Fractal Weyl laws for asymptotically hyperbolic manifolds (Q359555)

Let \((X,g)\) be a compact Riemannian manifold of dimension \(n\), \(h^2\Delta_g+V\) the Schrödinger operator for a real potential \(V\) and \(\sigma_X(h)\) its eigenvalue spectrum (each eigenvalue being counted with multiplicity). The semi-classical Weyl law gives the asymptotic \[ \#(\sigma_X(h)\cap[a,b])=(2\pi h)^{-n} \text{Vol}(\{\xi\in T^*X,a\leq\|\xi\|_g^2\leq b\})+{\mathcal O}(h^{-n+1}),\quad h\to 0^+. \] and the following complementary estimate \(\#(\sigma_X(h)\cap[-h,h])={\mathcal O}(h^{-n+1})\). The main result of this paper is a Weyl upper bound of the last type for resonance counting on some hyperbolic manifolds. Let \((M,g)\) be an asymptotically hyperbolic conformally compact manifold, let \(K_M\subset T_1^*M\) be the trapped set for the geodesic flow on the cotangent bundle \(T_1^*M\) with upper Minkowski dimension \(\text{dim}_{\mathcal M}(K_M)\) and let \(\mathcal R_M(h)=\{\sigma\in\mathbb C|\sigma \text{ pole of the resolvent } (h^2\Delta_g-s(n-1-s))^{-1}\}\) be the resonance spectrum (including multiplicity) of the Laplacian \(h^2\Delta_g\) where the meromorphic continuation for the resolvent kernel from \(\{\text{Re } s> n-1\}\) to \(\mathbb C\) is well known. Then the authors prove the estimate \[ \#(\{h^2s(n-1-s)-1| s\in{\mathcal R}_M(h)\}\cap D(h)) ={\mathcal O}(h^{-\nu}), \quad h\to0^+ \] where \(D(h)\) is the complex disk \(\{z\in\mathbb C, |z|\leq h\}\) and \(\nu>\nu_M=(\text{dim}_{\mathcal M}(K_M)-1)/2\) (with equality possible when \(K_M\) Minkowski dimension is pure). The conformal metric \(g\) is taken even (in \textit{C. Guillarmou}'s sense [Duke Math. J. 129, No. 1, 1--37 (2005; Zbl 1099.58011)]) and the trapped set \(K_M\) is supposed hyperbolic (e.g. \((M,g)\) negatively curved): the fractal nature of the set \(K_M\) explains the expression \textit{fractal Weyl law} in the title. In case of an hyperbolic convex cocompact manifold \(M=\Gamma\backslash\mathbb H^{n}\) (e.g. a Schottky manifold or a quasifuchsian manifold), there is a correspondence between resonances, scattering poles and zeroes of the Selberg zeta function \(Z_M(s)=\exp\left(-\sum_{\gamma\in{\mathcal P_M}}\sum_{m=1}^{+\infty} \frac1{m}\frac{\text{e}^{-sm\ell_\gamma}}{\text{det}(\text{Id}-P_\gamma)}\right)\) where \({\mathcal P_M}\) is the set of closed geodesics on \(M\), the closed geodesic \(\gamma\) being of length \(\ell_\gamma\) and with Poincaré return map \(P_\gamma\). Hence, the preceding spectral bound gives, for any fixed \(R>0\), the bound on the zero divisor \({\mathcal D}_M\) of the Selberg zeta function \[ \#((\text{i}t+ D(R))\cap {\mathcal D}_M)={\mathcal O}(|t|^{\nu_M}),\quad |t|\to+\infty, \] where \(\nu_M\) coincides with the Poincaré exponent \(\delta_\Gamma\) of the discrete group \(\Gamma\) or the Hausdorff dimension of the limit set of \(\Gamma\); the Minkowski dimension is pure, equal to the Hausdorff dimension of the trapped set \(K_M\). This upper bound was proved when \(\Gamma\) is Schottky by K. K. Lin, M. Zworski and the reviewer [\textit{L. Guillopé} et al., Commun. Math. Phys. 245, No. 1, 149--176 (2004; Zbl 1075.11059)]. The proof of the upper bound for resonances counting is based on two recent advances: the new approach by \textit{A. Vasy} [Invent. Math. 194, No. 2, 381--513 (2013; Zbl 1315.35015)] on the effective meromorphic extension of geometric resolvents (in an appendix, the authors give a presentation of such construction in the case of the 2-d hyperbolic cylinder), the framework on resonances counting in hyperbolic open systems introduced by \textit{J. Sjöstrand} and \textit{M. Zworski} [Duke Math. J. 137, No. 3, 381--459 (2007; Zbl 1201.35189)]. Let us say a few words on the proof. The first step is the Vasy's pseudodifferential construction on the double \(X\) of the conformal compactification \(\overline M\): Vasy introduces a Fredholm pseudodifferential operator \(V(z,h)=P(z,h)-\text{i}Q\) whose spectrum contains the poles (including multiplicity) of the resolvent \(\Delta(z,h)^{-1}\) with \(\Delta(z,h)=h^2(\Delta_g-(n-1)^2/4)-(z+1)^2\), the resolvent and the inverse \(V(z,h)^{-1}\) being meromorphically continued from \(\{\text{Im } z>0\}\) to \(\{\text{Im } z>-C_0h\}\). More precisely, the operator \(Q\) is supported away from \(\overline M\), the operator \(P(z,h)\) operates on \(M\) as some weighted conjugate of the shifted laplacian \(\Delta(z,h)\). The second step is to introduce the conjugate of Vasy's operator \(P_t(z,h)=\text{e}^{-tF}TV(z,h)T^{-1}\text{e}^{tF}\) with \(F,T\) convenient pseudodifferential operators and to prove the estimate in the semiclassical Sobolev scale \(H_h^s:=H_h^s(X)\) \[ \|u\|_{H_h^{1/2}}\leq Ch^{-1}\|(P_t(z,h)-\text{i}th(A_R+A_E))u\|_{H_h^{-1/2}},\quad u\in{\mathcal C}^\infty(X) \] where \(A_R\) is of finite rank \({\mathcal O}(h^{-\nu})\) and a convenient bound is proved for \(\|A_E\|_{H_h^{1/2}\to H_h^{-1/2}}\). Various techniques of microlocal or semi-classical analysis are used, with classical (or exotic) pseudodifferential and Fourier integral operators. Analysis near the trapped set and near infinity are separated through introduction of appropriate escape functions on the cotangent bundle and the Hamiltonian flows induced by principal symbols (semiclassical and classical) are studied carefully. The bulk of the proof is based on positive commutator and resolvent estimates.