduringthewhole editorialprocess.Weshalltakespecialcaretomaintainareasonablyshortaveragetimebetween thereceptionofapaperanditsacceptance,andbetweenitsacceptanceanditspublication.

Allmanuscriptssubmittedwillbe peerreviewed byatleastonereviewer.Finaldecisionsonthe acceptanceofmanuscriptsaretakenbytheeditorialboard,basedonthereviewer’sopinion.

Thecontentspublishedin Reports@SCM aresubject,unlessotherwiseindicatedinthetextorinthe graphicmaterial,toanAttribution-NonCommercial-NoDerivs3.0Spain(by-nc-nd)licensefrom CreativeCommons,thefulltextofwhichcanbeconsultedat https://creativecommons.org/licenses/ by-nc-nd/3.0/es/deed.en.Therefore,thegeneralpublicisauthorizedtoreproduce,distribute,and communicatetheworkaslongasauthorshipandthepublishingentityareacknowledged,andno commercialuseorderivativeworkismade.

©Theauthorsofthearticles

EditedbySocietatCatalanadeMatem`atiques,Institutd’EstudisCatalans(IEC) CarrerdelCarme47,08001Barcelona.

https://scm.iec.cat

Tel`efon:(+34)933248583 scm@iec.cat

Fax:(+34)932701180

Institutd’EstudisCatalans https://www.iec.cat informacio@iec.cat

https://revistes.iec.cat/index.php/reports ISSNelectronicedition:2385-4227

ANELECTRONICJOURNALOFTHE

Generatinguniformspanningtrees fromconditioned Bienaym´e–Galton–Watsontrees

AlbertVives UniversitatPolit`ecnica deCatalunya avivesbatalla@gmail.com

Resum (CAT)

Aquestarticleexploralageneraci´od’arbresgeneradorsuniformes(UST),fonamentalsencombinat`oriaiprobabilitat,ambaplicacionsenteoriadexarxesi f´ısica.UtilitzantprocessosdeBienaym´e–Galton–Watson(BGW)condicionatsa unnombrefixdev`ertexs,s’introdueixunm`etodepergenerararbresgeneradors uniformesis’examinenpropietatsestructuralscoml’al¸cadail’amplada.

Abstract (ENG)

Thisreportexploresuniformspanningtree(UST)generation,essentialincombinatoricsandprobabilitywithapplicationsinnetworktheoryandphysics.Using conditionedBienaym´e–Galton–Watson(BGW)processes,itintroducesamethod togenerateUSTs.Rigorousproofsshowthatconditioningonafixednumberof verticesensuresuniformdistributionandletusexaminestructuralpropertieslike heightandwidth.

Keywords: randomtrees,Bienaym´e–Galton–Watsontrees. MSC(2020): Primary05C05.Secondary60J80.

Received: February17,2025. Accepted: October2,2025. 1 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),1–12;DOI:10.2436/20.2002.02.45.

1.Introduction

Thegenerationofuniformspanningtrees(USTs)isacrucialproblemincombinatoricsandstochastic processes,withapplicationsinnetworkanalysis,randomstructures,andstatisticalphysics.Aspanning treeisaconnectedsubgraphofagraphthatincludesallverticeswithoutformingcycles.Whensampled uniformlyatrandom,eachspanningtreehasanequalprobabilityofbeingchosen,presentingmathematical challengesindefininganappropriaterandommodelsusceptibleofbeinganalyzed.Thisdrivesthesearch foreffectiveapproaches,andthisreportproposesnovelmethodsusingconditionedBienaym´e–Galton–Watson(BGW)processes.Theseprocesses,traditionallyusedtomodelpopulationgrowth,areadaptedto generatespanningtreesunderuniformdistributions,bridgingcombinatoricswithstochasticprocesses.

ThisworkprovidesdetailedmathematicalformulationsandrigorousproofsdemonstratingthatconditioningBGWprocessesonafixednumberofverticesresultsinuniformspanningtrees.Itexploresclassical distributionssuchasPoisson,Geometric,andBernoullibranchingtoillustratethemethod’sversatilityand correctness.Italsodelvesintostructuralcharacteristicsofrandomtrees,analyzingheightandwidthas primarymetricsforcomplexity.Resultsontheseparametersofferinsightsintotheasymptoticbehaviorof USTs,supportedbyprobabilisticboundsandexamples.Thesefindingscontributetoadeeperunderstanding ofUSTsandtheirgenerationthroughprobabilistictechniques.

2.Obtaininguniformspanningtrees fromBGWtrees

Dependingontheprobabilitydistributiongoverningthereproductionofsomeindividuals,onemayobtain differenttypesoftrees.Specifically,weaimtoprovethatifweconditionaBienaym´e–Galton–Watsontree tohave n verticesandfixcertainknowndistributions,weobtainvarioustypesofuniformrandomtrees. ThefirstdescriptionofrandomtreesfromconditioningaBGWprocessbyitstotalprogenycanbetraced backtoKolchin[3]andAldous[2].

Lemma2.1. LetZbeanonnegativeinteger-valuedrandomvariableandletXbea BGW(Z ) process. Let Tn denotetheclassoftreeswithnverticeswhichcanbegeneratedbytheprocessandletT ∈Tn be atreeintheclass.IftheprobabilitythatTisgeneratedbytheprocessdependsonlyonthenumberof verticesn,then

Proof. Letusdenote P(X = T )by f (n),where n isthenumberofverticesof T .Then,

(|X | = n)= |T

|f (n), sincethelastprobabilityisthesumoftheprobabilitiesofobtainingeachofthetreeswith n vertices.Now, whenconditionedonhaving n vertices,allpossibletreesareequallylikelytooccur:

= T ||

Wenowaimtoprovethat,byselectinganappropriateoffspringdistribution,itispossibletogenerate severalwell-knowntypesoftrees.

Theclassoflabeledtreeswith n nodesisalsocalledtheclassofCayleytrees,duetotheCayleyformula enumeratingthem,itsnumberbeing nn 2.WhileBienaym´e–Galton–Watsontreesarenaturallyrooted, Cayleytreesarenot;inthiscontextweconsiderCayleytreestoberootedbyfixingonedistinguishedvertex astheroot.

Theorem2.2. ConditioningaBienaym´e–Galton–Watsontreewithoffspringdistribution Poisson(1) on havingnverticesresultsinaCayleytreewithnverticesgenerateduniformlyatrandom.

Proof. EveryrootedCayleytreecanbegeneratedfromaPoisson(1)BGWprocessbyalabelingofits vertices.Let T beaspecificCayleytreewith n vertices.Toprovethateachtreecanbeobtaineduniformly, wefirstneedtoorderallsiblingsetsin T byincreasingvertexlabels.Let χ1,..., χn representthenumber ofchildrenofeachnode,wheretheverticesareindexedstartingfromtherootandthenrecursivelyvisiting thechildrenfromlefttoright.Thefirstrequirementforgenerating T isensuringthecorrectnumberof descendantsforeachvertex.Sincetheserandomvariablesaremutuallyindependent,theprobabilityof obtainingaspecificnumberofchildrenforallverticesistheproductoftheirindividualprobabilities.

Thesecondrequirementisassigningthecorrectlabeling,asweareconsideringlabeledtrees.With n vertices,thereare n!possiblewaystolabelthem.Moreover,thechildrenofthe i -thvertexcanbe permutedin χi !distinctwaysforeach i =1,..., n,resultinginthesametree.Thus,thefinalcalculation canbeexpressedclearlyasfollows:

Sincethelastprobabilitydependsonlyonthenumberofvertices,anditisknownthatthereexist nn 2 labeledtrees,byLemma 2.1 thetheoremisproved.

Theclass Bn offullbinarytreesof n verticesisthefamilyofunlabelledrootedplanetreeswhereevery nodehastwoorzerochildren.Beingplanemeansthattreeshavedistinguishedleftandrightsubtrees.The treesdepictedinFigure 1 areconsideredtobedistinct.

Afullbinarytreewith n nodeshasanoddnumber n ofverticesand m =(n +1)/2leaves.Thenumber ofsuchtreesisgivenbytheCatalannumber Cm 1

Theorem2.3. Conditioninga 2Bernoulli(1/2) Bienaym´e–Galton–Watsontreeonhavingnverticesresults inabinarytreewithnverticesgenerateduniformlyatrandom.

Reports@SCM, 10 (2025),1–12;DOI:10.2436/20.2002.02.45.

Figure1:Twodistinctbinarytrees.

GeneratinguniformspanningtreesfromconditionedBienaym´e–Galton–Watsontrees

Proof. Itisclearthatonecanobtaineverybinarytreefroma2Be(1/2)BGWtreeandviceversa.Let T be aparticularbinarytreewith n vertices.Itisknownthatinabinarytreethereare(n 1)/2internalnodes, and(n +1)/2leafs.ToapplyLemma 2.1,weneedtocalculate P(X = T ).Let χ ∼ 2Be(1/2).

Sincethelastprobabilitydependsjustonthenumberofvertices,byLemma 2.1 wehaveprovedthe theorem.

Theclass Pn oforderedplanetreeswith n verticesisthefamilyofunlabelledrootedtreeswherethe childrenofeverynodeareorderedfromlefttorightintheplane.

Theorem2.4. Conditioninga Geom(1/2) Bienaym´e–Galton–Watsontreeonhavingnverticesresultsinan orderedplanetreewithnverticesgenerateduniformlyatrandom,where Geom(1/2) denotesthegeometric distributionon {0,1,2,... }

Proof. ItcanbeobservedthateveryorderedplanetreecanbeobtainedfromaGeom(1/2)BGWtree,and conversely,everyGeom(1/2)BGWtreecorrespondstoanorderedplanetree.Let T beaspecificordered planetreewith n vertices.Togeneratesuchatree,itisnecessarytoaccountforbothinternalnodesand leaves.

Theprobabilityofaninternalnodehavingexactly k childrenis(1/2)k (1/2),wherethefirstfactor representstheprobabilityofsuccessfullyhaving k children,andthesecondfactoraccountsfortheprobability ofnoadditionalchildren.Foraleaf,therequirementissimplytohavenochildren,whichoccurswith probability1/2.

Consequently,theprobabilityofachievingthecorrectnumberofchildrenforeachvertexisindependent ofwhetherthevertexisaninternalnodeoraleaf.Thisprobabilityisgivenby(1/2)χi +1,where χi denotes thenumberofchildrenofvertex i .Thus,thefollowingconclusionnaturallyarises:

Furthermore,itisusuallyknownthatthenumberoforderedplanetreeswith n verticesis Cn 1,where Cn isthe n-thCatalannumber.Hence,byLemma 2.1,thetheoremhasbeenproved.

3.Studyingsomeparametersofrandomtrees

Thischapterisfocusedonthestudyofkeyparametersofrandomtrees:height,width,andthenumberof leaves.AllresultspresentedinthissectionarederivedfromthestudyL.Addario-Berry,L.Devroye,and S.Jansondidin[1].Theapproachin[1]appliesnotonlytorandomCayleytrees,buttoanyfamilyof treesarisingfromaBGW(χ)treeaslong χ hasexpectation1andfinitevariance(criticalBGWtrees). Furthermore,explicitproofsareprovidedforcertainconceptsthatareoftenassumedwithoutfurther justificationintheliterature.Theexplorationoftheseparametersoffersadeeperunderstandingofrandom trees,contributingnewinsightsintoestablishedresultsinthisfield.

https://revistes.iec.cat/index.php/reports

3.1.Somepreliminaries

TheBreadth-FirstSearch(BFS)onaBGWtreeisanalgorithmusedtoexplorethetreelevelbylevel, startingfromtheroot.BFSexploresthetreebyvisitingallnodesatonelevelbeforemovingtothenext one,ensuringthatnodesareprocessedinincreasingdistancefromtheroot.Hence,thissearchkeepsa queue Q with Qi nodesatthe i -thstep,with Q0 =1.Duringtheexplorationofavertex,itsoffspringare addedtothebackofthequeue.Then,onecaneasilyobtainthefollowingrecursion:

i = Qi 1 1+ χ

, where χi areindependentandidenticallydistributedcopiesoftheoffspringdistribution χ.Hence,bythis recursion, Qj =1+ ˜ Sj ,where ˜ Sj := j i =1(χi 1)= Sj j.Thetreeiscompletelyexploredwhen Qj =0. Inthiscase, ˜ Sn = 1.

Definition3.1. Let Tn bearandomtreewith n nodes.Thewidthofatree Tn,denotedby W (Tn),is themaximumnumberofnodesatanydepthlevel.Generally,let d (v )denotethedepthofanode v ina rootedtree T .Then width(Tn)=max k≥0 |{v ∈ V (Tn): d (v )= k}|

Wenowpresentkeylemmasthatarenecessaryforstudyingtheexpectedwidthofarandomtree.

Lemma3.2 (Raney’sLemma) Leta1, a2,..., an beasequenceofintegerssuchthat n i =1 ai = 1

ThenthereexistsauniqueindexssuchthatthecyclicpartialsumsSk = k 1 j =0 a(s+j )mod n fork = 1,2,..., n, satisfy:

(i)Sk > 0 for 1 ≤ k < n.

(ii)Sn = 1

Lemma3.3. SupposethattheindividualsinaBGWprocessreproduceaccordingtoarandomvariable χ, with E[χ]=1 and Var(χ) < ∞.Then,thereisaconstantc1 ∈ R suchthat,forallnsufficientlylarge, P( ˜ Sn = 1) ≥ c1n 1/2

Lemma3.4. Supposethat χi arei.i.d.,non-negativeandinteger-valuedrandomvariables,with E[χi ]=1 and Var[χi ] < ∞,andletSn = n i =1 χi .Then,foralln ≥ 1 andm ≥ 0, P(Sn = n m) ≤ c2 √n e c3m2/n , wherec2 > 0 andc3 > 0 arerealconstants.

Lemma3.5. Let χ beadiscreterandomvariabletakingvaluesinnonnegativeintegers.Supposethat E(χ)=1 and 0 < Var(χ) < ∞.Let T bea BGW(χ) tree.Then, P(|T| = n) ≥ n 3/2

Lemma3.6 (Bernsteininequality). LetX1, X2,..., Xn beindependentrandomvariablessuchthatXi E[Xi ] ≤ bforeveryi,whereb ∈ R.LetV := n i =1 Var(Xi ).Then, P n i =1 (Xi EXi ) ≥ t ≤ exp t 2 2V + 2bt 3

, 10 (2025),1–12;DOI:10.2436/20.2002.02.45.

GeneratinguniformspanningtreesfromconditionedBienaym´e–Galton–Watsontrees

3.2.Thewidth

Theorem3.7. Let χ beadiscreterandomvariabletakingvaluesinnonnegativeintegers.Supposethat E[χ]=1 and 0 < Var(χ) < ∞.LetTn bea BGW(χ) treewithnnodes.Then,

forallx ≥ 0 andn ≥ 1,wherec4 > 0 andc5 > 0

Proof. Wewillfollowtheproofstrategyoutlinedin[1].Let Zk bethenumberofindividualsinthe k-th levelofaBGWtree.Itisclearthatevery Zk issome Qj ,where Qj isthenumberofnodesinthequeue ofBFSinthe j-thiteration.Hence,

Asaresult,fortheconditionedBGWtree Tn, P(W ≥ x +1) ≤ P(max j Qj ≥ x +1)= P(max j Sj ≥ x | Sj ≥ 0, j < n, Sn = 1).

Now,weaimtosimplifythelastconditionedprobability.Ourfirstgoalistogetridoftheconditioning on Sj ≥ 0,where j < n.ByRaney’sLemma 3.2,conditioningon {Sj ≥ 0, j < n, Sn = 1} isequivalent toconditioningonlyon {Sn = 1}.However,maxj Sj maybechanged.Byconditioningon Sn = 1,we canwrite max j ≤n Sj =max j ≤n Sj min j ≤n Sj +1, andthelatterquantityischangedbyatmost1byarotationof˜ χi := χi 1 ∀i =1,..., n.Then P(max j ˜ Sj ≥ x | ˜ Sj ≥ 0, j < n, ˜ Sn = 1) ≤ P(max j ≤n ˜ Sj min j ≤n ˜ Sj ≥ x | ˜ Sn = 1).

Therefore,wenowhavemoretoolstobound P(W ≥ x +1).

P(W ≥ 2x +2) ≤ P(max j Qj ≥ 2x +2) ≤ P(max j ≤n ˜ Sj min j ≤n ˜ Sj ≥ 2x +1 | ˜ Sn = 1) ≤ P(max j ≤n Sj ≥ x | Sn = 1)+ P(min j ≤n Sj ≤−x 1 | Sn = 1).

Thelastinequalityisduetothefactthat,if y z ≥ 2x +1,theneither y ≥ x or z ≤−x 1,where x, y , z arerealnumbers.Furthermore,thereflection χi ↔ χn+1 i ,whichswaps Sj ↔−Sn Sn j ,shows thatthelastprobabilitiesarethesame.Hence, P(max j Qj ≥ 2x +2) ≤ 2 P(max j ≤n ˜ Sj ≥ x | ˜ Sn = 1).

Thelastexpressioncanbewrittenintermsofthefirstindexsuchthat ˜ Sj ≥ x.Sothat,letusdefine τ =min{j ≥ 0: ˜ Sj ≥ x}.Then,

P(max j Qj ≥ 2x +2) ≤ 2 P(τ< n | ˜ Sn = 1) =2 P( ˜ Sn = 1 | τ< n) P(τ< n) P( ˜ Sn = 1) , https://revistes.iec.cat/index.php/reports

wherethelastequalityisduetothedefinitionofconditionedprobability.Bydefinitionof τ , Sτ ≥ x.Then, ifwefixsome t < n and y ≥ x,byLemma 3.4 wehavethefollowing:

wherewehaveused Sn St = Sn t duetothefactthatthese n t randomvariablesarei.i.d.Then, by(1)andLemma 3.3 itisclearthatwecannowprovewhatweseek:

Theorem3.8. Let χ beadiscreterandomvariabletakingvaluesinnonnegativeintegers.Supposethat E[χ]=1 and 0 < Var(χ) < ∞.LetTn bea BGW(χ) treewithnnodes.Then,

[W (Tn)]= O(√n),

foralln ≥ 1.

Proof. Thisproofisnotprovidedin[1];however,weconsiderithighlyrelevanttothetopicathand.Therefore,wepresentarigorousproofofthisresultherein.Weaimtoprovethattheexpectedvalue E[W (Tn)] growsasymptoticallyatmostas √n,giventheinequality

Foradiscreterandomvariable W ,theexpectationcanbewrittenas

orequivalently

Thisequivalenceholdsbecause

andreorganizingtermsyieldsthealternativerepresentation.

Usingthegiveninequality

,wecanboundtheexpectationas

GeneratinguniformspanningtreesfromconditionedBienaym´e–Galton–Watsontrees

Thesum ∞ w =1 e c5w 2/n resemblesaRiemannsumandcanbeapproximatedbyanintegralforlarge n Wehave

Tocomputethisintegral,weusethesubstitution

du.Substitutingintotheintegralgives

Therefore,

Thus,thesumcanbeapproximatedas

Substitutingthisbackintotheboundfor

[W (Tn)],wefind

Thisshowsthat

3.3.Theheight

[W (Tn)]= O(√n).

Definition3.9. Theheightofatree T ,denotedby H(T ),isthemaximumdepthofanynodeinthetree. H(T )=max v ∈V (T ) d (v ).

Alexicographicdepth-firstsearch(DFS)isalineartimealgorithmfororderingtheverticesofalabelled graph.Ateachnode,itschildrenarevisitedinlexicographicalorder.Thechildrenofthefirstchildrenare exploredbeforegoingtoitssiblings.Thisideaisappliedrecursivelyforeachvertex.Let Q d i bethenumber ofnodesinthestackoftheDFSatthe i -thstep.Wedefine Q d 0 =1.Ateachstep,wegetridofavertex fromthequeueandadditschildren,whichwereadinthelexicographicalorder.Hence,sinceallindividuals reproducewiththesameprobabilitydistribution,thefollowingrecursionisclear: Q d i = Q d i 1 1+ χi

Thereverse-lexicographicdepth-firstsearchisavariationofthestandarddepth-firstsearch(DFS) where,insteadofvisitingthechildrenofanodeinlexicographicalorder(smallestfirst),wevisitthe childreninreverselexicographicalorder(largestfirst).Wewilldenoteby Q r i thenumberofnodesinthe queueofthealgorithmatthe i -thstep.

https://revistes.iec.cat/index.php/reports

Onecaneasilyobservethatinauniformlyrandomtree Tn,thelabelsoftheverticesdonotaffectthe probabilitydistributionofthelabeledtree.WhileBFSexploresnodeslevelbylevelandDFSgoesdeeper first,lexicographicandreverse-lexicographicDFSaremerelydifferentwaysoforderingthevisitstothe nodes.Therefore,theresultinglabelingsobservedalongtheseproceduresfollowthesamedistribution.

Now,wewillstandoutseveralpreliminarylemmas:

Lemma3.10. LetPbetheuniquepathinatreefromavertexwhichhasheighthtotheroot.The expectedvalueofnodesinPthathavemorethanonechildish q1,whereq1 =1 P(χ =1)

Bythepreviouslemma,theexpectednumberofnodesin P thathaveexactlyonechildis h p1,where p1 = P(χ =1).

3.4.AmodifiedBienaym´e–Galton–Watsontree

Tocompletethestudyoftheheightofarandomtree,wewillneedtointroduceanewconcept.

Letˆ χ bearandomvariablewiththesize-biaseddistribution

(ˆ χ

Itisclearthatthisisaprobabilitydistributionsince m P(˜ χ = m)= E[χ]=1,andthatˆ χ ≥ 1.

Let,for k ≥ 1, T (k) bethemodifiedBGWdefinedasfollows.Ithastwodifferenttypesofnodes: normalandmutant.Normalconstitutetheoffspringof χ,whilemutantnodeshaveoffspringaccordingto independentcopiesofˆ χ.Allchildrenofnormalnodesarealsonormal.Exactlyonechildofeachmutant nodeischosenatrandomanditiscalleditsheir.Ifthisheirhasdepthlessthan k,itismutant.Ifnot, itisnormal.Hence,thereareexactly k mutantnodes,whichtogetherwiththeheir v ∗ ofthelastmutant node,formapathfromtherootto v ∗ atdepth k.Thispathiswhatwecallthespine γ of ˆ T (k)

AnequivalentconstructionofthemodifiedBGWtreecanbedescribedasfollows:

1.Constructthespine:

• Thespineisapathconsistingof k nodes,startingfromtherootatdepth0andendingata nodeatdepth k

• Thesenodescorrespondtothemutantnodesintheoriginalconstruction.

2.Attachindependentsubtreestothenodesofthespine:

• Foreachnodealongthespine,exceptthelastoneatdepth k,attachanumberofindependentsubtrees.Thenumberofsuchsubtreesisdistributedasˆ χ 1,meaningitfollowsthe distributionˆ χ reducedby1.

• Forthenodeofthespineatdepth k,attachindependentsubtreesaccordingtothedistribution χ, whichcorrespondstothebehaviorofnormalnodes.

Hence,thespineformsthecentralpathfromtheroottodepth k,whilethesubtrees,attachedindependently,reflecttheprobabilisticstructureoftheoriginalmodel.

, 10 (2025),1–12;DOI:10.2436/20.2002.02.45.

GeneratinguniformspanningtreesfromconditionedBienaym´e–Galton–Watsontrees

Now,weaimtostudytheprobabilityofobtainingaparticulartree T fromamodifiedBGWtree ˆ T (k) Tostartwithit,itisnotdifficulttofindtheprobabilitythatagivenmutantnodehas m childrenandone ofthemisselectedasheir.Thisis:

Hence,sinceeverynormalnodehasdistribution χ andtheprocessconsistingoftakingamutantnodewith m childrenandselectingoneofthemtobemutantalsodistributesas χ,thefollowingisclear:

Then,theprobabilityofgettingaparticulartree T fromamodifiedBGWwithafixedspineisthe sameofobtainingthis T fromtheBGWtree.Sinceforeverynode v atdepth k,thereexistsoneunique pathfrom v totheroot,thereisabijectionbetweennodesatdepth k andspines.Therefore,summingfor allnodesatdepth k:

Inconclusion, ˆ T (k) hasthedistributionof T biasedby Zk ,whichisthesizeofthe k-thgeneration. Returningtothebroaderdiscussion,withthesepreliminariesinplace,wenowhavethenecessarytools toprovethefollowing.

Theorem3.11. Let χ beadiscreterandomvariabletakingvaluesinnonnegativeintegers.Supposethat E[χ]=1 and 0 < Var(χ) < ∞.LetTn bea BGW(χ) treeconditionalonhavingnnodes.Then,for alln ≥ 1 andh ≥ √n,

whereC2 > 0 andc2 > 0.

Proof. Wewillfollowtheproofapproachprovidedin[1].Let h betheheightof Tn.Wemayassume that h ∈ Z.Wewillbasetheproofonthenextobservation.Sincewehaveprovedthatthewidthis expectedtobe √n,wecansupposethereisavertex v ∈ V (Tn)with“large”height.Hence,therearetwo possiblecasesofobtainingatreewithheight h:eithertherearemanyedgesleavingtheuniquepath P from v totheroot,where v isavertexfromthe h-thlevel,ortherearemanyoftheancestorsof v with justonechild.

Inthefirstcase,theobjectiveistoboundmax(Q d j , Q r k )andconnectthisboundwiththeprobabilistic structureofthetree.Thequantities Q d j and Q r k measurehowmanyverticesaresimultaneouslyactive duringthelexicographicandreverse-lexicographicdepth-firstexplorations,respectively.Wheneveranode onthepathfromavertexatheight h totheroothasmorethanonechild,allofitsextrachildrenare addedtotheexplorationdatastructureandremainthereuntiltheyareprocessed.Thiscauses Q d j or Q r k to increase,andtheirmaximumsizethusreflectsthecumulativeeffectofbranchingalongthepath.Therefore, theheight h ofthetreecanbedirectlyrelatedtomax(Q d j , Q r k ):ifthetreehasheight h,thentheremust existanindex j with Q d j = h oranindex k with Q r k = h

Let p1 = P(χ =1)andlet q1 =1 p1.Let v ∈ V (Tn)suchthat h(v )= h.Let j (resp. k)betheindex of v inlexicographic(resp.reverse-lexicographic)order.Let X bethenumberofnodeswhichhavemore thanonechildin P.Eachancestorof v withmorethanonechildcontributestoatleastoneunitto Q d j or

https://revistes.iec.cat/index.php/reports

to Q r k .Wedistinguishtwocases,eithermax(Q d j , Q r k ) ≥ q1 3 h ormax(Q d j , Q r k ) < q1 3 h.Inthesecondcase,by theaboveremark,thenumberofancestorsof v withexactlyonechildisatleast(1

/3)h

Now,itisnotusefultothinkaboutthequeuesbecausewhenthealgorithmprocessesanodewhichhas justonechild,thesizeofthequeuedoesnotincrease,soitisnotagoodrepresentationfortheheightof thetree.Let S(h)bethesetoftrees T with |T | = n andcontaininganode v suchthat h(v )= h thathas atleast(p1 + q1/3)h ancestorsin P withexactlyonechild.Then,let

Then,wecanapplythesetwocasesdescribedtothecalculusofthefollowingprobability.Thefirsttwo termscorrespondtothefirstcase,whiletheremainingtermcorrespondstothesecondcase:

wherethelastinequalityhasbeenseenintheproofofTheorem 3.7

Then,weonlyneedtobound P(α):

Tojustifythelastinequality,recalltheconstructionof ˆ T (h):thespine γ consistsofthe h mutantnodes fromtheroottodepth h,andforeach i =0,..., h 1thenumberofoffspringofthe i -thmutantis distributedasˆ χi (thesize-biaseddistribution),oneofwhosechildrenisthenchosenuniformlytocontinue thespine.

Thereisaone-to-onecorrespondencebetweentheproperty“the i -thvertexonthespinehasexactly onechildintherealisedtree T ”andtheevent“ˆ χi =1intheconstructionof ˆ T (h)”:ifthe i -thmutant hasexactlyonechild,thennecessarilythatchildistheheir(soˆ χi =1),andconversely,ifˆ χi =1,thenthe i -thspinevertexhasexactlyonechildintherealisedtree.

Consequently,whenever ˆ T (h) equalssome T ∈ S withspine γT ,thenumberofspineverticeshaving exactlyonechildisatleast(p1 + q1/3)h bythedefinitionof S.Thereforetheevent T ∈S { ˆ T (h) = T with γT asspine}

iscontainedintheevent h 1 i =0 1{ ˆ χi =1} ≥ (p1 + q1/3)h , whichyieldsthedisplayedinequality.

Reports@SCM, 10 (2025),1–12;DOI:10.2436/20.2002.02.45.

GeneratinguniformspanningtreesfromconditionedBienaym´e–Galton–Watsontrees

Since 1 ˆ χi =1 areBernoulli(p1),byLemma 3.6,andtaking t = q1h/3, V = p1q1h, b = q1,wehavethe followingbound:

Furthermore,thefollowingequalityisclear:

ByLemma 3.5, P(|T| = n) ≥ n 3/2 andwecanestablishanupperboundfor P(δ).

