Numerical analysis of nonlinear eigenvalue problems

Eric Cancès, Rachida Chakir, Yvon Maday

Published 2009-05-11, updated 2009-06-05Version 2

We provide a priori error estimates for variational approximations of the ground state eigenvalue and eigenvector of nonlinear elliptic eigenvalue problems of the form $-{div} (A\nabla u) + Vu + f(u^2) u = \lambda u$, $\|u\|_{L^2}=1$. We focus in particular on the Fourier spectral approximation (for periodic problems) and on the $\P_1$ and $\P_2$ finite-element discretizations. Denoting by $(u_\delta,\lambda_\delta)$ a variational approximation of the ground state eigenpair $(u,\lambda)$, we are interested in the convergence rates of $\|u_\delta-u\|_{H^1}$, $\|u_\delta-u\|_{L^2}$ and $|\lambda_\delta-\lambda|$, when the discretization parameter $\delta$ goes to zero. We prove that if $A$, $V$ and $f$ satisfy certain conditions, $|\lambda_\delta-\lambda|$ goes to zero as $\|u_\delta-u\|_{H^1}^2+\|u_\delta-u\|_{L^2}$. We also show that under more restrictive assumptions on $A$, $V$ and $f$, $|\lambda_\delta-\lambda|$ converges to zero as $\|u_\delta-u\|_{H^1}^2$, thus recovering a standard result for {\em linear} elliptic eigenvalue problems. For the latter analysis, we make use of estimates of the error $u_\delta-u$ in negative Sobolev norms.