On the shift-invert Lanczos method for the buckling eigenvalue problem

We consider the problem of extracting a few desired eigenpairs of the buckling eigenvalue problem $Kx = λK_Gx$, where $K$ is symmetric positive semi-definite, $K_G$ is symmetric indefinite, and the pencil $K - λK_G$ is singular, namely, $K$ and $K_G$ share a non-trivial common nullspace. Moreover, in practical buckling analysis of structures, bases for the nullspace of $K$ and the common nullspace of $K$ and $K_G$ are available. There are two open issues for developing an industrial strength shift-invert Lanczos method: (1) the shift-invert operator $(K - σK_G)^{-1}$ does not exist or is extremely ill-conditioned, and (2) the use of the semi-inner product induced by $K$ drives the Lanczos vectors rapidly towards the nullspace of $K$, which leads to a rapid growth of the Lanczos vectors in norms and cause permanent loss of information and the failure of the method. In this paper, we address these two issues by proposing a generalized buckling spectral transformation of the singular pencil $K - λK_G$ and a regularization of the inner product via a low-rank updating of the semi-positive definiteness of $K$. The efficacy of our approach is demonstrated by numerical examples, including one from industrial buckling analysis.