(δ)= P(T∈ S)

forall h ≥ √n.Takingeverythingintoaccount,whatwehaveisthefollowingboundfortheheightofa BGWtree:

Now,taking

Havingprovedthat,weleftthefollowingresultwithoutproofduetoitsanalogywithTheorem 3.8 Theorem3.12. Let χ beadiscreterandomvariabletakingvaluesinnonnegativeintegers.LetTn bea BGW(χ) treeconditionalonhavingnnodes.Supposethat E[χ]=1 and 0 < Var(χ) < ∞.Then,for alln ≥ 1, E[H(Tn)]= O(√n).

References

[1]L.Addario-Berry,L.Devroye,S.Janson,SubGaussiantailboundsforthewidthandheightof conditionedGalton–Watsontrees, Ann.Probab. 41(2) (2013),1072–1087.

[2]D.Aldous,Thecontinuumrandomtree.I, Ann. Probab. 19(1) (1991),1–28.

[3]V.F.Kolchin, RandomMappings,TranslatedfromtheRussian,Withaforewordby S.R.S.Varadhan,Transl.Ser.Math.Engrg., OptimizationSoftware,Inc.,PublicationsDivision,NewYork,1986. https://revistes.iec.cat/index.php/reports

ANELECTRONICJOURNALOFTHE SOCIETATCATALANADEMATEM ` ATIQUES

Onthebasinsofattraction ofroot-findingalgorithms

DavidRosadoRodr´ıguez UniversitatdeBarcelona rosadodav4@gmail.com

Resum (CAT)

Elsalgoritmesdecercad’arrelss’hanutilitzatperresoldrenum`ericamentequacions nolinealsdelaforma f (x )=0.Aquestarticleestudialadin`amicadelafam´ılia parametritzadadeTraub Tp,δ aplicadaapolinomis,queabastadesdelm`etodede Newton(δ =0)finsaldeTraub(δ =1).Enscentremenpropietatstopol`ogiques delesconquesimmediatesd’atracci´odelspuntsfixosfinits,especialmentlaseva connectivitatielfetdeseracotadaono.Aquestsfetss´onclauperidentificar condicionsinicialsuniversalsqueassegurinlaconverg`enciaatoteslesarrelsde p

Abstract (ENG)

Root-findingalgorithmshavehistoricallybeenusedtonumericallysolvenonlinear equationsoftheform f (x )=0.Thispaperstudiesthedynamicsoftheparameterized Traubfamily Tp,δ appliedtopolynomials,rangingfromNewton’smethod(δ =0) toTraub’smethod(δ =1).Wefocusontopologicalpropertiesoftheimmediate basinsofattractionofthefinitefixedpoints,especiallysimpleconnectivityand unboundedness,whicharekeytoidentifyingauniversalsetofinitialconditions ensuringconvergencetoallrootsof p

Keywords: dynamicalsystems,root-findingalgorithms,Newton’smethod, Traub’smethod.

MSC(2020): Primary30D05,37F10,37F46.

Received: April27,2025. Accepted: August8,2025.

13 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),13–21;DOI:10.2436/20.2002.02.46.

1.Introduction

Solvingnon-linearequationsoftheform f (x )=0isacommonchallengeinvariousscientificfields, spanningfrombiologytoengineering.Whenalgebraicmanipulationisnotfeasible,iterativemethods becomenecessarytodeterminesolutions.Amongthese,Newton’smethodstandsoutasawidelyused approach,relyingonthelinearizationof f (x ).Itsiterativeschemeisgivenby:

n+1 = xn

(xn)

′(xn) , n ≥ 0.

Nevertheless,numerousnumericalmethodshaveproventobeefficientwhentheyconverge,includingthe oneconsideredhere,Traub’smethod.WhileNewton’smethodexhibitsquadraticconvergenceforsimple rootsofapolynomialwhentheinitialguessissufficientlyclosetotheroot,Traub’smethodachievescubic (local)convergence.Thismethodbelongstoaparametricfamilyofiterativeschemes,firstintroduced in[6, 12],knownasthe dampedTraub’sfamily.Itsiterativeformulaisgivenby:

0, where yn = xn f (xn ) f ′ (xn ) representsaNewtonstep,and δ isthedampingparameter.Notably,setting δ =0 recoversNewton’smethod,while δ =1correspondstoTraub’smethod.Itisimportanttomentionthateach iterationofTraub’smethodinvolvesadditionalcomputationscomparedtoNewton’smethod.Althoughthe questioncanbeexploredinothersettings,herewewillfocusonthecasewhere p(z )=0, z ∈ C

(xn) ,

Roughlyspeaking,whenwehaveagoodestimateofthesolutionstotheequation p(z )=0,iterative methodstendtoworkwell.However,challengesarisewhenthenumberofsolutionsof f islargeorwhenwe lackcontroloverthesesolutions.Thisisparticularlyproblematicwhenselectinginitialconditionstoinitiate thealgorithm.Insuchsituations,thestudyofdynamicalsystemsbecomesvaluable.Byexaminingthe topologicalcharacteristicsoftheimmediatebasinsofattractionassociatedwiththesolutionsof p(z )=0, wecangainvaluableinsightsandaidinaddressingthesechallenges.Anillustrationofthisisprovidedin[8]. Intheirwork,theauthorsusedsometopologicalresultsofthebasinsofattractiontoconstructauniversal andexplicitsetofinitialconditionsdenotedas Sd .Thisset,dependingonlyonthepolynomial’sdegree, allowsNewton’smethodtofindallrootsofapolynomial.Theexistenceoftheset Sd isguaranteedbythe followingkeypropertiesoftheimmediatebasinsofattractionsfortheNewton’smethod(firstprovedby Przytycki[10]andlatergeneralizedbyShishikura[11]).

Theorem1.1. Letpbeapolynomialofdegreed ≥ 2.Assumethatp(α)=0 andletNp bethe correspondingNewton’smap.Then,theimmediatebasinofattractionof α,denotedas A∗(α),isasimply connected,unboundedset.

Anaturalquestionthatcomesupnowiswhetherwecancreateasetsimilarto Sd forTraub’smethod. Ifthiswerepossible,itwouldprovideawaytofindalltherootsofapolynomialwithimprovedconvergence speed.Specifically,aspreviouslynoted,forsimplerootsofthepolynomial,thelocalconvergenceorder wouldbecubicinsteadofquadratic,leadingtofasterconvergence.Toachievethis,provinganequivalent toTheorem 1.1 forTraub’smethod,willprovidethenecessarytoolsforbuildingthe Sd like-set.Inarecent study[3],Theorem 1.1 wasprovedforTraub’smethodundercertainadditionalassumptions.Tobeprecise, theresearcherssuccessfullyestablishedthefollowingtheorem:

https://revistes.iec.cat/index.php/reports

Theorem1.2. Letpbeapolynomialofdegreed ≥ 2.Assumethatpsatisfiesoneofthefollowing conditions:

(i)d =2,or

(ii)itcanbewrittenintheformpn,β (z ):= z n β forsomen ≥ 3 and β ∈ C

Supposethatp(α)=0 andconsiderdampedTraub’smapTp,δ (z ):=

with

[0,1] Then A∗ δ (α) isasimplyconnectedunboundedset.

ThisarticleexploresthedampedTraub’smethodasaroot-findingalgorithm.Weprovidebackground tounderstandtheproofofTheorem 1.2 andpresenttworesultsthatbringusclosertoourgoal:proving anequivalentresulttoTheorem 1.1 forTraub’smethod.Specifically,weestablishthefollowingresult:

TheoremA. Letpbeapolynomialofdegreed ≥ 2.Assumethatp(α)=0 andletTp,δ bethe correspondingdampedTraub’smap.Then,for δ closeenoughtozero, A∗ Tp,δ (α) isanunboundedset.

Moreover,weanalyzethebehaviorofTraub’smethodforthepolynomialfamily pd (z )= z (z d 1). Notably,forHalley’sroot-findingalgorithm,thebasinofattractionof z =0isboundedwhen d =5,but weestablishthefollowingresult:

TheoremB. Letpd (z )= z (z d 1).Then, A∗ Tpd ,1 (0) isanunboundedsetforeveryintegerd > 0

2.Anintroductiontoholomorphicdynamics

Letusdenote ˆ C = C ∪{∞} the extendedcomplexplane or RiemannSphere

Definition2.1. Let R : ˆ C → ˆ C bearationalmap.Apoint z = z0 isa fixedpoint if R(z0)= z0 (resp. periodicofperiodp if R p (z0)= z0 forsome p ≥ 1and R n(z0) = z0 forall n < p).The multiplier of z0 is λ = R ′(z0)(resp. λ =(R p )′(z0)).

Thecharacterofthefixedorperiodicpointscanbedeterminedbyusingthemultiplier.Infact,the fixedorperiodicpoint z0 is attracting if |λ| < 1(superattracting if λ =0), repelling if |λ| > 1and indifferent if |λ| =1.

Definition2.2. Let R : ˆ C → ˆ C bearationalmapand z0 ∈ ˆ C beanattractingfixedpointof R.Wedefine the basinofattraction of z0 as

R (z0):= A(z0)= {z

C : R n(z ) n→∞ −−−→ z0}

Wedenoteby A∗(z0)theconnectedcomponentof A(z0)containing z0,andwerefertoitasthe immediate basinofattraction

Inwhatfollowsweomitthedependencewithrespecttotherationalmapunderconsideration,unless itismandatory.Itiseasytoseethat A(z0)isanopensetcontaining z0.Thereisavastbodyofresultson thistopic,andforageneraloverview,manyexcellentreferencesareavailable;see,forinstance,[2, 9, 5]. Toconcludethischapter,wepresentatheoremthatwillbeusefullater.Thistheoremstatethat,ina neighbourhoodofanattractingfixedpoint,themap lookslikeg (ζ)= λζ

, 10 (2025),13–21;DOI:10.2436/20.2002.02.46.

Onthebasinsofattractionofroot-findingalgorithms

Theorem2.3 (KoenigslinearizationTheorem). Letz0 ∈ C,Uneighbourhoodofz0 andf : U → C be aholomorphicfunctionsuchthatz0 isanattractingfixedpointwithmultiplier 0 < |λ| < 1.Thenthereis aconformalmap ζ = ϕ(z ) ofaneighbourhoodofz0 ontoaneighbourhoodofz0 whichconjugatesfto thelinearfunctiong (ζ)= λζ.Theconjugatingfunctionisunique,uptomultiplicationbyanonzeroscale factor.

3.Localdynamicsofthefamily T T

Recallthatif p isapolynomialofdegree d ≥ 2,thedampedTraub’smapappliedto p isdefinedas

where Np istheNewton’smapand δ ∈ C.Forourpurposes,itwillsufficetoconsider δ ∈ [0,1].Notice that,setting δ =0recoversthewell-knownNewton’smethod.Themap Np isthe universal root-finding algorithmanditsatisfiesthefollowingkeyglobaldynamicalproperties:

Proposition3.1. Letpbeapolynomialofdegreed ≥ 2.ThefollowingpropertiesregardingtheNewton’s maphold:

(i)Apointz = α isarootofpifandonlyifitisafixedpointofNp

(ii)ThesimplerootsofparesuperattractingfixedpointsofNp ,whilemultiplerootsareattractingfixed pointsofNp

(iii)Thepointz = ∞ istheonlyrepellingfixedpointofNp

Proof. (i) and (ii) arestraightforwardcomputations.For (iii),observethat Np (∞)=limz→∞ Np (z )= ∞, so z = ∞ isafixedpoint.Toseeitsnature,considerthetransformation ϕ : U → V suchthat ϕ(z )=1/z , where U isaneighbourhoodof z = ∞ and V isaneighbourhoodof z =0.Theconjugatemapis then Np (z )= ϕ(Np (ϕ 1(z )))= 1 Np (1/z ) ,sostudyingthebehaviourof Np at z =0revealsthecharacter of z = ∞ intheoriginalsystem.

LeveragingthepropertiesofNewton’smethod,particularlynotingthat z = ∞ istheonlyrepelling fixedpoint,Theorem 1.1 canbeestablished.Detailsoftheproofcanbefoundin[1].For δ =0,someof thepropertiesthatholdfor Np remainthesame,whileotherschangeslightly.Letussummarizethem:

Proposition3.2. Letpbeapolynomialofdegreed ≥ 2 and δ ∈ (0,1].Thefollowingpropertiesregarding thedampedTraub’smaphold:

(i)Ifz = α isarootofp,thenz = α isafixedpointofTp,δ .Theconverseisnotnecessarilytrue.

(ii)ThesimplerootsofparesuperattractingfixedpointsofTp,δ ,whilemultiplerootsareattracting fixedpointsofTp,δ

(iii)Thepointz = ∞ isarepellingfixedpointofTp,δ https://revistes.iec.cat/index.php/reports

Thecompleteanddetailedproofcanbefoundin[3].Noticethat,finitefixedpointsofthemethoddo notnecessarilycorrespondtozerosofthepolynomial,asisthecasewithNewton’smethod;seeFigure 1 foravisualillustration.Withallthisinformation,arecentstudy[3]successfullyestablishedTheorem 1.2

Figure1:Ontheleft,weillustratethedynamicalplaneofTraub’smethodappliedtothecubicpolynomial P(z )=(z 2 +0.25)(z 0.439).Basinsofattractioncorrespondingtorootsofthepolynomialare showninorange.Itisnotablethat Tp,1 exhibitsanattractingfixedpointlocatedat ζ ≈ 0.155,whosebasin isdepictedinblue,thatdoesnotcorrespondtoanyrootof P.Ontheright,wepresentthedynamicalplane ofNewton’smethodappliedto P.Here,itisevidentthattherearenofixedpointsotherthantheroots.

4.Themethodasasingularperturbation

Inthischapter,weprovetheunboundednessofimmediatebasinsofattractionwhen δ ≈ 0.Forsmall δ, thedampedTraub’smethodactsasasingularperturbationofNewton’smethod.A singularperturbation referstoabasefamily(withwell-understooddynamics)combinedwithalocalperturbationthatincreases themap’sdegreeandenrichesitsdynamics.Thisperturbationaffectsonlycertainregionsofthedynamical planewhentheparameterissmall[7].Here,Newton’smethodisthebasefamily,and p(Np (z ))/p′(z )is theperturbation.NoticethatthesingularperturbationoccursovertheJuliaset,asitinvolvesadding additionalpreimagesof z = ∞ tothezerosof p′(z ).Toestablishthemainresultofthesection,wewill firstpresentsomeauxiliaryresults.

Lemma4.1. Letp(z )= ad z d + + a1z + a0 beapolynomialofdegreed ≥ 2.Letqj bethezeros ofp′(z )=0,i.e.,thepolesofboththedampedTraub’smap,Tp,δ ,andNewton’smap,Np .Consider thecompactK = D(0, R) \∪j D(qj , ε) whereR > 0 and ε> 0 arepositivefixedconstants.Then,for everyz ∈ K,thereexistsaconstantCR,ε suchthat |p(Np (z ))/p′(z )|≤ C

Proof. Let z ∈ K .Thereexistsapositivevalue ηϵ > 0suchthat |p′(z )| >ηϵ .Moreover,since |z | < R, |p(z )|≤|ad |R d + ··· + |a1|R + |a0| := R ′.Hence,

Therefore,

Onthebasinsofattractionofroot-findingalgorithms

Lemma4.2. Letpbeapolynomialofdegreed ≥ 2.Letqj bethezerosofp′(z )=0,i.e.,thepoles ofboththedampedTraub’smap,Tp,δ ,andNewton’smap,Np ,andletz = α beazeroofp,i.e.,an attractingfixedpointforbothNp andTp,δ .ConsiderthecompactK = D(0, R) \∪j D(qj , ε′) whereR > 0 and ε′ > 0 arepositivefixedconstantssuchthat α ∈ K.Then,thefollowingstatementshold:

(i)ThereexistsacompactK ′ ⊂ KsuchthatK ′ ⊂A∗ Np (

), α ∈ K ′ and ∂K ′ ∩

K = ∅,satisfyingthat foreveryz ∈ K ′,thereisauniqueM ∈ N suchthat: ∀ε> 0,N M p (z ) ∈ D(α, ε/2)

(ii)Forthegiven ε> 0 andfor δ smallenough,thefollowingpropertyholds: ∀z ∈ K ′ , |N M p (z ) T M p,δ (z )| <ε/2.Inparticular,T M p,δ (z ) ∈ D(α, ε).

Proof. (i) Theexistenceofsuchacompactisguaranteedbythefactthat A∗ Np (α)isanopenset,unbounded andsimplyconnected.Since z = α isanattractingfixedpointfor Np ,theexistenceof M ∈ N isalso guaranteed.

(ii) Toprovetheresult,letusfirstestablishthefollowingclaim:For δ smallenough, ∀r > 0, ∃ρ> 0suchthatif |z1 z2| <ρ =

Toprovetheclaim,observethat,usingLemma 4.1 inthelastinequality, |Np (z1) Tp,δ (z2)|≤|Np (z1) Np (z2)| + δ p(Np (z )) p′(z ) ≤|Np (z1) Np (z2)| + δCR,ε′

Hence,sinceNewton’smapiscontinuousin K (inparticularitisalsoin K ′),thereexists ρ> 0suchthat if |z1 z2| <ρ,then |Np (z1) Np (z2)| < r /2.Setting δ = r 2CR,ε′ ,weobtainthedesiredbound.

Now,let z ∈ K .Toprovetheresult,weproceedasfollows:

1.Usingtheclaimwith z1 := N M 1 p (z )and z2 := T M 1 p,δ (z ),thereexists ηM > 0and δM > 0suchthat if |N M 1 p (z ) T M 1 p,δ (z )| <ηM ,then |N M p (z ) T M p,δM (z )| <ε/2.

2.Iteratingthealgorithm,weobtainsequences {ηM i }M 3 i =0 , {δM i }M 3 i =0 satisfyingthatif |N M i 1 p (z ) T M i 1 p,δ (z )| <ηM i ,then |N M i p (z ) T M i p,δi (z )| <ηM i +1

3.Weconcludethealgorithmwiththeexistenceof η2 > 0and δ2 > 0suchthatif |Np (z ) Tp,δ2 (z )| < η2,then |N 2 p (z ) T 2 p,δ2 (z )| <η3

Finally,toensurethat |Np (z ) Tp,δ (z )| <η2,wejustneedtotake δ1 = η2 CR,ε′ .Therefore,taking δ = min{δ1,..., δM },weobtainthatforevery z ∈ K : |T M p,δ (z ) N M p (z )| <ε/2.

TheoremA. Letpbeapolynomialofdegreed ≥ 2.Assumethatp(α)=0 andletTp,δ bethe correspondingdampedTraub’smap.Then,for δ closeenoughtozero, A∗ Tp,δ (α) isanunboundedset.

Proof. First,observethatfor δ closeenoughtozero(indeedforevery δ ∈ [0,1]), z = ∞ isarepelling fixedpointfor Tp,δ .ByKoenigslinearizationTheorem,inaneighborhoodof z = ∞,say D(∞, ε), Tp,δ is locallyconjugatedto g (ζ)= λζ,where λ isthemultiplierof z = ∞.Noticethat,if λ ∈ C,since |λ| > 1,

https://revistes.iec.cat/index.php/reports

pointsnear z = ∞ tendtomoveawayinaspiralshape,andif λ ∈ R,since |λ| > 1,pointsnear z = ∞ tendtomoveoutwardinaradialmanner.

Letusdefine R := 1 ε andconsiderthecompact K := D(0, R) \∪j D(qj , ε′)where qj arethepoles of Tp,δ ,i.e.,thezerosof p′(z )=0,and ε′ > 0isapositivefixedconstant.Wecanassumethat α ∈ K . Ifnot,wecanchooseasmallervaluefor ε (increasingthevalueof R)toensurethat α ∈ K ,makingthe neighborhoodwheretheKoenigscoordinatesapplysmaller.Since z = α isanattractingfixedpointfor both Np and Tp,δ ,thereexists η1, η2 > 0suchthat D(α, η1) ⊂A∗ Np (α)and D(α, η2) ⊂A∗ Tp δ (α).Setting η =min{η1, η2},wehavethat D(α, η) ⊂A∗ Np (α) ∩A∗ Tp δ (α).AccordingtoLemma 4.2(i),thereexistsa compact K ′ ⊂ K suchthat K ′ ⊂A∗ NP (α), α ∈ K ′ and ∂K ′ ∩ ∂K = ∅,satisfyingthatforevery z ∈ K ′ , thereisaunique M ∈ N suchthat,forevery z ∈ K ′ , N M p (z ) ∈ D(α, η/2) ⊂ D(α, η).Moreover,since thebasinsofattractionofNewton’smethodareunboundedandsimplyconnected,thereexistsaray τ connectingthefixedpoint z = α and z = ∞,includedin A∗ Np (α).Thisraycanbechosensuchthat itsrestrictionto K isincludedin K ′.Fromnowown,anyreferenceto τ willindicatetherayextending fromthepoint z = α totheboundaryoftheset K .Then,accordingtoLemma 4.2(ii),for δ small enoughand z ∈ K ′ , T M p,δ (z ) ∈ D(α, η),indicatingthat z ∈A∗ Tp δ (α).Then,either τ ⊂ K ′ ⊂A∗ Tp δ (α) orthereexists w ∈J (Tp,δ ) ∩ K ′.Inthelastcase,since w ∈ K ′ , T M p,δ (w ) ∈ D(α, η),incontradiction with w ∈J (Tp,δ ).Therefore, τ ⊂ K ′ ⊂A∗ Tp,δ (α).

Byconstruction,observethat ∂D(0, R)= ∂D(∞, ε),hence,theray τ ,whichendsat ∂D(0, R), connectswiththespiral(orthelineincase λ ∈ R)thatextendstowards z = ∞.Thus,wefoundaray thatconnectsthefixedpoint z = α to z = ∞,whichiscontainedwithin A∗ Tp δ (α).Thisprovesthatthe immediatebasinofattractionforthedampedTraub’smethodisunboundedwhen δ ≈ 0.

5.Traub’smethodappliedto

z z z(z d 1) (z 1) (z d 1)

Now,weaimtoexamineTraub’smethodappliedtothefamily pd (z )= z (z d 1).Thisfamilyisparticularly interestingbecause,forHalley’sroot-findingalgorithm,itwasfoundthatfor d =5,theimmediatebasin ofattractionof z =0isbounded.Therefore,provingthatthisisnotthecaseforTraub’smethodwould supporttheconjecturethattheimmediatebasinsofattractionofTraub’smethodareunbounded.Wehave beenabletoprovethattheimmediatebasinofattractionof z =0isunboundedforevery d .Toestablish theresult,wewillfirstpresentanauxiliaryresult.

Lemma5.1. Thesemi-linesz = re (2k +1)πi d ,r > 0 andk =0,1,..., d 1,areforwardinvariantunderTpd ,1

Firstofall,observethat

Hence,since e (2k +1)πi = 1,astraightforwardcomputationrevealsthat T

), where Rd (r ):= d (d +1)r 2d +1[(d +1)r d +1]d d d +1r (d +1)2 [(d +1)r d +1]d +2 .

, 10 (2025),13–21;DOI:10.2436/20.2002.02.46.

Onthebasinsofattractionofroot-findingalgorithms

TheoremB. Letpd (z )= z (z d 1).Then, A∗ Tpd ,1 (0) isanunboundedsetforeveryintegerd > 0

Proof. Consideronlythesemi-linesthatdonotcrossthe d throotsofunity,i.e., z = re (2k +1)πi d , r > 0and k =0,1,..., d 1.ByLemma 5.1,thesemi-linesareforwardinvariantunder Tpd ,1.Infact,wehavethat Tpd ,1(re (2k +1)πi d )= e (2k +1)πi d Rd (r ),where Rd (r ):= d (

Then,ifwecanprovethatforevery r > 0wehave0 < Rd (r ) < r ,wecanconcludethat A∗ Tpd ,1 (0)isan unboundedsetforevery d .Inthatcase,wecanalsostatethat A∗ Tpd ,1 (0)hasatleast d accessestoinfinity. Sincethedenominatorof Rd isalwayspositiveforevery r > 0,theinequality0 < Rd (r )isequivalentto

(1)

ExpandingthelastexpressionusingtheBinomialexpansion,weobtainthatinequality(1)becomes

Noticethat,since(d +1)d +1 d d > 0foreverypositiveinteger d ,weobtainthattheinequalityholdsfor every r > 0.

Theinequality Rd (r ) < r canbewrittenas Sd (r ) < 0,where Sd isdefinedas

UsingtheBinomialexpansion,wecanrewritethelastexpression:

Now,arrangingterms,

Observethat

Hence,allthecoefficientsofthepolynomial Sd arenegative.Therefore,wecanconcludethatfor r > 0, Sd (r ) < r ,whichcompletestheproof.

https://revistes.iec.cat/index.php/reports

Itstillneedstobeproventhattheimmediatebasinsofattractionofthe d throotsofunityare unbounded.Thisisamorechallengingpartoftheproof,asattemptingtoapplythesameargumentsused earlierleadstodifficultiesinestablishingboundsfortheexpressions.However,arecentstudyconfirmsthat thisholdsforallintegers d ≥ 2.Infact,thecaseofTraub’smethodappliedtothefamily p(z )= z (z d 1) hasalreadybeenfullyresolved[4].

6.Conclusions

Withthispaper,wearecontributingtowardsdemonstratingthattheimmediatebasinsofattractionof thedampedTraub’smethodareunboundedandsimplyconnectedsets.Wehavebeenabletoprovewith completegeneralitytheunboundednessofthemethodwhen δ ≈ 0andweanalyzeaparticularcase,the family pd (z )= z (z d 1).Ourfindingsindicatethatanalyzingthetopologicalpropertiesofthismethodis notastraightforwardandthatacomprehensiveproofrequiresdifferentapproachesfromthoseusedin[3].

References

[1]K.Bara´nski,N.Fagella,X.Jarque,B.Karpi´nska,ConnectivityofJuliasetsofNewton maps:aunifiedapproach, Rev.Mat.Iberoam. 34(3) (2018),1211–1228.

[2]A.F.Beardon, IterationofRationalFunctions. ComplexAnalyticDynamicalSystems,Grad. TextsinMath. 132,Springer-Verlag,NewYork, 1991.

[3]J.Canela,V.Evdoridou,A.Garijo,X.Jarque, Onthebasinsofattractionofaone-dimensional familyofrootfindingalgorithms:fromNewtontoTraub, Math.Z. 303(3) (2023),Paper no.55,22pp.

[4]J.Canela,A.Garijo,X.Jarque,Boundednessandsimpleconnectivityofthebasins ofattractionforsomenumericalmethods, Preprint(2025). https://arxiv.org/abs/ 2507.22704.

[5]L.Carleson,T.W.Gamelin, ComplexDynamics,UniversitextTractsMath.,Springer-Verlag, NewYork,1993.

[6]A.Cordero,A.Ferrero,J.R.Torregrosa, DampedTraub’smethod:convergenceandstability, Math.Comput.Simulation 119 (2016), 57–68.

[7]R.L.Devaney,Singularperturbationsofcomplexpolynomials, Bull.Amer.Math.Soc. (N.S.) 50(3) (2013),391–429.

[8]J.Hubbard,D.Schleicher,S.Sutherland,How tofindallrootsofcomplexpolynomialsbyNewton’smethod, Invent.Math. 146(1) (2001),1–33.

[9]J.Milnor, DynamicsinOneComplexVariable, Thirdedition,Ann.ofMath.Stud. 160,PrincetonUniversityPress,Princeton,NJ,2006.

[10]F.Przytycki,Remarksonthesimpleconnectednessofbasinsofsinksforiterationsofrationalmaps,in: DynamicalSystemsandErgodic Theory,BanachCenterPubl. 23,PWN—Polish ScientificPublishers,Warsaw,1989,pp.229–235.

[11]M.Shishikura,TheconnectivityoftheJuliaset andfixedpoints,in: ComplexDynamics,AK Peters,Ltd.,Wellesley,MA,2009,pp.257–276.

[12]J.E.V´azquez-Lozano,A.Cordero,J.R.Torregrosa,DynamicalanalysisoncubicpolynomialsofdampedTraub’smethodforapproximatingmultipleroots, Appl.Math.Comput. 328 (2018),82–99.

21 Reports@SCM, 10 (2025),13–21;DOI:10.2436/20.2002.02.46.

ANELECTRONICJOURNALOFTHE SOCIETATCATALANADEMATEM ` ATIQUES

Convergenceofgeneralized MITbagmodels

JoaquimDuraniLamiel CentredeRecercaMatem`atica jduran@crm.cat

Resum (CAT)

Estudiempropietatsespectralsdelsmodelsdebossadel’MITgeneralitzats. Aquestss´onoperadorsdeDirac {Hτ }τ ∈R actuantendominisde R3 ambcondicionsdefronteraquegenerenconfinament.Estudiantlaconverg`enciaenelsentit delaresolventdelsoperadors Hτ capalsoperadorsl´ımit H±∞ quan τ →±∞, provemquecertespropietatsespectralss’heredenalllargdelaparametritzaci´o. Aquestsresultats,obtingutsparcialmentaltreballdefidem`aster[3],s´onnousi s’hanpublicata[4].

