arXiv:1908.07672v3 [math.DG] 29 Dec 2020 sisie yta fteSuSmsnrsl o amncma harmonic for result Siu-Sampson the of [ABC in that by inspired is embed of deformations for CCY16] . [CCY15, Yang and Chanillo, functions. is pluriharmonic operator CR Paneitz of CR consists the kernel Then three. dimension of hoe 1.1. Theorem operat Paneitz CR Indeed: the manifolds. that prove CR we embeddable of paper, stability this the In and functions. curvature scalar P Tanaka-Webster deformations CR the embeddable the of of real-analytic nonnegativity under the preserved that is shown have [CCY16] Yang and in fsrcl suoovxra yesrae in hypersurfaces real suffi pseudoconvex under strictly stable of is tions condition nonnegativity Chiu the Chanillo, direction, that this proved three-m In CR embeddable operator. closed Paneitz any CR tive whether ask to 4. natural Section is [CMY17, it see d operator; of Paneitz manifolds CR de theor CR nonnegative is non-embeddable not mass closed it positive are CR importance; there that the great known and of [CCY12] been embeddability has global operator Paneitz CR the NIGatNme JP19J00063. Number Grant ENHI kernel Szegő manifold, hyperbolic complex asymptotically eew iea uln ftepofo hoe . ie nSec in given 1.1 Theorem of proof the of studi outline an been give has we Here operator Paneitz CR the of kernel the that Note o he-iesoa titypedcne Rmanifolds CR pseudoconvex strictly three-dimensional For 2010 e od n phrases. and words Key hswr a upre yJP eerhFlosi o Youn for Fellowship Research JSPS by supported was work This ONGTVT FTEC AET PRTRFOR OPERATOR PANEITZ CR THE OF NONNEGATIVITY ahmtc ujc Classification. Subject Mathematics atfrs utemr,w ics h oaihi singul logarithmic the discuss kernel. we Furthermore, forms. tact Q h RYmb rbe o medbeC aiod.W losh also We CR zero manifolds. with CR form contact embeddable a affirmat for of an existence problem gives th Yamabe result prove This CR we manifolds. the paper, CR this embeddable In for ity geometry. CR three-dimensional in Abstract. + piecrauet medbeC aiod ihn pseudo- no with manifolds CR embeddable to curvature -prime 6 hpe .] e ( Let 6.1]. Chapter 96, Let h ongtvt fteC aet prtrpasacrucial a plays operator Paneitz CR the of nonnegativity The MEDBEC MANIFOLDS CR EMBEDDABLE ( ,T M, RPnizoeao,C lrhroi ucin CR function, pluriharmonic CR operator, Paneitz CR 1 , 0 M ) 1. UATAKEUCHI YUYA eacoe medbesrcl suoovxCR pseudoconvex strictly embeddable closed a be ,T M, Introduction rmr 22;Scnay3V5 23,53C55. 32V30, 32V15, Secondary 32V20; Primary 1 , 0 1 M Q ea nTerm11 rm[HL75, From 1.1. Theorem in as be ) craueadgnrlz h total the generalize and -curvature . C 2 oevr ae Chanillo, Case, Moreover, . n ag[C1]have [CCY13] Yang and , cetssadJP KAK- JSPS and Scientists g rt fteSzegő the of arity ihuiompositivity uniform with inl ml deforma- small ciently h Rpluriharmonic CR the ]freape Hence example. for 1] h ongtvt of nonnegativity the , ri ongtv for nonnegative is or v ouinto solution ive nfl a nonnega- has anifold snonnegativ- is ongtv,adits and nonnegative, mnintrewith three imension isencon- Einstein m[M1] tis It [CMY17]. em pycnetdto connected eply s[i8,Sam86] [Siu80, ps al Rthree- CR dable in4 u proof our 4; tion nizoperator aneitz wthe ow db Case, by ed role Q -curvature, 2 YUYA TAKEUCHI

Lem95], it follows that M bounds a strictly pseudoconvex domain Ω in a two- dimensional complex projective manifold X. By using a Kähler form on X and a defining function of Ω, we have an asymptotically complex hyperbolic Kähler form ∞ ω+ on Ω; see Section 4 for details. For u C (M), take a harmonic extension u of ∈c c u to Ω. Since ω+ is Kähler, u satisfies dd u ω+ = 0, where d = (√ 1/2)(∂ ∂). ∧ − −e Then we have ddcu ddcu 0, and so ∧ ≤e e e edcu ddcu = ddcu ddcu 0. ZM ∧ ZΩ ∧ ≤ On the other hand, integratione by partse on Meyields thate

c c d u dd u = u(Pθu)θ dθ, ZM ∧ − ZM ∧ where θ is a contact form one M ande Pθ is the CR Paneitz operator. Thus we obtain the nonnegativity of Pθ. Moreover, if the equality holds, then u must be pluriharmonic, and u is CR pluriharmonic. e Theorem 1.1 has various applications. Let (M,T 1,0M) be a closed strictly pseu- doconvex CR manifold of dimension three. The CR Yamabe constant (M,T 1,0M) Y is defined by

(M,T 1,0M) = inf Scal θ dθ θ dθ =1 . Y θ  Z · ∧ Z ∧  M M 1 0 This value gives a CR invariant of (M,T , M). Theorem 1.1 and [CCY12, Theorem 1.4(b)] imply the following embeddability criterion. Corollary 1.2. Let (M,T 1,0M) be a closed strictly pseudoconvex CR manifold of dimension three with positive CR Yamabe constant. Then (M,T 1,0M) is embeddable if and only if the CR Paneitz operator is nonnegative. It is interesting to ask whether we can remove the positivity of the CR Yamabe constant from the assumption of Corollary 1.2. It is known [JL87] that the CR Yamabe constant satisfies (M,T 1,0M) (S3,T 1,0S3), Y ≤ Y where (S3,T 1,0S3) is the standard CR sphere in C2; note that (S3,T 1,0S3) > Y 0. Combining Theorem 1.1 with a consequence of the CR positive mass theo- rem [CMY17, Theorem 1.2], we have the following Corollary 1.3. Let (M,T 1,0M) be a closed embeddable three-dimensional strictly pseudoconvex CR manifold. Then (M,T 1,0M) = (S3,T 1,0S3) if and only if Y Y (M,T 1,0M) is CR equivalent to (S3,T 1,0S3). The CR Yamabe problem is to find a contact form θ satisfying

