ebook img

Global existence and convergence of $Q$-curvature flow on manifolds of even dimension PDF

0.33 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Global existence and convergence of $Q$-curvature flow on manifolds of even dimension

GLOBAL EXISTENCE AND CONVERGENCE OF Q-CURVATURE FLOW ON MANIFOLDS OF EVEN DIMENSION QUO´ˆCANHNGOˆ ANDHONGZHANG Abstract. Usinganegativegradientflowapproach,wegeneralizeandunifysomeexis- 7 tencetheorems fortheproblem ofprescribing Q-curvature firstbyBaird, Fardoun, and 1 Regbaoui(Calc. Var.2775-104)for4-manifoldswithapossiblesign-changingcurvature 0 candidatethenbyBrendle(Ann. Math. 158323-343)forn-manifoldswithevennwith 2 positivecurvaturecandidate tothecaseofn-manifoldsofallevendimensionwithsign- changingcurvaturecandidates. MakinguseoftheŁojasiewicz–Simoninequality,wealso n addresstherateoftheconvergence. a J 9 ] P 1. Introduction A On a closed manifold (Mn,g) of dimension n > 3, a formally self-adjoint geometric . h differentialoperatorA beingcalculatedwiththemetricgiscalledconformallycovariant g t a ofbidegree(a,b)if m A (ϕ)=e bwA (eawϕ) (1.1) gw − g [ for all ϕ C∞(Mn) where gw := e2wg is a conformal metric conformally to g, usually ∈ 1 denotedbyg [g]. w ∈ v Ofimportanceinconformalgeometryisthesecond-orderconformalLaplaciandefined 7 4 onclosedmanifoldsofdimensionn>3. Tobeprecise,ifwedenote 2 n 2 2 L := ∆ + − R g g g − 4(n 1) 0 − . where R is the scalar curvatureof g, then L is conformallycovariantof bidegree((n 1 g g 0 2)/2,(n+2)/2)because − 7 L (ϕ)=e (n+2)w/2L (e(n 2)w/2ϕ) (1.2) 1 gw − g − v: forallϕ C (Mn). Uponusingthechangeofvariableu4/(n 2) =e2w,Eq. (1.2)leadsusto ∞ − ∈ i X L (ϕ)=u (n+2)/(n 2)L (uϕ) gw − − g r a forallϕ C∞(Mn).Bysettingϕ 1,thescalarcurvatureRgobeystheso-calledconformal ∈ ≡ change 4(n 1) − n −2 ∆gu+Rgu=Rguu(n+2)/(n−2) (1.3) − whereg =u4/(n 2)g. Intheliterature,Eq. (1.3)forunknownuisknownastheprescribed u − scalarcurvatureproblemwiththeprescribedfunctionR . Thischallengingproblemhas gu already been captured much attention by many mathematicians in the last few decades. NoticethataspecialcaseofEq.(1.3)whenR isconstantisthefamousYamabeproblem. gu Sincethesetwoproblemswereinthecoreofconformalgeometry,wedonotmentionany particularreferencesinceitcanbeeasilyfoundintheliterature. Date:10thJanuary,2017at01:24. Keywordsandphrases. Łojasiewicz–Simon inequality, Q-curvature, negative gradient flow, closed mani- folds,evendimension. 1 2 Q.A.NGOˆ ANDH.ZHANG Thefirsthigh-orderexampleofconformaloperatorsisthefourth-orderPaneitzoperator P4for(M4,g),discoveredby[Pan83]. Thisoperatorisgivenasfollows 2 P4 =( ∆ )2 div R g 2Ric d , (1.4) g − g ·− g 3 g − · (cid:16)(cid:0) (cid:1) (cid:17) where d is the differential and Ric denotes the Ricci tensor. (Note that the subscript 4 denotestheorderofthePaneitzoperatorbeingused,notthedimensionoftheunderlying manifold.)Clearly,thePaneitzoperatorP4isconformallycovariantofbidegree(0,4)since