Abstract (ENG)

WestudyspectralpropertiesofgeneralizedMITbagmodels.TheseareDiracoperators {Hτ }τ ∈R actingondomainsof R3 withconfiningboundaryconditions.By studyingtheresolventconvergenceoftheoperators Hτ towardsthelimitingoperators H±∞ as τ →±∞,weprovethatcertainspectralpropertiesareinherited throughouttheparametrization.Theseresults,partiallyobtainedinthemaster’s thesis[3],arenewandhavebeenpublishedin[4].

Keywords: Diracoperator,spectraltheory,MITbagmodel,shapeoptimization,resolventconvergence. MSC(2020): Primary35P05,35Q40.Secondary47A10,81Q10.

Received: June18,2025.

Accepted: August4,2025.

https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),23–29;DOI:10.2436/20.2002.02.47.

23

ConvergenceofgeneralizedMITbagmodels

1.Introduction

Theequationthatgovernsallrelativisticquantumprocessesiscalled Diracequation.In R3,itisasystem offourcomplexvaluedlinearPDEsoffirstorderintimeandspacevariables.Foraspin-1/2freeparticle ofmass m ≥ 0,onecanwritetheDiracequationinmatricialformas

where α =(α1, α2, α3)and β aretheso-called Diracmatrices, αj := 0

givenbythe Paulimatrices

andwhere

istheso-called wavefunction oftheparticle.Here, ∇ =(∂1, ∂2, ∂3)denotesthegradientin R3,andas customaryweusethenotation α ·∇ = α1

.InCartesiancoordinates,thedifferential operatorintheright-handsideof(1)writesas

Noticethatifonediagonalizesthisoperator(takingintoaccountboundaryconditions),onecansolvethe time-dependentDiracequation(1)usingthemethodofseparationofvariables.Hence,thetime-dependent problemreducestoastationaryeigenvalueproblemoftheform ( i α ·∇ + mβ)φ = λφ inΩ, boundaryconditionsfor φ on ∂Ω, whereΩ ⊆ R3 isthedomainwheretheparticleevolves, φ :Ω → C4,andtheboundaryconditionstypically dependonphysicalconstraints.Theeigenvalues λ providerelevantinformationtounderstandtheevolution ofthesystem,hencethismotivatestheirstudyandunderstanding.Thisiswhatwedointhiswork,for someprescribedboundaryconditions.

https://revistes.iec.cat/index.php/reports

2.GeneralizedMITbagmodels

DiracoperatorsactingondomainsΩ ⊂ R3 with C 2 boundaryareusedinrelativisticquantummechanics todescribeparticlesthatareconfinedinabox.Theso-called MITbagmodel isaveryremarkableexample, whichwasintroducedinthe1970sasasimplifiedmodeltostudyconfinementofquarksinhadrons(like quarksupanddowninsideaproton).Itistheoperator H0 definedby

Dom(H0):= {φ ∈ H 1(Ω) ⊗ C4 : φ = i β(α ν)φ on ∂Ω}, H0φ :=( i α ·∇ + mβ)φ forall φ ∈ Dom(H0).

Here, ν denotestheoutwardunitnormalvectoron ∂Ω,and H 1(Ω)isthestandardSobolevspaceoffirst weakderivativesin L2(Ω),namely H 1(Ω):= {f ∈ L2(Ω): ∥f ∥H 1 (Ω) < ∞},where ∥f ∥H 1 (Ω) :=(

Forthesakeofnotation,inthesequelweshalldenote H 1(Ω) ⊗ C4 as H 1(Ω)4,andsimilarly L2(Ω) ⊗ C4 as L2(Ω)4 .

Motivatedbysomephysicalconsiderations,in[1]itwasstudiedthefamilyofDiracoperatorswith confiningboundaryconditionsdefinedfor τ ∈ R by Dom(Hτ ):= {φ ∈ H 1(Ω)4 : φ = i (sinh τ cosh τβ)(α ν)φ on ∂Ω}, H

NoticethattheMITbagmodelcorrespondsto τ =0—thiswasthemainreasonin[1]tocallthe operators Hτ in(2) generalizedMITbagmodels.For τ ∈ R,theoperator Hτ isself-adjointin L2(Ω)4 by[2,Proposition5.15].Moreover,from[1,Lemma1.2]weknowthatitsspectrum σ(Hτ )iscontained in R \ [ m, m]andispurelydiscrete.Inparticular,theessentialspectrum σess(Hτ )isemptyforall τ ∈ R Furthermore, λ ∈ σ(Hτ )ifandonlyif λ ∈ σ(H τ ).Thankstothisoddsymmetry,onecanreducethe studyofthespectralpropertiesofthegeneralizedMITbagmodelstothestudyof σ(Hτ ) ∩ (m,+∞) for τ ∈ R

Aspectralstudyofthemapping τ →Hτ wascarriedoutin[1],wherethefollowingresultwas shown.Initsstatement, ∆D denotestheself-adjointrealizationoftheDirichletLaplacianin L2(Ω),and σ( ∆D )denotesitsspectrum.

Theorem2.1 ([1,Theorem1.4]). Theeigenvaluesof Hτ canbeparametrizedbyincreasingrealanalytic functionsof τ .Moreover,if τ → λ(τ ) ∈ σ(Hτ ) ∩ (m,+∞) isacontinuousfunctiondefinedonan intervalI ⊂ R,thenthefollowingholds:

(i)IfI =(−∞, τ0) forsome τ0 ∈ R,then λ(−∞):=limτ ↓−∞ λ(τ ) existsandbelongsto [m,+∞).In addition, λ(−∞)= mif λ(τ ) ≤ min σ( ∆D )+ m2 forsome τ ∈ I, λD + m2 forsome λD ∈ σ( ∆D ) otherwise.

(ii)IfI =(τ0,+∞) forsome τ0 ∈ R,then λ(+∞):=limτ ↑+∞ λ(τ ) existsasanelementofthe set (m,+∞].Inaddition,if λ(+∞) < +∞,then λ(+∞)= λD + m2 forsome λD ∈ σ( ∆D ). 25

Reports@SCM, 10 (2025),23–29;DOI:10.2436/20.2002.02.47.

ConvergenceofgeneralizedMITbagmodels

ThisresultestablishesaclearconnectionbetweenthespectrumoftheDiracoperator Hτ as τ →±∞ andthespectrumoftheDirichletLaplacian ∆D .In[1,Remark4.4]itwasleftasanopenquestionto investigatewhichshouldbethelimitingoperatorsof Hτ as τ →±∞,andinwhichsensetheconvergence holdstrue.Theanswerwasdevelopedinthemaster’sthesis[3]andthenpublishedin[4].Inthepresent work,wereviewtheresultsobtained.

3.Convergenceas τ τ τ movesin R

Inordertoguesswhothelimitingoperatorsmightbe,wefirstmakeanobservation.Writing φ ∈ Dom(Hτ ) incomponents1 as φ =(u, v )⊺ ,theboundarycondition

= i (sinh τ cosh

rewritesas u = ie τ (σ · ν)v .Formally,thisequationforces u and v tovanishon ∂Ωinthelimits τ ↑ +∞ and τ ↓−∞,respectively.Thisleadstoconsidertheso-called Diracoperatorswithzigzagtypeboundary conditions studiedin[6],whicharedefinedby

Dom(H+∞):= {φ =(u, v )⊺ : u ∈ H 1 0 (Ω)2 , v ∈ L2(Ω)

—here H 1 0 (Ω)2 isthesubspaceoffunctionsin H 1(Ω)2 withzerotrace—,and

Dom(H−∞):= {φ =(u, v )⊺ : u ∈ L2(Ω)

From[6,Theorem1.1andLemma3.2]weknowthat H±∞ areself-adjointin L2(Ω)4 andthattheirspectra arecharacterizedbythespectrumoftheDirichletLaplacian.Morespecifically,

and ∓m ∈ σess(H±∞)isaneigenvalueofinfinitemultiplicity.

Observethatthedescription(5)of σ(H±∞)isinagreementwiththelimitingspectrumstatedin Theorem 2.1.Thisheuristicallymotivatestoproposetheoperators H±∞ definedin(3)and(4)asthe limitingoperatorsof Hτ ,as τ →±∞.Toseeinwhichsensetheconvergenceholdstrue,westudythe resolventconvergenceof Hτ to H±∞ as τ →±∞;see[8,Chapter8]forasurveyonresolventconvergence.

Theorem3.1 ([4,Theorem1.2]). Given τ ∈ R,let Hτ betheoperatordefinedin (2).Let H+∞ and H−∞ betheoperatorsdefinedin (3) and (4),respectively.Then, Hτ convergesto H±∞ inthestrongresolvent senseas τ →±∞.Thatis,foreveryf ∈ L2(Ω)4 lim τ →±∞ ∥((H±∞ λ) 1 (Hτ λ) 1)f ∥L2 (Ω)4 =0 forall λ ∈ C \ R (6)

1 Thenotation φ =(u, v )⊺ referstothedecompositionof φ :Ω → C4 inupperandlowercomponents,thatis,if φ = (φ1 , φ2 , φ3 , φ4 )⊺ with φj :Ω → C for j =1,2,3,4,then u =(φ1 , φ2 )⊺ and v =(φ3 , φ4 )⊺

https://revistes.iec.cat/index.php/reports

Aproofofthistheorembasedondirectlyestimatingthedifferenceofresolventsin(6)canbefound in[3,Section3.2],andanalternativeproofbasedonthenotionofstronggraphlimit[8,Definitionin p.293]canbefoundbothin[3,Section3.1]andin[4,Section2].Animmediateconsequenceofthis theoremisthefollowingresult,whichisanimprovementofitem (ii) inTheorem 2.1 forthefirstpositive eigenvalueof Hτ

Corollary3.2 ([4,Corollary1.3]). Forevery τ ∈ R,denotethefirstpositiveeigenvalueof Hτ in Ω by λΩ(τ ):=min(σ(Hτ ) ∩ (m,+∞)).Then, limτ ↑+∞ λΩ(τ )= ΛΩ + m2,where ΛΩ :=min σ( ∆D ) is thefirsteigenvalueoftheDirichletLaplacianin Ω

ItisremarkabletopointoutthatTheorem 3.1 doesnotensurethattheconvergencein(6)isuniformin theunitballof L2(Ω)4,butonlypointwiseforevery f ∈ L2(Ω)4.Actually,wenowjustifythattheconvergencecannotbeuniform—inthelanguageofresolvents,thismeansthat Hτ cannotconvergeto H±∞ in thenormresolventsenseas τ →±∞;see[8,Definitioninp.284]orTheorem 3.4 below—:indeed,ifthere wasconvergenceinthenormresolventsense,[9,Satz9.24]wouldleadtolimτ →±∞ σess(Hτ )= σess(H±∞), butthisisimpossiblesince σess(H±∞) = ∅ —recallthat ∓m isaneigenvalueofinfinitemultiplicity—and σess(Hτ )= ∅ forall τ ∈ R —because σ(Hτ )ispurelydiscrete.

Thisargumentshowsthattheessentialeigenvalue ∓m ∈ σess(H±∞)prevents Hτ fromconverging to H±∞ inthenormresolventsense.Itisthennaturaltoaskwhetherthenormresolventconvergence couldbeachievedif,insomesense,thestudywasrestrictedto σ(H±∞) \{∓m}.Anaffirmativeanswer holdstrueinthefollowingsense.Denote

ker(H±∞ ± m):= {ψ ∈ Dom(H±∞) ⊂ L2(Ω)4 :(H±∞ ± m)ψ =0}, ker(H±∞ ± m)⊥ := {φ ∈ L2(Ω)4 : ⟨φ, ψ⟩L2 (Ω)4 =0forall ψ ∈ ker(H±∞ ± m)}

Sinceker(H±∞ ± m)⊥ isaclosedsubspaceof L2(Ω)4,theorthogonalprojection

(7) isawell-definedboundedself-adjointoperatorin L2(Ω)4.Moreover,from(5)weknowthatker(H±∞ ± m)⊥ = {0} and,thus, ∥P±

L2 (Ω)4 →L2 (Ω)4 =1.

Theorem3.3 ([4,Theorem1.4]). Given τ ∈ R,let Hτ betheoperatordefinedin (2).Let H+∞ and H−∞ betheoperatorsdefinedin (3) and (4),respectively.Then, lim τ →±∞ ∥P±((H±∞ λ) 1

whereP± aretheorthogonalprojectionsdefinedin (7)

Aproofofthistheoremcanbefoundin[4,Section3].AswementionedafterCorollary 3.2,thedifference ofresolvents(H±∞ λ) 1 (Hτ λ) 1 doesnotconvergetozeroinoperatornormas τ →±∞.However, ifwewritethisdifferenceas

(H±∞ λ) 1 (Hτ λ) 1 =(P± +(1 P±))((H±∞ λ) 1 (Hτ λ) 1), thenTheorem 3.3 showsthattheeigenvalue ∓m isindeedtheonlyobstructionforhavingnormresolvent convergenceof Hτ to H±∞ as τ →±∞,since(1 P±)(L2(Ω)4)=ker(H±∞ ± m).

Althoughthemaininterestisthestudyoftheconvergenceof Hτ inaresolventsenseas τ →±∞,for thesakeofcompletenesswealsostudytheconvergencewhen τ approachesafinitevalue τ0 ∈ R

Reports@SCM, 10 (2025),23–29;DOI:10.2436/20.2002.02.47.

ConvergenceofgeneralizedMITbagmodels

Theorem3.4. Given τ ∈ R,let Hτ betheoperatordefinedin (2).Then,forevery τ0 ∈ R, Hτ converges to Hτ0 inthenormresolventsenseas τ → τ0.Thatis,

Aproofofthistheorembasedonthefactthattheresolventoperator(Hτ λ) 1 isrealanalyticin τ in aneighborhoodof τ0 —givenby[1,Lemma3.1]—canbefoundin[3,Section3.4].Analternativeproof basedonestimatingtheoperatornormofthedifferenceofresolventscanbefoundin[4,Section4].

4.Shapeoptimization

Ahotopenprobleminspectralgeometryistoprovethatthefirstpositiveeigenvalue λΩ(τ )of Hτ isminimal, amongallbounded C 2 domainsΩ ⊂ R3 withprescribedvolume,whenΩisaball;see[1,Conjecture1.8].

TheanalogousstatementforthefirsteigenvalueoftheDirichletLaplacian,ΛΩ :=min σ( ∆D ),isknown tobetrue,anditistheso-called Faber–Krahninequality —provenindependentlybyFaberin1923and Krahnin1925[5, 7],assertingthatΛΩ > ΛB wheneverΩ ⊂ R3 isaboundeddomainwithLipschitz boundarydifferentfromaball B withthesamevolume.

Asanapplicationoftheresultsobtainedin[3, 4]andpresentedinthispaper,weconcludewitha statementsupporting(butnotproving)theoptimalityoftheballfor λΩ(τ ).Ontheonehand, τ → λΩ(τ )is anincreasingandcontinuousfunctionin R,thatconvergesto m as τ →−∞ —byTheorem 2.1—and thatconvergesto ΛΩ + m2 as τ ↑ +∞,byCorollary 3.2;inparticular, τ → λΩ(τ )isbijectivefrom R to(m, ΛΩ + m2).Ontheotherhand,ifΩisnotaball,thenbytheFaber–Krahninequalitywehave (m, ΛB + m2) ⊊ (m, ΛΩ + m2).Therefore,thereexistsalargeenough τΩ ∈ R suchthat λΩ(τ ) ∈ ( ΛB + m2 , ΛΩ + m2)forall τ ≥ τΩ.Since λB (τ ) < ΛB + m2 forallsuch τ —byTheorem 2.1 and Corollary 3.2—,wegetthefollowing.

Proposition4.1. Let Ω ⊂ R3 beaboundeddomainwithC 2 boundary,andletBbeaballsuchthat |Ω| = |B|.If Ω isnotaball,thenthereexists τΩ ∈ R suchthat λB (τ ) <λΩ(τ ) forall τ ≥ τΩ

Itisveryremarkabletosaythatthelargeenough τΩ ∈ R ensuringtheoptimalityoftheballforthe firstpositiveeigenvalue λΩ(τ )intheregime τ ≥ τΩ dependsitselfonΩ.Hence,fromProposition 4.1 one can not ensurethatthereexistsalargeenough τ⋆ ∈ R forwhich λΩ(τ ) >λB (τ )forall τ ≥ τ⋆ and every bounded C 2 domainΩdifferentfromaball B withthesamevolume.Toproveordisprovetheexistence ofsuch τ⋆ alsoremainsasanopenandchallengingproblem.

Acknowledgements

TheauthorissupportedbytheSpanishgrantsPID2021-123903NB-I00andRED2022-134784-Tfundedby MCIN/AEI/10.13039/501100011033,byERDF“AwayofmakingEurope”,andbytheCatalangrant2021SGR-00087.ThisworkissupportedbytheSpanishStateResearchAgency,throughtheSeveroOchoaand Mar´ıadeMaeztuProgramforCentersandUnitsofExcellenceinR&D(CEX2020-001084-M),andmore specificallybythegrantCEX2020-001084-M-20-1.TheauthoracknowledgesCERCAProgramme/GeneralitatdeCatalunyaforinstitutionalsupport.

TheauthorthanksSocietatCatalanadeMatem`atiquesandInstitutd’EstudisCatalansforawarding witharunnerupthemaster’sthesis[3]inthe62ndeditionofthe ´ EvaristeGaloisAward.Theauthoralso thanksAlbertMasforhisguidance,dedication,andcommitment.

https://revistes.iec.cat/index.php/reports

References

[1]N.Arrizabalaga,A.Mas,T.Sanz-Perela,L.Vega,EigenvaluecurvesforgeneralizedMITbag models, Comm.Math.Phys. 397(1) (2023), 337–392.

[2]J.Behrndt,M.Holzmann,A.Mas,Self-adjoint Diracoperatorsondomainsin R3 , Ann.Henri Poincar´e 21(8) (2020),2681–2735.

[3]J.Duran,Spectralgapofgeneralized MITbagmodels,Master’sThesis,UniversitatPolit`ecnicadeCatalunya,2024. https://upcommons.upc.edu/handle/2117 /400748?locale-attribute=en

[4]J.Duran,A.Mas,Convergenceofgeneralized MITbagmodelstoDiracoperatorswithzigzag boundaryconditions, Anal.Math.Phys. 14(4) (2024),Paperno.85,23pp.

[6]M.Holzmann,Anoteonthethreedimensional diracoperatorwithzigzagtypeboundaryconditions, ComplexAnal.Oper.Theory 15(3) (2021),Paperno.47,15pp.

[7]E.Krahn, UbereinevonRayleighformulierte MinimaleigenschaftdesKreises, Math.Ann. 94 (1925),97–100.

[8]M.Reed,B.Simon, MethodsofModernMathematicalPhysics.I.FunctionalAnalysis,Second edition,AcademicPress,Inc.[HarcourtBrace Jovanovich,Publishers],NewYork,1980.

[9]J.Weidmann, LineareOperatoreninHilbertr¨aumen.Teil1,Grundlagen,Mathematische Leitf¨aden,B.G.Teubner,Stuttgart,2000.

[5]C.Faber, Beweiss,dassunterallenhomogenenMembranevongleicherFl¨acheund gleicherSpannungdiekreisf¨ormigedietiefstenGrundtongibt,Sitzungsber.-Bayer.Akad. Wiss.,Math.-Phys.Munich.(1923),169–172.

Reports@SCM, 10 (2025),23–29;DOI:10.2436/20.2002.02.47.

ANELECTRONICJOURNALOFTHE SOCIETATCATALANADEMATEM ` ATIQUES

MonodromyconjectureforNewton non-degeneratehypersurfaces

OriolBaezaGuasch UniversitatPolit`ecnica deCatalunya oriol.baeza@estudiantat.upc.edu

Resum (CAT)

AquesttreballestudialaConjecturaFortadelaMonodromia(SMC)enla versi´otopol`ogica.Despr´esd’introduirelsconceptesderesoluci´odesingularitats, polinomideBernstein–Satoilafunci´ozeta,esbocemelsresultatsinvolucratsen lademostraci´odelaSMCperasingularitatsNewtonnodegenerades(NND). Aquestaprovarequereix,per`o,hip`otesisaddicionalssobreelsnombresdelresidu, iconstru¨ımexemplesquemostrenquenopodenometre’s,laqualcosasuggereix quecalenaltrest`ecniquesperaatacarelcasgeneral.

Abstract (ENG)

ThisworkstudiestheStrongMonodromyConjecture(SMC)initstopologicalsetting.Afterintroducingtheconceptsofresolutionofsingularities,Bernstein–Sato polynomial,andthezetafunction,wesketchtheresultsinvolvedintheproofof theSMCforNewtonnon-degenerate(NND)singularities.Thisapproachrequires nonethelessadditionalhypothesisontheresiduenumbers,andweconstructexamplesshowingthattheycan’tbedropped,whichsuggeststhatnewtechniquesare neededtoattackthegeneralcase.

Keywords: monodromy,Bernstein–Satopolynomial,resolutionofsingularities,planecurves,Newtonnon-degenerate.

MSC(2020): Primary14B05,32S40.Secondary14H20,14J17,34M35.

Received: July1,2025.

Accepted: July18,2025.

31 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),31–42;DOI:10.2436/20.2002.02.48.

MonodromyconjectureforNewtonnon-degeneratehypersurfaces

1.Introduction

Themonodromyconjectureisaprobleminthefieldofsingularitytheoryinalgebraicgeometry,formulated bytheJapanesemathematicianIgusaintheseventies,whichrelatestwoinvariantsofasingularity.Onthe onehand,therootsofapolynomialarisingfromafunctionalequationsatisfiedbythesingularity(theso calledBernstein–Satopolynomial).Ontheotherhand,thepolesofthezetafunction(inoursetting,the topologicalversion),whichcontainsinformationaboutaresolutionofthesingularity.Moreprecisely,the conjecturepredictsthateverypoleofthiszetafunctionisarootoftheBernstein–Sato.

Althoughthegeneralcaseremainsopen,apositiveresulthasbeenprovenforsomespecialcases. Inparticular,itisknowntobetrueforplanecurves(Loeser’88),forNewtonnon-degenerate(NND) polynomialsmoduloanhypothesisonthesocalled residuenumbers (Loeser’90),aswellasincertain hyperplanearrangements,oralsosemi-quasihomogeneoussingularities.

BothinthecasesofplanecurvesandNNDpolynomials,amorecombinatorialapproachispossible, whichsimplifiessomecomputationsandallowstousesometechnicalcohomologicalresults.Nonetheless, fortheNNDcase,thiscomeswiththepriceofaddingtwohypothesisontheresiduenumbers.Wediscuss thepossibilityofremovingthehypothesis,andshowthattheydonotholdingeneral.Evenmore,we willseethatdivisorsnotsatisfyingthemcanstillcontributetothepolesofthetopologicalzetafunction, suggestingthatthisapproachwon’tworkforthegeneralcase.

2.Preliminaries

2.1.Complexzetafunctionandresolutionofsingularities

Beforestatingtheconjecture,wemustintroducethetwomainobjectsoftheproblem:theBernstein–Sato polynomialandthetopologicalzetafunction.Evenmore,togivesomecontextandmotivationofthe statement,wemustfirstbeginwiththecomplexzetafunction.

Thecomplexzetafunction,forapolynomial f andatestfunction ϕ (meaningacomplexfunction C∞ withcompactsupport),isdefinedas Z (s)= Z (f , ϕ; s) := Rn |f (x )|s ϕ

wheretechnicallywemustunderstandthisadistributioninthespaceoftestfunctions.Itcanbechecked that Z (s)convergesandisholomorphicinthesemiplane ℜ(s) > 0.Itsmeromorphiccontinuationandthe distributionofthepossiblepoleswasposedasaproblembyI.Gelfand[11, §3.I],andsolvedintwodifferent manners.

Ononehand,wecanusearesolutionofsingularities(guaranteedincharacteristic0byHironaka[12]). Recallthatanembeddedresolutionofapolynomial f isapropermorphism π : Y → X suchthat Y is smooth,therestrictionof π outsidethesingularlocusisanisomorphism,andthataroundeachpointin thepreimagewehaveaneighborhoodandachartoverwhich π∗f = u(y )y N1 i1 ··· y Nr ir with u(0) =0aunit and Ni ≥ 0integers.Fromthelocalexpression,wecanwritethepullbackdivisorgloballyas div(π ∗f )= j ∈J Nj Ej , https://revistes.iec.cat/index.php/reports

OriolBaezaGuasch

where(Ei )i ∈J aretheirreduciblecomponentsof π 1(f 1(0)),each Ei giveninlocalcoordinatesby {xi =0}, respectively.Anotherrelevantnumericalquantitythatwillappeararethecoefficientsintheglobalexpression ofthepullbackofthestandardvolumeform

Inthiscontext,wecanusetheresolutionasachangeofvariablesintheintegralanddeducethatthe polesofthecomplexzetafunctionareoftheform kj +ν Nj ,with ν anon-negativeinteger,and(kj , Nj )the numericaldataassociatedtotheexceptionaldivisors,introducedabove.

2.2.Bernstein–Satopolynomial

Ontheotherhand,wecanintroducetheBernstein–Satopolynomial,whichisananalyticalinvariant(and notatopologicalone)ofthesingularity.First,consider R := C[x1,..., xn ](ormoregenerallytheringof holomorphicfunctionsorevenformalpowerseries)anddenote D := R⟨∂1,..., ∂n ⟩ theWeylalgebra.All elementscommuteexceptfortherelations ∂i xi xi ∂i =1,andsoitiseasytoshowthatanyelement(a priorionlyalineardifferentialoperator)canbewrittenasafinitesum P = α,β aα,β x α∂β .Foramore gentleintroductionandmoredetailsonthepropertiesoftheWeylalgebra,wereferto[5].Next,consider thepolynomialring D [s] := D ⊗C C[s]withnewvariable s,andnotethatthefreemodule Rf [s] f s hasa naturalstructureofleft D [s]-modulegivenbytheproductrule.Indeed,everyelementofthemodulecan bewrittenas g f k f s forsome g (x , s) ∈ R[s],andtheactionofthepartialderivativesis

i g f k f s = ∂i g f k f s + sg f k +1 ∂f ∂xi f

Therelevantresultisthentheexistenceofsolutionstothefollowingfunctionalequation,whichwasfirst provenbyBernstein[2]forthecaseofpolynomials,andlaterbyKashiwara[13]andBj¨ork[3]inthecases ofholomorphicfunctionsandformalseries,respectively.

Theorem2.1 ([2]). Letf ∈ Rbeapolynomial.Then,thereexistsapolynomialP (s) ∈ D [s] anda polynomialbf ,P (s) ∈ C[s] suchthattherelation

holdsformallyinthe D -moduleRf [s] f s

Thesetofpolynomials bf ,P satisfyingsuchadifferentialequationasaboveformanidealin C[s],so wecanconsideritsmonicgenerator:the Bernstein–Satopolynomial of f denotedby bf (s).

Example2.2. For f = x 2 1 + ··· + x 2 n wehave bf (s)=(s +1)(s + n/2),astaking P(s)tobetheLaplacian operator(thoughtasaconstantpolynomialin D [s]),wehavetherelation ∂2 ∂x 2 1 + ··· + ∂2 ∂x 2 n f s +1 =4(s +1) s + n 2 f s 33

Reports@SCM, 10 (2025),31–42;DOI:10.2436/20.2002.02.48.

MonodromyconjectureforNewtonnon-degeneratehypersurfaces

Example2.3. For f = x 2 + y 3 in C[x , y ]wehavethefollowingrelation

anditcanbeprovedthat

Now,havingintroducedthefunctionalequation(1),theideaistouseittointegratebypartsand obtainameromorphiccontinuationof Z (s)tothewholecomplexplane,asforany r ∈ N wehave Z (s)= 1

Inthiscase,itcanbeseenthatthepolesofthezetafunctionareoftheform λ ν for λ arootof bf (s) and ν anon-negativeinteger.Bycomparingthiswiththepreviouscandidatequantitiesforthepoles,one arrivesatthefollowingresult.

Theorem2.4. EveryrootoftheBernstein–Satopolynomialbf isoftheform kj +ν Nj forsomej ∈ Jand ν anon-negativeinteger.

2.3.Monodromyconjecture

Altogether,therelationbetweenpolesofthecomplexzetafunctionandtherootsoftheBernstein–Sato isclear.Motivatedbythis,andaftercomputingsomeexamples,Igusaformulatedtheconjectureforthe polesofthe p-adiczetafunction,andlaterstatedinthetopologicalandmotivicsettingstoo.

