A Comparison of Algorithms for Predicting Optical Constants

Three programs were written to compute the optical constants of materials from normal incidence reflectance measurements. These include a Kramers-Kronig technique, a Fourier series algorithm, and an oscillator matching method. Each approach was used to calculate n and k for potassium bromide from measured reflectance data between 55-250 um. A 1% uncertainty in the reflectance data resulted in a 5%-10% uncertainty in the optical constants computed by the Kramers-Kronig analysis. The difficulty with this method of treating spectral bands beyond the measured region is also inherent to the Fourier series procedure. Although the Fourier technique correctly predicts the phase shift on reflectance, very low frequency data are needed to accurately compute the optical constants in the usualspectral region of interests. The oscillator matching method is a complex reflectance curve fitting routine which iteratively converges on n and k expressed as a sum of classical harmonic oscillators. As with the Kramers-Kronig analysis, the optical constants were found to be extremely sensitive to uncertainties in the measured reflectance, especially at high frequencies.