P4 (ϕ)=e 4wP4(ϕ) (1.5) gw − g forallϕ C (M4)whereg =e2wg. Uponlettingϕ 1,wededucethat ∞ w ∈ ≡ Q =e 4u(P4u+Q ), gw − g g wherethequantityQisgivenby 1 Q = (∆R R2+3 Ric 2). g −6 g− g | g| The quantity Q is called the Q-curvature associated with the Paneitz operator P4 on g (M4,g), which was discovered by Branson [Bra85]. One of interesting features of the Q-curvature,bytheGauss–Bonnet–Cherntheorem,isthatitobeysthefollowingidentity 1 Q dµ + W 2dµ =8π2χ(M), Z g g 4Z | g|g g M M whereχ(M) isthe Euler-characteristicof M4 andW is the Weyltensor. Since the Weyl g tensor W is conformally invariant, the total Q-curvature Q dµ is also conformally g M g g invariant;thatisthevalue R k := Q dµ 4 g g Z M is independent of the metric g but the conformal class [g] represented by g. (Note that the subscript 4 in k indicates the order of the Paneitz operator which then defines Q .) 4 g SimilartoAubin’sthresholdintheYamabeproblem,thereisathreshold16π2fork when 4 workingwithQ-curvaturesonM4. (Thenumber16π2,whichis3!vol(S4),isexactlyequal to S4QgS4dµgS4 beingcalculatedwithrespecttothestandardmetricgS4 onS4.) RFor general even dimension n > 4, Graham, Jenne, Mason and Sparling [GJMS92] discovered a similar operator Pn which is a conformally invariant self-adjoint operator g of order n with leading term ( ∆)n/2. This operator is commonly known as the GJMS operator. The work [GJMS92]−was based on an earlier work by Fefferman and Graham [FG85]. Moreprecisely,theGJMSoperatorisconformallycovariantofbidegree(0,n)in thesensethat Pn (ϕ)=e nwPn(ϕ) (1.6) gw − g forallϕ C (Mn)whereg =e2wg.Inaddition,thereiscurvaturequantityQnassociated ∈ ∞ w g with the operator Pn, analogue with the Q-curvature Q4, which also satisfies a similar g g conformalchange Q =e nu(Pnu+Q ). (1.7) gu − g g Inthispaper,weconsidertheproblemofprescribingQ-curvatureonaclosedmanifold of even dimension. Let us first describe the problem in details. Let (Mn,g ) be closed 0 manifoldsofevendimensionnwithbackgroundmetricg .Thisproblemcanbeformulated 0 as follows: Let f be a smooth function on M. The problem is to ask if there exists a pointwise conformal metric g [g ], that is g = e2ug for some u, such that f can be 0 0 ∈ realizedastheQ-curvatureofg.Thankstotheruleofconformalchange(1.7),thisproblem isequivalenttosolvingthehighordernonlinearequation P u+Q = fenu, (1.8) 0 0 Q-CURVATUREFLOWONMANIFOLDSOFEVENDIMENSION 3 where,forsimplicity,wesetP :=Pn andQ = Qn . 0 g0 0 g0 For f > 0 keeping sign, Brendle [Bre03] used a flow method to show that Eq.(1.8) hasasolutionprovidedthatP is(weakly)positivewithkernelconsistingoftheconstant 0 functionsand Q dµ =:k <(n 1)!ω , Z 0 g0 n − n M whereω isthevolumeofunitsphereinRn+1. HerebypositivityofP wemean n 0 u P u dµ >λ (P ) (u u)2dµ >0 (1.9) Z · 0 g0 1 0 Z − g0 M M forallsmoothfunctionuwhereλ (P )isthefirstnontrivialeigenvalueoftheoperatorP 1 0 0 andu = (vol(M,g )) 1 udµ . (Whenn = 4,thethreshold(n 1)!