Inthetopologicalversion,wehavelefttointroducethetopologicalzetafunction,firstdefinedbyDenef andLoeserin[8],whoformalizeditsdefinitionfromaheuristicargumenttakingalimitofthe p-adiczeta function,andshowingthatthefollowingexpressionisindependentofthechoiceofaresolution(thereis currentlynoknownintrinsicdefinition).

Definition2.5 (Topologicalzetafunction). Let f ∈ C[x1,..., xn ]beanon-constantpolynomial,andchoose aresolution π : Y → An C of {f =0}.The(global) topologicalzetafunction of f is

Then,itisclearfromtheexpressionsthatthepolesarestillrelatedwiththerootsoftheBernstein–Sato, astheybothcontaininformationfromaresolution.However,thedifficultylies,ononeside,indetermining whichpolesremaininthezetafunctionafterpossiblecancellation,andintheother,whatcandidatevalues inTheorem 2.4 areactuallyroots.Themonodromyconjecturepartlyanswersthis,bypredictingthatevery poleofthetopologicalzetafunctionisaroot.

Conjecture2.6 (Topologicalmonodromyconjecture). Letf ∈ C[x1,..., xn ] beanon-constantpolynomial. Ifs0 isapoleofZtop(f ; s),then (i) (standard) e 2πıℜ(s0) isaneigenvalueofthemonodromyoff : Cn → C atapointof {f =0} (ii) (strong) s0 isarootoftheBernstein–Satopolynomialbf (s)

https://revistes.iec.cat/index.php/reports

Inthiswork,wealwaysrefertothestrongversion,whichimpliesthestandard(orweak)onethanksto aresultbyMalgrange([20,Proposition7.1]).Sofar,theapproachesforthesolvedcasesdonotprovidea clearconceptualideawhytheresultshouldbetrue,apartfromtheanalogywiththecomplexzetafunction (see[22,Remark2.13]).

Akeycommonelementintheproofsoftheknowncasesisthestudyofperiodsofintegrals(see[18, 19]). Theseobjectsallowtorelateitsasymptoticbehaviorwiththemonodromyaction(thatis,theactionofa generatorofthefundamentalgroupofapunctureddiskaroundthesingularityofthehomologygroupsofa fiber Xt oftheMilnorfiber).Evenmore,inthiscontextMalgrangeisabletoprovethatcertainquantities appearingintheasymptoticexpressionsarerootsof bf (s).

Nonetheless,thisapproachrequirestheexistenceofanon-zerocohomologyclassin Hn (Xt , C).This resultispreciselywhatDeligneandMostowproveinthecaseofplanecurves([7,Proposition2.14]), andLoeserforthecaseofNewtonnon-degenerate([15,Th´eor`eme3.7])adaptingaresultbyEsnaultand Viehweg([9]).

Furthermore,inthecaseofplanecurves,thecohomologicalresultofexistenceonlyappliestorupture divisorsoftheresolution.Forthisreason,Loesercomplementsitwithacombinatorialstudyofa(minimal) resolutionofthesingularity,representedinthesocalled dualgraph.Inthisgraph,eachvertexrepresents arupturedivisor Ei , i ∈ J,oftheresolution,andtheedgesareaddedandmodifiedaftereach blowup.In thissituation,itispossibletostudythequotient ki /Ni andtheresiduenumbers ε(i , j )= kj Nj ki /Ni over thedivisorsviarecurrencesinthedualgraph.Inparticular,itcanbeproventhattheonlypossiblepoles contributingtothetopologicalzetafunctionarisefromtherupturedivisors,andsotheaforementioned cohomologicalresultisenough.

AsfortheNNDcase,inthenextsectionswewillseehowthetechnicalhypothesisrequiredforthe cohomologicalresultleadtoadditionalconditionsontheresiduenumbers.Wewillshowthattherelevant hypothesisisthesecondone,andwewillconstructexamplesthatprovethattheyarenotalwayssatisfied. Evenworse,wewillseethatthe bad divisorsviolatingitcancontributewithnon-zeroresiduetoapoleof thetopologicalzetafunction.

3.Newtonnon-degeneratepolynomials

3.1.Definitionandproperties

PolynomialswhichareNewtonnon-degeneratearesometimesalsocallednon-degeneratewithrespectto theNewtonpolytope,orsimplynon-degenerate.Theconditionthatthesepolynomialssatisfyhasamore combinatorialflavor,asitiseasierdescribedbyconsideringtheobjectsthatwewillintroducenext.

Weconsiderapolynomial f (x1,..., xn )= p∈Nn ap x p 1 x p n suchthat f (0)=0.Forbrevity,we willusemulti-indexnotationwhenconvenient f (x )= p∈Nn ap x p ,andwedefineitssupporttobe supp(f )= {p ∈ Nn | ap =0}

Definition3.1 (Newtonpolyhedron). Let f = p∈Nn ap x p ∈ C[x ]with f (0)=0.Wedefinethe global Newtonpolyhedron Γgl (f )of f astheconvexhullofsupp(f ).Also,wedefinethe localNewtonpolyhedron Γ(f )astheconvexhulloftheset p∈supp(f ) p +(R≥0)n

Wewillusetheterm face ofΓ(f )torefertoanyconvexsubset τ thatcanbeobtainedbyintersecting theNewtondiagramwithahyperplane H of Rn suchthatΓ(f )iscontainedinoneofthehalf-spaces definedby H.Notethatwealsoconsiderthetotalpolyhedronasaface.

Reports@SCM, 10 (2025),31–42;DOI:10.2436/20.2002.02.48.

MonodromyconjectureforNewtonnon-degeneratehypersurfaces

Definition3.2 (Non-degenerate). Wesaythat f isNewtonnon-degenerateat0ifforanyface τ ⊂ Γ(f ), thehypersurfacedefinedbythetruncation f τ := p∈τ ∩supp(f ) ap x p satisfiesthatthepolynomials xi

xi for i =1,..., n donotvanishatthesametimein(C \ 0)n

ThisclassofNNDpolynomialsisgeneralenoughtobeofinterest,whilealsoallowsamorecombinatorial treatmentoftheproblem.Wewillnowseehowthepolyhedronanditsdualfanencodetheinformationof agoodresolutionofthesingularity.Forthat,wefirstintroducesomenotation.

Definition3.3 (N, k, F ) Foravector a ∈ (R+)n ,wedefinethequantities N(a) :=inf x ∈Γ(f ){⟨a, x ⟩} and k(a) := n i =1 ai .Also,definethefirstmeetlocus F (a) := {x ∈ Γ(f ) |⟨a, x ⟩ = N(a)},whichisaproper faceofΓ(f )if a =0,and F (0)recoversthewholediagram.

Definition3.4 (Dualfan). For τ afaceofΓ(f ),wedefinethe coneassociated to τ as ∆τ := {a ∈ (R>0)n | F (a)= τ }/ ∼ , a ∼ a ′ iff F (a)= F (a ′).

ThecollectionoftheseconesforallfacesoftheNewtonpolytopeasthe dualfan. Asanexample,seethefollowingFigure 1,whereweconsidertheplanecurvedefinedby f = x 3 y 2 + 4xy +3x 2y ,andconstructtheNewtonpolyhedronwithlabeledfacesandcorrespondingtruncations(left), andtheassociateddualfan(right).

Figure1:Γ(f )anddualfanofthepolynomial f = x 3 y 2 +4xy +3x 2y

Next,werecallthefollowingpropertiesofcones.

Definition3.5 (Cone) A convexpolyhedralcone,or cone forshort,isaset

where V isan n-dimensionalvectorspaceover R,andthevectors {vi } arecalledthe generators ofthe cone.Thedimensionof C isdefinedtobethedimensionofthesmallestvectorspacecontainingit. Wesaythattheconeis simplicial ifitsgeneratingvectorsarelinearlyindependentover R.Moreover, wewillsayitis simplicialrational ifontopofthattheentriesofthevectorsareintegers.Wesaythatthe coneis regular (or simple)ifthesetofgeneratingvectorscanbeextendedtoabaseofthe Z-module Zn

Then,thekeytheoremfromtoricgeometrythatwewillneedisthefollowing.

Theorem3.6 ([14,pp.32–25]). Let ∆ beaconegeneratedbyvectorsv1,..., vr ∈ Rn \{0}.Thereexists afinitepartitionof ∆ incones ∆i ,suchthateachconeisgeneratedbyasubsetoflinearlyindependent vectorsof {v1,..., vr }.Moreover,if ∆ issimplicialrational,apartitioninregularconescanbeobtainedby introducingsuitablenewgeneratingrays.

Sucharegularsimplicialsubdivisioncanbeobtainedalgorithmically(seealso[1, §8.2.2]),andin particularapplyingittoalltheconesinthedualfan,wededucethenextresult.

Theorem3.7 ([1,Lemma8.7]). Thereexistsafanconsistingofregularsimplicialconeswhichisobtained asasubdivisionofthedualfanassociatedtotheNewtonpolyhedron.

Remark 3.8 Notice,however,thatwehaven’tclaimedanythingaboutuniqueness,astherecanexist multiplevalidsubdivisionsinsimplicialregularcones.

3.2.Resolutionandtopologicalzetafunctionfromthedualfan

Backtoresolutionofsingularities,weknowthankstotheresultbyHironakathatwealwayshaveone,and thatitcanbeobtainedasacompositionofblowups.However,inthissetupwewillobtainaresolution moredirectlyviaasingletoricblowup,whichwedefinenext(seeamoredetailedintroductionin[21] and[17],andalso[10, 6]foracompletetreatmentoftoricvarieties).

Definition3.9 (Toricblowup). Consideraunimodularintegral n × n matrix σ =(σi ,j ),anddefinethe toricblowup (or modification)associatedtoσ asthebirationalmorphism

).

Inparticular,ifwehavearegularsimplicialconeinΣ∗ ofmaximumdimensiongivenbyvectors {r1,..., rn }, wecanconsidertheunimodularmatrix σ =(r1 r2 rn )andassociatetoitthebirationalmap πσ : Cn σ → Cn ,whichweoughttothinkaboutasoneofthedifferentchartsoftheresolution.Withthat,weconstruct anon-singularvariety X asthequotientofthedisjointunion σ Cσ overalltheregularconeswiththe followingidentification.Twopoints x ∈ Cn σ and y ∈ Cn τ areidentifiedif,andonlyif,thebirational map πτ 1σ isdefinedatthepoint x and πτ 1σ (x )= y

Itcanbeverifiedthat X isnon-singular,andthemaps {πσ : Cn σ → Cn | σ regularsimplicialcone} glue intoaproperanalyticmap π : X → Cn

Definition3.10 (Associatedtoricblowup) Themap π : X → Cn iscalledthe toricblowup (ormodification) associated withΣ∗ attheorigin,whereΣ∗ isaregularsimplicialconesubdivisionofΣ.

Finally,wearriveattheresultjustifyingourclaimthatthegeometricpropertiesoftheNewtonpolyhedroncontainstheinformationofaresolutionforNNDsingularities.

Theorem3.11 ([21,p.101]). IffisNewtonnon-degenerate,thentheassociatedtoricblowup π : X → Cn isagoodresolutionoffasagermattheorigin. 37 Reports@SCM, 10 (2025),31–42;DOI:10.2436/20.2002.02.48.

MonodromyconjectureforNewtonnon-degeneratehypersurfaces

Inthesamecombinatorialspirit,itispossibletoobtainamoreexplicitexpressionforthetopological zetafunction.Forthat,letusfirstintroducethefollowingterms.

Definition3.12. Let τ beafaceinΓ(f ),andconsideradecompositionoftheassociatedcone∆τ = r i =1 ∆i insimplicialconesofdimensiondim∆τ = l suchthatdim(∆i ∩ ∆j ) < l ,forall i = j .Then,define

J(τ , s) := r i =1 J∆i (s),with J∆i (s)= mult(∆i ) (N(ai1 )s + k(ai1 )) (N(ail )s + k(ail ))

being ai1 ,..., ail ∈ Nn thelinearlyindependentprimitiveintegralvectorsthatgenerate∆i .Lastly,if τ = Γ(f ),werathertake J(τ , s)=1.

Remark 3.13 By[8,Lemme5.1.1],thedefinitionof J(τ , s)isindependentofthechoiceofthedecompositionof∆τ insimplicialcones.

Theorem3.14 ([8,Th´eor`eme5.3]) Letf ∈ C[x ] beapolynomialNewtonnon-degenerate,then

Z (s)= τ vertex of Γ(f ) J(τ , s)+ s s +1 τ faceof Γ(f ) dim τ ≥1 ( 1)dim τ (dim τ )!Vol(τ )J(τ , s).

Asthepolesof Z (s)arisefromthepolesoftheterms J(τ , s)inthesum,weseethattheyarestill either 1oroftheform k(a)/N(a)(seealso[16,Th´eor`eme5.3.1]),whichjustifiesthechoiceofnotation.

3.3.Sketchoftheproofandadditionalhypothesis

WearenowreadytosketchtheapproachbyLoeserintheproofofthemonodromyconjectureforthe Newtonnon-degeneratecase.Onerelevantsubtletyis,asstatedinRemark 3.8,thattheregularsubdivision isnotunique,andthereforesoaretheresiduenumbers.Forthatreason,Loeserdecidestoworkwith toric residuenumberscomputeddirectlyfromtheoriginaldualfan,withoutperformingaregularsubdivision.

Definition3.15 (Toricresiduenumbers) If τ , τ ′ aretwodistinctfacesofcodimension1oftheNewton polyhedronattheoriginof f ,wedenoteby β(τ , τ ′)thegreatestcommondivisoroftheminorsoforder2 ofthematrix(a(τ ), a(τ ′)),where a(τ )isaprimitiveintegralvectordefiningtheface τ .Additionally,set

whenever N(τ ) =0,whichisthecaseif τ isacompactface.

However,thedrawbackisthattheresolutionobtainedfromtheoriginaldualfanneednotbeminimal, andthesubsequentcomputationsrequirespecialcare,leadingtotheintroductionofthis β factorwhich appearsasthedegreeofafinitemorphismbetweensingulartoricvarieties. Then,thepositiveresultprovenbyLoeseristhefollowing.

OriolBaezaGuasch

Theorem3.16 ([16,Th´eor`eme5.5.1]). Letfbeacomfortablepolynomialverifyingf (0)=0,withNewton diagram Γ(f ),andNewtonnon-degenerate.Supposethatallcompactfaces τ0 verify (i) k (τ0) N (τ0) < 1

(ii)Foreveryface τ ofcodimension 1 of Γ(f ),distinctof τ0 andhavingnon-emptyintersectionwith τ0, wehave ε(τ0, τ ) / ∈ Z

Then,therealpartsofthepolesofthezetafunctionoffarerootsoftheBernstein–Satopolynomialoff.

Asexplained,thisisbasedonacohomologicalargumentofexistenceofanon-zeroclass,whichinturn forcesthetwostatedconditions,thesecondonebasicallyensuringthatthemonodromiesarenotidentity.

Remark 3.17 Loeseralreadypointsoutin[16,Remarque5.5.2.1]thatifonereplacesthecondition k (τ0) N (τ0) < 1with k (τ0) N (τ0) / ∈ N,thisisenoughtoprovetheweakversionoftheconjecture.

Asforthesecondhypothesis,althoughitisnotclearifitispossibletoremove,wecantrytorelaxit. Indeed,oneoughttoexpectthatnon-positiveresiduenumberscouldbeallowedtohappen,afterpossibly generalizingresultsinthespiritof[7,Proposition2.14]or[4,Proposition11.1].Therefore,wewillnext trytoconstructexampleswherethisisthecase,andanalyzethecontributionsofsuchdivisorsviolating thecondition.

3.4.Constructingcounterexamplestothesecondcondition

Tobegin,itshouldbementionedthatallexamplesstudiedhavebeencheckedtosatisfythetopological strongmonodromyconjecture(andthustheweakversiontoo).Nonetheless,therelevantfindingsare exampleswherethesecondconditiononthe(toric)residuenumbersisnotmet,andevenmore,where ε is apositiveinteger.

Evenmoreinterestingly,theseexampleshavebeenmotivatedgeometricallyfromtheNewtonpolyhedron.Moreprecisely, f isfirstconstructedbyintroducingmonomials x p + y q + z r for p, q, r largeenough integers.Then,addingsmallmixedmonomialsofthetype x s y t z u forsmallenoughintegers s, t, u,we obtainsmallcompactfacesclosetotheorigin.Bychoosingtheexponentsappropriately,wecanconstruct apolyhedronwhosefaceshavenormalvectorsasdesired.Inparticular,wecanfindpairsofadjacentfaces forwhichthecorrespondingdivisorsoftheresolutiongiveriseto(toric)residuenumbersthatareintegers.

Figure2:ExampleoftheconstructionofΓ(

MonodromyconjectureforNewtonnon-degeneratehypersurfaces

Asanillustrativeexample,weconsider f = x 5 + y 6 + z 7 + x 2yz + xyz 2 + xy 2z ,whoseNewtondiagram isdepictedintheaboveFigure 2.Theoriginalraysinthedualfanare: [(0,0,1),(0,1,0),(1,0,0),(1,1,1),(1,1,2),(1,2,1),(3,1,1),(6,5,14),(7,18,5),(23,7,6)], andaregularsubdivisionrequiresalmost400newrays.Also,theBernstein–Satopolynomialis

bf (s)=

andthe(local)topologicalzetafunction(whichisthesameexpressionastheglobalversionTheorem 3.14, butthesecondsumrunsonlyovercompactfaces)is

withpoles {−3/4, 4/5, 5/6, 6/7, 1}.Therefore,itisimmediatethatthestrongmonodromyconjectureholdsforthiscase.

However,wehavetheray(1,1,2)withcandidatepolevalue σ = k /N = 4/5appearing,andwhich isabaddivisor.Indeed,wehavethe(toric)residuenumber ε((1,1,2),(7,8,5))=2.

Inlightofthis,thenextnaturalquestioniswhetherwecandiscardthecontributionofthis bad divisor tothezetafunction,andworkonlywiththosesatisfyingbothconditions(inthespiritofthecaseofplane curves,whereastudyofthedualgraphallowedtodiscardnon-rupturedivisorsforwhichwedon’thave thecohomologicalresult).

Followingtheaboveexample,thetwodivisorswithcandidatepolevalue 4/5arethebaddivisor(1,1,2) andalso(1,2,1).Sowecomputethecontributionsofeachdivisortothe(local)topologicalzetafunction, bysummingonlythetermsfromcones∆i thatcontaintherayasoneofitsgeneratingrays.Asawarning, foraray a thisisnotthesameassimplytakingthetermswhereafraction 1 N (a)s +k (a) appears,asitcan happenthatanotherray a′ givesrisetothesamecandidatepole.

Inparticular,wefindthattheresiduesoftheindividualcontributionsare

sowecan’tdiscardthebaddivisor.

Remark 3.18 Anotherremarktopointhereisthatthetotalresiduewon’tnecessarilybethesumofthese residues.Indeed,therecanbeconeswherebothraysappearasgeneratingrays(thishappensprecisely iftheassociateddivisorsintersect),andinthatcasewewouldneedtosubtractthedoublycounted contribution.Moregenerally,aninclusion-exclusionformulashouldbeappliedinordertocomparethe separatecontributionsandthetotalonefordivisorswiththesamecandidatepolevalue.

Altogether,thereasoningandstudyofthepresentedexampleconfirmsthatwecan’tomittheadditional hypothesisontheresiduenumbersrequiredfortheproofoftheconjectureintheNewtonnon-degenerate caseand,evenworse,thatthepossibledivisorsnotsatisfyingitcanindeedcontributewithnon-zero residuetothepolesofthetopologicalzetafunction.Inotherwords,theapproachbyLoeserbasedonthe constructionofanon-zerocohomologyclassviathementionedresultsdoesnotallowtoextendtheproof oftheconjecturetothegeneralcase.

https://revistes.iec.cat/index.php/reports

OriolBaezaGuasch

Acknowledgements

TheauthorwaspartiallysupportedbyCFISMobilityProgram,anErasmus+grantco-fundedbythe EuropeanUnionandaSantanderErasmusscholarship.TheauthorwouldliketothankGuillemBlanco Fern´andezandJosep ` AlvarezMontanerfortheirsupervisionoftheproject,andKULeuvenforhostinghim duringitsdevelopment.

References

[1]V.I.Arnold,S.M.Gusein-Zade,A.N.Varchenko, SingularitiesofDifferentiableMaps. Volume2.MonodromyandAsymptoticsofIntegrals,TranslatedfromtheRussianbyHugh PorteousandrevisedbytheauthorsandJames Montaldi,Reprintofthe1988translation,Mod. Birkh¨auserClass.Birkh¨auser/Springer,New York,2012.

[2]J.H.Bernstein,Analyticcontinuationofgeneralizedfunctionswithrespecttoaparameter, FunctionalAnal.Appl. 6(4) (1972),273–285.

[3]J.-E.Bj¨ork,Dimensionsoveralgebrasofdifferentialoperators,D´epartementdemath´ematiques(1973).

[4]G.Blanco,Bernstein–Satopolynomialofplane curvesandYano’sconjecture,PhDThesis,UniversitatPolit`ecnicadeCatalunya,2020.

[5]F.J.CastroJim´enez,ModulesovertheWeyl algebra,in: AlgebraicApproachtoDifferential Equations,WorldScientificPublishingCo.Pte. Ltd.,Hackensack,NJ,2010,pp.52–118.

[6]D.A.Cox,J.B.Little,H.K.Schenck, ToricVarieties,Grad.Stud.Math. 124,AmericanMathematicalSociety,Providence,RI,2011.

[7]P.Deligne,G.D.Mostow,Monodromyofhypergeometricfunctionsandnonlatticeintegral monodromy, Inst.Hautes ´ EtudesSci.Publ. Math. 63 (1986),5–89.

[8]J.Denef,F.Loeser,Caract´eristiquesd’Euler–Poincar´e,fonctionszˆetalocalesetmodifications

analytiques, J.Amer.Math.Soc. 5(4) (1992), 705–720.

[9]H.Esnault,E.Viehweg, LecturesonVanishing Theorems,DMVSem. 20,Birkh¨auserVerlag, Basel,1992.

[10]W.Fulton, IntroductiontoToricVarieties,Ann. ofMath.Stud. 131,WilliamRoeverLectures Geom.,PrincetonUniversityPress,Princeton, NJ,1993.

[11]I.Gelfand,Someaspectsoffunctionalanalysisandalgebra,in: ProceedingsoftheInternationalCongressofMathematicians,Vol.1 (Amsterdam,1954),ErvenP.NoordhoffN.V., Groningen,1957,pp.253–276.

[12]H.Hironaka,Resolutionofsingularitiesofanalgebraicvarietyoverafieldofcharacteristiczero: II, Ann.ofMath.(2) 79(2) (1964),205–326.

[13]M.Kashiwara, B-functionsandholonomicsystems.Rationalityofrootsof B-functions, Invent.Math. 38(1) (1976/77),33–53.

[14]G.Kempf,F.Knudsen,D.Mumford,B.SaintDonat, ToroidalEmbeddings1,LectureNotes inMath. 339,Springer-Verlag,Berlin-New York,1973.

[15]F.Loeser,Fonctionsd’Igusa p-adiquesetpolynˆomesdeBernstein, Amer.J.Math. 110(1) (1988),1–21.

[16]F.Loeser,Fonctionsd’Igusa p-adiques,polynˆomesdeBernstein,etpoly`edresdeNewton, J. ReineAngew.Math. 412 (1990),75–96.

41 Reports@SCM, 10 (2025),31–42;DOI:10.2436/20.2002.02.48.

MonodromyconjectureforNewtonnon-degeneratehypersurfaces

[17]J.MacLaurin,Theresolutionoftoricsingularities,PhDThesis,SchoolofMathematics,The UniversityofNewSouthWales,2006.

[18]B.Malgrange,SurlespolynˆomesdeI.N.Bernstein,in: S´eminaireGoulaouic–Schwartz1973–1974: ´ Equationsauxd´eriv´eespartiellesetanalysefonctionnelle,Exp.No.20, ´ EcolePolytechnique,CentredeMath´ematiques,Paris,1974, 10pp.

[19]B.Malgrange,Int´egralesasymptotiqueset monodromie, Ann.Sci. ´ EcoleNorm.Sup.(4) 7 (1974),405–430.

[20]B.Malgrange,PolynˆomesdeBernstein–Satoet cohomologie´evanescente, Ast´erisque 101-102 (1983),243–267.

[21]M.Oka,Geometryofplanecurvesviatoroidal resolution,in: AlgebraicGeometryandSingularities (LaR´abida,1991),Progr.Math. 134, Birkh¨auserVerlag,Basel,1996,pp.95–121.

[22]W.Veys,Introductiontothemonodromyconjecture,in: HandbookofGeometryandTopologyofSingularitiesVII,Springer,Cham,2025, pp.721–765.

https://revistes.iec.cat/index.php/reports

ANELECTRONICJOURNALOFTHE SOCIETATCATALANADEMATEM ` ATIQUES

GerardCastroL´opez ETHZurich gcastro@student.ethz.ch

Resum (CAT)

Els´unicsexemplesexpl´ıcitsdev`ortexsenrotaci´ouniformealesequacionsd’Euler2D s´onelscerclesilesel lipses.Elsaltresexemplesdelsqualsesconeixenpropietats quantitativess´onpropersaaquests.

Enaquestarticlepresenteml’exist`enciadev`ortexsnoconvexos,llunydelsr`egims pertorbatius,podentobtenir-neunadescripci´oquantitativaprecisa.Perdemostrarhoutilitzemunacombinaci´od’an`alisiit`ecniquesdedemostraci´oassistidaper ordinador.

Abstract (ENG)

Theonlyexplicitexamplesofuniformlyrotatingvortexpatchestothe2DEuler equationsarecirclesandellipses.Theotherexamplesforwhichquantitativepropertiesareknownareclosetotheseones.

Inthispaperwepresenttheexistenceofnon-convexones,farfromtheperturbative regimes,beingabletoobtainaprecisequantitativedescription.Toproveitweuse acombinationofanalysisandcomputerassistedproofstechniques.

Keywords: PDE,fluidmechanics,2DEuler,vortexpatches,computerassistedproofs.

MSC(2020): Primary35Q31,57P10.Secondary65G30,76B03.

Received: July2,2025.

Accepted: August10,2025.

43 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),43–50;DOI:10.2436/20.2002.02.49.

Existenceofnon-convexrotatingvortexpatchestothe2DEulerequation

1.Introduction

Thebehaviorofideal,incompressiblefluidsisgovernedbytheEulerequations,asetofpartialdifferential equationsderivedbyLeonhardEulerinthe1750s.Intwodimensions,theseequationscanbeformulated intermsofvorticity, ω,whichmeasuresthelocalrotationofthefluidinthefollowingway:

where K (y)= y⊥ 2π|y|2 .Aspecialclassofsolutions,knownas vortexpatches,occurswhentheinitialvorticity isconstantwithinadefinedregion D0 andzeroelsewhere.Thevorticityremainsconstantalongparticle trajectories,sothepatchsimplydeformsovertime.Amoreexhaustiveintroductionwithmoredetailed computationscanbefoundin[10].

Thisstudyfocusesonaparticulartypeofvortexpatchcalleda V-state,whichisapatchthatrotates withauniformangularvelocity(Ω)withoutchangingitsshape.Mathematically,ifapatchisdefinedbya domain D0 attime t =0,itsevolutionisgivenby D(t)= M(Ωt)D0,where M isarotationmatrix.

Theboundaryofthepatch ∂D0,parametrizedinpolarcoordinatesbythefunction R(α),mustsatisfy thefollowingnon-localintegro-differentialequation:

where F [R]isanintegraloperatorgivenby:

1.1.Previouswork

Thecircleisatrivialsolution,andin1874,Kirchhoff([9])provedthatellipsesarealsoV-states.Apartfrom these,nootherexplicitsolutionsareknown.Inthe1970sand80s,numericalstudiesbyDeem–Zabusky[3] andothersrevealedfamiliesofV-stateswithm-foldsymmetrybifurcatingfromthecircle,suggestingarich varietyofsolutionsbeyondtheclassicalexamples.

Theoreticalprogressfollowed,withBurbea[1]andHmidi,Mateu,andVerdera[6]provingtheexistence oflocalbifurcationbranchesfromthediskforeveryintegersymmetry m ≥ 3.Morerecently,global bifurcationcurveswereconstructedbyHassainia,Masmoudi,andWheeler[5].However,thesepowerful theoreticaltoolsprovideexistencebutlackquantitativeinformationaboutthesolutionsfarfromtheinitial bifurcationpoint.Thisisacriticallimitationbecauseprovingpropertieslikenon-convexityrequiresprecise knowledgeofthesolution’sshape.Ageneralargumentforallm-foldbrancheswillfail,asthebranch for m =2isknowntobeconvexandfor m =3itisexpected.

Thispaperaddressesthisgapbyprovidingthefirstpositiveexistenceresultwithquantitativeinformation foraV-statefarfromtheperturbativeregionsaroundthecircleorellipses.Weprovetheexistenceofa non-convex,6-foldsymmetricV-state.

