Singular equivariant asymptotics and Weyl's law. On the distribution of eigenvalues of an invariant elliptic operator (Q2629830)

The article under review is the third part of a series of papers concerning the equivariant asymptotic Weyl's law of invariant elliptic operators [the author et al., J. Funct. Anal. 255, No. 4, 777--818 (2008; Zbl 1149.47038); J. Funct. Anal. 256, No. 1, 91--128 (2009; Zbl 1157.47033)]. Here, the author considers a closed Riemannian manifold in place of a bounded domain in the Euclidean space. Let \(M\) be a compact connected \(n\)-dimensional Riemannian manifold without boundary. Let \(P_0\) be an elliptic classical pseudodifferential operator of order \(m\) on \(M\), with principal symbol \(p(x,\xi)\), which is positive and symmetric. Let \(P\) denote its self-adjoint extension. Let \(G\) be a compact connected Lie group acting effectively and isometrically on \(M\) such that \(P\) commutes with the induced representation of \(G\) in \(L^2(M)\). Then \[ L^2(M)\simeq \bigoplus_{\chi\in\widehat G} L^2(M)(\chi) \] as a representation of \(G\), and \(L^2(M)(\chi)\) is preserved by \(P\) for any \(\chi\). Let \[ N_\chi(\lambda) =(\dim\chi)\sum_{t\leq \lambda}\text{mult}_{\chi}(t), \] where \(\text{mult}_{\chi}(t)\) stands for the multiplicity of the eigenvalue \(t\) of \(P\) restricted to \(L^2(M)(\chi)\). The main result in the paper is the following Weyl's law, including an estimate for the remainder: \[ N_\chi(\lambda)=\frac{\dim\chi\, \dim V_\chi^H}{(n-\kappa)(2\pi)^{n-\kappa}} \text{vol}\big((\Omega\cap S^*M)/G\big)\, \lambda^{\frac{n-\kappa}{m}} + O\Big(\lambda^{(n-\kappa-1)/m}(\log\lambda)^{\Lambda}\Big), \] as \(\lambda\to+\infty\), provided \(n-\kappa\geq1\), where \(\kappa\) is the dimension of a \(G\)-orbit of principal type, \(H\) is a principal isotropy group, \(S^*M=\{(x,\xi)\in T^*M: p(x,\xi)=1\}\), \(\Omega=\mathbb J^{-1}(0)\) denotes the zero level of the canonical symplectic momentum map \(\mathbb J:T^*M\to \mathfrak{g}^*\), and \(\Lambda\) is a natural number which is bounded by the number of orbit types of the \(G\)-action on \(M\). The main tools are resolution of singularities and Fourier integral operators.