Efficient computation of magnetic resonance spectra and related correlation functions from stochastic Liouville equations

It is found that a very powerful method for computing slow-motional ESR (and NMR) spectra may be developed by the use of the Lanczos algorithm (LA) modified to tridiagonalize complex-symmetric matrices. It leads to at least order of magnitude reductions in computation time and in computer storage requirements than the commonly used Rutishauser algorithm. This permits the rapid analysis of spectral problems that were previously too difficult. The formal similarity of these problems to those from Fokker-Planck equations in molecular dynamics suggests that the LA should also be a very powerful method for the latter.