Novel information theory techniques for phonon spectroscopy

Abstract

The maximum entropy method (MEM) and spectral reverse Monte Carlo (SRMC) techniques are applied to the determination of the phonon density of states (PDOS) from heat-capacity data. The approach presented here takes advantage of the standard integral transform relating the PDOS with the specific heat at constant volume. MEM and SRMC are highly successful numerical approaches for inverting integral transforms. The formalism and algorithms necessary to carry out the inversion of specific heat curves are introduced, and where possible, I have concentrated on algorithms and experimental details for practical usage. Simulated data are used to demonstrate the accuracy of the approach. The main strength of the techniques presented here is that the resulting spectra are always physical: Computed PDOS is always positive and properly applied information theory techniques only show statistically significant detail. The treatment set out here provides a simple, cost-effective and reliable method to determine phonon properties of new materials. In particular, the new technique is expected to be very useful for establishing where interesting phonon modes and properties can be found, before spending time at large scale facilities.