Scal = (M,T 1,0M), θ dθ = 1; Y ZM ∧ such a contact form is called a CR Yamabe contact form. Recently, Cheng, Mal- chiodi, and Yang [CMY19] have found that there exist no CR Yamabe contact forms on the Rossi sphere, which is an example of non-embeddable CR manifolds. Nonetheless, the CR Yamabe problem is solvable affirmatively for embeddable three-dimensional CR manifolds. Theorem 1.4. There exists a CR Yamabe contact form on any closed embeddable strictly pseudoconvex CR manifold of dimension three. NONNEGATIVITY OF THE CR PANEITZ OPERATOR 3

We will also apply Theorem 1.1 to the zero CR Q-curvature problem. In a study of the Szegő kernel, Hirachi [Hir93] has introduced a pseudo-Hermitian invariant

1 11 Qθ = (∆bScal 2 Im A ), 6 − 11, which is now called the CR Q-curvature. Note that this name stems from the fact that this invariant coincides with a CR analog of the Q-curvature; see [FH03]. Since Qθ is expressed as a divergence, its integral is identically zero. Moreover, the CR Q-curvature itself is identically zero for pseudo-Einstein contact forms; that is, contact forms satisfying the equality Scal √ 1A 1 =0. 1 − − 11, The existence of such a contact form is equivalent to that of a Fefferman defining function if M bounds a strictly pseudoconvex domain [Hir93,HPT08]; in particular, it always exists if the CR manifold is a real hypersurface in C2. For these reasons, it is natural to ask whether any CR manifold admits a contact form with zero CR Q-curvature. Some authors have tackled this problem via the CR Q-curvature flow [CCC07,SC11,CKS19]. In this paper, however, we will take a more functional- analytic approach.

Theorem 1.5. Let (M,T 1,0M) be a closed embeddable strictly pseudoconvex CR manifold of dimension three. There exists a contact form θ on M with zero CR Q-curvature. Moreover, if θˆ is also such a contact form, then θˆ = eΥθ for a CR pluriharmonic function Υ. In particular, in the case where M admits a pseudo- Einstein contact form, θ is pseudo-Einstein if and only if Qθ =0.

The existence of such a contact form gives a generalization of the total Q-prime curvature to embeddable CR manifolds with no pseudo-Einstein contact forms. Let (M,T 1,0M) be a closed strictly pseudoconvex CR manifold of dimension three and ′ θ a contact form on M. The Q-prime curvature Qθ is defined by

′ 1 2 2 Q = Scal 2 A + ∆bScal. θ 2 − | 11| Case and Yang [CY13] have proved that the integral of the Q-prime curvature

′ 1,0 ′ Q (M,T M)= Qθθ dθ ZM ∧ is independent of the choice of a contact form θ with zero CR Q-curvature, and defines a CR invariant of (M,T 1,0M), called the total Q-prime curvature. Hence, if (M,T 1,0M) is embeddable, then the total Q-prime curvature is always well- defined and gives a CR invariant. The total Q-prime curvature and the CR Yamabe constant satisfy the following inequality, which is a CR analog of [Gur99, Equation (1.4)].

Theorem 1.6. Let (M,T 1,0M) be a closed embeddable three-dimensional strictly pseudoconvex CR manifold. Then ′ 1 (1.1) Q (M,T 1,0M) (M,T 1,0M)2 ≤ 2Y with equality if and only if M admits a pseudo-Einstein contact form with vanishing 1,0 2 Tanaka-Webster torsion. In particular if c1(T M) = 0 in H (M, R), then the 6 strict inequality holds. 4 YUYA TAKEUCHI

Note that there exists a two-dimensional strictly pseudoconvex domain whose 1,0 2 boundary M satisfies c1(T M) = 0 in H (M, R) [EO08, Theorem 6.2]. In partic- 6 ular, the strict inequality in (1.1) holds for this M. We also mention the case where the equality in (1.1) holds. A CR manifold having a pseudo-Einstein contact form with vanishing Tanaka-Webster torsion is known as a Sasakian η-Einstein manifold. Such a manifold plays a crucial role in CR geometry; see [Wan15,CG17,Tak18] for example. Theorem 1.6 gives a generalization of [CY13, Theorem 1.1] for embeddable CR manifolds.

Corollary 1.7. Let (M,T 1,0M) be a closed embeddable strictly pseudoconvex CR manifold of dimension three with nonnegative CR Yamabe constant. Then ′ ′ Q (M,T 1,0M) Q (S3,T 1,0S3) ≤ with equality if and only if (M,T 1,0M) is CR equivalent to (S3,T 1,0S3).

We will also apply our results to the logarithmic singularity of the Szegő kernel. Let Ω be a strictly pseudoconvex domain in a two-dimensional and θ a contact form on ∂Ω. Define the Hardy space θ(Ω) as the set consisting H of the boundary values of holomorphic functions on Ω that are L2 with respect to the volume form θ dθ on ∂Ω. The Szegő kernel Sθ(z, w) is the reproducing kernel ∧ of θ(Ω). Fix a defining function ρ of Ω. The boundary behavior of Sθ on the H diagonal is given by −2 (1.2) Sθ(z, z)= φθρ + ψθ log( ρ), − where φθ and ψθ are smooth functions on Ω. Moreover, the Tayler coefficients of ψθ at a given boundary point are uniquely determined by the behavior of θ near the point [BdMS76]. Hirachi [Hir93, Main Theorem] has proved that 1 (1.3) ψθ ∂Ω = Qθ. | 4π2

Hence ψθ = O(ρ) if and only if θ is of zero CR Q-curvature. It is natural to ask when ψθ vanishes to higher-order on the boundary. Theorem 1.8. Let Ω be a two-dimensional strictly pseudoconvex domain whose boundary admits a pseudo-Einstein contact form. For a contact form θ on ∂Ω, the 2 3 function ψθ is of O(ρ ) (resp. O(ρ )) if and only if θ is pseudo-Einstein and ∂Ω is obstruction flat (resp. spherical).

