Evaluating matrix power series with the Cayley-Hamilton theorem

Abstract

The Cayley-Hamilton theorem is used to implement an iterative process for the efficient numerical computation of matrix power series and their differentials. In addition to straight-forward applications in lattice gauge theory simulations e.g. to reduce the computational cost of smearing, the method can also be used to simplify the evaluation of SU(N) one-link integrals or the computation of SU(N) matrix logarithms.

Comments: 18 pages, 3 algorithms, 3 figures, performance and accuracy discussion added