Length orthospectrum of convex bodies on flat tori (Q6056815)

This paper is devoted to a very deep study of the geodesics of the flat torus \(\mathbb{T}^d=\mathbb{R}^d/\mathbb{Z}^d\) (with \(d\ge 2\)) connecting orthogonally two convex bodies \(\Sigma_1\), \(\Sigma_2\subset\mathbb{T}^d\), where each \(\Sigma_j\) is either a point or the projection of a strictly convex compact subset \(K_j\subset\mathbb{R}^d\) with nonempty interior and oriented smooth boundary having all its sectional curvatures positive. Let \(\mathcal{P}_{K_1,K_2,T_0}\) be the set of geodesic arcs \(\gamma\) directly orthogonal to \(\Sigma_1\) and \(\Sigma_2\) of length \(\ell(\gamma)\) larger than \(T_0\). It turms out that if \(T_0\) is large enough than the number of geodesics in \(\mathcal{P}_{K_1,K_2,T_0}\) of length at most \(T>T_0\) is finite (of the order of \(T^d\)); this allows the authors to associate to this set of geodesics several analytic objects, whose behaviour is dictated by the geometry of the convex bodies. For instance, the authors introduce the generalised Epstein zeta function \[ \zeta_\beta(K_1,K_2,s)=\sum_{\gamma\in\mathcal{P}_{K_1,K_2,T_0}}\frac{e^{i\int_\gamma\beta}}{\ell(\gamma)^s}\;, \] where \(\beta\) is a closed real-valued 1-form on \(\mathbb{T}^d\). When \(\Sigma_1\) and \(\Sigma_2\) are points, this function reduces, up to a multiplicative factor, to the classical multi-dimensional zeta function introduced by \textit{P. Epstein}, [Math. Ann. 56, 615--644 (1903; JFM 34.0461.02)]. The authors prove that: \\ (i) if the cohomology class \([\beta]\) of \(\beta\) does not belong to \(H^1(\mathbb{T}^d,\mathbb{Z})\) then \(s\mapsto\zeta_\beta(K_1,K_2,s)\) extends holomorphically to~\(\mathbb{C}\); and\\ (ii) otherwise, \(s\mapsto\zeta_\beta(K_1,K_2,s)\) extends meromorphically to~\(\mathbb{C}\), with poles, all simple, in \(\{1,\ldots,d\}\).\\ Furthermore, the residues can be computed explicitly; in particular, when \(\beta=0\) they can be expressed in terms of the intrinsic volumes of the convex body \(K_1-K_2\). Analogously, the authors define the generalised Poincaré series \[ Z_\beta(K_1,K_2,s)=\sum_{\gamma\in\mathcal{P}_{K_1,K_2,T_0}}\exp\left(i\int_\gamma\beta-s\ell(\gamma)\right)\;, \] which is holomorphic on \(\{\Re(s)>0\}\). It turns out that \(s\mapsto Z_\beta(K_1,K_2,s)\) extends continuously to the imaginary axis outside the points \(\{\pm i\|\xi-[\beta]\|\mid\xi\in\mathbb{Z}^d\}\); moreover, the authors are able to describe precisely the behaviour of \(Z_\beta(K_1,K_2,s)\) at the singularities. The extension to the imaginary axis of the generalised Poincaré series is strictly related to the Fourier transform \(\widehat{T}_{\beta,K_1,K_2}\) of the counting measure \[ T_{\beta,K_1,K_2}(t)=\sum_{\gamma\in\mathcal{P}_{K_1,K_2,T_0}}e^{i\int_\gamma\beta}\delta\bigl(t-\ell(\gamma)\bigr)\;. \] The authors prove that the singular support of \(\widehat{T}_{\beta,K_1,K_2}\) is contained in the spectrum of the magnetic Laplacian \(\Delta_{-[\beta]}=(\partial_x-i[\beta])^2\) and they explicitly describe the singularities. All these results (and many others) are obtained via a very deep study of the analytic properties of the geodesic vector field \(V=\theta\cdot\partial_x\) on the unit tangent bundle \(S\mathbb{T}^d=\{(x,\theta)\in\mathbb{T}^d\times\mathbb{S}^{d-1}\}\). More precisely, the authors reduce the study of the sums appearing in the previous formulas to a detailed study of integrals of the form \[ \hat\chi(-iV)=\int_{\mathbb{R}} \chi(t)e^{-tV*}|dt| \] where \(\chi\) is a nice enough function (e.g., \(\chi(t)=t^{-s}\) or \(e^{-ts}\)) and \(e^tV\) is the geodesic flow. The study is performed in suitable anisotropic Sobolev spaces of distributions, defined using tools from harmonic analysis, and it depends on a very precise description of the asymptotic expansion as \(t\to+\infty\) of the pullback operator \(e^{-tV*}\). Finally, it should be mentioned that the analysis carried out in this very interesting paper is actually performed with respect to any translation invariant Finsler structure on \(\mathbb{T}^d\), not only with respect to the standard Euclidean structure.