ω fork coincides 0 − M g0 − n n with 16π2, the thresholRdfork .) More precisely,givenanyinitialdata u , he considered 4 0 theevolutionofmetricsasfollow ∂g(t) = Q MQ0 dµg0 f g(t) fort>0, Then,Brendleshowedg∂th(t0a)tg=iev−2eun(cid:18)0ga0gn.(yt)−u0RtRhMeflfodwµg(01.1(cid:19)0)hasasolutionwhichexistsf(o1r.1a0l)l timeandconvergesatinfinitytoametricg with ∞ Q dµ Q = M 0 g0 f. g R ∞ M f dµg0 R Later on, Baird et al. [BFR06] consideredthe problemof prescribing Q-curvatureon 4- manifolds, where they allow f to change sign. They adoptan abstract negativegradient flowwhichisdifferentfromBrendle’storealizeit. Preciselyspeaking,theyconsideredthe functional J[u]= u P u dµ +2 Q u dµ , Z · 0 g0 Z o g0 M M ontheSobolevspaceH = H2(M,g )undertheconstraint 0 u X := u H : e4uf dµ = Q dµ . ∈ (cid:26) ∈ ZM g0 ZM 0 g0(cid:27) Then,theythoughtofXasahypersurfaceoftheSobolevspaceHandstudiedthenegative gradientflowforconformalfactorswithrespectivetothishypersurfacebelow ∂u= XJ(u) t −∇ (1.11) u(0)=u X.  0 ∈ Clearly,iftheflow(1.11)existsforalltimeandconvergesatinfinity,thenthelimitfunction u yieldsasolutionof(1.8)withn=4. ∞ Forgeneralmanifoldsof even dimension,there were two waysto improveBrendle’s. First,byusingavariationalapproach,Bairdetal.[BFR09]improvedBrendle’sbyrelaxing thepositivitypropertyofP andallowing f >0.However,theykeptusingtheassumption 0 that P has kernel consisting of constant functions. Note that positivity property of the 0 operator P0 on (Sn,gSn) is no longer needed as it automatically holds; see Appendix C. In another work of the same research group, Fardoun et al. [FR12] improvedBrendle’s byassumingthepositiveoperatorP hasnon-trivialkernelinthesensethatitskernelhas 0 dimensionatleast2. Inspiredby the work [BFR06] for sign-changingcandidatesin low dimensionand by [Bre03] for positive candidatesin high dimensions, the aim of this paper is to study the prescribing Q-curvatureproblemonall evendimensionalmanifoldswherethe candidate curvature f is allowed to change sign. The method here we used is again the negative 4 Q.A.NGOˆ ANDH.ZHANG gradient flow; see also [NX15]. However, instead of the abstract negativegradient flow method(1.11),weusingaflowwhichisananalogueof (1.10). Inthefollowing,wewill describeourflowmethodandstatethemainresultofthepaper. Asin[BFR06],weconsiderthefollowingenergyfunctionalforconformalfactors n E[u]= u P u dµ +n Q u dµ (1.12) 2Z · 0 g0 Z 0 g0 M M onasubsetoftheSobolevspaceHn/2(M,g )givenbythefollowingconstraint 0 u u Hn/2(M,g ): fenu dµ = Q dµ =:Y. (1.13) ∈ ∈ 0 Z g0 Z 0 g0 n M M o Because P has the leading term ( ∆)n/2, it is not hardto see thatthere exists a positive 0 − constantCdependingonlyonM,suchthatforanyu Hn/2(M,g )wehave 0 ∈ C 1 u ( ∆ )n/2udµ 6 u P u dµ 6C u ( ∆ )n/2udµ . (1.14) − Z · − 0 g0 Z · 0 g0 Z · − 0 g0 M M M NoticethatanusualnormforHp(M,g )isgivenasfollows 