Duetolengthconstraints,mostproofsareomittedfromthisarticle.Amorethoroughexplanationof theresultsisdetailedbyG´omez-Serranoandthepresentauthorin[2].

https://revistes.iec.cat/index.php/reports

GerardCastroL´opez

1.2.Results

Theorem1.1 (Maintheorem). ThereexistsananalyticsolutionR (x) oftheV-stateequation (1),with Ω= 1537 3750 ,suchthatitsassociatedvortexpatchD ⊂ R2 isnon-convexandhas 6-foldsymmetry.See Figure 1

Corollary1.2. Thereexists δ> 0 suchthatforanyangularvelocityin (Ω δ,Ω+ δ) thereexistsan analyticsolutionRto (1) with 6-foldsymmetry,where Ω istheangularvelocitygivenbyTheorem 1.1

Figure1:Theboundary ∂D isacurvecontainedintheplottedline.

ProofofTheorem 1.1 InSection 2 weprovetheexistenceofasolutioncloseto R0,inSection 3 weprove theanalyticityandnon-convexityandinthelastSection 4 weexplainthecomputerassistedpartsofthe proof.

2.Existence

Ourapproachistoformulatetheproblemasafixedpointequationandsolveitusingacombinationof analyticalestimatesandrigorousnumericalcomputations.Thecoreideaistofindanapproximatesolution andthenprovethatatruesolutionexistsinitsvicinity.

2.1.Fixedpointequationaroundtheapproximatesolution

Thefirststepistofindahighlyaccurateapproximatesolution, R0(x).WedefineitasatruncatedFourier serieswith6-foldsymmetry(m =6):

, 10 (2025),43–50;DOI:10.2436/20.2002.02.49.

Existenceofnon-convexrotatingvortexpatchestothe2DEulerequation

with N0 =30.Thecoefficients ck andtheangularvelocityΩarechosentomaketheerrorofthis approximation, E [0](x )= R ′ 0(x)R0(x) F [R0](x),assmallaspossible.Weseekatruesolution R(x)asa perturbationof R0(x):

Here, u isafunctionin L2([0, π/m])and˜ u isitsodd,2π/m periodicextension.Thisformulationisdesigned tofixthescalingandrotationsymmetriesoftheproblemwhilepreservingthem-foldsymmetry.

TakingtheFr´echetderivativeoftheequation(1),wecanwritetheequationfortheperturbationas Lv = E [0]+ NL[v ],where L isalinearoperatorand NL[v ]= O(v 2)isnonlinear.Usingtheexpression(2) andthesymmetriesof˜ u,wewritethisequationintermsof u inthefollowingway:

[0, π/m].

• L isalinearoperatordefinedas Lu := u + π/m 0 K (x, y )u(y ) dy ,where K isa L2 0, π m 2 function thatdependsontheapproximatesolution R0 inanontrivialway.

• E0 istheerroroftheapproximatesolution R0 aswehavementionedbefore.

• NL isanonlinearoperatorthatcontainshigher-ordertermsoftheperturbation u

Ifweprovethat L isinvertible,theproblemisnowreducedtoshowingthattheoperator Gu := L 1(E [0]+ NL[u])hasafixedpointinasmallballaroundtheorigininanappropriatefunctionspace.We choosethespace L2([0, π/m])andseekasolution u withinaballofradius ϵ =3 · 10 5

2.2.Invertingthelinearoperator

Adirectinversionoftheoperator L isnotfeasible,sinceitdependsverynonlinearly(andnon-locally) on R0.Instead,weuseacomputerassistedapproach.Themainideatoprovetheinvertibilityof L isto firstapproximate L by LF =Identity+FiniteRank,thenprovetheinvertibilityof LF andfinallyprovethat theapproximationerror L LF issmallenough,making L invertibleviaaNeumannseries.

Definition2.1. Let {en(x)}n bethenormalizedFourierbasisof Xm = L2 0, π m .Alsolet N =201,and EN =span{en}N n=1 bethesubspacegeneratedbythefirst N vectors.Similarly,let E ⊥ N beitsorthogonal subspace.Wewillalsodefine LF = I + KF : Xm → Xm where KF : EN → EN isgivenby KF [u]= π m 0 KF (x, y )u(y ) dy with KF (x, y ):= N k ,l =1 Ak ,l ek (x)el (y ), where Ak ,l isfiniteexplicitmatrixveryclosetotheprojectionoftheoperator K (x, y )tothe EN subspace.

TheproofofthenextlemmaiscomputerassistedandwillbeexplainedinSection 4

Lemma2.2. ThematrixI + Aisinvertibleandsatisfies ∥(I + A) 1∥2 ≤ C2,withC2 :=8.8

Lemma2.3. TheoperatorLF isinvertibleand ∥L 1 F ∥2 < C2 withC2 :=8.8

https://revistes.iec.cat/index.php/reports

Proof. Let PN betheprojectionoperatoronto EN .Usingthat KF =

, LF decouples in EN E ⊥ N as

=

thenas EN isafinitedimensionalvectorspace,toinvert IEN + A wehavetoinvertthecorresponding matrix,andtheidentityin E ⊥ N istriviallyinverted,so

Wecanconcludethat LF isinvertible.Moreover

henceitsnormisboundedby C2 becausebyLemma 2.2, ∥

TheproofofthenextlemmaisagaincomputerassistedanditwillbeexplainedinSection 4

Lemma2.4. TheerrorofapproximatingtheoperatorLbyLF satisfies ∥

Wecannowstateandprovethemainresultofthissubsection.

,withC3 :=0.085

Proposition2.5. ThelinearoperatorL : Xm → Xm isinvertible.Moreover ∥L 1∥2 ≤ C1 =35

Proof. Usingthat LF isinvertible,wecanwrite

Wecantheninvert I + L 1 F (L LF )usingaNeumannseriesbecauseduetoLemmas 2.3, 2.4 wehavethat

isalsoinvertible,wecanconcludethat L isinvertibleand

2.3.Solvingthefixedpointequation

ThegoalistouseBanachFixedPointtheoremtoprovetheexistenceofsolutions.Forthisweneedcontrol overtheLipschitznormof G .Wearegoingtostateapropositionthattogetherwiththeestimateson L 1 isgoingtoallowustoprovetheexistenceofafixedpoint.

Proposition2.6. Wehavethefollowingboundsonthenonlinearterms:

for ϵ0 :=3 · 10 7 andsomeexplicitpositiveconstantsC

Existenceofnon-convexrotatingvortexpatchestothe2DEulerequation

Theproofoftheboundontheerroroftheapproximatesolution,thefirstboundoflastlemma,is computerassisted.Therestofthemareestimatedbyhand.

Thenextpropositionisthemainoneofthissection.

Proposition2.7. TheoperatorG [u]:= L 1(E [0]+ NL[u]) hasafixedpointintheballofradius ϵ

Proof. Wecheckthattheoperatormapstheball Bϵ (0)intoitself:

andthatitisLipschitz

Inbothinequalitieswehavecheckedthattheexplicitvaluesoftheconstantssatisfythem.Weconclude byBanachFixedPointtheorem.

Oncewehaveproventheexistenceofthissolutionitispossibletocheckthatindeed R = R0 + x 0 ˜ u satisfiesequation(1).

3.Non-convexityandimprovedregularity

3.1.Proofofnon-convexity

Toprovethis,wefirstnotethatusingH¨olderinequalityandthebound

L2 ≤ ϵ wegetthefollowing enclosureforthesolution R(x)

Thequantitativecontrolwehaveonthesolutionallowsustorigorouslyproveitsnon-convexity.Let’s define P0, P1 thepointsontheboundaryofthepatchatangle0,2π/m.Wewillthenprovethat R(π/m) < P0 +P1 2 sothemidpointdoesnotbelongtothedomainofthepatch D

wheretheexplicitinequalityinthemiddleischeckedusingintervalarithmeticwiththecomputer.

3.2.Furtherregularity

Weinitiallyprovedthat R isan H 1 function.However,wecanshowthatitismuchsmoother.

https://revistes.iec.cat/index.php/reports

1. C ∞ Regularity:Weuseabootstrappingargument.ByrewritingtheV-stateequation,weshowthat ifthe k-thderivative ∂k R isbounded,then ∂k +1R isalsobounded,andbyinduction, R is C ∞.The argumentinvolvescarefullydifferentiatingthenon-localequationandcontrollingthesingularitiesin theintegralkernelstoobtainthefollowinginequality:

forasufficientlysmall δ> 0.

2. Analyticity:Toprovethattheboundaryisanalytic,werephrasetheproblemasafreeboundary ellipticproblemforthestreamfunction ψ.Thenwecanexplicitlycheckthatthegradientisnot vanishingintheboundary.Oncewehavethistheanalyticityisadirectconsequencedueto[8, Theorem3.1’].

4.Computerassistedestimates

Theproofofthemaintheoremhasstronglyreliedinusingthecomputertoperformdifficultcomputations. Wewillexplaininwhichstepsandhowwewereabletoobtaintherequiredbounds.Wereferto[4]fora thoroughoverviewofcomputerassistedproofsinPDEs.

Toperformrigorouscomputationswithacomputer,wecannotusefloating-pointarithmetic,asthe resultswouldnotberigorousbecausewehavenocontrolovertheroundingerror.Therefore,wemustuse intervalarithmetic.Intervalarithmetictreatsnumbersasintervals,wheretheintervalboundsarenumbers representablewithafixednumberofdigits.Inthisarithmetic,operationsaredefinedbetweenintervalsin suchawaythatforallnumbersbelongingtoaninputinterval,theresultoftheoperationisguaranteed tobeincludedinthefinaloutputinterval.Thismethodallowsforthepropagationoferrorstoultimately obtainanintervalthatcontainstheexactresult.

Weusedthe C++ library Arb forthesecomputations[7].Thekeycomputerassistedstepswere:

• Boundingthematrixinverse:Findingarigorousupperboundfor ∥(I + A) 1∥2 bycomputinga lowerboundfortheminimumeigenvalueofthesymmetricmatrix(I + A)T (I + A).

• Boundingintegrals:Usinghigh-orderquadratureruleswithrigorouserrorboundstocomputenorms offunctionsinvolvingdifficultintegrals,suchastheerroroftheapproximatesolution ∥E [0]∥L2 and theapproximationerrorofthelinearoperator

.Specialcarewastaken tohandlethelogarithmicsingularitiesintheintegrands.

• Verifyingfinalinequalities:CheckingtheconditionsfortheBanachFixedPointtheoremandthe finalnon-convexityinequalitybypluggingintherigorouslycomputedboundsforallconstants.

Thenumericsrequiredanontrivialamountofcomputingsize;forinstance,boundingtheerrorofthe approximatesolutiontotheorderof10 7 tookabout9hoursin64CPUs,andtheboundingoftheerror ofthekernelapproximationtook29hoursinalso64CPUs.

, 10 (2025),43–50;DOI:10.2436/20.2002.02.49.

Existenceofnon-convexrotatingvortexpatchestothe2DEulerequation

Acknowledgements

IwouldliketoexpressmysinceregratitudetoJavierG´omezSerranoforhisoutstandingsupervisionof myBachelor’sthesis,whichledtothisarticle.IalsowishtowholeheartedlythanktheDepartmentof MathematicsatBrownUniversityforhostingmeduringthedevelopmentofthisproject.

ThisarticleispublishedinconnectionwiththeEmmyNoetherPrize.Iamdeeplyhonoredtobea recipientandthankthejuryfortheirworkandconsideration.

IacknowledgetheCFISMobilityProgramforpartiallyfundingthisresearch,withspecialthankstothe Fundaci´oPrivadaMir-Puig,theCFISpartners,andthedonorsofthecrowdfundingprogram.Iwaspartially supportedbyaMOBINTgrantfromtheGeneralitatdeCatalunya,anErasmus+grantco-fundedbythe EuropeanUnion,andNSFgrantsDMS-2245017,DMS-2247537,andDMS-2434314.IalsothankBrown Universityforprovidingthecomputingfacilitiesusedinthisresearch.

References

[1]J.Burbea,Motionsofvortexpatches, Lett. Math.Phys. 6(1) (1982),1–16.

[2]G.Castro-L´opez,J.G´omez-Serrano,Existence ofanalyticnon-convexV-states, Comm.Math. Phys. 406(9) (2025),Paperno.217.

[3]G.S.Deem,N.J.Zabusky,Vortexwaves:stationary“Vstates,”interactions,recurrence, andbreaking, Phys.Rev.Lett. 40(13) (1978), 859–862.

[4]J.G´omez-Serrano,Computer-assistedproofsin PDE:asurvey, SeMAJ. 76(3) (2019),459–484.

[5]Z.Hassainia,N.Masmoudi,M.H.Wheeler, Globalbifurcationofrotatingvortexpatches, Comm.PureAppl.Math. 73(9) (2020),1933–1980.

[6]T.Hmidi,J.Mateu,J.Verdera,Boundaryregularityofrotatingvortexpatches, Arch.Ration. Mech.Anal. 209(1) (2013),171–208.

[7]F.Johansson,Arb:efficientarbitrary-precisionmidpoint-radiusintervalarithmetic, IEEE Trans.Comput. 66(8) (2017),1281–1292.

[8]D.Kinderlehrer,L.Nirenberg,Regularityinfree boundaryproblems, Ann.ScuolaNorm.Sup. PisaCl.Sci.(4) 4(2) (1977),373–391.

[9]G.Kirchhoff. Mechanik,Vorlesungen¨uber mathematischePhysik,Vol.1,B.G.Teubner, Leipzig,1874.

[10]A.J.Majda,A.L.Bertozzi, VorticityandIncompressibleFlow,CambridgeTextsAppl. Math. 27,CambridgeUniversityPress,Cambridge,2002.

https://revistes.iec.cat/index.php/reports

ANELECTRONICJOURNALOFTHE SOCIETATCATALANADEMATEM ` ATIQUES

PedroL´opezSancha UniversitatPolit`ecnica deCatalunya pedro.lopez.sancha@upc.edu

Resum (CAT)

Unaestrat`egiaperestudiarvarietatsalgebraiques´esconstruirinvariantsalgebraics quemesurinlessevessingularitats.Sobreelsnombrescomplexos,destaquenels idealsmultiplicadorsielsnombresdesalt.Encaracter´ısticapositiva,lesseves contrapartss´onelsidealsdetestiels F -nombresdesalt.Enaquestprojecte, calculemelsidealsdetesti F -nombresdesaltdecorbesplanesquasi-homog`enies, aix´ıcomdelessevesdeformacionsanombredeMilnorconstant,perunaquantitat infinitadecaracter´ıstiques p > 0.Enaquestscasos,veiemqueelsidealsdetest s´onlareducci´om`odul p delsidealsmultiplicadors.

Abstract (ENG)

Acommonapproachtostudyingalgebraicvarietiesisthroughalgebraicinvariants thatmeasuretheirsingularities.Overthecomplexnumbers,acelebratedexample ofsuchinvariantsincludethemultiplieridealsandthejumpingnumbers.Inpositive characteristic,theircounterpartsarethetestidealsand F -jumpingnumbers.Inthis work,wecomputethetestidealsand F -jumpingnumbersofquasi-homogeneous planecurves,aswellastheirone-monomialconstantMilnornumberdeformations, forinfinitelymanycharacteristics p > 0.Inthesecases,weseethatthetestideals arethemodulo p reductionofthemultiplierideals.

Keywords: testideals,F-jumpingnumbers,quasi-homogeneousplane curve,constantMilnornumberdeformations. MSC(2020): Primary13A35,14F10.Secondary14B05.

Received: July2,2025. Accepted: July18,2025.

51 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),51–61;DOI:10.2436/20.2002.02.50.

Idealsof pe -throotsofplanecurvesinpositivecharacteristic

1.Introduction

Achallengethatliesattheheartofmodernalgebraicgeometryistheclassificationofalgebraicvarieties which,inparticular,encompassesthecharacterizationoftheirsingularities.Themostcommonapproach tothisproblemistoconstructalgebraicandgeometricinvariantstoquantifythesingularities.

Inbirationalalgebraicgeometryoverthecomplexnumbers,ormoregenerally,overfieldsofcharacteristic zero,onecantakeadvantageoftheexistenceofaresolutionofsingularitiestoconstructinvariants.A celebratedexampleofsuchinvariantsisthefamilyofmultiplierideals.Givenahypersurfacedefinedbythe vanishinglocusofapolynomial f ,themultiplierideals J (f λ)formafamilyofidealsindexedbynonnegative realnumbers λ ∈ R≥0.Thesegiveadescendingchainofideals J (f λ) ⊇J (f µ)whenever λ ≤ µ,which inadditionisright-semicontinuous,meaningthat J (f λ)= J (f λ+ε)forsome ε> 0.Thevalues λ> 0 wherethechainjumps,thatis, J (f λ ε) ⊋ J (f λ)forarbitrarilysmall ε> 0,areknownasthejumping numbersof f .Thesmallestjumpingnumberamongthemiscalledthelog-canonicalthreshold.Bymeans oftheresolutionofsingularities,onecanshowthissetrationalanddiscrete.Referencesaremadeto[6].

Themultiplieridealsandjumpingnumbersencodethesingularitiesofthehypersurfacedetermined by f insubtleways.Suppose,forinstance,thatthevanishinglocusof f isacurve C inthecomplex planewithasingularityattheorigin.Ifonedeforms C intoanewcurve C ′ whilepreservingtheanalytic typeofthesingularity,thentheentirefamilyofmultiplieridealsremainsunchanged.Onthecontrary,if thedeformationonlypreservesthetopologicaltype,thenthejumpingnumbersstillagree,althoughthe multiplieridealsmay,ingeneral,differ.

Overfieldsofpositivecharacteristic p > 0,thereisnoresolutionofsingularitiesavailableforvarieties ofarbitraryfinitedimension.Inthissetting,theFrobeniusendomorphism,or p-thpowermap,servesasa substitutetool.Thetestideals,whichplaytheroleofthemultiplierideals,wereintroducedbyHochster andHunekeasanauxiliarytoolintightclosuretheory[5],andwerelaterrefinedbyHaraandYoshida[4]. WeshalladopttheconstructionofBlickle,Mustat¸˘a,andSmith(seeDefinition 2.9),whichgeneralizes earlierdefinitions[3].Inbrief,thetestideals τ (f λ)ofapolynomial f areanested,right-semicontinuous familyofidealsindexedoverthenonnegativerealnumbers λ ∈ R≥0.Thespotswherethechainoftest ideals“jumps”arethe F -jumpingnumbersof f ,thesmallestofwhichisthe F -purethreshold.

Itisawell-establishedfactduetoMustat¸˘a,Takagi,andWatanabe,thatif f isapolynomialdefined overtheintegers,thelog-canonicalthresholdof f canberecorveredfromthe F -purethresholdsofthe reductions fp of f moduloaprime p,as p →∞ [9].Aprofoundconjectureinarithmeticgeometry,the weakordinarityconjecture,proposesafurtherconnection,namely,thatthetestidealsofthereductions fp canbecalculatedbyreducingthemultiplieridealsmodulo p,forallprimesinaZariski-denseset[8].In thissense,thetheoriesofmultiplieridealsincharacteristiczeroandtestidealsinpositivecharacteristic arecloselyanalogous.

Testidealsand F -jumpingnumbersarenotoriouslydifficulttocompute.Inthefewcaseswhereexplicit descriptionsareknown—suchasellipticcurves,diagonalhypersurfaces,determinantalidealsofmaximal minors,oridealsinvariantundertheactionofasubgroupofalineargroup,thecalculationsrelyonthe arithmeticorcombinatorialpropertiesofthevariety.Anaiveyeteffectiveapproachtoobtain τ (f λ)isto calculatetheideals pe -throotsof f ,denoted Ce R · f n (seeDefinition 2.5),forafixedinteger e ≥ 0.As n ≥ 0rangesoverthenaturalnumbersintegers,oneobtainsdescendingchainofideals

e R · f 0 ⊇Ce R · f ⊇Ce R · f 2 ⊇···⊇C e R · f n ⊇Ce R · f n+1 ⊇··· which,inessence,containsallthetestidealsof f ,andcodifiesthe F -jumpingnumbers.

https://revistes.iec.cat/index.php/reports

Inthiswork,webeginbystudyingquasi-homogeneousplanecurves C overaperfectfield K ofcharacteristic p > 0,thatis,curvesin K 2 givenasthevanishinglocusofapolynomialoftheform f = x a + y b , with a, b ≥ 2.Forthese,weobservetheidealsof pe -throots,andconsequentlythetestidealsaremonomial idealsforsufficientlybigcharacteristics p ≫ 0.Todeterminethe F -jumpingnumbers,weposealinear integerprogrammingproblem,andprovideitssolution.

Wethenturntodeformations C ′ oftheoriginalcurve C ,whicharecurvesgivenasthezerosof polynomials g = f + i ti x αi y βi .Werestrictourselves,however,toone-monomialdeformations g = f +tx αy β .Fromthealgebro-geometricstandpoint,itismostnaturaltoconsiderdeformationsthatpreserve thesingularitytypeof C attheorigin,namely,constantMilnornumberdeformations.Inthissetting,we againfindthatthe pe -throotsandtestidealsaremonomial.

Forallsufficientlylargeprimessuchthat p ≡ 1(mod ab),wedescribeexplicitlythechainsof pe -th roots,testideals,and F -jumpingnumbers,andobservethattheycoincidefor f and g .Finally,wenote thatthetestidealsinthissettingariseasreductionsmodulo p ofthecorrespondingmultiplierideals.

2.Invariantsofsingularitiesincharacteristic p > 0

Throughout,let R denotearingofcharacteristic p > 0,andlet F : R → R, f → f p ,betheFrobeniusendomorphismof R.Foranonnegativeinteger e ≥ 0,the e-thiteratedFrobeniusistheendomorphism F e : R → R, f → f pe .Inthissectionweintroduceinvariantsofsingularitiesinpositivecharacteristic ofinteresttous.Often,thesearereferredtoas F -invariantsfortheyoriginatefromtheactionoftheFrobeniuson R

Restrictionofscalarsalong F e endows R withanexotic R-modulestructuredenoted F e ∗ R.Itselements arewrittenas F e ∗ x for x ∈ R.Asanabeliangroupwithrespecttoaddition, F e ∗ R isisomorphicto R

Theactionof R on F e ∗ R isgivenbyrestrictionofscalars: r F e ∗ x := F

(

ANoetherianring R ofcharacteristic p > 0issaidtobe F -finiteprovided F

x),for

R

∗ R isafinite R-modulefor some e ≥ 1(equiv.all e ≥ 1).

Example2.1. Let R = K [x1,..., xn]beapolynomialringoverafield K ofcharacteristic p > 0.If K is perfect,i.e.theFrobenius F : K ≃ −→ K isanautomorphismof K ,then F e ∗ R splitsas

R ≃

therefore R isan F -finitering,and F e ∗ R isafinitefreemodulewithbasis

Inthesequel,wewillrefertothisasthestandardbasisof F e ∗ R

2.1.Frobeniuspowersand pe epe -throotsofideals

Let R bearegular F -finiteringofcharacteristic p > 0.

Definition2.2. Let I beanidealof R.Foraninteger e ≥ 0,the e-thFrobeniuspowerof I istheideal I [pe ] = F e (I )R =(f pe | f ∈ I ).

, 10 (2025),51–61;DOI:10.2436/20.2002.02.50.

Idealsof pe -throotsofplanecurvesinpositivecharacteristic

Remark 2.3 Onechecksthatif I =(fλ | λ ∈ Λ)isageneratingsetfor I ,then I [pe ] =(f pe λ | λ ∈ Λ), hence I [pe ] ⊆ I pe .Thereversecontainmentholdswhen I =(f )isprincipal,thatis,(f )[pe ] =(f )pe ,thus Frobeniuspowersandregularpowerscoincideforprincipalideals.

Asortof“converse”operationtotheFrobeniuspoweristheidealof pe -throotsofanideal I .These wereintroducedin[1]intheprincipalcaseunderthenotation Ie (f ),andlateronexploitedin[3]togive analternativedefinitionofthetestideals(seeSection 2.2),usingthenotation I [1/p

]

Definition2.4. ACartieroperatoroflevel e ≥ 0isan R-linearmap F e ∗ R → R.ThesetofCartieroperators oflevel e ≥ 0hasanatural R-modulestructure,whichwedenoteby Ce R :=HomR (F e ∗ R, R).

Definition2.5. Let I beanidealof R.Foraninteger e ≥ 0,theidealof pe -throotsof I istheideal Ce R · I =(φ

Remark 2.6 If R isaregularand F -finite, Ce R · I ischaracterizedasthesmallestidealof R inthesenseof inclusionsuchthat I ⊆ (Ce R · I )[pe ]

Proposition2.7 ([1],[3,Proposition2.5]). SupposethatF e ∗ RisafreeR-modulewithbasisF e ∗ x1,..., F e ∗ xn.

ForanidealI =(f1,..., fm) ofR,let

betheexpressioninthebasisoffi ,i =1,..., m.Then Ce R · I =(fij | 1 ≤ i ≤ m,1 ≤ j ≤ n)

Wecollectbelowafewfactsabout pe -throotsthatwillbeusefullateron;foraproof,wereferthe readerto[3,Lemma2.4].

Lemma2.8. LetI,JbeidealsofR,andd , e ≥ 0 benonnegativeintegers.

(i)IfI ⊆ J,then Ce R · I ⊆Ce R · J.

(ii)OnehasthatJ · (Ce R · I )= Ce R · (I · J [pe ])

(iii)Onehasthat Ce R · I = Cd +e R · I [pd ].Inparticular,ifI =(f ) isprincipal,then

2.2.Testideals,

F F F -jumpingnumbers,and ν ν ν-invariants

Let R denoteregular F -finiteringofcharacteristic p > 0.Wenowintroducethegeneralizedtestidealsof anidealin R,asdefinedin[3],alongwiththeirassociatedinvariants.Throughout,wedenoteby ⌈x⌉ the ceilingofarealnumber.

Definition2.9. Fixanideal I of R.Thetestidealof I withexponent λ ∈ R≥0 is

Remark 2.10 Itcanbeshownthatthe pe -throotsappearingontheright-handsidegiveanascending chainofideals,whicheventuallystabilizesbecause R isNoetherian,therefore

Since pe -throotspreserveinclusions(Lemma 2.8),sodotestideals,thatis, τ (I λ) ⊇ τ (I µ)whenever λ ≤ µ.Itfollowsthetestidealsgiveadescendingfamilyofidealsobtainedas λ rangesoverthenonnegative realnumbers.Thischainisrightsemi-continuousinthefollowingsense:

Theorem2.11 ([9,Remark2.12],[3,Corollary2.16,Theorem3.1]). LetIbeanidealofR.

(i)Foreach λ ≥ 0,thereexists ε> 0 suchthat τ (I λ)= τ (I λ+ε)

(ii)Thereexistrealnumbers λ> 0 suchthat τ (I λ ε) ⊋ τ (I λ) forall ε> 0.

Definition2.12. Apositiverealnumber λ> 0isan F -jumpingnumberofanideal I of R provided τ (I λ ε) ⊋ τ (I λ),forall ε> 0.

The F -purethresholdof I ,writtenfpt(I ),isdefinedastheinfimumamongthe F -jumpingnumbersof I Itischaracterizedbyfpt(I )=sup {λ> 0 | τ (I λ)= R}

Theorem2.13 (Skoda’stheorem,[3,Proposition2.25]). SupposethatIisanidealgeneratedbynelements.Then τ (I λ)= I · τ (I λ 1) foreveryrealnumber λ ≥ n.

Theresultbelowshowsthatthelog-canonicalthresholdofapolynomialwithintegercoefficientscan berecoveredfromthe F -purethresholdsofthereductionsmodulo p:

Theorem2.14 ([9,Theorem3.4]). Letf ∈ C[x1,..., xn] beapolynomialwithrationalcoefficients.Fora primenumberp > 0,letfp ∈ Fp [x1,..., xn] bethereductionoffmodulop.Then

lct(f )=lim p→∞ fpt(fp ).

Itisconjecturedthatthestatementabovegeneralizes,inasense,tomultiplierandtestideals.Before announcingit,letusremarkthatif s isaZariski-closedpointinthespectrumofafinitelygenerated Z-algebra A,thatis, s isamaximalidealof A,thenthequotient A/sA haspositivecharacteristic.

Conjecture2.15 (Weakordinarityconjecture,[8,Conjecture1.2]) LetIbeanidealinthepolynomial ring C[x1,..., xn].SupposethatIcanbegeneratedbyelementsinafinitelygenerated Z-subalgebraA of C[x1,..., xn].ThenthereexistsaZariski-densesubsetSof Spec Aconsistingofclosedpointssuchthat J (I λ)s = τ (I λ s ), forall λ ≥ 0, foreverys ∈ S,whereI λ s , J (I λ)s denotetheimageunderA → A/sAofI λ,and J (I λ),respectively.

Closelyrelatedtothe F -jumpingnumbersarethe F -thresholdsintroducedin[9].Theirdefinitionis basedonadifferentfamilyofinvariants,namely,the ν-invariants,whichareofinterestbythemselves.