Note that the above theorem has been obtained by Hirachi in the case when Ω is a domain in C2 and ∂Ω has transverse symmetry [Hir93]. It is interesting to ask what happens when ∂Ω has no pseudo-Einstein contact forms. We also mention a 2 recent progress by Curry and Ebenfelt [CE18,CE19] on the question: If ψθ = O(ρ ), 3 does ψθ = O(ρ ) hold? This paper is organized as follows. In Section 2, we recall some basic definitions and facts on CR manifolds. In Section 3, we discuss a CR analog of dc, which is closely related to CR pluriharmonic functions and plays an important role in the proofs of our results in this paper. Section 4 is devoted to proofs of Theorems 1.1 and 1.4. In Section 5, we tackle the existence problem of a contact form with zero CR Q-curvature. Section 6 deals with the logarithmic singularity of the Szegő kernel. NONNEGATIVITY OF THE CR PANEITZ OPERATOR 5

2. Preliminaries Let M be a smooth three-dimensional manifold without boundary. A CR struc- ture is a complex one-dimensional subbundle T 1,0M of the complexified tangent bundle TM C such that ⊗ T 1,0M T 0,1M =0, ∩ where T 0,1M is the complex conjugate of T 1,0M in TM C. An important example ⊗ of a three-dimensional CR manifold is a real hypersurface M in a two-dimensional complex manifold X; this M has the canonical CR structure

1,0 1,0 T M = T X M (TM C). | ∩ ⊗ ∞ 0,1 ∗ Introduce an operator ∂b : C (M) Γ((T M) ) by → ∂bf = (df) 0,1 . |T M A smooth function f is called a CR if ∂bf =0. A CR pluriharmonic function is a real-valued smooth function that is locally the real part of a CR holomorphic function. We denote by the space of CR pluriharmonic P functions. A CR structure T 1,0M is said to be strictly pseudoconvex if there exists a nowhere-vanishing real one-form θ on M such that θ annihilates T 1,0M and √ 1dθ(Z, Z) > 0, 0 = Z T 1,0M; − − 6 ∈ we call such a one-form a contact form. Denote by T the Reeb vector field with respect to θ; that is, the unique vector field satisfying

θ(T )=1, ιT dθ =0.

1,0 Let Z1 be a local frame of T M, and set Z1 = Z1 . Then (T,Z1 ,Z1 ) gives a local frame of TM C, called an admissible frame. Its dual frame (θ,θ1,θ1) is called an ⊗ admissible coframe. The two-form dθ is written as dθ = √ 1l θ1 θ1, − 11 ∧ 11 where l11 is a positive function. We use l11 and its multiplicative inverse l to raise and lower indices. A contact form θ induces a canonical connection , called the Tanaka-Webster ∇ connection with respect to θ. It is defined by

1 1 1 1 T =0, Z1 = ω1 Z1 , Z1 = ω1 Z1 ω1 = ω1 ∇ ∇ ∇   with the following structure equations: dθ1 = θ1 ω 1 + A1 θ θ1, ∧ 1 1 ∧ 1 1 dl11 = ω1 l11 + l11ω1 .

The tensor A11 = A11 is called the Tanaka-Webster torsion. We denote the com- ponents of a successive covariant derivative of a tensor by subscripts preceded by a comma, for example, K ; we omit the comma if the derivatives are applied to a 11,1 function. We use the index 0 for the component T or θ in our index notation. The commutators of the second derivatives for u C∞(M) are given by ∈ (2.1) u u = √ 1l u , u u = A u1; 11 − 11 − 11 0 01 − 10 11 6 YUYA TAKEUCHI see [Lee88, (2.14)]. Define the sub-Laplacian ∆b by

1 1 (2.2) ∆bu = u u − 1 − 1 for u C∞(M). From (2.1), it follows that ∈ 1 1 (2.3) ∆bu = 2u √ 1u = 2u + √ 1u . − 1 − − 0 − 1 − 0 1 1 The curvature form Ω1 = dω1 of the Tanaka-Webster connection satisfies (2.4) Ω 1 = Scal l θ1 θ1 A 1θ θ1 + A 1θ θ1, 1 · 11 ∧ − 11, ∧ 11, ∧ where Scal is the Tanaka-Webster scalar curvature. Let θˆ = eΥθ be another contact form, and denote by T the corresponding Reeb vector field. The admissible coframe corresponding to (T,Z ,Z ) is given by b 1 1 (2.5) (θ,ˆ θˆ1 = θ1 + √ 1Υ1θ, θˆ1 = θ1 √ b1Υ1θ). − − − Under this conformal change, the sub-Laplacian ∆b with respect to θˆ satisfies Υ 1 1 (2.6) e ∆bu = ∆bu Υ u bΥ u ; − 1 − 1 see [Sta89, Lemma 1.8] for example.b An important example of strictly pseudoconvex CR manifolds is the boundary of a strictly pseudoconvex domain. Let Ω be a relatively compact domain in a two-dimensional complex manifold X with smooth boundary M = ∂Ω. Then there exists a smooth function ρ on X such that Ω= ρ−1(( , 0)),M = ρ−1(0), dρ =0 on M; −∞ 6 such a ρ is called a defining function of Ω. A domain Ω is said to be strictly pseudoconvex if we can take a defining function ρ of Ω such that ddcρ defines a Kähler form near M. The boundary of a strictly pseudoconvex domain is a closed c strictly pseudoconvex real hypersurface; the one-form d ρ M is a contact form on | M.

3. CR pluriharmonic functions and pseudo-Einstein condition Let (M,T 1,0M) be a strictly pseudoconvex CR manifold of dimension three and θ a contact form on M. We first discuss a CR analog of dc. Such an operator was first introduced by Rumin [Rum94] in his study of the Rumin complex. Here, however, we define it as an operator taking values in Ω1(M), which implicitly appears in the proof of [Hir14, Lemma 3.2].