0 u 2 = ( ∆ )p/2u2dµ + u2dµ . (1.15) k kHp(M,g0) Z | − 0 | g0 Z g0 M M However,inviewof(1.14)andsincetheGJMSoperatorP ispositivewithkernelconsist- 0 ingofconstantfunctionswecandefineanequivalentnormonHn/2(M,g )asthefollowing 0 u 2 = u P u dµ + u2 dµ . (1.16) k kHn/2(M,g0) ZM · 0 g0 ZM g0 Now, we define our negativegradient flow for the functional(1.12) under the constraint (1.13)asfollows ∂ g(t)= 2(Q λ(t)f)g(t), (1.17) ∂t − g(t)− wheretheparameterλischosen,insteadofpreservingthevolumeofmetricg(t)in(1.10), tofixthequantity f dµ foralltime. Adirectcalculationfor d f dµ =0shows M g(t) dt M g(t) R R f(λ(t)f Q )dµ =0. (1.18) g(t) g(t) Z − M Hence,thenaturalchoiceofλis fQ dµ λ(t)= M g(t) g(t). (1.19) R f2 dµ M g(t) R Consequently,the constraint(1.13) is preservedalongthe flow. Up to thispoint, we can stateourmainresultasfollows. Theorem 1.1. Let (M,g ) be a compact, oriented n-dimensional Riemannian manifold 0 withneven. AssumethattheGJMSoperatorP ispositivewithkernelconsistingofcon- 0 stantfunctions.Moreover,assumethatthemetricg satisfies 0 Q dµ <(n 1)!ω . Z 0 g0 − n M Also,wechoosetheinitialmetricg(0)=e2u0g0withu0 Y andlet f beasmoothfunction ∈ onMsuchthat (i) sup f(x)>0 if Q dµ >0, x∈M ZM 0 g0  (ii) supx∈M f(x)>0andinfx∈M f(x)<0 ifZMQ0dµg0 =0,  (1.20) (iii) supx∈M f(x)6C0and infx∈M f(x)<0 ifZMQ0dµg0 <0,  Q-CURVATUREFLOWONMANIFOLDSOFEVENDIMENSION 5 where C is a positive constant depending only on f = max( f(x),0),u ,g and M. 0 − 0 0 − Then, the flow (1.17) has a solution which is defined for all time and converges in C - ∞ topologytoametricg with Q = λ f astimetendstoinfinity,whereλ isaconstant and, in addition, equa∞lto 1 if ∞ Q d∞µ , 0. Moreover, if we write g(t)∞= e2u(t)g and M 0 g0 0 g = e2u g , thentheconvergeRnceg(t) g isnotworsethanapolynomialrateinthe se∞nsetha∞tth0ereexisttwopositiveconsta→ntsB∞andβsuchthat u(t) u 6 B(1+t) β C (M) − k − ∞k ∞ for allt > 0. Finally, if Q dµ < 0 and f 6 0, then the convergenceg(t) g is M 0 g0 → ∞ exponentiallyfast. R WenowconsiderthecaseM =Sn,thestandardn-sphere.LetGbeagroupofisometries ofSn. Recallthatafunction f issaidtobeG-invriantifitsatisfies f(σx)= f(x) for all σ G and x Sn. Furthermore, we say that a conformal metric g to gSn is G- invarianti∈fuisaG-in∈variantfunctionwhengiswrittenasg = e2ugSn. LetΣbethesetof fixedpointsofG,thatis, Σ= x Sn :σx= xforallσ G . { ∈ ∈ } Ournextresultisasfollows. Theorem 1.2. Let f be a smooth and G-invariant function on Sn with sup f > 0. x Sn Choosetheinitialmetricg(0)=e2u0gSn withu0 Y beingG-invariant.Ifeither∈ ∈ (a) Σ= or ∅ (b) E[u ] sup f(x)6(n 1)!exp 0 , x∈Σ − n− (n−1)!