Maximum entropy method and spectral RMC
In experimental physics, it is usual to fit models to data to obtain useful physical information. The model may be quite detailed, and include information on the energies of microscopic states in the system (from which a continuous spectrum can be derived). In this kind of modelling, there are many more unknowns than data points, and many sets of microscopic states could give a good fit to the data. However, many of those sets are unphysical (for instance predicting negative numbers of states).
Information theory techniques such as the maximum-entropy method provide a way to choose a physical result from the infinite number of models with a small chi-squared value. I have been involved with developing methods for mobility spectrum analysis of semiconductor devices, which may be used on real systems. I have also developed a technique for extracting the phonon density of states from specific heat measurements. In that paper, I also developed a spectral reverse Monte-Carlo technique, which is an alternative to maximum entropy, without the assumptions of the entropy functional.
- Application of Bryan's algorithm to the mobility spectrum analysis of semiconductor devices.D.Chrastina, J.P.Hague and D.Leadley. J. Appl. Phys. 94, pp6583-6590 (2003). (8 pages) [arXiv]
- High mobility SiGe heterostructures fabricated by low-energy plasma-enhanced chemical vapor deposition. H. von Kaenel, B. Roessner, D. Chrastina, G. Isella, J. P. Hague and M. Bollani. Microelectronic Engineering, 76 , pp279-284 (2004). (6 pages)
- Determining the phonon DOS from specific heat measurements via maximum entropy methods. J.P.Hague. J. Phys: Condens. Matter. 17, pp2397-2405 (2005) (9 pages) [arXiv]