Answers are grounded in this paper and cite their sources.
In this work, we investigate Krylov‐subspace iteration schemes for the linear systems arising in the Gauss integration of linear Poisson systems. Exploiting the relation between Gauss Runge–Kutta methods and diagonal Padé approximations, the ‐Arnoldi process yields approximations that preserve the quadratic Hamiltonian at the iterative level. We show that this preservation property is accompanied by an induced geometric structure: the projected dynamics defines a reduced Poisson structure matrix, and the corresponding Padé updates preserve this structure. In the case of a non‐singular Poisson structure matrix, this further implies preservation of the associated induced two‐form. The results thereby extend structure‐preserving Krylov techniques beyond the canonical symplectic setting to non‐canonical linear Poisson systems, including singular cases. Numerical experiments confirm the predicted structure‐preserving properties up to roundoff accuracy and demonstrate that the Krylov iteration may be terminated according to the accuracy requirements of the underlying time integrator without compromising the relevant structures.
No takes yet. Share an insight, caveat, or question.
Maier et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: