Mean curvature flow: proceedings of the john h. barrett memorial lectures held at the university of

Page 1


MeanCurvatureFlow:ProceedingsoftheJohnH. BarrettMemorialLecturesheldattheUniversity ofTennessee,Knoxville,May29–June1,2018 TheodoraBourni(Editor)

https://ebookmass.com/product/mean-curvature-flowproceedings-of-the-john-h-barrett-memorial-lectures-held-atthe-university-of-tennessee-knoxvillemay-29-june-1-2018-theodora-bourni-editor/

Instant digital products (PDF, ePub, MOBI) ready for you

Download now and discover formats that fit your needs...

A3N2M: Approximation, Applications, and Analysis of Nonlocal, Nonlinear Models: Proceedings of the 50th John H. Barrett Memorial Lectures Tadele Mengesha

https://ebookmass.com/product/a3n2m-approximation-applications-andanalysis-of-nonlocal-nonlinear-models-proceedings-of-the-50th-john-hbarrett-memorial-lectures-tadele-mengesha/ ebookmass.com

Trapped: Brides of the Kindred Book 29 Faith Anderson

https://ebookmass.com/product/trapped-brides-of-the-kindredbook-29-faith-anderson/

ebookmass.com

Patristic Theology: The University Lectures of Fr. John Romanides Romanides

https://ebookmass.com/product/patristic-theology-the-universitylectures-of-fr-john-romanides-romanides/

ebookmass.com

Work, Retire, Repeat: The Uncertainty of Retirement in the New Economy Ghilarducci

https://ebookmass.com/product/work-retire-repeat-the-uncertainty-ofretirement-in-the-new-economy-ghilarducci/

ebookmass.com

Our

Forever Place Maren Hill

https://ebookmass.com/product/our-forever-place-maren-hill/

ebookmass.com

Human Resource Management in the Pornography Industry: Business Practices in a Stigmatized Trade 1st ed. Edition

David M. Kopp

https://ebookmass.com/product/human-resource-management-in-thepornography-industry-business-practices-in-a-stigmatized-trade-1st-ededition-david-m-kopp/ ebookmass.com

Bratva Beauty (The Ivankov Brotherhood Book 4) Sabine Barclay & Celeste Barclay

https://ebookmass.com/product/bratva-beauty-the-ivankov-brotherhoodbook-4-sabine-barclay-celeste-barclay/

ebookmass.com

Careless (Bastards of Grove Hill Book 2) Zepphora .

https://ebookmass.com/product/careless-bastards-of-grove-hillbook-2-zepphora-4/

ebookmass.com

Wicked Heiress (Princes of Devil's Creek Book 5) Jillian Frost

https://ebookmass.com/product/wicked-heiress-princes-of-devils-creekbook-5-jillian-frost/

ebookmass.com

Pentecostal Politics in a Secular World: The Life and Leadership of Lewi Pethrus 1st ed. Edition Joel Halldorf

https://ebookmass.com/product/pentecostal-politics-in-a-secular-worldthe-life-and-leadership-of-lewi-pethrus-1st-ed-edition-joel-halldorf/

ebookmass.com

Mean Curvature Flow

De Gruyter Proceedings in Mathematics

Mean Curvature Flow

Proceedings of the John H. Barrett Memorial Lectures held at the University of Tennessee, Knoxville, May 29 – June 1, 2018

Editors

Prof. Dr. Theodora Bourni

University of Tennessee Department of Mathematics

1403 Circle Drive

Knoxville TN 37996-1320

USA

Prof. Dr. Mat Langford

University of Tennessee Department of Mathematics 1403 Circle Drive

Knoxville TN 37996-1320

USA

ISBN 978-3-11-061818-1

e-ISBN (PDF) 978-3-11-061836-5

e-ISBN (EPUB) 978-3-11-061822-8

Library of Congress Control Number: 2020933595

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de.

© 2020 Walter de Gruyter GmbH, Berlin/Boston Printing and binding: CPI books GmbH, Leck www.degruyter.com

Foreword

TheJohnBarrettMemorialLectureswereestablishedin1970asatributetoDr.John H.Barrett,anexpertonoscillationanddisconjugacytheoryoflinearordinarydifferentialequationsandinfluentialMathematicsDepartmentHeadattheUniversity ofTennessee.OriginallydedicatedtotopicsinBarrett’sfield,ordinarydifferential equations,thelectureshavebeenheldannuallysince1970.Sincethe1980’s,however, thelecturethemeshavetraversedthemathematicallandscape—frommathematics educationthroughcomputationalandappliedmathematics,discretemathematicsand stochasticstogeneralrelativity,nonlinearpartialdifferentialequationsandtopological quantumfieldtheory.

Duringtheirhistory,theBarrettLectureshavebeengracedbyasuccessionof highlydistinguishedmathematicians,withlecturesgivenbyGarrettBirkoff(1972);Sir MichaelAtiyah,IsadoreSingerandKarenUhlenbeck(1989);SirJohnBallandLawrence C.Evans(1991);SergiuKlainerman,Fang-HuaLinandMichaelStruwe(1995);Alice Chang,TobiasColdingandKarstenGrove(2000);JohnBaez(2006);RichardSchoen, RobertWaldandIgorRodnianski(2011);andFernandoMarquesandAndréNeves (2013)tonameafew.

Inrecentyears,geometricanalysishasalsobeenrepresented,withworkshopson generalrelativityin2011andtheproofoftheWillmoreconjecturein2013.Marques andNevesacknowledgedtheimportantroleplayedbytheBarrettLecturesindisseminatingtheirproofoftheWillmoreconjecturetothemathematicalcommunityintheir InternationalCongressofMathematiciansaddressesin2014.

TheLectureshavebeenpartiallysupportedbyagrantfromtheMathematicsDepartmentoftheUniversityofTennesseesincetheirinauguration.Inrecentyears,theyhave alsobeensupportedbytheCollegeofArtsandSciencesandtheOfficeofResearchand EngagementatUTKand,since2017,havereceivedsupportfromtheBarrettMemorial Endowment,anendowmentfundstartedbyformerUniversityofTennessee,Knoxville mathematicsdepartmentheadJohnBradleyandhiswifeBeverley.TheLectureshave alsobeenconsistentlysupportedbytheNationalScienceFoundation¹.

2018JohnH.BarrettMemorialLectures:

Geometricanalysisisoneofthemostactiveandexcitingareasinpuremathematics todayandgeometricflowsinparticularhaveprovedtobeapowerfultoolintheanalysis ofalargenumberofimportantproblemsindifferentialgeometry,imageprocessing

1 The 2018 Lectures were supported partly by the NSF grant DMS-1812058.

https://doi.org/10.1515/9783110618365-202

andmathematicalphysics,leadingtoaprofoundimpactoneachofthesefields.They alsoarisenaturallyinvariousphysicalcontextssuchasthermomechanics,annealing metals,crystalgrowth,flamepropagation,wearingprocessesandconformalfield theory. Meancurvatureflow,thethemeofthe2018JohnBarrettMemorialLectures,is inmanywaystheprototypicalgeometricflowequation.

TheLecturestookplaceduringthefourdayperiod29May-1June,2018atthe UniversityofTennessee,Knoxville.Theyhadtheformatofaworkshoponrecent developmentsintheareaofmeancurvatureflowaimedprimarilyatgraduatestudents andearlycareerresearchers.Furthertalksweregiveninthebroaderareaofgeometric flows(whichincludes,forexample,theRicci,YamabeandGausscurvatureflows) includingaseriesofshorttalksbygraduatestudentsandearlycareerresearchers.

Theintroductoryone-dayworkshop,consistingof300minutesoflecturesandadditionalinformaldiscussion,introducedparticipants(overhalfofwhomweregraduate orundergraduatestudents)tobasicmethodsinthefieldofmeancurvatureflow,and geometricflowsmoregenerally.Duringthefollowingthreedays,participantsheard surveylecturesintroducingrecentdevelopmentsandmajorresearchtrendsinthe meancurvatureflow,deliveredbyleadingexpertsinthefield.Thesupplementaryinvitedandshorttalksprovidedfurtherinsightintocurrentresearchtrends.TheBradley Lecturer,BruceKleiner,presentedhisrecentresolution,withRichardBamler,ofthe generalizedSmaleconjecture(apartfromthecaseof ℝP3).

UniversityofTenneseeKnoxville,June2018TheodoraBourniandMatLangford

Contents

Foreword | V

BrianAllen

IntroducingMeanCurvatureFlow 1

JacobBernstein

Self-similarsolutionsofmeancurvatureflow | 26

SigurdAngenentandNatasaSesum Ancientsolutionsingeometricflows | 47

IanM.Adelstein

AnextensiontotheMorseenergygradientflow 67

BeomjunChoi

Regularityofnon-compactinversemeancurvatureflow | 71

FriederikeDittberner

Areapreservingcurveshorteningflow | 75

ChristineGuenther

SecondOrderRenormalizationGroupFlow 81

Siao-HaoGuo

AnalysisofVelàzquez’ssolutiontothemeancurvatureflowwithatypeII singularity | 92

RobertHaslhofer

Somerecentapplicationsofmeancurvatureflowwithsurgery 93

BrettKotschwar

Identifyingshrinkingsolitonsbytheirasymptoticgeometries 99

SajjadLakzian

GeometricsingularitiesundertheGigli-Mantegazzaflow | 109

StephenLynch

Pinchedancientsolutionstohighcodimensionmeancurvatureflow | 116

AlexanderMramor

Ontheunknoteddnessofselfshrinkersin ℝ3 120

XuanHienNguyen

Gluingconstructionsforself-translatingandself-shrinkingsurfacesundermean curvatureflow | 123

ShengwenWang(jointworkwithJacobBernstein)

Thelevelsetflowofahypersurfacein ℝ4 oflowentropydoesnotdisconnect 132

KaWaiWong

ApplicationofMeanCurvatureFlowforsurfaceparametrizations 135

IntroducingMeanCurvatureFlow

Abstract: WewillstartbydefiningMeanCurvatureFlow(MCF)andlookingatmany importantexamples.Thenwewillmoveontodiscusstheavoidanceprinciple,scaling properties,evolutionequationsandfirstconsequences,convergencetoaroundpoint forconvexsolutions,Huisken’smonotonicityformula,singularityanalysis,anda surveyofresultsforTypeIandTypeIIsolutions.Emphasiswillbeongivingproof ideaswithreferencesratherthanshowingallofthedetailsinordertobeabletosurvey asmanyimportantpropertiesofMCFaswecanatacomfortablepace.

MSC 2010: 53C44, 53A10, 35K40

1 DefinitionofMCFandExamples

WestartbydefiningthenotionofaMeanCurvatureFlowwhichwillbethemainobject ofstudyforthisentireintroduction.

Definition1.1. Let Σn beasmooth,orientedmanifoldandlet

beaoneparameterfamilyofsmoothembeddings(immersions)thatsatisfy

wherewechoose ν tobetheoutwardpointingnormalvectorto Σt = φ(Σ, t).Asolution to (1.1) iscalledaMeanCurvatureFlow(MCF)where Σ0 istheinitialembeddingof Σ fortheflow.

MeanCurvatureFlow(MCF)isanalogoustotheheatequationonaclosedhypersurface. Forinstance,wecanconsiderheatflowonacirclewhichcanbemodeledbyafunction u : [ L, L]×[0, T]→ℝ withperiodicboundaryconditions u(L)= u( L) and ux (L)= ux ( L).Inthiscasethesolutionwillexistforalltime,theheatwillbecomeuniformover time,andequaltotheaveragetemperatureattime t = 0.Insteadofthetemperature becominguniformovertimetheexpectationunderMCFisthattheprincipalcurvatures willbecomeuniformovertime.Seefigure1foranexampleofanellipsoidwhich becomesmoresphericalovertimeunderMCF.

MCFisanonlinearequationandhencewewillfindthatforcertaininitialhypersurfacestheflowwillencountersingularities,whichcausetheflowtonolongerbe defined,beforetheflowhashadachancetocausethecurvaturetobecomeuniform.

Brian Allen, United States Military Academy, e-mail:Brian.Allen@usma.com https://doi.org/10.1515/9783110618365-001

Inthecasewheresingularitiesoccurourgoalwillbetoclassifythegeometryofthe singularitiesthatcanoccur.Thisisdonebyzoominginonthesingularityviaarescaling andobtainingdetailedestimatesontheflowbehavior.Afteradetailedunderstanding ofsingularitiesareobtainedthenwecanhopetodefineawaytocontinuetheflow throughasingularity.Inthisintroductionwewillgetaglimpseofhowtomakethese intuitiveideasandexpectationsaboutthebehaviorofMCFrigorousbysurveying classicresultsinthefield.

Fig.1: ApillshapedsurfaceevolvingunderMCFwhichbecomesmoresphericalovertime.

Remark 1.2. When Σ = S1 thenDefinition1.1iscalledCurveShorteningFlow(CSF).

Remark 1.3. SometimesMCFisdefinedasasolutionto

where ⊥ representsorthogonalprojectionto Σt insideof ℝn+1.Thedifferencebetween (1.1) and (2) isthat (2) allowsformotionindirectiontangentto Σt whichproduces reparameterizations.Fromageometricpointofviewthesetwosolutionsarethesame butfromananalyticalpointofviewthesetwodefinitioncanproducesubtledifferenceswhichwewillhavetobecarefultokeeptrackof.Wewillseeinexample1.4the importanceofthisdistinctionforgraphsover ℝn .

Let {x1 ,..., xn } belocalcoordinateson Σ withcorrespondingbasisvectors { ∂ ∂x1 ,..., ∂ ∂xn }. Forafixedtime t ∈ [0, T] wecandefinethepushforwardvectorsthroughthemap φ(⋅, t) by { ∂φ ∂x1 ,..., ∂φ ∂xn }.Nowifwelet ⟨⋅, ⋅⟩ denotetheusualinnerproducton ℝn+1 we candefinedtheinducedmetricon Σt by,

(3) andthesecondfundamentalform,

Thisallowsustodefine,

Example1.4. Let Σ = ℝn andconsiderthefunction u : ℝn ×[0, T] → ℝ sothat

φ(x1 ,..., xn , t)=(x1 ,..., xn , u(x1 ,..., xn , t) isagraphoveraplanein ℝn+1.Onecan showthat Σt willremainagraphoveraplaneforatleastashorttimeandforthe solutionsof(2)areequivalenttosolutionsofthescalarPDE,

where D and div representthegradientanddivergencein ℝn+1.Equation (7) follows fromformulasfor gij , ν, Aij , H forgraphs.Againboth (1) and (2) wouldgivethesame geometricsolutionbutinthiscase (2) allowsustoreparameterizesothat Σt remains thegraphofthesolutionof (7).Studying (7) givesustheadvantageofapplyingPDE resultsandtechniquesforscalarPDEson ℝn .

