The Hadamard variation of the ground state value of some quasi-linear elliptic equations (Q1340415)

The author extends a classical result of Hadamard on the first eigenvalue of the Laplacian over varying domains to nonlinear eigenvalues of \(p\)- Laplacian operators. Specifically, suppose \(\Omega\) is a bounded domain in \(\mathbb{R}^n\) with smooth boundary, with exterior unit normal \(\nu(x)\) at each point \(x\in \partial\Omega\) and fix a real, smooth function \(\rho\) defined on \(\partial\Omega\). For all (sufficiently) small \(\varepsilon\geq 0\), let \(\Omega_\varepsilon\) be the domain with boundary \(\partial\Omega_\varepsilon= \{x+ \varepsilon\rho(x) \nu(x)|x\in \partial\Omega\}\). The eigenvalue \(\lambda(\varepsilon)\) is the infimum of the \(L^p(\Omega_\varepsilon)\) norm of \(Du\), as \(u\) varies over all nonnegative functions in \(W^{1,p}_0(\Omega_\varepsilon)\) with \(L^{q+ 1}(\Omega_\varepsilon)\) norm normalized to be 1. If \(p\geq 2\) and \(p\leq q+ 1< p^*\), where \(p^*= \infty\) if \(p\geq n\) and \(p^*= np/(n- p)\) if \(p< n\), and if there is a unique function \(u_0\) such that the infimum \(\lambda(0)\) is attained only for \(u= u_0\), then \(\lambda(\varepsilon)= \lambda(0)+ A\varepsilon+ o(\varepsilon)\), where the coefficient \(A\) is given by \[ A= (1- p) \int_{\partial\Omega} \Biggl|{\partial u_0\over \partial\nu} (x)\Biggr|^p \rho(x) d\sigma(x), \] and \(\sigma\) is the surface measure on \(\partial\Omega\). A key element of the proof is a transformation which reduces the minimization problem over varying domains to one over a fixed domain.