Maximum-Likelihood Parameter Estimation in Terahertz Time-Domain Spectroscopy

We present a maximum-likelihood method for parameter estimation in terahertz time-domain spectroscopy. We derive the likelihood function for a parameterized frequency response function, given a pair of time-domain waveforms with known time-dependent noise amplitudes. The method provides parameter estimates that are superior to other commonly used methods and provides a reliable measure of the goodness of fit. We also develop a simple noise model that is parameterized by three dominant sources and derive the likelihood function for their amplitudes in terms of a set of repeated waveform measurements. We demonstrate the method with applications to material characterization.