Definition2.16. Let I , J beidealsof R suchthat I ⊆ √J,where √J denotestheradicalof J.The ν-invariantoflevel e ≥ 0of I withrespectto J is

ν J I (pe )=max {n ≥ 0 | I n ⊆ J [pe ]}

Letusdenoteby ν• I (pe )thesetof ν-invariantsoflevel e ≥ 0of I

Reports@SCM, 10 (2025),51–61;DOI:10.2436/20.2002.02.50.

Idealsof pe -throotsofplanecurvesinpositivecharacteristic

Definition2.17. Let I , J beidealsof R suchthat I ⊆ √J.The F -thresholdof I withrespectto J is c J (I )=lim

p

Theorem2.18 ([3,Corollary2.30,Theorem3.1]). ThesetofF-jumpingnumbersofanidealIcoincides withthesetofF-thresholdsc J (I ) ofIobtainedasJrangesoveralltheidealsofRsatisfyingI ⊆ √J.In particular,bothsetsarerationalanddiscrete.

Incomputing F -thresholds,differentideals J, J ′ containing I intheirradicalmaygiveraisetothesame F -threshold.Instead,however,onecanlookatthespotswherethechainofidealsbelowjumps:

Definition2.19 ([10,Proposition4.2]) Thesetof ν-invariantsoflevel e ≥ 0ofanideal I of R is

2.3.Chainsof pepepe -throotsand rr -invariants

Throughthissection,let R = K [x1,..., xd ]denoteapolynomialringoverafield K ofcharacteristic p > 0.We areinterestedinobtainingthetestideals τ (f λ)ofapolynomial f ∈ R.Skoda’stheorem(seeTheorem 2.13) showsthat τ (f λ)=(f ) τ (f λ 1)foreveryrealnumber λ ≥ 1,henceitsufficestolookattestideals with0 <λ< 1.Inthiscase,byRemark 2.10,onehasthat τ (f λ)= Ce R · f r forsomeinteger r ≤ pe .Aside fromtestideals,wearekeenon pe -throots Ce R · f n.Bywriting n uniquelyas n = spe + r ,with s ≥ 0, 0 ≤ r < pe ,itfollowsfromLemma 2.8 that Ce R · f n =(f )s ·Ce R · f r .Altogether,thisshowsitisenoughto considertheidealsinchainsoftheform

Definition2.20. Wereferto(1)asthechainofidealsof pe -throotsof f

Lemma2.21. Forapolynomialf ∈ R,onehasthat ν• f (pe )=(ν• f (pe ) ∩ [0, pe ))+ pe Z≥0

Notation2.22. Ifu =(u1,..., ud ) ∈ Zd ≥0 isamulti-index,weletx u bethemonomialx u = x u1 1 x ud d .

Perhaps,themoststraightforwardwaytodetectajump Ce R · f r ⊋ Ce R · f r +1 inchain(1)istotest ifamonomial x u in Ce R · f r dropsfrom Ce R · f r +1,whichmeans r isa ν-invariantattachedto f andthe monomial x u .Whilethistechniqueisstandardinthefield,tothebestofourknowledge,theinvariant“r ” hasnotbeenassignedanameintheliterature.Wethereforeintroducethefollowingdefinition:

Definition2.23. Wedefinethe r -invariantoflevel e ≥ 0of f withrespecttoamonomial x u by r e R (f ; x u )=sup {n ∈ Z | x u ∈Ce R · f n}

Lemma2.24. Letf ∈ Rbeapolynomial.Supposethatfisnotamonomial.

(i)Forallmonomialsx u inR, 0 ≤ r e R (f , x u ) ≤ pe 1

(ii)Everyr-invariantoffisa ν-invariant.

https://revistes.iec.cat/index.php/reports

Wheneveryidealinthechainof pe -throotsof f ismonomial,theconversetoLemma 2.24 holds.In spiteofhowrestrictivethislatterconditionmayseem,wewillcomeacrossitinSection 3

Lemma2.25. Letf ∈ Rbeapolynomialthatisnotamonomial.Supposethateveryidealinthechain ofpe -throotsoffismonomial.

(i)Onehasthat ν• f (pe ) ∩ [0, pe )= {r e R (f ; x u ) | x u }

(ii)Foraninteger 0 ≤ n < pe ,onehasthat Ce R · f n =(x u | r e R (f ; x u ) ≤ n)

3.Idealsof p ep ep e -throotsofplanecurves

Inthissection,wedescribetheinvariantspreviouslyintroduced,forquasi-homogeneousplanecurvesdefined overperfectfieldsofcharacteristic p > 0,forinfinitelymanyprimes,andtheirone-monomialconstant Milnornumberdeformations.Throughout,let ⌊x⌋ and ⌈x⌉ bethefloorandceilfunctions,respectively.

Remark 3.1 Let R = K [x1,..., xd ]beapolynomialringoveraperfectfield K ofcharacteristic p > 0, so F e ∗ R isafree R-modulewithstandardbasis {F e ∗ x i1 1 x id d | 0 ≤ i1,..., id < pe } (Example 2.1).Given amonomial x u1 1 x ud d ,writeeachexponentuniquelyas ui = si pe + ri ,with si ≥ 0,0 ≤ ri < pe ,for i =1,..., d.Then

e ∗ (x u1 1 x ud n )= x s1 1 x sd n F e ∗ (x r1 1 x rd n )

isthebasisexpressionof x u1 1 ··· x ud d .Notethat si = ⌊ui /pe ⌋,and ri istheonlyinteger0 ≤ ri < pe with ui ≡ ri (mod pe ).Thiscalculationextendslinearlytopolynomialsof R

Definition3.2. Inthesettingabove,wesaythemonomials x u1 1 x ud d , x v1 1 x vd d appearwiththesame basiselementif F e ∗ x u1 1 ··· x ud d , F e ∗ x v1 1 ··· x vd d lieinthesamerank-onefree R-submoduleof F e ∗ R spannedby anelementofthestandardbasisof F e ∗ R.Thisisequivalentto ui ≡ vi (mod pe ),for i =1,..., d

3.1.Idealsof

pe epe -throotsofquasi-homogeneousplanecurves

Definition3.3. Let K beafield.Aquasi-homogeneousplanecurvedefinedover K isthevanishinglocus in K 2 ofapolynomialoftheform f = x a + y b ,where a, b ≥ 2.Forsimplicity,wewillrefertothe binomial f = x a + y b asthequasi-homogeneousplanecurve.

Fromnowon,weworkoverthepolynomialring R = K [x, y ],with K perfectofcharacteristic p > 0. Weremark,however,thatallresultsremainvaliduponrelaxingtheassumptionon K tomerelybeing F -finite,i.e.that K /K pe beafiniteextensionforsome e > 0(equiv.all e > 0).

Proposition3.4. Letf = x a + y b beaquasi-homogeneousplanecurve.Supposethatpdoesnotdividea orb.Foreveryinteger 0 ≤ n < pe ,onehasthat

Ce R · f n = x ⌊ai /pe ⌋y ⌊bj /pe ⌋ i + j = nand n i, j ̸≡ 0(mod p)

Inparticular,everyidealinthechainofpe -throotsoffismonomial.

, 10 (2025),51–61;DOI:10.2436/20.2002.02.50.

Idealsof pe -throotsofplanecurvesinpositivecharacteristic

Remark 3.5 InviewofProposition 3.4, Ce R · f n containsthemonomial x u y v ,ifandonlyifthereexistsa pair(i , j)ofnonnegativeintegerssuchthat:

,and

Thefirsttwoconditionsareequivalentto ai ≤ (u +1)pe 1,and bj ≤ (v +1)pe 1,respectively.Asa result,the r -invariant r e R (f ; x u y v )isthesolutiontothefollowinglinearintegerprogrammingproblem:

Asolution ≥ pe hasnomeaningbyLemma 2.24,henceitshouldbethoughtofas x u y v ∈Ce R ·

Onecangivegeneralboundsfortheoptimumof(P1).Toobtainasolution,however,itishelpfulto makeassumptionsonthecongruenceclassof p modulo ab.Inwhatfollows,weprovidethesolutionunder suchadditionalassumption.Lateron,inSection 3.3,westudytheconsequencesonthe F -invariants.

Lemma3.6. Letf = x a + y b beaquasi-homogeneousplanecurve.Supposethatp ≡ 1(mod ab).The r-invariantr e R (f ; x u y v ) ofamonomialx u y v is

ifbu + av < ab a b, pe 1 ifbu + av ≥ ab a b

3.2.One-monomialdeformationsofquasi-homogeneousplanecurves

Let C ⊆ K 2 beaplanecurvegivenasthezerolocusofapolynomial h ∈ K [x, y ].Suppose C passesthrough theorigin.ThentheMilnornumberof C isdefinedas µ =dimK K [x, y ]/(∂x h, ∂y h).When K = C,and h isaquasi-homogeneousplanecurve x a + y b ,theMilnornumberdeterminesthetopologicaltypeofthe singularityof C attheoriginunderdeformations,inthefollowingsense:

Theorem3.7 ([7]). Letf = x a + y b beaquasi-homogeneousplanecurvedefinedover C,andga deformationg = x a + y b + t1x α1 y β1 + + tnx αn y βn ,witht1,..., tn ∈ C.Supposeeverydeformation monomialx αi y βi onwhichgissupported(i.e.ti =0)satisfies 0 ≤ αi < a 1, 0 ≤ βi < b 1,and aβi + bαi > ab.ThenfandghavethesameMilnornumber.

Definition3.8. AconstantMilnornumberdeformation,or µ-constantdeformationofaquasi-homogeneous planecurve f = x a + y b definedover K ,isthevanishinglocusin K 2 ofapolynomialoftheform g = x a + y b + n i =1 ti x αi y βi ,where ti ∈ K , with0 ≤ αi < a 1,0 ≤ βi < b 1,and aβi + bαi > ab,for i =1,..., n

https://revistes.iec.cat/index.php/reports

(P1)

PedroL´opezSancha

Hereinafter,weconsiderone-monomial µ-constantdeformationsofquasi-homogeneousplanecurves definedoveraperfectfield K ofcharacteristic p > 0,thusweworkoverthepolynomialring R = K [x, y ].

Proposition3.9. Letg = x a + y b + tx αy β bea µ-constantdeformationoff = x a + y b .Supposethat pdoesnotdivideaorb,andp > aβ + bα ab.Foreveryinteger 0 ≤ n < pe ,onehasthat Ce R · g n = x ⌊(ai +α

/pe ⌋ i + j + k = nand n i, j, k ̸≡ 0(mod p)

Proposition3.10. Letg = x a + y b + tx αy β bea µ-constantdeformationofthequasi-homogeneous planecurvef = x a + y b .Supposethatpdoesnotdivideaorb,andp > aβ + bα ab.Onehasthat Ce R · f n ⊆Ce R · g n foreveryinteger 0 ≤ n < pe

Remark 3.11 ByProposition 3.9,givenaninteger0 ≤ n < pe ,amonomial x u y v isin Ce R · g n ifandonly ifthereexistsatriple(i, j, k)ofnonnegativeintegerssuchthat: ai + αk pe ≤ u,and bj + βk pe ≤ v ,and n i, j, k ̸≡ 0(mod p),and i + j + k = n

Oneseesthefirsttwoconditionsareequivalentto ai + αk ≤ (u +1)pe 1,and bj + βk ≤ (v +1)pe 1. Itfollowsthat r e R (g ; x u y v )isthesolutiontothefollowinglinearintegerprogrammingproblem:

maximize: i + j + k, subjectto: ai + αk ≤ (u +1)pe 1, bj + βk ≤ (v +1)pe 1, i +j +k i ,j ,k ̸≡ 0(mod p), i, j, k ∈ Z≥0

ByProposition 3.10,asolutionof(P2)isboundedbelowbyasolutionof(P1).AsinRemark 3.5,a solution ≥ pe mustbethoughtofas x u y v ∈Ce R · g pe 1

Lemma3.12. Letg = x a + y b + tx αy β bea µ-constantdeformationofaquasi-homogeneousplane curvef = x a + y b .Supposethatp ≡ 1(mod ab).Foramonomialx u y v ,onehasthat r e R (f ; x u y v )= r e R (g ; x u y v ).

3.3. F F F -invariantsofquasi-homogeneousplanecurvesanddeformations

Toconclude,weusethe r -invariantsofquasi-homogeneousplanecurvesandtheirone-monomial µ-constant deformationstocomputetheir F -invariantswhen p ≡ 1(mod ab).

Proposition3.13. Lethbeeitherthequasi-homogeneousplanecurvef = x a + y b ,orthe µ-constant deformationg = x a + y b + tx αy β .Supposethatp ≡ 1(mod ab)

, 10 (2025),51–61;DOI:10.2436/20.2002.02.50.

(P2)

Idealsof pe -throotsofplanecurvesinpositivecharacteristic

(i)Thepe -throotofhn,with 0 ≤ n < pe ,is

e R · hn = x u y v u +1 a + v +1 b ≤ n pe 1

(ii)The ν-invariantsofhofleveleare

Theorem3.14. Lethbeeitherthequasi-homogeneousplanecurvef = x a + y b ,orthe µ-constant deformationg = x a + y b + tx αy β .Supposethatp ≡ 1(mod ab),andp > aβ + bα ab.

(i)TheF-jumpingnumbersofhare

(ii)Thetestidealofhwithexponent λ ∈ (0,1) is

Remark 3.15 Foraprimenumber p sufficientlylargewith p ≡ 1(mod ab),Proposition 3.13 andTheorem 3.14 showthataquasi-homogenousplanecurveandaone-monomial µ-constantdeformationhavethe samechainsof pe -throots, ν-invariants,and F -jumpingnumbers.Furthermore,theirtestidealscoincide for λ ∈ (0,1).

Remark 3.16 Let f = x a + y b beaquasi-homogeneousplanecurve,oraone-monomial µ-constantdeformation g = x a +y b +tx αy β , t ∈ Z,definedover C.Considerthesets Xf = {p | p ≡ 1(mod ab), p prime}, Xg = {p | p ≡ 1(mod ab), p > aβ + bα ab, p prime} assubspacesofSpec Z,whichareinfiniteby Dirichlet’stheoremonarithmeticprogressions.IntheZariskitopologyon Z,anonemptyopensubsetis thecomplementoftheunionoffinitelymanypoints,andhencemustintersectsboth Xf and Xg .Itfollows thatthesesetsaredenseinSpec Z

Chooseaprime p ∈Xf ,anddenoteby fp thereductionmodulo p of f along Z[x, y ] → Fp [x, y ]. Similarly,let gp bethereductionof g ,with p ∈Xg .Themultiplierideals J (f λ), J (g λ)aregeneratedby polynomialsover Z,andcanbeobtainedwithanalgorithmproposedbyBlancoandDachs-Cadefau[2]. Aftercomputingtheirreductions J (f λ)p , J (g λ)p ,oneseestheycoincidewiththetestideals,namely J (f λ)p = τ (f λ p ),and J (g λ)p = τ (g λ p ),forall λ.Inconsequence,theweakordinarityconjecture(Conjecture 2.15)holdsforquasi-homogeneousplanecurvesandtheirone-monomial µ-constantdeformations.

Acknowledgements

Theauthorwouldliketoexpresshissinceregratitudetohisadvisor,Josep ` AlvarezMontaner,andtothe master’sthesiscommitteeforevaluatingtheprojectfromwhichthisworkoriginates.Healsowishesto thanktheInstitutd’EstudisCatalansand,inparticular,theSocietatCatalanadeMatem`atiquesandthe jurymembersforawardingthe ´ EvaristeGaloisPrizetothemaster’sthesisproject.Finally,theauthoris gratefultotheeditorsandtheanonymousrefereefortheirvaluablefeedbackandsuggestions.

https://revistes.iec.cat/index.php/reports

References

[1]J. ` Alvarez-Montaner,M.Blickle,G.Lyubeznik, Generatorsof D-modulesinpositivecharacteristic, Math.Res.Lett. 12(4) (2005),459–473.

[2]G.Blanco,F.Dachs-Cadefau,Computingmultiplieridealsinsmoothsurfaces,in: Extended AbstractsFebruary2016—PositivityandValuations,TrendsMath.Res.Perspect.CRM Barc. 9,Birkh¨auser/Springer,Cham,2018, pp.57–63.

[3]M.Blickle,M.Mustat¸ˇa,K.E.Smith,Discretenessandrationalityof F -thresholds,Special volumeinhonorofMelvinHochster, Michigan Math.J. 57 (2008),43–61.

[4]N.Hara,Ken-IchiYoshida,Ageneralizationof tightclosureandmultiplierideals, Trans.Amer. Math.Soc. 355(8) (2003),3143–3174.

[5]M.Hochster,C.Huneke,Tightclosure,invarianttheory,andtheBrian¸con–Skodatheorem, J.Amer.Math.Soc. 3(1) (1990),31–116.

[6]R.Lazarsfeld, PositivityinAlgebraicGeometry II.PositivityforVectorBundles,andMultiplier

Ideals,Ergeb.Math.Grenzgeb.(3) 49 [Results inMathematicsandRelatedAreas.3rdSeries. ASeriesofModernSurveysinMathematics], Springer-Verlag,Berlin,2004.

[7]LˆeD˜ungTr´ang,C.P.Ramanujam,TheinvarianceofMilnor’snumberimpliestheinvariance ofthetopologicaltype, Amer.J.Math. 98(1) (1976),67–78.

[8]M.Mustat¸˘a,V.Srinivas,Ordinaryvarietiesand thecomparisonbetweenmultiplieridealsand testideals, NagoyaMath.J. 204 (2011),125–157.

[9]M.Mustat¸˘a,S.Takagi,Kei-ichiWatanabe, F-thresholdsandBernstein–Satopolynomials, in: EuropeanCongressofMathematics,EuropeanMathematicalSociety(EMS),Z¨urich, 2005,pp.341–364.

[10]E.Quinlan-Gallego,Bernstein–Satotheoryfor arbitraryidealsinpositivecharacteristic, Trans. Amer.Math.Soc. 374(3) (2021),1623–1660.

, 10 (2025),51–61;DOI:10.2436/20.2002.02.50.

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

MiguelNasarreBudi˜no UniversitatAut`onoma deBarcelona miguel.nasarre.budino@gmail.com

Resum (CAT)

Primerenscentremaestendrelanoci´odes`eriedeFourier,estudiantcompodem representarfuncionsdequadratintegrablesobrevarietatsdeRiemann.Perferaix`o ensajudemdelteoremadeHodge,queenspermetr`atrobarbasesd’aquestsespais apartirdellaplaci`a.

Despr´esveiemcomaquestm`etodeespotutilitzarpertrobarless`eriesdeFourier perafuncionsperi`odiques,iperalesfuncions L2(S2),elcasprincipal.Tamb´e discutimcomestimarl’errorennorma L2,iimplementemtoteslesf´ormulesquees trobenal’articleaunprogramaperpodervisualitzarelsresultatsobtinguts.

Abstract (ENG)

FirstwewillfocusinextendingthenotionofFourierseries,studyinghowcanwe representfunctionsofintegrablesquareoverRiemannianmanifolds.Todothiswe willusetheHodgetheorem,thatwillallowustofindbasisofthesespacesthrough theLaplacian.

ThenwewillseehowthismethodcanbeusedtofindtheFourierseriesforperiodic functions,andforthefunctions L2(S2),ourmaincaseofstudy.Wewillalsodiscuss howtoestimatethe L2-error,andweimplementalltheformulasfoundinthearticle inaprogramtobeabletovisualizetheobtainedresults.

Keywords: Laplacian,Riemannianmanifold,basis,sphericalharmonics, Fourierseries,L2-errorestimates.

MSC(2020): Primary55M25,57P10.Secondary55P15,57R19,57N15.

Received: July29,2025.

Accepted: August25,2025.

63 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),63–74;DOI:10.2436/20.2002.02.51.

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

1.Introduction

Thisarticlewillfollowthestepsthatweresetinmybachelor’sthesisandexplainhowtobuildFourier seriestobeabletorepresentstar-shapedsurfaces.

InparticularwewilllookattheconstructionoftheLaplacianoperator∆inducedbyametric g ,on asetoffunctions L2(X ),where X isaRiemannianmanifold,andhowwecanusetheHodgetheoremto lookforbasisofthefunctionspace {ψi }i ∈Λ ∈ L2(X )intheeigenvectorsoftheLaplacian∆ψi = λi ψi

Wewillprimarilyfocusourattentiononthecasefor S2,wheresuchbasiswillbegivenbythespherical harmonics,andwewillstudyhowtoapplytheseformulasforcomputationaluse,andalsohowtoapriori estimatethe L2-errorforacertainamountofcoefficients.

Allofthisculminatinginadesktopapplication,thatgivenastar-shapedsurfacetriangulationwilluse theformulasexplainedonthearticletorepresentthatshapeasaFourierseries,provingtheapplicability ofthemathematicsexploredonmythesis.

2.ConstructionoftheLaplacianonaRiemannian manifold

Given(X , g )acompactRiemannianmanifold,forfurthercalculationswewillassume

aparameterizationwithadenseimage, Φ(U)= X

Definition2.1. GiventhepriormanifoldwedefinetheHilbertspace L2(X )= {f : X −→ R, f integrablesquare} withthescalarproductfor f , h ∈ C ∞(X ), ⟨

where dVg isthedifferentialinducedbythemetric.

TobuildtheLaplacianwewillworkwiththedensesubset C ∞(X ) ⊂ L2(X ).

Definition2.2. LetΩ0(X )= C ∞(X )bethespaceofdifferential0-formsof X .Wedefine Ω1(X )= {ω = f1 dx1 + + fn dxn | fi ∈ Ω0(X )} thespaceofdifferential1-formsof X https://revistes.iec.cat/index.php/reports

Definition2.3. Wedefinethedifferentialoperator d as d :Ω0(X ) −→ Ω1(X )

WeintendtousethedifferentialoperatortobuildtheLaplacianasaself-adjointoperator,∆= d ∗ ◦ d ThereforeweneedtoalsobuildaHilbertspacewithawelldefinedscalarproductonΩ1(X ).

Definition2.4. Let L2(X )= A0(X ),thenwedefine A1(X )= {ω = f1 dx1 + + fn dxn | fi ∈ A0(X )}

thesetof L2 1-formsof X ,andwesee Ω1(X )= A1(X ).

Nowwewanttobuildthetensor g ∗ : T ∗X ⊕ T ∗X → R

Let e1,..., en ∈ Tx X beanorthonormalbasisfor g ,thenthedualbasisofthe1-forms ω1,..., ωn ∈ T ∗ x X sothat ωi (ej )= δij willbeanorthonormalbasisfor g ∗ of T ∗X .Throughthisbasis g ∗ iswelldefined.

Definition2.5. Withwhatwehaveseensofarwecanintroducethenthescalarproductin A1(X )with thistensor.Given ω, η ∈ Ω1(X ),

(x ), η(x

ThisscalarproductsatisfiesthepropertiesofaHilbertspace.Therefore,wehaveanotherHilbertspace in A1(X ).

Nowwecanusethedefinitionoftheadjointoperatortofindtheadjointofthedifferential,andfinally buildtheLaplacianoperator.

3.TheHodgetheorem

Nowthatwehaveaself-adjointoperatorinsideaHilbertspace,wewillbeusingtheHodgetheorem,specific forthiscase,tobackupourclaimthatwecanfindbasisforthisfunctionspaceusingtheLaplacian.

Inthisarticlewewilljustannouncethetheoremwithoutfurtherproof.Foraproperproofaswellas moreindepthinformationpleasereferto[5,p.32],wheretheHodgetheoremisestablishedandproven. Theorem3.1. Let(M,g)beacompactRiemannianmanifold,orientedandconnected.Thereexistsan orthonormalbasisforL2(M, g ) thatconsistsofeigenvectorsoftheLaplacian.Alltheeigenvaluesarereal andpositive,exceptfor 0 whichisaneigenvalueofmultiplicity 1.Everyeigenvaluehasfinitemultiplicity, andtheyonlyaccumulateatinfinity.

Oncewefindthoseeigenvectors {ψi }i ∈Λ ∈ L2(X )thatformabasis,wecanwriteanygivenelement f ∈ L2(X )asitsFourierseries

(1) wherethecoefficientsaretheonesobtainedthroughthescalarproduct

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

4.ConstructionofthecommonFourierseries

Tobeginwiththeexamplesletusfocusonthecasefor X = S1,withtheparameterization

Φ:(0,2π) −→ S1 ⊂ R2 x −→ (cos x ,sin x ).

Wewillbeusingthemetric

,givingusthefollowingscalarproduct,given

Andfollowingthedefinitionfor g ∗ weobtainthescalarproductfor A1(S1).Given

), with ω(x )= f (x ) dx and η(x )= h(

Thenapplyingthedefinitionoftheadjointoperatorweobtain,givenany f ∈ Ω0(S1)andany ω ∈ Ω1(S1)with ω = h(x ) dx ,

IntegratingbypartsandaftersomealgebraweendupobtainingtheexpressionfortheLaplacian

Nowwewanttolookfortheeigenvectorsofthisoperator.Thereforewewanttoobtainasetof orthonormalfunctions {ψi }i ∈Λ ∈ L2(S1)thatsatisfytheequation∆ψi = λi ψi .Ifwedeveloptheprior expression,wesee

givingustheverynaturalorthonormaleigenvectors

0(x )=1, ψn(x )= √2cos(nx ), ϕn(x )= √2sin(nx ).

Aswecaneasilyprovebysolvingthedifferentialequation,theseareallthesolutionsdespitelineal combinationsofthem.Therefore,nowasdescribedinthepriorsectionwecancreatetheFourierseries https://revistes.iec.cat/index.php/reports

usingthisbasis,givingusthecommonexampleforrealFourierseriesweallknow.Givenany f ∈ L2(S1), wecanwriteitas

Wecanalsoconsiderthecasefor L2(S1 , C).Inthisspacealmostalltheprocessisexactlythesame, andweobtainthesameself-adjointoperator.Theonlydifferenceisthatwhendefiningthescalarproduct, thisonerequiresasymmetrybytheconjugate,thereforewedefinegiven f ,

L2(S1 , C),

Forthisparticularexample C ∞(S1 , C),themostcommonbasiswillbe

Finally,givenany f ∈ L2(S1 , C),wecanwriteitas

5.Casefor L2L2L2( ( (S2)2)2),thesphericalharmonics

Tostartwiththiscasewewillalsostartdefiningtheparameterization,

)

ThroughthisparameterizationwecanfindaRiemannianmetricinducedfrom R3,givingus

Thereforethescalarproductwillbe

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

Andfollowingthedefinitionfor g ∗ weobtainthescalarproductfor A1(S2),given

weobtain

Applyingthedefinitionoftheadjointoperatorweobtain,given

AftersomecalculationsweendupobtainingthefollowingexpressionfortheLaplacian,

Atthispointweintroducethesphericalharmonics,whicharefunctionswiththefollowingexpression.

Definition5.1. Let ℓ ∈ N∪{0} and m ∈ Z with |m|≤ ℓ.Wedefinethefunction Y m ℓ : S2 → R parameterized withthecoordinateswehavebeenusingsofaras

(2ℓ +1)(ℓ m)!

π(ℓ +

where P m ℓ aretheassociatedLegendrepolynomials.Theyaredefinedasthecanonicalsolutionsofthe generalLegendreequation

ThesefunctionsaretheorthonormalbasisobtainedbytheeigenvectorsoftheLaplacian.Unfortunately giventhelengthconstrainsofthisarticletheproperproofofsuchclaimisnotpossible.Fortheinterested readerIstronglyrecommendcheckingthefullthesisforaveryinterestingproofaswellasamoreindepth explanationofallthestepswehavedonebeforeandalltheresultswehaveseen.

NeverthelessIwilltrytogiveasomewhatsatisfyingoverviewoftheprooffoundinthethesis.Firstwe startbyprovingthatthesphericalharmonicsareorthonormal;forthisandfurtherstepsaheadweuse[4, Chapter14],[1,pp.331–341],wherealotofthepropertiesoftheLegendrepolynomialsarefoundaswellas somerecurrencesusefulfortheircomputation.InthiscaseweusetheLegendrepolynomialsorthogonality,

Whichfinallyendsuphelpingusfindtheorthonormalityresult

ThenweevaluatetheLaplacianofthesphericalharmonics.Tosimplifytheexpressionweusethe substitution v =cos θ,where v ′ = sin θ = √1 v 2;withsomerearrangementwecanuse(4)to getridofoneofthederivatives,allowingustosimplifytheexpressionandwithsomealgebraweobtain ∆Y m ℓ = ℓ(ℓ +1)Y m ℓ ,whichprovesthatthesphericalharmonicsareeigenvectorsoftheLaplacianwith eigenvalues λℓ = ℓ(ℓ +1).

Finallywehavetoprovethattheseeigenvaluesaresufficienttomakeabasisfor L2(S2).Forthisfinal stepweinspireonthebeautifulprooffoundinChapter7ofthebook Groupesetsym´etries [2].Webegin withtheStone–Weierstrasstheorem.