ωno then the flow (1.17) has a solution which is defined for all time and converges in C - ∞ topology to a G-invariant conformal metric g with Q = f as time tends to infinity. Moreover,ifwewriteg(t)=e2u(t)gSn andg∞ =e∞2u∞gSn,th∞entheconvergenceg(t)→g∞is notworsethanapolynomialrateinthesensethatthereexisttwopositiveconstantsBand βsuchthat u(t) u 6 B(1+t) β C (M) − k − ∞k ∞ forallt>0. RecallthatonSn, theGJMSoperatorP ispositivewithkernelconsistingofconstant 0 functions.AsadirectconsequenceofTheorems1.1and1.2,wehave Corollary 1.3. Let (M,g ) be a compact, oriented n-dimensional Riemannian manifold 0 withneven. AssumethattheGJMSoperatorP ispositivewithkernelconsistingofcon- 0 stantfunctions. Let f beanon-constant,smoothfunctionon M. Then,thefollowinghold true. (i) If Q dµ < 0, then there exists a conformal metric to g with associated M 0 g0 0 Q-Rcurvature f given that f satisfies inf f(x) < 0 and there exists a positive x M constantCdependingonlyon f ,g and∈Msuchthatsup f(x)6C. Inpartic- − 0 x M ular,if f 60,thereexistsaconformalmetrictog suchth∈at f canberealizedits 0 Q-curvature. (ii) If Q dµ = 0, then there exists a conformal metric to g with associated M 0 g0 0 Q-Rcurvature αf for α 1,1 providedsup f(x) > 0, inf f(x) < 0, and f dµ ,0. ∈ {− } x∈M x∈M M g0 (iii) IRf 0 < Q dµ < (n 1)!ω , thenthere exists a conformalmetric to g with M 0 g0 − n 0 associaRtedQ-curvature f ifandonlyifsup f(x)>0. x M ∈ 6 Q.A.NGOˆ ANDH.ZHANG (iv) LetM =Snand f beaG-invariantfunctiononSnwithsup f >0. Ifeither x Sn (a) Σ= or ∈ (b) there∅existy Σandr >0suchthat 0 0 ∈ sxuΣp f(x)6max(cid:18)Z−Sn f ◦φy0,r0 dµSn,0(cid:19), (1.21) ∈ thenthereexistsaG-invariantconformalmetrictog withassociatedQ-curvature 0 f. Inparticular,if sup f(x)6max f dµSn,0 , x Σ Z−Sn ∈ (cid:0) (cid:1) thenthereexistsaG-invariantconformalmetrictog withassociatedQ-curvature 0 f. InviewofCorollary1.3(ii)above,weareunabletogetridofthemultipleαin(ii),due tothe factthatwe cannotconfirmthe signofλ . However,notethatifsup f(x) > 0 x M and f dµ <0,thenGeandXu[GX08]succ∞essfullyremovedthemultiple∈αbyusing M g0 avarRiationalapproach.Note,however,thattheauthorsin[BFR06,Corollary2.4]wasable toconfirmthesignofλ insomecases. Theresultin[BFR06]roughlysaysthatifthere exists some u C2(M)∞with small ∆u2dµ , then there exists a conformal metric g ∈ M| | g0 withitsQ-curvatureequalto sgn( R f dµ )f. Theproofmakesuseof(n/2)2=n;hence e − M eg0 it is not easy to generalize such aRresult for the general operator P because the precise 0 formulaforP isnotclear. 0 Beforeclosingthissection, we notethat, asin [BFR06] and[Bre03], the metricg in 0 Theorem1.1islimitedtotheso-called“subcritical”inthesense k < (n 1)!ω . When n n − k >(n 1)!ω ,theanalysisisdelicatesinceblow-upphenomenamayoccurs.Thiscanbe n n easilyse−enbyconsideringthefamilyofconformaldiffeomorphismofSn. Onthestandard model Sn, hence k = (n 1)!ω , a lot of attempts by different methods have already n n − beenperformed,forinterested readers, we referto [WX98, Bre03, Bre06, MS06, CX11, Ho12]andthereferencestherein.Anothersourceofusefulinformationfortheproblemof prescribingQ-curvaturecanalsobefound,forexample,in[DM08,LLL12,GM15]. 