The asymptotic behavior of the principal eigenvalue in a singular perturbation problem with invariant boundaries (Q1095495)

Consider the deterministic dynamical system (1) \(\dot x(t)=B(x(t))\) on \({\mathbb{R}}^ d\), B is bounded and \(C^ 2\). The stochastic differential equation \[ (2)\quad dX_{\epsilon}(t)=(B(X_{\epsilon}(t))+\epsilon^ 2b(X_{\epsilon}(t)))dt+\epsilon \sigma (X_{\epsilon}(t))dW(t) \] is a diffusion random perturbation of (1), B and the matrix function \(\sigma\) are smooth and \(t\mapsto W(t)\) is a d-dimensional Wiener process. The generator of (2) is \(L_{\epsilon}=<B,\nabla >+\epsilon^ 2L\), where \[ L=\sum^{d}_{i=1}b^ i(x)\partial /\partial x_ i+\sum^{d}_{i,j=1}a^{ij}(x)\partial^ 2/\partial x_ i\partial x_ j,\quad (a^{ij}(x))_{ij}=\sigma (x)\sigma^*(x). \] Let \(G\subset {\mathbb{R}}^ d\) be a connected bounded domain with \(C^ 2\) boundary \(\partial G\). Denote by \(\lambda_{\epsilon}(G)\) the principal eigenvalue of \(L_{\epsilon}\) corresponding to zero boundary conditions on \(\partial G.\) The paper generalizes results of the second author on the asymptotic behaviour of \(\lambda_{\epsilon}(G)\) for \(\epsilon\to 0\) [J. Differ. Equations 37, 108-139 (1980; Zbl 0413.35010) and J. Anal. Math. 38, 255- 286 (1980; Zbl 0475.58019)]. In particular, the results of the present paper allow to treat the case when parts or all of the boundary \(\partial G\) are invariant for (1). This was excluded before.