Theorem5.2. LetXbeacompactHausdorffspaceandAasub-algebraofthespaceofcontinuous functionsfromXtotherealnumbersC 0(X ),thathasaconstantnon-zerofunction.ThenAisdense inC 0(X ) ifandonlyifitseparatespoints.

Thistellsusthatthepolynomialsaredensefor C 0(X )aslongas X isacompactHausdorffspace. Thereforetheproofcentersitselfinprovingthattheharmonicpolynomialsareabasisfortherestrictionto S2 ofallpolynomialsin R3,morespecificallythereareatotalof2ℓ +1homogeneousharmonicpolynomials ofdegree ℓ,andalltogethertheyformabasisofthepolynomialsin R3 restrictedtothesphere.And thereforeapplyingtheStone–Weierstrasstheoremtheyareabasisforallcontinuousfunctionsrestricted tothesphere.

Thesecondpartoftheproofusestheresultsweobtainedearlierfromthesphericalharmonicstoprove thattheseareindeedtherestrictionto S2 ofhomogeneousharmonicpolynomialsofdegree ℓ,andsincewe have2ℓ +1sphericalharmonicsforevery ℓ value,weconcludethattheseareabasisfor L2(S2).

6.Aprioriestimationsofthe

L2L2L2-error

Itwouldbeinterestingforfurtherapplicationsoftheformulaswehaveseensofarifwehadanapriori estimateofhowmanycoefficientsweneedtocomputetoobtainadesiredrelative L2-error.Forasubset offrequencies L ⊂ Λwedefinethiserroras

=

f || , where f L = i ∈L⟨f , ψi ⟩ψi

Todothiswewillbeusingthemethodsfoundin[3].Thispaperwasdevelopedbymytutorandisthe initialinspirationbehindthethesis,forproofofthetheoremsthatwillfollowIstronglyrecommendlooking attheirarticle,oritcanalsobefoundinthethesis.

Theorem6.1. Letf ∈ A0(X ) beafunctionsothatdf ∈ A1(X ) and ∆f ∈ A0(X ) arewelldefined.For every ϵ> 0 existsafinitesubsetLf (ϵ) ⊂ Λ thatonlydependsof ||f ||, ||df ||, ||∆f || and ϵ sothat

Actually,Lf (ϵ) canbechosenbythepreimage λ :Λ → R ofthecompactinterval [Lf (ϵ), L+ f (ϵ)],where L

, 10 (2025),63–74;DOI:10.2436/20.2002.02.51.

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

Theorem6.2. AssumewealreadycomputedtheFouriercoefficientsoffforagivensubsetI ⊂ Λ.Then theinequality (5) isalsosatisfiedforthenewsubsetofcoefficients

TheseformulaswillbeusedintheprogramsthatcalculateanddisplaytheFouriercoefficients,inorder toseetheirperformanceonarealscenario,andtheresultswillbediscussedattheend.

7.Formulasforcomputationaluse

InthissectionwewillexplaintheformulasimplementedintheprogramtocalculatetheFourierseries forthecase L2(S2).Forourparticularcasewewillsupposethatwehaveatriangulation {Ti }N i =0 ofa star-shapedsurface S ofstrictlypositiveradius.Wewillexpressthissurfacebyitsradius r : S2 → R+ given anypointofthesphere.TocomputetheFouriercoefficientsforthisfunctionwewillneedtoapply(2), andtheexpressionwillbeasfollows,

where T p i ⊂ S2 isthetriangle Ti projectedontotheunitsphere.Ifthetriangulationisfineenough,wecan approximatethat Y m ℓ isconstantovertheentiresurfaceofthetriangle,andsincethemeanradiusofthe triangleistheradiusinthebarycenter,wewillrewritethepriorexpressionas

where T i isthebarycenterofthetriangle Ti ,and A(T p i )istheareaofthesphericaltriangle T p i Nowwewilllookathowtoevaluateeverytermofthisformula.Startingwith ||T ||,thisisjustthe normofthemeanofthethree v0, v1, v2 ∈ R3 pointsthatmakethetriangle, ||T || = || v0 +v1 +v2 3 ||

Thentocalculatetheareaofthesphericaltriangle A(T p )wewillusetheformula A = α0 + α1 + α2 π, where αi aretheanglesofthesphericaltriangle.Tofindtheseangleswecanusethetangentlinestothe sphere u = (vi ×vj )×vi ||(vi ×vj )×vi || ,andfindtheanglesthroughthescalarproductbetweenthem.

Finallywehavetocomputethesphericalharmonicatthebarycenteroftheprojectedtriangle.Inthis casewewillexplainhowtheprogramcomputes(3)foranygivenpoint(x , y , z ) ∈ S2 ⊂ R3.Tocompute thetrigonometricfunctionscos(mφ)andsin(mφ)wewillusethefactthatweknowcos φ = x √1 z 2 and sin φ = y √1 z 2 andtheChebyshevpolynomial,thatgiveus

Tn(cos θ)=cos(nθ), Un 1(cos θ)sin θ =sin(nθ).

https://revistes.iec.cat/index.php/reports

Tocomputethemwewillusethefollowingrecurrences,

0(x )=1,

1(x )= x ,

ForfurtherinformationabouttheChebyshevpolynomialsanditsrecurrencesyoucancheck[1,Chapter22],or[4,Chapter18].

TocomputetheassociatedLegendrepolynomials,wewillusethefollowingrecurrences,whichcanbe foundinthecitedreferences,

Morepreciselywewilluse(8)tocomputeuntil P m m ,then(9)tocompute P m m+1,andfinally(10)tocompute until P m ℓ ,usingthecos θ = z thatwealreadyknow.

NowwehaveaprecisewayofcalculatingeverytermoftheFourierseries,sincewejustneedtorepeat thesecalculationsforeverytriangleofourtriangulationandwewillgettheFouriercoefficient,andthis wayishowithasbeenimplementedinourprogram.Finallytoplotthesurfacewejustapply(1),

Itisalsointerestingtoseehowwecomputetheformulafortheestimated L2-errorgiventhetriangulation.Asseenintheformulas(6)and(7)allweneedtocomputearethenormssquared.Forthefirst normwewillusethefollowingformula,

Forthenormofthedifferentialwewillusethescalarproductasdescribedgivingusthefollowing expression,

whichcanbeapproximatedasshowninthethesisbythefollowingsum,

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

WearejustmissinganexpressionfortheLaplaciannorm.Inthiscasewewillbeusingtheformula discussedin[6],thiswillgiveusavaluefortheLaplacianineveryvertexofthetriangulation.

Let {vi }V i =1 bethesetofvertexofourtriangulation,wedenote {pi }V i =1 theirprojectionsontheunit spheresothat pi = v p i .Foreveryvertex pi with n neighbors,wedenote N(i )= i1,..., in thesetofindex oftheneighborvertexof pi .Wewillassumetheyareorderedcounterclockwiseasseenfromoutsidethe sphereabove pi .ThenwecanapproximatetheLaplacianinthisspotas

where αij and βij aretheanglesoftheadjacenttrianglestothesegment

Finallysincewewanttocompute ||∆r ||.Wewillconsiderthateachoneofourvertextakesaregion s(pi ) ⊂ S2 equivalenttoonethirdoftheareaofthesphericaltrianglessurroundingit.Thereforethe areaofallthetriangleswillbeequallysharedbetweenthevertex,andweobtaintheexpression

8.Conclusionandresults

Asmentionedpreviouslythisthesiswasbuiltaroundtheideaofcreatingaprogramthatcanreplicateall theformulasexplainedhereandcanshowsomeinterestingresults.Fortheinterestedreaderthisprogram doesexistandcanbefoundat https://github.com/MiquelNasarre/FourierS2

Itissatisfyingtoseethatalltheseformulasactuallyworkandcanproducesomeinterestingresults.This canbeseeninFigure 3,thatshowsthesphericalharmonicsasdepictedbytheprogram,andFigure 1,that showsabasicexampleoftheprogram’sfunctionality.

Figure1:Programtryingtorecreateacubetriangulationwith ℓ from0to4.

Alsothisprogramallowsustoseesomeclearlimitationsoftheformulas.Forexampleifyoutrytogo toodeepandyourtriangulationisnotfineenoughtheapproximationstocalculatethecoefficientswillnot beasgood,givingyousomeweirdlookingshapesintheprocess,asseeninthemiddleshapeofFigure 2. Thislimitationthoughcanbeeasilysolvedbydividingthetrianglesinthetriangulation,asshownbythe lastshapeofFigure 2,wheretheshapeisvisiblybetterdefinedandthe L2-errorislower.

https://revistes.iec.cat/index.php/reports

MiguelNasarreBudi˜no

Figure 2:ProgramcreatingtheFourierseries ℓ ≤ 20of example.dat withoutthesubdivisionandwith foursubdivisionsofthetriangles.

Anotherlimitationthatcannotbesolvedeasilyistheerrorformulas,duetotheamountofapproximationsinvolvingtheentireprocesstheseformulashaveprovennottobeveryreliableforthecasein L2(S⊭) althoughtheyhaveshowngreatresultsforthecommonFourierseries.

OverallIamverysatisfiedoftheresultsobtainedbytheprogramaswellasallthemathematical backgrounddevelopedinthethesistobackitup,Ihopethisarticleisusefulforsomeonewhodecidesto undertakeasimilarcaseofstudyinthefuture.

from0to5.

References

[1]M.Abramowitz,I.A.Stegun, Handbook ofMathematicalFunctionswithFormulas, Graphs,andMathematicalTables,Forsale bytheSuperintendentofDocuments,National BureauofStandardsAppliedMathematicsSeries 55,U.S.GovernmentPrintingOffice, Washington,DC,1964.

[2]Y.Kosmann-Schwarzbach, Groupesetsym´e-

tries.Groupesfinis,groupesetalg`ebresdeLie, repr´esentations,Secondedition,Les ´ Editionsde l’ ´ EcolePolytechnique,Palaiseau,2006.

[3]D.Mar´ın,M.Nicolau,Apriori L2-errorestimatesforapproximationsoffunctionson compactmanifolds, Mediterr.J.Math. 12(1) (2015),51–62.

Reports@SCM, 10 (2025),63–74;DOI:10.2436/20.2002.02.51.

73

Figure 3:Sphericalharmonicsasshownbytheprogramwith ℓ

UseofFourierseriesin S2 toapproximatestar-shapedsurfaces

[4]F.W.J.Olver,D.W.Lozier,R.F.Boisvert, C.W.Clark(ed.), NISTHandbookofMathematicalFunctions,U.S.DepartmentofCommerce,NationalInstituteofStandardsand Technology,Washington,DC;CambridgeUniversityPress,Cambridge,2010. dlmf.nist. gov

[5]S.Rosenberg, TheLaplacianonaRiemannian

Manifold.AnIntroductiontoAnalysisonManifolds,LondonMath.Soc.Stud.Texts 31,CambridgeUniversityPress,Cambridge,1997.

[6]G.Xu,DiscreteLaplace–Beltramioperatoron sphereandoptimalsphericaltriangulations, Internat.J.Comput.Geom.Appl. 16(1) (2006), 75–93.

https://revistes.iec.cat/index.php/reports

ANELECTRONICJOURNALOFTHE SOCIETATCATALANADEMATEM ` ATIQUES

ExtendedAbstracts

JaumeCapdevilaJov´e

DensitiesforHausdorffmeasureandrectifiability.Besicovitch’s1/2-conjecture

LuisPabloColmenar Fusiontheoremsandapplications

ManuelGarc´ıaGarc´ıa

Atypicalvaluesofcomplexpolynomialfunctions

MarinaGarcia-Romero

Exploringtheprinciplesofcoexistenceininvader-drivenreplicatordynamics

MarioGuill´en,PedroTradacete UniquepredualsandfreeobjectsinBanachspaces

VicentMirallesLluch Idempotentelementsofthegroupalgebra

AlbertoRodr´ıguezDur`a OnnilpotencyinbracesandtheYang–Baxterequation

FrancescTimonerVaquer Densityofhyperbolicityinfamiliesofcomplexrationalmaps

https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),75–92.

SCMMasterThesisDay

LastOctober3,withanotableattendance,wecelebratedthethirdSCMTFMday.Thisisanactivity organizedbytheCatalanMathematicalSociety(SCM)whichaimstofacilitatethosewhohavejustgraduatedfromamaster’sdegreeinmathematicsataCatalanuniversityorfromthecommonlinguisticarea (XarxaVives)topresenttheirFinalMaster’sThesis.Thisinteruniversitaryactivityitisaboutgivingyoung master’sgraduatestheopportunitytoparticipateandpresenttheirfirstcommunicationataworkshop,to energizethecommunityofyoungmathematiciansinthecountrythatstarttheresearch,toinformabout theconvocationoftheGaloisawardsandaboutthemagazine Reports@SCM,andtospreadthewordabout themathematicsmaster’sprogramsoftheuniversitiesoftheVivesNetworktostudentsinthefinalyear ofthemathematicsdegreeattendingtheday.

ThedaywasheldattheheadquartersoftheInstitutd’EstudisCatalansandhadtheparticipationas speakersofeightstudents,andalsowiththepresentationofthetwomaster’sthesesawardedwiththe EvaristeGalois2025prize(PedroL´opezandJoaquimDuran,winnerandrecipient),anawardgivenby theSCMtothebestfinalmaster’sthesisofthepreviousyear,inthiscase,2024.Infact,oneofthetwo winnersofthe2025GaloisawardpresentedtheirTFMinthe2024editionoftheSCMTFMday.

ThescientificcommitteeofthedaywasEnricCosme(co-coordinatoroftheMaster’sinMathematicalResearchoftheUV-UPV),SimoneMarchesi(editorinchiefof Reports@SCM),XavierMassaneda(coordinator oftheMaster’sinAdvancedMathematicsoftheUB-UAB),JordiSaludes(coordinatoroftheMaster’sin AdvancedMathematicsoftheUPC)andPabloSevilla(co-coordinatoroftheMaster’sinMathematical ResearchoftheUV-UPV).TheorganizingcommitteewasMontserratAlsina(presidentoftheSCM),JosepVives(vice-presidentoftheSCM)and ` OscarBur´esandPhilipPita,currentPhDstudentsandformer participantsinpreviouseditionsoftheday.

Reports@SCM collectsinthisissuetheextendedabstractsofthepresentationsoftheday.

https://revistes.iec.cat/index.php/reports

DensitiesforHausdorffmeasureand rectifiability.Besicovitch’s1/2-conjecture

JaumeCapdevilaJov´e

UniversitatAut`onoma deBarcelona jaume.capdevila.jove@uab.cat

Abstract

Resum (CAT)

Enaquesttreballestudiemundelsconceptescentralsdelateoriageom`etricadela mesura,eldeconjuntrectificable,ilasevarelaci´oamblesdensitatsperlamesura deHausdorff.Enaquestainteracci´ohihaundelsproblemesobertsm´esantics delateoria:laconjectura-1/2deBesicovitch.Estudiemunaselecci´oderesultats rellevants,desdelsarticlespionersdeBesicovitch[1]finsalamilloradePreiss iTiˇser[7].Despr´es,presentemunacontribuci´ooriginal:generalitzema Rn un exempledonatoriginalmentperBesicovitchenelpla,demostrant-nelespropietats clauiestenentaix´ıunacotainferiordelaconjecturaadimensi´oarbitr`aria.

Keywords: geometricmeasuretheory,Hausdorffmeasure,rectifiability, Besicovitch’s 1/2-conjecture.

Oneofthemainconceptsofgeometricmeasuretheoryisthatof m-rectifiablesubsetsof Rn,givenintegers0 < m ≤ n.Theyappearasageneralizationofthenotionof“nice” m-dimensionalsurfaces,suchas C 1 submanifolds,orLipschitzgraphs.Theyaresetswhich,uptoasetofzero Hm-measure,arecontainedin acountableunionofimagesofLipschitzmapswithdomainin Rm (where Hm denotesthe m-dimensional Hausdorffmeasure).Forexample,for m =1,the1-rectifiablesetsarethosewhicharecontainedina countableunionofrectifiablecurves,againuptoasetofzero H1-measure.Ontheothersideofthecoin, wehavethepurely m-unrectifiablesets,whicharethosethatcontainno m-rectifiablesubsetofpositive Hm-measure.Oneofthegoalsofgeometricmeasuretheoryistocharacterizerectifiabilityintermsofother geometricoranalyticalproperties.

Tothatend,oneofthebasictoolsisthatofthedensitiesfortheHausdorffmeasure.Considera set E ⊂ Rn suchthat0 < Hs (E ) < ∞ forsome0 ≤ s ≤ n,whichwecallan s-set.Onedefinesthe upperandlower s-densitiesof E atapoint x ∈ Rn,denotedasΘ∗s (E , x )andΘs ∗(E , x )respectively,as thelimsupandliminfas r → 0of

Hs (E ∩ Br (x )) (2r )s

When bothquantitiescoincide,thelimitiscalledthe s-densityof E at x ThedensitiesfortheHausdorffmeasureandthenotionofrectifiabilityareintimatelyconnected.Oneof themostimportanttheoremsinthisdirectionstatesthatan m-set E ⊂ Rn is m-rectifiableifandonlyifthe m-densityof E existsandisequalto1at Hm-almostallpointsof E .Thisisknownasthecharacterizationof

77 https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),77–78.

DensitiesforHausdorffmeasureandrectifiability.Besicovitch’s1/2-conjecture

rectifiabilityintermsofdensities.ThislineofstudywasinitiatedinthepioneeringworkofBesicovitch[1] in1938,whereheestablishedtheresultfor1-setsintheplane,i.e.,thecase m =1and n =2.It wasextendedtoarbitrarydimensionindifferentstages,withtheworkofMoore[6],Marstrand[4]and Mattila[5].

Anotherpointofconnectionbetweenthetwotopicsinvolvesthelowerdensityalone.Itwasprovenby BesicovitchinthesamearticlethatifΘ1 ∗(E , x ) > 3/4for H1-almostallpointsofa1-set E ,then E is automatically1-rectifiable.Followingthisidea,wedefinethefollowingcoefficient:

m(Rn):=min{σ> 0:forany m-set E ⊂ R

ThepreviouslystatedresultofBesicovitchtranslatestothebound σ1(R2) ≤ 3/4.Moreover,inthesame articlein1938heprovidedanexampleofapurely1-unrectifiableset P whichsatisfiesΘ1 ∗(P, x )=1/2 at H1-almostall x ∈ P;aformalproofofthisfactappearedlaterinapaperbyDickinson[3]in1939. Thisway,theyprovedthelowerbound σ1(R2) ≥ 1 2 .Withthisinmind,Besicovitchconjecturedthatthe exactvalueof σ1(R2)is1/2,whichisnowknownas Besicovitch’s 1/2-conjecture

Furtherimprovementstothisboundhavebeenobtainedsincethen.In1992,PreissandTiˇser[7]refined theestimateto σ1(Rn) ≤ (2+ √46)/12 < 59/80,whichholdsforall n ≥ 2(forallmetricspaces,infact). Recently,in2024,CamilloDeLellisetal.[2]establishedthat σ1(Rn) ≤ 7/10,whichiscurrentlythebest knownupperbound.

Inhigherdimensions(for m > 1),nogoodupperboundsareknownfor σm(Rn).Ontheotherhand, thesamelowerboundremainsvalid;inthiswork,wegeneralizeBesicovitch’sexampleintheplaneto arbitrarydimensions,therebyshowing σm(Rn) ≥ 1 2 ,forany0 < m < n

Thisisanoriginalcontributionfromthiswork.

References

[1]A.S.Besicovitch,Onthefundamentalgeometricalpropertiesoflinearlymeasurableplanesetsof points(II), Math.Ann. 115(1) (1938),296–329.

[2]C.DeLellis,F.Glaudo,A.Massaccesi,D.Vittone,Besicovitch’s1/2problemandlinearprogramming,Preprint(2024). arXiv:2404.17536

[3]D.R.Dickinson,Studyofextremecaseswithrespecttothedensitiesofirregularlinearlymeasurableplanesetsofpoints, Math.Ann. 116(1) (1939),358–373.

[4]J.M.Marstrand,Hausdorfftwo-dimensional measurein3-space, Proc.LondonMath.Soc.(3) 11 (1961),91–108.

[5]P.Mattila,Hausdorff m regularandrectifiable setsin n-space, Trans.Amer.Math.Soc. 205 (1975),263–274.

[6]E.F.Moore,Densityratiosand(ϕ,1)rectifiabilityin n-space, Trans.Amer.Math.Soc. 69 (1950),324–334.

[7]D.Preiss,J.Tiˇser,OnBesicovitch’s 1 2 -problem, J.LondonMath.Soc.(2) 45(2) (1992), 279–287.

https://revistes.iec.cat/index.php/reports

Fusiontheoremsandapplications

Resum (CAT)

Enteoriadegrupsfinits,moltsresultatscl`assicsimpliquensubgrupsdeSylow.Una direcci´onatural´esgeneralitzar-losmitjan¸cantsubgrupsdeHall.Enaquesttreball, mostremcomunresultatdeWielandtpermetfer-hoefica¸cment.Presentemdues aplicacions:unarelacionadaambelteoremadefusi´od’Alperin,iunaaltraambel subnormalitzador,unconceptemenysconegutper`oambconnexionsrecentsamb lateoriadecar`acters.

Abstract

Keywords: fusioningroups,subnormalizer.

Infinitegrouptheory,manyresultsareformulatedintermsofSylowsubgroupsandrelyheavilyonthe classicalSylowtheorems.Theseresultsarecentraltothelocal-globalphilosophyofthesubject,wherelocal propertiesofsubgroupsprovidevaluableinformationaboutthestructureofthewholegroup.

Wheneversuchtheoremsareestablished,anaturallineofinquiryarises:cantheseresultsbegeneralized beyondSylowsubgroups?OnepromisingdirectioninvolvesreplacingSylowsubgroupswithHallsubgroups, whicharemoregeneralbutretainmanydesirablepropertieswhentheyexist.However,suchgeneralizations oftenrequiremoresophisticatedtools,sincethetheoryofHallsubgroupsisnotasrobustorwidelyapplicable asSylowtheoryingeneralfinitegroups.

Inthiswork,wefocusonaclassicalbutperhapsunderappreciatedresultbyWielandt,whichprovesto beapowerfulinstrumentinextendingcertainSylow-basedstatementstomoregeneralcontextsinvolving Hallsubgroups.Wielandt’stheoremoffersaunifyingperspectivethatopensthedoortonewapplications. Wepresenttwomainapplicationsofthisapproach.ThefirstconcernsAlperin’sfusiontheorem,first provedbyAlperinin[1],afundamentalresultdescribinghowconjugacyinaSylow p-subgroupiscontrolled intermsofthelocalstructure.Thistheoremisimportantinsomeconjecturesinrepresentationtheoryand charactertheory.WewillshowhowWielandt’sresultcanbeusedtoextendaspectsofthistheorembeyond theSylowsubgroups,providingamoreflexibleframeworkforstudyingfusionphenomena.

Thesecondapplicationinvolvesalesswell-knownconcept:thesubnormalizerofasubgroup.Thisnotion, mainlystudiedbyCarloCasoloin[2],triestogeneralizetheconceptofnormalizer.Subnormalizersofferan alternativelensthroughwhichonecanexaminetheinternalstructureofafinitegroup.Recentdevelopments

https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),79–80.

Fusiontheoremsandapplications

showthatthisconceptisnotmerelytechnical:itisconnectedtonewconjecturesincharactertheoryand mayleadtofreshinsightsintotheinterplaybetweensubgroupstructureandrepresentationtheory.

Bothapplicationsillustratehowclassicaltools,whenviewedfromamodernperspective,canbeeffectivelyrepurposedtoapproachcontemporaryproblemsingrouptheory.Theideaswepresenthighlightthe ongoingrelevanceofresultslikeWielandt’stheoremanddemonstratethevalueofre-examiningclassical resultsthroughnewconceptualframeworks.

Acknowledgements

IwouldliketoexpressmysinceregratitudetoAlexanderMoret´oandNoeliaRizofortheirvaluableguidance, insightfulcomments,andcarefulcorrectionsthroughoutthedevelopmentofthiswork.

References

[1]J.L.Alperin,Sylowintersectionsandfusion, J. Algebra 6 (1967),222–241.

[2]C.Casolo,Subnormalizersinfinitegroups, Comm.Algebra 18(11) (1990),3791–3818.

https://revistes.iec.cat/index.php/reports

Atypicalvalues ofcomplexpolynomialfunctions

ManuelGarc´ıaGarc´ıa UniversitatdeVal`encia magarg25@alumni.uv.es

Resum (CAT)

Desde1983,ambeltreballdeBroughton,s’hanintrodu¨ıtdiversescondicionsde regularitatal’infinitperaunpolinomicomplex f quegaranteixenl’abs`enciade valorscr´ıticsal’infinit,´esadir,devalorsat´ıpicsde f quenos´onvalorscr´ıtics.En aquesttreballrecollimlescondicionsderegularitatm´esrellevantsiestudiemles relacionsquehihaentreelles.Enparticular,responemaduespreguntesobertes proposadesperD˜ungTr´angLˆeiJ.J.Nu˜no-Ballesterosa[3].

Abstract

Keywords: complexpolynomials,atypicalvalues,criticalvalues.

Thetopologyofcomplexpolynomialfunctions f : Cn −→ C hasbeenobjectofconsiderablestudyin recentdecades.Inparticular,acentralgoalistounderstandhowthetopologyofthefibers f 1(c), c ∈ C, changes.Inthiscontext,theconceptof locallytrivialfibrations playsakeyrole.Specifically,if f islocally atrivialfibrationat c ∈ C,thenthetopologyofthefibersnear c remainsunchanged.Thepoints c ∈ C where f failstobelocallyatrivialfibrationarecalledatypicalvaluesof f .Thesetofallatypicalvaluesof f isdenotedbyAtyp f .In[4],Thomprovedthefinitenessofthesetofatypicalvalues.However,determining preciselythissetisamajoropenproblem.

Amongtheatypicalvalues,onehasthecriticalvalues, i.e., f (Σf ) ⊂ Atyp f ,whereΣf isthesetof points x ∈ Cn wheredfx =0.Ingeneral,thisinclusionisstrict.Overthepastdecades,several regularity conditionsatinfinity for f havebeenintroducedinordertoguaranteetheequality f (Σf )=Atyp f

Thefirstoneisthenotionof tameness,whichwasintroducedbyBroughtonin[1]and[2].In[5],Tib˘ar compilessomeotherregularityconditionsatinfinity,suchasthe MalgrangeCondition (whichgeneralizes thenotionoftameness)andthe ρ-regularityatinfinity,where ρ isa controlfunction.Thefollowingchain ofimplicationsiswell-known:

f istame=⇒ f 1(c)satisfiestheMalgrangeCondition =⇒ f 1(c)is ρE -regularatinfinity.

Mostrecently,in[3],D˜ungTr´angLˆeandJ.J.Nu˜noBallesterosintroducedthenotionof atypicalvalues frominfinity.Inthispaper,theygeneralizetheBroughton’sGlobalBouquetTheoremin[2].Thepaper

https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),81–82.

Atypicalvaluesofcomplexpolynomialfunctions

concludesbyposingseveralopenquestionsaimedatgainingadeeperunderstandingofatypicalvalues frominfinity.Namely,

1.Isittruethat,if f istame,then f doesnothaveatypicalvaluesfrominfinity?

2.Doesafiber f 1(c)whichsatisfiestheMalgrangeConditioncorrespondtoavalue c whichisnotan atypicalvaluefrominfinity?

Inthisworkwereviewallthesedefinitionsandexplainourmaincontribution:

f 1(c)is ρ-regularatinfinity=⇒ c isnotanatypicalvaluefrominfinity.

Usingthisresult,weobtainanextensionofthepreviouschainofimplications:

f istame=⇒ f 1(c)satisfiestheMalgrangeCondition

=⇒ f 1(c)is ρE -regularatinfinity =⇒ c isnotanatypicalvaluefrominfinity.

Thisanswersthefirstquestionoftheauthorsin[3]andgivestherightimplicationforthesecondone.The otherimplicationremainsopen.

Acknowledgements

TheauthorwouldliketodeeplythankProfessorJuanAntonioMoyaP´erezforhisencouragementduringthis Master’sThesis.TheauthorwouldliketoexpresshissinceregratitudetoProfessorJ.J.Nu˜no-Ballesteros forhisvaluableadviceandcomments.

References

[1]S.A.Broughton,Onthetopologyofpolynomial hypersurfaces,in: Singularities,Part1 (Arcata, Calif.,1981),Proc.Sympos.PureMath. 40, AmericanMathematicalSociety,Providence,RI, 1983,pp.167–178.

[2]S.A.Broughton,Milnornumbersandthetopologyofpolynomialhypersurfaces, Invent.Math. 92(2) (1988),217–241.

[3]D˜ungTr´angLˆe,J.J.Nu˜noBallesteros,Are-

markonthetopologyofcomplexpolynomial functions, Rev.R.Acad.Cienc.ExactasF´ıs. Nat.Ser.AMat.RACSAM 113(4) (2019), 3977–3994.

