How do I select the right model for the optical constants?

Selecting appropriate optical constant models with the right number of parameters is difficult and requires some experience. The choice depends – first of all – on the material and also on the spectral range of interest.

In the infrared the choice is rather easy: Use a constant refractive index, and add for each clearly visible vibrational mode a Kim oscillator. If you have free charge carriers (i.e. you deal with a metal or a doped semiconductor) you should add a simple Drude model (in case of highly doped semiconductors use an enhanced Drude model instead).

In the NIR, visible or UV the right choice is not so simple. Here you can be guided by the following rules of thumb:

Metals:
Use a constant to represent high energy interband transitiions. For every interband transition causing structure in the wanted spectral range use a Tauc-Lorentz or an OJL term. Take a Drude model for the free electrons.

Semiconductors (crystalline):
Combine a constant and Tauc-Lorentz terms for interband transitions. Add a Drude model for low doped materials. In the case of high doping levels use an extended Drude model.

Semiconductors (amorphous):
Setup a model like for crystalline semiconductors (see above) but replace the Tauc-Lorentz terms by OJL terms.

Oxides and Nitrides:
Combine a constant and an OJL term. If the OJL term alone cannot generate the right absorption and dispersion, you should add a harmonic oscillator in the far UV (resonance frequency 60000 1/cm, oscillator strength in the range 10000 … 100000 1/cm, damping 1 1/cm). In the fitting procedure the harmonic oscillator parameters should be kept fixed with the exception of the oscillator strength.

Organic materials:
Combine a constant, an OJL model and Kim oscillators. Whereas the OJL model is used to describe the ‘fundamental absorption’ in the UV, the Kim oscillators should represent absorption bands in the visible and NIR.

!!! Please note that OJL and Tauc-Lorentz terms may not be inserted in the susceptibility list as the other models. Instead, they have to be used within a KKR susceptibility object. !!!