An implicitly restarted symplectic Lanczos method for the symplectic eigenvalue problem (Q2706283)

Symplectic Lanczos methods are discussed for the eigenvalue problem \(Mx=\lambda x\). Here the matrix \(M\in {\mathbb{R}}^{2n\times 2n}\) is large, sparse and symplectic, i.e., \(MJM^T=J\) with NEWLINE\[NEWLINE J^{2n,2n}=\left[\begin{matrix} 0& I^{n,n} \cr I^{n,n} & 0 \end{matrix}\right] NEWLINE\]NEWLINE and \(I^{n,n}\) is the \(n\times n\) identity matrix. The Lanczos vectors are constructed to form a symplectic basis and to reduce the symplectic matrix into a butterfly form. The authors determine stopping criteria and conditions for the symplectic Lanczos method terminating prematurely and propose an implicitly restarted symplectic Lanczos scheme. This method is efficient for computing a few extremal eigenvalues of large, sparse and symplectic matrix. Numerical examples confirm the conclusion.