[4]R.Thom,Ensemblesetmorphismesstratifi´es, Bull.Amer.Math.Soc. 75 (1969),240–284.

[5]M.Tib˘ar, PolynomialsandVanishingCycles, CambridgeTractsinMath. 170,CambridgeUniversityPress,Cambridge,2007.

https://revistes.iec.cat/index.php/reports

Exploringtheprinciplesofcoexistence ininvader-drivenreplicatordynamics

MarinaGarcia-Romero

UniversitatPolit`ecnica deCatalunya(UPC) marinagr99@gmail.com

Abstract

Resum (CAT)

Enaquesttreball,utilitzemla“replicatorequation”perexplorarunadeles q¨uestionsfonamentalsdelabiologiaevolutivail’ecologia:comesgeneraies mant´elabiodiversitat?Centrant-nosenelssistemes“invader-driven”,enqu`eles interaccionso“fitnesses”delesesp`eciesestandeterminadesperl’esp`ecieinvasora independentmentdel’esp`ecieenva¨ıda,busquemrelacionarles“fitnesses”amb lesesp`eciesquecoexisteixenalsestatsfinalsd’equilibri.Descobrimelmecanisme queregeixlaselecci´od’esp`eciessuperviventsiquemaximitzalaresist`enciadel sistemaenverslesinvasionsexternes,itrobemqueelnombremitj`ad’esp`eciesque coexisteixencreixambelnombreiniciald’esp`ecies.

Keywords: biologicalmodelling,replicatorequation,pairwiseinvasion fitnessmatrix,multi-speciessystem,coexistence,invasionresistance.

Studyingthenon-linearandoftencomplexdynamicsoflargesystemsofinteractingspecies,competingor cooperatingbetweenthem,canhelptodiscovertheprinciplesthat,inecosystems,leadsomespeciesto surviveandcoexist,whileothersgoextinct,tobetterunderstandofoneofthecentralquestionsinecology andevolutionarybiologythatremainsunsolved,whichishowbiodiversityisgeneratedandmaintained. Intheearly1970s,ecologistswidelyacceptedthatthestabilityandresilienceobservedinrichecosystems wereenhancedbycomplexityandbiodiversity,untilin1972theparadigmshiftedcompletelywhenRobert Maymathematicallyshowedthatrandomcomplexitytendstodestabilisesystemdynamics[3].Thisraised acontradictionbetweenobservationandtheoryknownastheecologyparadoxordiversity-stabilitydebate, highlightingtheneedforsomehiddenstructureorpatterninnature,suchastheantisymmetricpreypredatorinteractions[1].Inthiswork,westudyinvader-driveninteractionsasapotentialmechanismfor thestabilizationoflargecomplexsystemsandwefindthat,undercertainassumptions,invader-driven systemsleadtothecoexistenceofspecies.

Weusethereplicatorequationasatheoreticalframework[4],whichoriginatedingametheorybuthas beenwidelyappliedtobiologyandepidemiology[2].Givenasystemwith N species, S = {1,2,..., N}, considerthepairwiseinvasionfitness λj i fromspecies i to j ,with i , j ∈ S and λi i =0,andtheinvasion fitnessmatrixΛ=(λj i )i ,j ∈S .Then,thereplicatorequationmodelsthetimeevolutionofthespecies frequencies z =(z1, z2,..., zN ),with i ∈S zi =1and0

Exploringtheprinciplesofcoexistenceininvader-drivenreplicatordynamics

wheretheconstantΘ ≥ 0isthespeedofdynamicsand Q(z)istheglobalmeanfitnessorsystemresistance toexternalinvasion.Inparticular,wefocusoninvader-drivensystems,inwhichpairwiseinteractionsare determinedbytheinvadingspeciesregardlessoftheinvadedone,i.e., λj i = λi (1 δj i ) ∀ i , j ∈ S ,soeach species i ischaracterizedbyitsactivetrait λi .Westudytheequilibriumstates z∗ tounderstandhow fitnessesinthecase λi > 0 ∀ i ∈ S relatetothesetofsurvivingorcoexistingspeciesatequilibrium, S ∗ = {i ∈ S | z ∗ i > 0}⊆ S ,findingthat Q∗ = Q(z∗)playsacrucialroleinthespeciesselectionprocess.

Weprovethatlocallyasymptoticallystableequilibriaarealwayscomposedofthetop n = |S ∗| species withoutgaps,withfitnesses λN ≤···≤ λn ≤···≤ λ2 ≤ λ1 and2 ≤ n ≤ N,and,furthermore,wefind numericalevidencethateachsystemcontainsjustoneoftheseequilibria,whichinturnisaglobalattractor (anyinitialconditionwithallspeciespresenttendsasymptoticallytowardsit).Therefore,foreach S there isaunique S ∗ thatcanbeasymptoticallyreachedbythedynamicsand,hence,auniquesetofspecies characterizedby n thatendupcoexisting.Wediscoverthemechanismrulingthespeciesselection(see Figure 1),whichstartingwithtwospeciesiterativelyaddsaspecies i ∈ S if Q∗ i 1 <λi ,thatis,ifit caninvadetheprevious i 1,untilsomespecies n meetsthecondition λn+1 < Q∗ n <λn.Moreover,we provethatineachstep Q∗ increases, Q∗ i 1 < Q∗ i ,sothisbiologicalprocesstendstomaximizethesystem resistancetoinvasion.

Lastly,usingthismechanismwecreateanalgorithmthatallowstofind n forseveralinvader-driven systemsgeneratedrandomlywith λi ∼U [0,1],from N =5to N =500.Fittingthedatawefindthatthe meannumberofcoexistingspeciesincreasesaccordingto¯ n =1.381√N,suggestingthatinvader-driven interactionscouldbeapotentialmechanismthroughwhichecosystemsstabilizeandmaintainbiodiversity.

Figure1:Speciesselectionmechanismininvader-drivensystems,0 <λN ≤···≤ λn ≤···≤

Acknowledgements

IwanttothankEridaGjiniforwelcomingmeintohergroupandforalltheinterestingdiscussions,andSten Madec,NicolaCinardi,Tom´asFreire,Tom´asCamolasandThaoLe,forcreatinganenrichingenvironment. IalsowanttothankJos´eTom´asL´azaroOchoaforhisco-supervision,supportandimportantcontributions.

References

[1]T.Chawanya,K.Tokita,Large-dimensional replicatorequationswithantisymmetricrandom interactions, J.Phys.Soc.Japan 71(2) (2002), 429–431.

[2]S.Madec,E.Gjini,Predicting N-straincoexistencefromco-colonizationinteractions:epidemiologymeetsecologyandthereplicatorequation,

Bull.Math.Biol. 82(11) (2020),Paperno.142, 26pp.

[3]R.M.May,Willalargecomplexsystembestable?, Nature 238(5364) (1972),413–414.

[4]P.D.Taylor,L.B.Jonker,Evolutionarilystable strategiesandgamedynamics, Math.Biosci. 40(1-2) (1978),145–156.

https://revistes.iec.cat/index.php/reports

Uniquepredualsandfreeobjects inBanachspaces

MarioGuill´en

UniversitatPolit`ecnica deVal`encia(UPV)

Resum (CAT)

PedroTradacete

Matem´aticas(ICMAT) mguisan3@alumni.upv.es

Abstract

EstudiemquanunespaideBanacht´eun´unicpredual,centrant-nosprimerenles funcionsholomorfesacotadesaldiscunitatianalitzantlademostraci´od’Ando. Consideremcomestendreelresultatadiversesvariables,onapareixendificultats t`ecniques.Tamb´etractemdiferentscondicionssuficientspergarantirlaunicitati elcasdereticlesdeBanach.

Keywords: uniquepredual,Banachspace,boundedholomorphicfunctions, Property(X),L-embeddedspace,freeBanachlattice.

A predual ofaBanachspace X isaBanachspace Y suchthatthereexistsanisomorphism Y ∗ → X .When X admitsonlyonesuchspace Y uptoisometricisomorphism,wesaythat X hasa uniquepredual.The problemofdeterminingwhenaBanachspacehasauniquepredualisacentraloneinfunctionalanalysis, startingintheworksofDixmier(1948)andNg[5]andstudiedbySakai,Ando,Godefroy,Pfitzner,and others.ClassicalexamplesofspaceswithuniquepredualsincludevonNeumannalgebrasbySakai(1971), thespaceofboundedholomorphicfunctionsonthecomplexdisk H ∞(D)byAndo[1],andseparable L-embeddedBanachspacesbyPfitzner[6].Godefroy’ssurvey[3]remainsakeyreferencesummarizing thesedevelopmentsandlistingopenproblems.

Thisprojectrevisitstheuniquenessproblemwithemphasisonthespaceofboundedholomorphic functions, H ∞(U),definedonanopensubset U ⊂ Cn.WereviewAndo’soriginalproofoftheuniqueness ofthepredualof H ∞(D),whichidentifiesthespace L1(T)/H 1 0 (T)asitsuniquepredual.Wepresenta detailedprooffollowingbothAndo’soriginalformulation[1]andalateronefromGarnett[2],fillingseveral gapsleftunprovedintheliterature.

WeextendAndo’sresulttothecasewhere U isadisjointunionofsimplyconnectedopensubsets ofthecomplexplane.UsingMujica’snotionoftheholomorphicfreeBanachspace G ∞(U)givenin[4], characterizedbytheuniversalproperty

weexplicitlyconstructanisometricisomorphism

https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),85–86.

UniquepredualsandfreeobjectsinBanachspaces

where U = α∈A Uα.Thisallowsustoprovethat H ∞(U)hasauniquepredualwhenever U issucha disjointunion,thusgeneralizingAndo’sresult.

Wethenattempttoextendtheresulttoseveralcomplexvariables,considering H ∞(Tn)and H ∞(Bn). Followingadifferentpathofprovinguniquenessintheone-dimensionalcase,wereducetheproblemof provingthat H ∞(B n)hasauniquepredualtoshowingthat BH 1 0 (Sn ) is ∥·∥p -closedforsome p ∈ (0,1). However,thishigher-dimensionalsettingpresentsseveralchallenges.Whenpassingfromonevariableto several,thespace H 1 0 (Sn)isnolongercontainedintheHardyspace H 1(Bn),sopre-compactnessarguments fromHardyspacetheorycannotbeapplied.Consequently,noconclusiveresultswereobtainedinthis direction.

Thisworkalsoreviewstechniquesguaranteeinguniquenessofpredualsinbroadersettings.Twosufficientconditionsarerevisited: Property(X) (Godefroy–Talagrand,1980),whichensuresthat X isthe uniquepredualof X ∗,andthenotionof L-embeddedspaces,forwhichseparablecasesweresolvedby Pfitzner[6].Finally,weexploretheBanachlatticesetting,introducingthe freeBanachlattice FBL[E ] (Avil´es–Rodr´ıguez–Tradacete,2015).Althoughweexplorepossibledefinitionsforpredualequivalencein thissetting,wefinddifficulties,particularlybecausethespaceoflatticehomomorphismsisnotavector space,thusleavingthisproblemforfutureresearch.

Acknowledgements

ThisprojecthasbeendevelopedundertheJAEIntroScholarshipfromtheConsejoSuperiordeInvestigacionesCient´ıficas(CSIC).

References

[1]T.Ando,Onthepredualof H ∞ , Comment. Math.Spec.Issue 1 (1978),33–40.

[2]J.B.Garnett, BoundedAnalyticFunctions,Revisedfirstedition,Grad.TextsinMath. 236, Springer,NewYork,2007.

[3]G.Godefroy,Existenceanduniquenessofisometricpreduals:asurvey,in: BanachSpaceTheory (IowaCity,IA,1987),Contemp.Math. 85, AmericanMathematicalSociety,Providence,RI, 1989,pp.131–193.

[4]J.Mujica,Linearizationofboundedholomorphic mappingsonBanachspaces, Trans.Amer.Math. Soc. 324(2) (1991),867–887.

[5]K.F.Ng,OnatheoremofDixmier, Math.Scand. 29 (1971),279–280.

[6]H.Pfitzner,SeparableL-embeddedBanach spacesareuniquepreduals, Bull.Lond.Math. Soc. 39(6) (2007),1039–1044.

https://revistes.iec.cat/index.php/reports

UniversitatPolit`ecnica deVal`encia vicent@vmiralles.com

Idempotentelements ofthegroupalgebra

Resum (CAT)

L’objectiud’aquesttreball´esestudiarelselementsidempotentscentralment primitiusdel’`algebradegrupidesenvoluparunm`etodeperalseuc`alculenelcas decossosfinits.Apartirdelateoriaderepresentacionsdegrupsfinitsideresultats sobrem`oduls,`algebresiextensionsdecossos,s’introdueixelconceptedecos d’escissi´operaungrup.Finalment,s’exploracoml’acci´odeGaloissobrel’`algebra degrupdefinidasobreaquestscossospermetobtindreaquestsidempotentsdelcos original.

Abstract

Keywords: idempotentelements,splittingfields,groupalgebra,Galois action,finitefields.

Thisworkfocusesonthestudyofidempotentelementsofgroupalgebras,withparticularemphasison centrallyprimitiveidempotents.Theseelementsarefundamentalbecausetheyallowthealgebratobe decomposedintosimplerblocks.Theimportanceofcentrallyprimitiveidempotentsliesinthefactthat eachofthemgeneratesoneoftheseblocksand,moreover,theyformabasisforthecentreofthealgebra, whichcompletelydefinesitsstructure.

Themainobjectiveistodevelopanexplicitandpracticalmethodforcalculatingtheseidempotents overfieldswhosecharacteristicdoesnotdividetheorderofthegroup(whichwewillassumetobefinite), andwhichareoftennotalgebraicallyclosed.Thisisnoeasytask,sincemanyresultsinrepresentation theoryrelyonthelatterproperty(see[3])andarenotvalidinamoregeneralcontext.Forthisreason,we resorttotheconceptofasplittingfieldforagroup(see[1, 2]),whichgeneralisesthealgebraicallyclosed field,providingatheoreticalframeworkthatguaranteesthevalidityofmanyclassicalresults,includingthe expressionoftheseidempotents.

Themethodwedevelop,oftenknownasGaloisdescent,consistsofexploitingtheexpressionofcentrally primitiveidempotentsofthegroupalgebraoverasplittingfield.TheideaistoconsiderafiniteGalois extensionoftheoriginalfieldthatisasplittingfieldforthegroup;inthisextension,theGaloisgroup actsontheseidempotents.Theexpressionoftheseidempotentsisknownsincetheyaredefinedovera splittingfield,anditcanbeshownthatthesumoftheorbitsresultingfromthisactionultimatelygives usthecentrallyprimitiveidempotentswearelookingforintheoriginalfield(see[4]).Thismethodis significantlysimplerthanotherapproaches,suchastheonein[5],whichreliesonthecomputationof

87 https://revistes.iec.cat/index.php/reports Reports@SCM, 10

Idempotentelementsofthegroupalgebra

adivisionring’sdimension—agenerallynon-trivialtask.Furthermore,weprovethatbothmethodsare equivalentbysummingoverageneralorbittoobtaintheexpressiongivenin[5].

Toillustratetheprocedure,weconcludewithadetailedapplicationtofinitefields,wheretheefficiency ofourapproachbecomesparticularlyevident.Thepracticalityofthemethodliesnotonlyinthesimplicity oftheorbitcomputations—thankstothecyclicnatureoftheGaloisgroupgeneratedbytheFrobenius automorphism—butalsointhetheoreticalresultspreviouslydevelopedinthiswork,whichdirectlyprovide thecorrespondingsplittingfields.Inthisexample,wefirstestablishtheidentificationbetweencharacters overthesplittingfieldandordinarycharactersviaBrauercharacters(seeagain[3]),andthencarryout theexplicitcomputationoftheidempotents,therebydemonstratingtheapplicabilityandstrengthofour self-containedapproach.

Acknowledgements

IwouldliketoexpressmygratitudetoMar´ıaJos´eFelipeRom´an,V´ıctorManuelOrtizSotomayorandXaro SolerEscriv`aforallowingmetodevelopthiswork,aswellasfortheirguidanceandcorrections.

References

[1]C.W.Curtis,I.Reiner, RepresentationTheory ofFiniteGroupsandAssociativeAlgebras,Pure Appl.Math. XI,IntersciencePublishers(adivisionofJohnWiley&Sons,Inc.),NewYorkLondon,1962.

[2]K.Doerk,T.Hawkes, FiniteSolubleGroups,De GruyterExp.Math. 4,WalterdeGruyter&Co., Berlin,1992.

[3]I.M.Isaacs, CharacterTheoryofFiniteGroups, PureAppl.Math. 69,AcademicPress[Har-

courtBraceJovanovich,Publishers],NewYorkLondon,1976.

[4]G.Karpilovsky, GroupRepresentations.Volume1.PartB:IntroductiontoGroupRepresentationsandCharacters,North-HollandMath. Stud. 175,North-HollandPublishingCo.,Amsterdam,1992.

[5]K.Lux,H.Pahlings, RepresentationsofGroups. AComputationalApproach,CambridgeStud. Adv.Math. 124,CambridgeUniversityPress, Cambridge,2010.

https://revistes.iec.cat/index.php/reports

Onnilpotencyinbraces andtheYang–Baxterequation

AlbertoRodr´ıguezDur`a UniversitatdeVal`encia albertord.bg@gmail.com

Resum (CAT)

Lesbridess´onestructuresalgebraiquesquepermetenestudiarlessolucionsno degeneradesdel’equaci´odeYang–Baxter(EYB).Cadabridaadmetunasoluci´o nodegeneradai,rec´ıprocament,totasoluci´od’aquesttipusest`adeterminada perunabridaassociada.Aix´ı,laclassificaci´odelessolucionsnodegenerades dep´endel’an`alisiestructuraldelesbrides.Lesseuespropietatsalgebraiqueses corresponenamblesdelessolucions,ilanilpot`enciapermetdescriureelcar`acter multipermutacionald’aquestesestructures.

Keywords: braces,Yang–Baxter,nilpotency.

Abstract

TheYang–Baxterequation(YBE)isafundamentalequationintheoreticalphysics,arisingindependently intheworksofC.N.Yang(1967),NobelLaureateinPhysics,andR.J.Baxter,withintheframeworksof quantumfieldtheoryandthestudyofintegrablemodelsinstatisticalmechanics,respectively.

TheformulationoftheYBEisstronglyinspiredbythecelebratedReidemeistermoves(cf.[3]).

Consequently,thestudyofYBEsolutionshasgainedsignificantrelevanceinrecentdecades,both becauseofitsintrinsicimportanceanditsapplicationsinbraidtheory,braidedgroups,quantumgroups, cryptography,andnoncommutativegeometry.

ThemultidisciplinarycontextoftheYBEhasgeneratedgreatinterestinthesearchforandclassification ofitssolutions.

OpenProblem. TofindandclassifythesolutionsoftheYang–Baxterequation.

GiventheHerculeannatureofthistask,theFieldsMedalistV.G.Drinfeld([2])proposedfocusing ontheso-calledset-theoreticsolutionsoftheYBE,atypeofcombinatorialsolutionwhosegeometricand symmetriccharacternaturallygivesrisetoalgebraictechniques.

Inthiswork,weundertakeathoroughanalysisofthealgebraicpropertyofnilpotencyinbraces,as aclearandsignificantexampleofthetranslationofalgebraicpropertiesintoclassificatorypropertiesof YBEsolutions.Westudytheso-calledlateralnilpotenciesinbraces,whichhaveadistinctimpactboth

https://revistes.iec.cat/index.php/reports Reports@SCM, 10 (2025),89–90.

OnnilpotencyinbracesandtheYang–Baxterequation

onthestructuralanalysisofbracesandontheclassificationofsolutions.Inthiscontext,akeyconcept ofnilpotencyinbraces—onethathasrecentlyemergedandhasadecisiveimpactbothstructurallywithin bracesandclassificatorilywithinYBEsolutions—iscentralnilpotencyinbraces.Thistypeofnilpotency ariseswiththeaimofunifyingbothlateralnilpotenciesinbraces,asshownin[1],whereitisdemonstrated thatcentralnilpotencyinbracescanberegardedasthetrueanalogue,withinbracetheory,ofgroup nilpotency.

Withingrouptheory,thelocalstudyofnilpotencyor p-nilpotencyassociatedwithaprime p has undergonesubstantialdevelopmentfollowingtheseminalworksofHallandHigman(cf.[4]).Akeyconcept inthiscontextisthe p-Fittingsubgroupofafinitegroup,thelargestnormal p-nilpotentsubgroupofthe group.

Themainobjectiveandcontributionofthisworkistheintroductionandanalysisofcentral p-nilpotency infinitebraces.Weconductacomprehensivestructuralstudyofcentral p-nilpotencyinbraces,allowing ustodefineanappropriate p-Fittingideal.Thiscontributionisoriginalwithinthetheoryandisintended toinspirefurtherdevelopmentsinthisfield.

Acknowledgements

TheauthorgratefullyacknowledgesProfessorAdolfoBallesterosandProfessorVicentP´erez(Universitat deVal`encia)forinsightfulguidanceandconstructivefeedbackthroughoutthisresearch.

References

[1]A.Ballester-Bolinches,R.Esteban-Romero, M.Ferrara,V.P´erez-Calabuig,M.Trombetti, CentralnilpotencyofleftskewbracesandsolutionsoftheYang–Baxterequation, PacificJ. Math. 335(1) (2025),1–32.

[2]V.G.Drinfeld,Onsomeunsolvedproblemsin quantumgrouptheory,in: QuantumGroups (Leningrad,1990),LectureNotesinMath. 1510 Springer-Verlag,Berlin,1992,pp.1–8.

[3]P.Etingof,T.Schedler,A.Soloviev,SettheoreticalsolutionstothequantumYang–Baxterequation, DukeMath.J. 100(2) (1999), 169–209.

[4]P.Hall,G.Higman,Onthe p-lengthof p-soluble groupsandreductiontheoremsforBurnside’s problem, Proc.LondonMath.Soc.(3) 6 (1956), 1–42.

https://revistes.iec.cat/index.php/reports

Densityofhyperbolicityinfamilies ofcomplexrationalmaps

FrancescTimonerVaquer

UniversitatdeBarcelona francesc.timoner-vaquer@utoulouse.fr

Abstract

Resum (CAT)

Undelsproblemesobertscentrals´esladensitatd’hiperbolicitat.Enaquesttreball hoinvestiguemenladin`amicacomplexaunidimensional,iensconcentremenelcas polin`omic(casparticulard’unafunci´oracional)comamodelonelsmecanismes principalspodenserexposatsicomprovatsendetall.Laviaprocedimental´esclara: primer,laconstrucci´odepecesdepuzleenunentorndelconjuntdeJulia;segon, l’´usd’aquestesperdefinirunafunci´odecaixacomplexa;ifinalment,l’aplicaci´ode teoremesderigidesaaaquestes.Aquestproc´estradueixlainformaci´ocombinat`oria enrigidesaperalspolinomis,demostrantqueunpolinominorenormalitzablepot seraproximatperunpolinomihiperb`olic.

Keywords: complexdynamics,holomorphicdynamics,rationalmaps, hyperbolicity,renormalisation,complexboxmapping.

A rationalmap isaholomorphicanalyticfunction f : C → C ontheRiemannspherethatcanbewrittenas thequotientoftwocoprimepolynomials;equivalently, f (z )= P(z ) Q(z ) , where P, Q arecomplexpolynomials ofsomedegree.Thedegreeof f isdefinedas d =max(deg P,deg Q),andweassumethat d ≥ 2.In theparticularcasewhere Q isaconstant, f isjustapolynomial.Rationalmapsofdegree d ≥ 2forma finite-dimensionalspace,soexploringthisparameterspaceisfeasible.Everyrationalmapofdegree d ≥ 2 has2d 2criticalpoints(countingmultiplicity),andnearthesepoints,themapbehaveslike z → z k ,soit ishighlycontractingandfailstobeinjective.Awayfromthecriticalpoints, f isalocalhomeomorphism.

Definition (Hyperbolicrationalmap). Arationalmapissaidtobe hyperbolic ifallitscriticalpointsare inthebasinsofattractingperiodicpoints.

Conjecture (Densityofhyperbolicity). Thehyperbolicrationalmapsformanopenanddensesetinthe spaceofallrationalmapsofagivendegree.

Definition. Wesaythatamapis non-renormalisable ifitdoesnotadmitanypolynomial-likerestriction foranyiterationwithconnectedfilled-inJuliaset.

Themaingoalofthisthesisistodeconstruct,understandalldetailsandprovethefollowingtheorem:

MainTheorem (Theorem1.3in[3]). Letfbeanon-renormalisablepolynomialofdegreed ≥ 2,without neutralperiodicpoints.Then,fcanbeapproximatedbyasequenceofhyperbolicpolynomials (gi ) ofthe samedegree.

91 https://revistes.iec.cat/index.php/reports

, 10 (2025),91–92.

Densityofhyperbolicityinfamiliesofcomplexrationalmaps

Totackletheproblemwereviewandcombineseveralfundamentaltools:puzzlepiecedecompositions sowecanconsiderreturnsandtracksymbolicallycriticalorbits,B¨ottchercoordinatesnearinfinitythat linearizeescapingbehaviour,holomorphicmotionstofollowdynamicalobjectsacrossparameters,and quasi-conformalconjugaciestotransfergeometriccontrolbetweenmaps.Also,wesupposeourmapisnonrenormalisable:forarationalmap,onedemandsthatitscriticalorbitsdonotreturninsmallneighbourhoods ina“periodicway”.Thesemapsareoftenrigid,inthesensethattheircombinatorialstructuredetermines theirgeometry.Finally,andmostimportantly,wemakeuseofcomplexboxmappingsasaninducedmap definedonadisjointunionoftopologicaldiscsthatcapturesreturndynamicsofcriticalorbitsinsidea controllabledomain(“upgrade”ofthefamouspolynomial-likemaps).Theseareflexibleenoughtoencode bothlocalrenormalisationbehaviourandglobalcombinatorialconstraints.

Theseminalpaper[3](ourmainreference)lacksexplicittechnicalassumptionsinitsstatementsand proofs.Thismakessomeoftheirstatements,aswritteninthatpaper,notentirelycorrect.Someassumptionswereimplicitornotconsidered,forexample,thedynamicallynaturalpropertyofcomplexbox mappings.Someparts,claimedtobestraightforward,arenot.In[1]theyclarifiedandfixedsomeresults onrigidityofpolynomialsandboxmappings,butthetheoremstatedaboveremainsunclear.Soforthe firsttimeintheliteratureofcomplexdynamics,weprovidedetailedexplanationsforeachpartoftheproof ofthattheorem.Weconsidertheimplicationsandensurevalidity,especiallywhenconsideringthedynamicallynaturalpropertyofboxmappings.Ouraimistoreviewtheexistingliterature([4, 2]),emphasising crucialaspects,andcomprehensivelyunderstandthetoolsrequiredforthetheorem’sproof.Weaimto encapsulatethemina“black-box”andusethemtoadvanceresearch,forinstance,toestablishthedensity ofhyperbolicityinotherfamiliesofrationalmaps.Webelievethismeticulousdeconstructionandattention todetailcansignificantlycontributetothegeneralpublic’sunderstandingofthesubjectmatter.

TheproofoftheMainTheoremliesonaconstructionofdynamicallynaturalboxmappingsfornonrenormalisablepolynomialswithoutneutralperiodicpointstogetherwithaverificationofthehypotheses neededtoinvokerigiditytheorems.Inrigidfamilies,topologicallyconjugatemapsareautomaticallymore regular(e.g.,quasi-conformalorconformalincomplexdynamics).Byanotherofthemaintheoremsneeded (Theorem6.1in[1]),combinatoriallyequivalentnon-renormalisabledynamicallynaturalcomplexboxmappingsarerigid,andhence,quasi-conformallyconjugate.Thisresult,alongwithotherknownorbasic notions,leadstothequasi-conformalrigidityofnon-renormalisablepolynomials.Consequently,theoriginal polynomialisapproximated,intheuniformtopologyoncompactsets,byhyperbolicpolynomials;hence densityofhyperbolicityholdsfortheclassconsidered.

References

[1]T.Clark,K.Drach,O.Kozlovski,S.vanStrien, Thedynamicsofcomplexboxmappings, Arnold Math.J. 8(2) (2022),319–410.

[2]K.Drach,D.Schleicher,RigidityofNewtondynamics, Adv.Math. 408 (2022),partA,Paper no.108591,64pp.

[3]O.Kozlovski,S.vanStrien,Localconnec-

tivityandquasi-conformalrigidityofnonrenormalizablepolynomials, Proc.Lond.Math. Soc.(3) 99(2) (2009),275–296.

[4]C.T.McMullen,D.P.Sullivan,Quasiconformal homeomorphismsanddynamics.III.TheTeichm¨ullerspaceofaholomorphicdynamicalsystem, Adv.Math. 135(2) (1998),351–395.

https://revistes.iec.cat/index.php/reports

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.