PROCEEDINGSOFTHE AMERICANMATHEMATICALSOCIETY Volume00,Number0,Pages000–000 S0002-9939(XX)0000-0 RIGIDITY OF ENTIRE SELF-SHRINKING SOLUTIONS TO KA¨HLER-RICCI FLOW ON COMPLEX PLANE 6 1 0 WENLONGWANG 2 (Communicated byLeiNi) t c O Abstract. We show that every entire self-shrinking solution on C1 to the 4 K¨ahler-Ricciflowmustbegenerated fromaquadraticpotential. 2 ] G 1. Introduction D In this short note, we prove the following result. . h t Theorem 1.1. Suppose that u(x) is an entire smooth subharmonic solution on Rn a m to the equation 1 [ (1.1) ln∆u= x Du u, 2 · − 2 then u is quadratic. v 8 For n = 2, up to an additive constant, equation (1.1) is equivalent to the one- 6 dimensional case of the complex Monge-Amp`ere equation 3 0 1 0 (1.2) lndetui¯j = 2x·Du−u . 2 on Cm. Any entire solution to (1.2) leads to an entire self-shrinking solution 0 6 x v(x,t)= tu 1 − √ t : (cid:18) − (cid:19) v to a parabolic complex Monge-Amp`ere equation i X vt =lndet vi¯j r a onCm ( ,0), where zi =xi+√ 1xm+i(cid:0). N(cid:1)ote that aboveequation of v is the × −∞ − potentialequationoftheK¨ahler-Ricciflow∂tgi¯j =−Ri¯j. Infact,thecorresponding metric ui¯j is a shrinking K¨aher-Ricci (non-gradient) soliton. Assuming a certaindecay of ∆u–a specific completeness condition, Q.Ding and (cid:0) (cid:1) Y.L. Xin have proved Theorem (1.1) in [2]. Under the condition that the K¨ahler metric ui¯j iscomplete,rigiditytheoremforequation(1.2)hasbeenobtainedbyG. Drugan,P.LuandY.Yuanin[3]. Similarrigidityresultsforself-shrinkingsolutions (cid:0) (cid:1) toLagrangianmeancurvatureflowsinpseudo-Euclideanspacewereobtainedin[1], [2], [4] and [5]. 2010 Mathematics Subject Classification. Primary53C44,53C24. Theauthor ispartiallysupportedbyCSC(ChinaScholarshipCouncil). (cid:13)cXXXXAmericanMathematicalSociety 1 2 WENLONGWANG Ourcontributionisremovingextraassumptionsfortherigidityofequation(1.1). As in [3] and [2], the idea of our argument is still to prove the phase–ln∆u is constant. Then the homogeneity of the self-similar term on the right-hand side of equation (1.1) leads to the quadratic conclusion. However, it’s hard to con- struct a barrier function as in [3] or to find a suitable integral factor as in [2] withoutcompletenessassumption. Takingadvantageofthe conformalityofthelin- earized equation (1.1), we establish a second order ordinary differential inequality for M(r)=max ln∆u(x) in the sense of comparison function. Then we prove |x|=r that M(r) blows up in finite time by Osgood’s criterion unless ln∆u is constant. 2. proof Proof. Define the phase by 1 φ(x)= x Du(x) u(x). 2 · − Taking two derivatives and using (1.1), we have eφ (2.1) ∆φ= x Dφ. 2 · Define M(r):[0,+ ) R by ∞ → M(r)= maxφ(x). |x|=r Assuming φ(x) is not a constant, we prove that M(r) blows up in finite time. Since M is locally Lipschitz, it is differentiable a.e. in [0,+ ). For all r > 0, there exists a corresponding angle θ Sn−1 satisfying ∞ r ∈ (2.2) M(r)=φ(r,θ ). r For r′ > 0 small enough, we have M(r+r′) φ(r +r′,θ ) and M(r r′) r ≥ − ≥ φ(r r′,θ ). It follows that r − M(r+r′) M(r) φ(r+r′,θ ) φ(r,θ ) r r − − r′ ≥ r′ and M(r) M(r r′) φ(r,θ ) φ(r r′,θ ) − − r − − r . r′ ≤ r′ Letting r′ 0 in above two equations, we have → ∂φ limM′ (r) (r,θ ) limM′ (r). − ≤ ∂r r ≤ + So if r >0 is a differential point of M, we have ∂φ (2.3) M′(r)= (r,θ ). r ∂r Because of the maximality of φ(r,θr) among θ Sn−1, we have ∆Sn−1φ(r,θr) 0. ∈ ≤ Plugin this inequality into (2.1), we obtain ∂2φ n 1∂φ r ∂φ (2.4) (r,θ )+ − (r,θ ) exp[φ(r,θ )] (r,θ ). ∂r2 r r ∂r r ≥ 2 r · ∂r r Fixing a positive R0, for any r [0,R0], θ Sn−1, t [0,1], we have the ∈ ∈ ∈ following Taylor’s expansion ∂φ 1∂2φ φ(r+t,θ) φ(r,θ) (r,θ ) t+ (r,θ ) t2 Ct3. − ≥ ∂r r · 2∂r2 r · − RIGIDITY OF ENTIRE SELF-SHRINKING SOLUTIONS TO KA¨HLER-RICCI FLOW ON COMPLEX PLANE3 HereCisaconstantdependsonlyonR0,infactwecanchooseC =max|x|≤R0+1|D3φ(x)|. We evaluate above inequality