2. Theflowequationanditsenergyfunctional In this section, we first derive the flow equations for the conformal factor u and the curvatureQandthenshowthattheenergyfunctionalE isnon-increasing. Since the evolution equation (1.17) preserves the conformal structure, we may write g(t) = e2u(t)g forsomereal-valuedfunctionu. Then,weclearlyhavethefollowingequa- 0 tionfortheconformalfactoru ∂ u(t)=λ(t)f Q (F) ∂t − g(t) forallt>0. Thanksto(1.7),wealsodeducethat ∂ u= P u e nuQ +λ(t)f ∂t − g(t) − − 0 forallt>0. Byeasycalculations,wecanobtainevolutionequationsfor∂Q and∂λ(t) t g(t) t alongtheflow. Thisisthecontentofthefollowinglemma. Lemma2.1. Supposethatu(x,t)isasolutionoftheflow(F). Then,wehavethefollowing claims: ∂ Q = P (Q λ(t)f)+nQ (Q λ(t)f) ∂t g(t) − g(t) g(t)− g(t) g(t)− and λ(t)= f2 dµ −1 P f nλ(t)f2 λ(t)f Q dµ ′ g(t) g(t) g(t) g(t) Z Z − − (cid:16) M (cid:17) M(cid:0) (cid:1)(cid:0) (cid:1) Q-CURVATUREFLOWONMANIFOLDSOFEVENDIMENSION 7 where, as always, P = e nuP is the GJMS operatorwith respective to the conformal g(t) − 0 metricg(t). Proof. Sincethisisjustaroutinecalculation,weleavetheprooftothereader. (cid:3) Lemma2.2. Supposethatu(x,t)isasolutionoftheflow(F). Then,onehas dE[u] = n (Q λ(t)f)2dµ . g(t) g(t) dt − Z − M Inparticular,theenergyfunctionalE[u]isnon-increasingalongtheflow. Proof. Itfollowsfromtheflowequationforu,(1.7),and(1.18)that dE[u] n = u P u+u P u dµ +n Q u dµ dt 2Z t· 0 · 0 t g0 Z 0 t g0 M M = n (Q λ(t)f)Q dµ g(t) g(t) g(t) − Z − M = n (Q λ(t)f)2dµ +nλ(t) f(λ(t)f Q )dµ g(t) g(t) g(t) g(t) − Z − Z − M M = n (Q λ(t)f)2dµ . g(t) g(t) − Z − M Theproofiscomplete. (cid:3) Before going further, let us recall some important inequalities that will also be used in the present paper. First, we recall the Gagliardo–Nirenberg interpolation inequality, appliedtoanyLs-integrablefunctionϕ, p(1 α)/r pα/q jϕpdµ 6Cp mϕrdµ − ϕqdµ (2.1) ZM|∇ | g0 (cid:18)ZM|∇ | g0(cid:19) (cid:18)ZM| | g0(cid:19) forsomeC >0with0<α<16q,r6+ satisfying1/p= j/n+(1/r m/n)α+(1 α)/q. ∞ − − SincethehighorderGJMSoperatorP isself-adjointandpositivewithkernelconsist- 0 ingofconstantfunctions,wecanapplyAdam’sinequality[Ada88,Theorem2]toget n!ω (u u)2 exp n − dµ 6C (2.2) ZM (cid:16) 2 Mu·P0u dµg0(cid:17) g0 A R forsomeconstantC >0. Hereandthroughoutthepaper,bywwemeantheaverageofw A calculatedwithrespecttog ,thatisw=(vol(M,g )) 1 wdµ . (Intherestofourpaper, 0 0 − M g0 byvol(K)wemeanthevolumeofK MbeingcalculaRtedwithrespecttothebackground ⊂ metricg . IfwewanttoemphasizethemetricgbeingusedtocalculatethevolumeofK, 0 we shall write vol(K,g).) A detailed explanation for the validity of (2.2) can be found, for example, in [Bre03, page 330]. As a consequence of (2.2), we obtain the following Trudinger-typeinequality α2 exp α(u u) dµ 6C exp u P u dµ (2.3) ZM (cid:0) − (cid:1) g0 A (cid:16)2n!ωn ZM · 0 g0(cid:17) forallrealnumberα. 3. BoundednessofuinHn/2for06t6T Inthissection,wetrytoshowthatuisuniformlyboundedinHn/2(M,g ). Toseethis, 0 wewillsplitourargumentintofourcasesdependingonthesignandvalueofQ . 