In this thesis collision-induced absorption (CIA) coefficients are computed using molec-ular dynamics (MD) simulations. Part I is dedicated to the theoretical frame of the method, from the classical theory radiation to the derivation of an absorption coefficient. The second part is a on the implementation of the method in the in-house software Spa-CIAL (Spectra of Collision-Induced Absorption with LAMMPS). This package is split in two parts: the molecular dynamics part being treated with the open source package LAMMPS, and the post-processing for the computation of the collision-induced absorp-tion with a Python code. The post-processing has been developed in two distinct ways each of them presenting different properties. The first one, based on what has been done previously, is designed to compute the dipole auto-correlation function (ACF) to obtain the CIA spectra after Fourier transformation. Many improvements has been made like the time averaging method is used in order to considerably increase the statistics requiring reasonable resource needs. The use of the fast Fourier transform algorithm (FFT) and the apodization procedure are also used for better accuracy of the results. The reformulation of the equations, especially with the Wiener-Kintchine (WK) theorem, gives a completely new implementation for which the CPU intensive computation of the dipole ACF is no longer needed. Instead, the contributions to the CIA spectrum are computed for each pair separately. In addition to improve significantly the performance of the code, it is now possible to separate the free-free and the bound-bound contributions. The comparison with the previous method (ACF) for the Ar-Xe system has shown a good accordance thus validating this new implementation. This great progress paves the way for the classical study of the dimers features in the absorption coefficient. The programs developed in this work can be adapted to handle molecular gas mixtures that are relevant in studies of radiative transfer in planetary atmospheres.