Jacobi's Algorithm on Compact Lie Algebras

From MaRDI portal





This paper presents a cyclic version of \textit{N. J. Wildberger}'s algorithm [Proc. Am. Math. Soc. 119, No. 2, 649--655 (1993; Zbl 0788.22012)] for diagonalizing matrices in compact Lie algebras and proves its local quadratic convergence. The abstract framework covers Jacobi-like algorithms for skew-symmetric, skew-Hermitian, skew-symmetric Hamiltonian, and skew-Hermitian Hamiltonian matrices. The idea of the algorithm is as follows. Proceeding from a real root space decomposition of a compact Lie algebra \(\mathfrak g\), a cyclic sweep consists of sequentially transforming each root space component of a given matrix \(A \in \mathfrak g\) such that a certain cost function is minimized. These cyclic sweeps are repeated until the root space components of \(A\) are sufficiently small, i.e., when \(A\) has converged to an element of the torus algebra. The authors provide a detailed illustration of this algorithm for skew-Hermitian Hamiltonian matrices and demonstrate its quadratic convergence by several numerical experiments. A careful introduction makes the results of this paper accessible to readers with little background in Lie algebras.











This page was built for publication: Jacobi's Algorithm on Compact Lie Algebras

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4650982)