EckerandHuisken[9,10]wereabletoshowthatif Σ0 isalocallyLipschitzgraph over ℝn thenasmooth,longtimesolutionexists Σt , t ∈[0, ∞) whichremainsagraph over ℝn .Thisisparticularlyinterstingsincetheydontorequiregrowthassumptions ontheinitialfunction u(⋅, 0) whicharenecessaryundertheheatequation.Under additionalassumptionsontheinitialhypersurface,EckerandHuiskenwereableto studythelongtermbehaviorofthesolution.

Remark 1.5. ShorttimeexistencetoMCFcanbeshownbywriting Σt asagraphover theinitialhypersurface Σ0 andobtainingascalarPDEdefinedon Σ.Withcareful attentiontodetailonecanthenappealtostandardPDEresultsforshorttimeexistence ofquasilinear,parabolicPDE.See[23]forthedetails.

1.1 SelfShrinkingSolutions

InthissubsectionwewillseeourfirstexamplesofexplicitsolutionsofMCF.

Example1.6. Let Σ = Sn beasphereandconsider Σ0 = Sn r0 ⊂ ℝn+1 isasphereof radius r0 > 0.Onecanobservethat Σt willremainasphereunderMCFwithradius solvingthefollowingdifferentialequation: { { { dr dt =− n r , r(0) = r0 .

Noticethat(8)isseparableandhasthesolution, r(t)=√r2 0 2nt,

Fig.2: AspherecontractingtoapointinfinitetimeunderMCF.

Fig.3: Aselfshrinkingcylindercontractingtoaline.

fromwhichwecanobservethatthefinalexistencetimeis,

Wecanalsonoticethat

Itisimportanttonote (10), (11) and (12) sincewewillseethesequantitiesshowup againwhenestimatingmoregeneralsolutions.

Example1.7. Let Σ =ℝk × Sn k , n > k ≥ 1 beacylinderandconsider Σ0 =ℝk × S

⊂ ℝn+1 isacylinderofradius r0 > 0.Onecanobservethat Σt willremainacylinderunder MCFwithradiussolvingthefollowingdifferentialequation:

Noticethat(13)isseparableandhasthesolution,

fromwhichwecanobservethatthefinalexistencetimeis,

BoththesphereandthecylinderareexamplesofspecialsolutionstoMCFcalledself shrinkers.

Fig.4: Aselfshrinkingtoruswhichendsatapoint.

Definition1.8. WesaythatasolutiontoMCF, Σt ,isa selfshrinker if, Σt =√T tΣ0 , t ∈[0, T]. (16)

Noticethatbedefinitionaselfshrinkermustendinasingularityattime T.Nowwe willseethatselfshrinkerscanbefoundbystudyingaparticularellipticPDE.

Theorem1.9. Onecanshowthat Σt isaselfshrinkerifandonlyifthemeancurvature of Σ0 satisfiestheequation,

Σ0 (p)=−

φ(p, 0) x0 , ν0⟩ 2 , (17) where x0 ∈ℝn+1 , p ∈ Σ0,and ν0 istheoutwardpointingnormalto Σ0.

Equation17isanellipticPDEincoordinatesandexistenceofsolutionsisimportant tounderstandingspecialsolutionsofMCF.Byimposingvarioussymmetriesonthe equationforaselfshrinkerwecanoftenreducetheobjectiveoffindingselfshrinkers downtostudyinganODEaswewillseeinExample1.11.

Remark 1.10. Sincewecanalwaystranslatetime,i.e.define s = T t, s ∈( ∞, 0],and translatein ℝn+1,i.e.move x0 totheorigin 0 ∈ℝn+1,wecanequivalentlydefineaself shrinkerbythepropertythat,

withdefinedequation,

Example1.11. In[3],Angenentfoundaselfshrinkingtoruswhichisrotationally symmetricbysolvingtheODEwhichresultsfromimposingrotationalsymmetryon theselfshrinkerequation (1.9).Thequalititativebehaviourofthissolutionisdepicted infigure4andthecurvewhichisrotatedinordertogeneratethisspecialtorusis depictedinfigure5.In[3],Angenentalsofoundafamilyofimmersedselfshrinking torussolutionsofMCF.

Example1.12. Infigure6weseeanexampleofarotationallysymmetrictorusin ℝ3 whichshrinkstoacircleatthefinaltime T.ThisisanexampleofasolutionsofMCF whichdoesnothavenon-negativemeancurvatureeverywhere.

Fig.5: ThecurvewhichisrotatedaroundtheverticalaxisinordertoproduceAngenent’storus (Figurefrom[7]).

Fig.6: Torusshrinkingtoacircle.

1.2 AvoidancePrinciple

NowthatwehavegatheredausefulcollectionofexamplesolutionsofMCFwewill nowseetheirusefulnessforstudyinggenericsolutionsofMCF.

Theorem1.13 (AvoidancePrinciple). Anytwosmooth,compactsolutionsofMCFwhich areinitiallydisjointwillremaindisjointunderMCF.

