A classical way of describing a dielectric function employs sums of contributions from damped harmonic oscillators. Each term leads to a maximum in the imaginary part of the dielectric function at the transversal optical (TO) resonance frequency of the corresponding oscillator. In contrast, the peak maxima of the negative imaginary part of the inverse dielectric function are attributed to the so-called longitudinal optical (LO) oscillator frequencies. The shapes of the corresponding bands resemble those of the imaginary part of the dielectric function. Therefore, it seems natural to also employ sums of the contributions of damped harmonic oscillators to describe the imaginary part of the inverse dielectric function. In this contribution, we derive the corresponding dispersion relations to investigate and establish the relationship between the transversal and longitudinal optical oscillator strength, which can differ, according to experimental results, by up to three orders of magnitude. So far, these differences are not understood and prevent the longitudinal optical oscillator strengths from proper interpretation. We demonstrate that transversal and longitudinal oscillator strengths should be identical for a single oscillator and that the experimental differences are in this case due to the introduction of a dielectric background in the dispersion formula. For this effect we derive an exact correction. Based on this correction we further derive a modified Kramers-Kronig sum rule for the isotropic case as well as for the components of the inverse dielectric function tensor. For systems with more than one oscillator, our model for the isotropic case can be extended to yield oscillator strengths and LO resonance wavenumber for uncoupled LO modes with or without dielectric background...
Download