Lemma 3.1. The differential operator

c ∞ 1 √ 1 1 1 1 (3.1) dCR : C (M) Ω (M); u − u1θ u1θ + ∆bu θ → 7→ 2  −  2 · is independent of the choice of θ.

ˆ Υ c Proof. Consider the conformal change θ = e θ. The independence of dCR in θ follows from the transformation laws of an admissible coframe (2.5) and the sub- Laplacian (2.6). 

As in complex geometry, CR pluriharmonic functions are smooth functions an- c nihilated by ddCR. NONNEGATIVITY OF THE CR PANEITZ OPERATOR 7

Lemma 3.2. For u C∞(M), ∈ (3.2) ddc u = P u θ θ1 + P u θ θ1, CR 1 · ∧ 1 · ∧ where P u = u 1 + √ 1A u1, P u = u 1 √ 1A u1. 1 1 1 − 11 1 1 1 − − 11 c In particular, u is a CR pluriharmonic function if and only if ddCRu =0. Proof. We first show the equality (3.2). From (2.2) and (3.1), it follows that

c 1 √ 1 √ 1 1 1 dd u = (∆bu) − u + − A u θ θ CR −2 1 − 2 10 2 11  ∧

1 √ 1 √ 1 1 1 + (∆bu) + − u − A u θ θ . −2 1 2 10 − 2 11  ∧ Hence it suffices to show

1 √ 1 √ 1 1 P u = (∆bu) − u + − A u ; 1 −2 1 − 2 10 2 11 the other part is the complex conjugate of this equality. By using (2.1) and (2.3), we have

1 √ 1 √ 1 1 1 √ 1 √ 1 1 (∆bu) − u + − A u = u + − (u u )+ − A u −2 1 − 2 10 2 11 1 1 2 01 − 10 2 11 = u 1 + √ 1A u1 1 1 − 11 = P1 u. The latter statement is a consequence of [Lee88, Proposition 3.4].  Each u defines a class [dc u] in H1(M, R). The following ∈ P CR lemma is obtained from [Lee88, Lemma 3.1]. Lemma 3.3. For u , the cohomology class [dc u] is equal to zero if and only ∈P CR if u is the real part of a CR holomorphic function globally. Assume that M is realized as the boundary of a two-dimensional strictly pseu- c doconvex domain Ω. Take a defining function ρ of Ω with θ = d ρ M . | Lemma 3.4. For each u C∞(M), there exists a smooth extension u of u to Ω c ∈ such that dd u M has vanishing (1, 1)-part. Moreover, such a u is unique modulo 2 c | c e O(ρ ), and d u M coincides with dCRu. e | e ′ ′ Proof. Take ae smooth function u on Ω with u M = u. Then | c ′ √ 1 1 1 d u M = − (u θ u θ )+ λθ | 2 1 − 1 ∞ c ′ for some λ C (M). Hence the (1, 1)-part of dd u M is given by ∈ | √ 1 − u + u +2λl θ1 θ1. 2 11 11 11 ∧  c ∞ On the other hand, the (1, 1)-part of dd (ρv) M for v C (Ω) coincides with | ∈ 1 1 √ 1v M l θ θ . − | 11 ∧ −1 c ′ If we choose v so that v M =2 ∆bu λ, the (1, 1)-part of dd (u +ρv) M vanishes, | − c | c which gives the existence of u. From the construction, it follows that d u M = d u. | CR The uniqueness of u modulo O(ρ2) is a consequence of the above computation of c e e the (1, 1)-part of dd (ρv) M .  e | 8 YUYA TAKEUCHI

Set W = Scal √ 1A 1, W = W = Scal + √ 1A 1, 1 1 − − 11, 1 1 1 − 11, and define a two-form W (θ) on M by (3.3) W (θ)= W θ θ1 + W θ θ1. 1 ∧ 1 ∧ A contact form θ is said to be pseudo-Einstein if W (θ)=0. Lemma 3.5. The two-form W (θ)/2π is closed and gives a representative of the first 1,0 1,0 2 1,0 Chern class c1(T M). In particular, c1(T M)=0 in H (M, R) if (M,T M) admits a pseudo-Einstein contact form. Proof. It follows from (2.4) that W (θ)= √ 1Ω 1 d(Scal θ)= √ 1dω 1 d(Scal θ). − 1 − · − 1 − · 1,0 Hence W (θ)/2π is closed and represents the cohomology class c1(T M). 

c The two-form W (θ) depends on the choice of θ, but ddCR appears in its trans- formation law under conformal change. Lemma 3.6. For another contact form θˆ = eΥθ, W (θˆ)= W (θ) 3ddc Υ. − CR In particular, in the case that θ is pseudo-Einstein, so is θˆ if and only if Υ is CR pluriharmonic. Proof. Under the conformal change θˆ = eΥθ, eΥW = W 3P Υ; 1 1 − 1 see [Hir93, Lemma 5.4] for example.c Hence W (θˆ)= W θˆ θˆ1 + W θˆ θˆ1 = W (θ) 3ddc Υ, 1 ∧ 1 ∧ − CR which establishes the formula.c c 

4. CR Paneitz operator in dimension three Let (M,T 1,0M) be a closed three-dimensional strictly pseudoconvex CR manifold and θ a contact form on M. The CR Paneitz operator Pθ is defined by 1 1 Pθu = (P1 u), = (P1 u), . The CR Paneitz operator is formally self-adjoint and annihilates CR pluriharmonic c functions [GL88, Section 2]. We first show an integral formula for Pθ by using dCR. Lemma 4.1. For u C∞(M), ∈ c c u(Pθu)θ dθ = dCRu ddCRu. ZM ∧ − ZM ∧ c Proof. From the definition of dCR (3.1) and Lemma 3.2, it follows that

c c √ 1 1 1 1 1 dCRu ddCRu = − u1(P1 u)θ θ θ u1(P1 u)θ θ θ ZM ∧ 2 ZM h ∧ ∧ − ∧ ∧ i 1 1 1 = u (P1 u)+ u (P1 u) θ dθ 2 ZM h i ∧

= u(Pθu)θ dθ, − ZM ∧ NONNEGATIVITY OF THE CR PANEITZ OPERATOR 9 which completes the proof. 