OneshouldseeEcker’sbook[8]orMantegazza’sbook[23]foradiscussionoftheproof ofTheorem1.13.Animportantconsequenceoftheavoidanceprincipleisthatcompact solutionstoMCFmustexistforafinitetime T <∞.Onecanseethisbychoosing r0 > 0 largeenoughsothat Σ0 ⊂ Sn r0 andhencebetheavoidanceprinciple Σt ⊂ Sn r(t) which implies T ≤ r2 0 2n .Thisrealizationmotivatesthequestionofwhatthesesingularitiesof MCFwilllooklikefordifferentcompactinitialhypersurfaces Σ0?

Fig.7: Adumbellshapedinitialsurfacedevelopinganeckpinchsingularity[2].

Example1.14. Onecanusetheavoidanceprinciple,thespheresolution,andthe angenenttorusinordertoshowthataneckpinchmustoccurforadumbellshapeas giveninfigure7.ThisargumentwasgivenbyAngenentin[3]whereoneplacestwo spheresinsidethebulbsofthedumbellandplacesanAngenenettorusontheoutside oftheneck.Iftheradiiofthebulbsandneckofthedumbellarechosenproperlythen onecanshowthattheexistencetimefortheself-shrinkingtorusisstrictlylessthan thatofthespheres.Bytheavoidanceprinciple,allfoursolutionsofMCFmustremain disjointandhencethesphereswillensurethatthebulbsremainun-collapsedwhereas theneckmusthaveshrunktoapointatthesametimeorbeforetheshrinkingtorus. HenceaneckpinchlikeinExample1.14musthaveoccurred.

Fig.8: AdumbellwithspheresinsidethebulbsofthedumbellandaAngenenttorussurroundingthe neckwhichcanbeusedtoshowasingularitymustoccur(Figurefrom[6]).

Inthecaseoftheneckpinchexample1.14itseemsreasonablethatifoneweretozoom inonthesingularityas t → T onewouldseeahypersurfacewhichlooksmoreand morelikeacylinder.Oneveryimportantresearchdirectionistoclassifysingularities ofMCFbyrescalingtheflowinordertozoomin.Wenowdescribeoneveryimportant waytorescaleasolutionofMCF.

1.3 RescalingMCF

Ifwedefine φ(p, t)= λφ(p, t), λ ∈ℝ then

Itisagoodexerciseingeometricanalysiscalculationstochecktherescaledequations givenabove.

Ifwedefinetheparabolicrescaling,

then,

(27) andhence φ isalsoasolutionofMCF.

Remark 1.15. Noticethatforlargeλ,ifthesingularityfor Σt occurredat p ∈ℝn+1 then therescaling,

(28)

hastheeffectofzoominginonthesingularitythatisoccurringat p overalargerportion oftime.Thisisnottheonlywaytorescalesolutionsinordertoclassifysingularities andwewillseeanothercommonrescalinglater.

1.4 TranslatingSolutions

InthissubsectionwewilllookatanotherspecialfamilyofsolutionstoMCFwhich evolveinasimpleway.Westartwithanexampleofthegrimreapersolution.

Example1.16. ThegrimreapersolutionofMCF, Γ1 t ⊂ℝ2,isgivenasthegraphofthe function u : ( π/2, π, 2)×ℝ→ℝ definedas, u(y, t)=− log(cos(y))+ ty ∈( π/2, π/2). (29)

Thissolutionevolvessimplybytranslating Γ0 ⊂ℝ2 inthe x-directionofthe xy-plane andisdefinedforallvaluesof t ∈ℝ.Wecanusethegrimreapersolutiontodefine higherdimensionalsolutionsofMCFbydefining Σn t = Γt ×ℝn 1 whichalsoevolvesby translating.

Fig.9: Grimreapercurveboundedbyapairofhorizontallines.

Ifwelet Πt bethegrimreaperthenwecanconstructothertranslatingsolitons

n t = Πt ×ℝn 1 . (30)

ThegrimreaperisanexampleofaspecialsolutiontoMCFcalledatranslatingsoliton.

Definition1.17. WesayasolutiontoMCFisatranslatingsolitonif,

t = Σ + tv, (31) for v ∈ℝn+1 , t ∈ℝ.

Similartoselfshrinkingsolutions,translatingsolitonscanalsobecharacterizedbyan ellipticequation.

Theorem1.18. Σt isatranslatingsolitonifandonlyif HΣ0 =⟨ν0 , v⟩. (32)

Example1.19. Anotherexampleofatranslatingsolitonisgivenbythebowlsoliton whichisarotationallysymmetric,paraboloidshapedhypersurfacewhichevolvesby translating(Seefigure10).

WewillseethattheselfshrinkersandthetranslatingsolitonsareimportantforclassifyingsingularitiesofMCF.

Fig.10: Paraboloidshapedbowlsolitonwhichevolvesbytranslation.

2 EvolutionEquationsandFirstConsequences

Inthissectionwewilllookatimportantevolutionequationsforgeometricquantities underMCF.Whentheseevolutionequationsarecombinedwiththemaximumprinciple wewillbeabletoderiveimportantgeometricestimateswhichallowustocontrolMCF solutionsforspecifiedclassesofinitialhypersurfaces.

Tobeginwecancalculatetheevolutionofsomebasicgeometricobjectsunder MCF.Byusingthedefinitionof,

,wecanderivethefollowingtime derivatives:

where ∇ isthecovariantderivativeof Σt .Oneshouldseetheappendixof[8]fora derivationofthesetimederivatives.Itisalsoagoodexercisetoderivetheseequations byusingthedefinitionofMCF.