at (r,θ ), where r is a differentiable point and θ r r is the corresponding critical angle. Using M(r+t) φ(r+t,θ ), (2.2), (2.3) and r ≥ (2.4), we have the following inequality that only involves M, namely 1 2(n 1) M(r+t) M(r) M′(r)t+ exp[M(r)]r − M′(r)t2 Ct3. − ≥ 4 − r − (cid:26) (cid:27) Or equivalently, (2.5) 1 1 2(n 1) [M(r+t) M(r) M′(r)t] exp[M(r)]r − M′(r) Ct. t2 − − ≥ 4 − r − (cid:26) (cid:27) Becauseinequality(2.5)holdsforr [0,R0]a.e.,wecanintegrateitwithrespect ∈ to r over any subinterval [a,b] [0,R0], and get the following inequality for every ⊂ t [0,1], ∈ 1 b+t a+t (2.6) M(r)dr M(r)dr [M(b) M(a)]t t2 (Zb −Za − − ) 1 b 2(n 1) exp[M(r)]r − M′(r)dr C(b a)t. ≥4 − r − − Za (cid:26) (cid:27) Choosing differentiable points a and b and letting t 0 in (2.6), we have → 1 b 2(n 1) (2.7) M′(b) M′(a) exp[M(r)]r − M′(r)dr. − ≥ 2 − r Za (cid:26) (cid:27) Since R0 can be arbitrarily large, in fact (2.7) holds for all differentiable points a,b R+. ∈ We claim there exists l0 > 0 such that M′(r) > 0 at every differentiable point in [l0,+ ). Otherwise, there exist an increasing sequence of differentiable points rk R∞+, and a sequence of corresponding critical angles θk Sn−1 such that { }⊂ { }⊂ ∂φ M′(r )= (r ,θ ) 0, and lim r =+ . k k k k ∂r ≤ k→∞ ∞ ThenaccordingtoHopf’slemma,weknowφ(x)isconstantinB (0). Sincer can rk k be arbitrarily large, φ(x) is in fact a constant on the whole Rn, which contradicts our assumption. Sothereexistsacertainl0 >0,suchthatM′(r)>0holdsa.e. in[l0,+ ). Then ∞ M(r)monotonicallyincreaseson[l0,+ ). Whena l1 ,l0+n+2exp[ M(l0)], ∞ ≥ − we have b 2(n 1) b (2.8) exp[M(r)]r − M′(r)dr >2 exp[M(r)]M′(r)dr − r Za (cid:26) (cid:27) Za =2 exp[M(b)] exp[M(a)] . { − } Combining (2.7) and (2.8), we obtain (2.9) M′(b) M′(a) exp[M(b)] exp[M(a)]. − ≥ − Above inequality holds for all differentiable points a,b [l1,+ ). Choosing a ∈ ∞ differentiable point l2 l1, then M′(r) M′(l2)>0 holds a.e. in [l2,+ ). Thus ≥ ≥ ∞ M(r) + as r + . → ∞ → ∞ 4 WENLONGWANG Then according to Osgood’s criterion, M(r) blows up in finite time, which con- tradicts the assumption that φ(x) is entire. Hence, we conclude φ(x) is constant. Using φ(x)= 1x Du(x) u(x), we have 2 · − 1 x D[u(x)+φ(0)]=u(x)+φ(0). 2 · Finally, it follows from Euler’s homogeneous function theorem that smooth u(x)+φ(0) is a homogeneous order 2 polynomial. (cid:3) Remark 2.1. From the proof, it’s not hard to see that the theorem also holds for ∆u=f(x Du 2u) · − if f C1(R) is convex, monotone increasing, and f−1 L1([d,+ )) for a certain d ∈R. Integrability condition for f−1 is necessary.∈Otherwis∞e, we have such ∈ counterexample: f(x) x and ≡ x1 1 s2 1 x2 u(x)= x21−1 0 s2(exp 2 −1)ds− x1(exp 21 −1)−x1. Z (cid:0) (cid:1) Acknowledgement. I would like to sincerely thank Professor Yu Yuan for sug- gesting this problem to me and for many heuristic discussions. The result was obtained when I was visiting the University of Washington. I would also like to thankCSC(ChinaScholarshipCouncil)foritssupportandtheUniversityofWash- ington for its hospitality. I am grateful to the referees for useful suggestions. References 1. A.Chau,J.Y.ChenandY.Yuan,Rigidityofentireself-shrinkingsolutionstocurvatureflows, J.ReineAngew.Math.,664(2012), 229–239. 2. Q. DingandY.L. Xin,The rigiditytheorems for Lagrangian selfshrinkers, J. ReineAngew. Math.,DOI:10.1515/crelle-2012-0081. 3. G.Drugan,P.LuandY.Yuan,Rigidityofcompleteentireself-shrinkingsolutionstoK¨ahler- Ricciflow,Int.Math.Res.Not.,DOI:10.1093/imrn/run051. 4. R.L. Huang, Lagrangian mean curvature flow in pseudo-Euclidean space, Chin. Ann. Math. Ser.B,32(2)(2011), 187-200. 5. R.L.HuangandZ.Z.Wang,Ontheentireself-shrinkingsolutionstoLagrangianmeancurva- tureflow,Calc.Var.andPDE,41(2011), 321–339. School of Mathematical Sciences, Peking University, Science Building in Peking University,No5. YiheyuanRoad, Beijing,P.R.China 100871 E-mail address: [email protected]