At least if we talk about infrared, solids and stronger oscillator stregths there are a couple of things that have to be added to Rafik's contribution. First of all, since Lambert-Beer is often not applicable, it is either a transmittance or reflectance spectrum that should be analyzed (and not pseudo-absorbance spectra). Those spectra are functions of the dielectric function tensor. Let's assume for simplicity a cubic crystal. Then the dielectric tensor becomes a scalar function and it is this function that shows bands with usually Lorentz-profiles (at least, if the oscillators are not extremely strong). The band shapes in the reflectance spectra or transmittance spectra can be quite variable once anisotropy comes into play or due to orientational averaging in case of isotropic samples. See e.g. the link below and references therein.
Article Modelling IR-spectra of single-phase polycrystalline materia...
The following publication covers fully the answer to your question:
The Gaussian profile works well for solid samples, powders, gels or resins. The
Lorentzian profile works best for gases, but can also fit liquids in many cases. The best functions for liquids are the combined G-L function or the Voigt profile. The Voigt profile is similar to the G-L, except that the line width Δx of the Gaussian and Lorentzian parts are allowed to vary independently. The treatment thus far applies equally to IR or Raman peaks. However, Raman spectra may also contain signals
due to fluorescence. Fluorescence involves an electronic transition combined with a vibrational transition. The higher energy needed to boost electrons (visible or UV
radiation) explains why this is rare in the IR, but more pronounced with high frequency lasers (depending upon the sample). Removal of the fluorescence signal can
sometimes be accomplished using background removal routines, but curve fitting can also be used, especially when the fluorescence is relatively narrow.
Curve Fitting in Raman and IR Spectroscopy:
Basic Theory of Line Shapes and Applications
Michael Bradley, Thermo Fisher Scientific, Madison, WI, USA
Conclusion
Curve fitting opens great power to the end user. The interpretation of protein peaks relies upon curve fitting to extract information about protein structure. The tremendous field of liquid dynamics is accessible, and calibrations based on
peak areas, rather than just heights, can be obtained. The fitting routine in OMNIC Peak Resolve was designed to be easy to use, rapidly convergent, and flexible. It provides an additional tool with excellent utility in many fields.
At least if we talk about infrared, solids and stronger oscillator stregths there are a couple of things that have to be added to Rafik's contribution. First of all, since Lambert-Beer is often not applicable, it is either a transmittance or reflectance spectrum that should be analyzed (and not pseudo-absorbance spectra). Those spectra are functions of the dielectric function tensor. Let's assume for simplicity a cubic crystal. Then the dielectric tensor becomes a scalar function and it is this function that shows bands with usually Lorentz-profiles (at least, if the oscillators are not extremely strong). The band shapes in the reflectance spectra or transmittance spectra can be quite variable once anisotropy comes into play or due to orientational averaging in case of isotropic samples. See e.g. the link below and references therein.
Article Modelling IR-spectra of single-phase polycrystalline materia...
I attach links to my reviews on line profiles in liquids/amorphous solids, in which the problem is treated in terms of time correlation functions(both are in my profile on ResearchGate). These reviews may be helpful from the point of view of both the general character of an approach described and curve fitting as well.
Chapter Novel Approaches in Spectroscopy of Interparticle Inter-Acti...
Article Spectroscopy of interparticle interactions in ionic and mole...
One more thing: In case of particulate matter the positions and the shapes of IR bands are depending on the shape of the particles... see link. This is of particular importance for strong bands and what you get is neither Gaussian nor Lorentzian (for such bands even the dielectric function shows deviations from Lorentzian shapes).
Article Huffman: Absorption and Scattering of Light by Small Particles