2.1 GradientFlowofArea

Ifwelet φ : Σ ×( ε, ε) → ℝn+1 beaoneparameterfamilyofsmoothembeddings whichsatisfiestheequation

forsomespeedfunction f : Σ ×( ε, ε)→ℝ.Thenwecancalculatethefirstvariation ofarea

Oneshouldsee[22]foradetailedderivationofthevariationofareaformula.Notice thatthecriticalpointsfortheareafunctionalsatisfytheequation H = 0 whichare calledminimalsurfaces.MinimalsurfacesaretrivialexamplesofsolutionsofMCF whichdonotchangethroughouttheirevolution.

Bypluggingin f =−H wefindtheevolutionforareaunderMCF,

andhencewenoticethatMCFalwaysdecreasesareaandisthegradientflowforthe areafunctional.

Remark 2.1. CommonrescalingsoftheMCFaimtokeeptheareaorenclosedvolume of Σt fixedalongtheevolution.Noticethatifwechoosethespeedin(5)tobe,

thenunderthecorrespondingflowwewouldfind,

Hencetheflowdefinedbythespeed (8) wouldkeeptheareaof Σt fixedalongthe evolutionandcanbeshowntobecloselyrelatedtoMCF.Asimilarrescalingwasused byHuiskeninhisoriginalMCFpaper[16]studyingconvexsolutionstoMCF.

2.2 Hamilton’sMaximumPrinciple

InordertofurtherunderstandhowgeometricquantititesevolveunderMCFwewill usethemaximumprincipleinordertoderiveestimates.WenowintroduceHamilton’s maximuprinciplewhichisparticularlyimportantwhenquantitiesareeasiertocontrol atamaximumoraminimum.

Theorem2.2 (Hamilton’sMaximumPrinciple). Assume g(t), t ∈ [0, T] isasmooth familyofRiemannianmetricsonaclosedmanifold M.Let f : M ×[0, T)→ℝ beasmooth

functionsatisfying,

where X isacontinuousvectorfieldand b isalocallyLipschitzfunction.Setting,

wehavethat umax islocallyLipschitz,hencedifferenriableatalmosteverytime t ∈[0, T), andateverydifferntiabletimesatisfies,

Thisimpliesthatif g : [0, T󸀠)→ℝ isasolutionto

OneshouldseeHamiltonswork[14, 15]andMantegazza’sbook[23]fortheproof ofthistheorem.Oneimportantaspectofthismaximumprincipleisthatyouonly needtomakethecomparisontotheODEfor g atpointsofmaximumfor f .Thisis importantsincethereisoftenaspecialgeometrythatcanbeexploitedatmaximums ofgeometricquantities.Wewillillustratethisbyprovingthefollowingweakerversion oftheavoidanceprincipleTheorem1.13.

Theorem2.3. If Σ0 ⊂ Sn r0 for r0 > 0 then Σt ⊂ Sn r(t) where r(t)=√

Proof. Considerthefunction

andcompute,

Nowwenotethatatamaxfor φ onecanshowthat φ and ν areparellelvectorsand H ≥ n |

| whichyieldsthesimplerequation

Nowbyintegrating(16)wefind,

whichyieldstheestimate,

fromwhichtheresultfollows.

2.3 MaximumPrincipleEstimates

Inordertoobtainfurtherestimatesofgeometricquantitiesviathemaximumprinciple wewillneedafurtherlistofevolutionequations.

Wenotethatif T and J aretensorsthen T ∗ J representssomeundeterminedtraceof T and J.Oneshouldsee[8]and[23]fordetailedderivationsoftheseevolutionequations. Asafirstconsequenceoftheseevolutionequationsweseethatmeanconvexityis preservedunderMCF.Wesayahypersurface Σ ⊂ℝn+1 ismeanconvexif H ≥ 0.

Theorem2.4. If Σ0 ismeanconvexthen Σt remainsmeanconvexunderMCF.

See[23]foraproofofthisclaim.Nowthatweknowthat H ≥ 0 ispreservedwecanuse thisinformationtogetamoredetailedestimateofmeancurvature.

Theorem2.5. If Σ0 ismeanconvexwith H0 = minΣ0 H > 0 then

(25) If H0 = 0 thenasimilarestimateholdswith H0 replacedwith H󸀠 = min

H for t

∈[0, ε).

Proof. Notethat |A|2 ≥ H2 n andsobycombiningwiththeevolutionequationfor H we find,

Henceatalmostevery t ∈[0, T] wefind,

whichbyintegratingyields,

Theorem2.6. If Σ0 satisfies |A|≤ CH forsome C > 0 then |A|≤ CH on Σt for t ∈[0, T).

Thisresultfollowsbycombiningtheevolutionequationfor |A|2 withtheevolution equationfor H2.Oneshouldsee[16]or[23]forthedetails.Theorem2.6impliesthatif |A| blowsupatacertainratethen H wouldalsoblowupwithatleastthatsamerate. Thenextresultsaysthatif A isnotboundedthen maxΣt |A| mustblowupwithsome minimumrate.