0 8 Q.A.NGOˆ ANDH.ZHANG 3.1. Thecase Q dµ = 0. Itfollowsfromourchoiceoftheinitialdatau in(1.13) M 0 g0 0 and(1.18)thatR f dµ =0forallt>0,whichimpliesthefollowinglemma. M g(t) R Lemma3.1. Alongtheflow(F),thevolumeofthemetricgispreserved. Hence,ifweset V0 = Menu0 dµg0 then Menu dµg0 =V0forallt. R R Proof. Clearly,wehave d enu dµ =n u dµ =n λ(t)f Q dµ dtZ g0 Z t g(t) Z − g(t) g(t) M M M (cid:0) (cid:1) =n λ(t) f dµ Q dµ =0. Z g(t)−Z 0 g0 (cid:16) M M (cid:17) Hence enu dµ isconstantalongtheflow;thisprovestheassertion. (cid:3) M g0 R Lemma3.2. ThereexistsauniformconstantC >0suchthat u 6C k kHn/2(M,g0) foralltimetinthemaximalintervalofexistence. Proof. FromLemma2.2,itfollowsthat n u P u dµ +n Q (u u) dµ 6E[u ]. (3.1) 2Z · 0 g0 Z 0 − g0 0 M M UsingthePoincare´-typeinequality(1.9),weget 1 (u u)2 dµ 6 u P u dµ , ZM − g0 λ1(P0)ZM · 0 g0 whichtogetherwithYoung’sinequalityimplythat n n n Q (u u) dµ 6 u P u dµ + Q2 dµ . (3.2) (cid:12)(cid:12) ZM 0 − g0(cid:12)(cid:12) 4ZM · 0 g0 λ1(P0)ZM 0 g0 (cid:12) (cid:12) Bysubstituti(cid:12)ng(3.2)into(3.1),w(cid:12)ehave (cid:12) (cid:12) 4 4 u P u dµ 6 E[u ]+ Q2 dµ , (3.3) ZM · 0 g0 n 0 λ1(P0)ZM 0 g0 Hence,inviewof (1.16),tobound u ,itsufficestobound udµ . ByJensen’s k kHn/2(M,g0) M g0 inequalityandLemma3.1,wehavethat R enu 6 enu dµ =V0/vol(M). Z− g0 M Therefore, 1 u6 log(V0/vol(M)). n Toboundufrombelow,weapplytheTrudinger-typeinequality(2.3)and(3.3)toget 2E[u ] 2n exp n(u u) dµ 6C exp 0 + Q2 dµ . ZM (cid:0) − (cid:1) g0 A (cid:16)(n−1)!ωn λ1(n−1)!ωn ZM 0 g0(cid:17) Since enu dµ =V0,weconcludethat M g0 R C 2E[u ] 2n exp nu 6 A exp 0 + Q2 dµ . (cid:0)− (cid:1) V0 (cid:16)(n−1)!ωn λ1(n−1)!ωn ZM 0 g0(cid:17) Hence,weget 1 C 2 2E[u ] u> log A + Q2 dµ + 0 . −(cid:16)n (cid:16)V0(cid:17) λ1(n−1)!ωn ZM 0 g0 n!ωn (cid:17) Bycombiningalltheestimatesabove,wethuscompletetheproof. (cid:3) Q-CURVATUREFLOWONMANIFOLDSOFEVENDIMENSION 9 3.2. Thecase Q dµ > 0. Inthiscase,itfirstfollowsfromourchoiceofinitialdata M 0 g0 u and(1.18)thRat fenu dµ = Q dµ forallt>0,whichimpliesthat 0 M g0 M 0 g0 R R Q dµ = fenu dµ 6(sup f) enu dµ . ZM 0 g0 ZM g0 M ZM g0 Hence, 1 enu dµ > Q dµ . (3.4) Z g0 sup f Z 0 g0 M M M Itfollowsfrom(2.3)and(3.4)that n/2 u P u dµ >log exp n(u u) dµ logC (n−1)!ωn ZM · 0 g0 ZM (cid:0) − (cid:1) g0 − A 1 >log Q dµ nu. (cid:18)CAsupM f ZM 0 g0(cid:19)− Thus,wehave n/2 1 1 u> u P u dµ + log Q dµ . (3.5) −n!ωn ZM · 0 g0 n (cid:18)CAsupM f ZM 0 g0(cid:19) Using(3.5)andLemma2.2,weget E[u ]>E[u] 0 n = u P u dµ +n Q (u u) dµ +nu Q dµ 2Z · 0 g0 Z 0 − g0 Z 0 g0 M M M >n 1 MQ0 dµg0 u P u dµ +n Q (u u) dµ (3.6) 2(cid:18) − R(n−1)!ωn (cid:19)ZM · 0 g0 ZM 0 − g0 1 + Q dµ log Q dµ . ZM 0 g0 (cid:18)CAsupM f ZM 0 g0(cid:19) ByYoung’sinequalityand(1.9),wehave n Q (u u) dµ 6 n 1 MQ0 dµg0 u P u dµ (cid:12)(cid:12)(cid:12) ZM 0 − g0(cid:12)(cid:12)(cid:12) 4(cid:18) − R(n−1)!