Curvature dimension inequalities and subelliptic heat kernel gradient bounds on contact manifolds (Q2436554)

This paper provides an approach to synthetic Ricci curvature bounds associated to the sub-Laplacian on a contact Riemannian manifold, not necessarily of Sasakian type. Applications include Bonnet-Myers type compactness theorems, volume estimates for metric balls, gradient bounds and stochastic completeness for the associated heat semigroup, and spectral gap estimates. Let \({\mathbb M}\) be a smooth manifold of dimension \(2n+1\) equipped with a contact form \(\theta\) and a Riemannian metric \(g\). Denote by \(L\) the canonical subelliptic contact sub-Laplacian operator. The relevant curvature-dimension assumption involves the usual carré du champ operator \[ \Gamma(f,g) = \tfrac12(L(fg)-fLg-gLf) \] and its second-order analogue \[ \Gamma_2(f,g) = \tfrac12(L\Gamma(f,g) - \Gamma(f,Lg) - \Gamma(Lf,g)), \] together with another bilinear form \(\Gamma^Z\) and corresponding second-order analogue \(\Gamma_2^Z\) corresponding to the differentiation in the direction of the Reeb vector field \(Z\) on \({\mathbb M}\). In [Math. Ann. 358, No. 3--4, 833--860 (2014; Zbl 1287.53025)] the authors developed a sub-Riemannian curvature-dimension theory associated to such data under the following intertwining assumption \[ \Gamma(f,\Gamma^Z(f)) = \Gamma^Z(f,\Gamma(f)), \qquad f \in C^\infty_0({\mathbb M}). \] For contact manifolds \({\mathbb M}\), this assumption is equivalent to the condition that \({\mathbb M}\) be Sasakian. The main purpose of the paper under review is to develop a corresponding curvature-dimension theory in the absence of the Sasakian assumption. The authors introduce the following curvature-dimension inequality \(CD(\rho_1,\rho_2,\rho_3,\kappa,m)\): for all smooth, compactly supported functions \(f\) and all \(\nu > 0\), \[ \Gamma_2(f) + \nu \Gamma_2^Z(f) \geq \frac1{m} (Lf)^2 + \left( \rho_1 - \frac\kappa\nu \right) \Gamma(f) + (\rho_2 - \rho_3\nu^2) \Gamma^Z(f). \] They show that suitable bounds on the Ricci curvature and several other associated geometric tensors imply the validity of the inequality \(CD(\rho_1,\rho_2,\rho_3,1,2n)\) for a suitable choice of \(\rho_1,\rho_2,\rho_3\), and give a partial converse. The key parameter here is \(\rho_1\), which corresponds to a lower bound on Ricci curvature. In the last two sections of the paper the authors assume the validity of the curvature-dimension inequality \(CD(\rho_1,\rho_2,\rho_3,\kappa,\infty)\) for some \(\rho_1 \in {\mathbb R}\), \(\rho_2>0\), \(\rho_3>0\) and \(\kappa>0\) as well as some additional technical hypotheses. In this setting they obtain an exponential volume growth estimate for metric balls and stochastic completeness of the heat semigroup generated by \(L\) (see Theorem 4.4). Moreover, for a suitable choice of \(C\geq 0\), if \[ \rho_1 > \sqrt{\frac{\rho_3}{\rho_2}} \kappa + C \sqrt{\frac{\rho_2}{\rho_3}}, \] then \({\mathbb M}\) is compact. The authors conjecture that the same conclusion should hold under the weaker assumption \[ \rho_1 > \sqrt{\frac{\rho_3}{\rho_2}} \kappa. \] In the final section, they prove that if \(\rho_1 > \sqrt{\frac{\rho_3}{\rho_2}} \kappa\), then the operator \(-L\) has a spectral gap of size \(\lambda_1 \geq (\rho_1\rho_2 - \kappa \sqrt{\rho_2\rho_3})/(\rho_2+\kappa)\).