Theorem2.7. If A isnotboundedas t → T <∞, Σ0 compact,thenthesolutionofMCF, Σt ,mustsatisfy,

(29) forall t ∈[0, T).

WecanobtainfurtherestimatesforsolutionsofMCFbyintroducingHamilton’stensor maximumprinciple.Webeginwithanimportantdefinitionwhichwillbeusedasan assumptioninthemaximumprinciple.

Definition2.8. Let A beasymmetric 2-tensorand g aRiemannianmetricon M.Let B beasymmetric 2-tensorsothat B = B(A, g).Wesaythat B satisfiesthe nulleigenvector assumption ifforanynullvector V for A,i.e. A(V, ⋅)= 0,wehavethat B(V, V)≥ 0.

Theorem2.9. Let M beaclosedmanifoldand g(t) asmoothfamilyofRiemannian metricson M.Let T beasymmetric 2-tensorwhichsatisfiestheevolutioninequality,

(30) where X(t) isatimedependentvectorfieldand

(31)

isasymmetric 2-tensorwhichislocallyLipschitz.Supposethat B satisfiesthenulleigenvectorassumptionand T isnonnegativedefiniteon Σ0.Then T isnonnegativedefiniteon Σt forall t ≥ 0.

OneshouldseethebookofChow,Lu,andNi[4]foraproofofthisresultaswell asapplicationstoRicciflow.Wenowsurveyafewgeometricconditionswhichare preservedunderMCFwhichcanbeshownbyusingthetensormaximumprinciple.

Theorem2.10. If A isnonnegativedefinite(convex)on Σ0 then A isnonnegativedefinite (convex)on Σt .

Proof. TheproofthatconvexityispreservedfollowsfromHamilton’stensormaximum principle,Theorem2.9,appliedtotheevolutionequationfor Aij .Inthiscasewesee that,

Nowassumethat v ∈ Tp Σt isanulleigenvectorfor A,i.e. Aij vi = 0 incoordinates.Then wecancalculate,

andhence B satisfiesthenulleigenvectorassumptionwhichimpliesthat A remains nonnegativedefinitealongtheflow.

Thisresultsaysthatconvexityof Σ0 ispreservedalongasolutionofMCFwhichwas usedbyHuiskeninhisoriginalMCFpaper[16].

Theorem2.11. If A ≥ αHg, α > 0 on Σ0 then A ≥ αHg on Σt .

ThisresultshowsthateigenvaluepinchingispreservedunderMCFwhichwasalso provedandusedbyHuiskenin[16].

Theorem2.12. Let λ1 ,..., λn betheeigenvaluesof A inincreasingorder.If λ1+...+λk ≥ 0 on Σ0 then λ1 + ... + λk ≥ 0 on Σt .

Thisresultshowsthat k-convexityispreservedalongsolutionsofMCF.Huiskenand Sinestrari[21]usedthefactthat 2-convexityispreservedunderMCF,combinedwith importantestimatesof 2-convexsolutions,inordertodefineasurgeryprocedurewhich allowsonetoclassifythetopologyof 2-convexhypersurfacesof ℝn+1 .

2.4 ContinuationCriterion

Inthissectionwewouldliketogiveacriterionfor T tobethefinalexistencetime. Tothisendwefirstshowthatif |A| isboundedon [0, T) thenallofitshigherorder covariantderivativesareboundedaswellonthesametimeinterval.

Theorem2.13. If Σ0 isclosedand |A|2 isboundedon Σt forall t ∈[0, T) then, |∇k A|2 ≤ Ck , (34) on Σt forall t ∈[0, T) forall k ∈ℕ.

Proof. Onewaytoprovethisresultisbyinductionon k,thenumberofcovariant derivativesof A.Inthiscasethebasecaseisassumedinthestatementsince |A|2 ≤ C. Fortheinductivestepweassumethat |∇j 1 A|2

andnote thatwehavetheevolutionequationfor |∇k A|2 ,

wherewehaveusedtheinductionhypothesisin (37).Nowwenoticethattheevolutionequationfor |∇k 1 A|2 hasagoodnegative |∇k A|2 term.Hencewecandefine f

andcalculate,

Nowbyapplyingthemaximumprincipleto(38)wefindthat,

Thisresultallowsustoproveacontinuationcriterionfortheflowwhichimpliesthat T canonlybeasingulartimeforMCFif max Σt |A| blowsupas t → T.

Theorem2.14 (ContinuationCriterion). If Σ0 iscompactand |A|≤ C on

t for t ∈[0, T), T <∞ then T cannotbeasingulartimeforthemeancurvatureflow Σt .

ProofIdea. Assumethat T isasingulartimefortheflow Σt .Byassumptionweknow that |A| isboundedon [0, T) andhencebyTheorem2.13weknowthat |∇k A| isbounded forall t ∈[0, T) and k ∈ℕ.Theboundon |A| alsoimpliesthat H isboundedwhich impliesthatforevery 0 ≤ s < t < T

Thisshowsthatthemaps φ(⋅, t) areLipschitzintimeandhenceuniformlyconvergeto acontinuousmap φT : Σ →ℝn+1 as t → T.