Assume that (M,T 1,0M) is embeddable; this means (M,T 1,0M) can be em- bedded in CN for large N. Without loss of generality, we may assume that M is connected. Harvey and Lawson [HL75, Theorem 10.4] have shown that M bounds a two-dimensional strictly pseudoconvex Stein space. Lempert [Lem95, Theorem 8.1] has proved that, in this case, M can be realized as the boundary of a strictly pseudoconvex domain Ω in a two-dimensional complex projective manifold X. Take a Kähler form ω on X and a defining function ρ of Ω. For sufficiently large N > 0, the real (1, 1)-form c ω+ = Nω 2dd log( ρ) − − defines an (even) asymptotically complex hyperbolic metric on Ω; see [EMM91, GSB08,Mat16] for details. Denote by ∆+ the (nonnegative) Riemannian Laplacian for ω+.

Lemma 4.2. Let F be a smooth function on Ω with F M = u. The condition 2 c | ∆+F = O(ρ ) is equivalent to that dd F M has vanishing (1, 1)-part. In particular, c c 2 | d F M = d u if ∆+F = O(ρ ). | CR Proof. We first note that c 2 4dd F ω+ = (∆+F )ω . − ∧ + It follows from the definition of ω+ that − − ω2 = 8ρ 3dρ dcρ ddcρ + O(ρ 2). + − ∧ ∧ 1,0 Take a local frame Z1 of T X Ker ∂ρ, and set Z1 = Z1 . Since M is strictly c ∩ pseudoconvex, dd ρ(eZ1 , Z1 ) is non-zero near M, ande e e e c c −2 dd F (Z1 , Z1 ) c c −1 dd F ω+ =2ρ dρ d ρ dd ρ + O(ρ ). ∧ c ∧ ∧ dd ρ(Ze1 , Ze1 ) This proves the former statement. Thee lattere one is a consequence of Lemma 3.4. 

We need a solution to the Dirichlet problem for ∆+. Near M, the Laplacian ∆+ satisfies 2 ∂ ∂ (4.1) ∆+ = ρ +2ρ + (a diff. op. increasing the order in ρ); − ∂ρ ∂ρ c.f. [GL88, Proposition 2.1].

Proposition 4.3. For each u C∞(M), there exist smooth functions F and G on ∈ 2 Ω such that F M = u and u = F + Gρ log( ρ) satisfies ∆+u =0. | − Proof. For a smooth functione f on Ω, (4.1) implies that e k k k+1 ∆+(fρ )= k(2 k)fρ + O(ρ ), − k k k k+1 ∆+(fρ log( ρ)) = k(2 k)fρ log( ρ)+2(1 k)fρ + O(ρ log( ρ)). − − − − − By using an inductive argument and Borel’s lemma, we obtain smooth functions F0 and G on Ω such that F0 M = u and | 2 ∞ ∞ H = ∆+(F0 + Gρ log( ρ)) ρ C (Ω), − ∈ 10 YUYA TAKEUCHI where ρ∞C∞(Ω) is the space of smooth functions on Ω that vanish to infinite order on M. Epstein, Melrose, and Mendoza [EMM91] have proved that ∆+ has a bounded inverse R, and it maps ρ∞C∞(Ω) to ρ2C∞(Ω); see also [GSB08, Section 2 5.1]. If we set F = F0 R(H), then we have F M = u and ∆+(F + Gρ log( ρ)) = − | − 0. 

2 c c Note that ∆+F = O(ρ ), and so d F M = d u by Lemma 4.2. Now we give a | CR proof of the nonnegativity of the CR Paneitz operator. Proof of Theorem 1.1. Fix a smooth function u on M, and let u = F +Gρ2 log( ρ) c − be as in Proposition 4.3. Since dd u ω+ = 0, we have ∧ e ddcu ddcu 0 e ∧ ≤ with equality if and only if ddcu = 0. The one-form dc(Gρ2 log( ρ)) can be ex- e e − tended continuously to Ω and vanishes along M. On the other hand, e (4.2) ddc(Gρ2 log( ρ)) ρ2 log( ρ)ddcG + (2ρ log( ρ)+ ρ)d(Gdcρ) (mod dρ). − ≡ − − The right hand side is continuous up to M and equal to zero on M. By using Stokes’ theorem, we have

c c c c dCRu ddCRu = d F dd F ZM ∧ ZM ∧ = lim dcu ddcu →+0 ǫ Zρ=−ǫ ∧ e e = lim ddcu ddcu →+0 ǫ Zρ<−ǫ ∧ 0. e e ≤ Hence Lemma 4.1 yields the nonnegativity of the CR Paneitz operator. If the equality holds, then u must be pluriharmonic on Ω, and so ddcu dρ = 0 on Ω. ∧ From (4.2), it follows that e e ddc(Gρ2 log( ρ)) dρ = ρ2 log( ρ)ddcG + (2ρ log( ρ)+ ρ)d(Gdcρ) dρ − ∧ − − ∧   is continuous up to M and vanishes along M. Therefore we have ddcF dρ = 0 c c ∧ on M, which implies dd F M = dd u = 0. In particular, u is annihilated by Pθ if | CR and only if u is CR pluriharmonic.  From the above proof, we obtain the following Corollary 4.4. Any CR pluriharmonic function u on M admits a smooth pluri- c c harmonic extension u to Ω, and d u M = d u holds. | CR Proof. Let u be a CRe pluriharmonice function on M, and u = F + Gρ2 log( ρ) − be as in Proposition 4.3. From the proof of Theorem 1.1, it follows that u is e pluriharmonic on Ω; in other words, ddcu = 0. On the other hand, ddcu is equal to e 2G log( ρ)dρ dcρ modulo a two-form continuous up to the boundary. Thus we − ∧ e ∞ ∞ e have G M = 0. The construction of G implies that G ρ C (Ω); in particular, u | c ∈ c c is smooth up to the boundary. The equality d u M = dCRu follows from dd u =0 | e and Lemma 3.4.  e e We apply Corollary 1.3, which follows from Theorem 1.1 and [CMY17, Theorem 1.2], to give an affirmative solution to the CR Yamabe problem for embeddable CR manifolds. NONNEGATIVITY OF THE CR PANEITZ OPERATOR 11

Proof of Theorem 1.4. If (M,T 1,0M) < (S3,T 1,0S3), then there exists a CR Y Y Yamabe contact form on M [JL87, Theorem 3.4]. On the other hand, if (M,T 1,0M)= Y (S3,T 1,0S3), then (M,T 1,0M) is CR equivalent to (S3,T 1,0S3) by Corollary 1.3, Y and a solution of the CR Yamabe problem exists [JL87, Theorem 7.2]. 

