Combining Spectral Amplitude and Phase Improves Terahertz Hyperspectral Imaging

We developed a method to analyze sample composition via hyperspectral imaging with terahertz time-domain spectroscopy (THz-TDS). Historically, THz spectral amplitude data alone have been most often used for hyperspectral analysis. We show advantages of simultaneously using both the amplitude and phase angle components of the THz transmission data in the analysis. Use of the complex transmission data improves the integrity of the sample component differentiation with respect to the sample composition. We incorporate principal component analysis and other simple machine learning algorithms into our data analysis methodology to automate the process of interpreting hyperspectral data and isolating unique components. As a demonstration, we automatically distinguish the structure of both a split-ring resonator and a sugar pellet sample.