ωn (cid:19)ZM · 0 g0 (3.7) (cid:12) (cid:12) n Q2 dµ (cid:12) (cid:12) + M 0 g0 R λ (P ) 1 ( Q dµ /(n 1)!ω 1 0 − M 0 g0 − n (cid:0) R (cid:1) (cid:1) Since Q dµ < (n 1)!ω , by substituting (3.7) into (3.6), we conclude that there M 0 g0 − n existsaRnuniformconstantC >0suchthat 2 u P u dµ 6C . (3.8) Z · 0 g0 2 M Substituting(3.8)into(3.5)givesthatthereexistsanuniformconstantC suchthat 3 u>C . (3.9) 3 To get an upper bound for u, we notice that (3.2) still holds in this case. Hence by substituting (3.2) into the first inequality in (3.6), the positivity of P and the inequality 0 Q dµ >0,weget M 0 g0 R (n/λ (P )) Q2 dµ +E[u ] u6 1 0 M 0 g0 0 . (3.10) R n Q dµ M 0 g0 R By plugging (3.8), (3.9) and (3.10) into (1.9), we conclude that there exists a uniform constantC >0suchthat 4 u2 dµ 6C . (3.11) Z g0 4 M Theassertionisthenprovedbycombining(3.8)and(3.11). 10 Q.A.NGOˆ ANDH.ZHANG 3.3. Thecase Q dµ <0. Toachieveourgoal,wefirstneedananalogueof[BFR06, M 0 g0 Lemma4.1]whRoseproofisprovidedinAppendixBforcompleteness. Lemma 3.3. Let K be a measurablesubsetof M with vol(K) > 0. Then there exist two constantsα>1dependingonMandg andC >1dependingonM,g ,andvol(K)such 0 K 0 that α enu dµ 6C exp α u 2 max enu dµ ,1 . ZM g0 K (cid:0) k 0kHn/2(M,g0)(cid:1) (cid:26)(cid:16)ZK g0(cid:17) (cid:27) NotethatLemma3.3holdsregardlessofthesignof Q dµ < 0. Withthehelpof M 0 g0 thislemma,weareabletoshowtheuniformboundofthRevolumeofthetimemetricg(t). Lemma3.4. Supposetheflow(F)existson[0,T)forsomeT >0. Also,assumethatthere existsaconstantC dependingon f ,u , g , and M suchthatsup f 6 C . Then,there 0 − 0 0 M 0 existsauniformconstantγ>0suchthat enu dµ 6γ Z g0 M forallt [0,T). ∈ Proof. Let f+ =max(f,0)anddefine 1 K = x M : f(x)6 inf f . ∈ 2 M n o Notice that K , due to the fact that inf f < 0. Hence, by the continuity of f, we M ∅ concludevol(K)>0. Sinceu Y,wehave 0 ∈ Z Q0 dµg0 =−Z fenu0 dµg0 =Z f−enu0 dµg0 −Z f+enu0 dµg0, M M M M whichimpliesthat 1 (−infM f)ZMQ0 dµg0 6ZMenu0 dµg0. From(2.3)andYoung’sinequality,itfollowsthat n n ZMenu0 dµg0 6CAexp(cid:18)2(n−1)!ωn ZMu0·P0u0 dµg0 + vol(M)ZMu0 dµg0(cid:19) n!ω n 6C exp n exp u P u +u2 dµ (3.12) A (cid:16)2vol(M)(cid:17) (cid:16)2(n−1)!ωn ZM(cid:2) 0· 0 0 0(cid:3) g0(cid:17) 6C exp C u 2 , 2 2k 0kHn/2(M,g0) (cid:16) (cid:17) whereC isaconstantdependingon M andg ,whichwemayassumetobegreaterthan 2 0 1. Therefore,wecanget 1 Q dµ 6C exp C u 2 . (3.13) −infM f ZM 0 g0 2 (cid:0) 2k 0kHn/2(M,g0)(cid:1) NowchoosetheconstantC inthelemmasuchthat 0 inf f C = − M exp (α(C +1) C ) u 2 , 0 8αCKC2α−1 · (cid:0)− 2 − 2 k 0kHn/2(M,g0)(cid:1) whereC >1andα>1areconstantsofLemma3.3. Moreover,set K γ=2C (8C )αexp (C +1)α u 2 . k 2 2 k 0kHn/2(M,g0) Then,weclaimthat (cid:0) (cid:1) enu dµ 6γ, (3.14) Z g0 M forallt [0,T).Toseethis,welet ∈ I = t [0,T): enu(s) dµ 6γforalls [0,t] . ∈ Z g0 ∈ n M o

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.