5. CR Q-curvature in dimension three In this section, we discuss the zero CR Q-curvature problem in dimension three and some applications. Let (M,T 1,0M) be a closed three-dimensional strictly pseu- doconvex CR manifold and θ a contact form on M. Hirachi [Hir93] has introduced the CR Q-curvature Qθ by

1 11 Qθ = (∆bScal 2 Im A ). 6 − 11, He has derived also the divergence formula

1 1 1 1 Qθ = W = W −3 1, −3 1, and its transformation law under conformal change θˆ = eΥθ: 2Υ e Qθˆ = Qθ + PθΥ; see [Hir93, Lemma 5.4]. In particular, the CR Q-curvature vanishes for pseudo- Einstein contact forms. We first discuss orthogonality relations between Qθ and . The integral of the P product of a CR pluriharmonic function and the CR Q-curvature has a cohomolog- ical expression.

Proposition 5.1. For u , ∈P c 1,0 3 [dCRu] c1(T M), [M] = uQθθ dθ. h ∪ i 2π ZM ∧ c Proof. The definitions of dCR (3.1) and W (θ) (3.3) yield that

c √ 1 1 1 1 1 dCRu W (θ)= − u1W1 θ θ θ u1W1 θ θ θ ZM ∧ 2 ZM h ∧ ∧ − ∧ ∧ i 1 1 1 = u W1 + u W1 θ dθ 2 ZM h i ∧

=3 uQθθ dθ. ZM ∧ 1,0 Since W (θ)/2π is a representative of c1(T M), we have the desired conclusion. 

Lemma 3.3 gives the following

Corollary 5.2. If u is the real part of a CR holomorphic function globally, then Qθ is orthogonal to u. Similarly, the vanishing of the first Chern class implies the orthogonality of the CR Q-curvature to . P 1,0 2 Corollary 5.3. If c1(T M)=0 in H (M, R), then Qθ is orthogonal to . P Now we give a complete solution to the zero CR Q-curvature problem for em- beddable CR manifolds. To this end, we recall some functional-analytic properties of the CR Paneitz operator obtained by Hsiao. 12 YUYA TAKEUCHI

Theorem 5.4 ([Hsi15, Theorem 1.2]). Let (M,T 1,0M) be a closed embeddable strictly pseudoconvex CR manifold of dimension three and θ a contact form on M. The maximal closed extension of Pθ is self-adjoint with closed range. Moreover, if ⊥ ∞ u Dom Pθ (Ker Pθ) satisfies Pθu C (M), then u must be smooth. ∈ ∩ ∈

Denote by Πθ the orthogonal projection to Ker Pθ. We write Gθ for the partial 2 inverse of Pθ; that is, Gθ is a bounded linear operator from L (M) to Dom Pθ ⊥ ∩ (Ker Pθ) satisfying

2 u = Πθu + PθGθu u L (M), ∈ u = Πθu + GθPθu u Dom Pθ. ∈ ∞ From the latter statement of Theorem 5.4, it follows that Gθ maps Ran Pθ C (M) ∞ ∞ ∩ to itself, and Πθ defines a linear operator C (M) Ker Pθ C (M); in particular, ∞ → ∩ = Ker Pθ C (M) is dense in Ker Pθ. P ∩

Proof of Theorem 1.5. Fix a contact form θ0 on M. If Qθ0 is orthogonal to , then P

θ = exp( Gθ0 Qθ0 ) θ0 − · is a contact form with zero CR Q-curvature. Hence it suffices to show

c 1,0 [d u] c1(T M), [M] =0 h CR ∪ i for any u . Take a smooth pluriharmonic extension u of u to Ω by Corollary 4.4. ∈P Fix a Hermitian metric on T 1,0Ω, and denote by Ψ the corresponding first Chern e form. By applying Stokes’ theorem, we have

c c c 1,0 0= dd u Ψ= dCRu Ψ M = [dCRu] c1(T M), [M] ; ZΩ ∧ ZM ∧ | h ∪ i

e 1,0 1,0 here we use the fact that Ψ M is a representative of c1(T M) = c1(T Ω M ). If | | θˆ = eΥθ is also a contact form with zero CR Q-curvature, then Υ is annihilated by Pθ, and so it is CR pluriharmonic by Theorem 1.1. 

′ Case and Yang [CY13] have defined the Q-prime curvature Qθ by

′ 1 2 2 Q = Scal 2 A + ∆bScal. θ 2 − | 11| For any contact forms θ and θˆ = eΥθ, the integral of the Q-prime curvature satisfies

′ ˆ ˆ ′ (5.1) Qθˆθ dθ = Qθθ dθ +6 [Υ(PθΥ)+2QθΥ]θ dθ; ZM ∧ ZM ∧ ZM ∧ see [CY13, Proposition 6.1]. In particular, a discussion in [CY13] yields that the integral of the Q-prime curvature

′ 1,0 ′ Q (M,T M)= Qθθ dθ ZM ∧ is independent of the choice of a contact form θ with zero CR Q-curvature, and defines a CR invariant of (M,T 1,0M), which we call the total Q-prime curvature. By Theorem 1.5, the total Q-prime curvature is always well-defined for closed em- beddable CR three-manifolds. Now we give a proof of Theorem 1.6, which relates the total Q-prime curvature with the CR Yamabe constant. NONNEGATIVITY OF THE CR PANEITZ OPERATOR 13

