On the Hadamard formula for nonsmooth domains (Q860746): Difference between revisions

Let \(k_1\) and \(k_2\) denote two compact, selfadjoint nonnegative operators acting in Hilbert spaces \(H_1, H_2,\) respectively. The bounded operator \(S: H_1 \to H_2\) is a measure for the proximity of these two operators. The author proves the following asymptotic formula \[ \frac{1}{\mu_{mj}} = \frac{1}{\lambda_m} + \omega_{m,j} + O(\sigma_m^2(\rho_m+\sigma_m^2)) \] for the eigenvalue \(\lambda_m^{-1}\) of \(K_1\) with the corresponding eigenspace \(X_m\), where \(\mu_{mj}^{-1},j=1,\dots,\dim X_m,\) are eigenvalues of \(K_2\) located near \(\lambda_m^{-1}\) and \(\omega_{mj}\) are eigenvalues of an eigenvalue problem given by \(S,K_1,K_2\) and the parameters \(\rho_m,\sigma_m\) are constants in some inequalities. Basing on this result an asymptotic result is given for eigenvalues of an elliptic \((M \times M)-\)matrix operator of order \(2m\) for two bounded domains with nonempty intersection. Further, for the Laplace operator in Lipschitz domains the following asymptotic formula for the first eigenvalues \(\lambda_1\) and \(\mu_1\) of the Dirichlet-Laplacian in \(\Omega_1\) and \(\Omega_2\), respectively is a consequence of this general approach \[ \frac{1}{\mu_1} = \frac{1}{\lambda_1}- \frac{1}{\lambda_1}\int_{\Omega_1 \setminus \Omega_2}| \nabla\varphi| ^2\,dx+ \frac{1}{\lambda_1}\int_{\Omega_2}(| \nabla\psi| ^2-| \nabla\phi| ^2)dx+O(d^{1+\delta}), \] here \(\delta\) is a positive constant depending on the Lipschitz constants of domains, \(\varphi\) is the eigenfunction corresponding to \(\lambda_1\) and \(\phi, \psi \) are solutions of a linear boundary value problem. This asymptotic result is a generalization of the well-known Hadamard variational formula. For sufficiently smooth domains and an appropriate choice of the operator \(S\) Hadamard's classical result follows.