Bayesian spectroscopy in solid-state photo-physics

We propose Bayesian spectroscopy (BS) both for the spectral analyses of optical spectra and for normal mode analyses of coherent phonon (CP) signals in solid-state materials. In optical spectra due to elementary electronic excitations, multiple excitonic absorption peaks appear and are overlapped with other background optical transitions, which makes their spectral decomposition difficult. We demonstrate that BS can be used to estimate the band gap and the excitonic binding energies with high accuracies by the introduction of physical laws related with the excitonic transitions. For normal mode analyses of CP signals, we show that BS can estimate the frequency and initial vibrating phase of the normal mode with high accuracies. In contrast to the Fourier transform spectrum which has a broad spectral width on account of the damping behaviors of the CP signals, BS with a physical model of a damped oscillation improves the estimation accuracy of the normal mode frequency by two orders of magnitude compared to the Fourier transform method. These results indicate the significant advantages of BS over conventional methodologies such as the least-squares and Fourier transform methods. (paper)