Spectrum of the Laplace operator and periodic geodesics: thirty years after (Q2477566)

The author of this paper is interested in giving an alternative proof of the so-called semi-classical trace formula (SCTF) for semiclassical Schrödinger operators. He emphasizes that after more than 30 years his original works on this subjects [see Zbl 0272.53034, Zbl 0281.53036], which are very close to Feynman path integral intuition, the proof is still mathematically difficult to follow. He underlines that the work by \textit{J. Duistermaat} and \textit{V. Guillemin} on this subject [see Zbl 0307.35071] is also difficult to read. Let \(\widehat H_h\) denote a Schrödinger operator depending on the parameter \(h\), simulating the ``Planck constant''. Then, the SCTF aims to compute for very small \(h\), the trace of a suitable function \(f(\widehat H_h)\) of \(\hat H_h\) by means of the spectrum \(S_p(\widehat H_h)\) of \(\widehat H_h\), i.e., \[ f(\widehat H_h)=\sum_jf(E_j(h)), \] where \(E_j(h)\in S_p(\widehat H_h)\), or by means of the \(2\)-points Schwartz kernel \([f(\widehat H_h)](x,y)\) of \(f(\widehat H_h)\), i.e., by the following integral representation: \[ f(\widehat H_h)= \int_X[f(\widehat H_h)](x,x)|dx|, \] where \(|dx|\equiv \sqrt{\det(g_{ij})}dx^1\wedge\cdots\wedge dx^n\). For example, if \(\widehat H_h={{1}\over{2}}h^2\triangle_g\), where \(\triangle_g\) is the Laplace operator on a compact Riemannian manifold \((X,g)\), then one tries to find exact or approximate formulas for the kernel \([f(\widehat H_h)](x,y)\) using the methods of partial differential equations. The author uses for \(f\) the function \(f(E)=e^{-itE/h}\) which corresponds to the Schrödinger equation. The goal is to prove the SCTF without using the full symbolic calculus of Fourier integral operators (Duistermaat and Guillemin), but to follow a more simple approach. This uses the time-dependent Schrödinger equation, some facts about the geodesic flow, the stationary phase approximation and the metaplectic representation as a computational tool. In fact the author shows that the calculus of determinants can be achieved by using the metaplectic representation. In particular he gives formulae for some determinants of discrete Sturm-Liouville operators, that coincide in the continuous case with the previous ones obtained by \textit{S. Levit} and \textit{U. Smilansky} [see Zbl 0374.34016]. After a detailed introduction where the purpose of the paper is well presented, it splits into 12 more sections. 1. About the history. (Here are recalled the works by Selberg, Huber, Gutzwiller, Balian and Bloch, Kac, Chazarain, Duistermaat and Guillemin, who first gave some important contributions on SCTF in the 1970s. More recent developments are also considered.) 2. The trace formula for a quantized twist. 3. Schrödinger operators. 4. Classical dynamics. 5. Statement of SCTF for Schrödinger operators. 6. A proof of SCTF for Schrödinger operators. 7. From the Schrödinger trace to the wave trace (i.e., with function \(f(E)=e^{-it\sqrt{2E/h}}\)). 8. Degenerate case. 9. An example: rational harmonic oscillator. 10. Applications to the inverse spectral problem. 11. Sturm-Liouville determinants and the metaplectic representation. (The metaplectic representation is a natural representation of the metaplectic group \(M_p(d)\) in \(L^2({\mathbb R}^d)\). This is a two-fold covering of the symplectic group \(S_p(d)\).) 12. Recent progress. 13. Open problems. (Here the author emphasizes two questions: 1) Are they isospectral Riemannian manifolds with different length spectra and conversely? (\textit{H. Huber} [see Zbl 0089.06101] interpreted SCTF as a formula relating eigenvalues of the Laplace operator and lengths of closed geodesics on a closed surface of a curvature \(-1\). This comes of the term lengths spectrum.) 2) Another open problem emphasized by the author is about the implications of SCTF for eigenvalues statistics. Reviewer's remark: It may be useful to emphasize that despite of SCTF's are very important tools in the spectral representation of classic objects, their systematic using as a quantum integration is in general unjustified, since it is unrelated with quantum dynamics. (On these subjects see the works on quantum PDE's of the reviewer of this paper.)