A Bayesian method for analysing relaxation spectra

The knowledge of electrical and mechanical properties of material, relies on a precise analysis of the relaxation spectra. We explore the ability of a Bayesian method to achieve an accurate estimation of spectral parameters. We implemented a parallel-tempering Markov-chain Monte Carlo algorithm and used it to fit simulated and measured spectra. An exhaustive testing of the code shows that it presents an extremely good performance, accurately fitting complex spectra under strong noise and overlapping components. We conclude that this technique is quite suitable for relaxation spectra analysis, complementing classical methods.