Proof of Theorem 1.6. Fix a contact form θ with zero CR Q-curvature. Theo- rem 1.4 implies that there exists Υ C∞(M) such that θˆ = eΥθ satisfies ∈ Scal = (M,T 1,0M), θˆ dθˆ =1. Y ZM ∧ Then d 2 ′ ˆ ˆ 1 2 ˆ ˆ ˆ ˆ Qθˆθ dθ = (Scal) θ dθ 2 A11 θ dθ ZM ∧ 2 ZM ∧ − ZM ∧ 1 d b (M,T 1,0M)2 ≤ 2Y

On the other hand, it follows from (5.1) and the nonnegativity of Pθ that

′ ′ ˆ ˆ 1,0 Qθˆθ dθ = Q (M,T M)+6 Υ(PθΥ)θ dθ ZM ∧ ZM ∧ ′ Q (M,T 1,0M). ≥ Therefore we get the desired inequality. Moreover, if the equality holds, then θˆ has vanishing Tanaka-Webster torsion, and it is a pseudo-Einstein contact form since Scal is constant. Conversely, suppose that (M,T 1,0M) admits a pseudo-Einstein contact form θ with vanishing Tanaka-Webster torsion; in this case, the Tanaka- d Webster scalar curvature is constant. Without loss of generality, we may assume that θ dθ = 1. From Theorem 1.4 and [Wan15, Theorem 4], it follows that θ M ∧ is a CRR Yamabe contact form, and so the equality holds. 

Theorem 1.6 gives a generalization of [CY13, Theorem 1.1] to embeddable CR manifolds.

Proof of Corollary 1.7. Since 0 (M,T 1,0M) (S3,T 1,0S3), we have ≤ Y ≤ Y (M,T 1,0M)2 (S3,T 1,0S3)2 Y ≤ Y ′ =2Q (S3,T 1,0S3); here the last equality follows from Theorem 1.6 and the fact that the standard con- tact form on S3 is a pseudo-Einstein contact form with vanishing Tanaka-Webster torsion. Moreover, if the equality holds, then (M,T 1,0M) = (S3,T 1,0S3), and Y Y (M,T 1,0M) is CR equivalent to (S3,T 1,0S3) by Corollary 1.3. 

6. Logarithmic singularity of the Szegő kernel Let Ω be a strictly pseudoconvex domain in a two-dimensional complex manifold X. We recall some facts on Fefferman defining functions; see [HPT08, Section 2D] for example. For a local coordinate z = (z1,z2) of X, set ϕ ∂ϕ/∂zb z[ϕ]= det J − ∂ϕ/∂za ∂2ϕ/∂za∂zb If w = F (z) is another local coordinate, then ′ 2 w[ϕ]= det F z[ϕ], J | | J where F ′ is the holomorphic Jacobi matrix of F . A Fefferman defining function is a defining function ρ of Ω such that

c c 3 (6.1) dd log z[ρ]= dd O(ρ ) J 14 YUYA TAKEUCHI for any local coordinate z. The condition (6.1) is equivalent to that, for any p ∂Ω, ∈ there exists a local coordinate z near p such that 3 z[ρ]=1+ O(ρ ). J c It is known that θ is a pseudo-Einstein contact form on ∂Ω if and only if θ = d ρ ∂Ω | for a Fefferman defining function ρ of Ω; see [Hir93, Lemma 7.2] and [HPT08, Proposition 2.10] for example. We say ∂Ω to be obstruction flat if we can take c a defining function ρ of Ω such that dd log z[ρ] vanishes to infinite order near a J given point on ∂Ω. Now we give a necessary and sufficient condition for the vanishing of ψθ in (1.2) to higher-order on the boundary.

Proof of Theorem 1.8. From (1.3), it follows that ψθ ∂Ω = 0 is equivalent to Qθ = | 0. Since ∂Ω has a pseudo-Einstein contact form, ψθ ∂Ω = 0 if and only if θ is | pseudo-Einstein by Theorem 1.5. In what follows, we suppose that θ is pseudo- c Einstein. Take a Fefferman defining function ρ with θ = d ρ ∂Ω. For a point | 3 p ∂Ω, take a local coordinate z near p such that z[ρ]=1+ O(ρ ). Hirachi, ∈ J Komatsu, and Nakazawa [HKN93] have computed ψθ up to second order in this setting. From [Gra87, Proposition 1.8] and [HKN93, Proposition 1 and Theorem 2 1], it follows that ψθ = O(ρ ) if and only if ∂Ω is obstruction flat. On the other 3 hand, we derive from [HKN93, Remark 2] that ψθ = O(ρ ) if and only if ∂Ω is spherical. 

Acknowledgements The author is grateful to Kengo Hirachi for valuable suggestions on the logarith- mic singularity of the Szegő kernel. He would like to express his gratitude to Jeffrey Case, Sagun Chanillo, and Paul Yang for helpful comments. He would like to thank the referees for their comments also, which are helpful for the improvement of the manuscript.

References

[ABC+96] J. Amorós, M. Burger, K. Corlette, D. Kotschick, and D. Toledo, Fundamental groups of compact Kähler manifolds, Mathematical Surveys and Monographs, vol. 44, Amer- ican Mathematical Society, Providence, RI, 1996. [BdMS76] L. Boutet de Monvel and J. Sjöstrand, Sur la singularité des noyaux de Bergman et de Szegő, Journées: Équations aux Dérivées Partielles de Rennes (1975), 1976, pp. 123– 164. Astérisque, No. 34–35. [CCC07] S.-C. Chang, J.-H. Cheng, and H.-L. Chiu, A fourth order curvature flow on a CR 3-manifold, Indiana Univ. Math. J. 56 (2007), no. 4, 1793–1826. [CCY12] S. Chanillo, H.-L. Chiu, and P. Yang, Embeddability for 3-dimensional Cauchy- Riemann manifolds and CR Yamabe invariants, Duke Math. J. 161 (2012), no. 15, 2909–2921. [CCY13] S. Chanillo, H.-L. Chiu, and P. Yang, Embedded three-dimensional CR manifolds and the non-negativity of Paneitz operators, Geometric analysis, mathematical relativity, and nonlinear partial differential equations, 2013, pp. 65–82. [CCY15] J. S. Case, S. Chanillo, and P. Yang, A remark on the kernel of the CR Paneitz operator, Nonlinear Anal. 126 (2015), 153–158. [CCY16] J. S. Case, S. Chanillo, and P. Yang, The CR Paneitz operator and the stability of CR pluriharmonic functions, Adv. Math. 287 (2016), 109–122. [CE18] S. N. Curry and P. Ebenfelt, Bounded strictly pseudoconvex domains in C2 with ob- struction flat boundary, 2018. arXiv:1803.09053. NONNEGATIVITY OF THE CR PANEITZ OPERATOR 15