Nowwecanusethehigherorderboundson |∇k A| toarguethat φT isasmooth immersionandhencebyshorttimeexistencewecanextendthesmoothflowpast T

whichcontradictsthefactthat T wasasingulartimefortheflow.Oneshouldnotethat alotoftechnicalworkgoesintothislaststepwhichwehavesimplysketchedtheidea ofhere.Oneshouldsee[23]or[16]formoredetails.

2.5 SurveyofFirstMCFResults

InthissectionwesurveyresultsforMCFinthecasewheretheoriginalintuitionthat MCFshouldcausethecurvaturetobecomeuniformovertimeplaysoutperfectly.In thesecasesthesingularitiesarewellbehavedandcanbeuniformlyrescaledaway.

Theorem2.15 (GageandHamilton[11,12]). UndertheCurveShorteningFlow(CSF)a convex,closedcurvein ℝ2 smoothlyshrinkstoapointinfinitetime.Afterrescalingin ordertokeepthelengthconstantthesolutionsmoothlyconvergestoacircle.

Theorem2.16 (Husiken[16]). UnderMCFacompact,convexhypersurfacein ℝn+1 , n ≥ 2 smoothlyshrinkstoaroundpointinfinitetime.Afterrescalinginordertokeep areaconstantthesolutionsmoothlyconvergestoasphere.

Theorem2.17 (Grayson[13]). Let γt betheCSFofaclosed,embedded,smoothcurvein ℝ2 withmaximalintervalofsmoothexistence [0, T).Then ∃τ < T suchthat γτ isconvex.

NoticethatbycombiningTheorem2.15withTheorem2.17weseethatanyCSFofa closed,embedded,smoothcurvein ℝ2 willshrinktoaroundpointinfinitetime.This isasurprisingresultandisspecialtothecaseofcurves.Inthecaseofanon-convex hypersurfacein ℝn+1 weknowthatsingularitiescandevelopingeneralandwillbethe focusofthenextsection.

Remark 2.18. ItisinterestingtonotethattheproofofTheorem2.16givenbyHuisken in[16]doesnotapplytothecaseofcurvesin ℝ2.ThisisbecauseHusikenleverages theevolutionequationforthetracefreesecondfundamentalform A∘ = A 1 n Hg.If onelets λ1 ,..., λn betheeigenvaluesof A thenonecanshowthat, |A∘|2 = 1 2 n ∑ i<j (λi λj )2 , (43)

andhenceifonecanshowthat |A∘|2 → 0 as t → T thenonewillconcludethatthe principalcurvaturesarebecomingasymptoticallyidenticalas t → T.Noticethatif n = 1 thenthetracefreesecondfundamentalformistrivialandhenceitisonlyuseful for n ≥ 2.

3 SingularityAnalysis

Oneshouldrememberthedumbellexample1.14whereasingularityoccursonacylindricalneck.OurgoalnowistostudysolutionsofMCFwherewewillbeabletoshow thattheonlytypeofsingularitiesthatcanoccurarespheresorgeneralizedcylinders afterrescaling.

3.1 TypeIvs.TypeIISolutions

RememberTheorem2.7whichstatesthatforcompactsolutionsofMCF, Σt ,wehave that

Wewillnowrestrictoutattentiontosolutionswhose maxΣt |A| isboundedabove bythesamerate.

Definition3.1. Let T bethemaximumexistencetimeforthesolutionofMCF, Σt .We willsaythat Σt is TypeI if ∃C > 1 sothat

Wesaythat Σt is TypeII ifinstead

NowthatwehavedefinedthenotionofTypeIandTypeIIsolutionsofMCFitisnatural toask:DoesaTypeIIsolutionofMCFexist?

BymodifyingthedumbellexampleonecanshowthatindeedaTypeIIsolution doesexist.Wewillgiveaninformaldiscussionoftheidea.Imagineadumbellwith onebulbofradius r1,theotherbulbofradius 1,andaneckofradius r2.

Ifwechoose r1 relativelysmalland r2 closeto 1 thenoneexpectstheMCFto eventuallybecomeconvexandeventuallyconvergetoaroundspherebyHuisken’s result[16].Ifwechoose r1 closeto 1 and r2 relativelysmallthenoneexpectstheMCF todevelopaneckpinchasinexample1.14whichwewillbeabletoshowisTypeI. Onecanshowthatthereisasolutioninbetweenthefirsttosituations,withcarefully chosen r1 and r2 value,sothattheMCFdevelopsaTypeIIsolution.

InthissectionwewillfocusonTypeIsolutionsofMCF.Oneshouldnoticethatin theexampleabovehowwehadtofinetunetheparametersinordertofindaTypeII solutionofMCF.Thisraisesthequestion:HowgenericareTypeIIsolutionsofMCF? ThisquestionhasbeenextensivelystudiedbyColdingandMinicozzi[5]wherethey showthatTypeIsolutionsofMCFaregeneric.Fortheprecisemeaningofgeneric solutionsoneshouldsee[5].

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.