Sharp exponential decay rates for anisotropically damped waves (Q6038730)

The authors consider the generalized damped wave equation \[ \left\{\begin{array}{l} \partial^2_tu-\Delta_gu+2W\partial_tu=0,\\ (u,\partial_tu)=(u_0,u_1) \end{array} \right. \] on a smooth compact Riemannian manifold \((M,g)\). Here \(\Delta_g\) is the Laplace-Beltrami operator (such that \(\Delta_g\leq0\)), \((u_0,u_1)^T\in H^1(M)\oplus L^2(M)\), \(W\,:\,L^2(M)\to L^2(M)\) is bounded, nonnegative and belongs to the space \(\Psi^0_{cl}(M)\) of classical pseudodifferential operators on \(M\) of order zero with polyhomogeneous symbol expansions. The main result of the paper (Theorem 2) states that the best exponential decay rate for solutions of~(1) is given by \[ \alpha:=2\min\{ -D_0,L_\infty \}, \] where \(D_0\) is the spectral abscissa, i.e. \(D_0:=\lim_{R\to0+}\sup\left\{ \text{Re}(\lambda)\,:\,|\lambda|>R,\,\lambda\in\text{Spec}(A_W) \right\}\), \(A_W:=\left( \begin{array}{cc} 0 & \text{Id}\\ \Delta_g & -2W \end{array} \right)\) is the generator of the solution semigroup for~(1), and \(L_\infty\) is the long-time limit of the time average of the damping along geodesics, i.e. \(L_\infty:=\lim_{t\to\infty}\inf_{(x,\xi)\in S^*M}\frac1t \int_0^t w(\varphi_s(x,\xi))ds\) with \(w:=\) principal symbol of \(W\) and \(\varphi_s:=\) the lift of the geodesic flow to \(T^*M\), \(S^*M:=\{(x,\xi)\in T^*M\,:\,|\xi|_g=1 \}\) (see details in the paper). The authors prove this result and, moreover, show that it implies the following statement (Theorem 1 in the paper): All solutions \(u\) to~(1) with \(W\in \Psi^0_{cl}(M)\) satisfy \[ E(u,t)\leq C e^{-\beta t}E(u,0) \] for some \(C,\beta>0\) and all \(t\geq0\) if and only if \(W\) satisfies Assumption 1 (Anisotropic Geometric Control Condition) and Assumption 2 (that the kernel of \(W\) does not contain any nontrivial eigenfunction of \(\Delta_g\)) which are formulated in the paper. Here \(E(u,t)\) is the energy \[ E(u,t):=\frac12\int_M |\nabla_g u(t,x)|^2+|\partial_t u(t,x)|^2\, dv_g(x), \] and \(dv_g\) is the Riemannian volume form on \(M\). Moreover, the authors provide some examples such that Assumptions 1 and 2 are satisfied.