[CE19] S. N. Curry and P. Ebenfelt, Bounded strictly pseudoconvex domains in C2 with ob- struction flat boundary II, Adv. Math. 352 (2019), 611–631. [CG17] J. S. Case and A. R. Gover, The P ′-operator, the Q′-curvature, and the CR tractor calculus, 2017. arXiv:1709.08057, to appear in Ann. Sc. Norm. Super. Pisa Cl. Sci. [CKS19] S.-C. Chang, T.-J. Kuo, and T. Saotome, Global existence and convergence for the CR Q-curvature flow in a closed strictly pseudoconvex CR 3-manifold, 2019. arXiv:1905.01783. [CMY17] J.-H. Cheng, A. Malchiodi, and P. Yang, A positive mass theorem in three dimensional Cauchy-Riemann geometry, Adv. Math. 308 (2017), 276–347. [CMY19] J.-H. Cheng, A. Malchiodi, and P. Yang, On the Sobolev quotient of three-dimensional CR manifolds, 2019. arXiv:1904.04665. [CY13] J. S. Case and P. Yang, A Paneitz-type operator for CR pluriharmonic functions, Bull. Inst. Math. Acad. Sin. (N.S.) 8 (2013), no. 3, 285–322. [EMM91] C. L. Epstein, R. B. Melrose, and G. A. Mendoza, Resolvent of the Laplacian on strictly pseudoconvex domains, Acta Math. 167 (1991), no. 1-2, 1–106. [EO08] J. B. Etnyre and B. Ozbagci, Invariants of contact structures from open books, Trans. Amer. Math. Soc. 360 (2008), no. 6, 3133–3151. [FH03] C. Fefferman and K. Hirachi, Ambient metric construction of Q-curvature in conformal and CR geometries, Math. Res. Lett. 10 (2003), no. 5-6, 819–831. [GL88] C. R. Graham and J. M. Lee, Smooth solutions of degenerate Laplacians on strictly pseudoconvex domains, Duke Math. J. 57 (1988), no. 3, 697–720. [Gra87] C. R. Graham, Scalar boundary invariants and the Bergman kernel, Complex analysis, II (College Park, Md., 1985–86), 1987, pp. 108–135. [GSB08] C. Guillarmou and A. Sá Barreto, Scattering and inverse scattering on ACH manifolds, J. Reine Angew. Math. 622 (2008), 1–55. [Gur99] M. J. Gursky, The principal eigenvalue of a conformally invariant differential opera- tor, with an application to semilinear elliptic PDE, Comm. Math. Phys. 207 (1999), no. 1, 131–143. [Hir14] K. Hirachi, Q-prime curvature on CR manifolds, Differential Geom. Appl. 33 (2014), no. suppl., 213–245. [Hir93] K. Hirachi, Scalar pseudo-Hermitian invariants and the Szegő kernel on three- dimensional CR manifolds, Complex geometry (Osaka, 1990), 1993, pp. 67–76. [HKN93] K. Hirachi, G. Komatsu, and N. Nakazawa, Two methods of determining local invari- ants in the Szegő kernel, Complex geometry (Osaka, 1990), 1993, pp. 77–96. [HL75] F. R. Harvey and H. B. Lawson Jr., On boundaries of complex analytic varieties. I, Ann. of Math. (2) 102 (1975), no. 2, 223–290. [HPT08] P. D. Hislop, P. A. Perry, and S.-H. Tang, CR-invariants and the scattering operator for complex manifolds with boundary, Anal. PDE 1 (2008), no. 2, 197–227. [Hsi15] C.-Y. Hsiao, On CR Paneitz operators and CR pluriharmonic functions, Math. Ann. 362 (2015), no. 3-4, 903–929. [JL87] D. Jerison and J. M. Lee, The Yamabe problem on CR manifolds, J. Differential Geom. 25 (1987), no. 2, 167–197. [Lee88] J. M. Lee, Pseudo-Einstein structures on CR manifolds, Amer. J. Math. 110 (1988), no. 1, 157–178. [Lem95] L. Lempert, Algebraic approximations in analytic geometry, Invent. Math. 121 (1995), no. 2, 335–353. [Mat16] Y. Matsumoto, GJMS operators, Q-curvature, and obstruction tensor of partially in- tegrable CR manifolds, Differential Geom. Appl. 45 (2016), 78–114. [Rum94] M. Rumin, Formes différentielles sur les variétés de contact, J. Differential Geom. 39 (1994), no. 2, 281–330. [Sam86] J. H. Sampson, Applications of harmonic maps to Kähler geometry, Complex differ- ential geometry and nonlinear differential equations (Brunswick, Maine, 1984), 1986, pp. 125–134. [SC11] T. Saotome and S.-C. Chang, The Q-curvature flow in a closed CR 3-manifold, Pro- ceedings of the 15th International Workshop on Differential Geometry and the 4th KNUGRG-OCAMI Differential Geometry Workshop [Volume 15], 2011, pp. 57–69. 16 YUYA TAKEUCHI

[Siu80] Y. T. Siu, The complex-analyticity of harmonic maps and the strong rigidity of com- pact Kähler manifolds, Ann. of Math. (2) 112 (1980), no. 1, 73–111. [Sta89] N. K. Stanton, Spectral invariants of CR manifolds, Michigan Math. J. 36 (1989), no. 2, 267–288. [Tak18] Y. Takeuchi, Ambient constructions for Sasakian η-Einstein manifolds, Adv. Math. 328 (2018), 82–111. [Wan15] X. Wang, On a remarkable formula of Jerison and Lee in CR geometry, Math. Res. Lett. 22 (2015), no. 1, 279–299.

Department of Mathematics, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan Email address: [email protected], [email protected]