In principle, MD or Monte Carlo simulations can accurately estimate EOS
properties for arbitrary gas compositions and arbitrary thermodynamic
conditions. In practice, people rarely do so. Here are the reasons that I
can think of, in order of decreasing importance:
1. Interatomic potentials;
a. These are often developed for condensed phases and so may not be
particularly accurate for gases. For example, the second virial coefficient
at 300K may be off by 10% or more.
b. They are usually developed for pure species. Mixture interactions are
computed using binary mixing rules, analogous to those used in
equations-of-state.
2. Statistical uncertainty: These methods are based on statistical sampling,
and so the accuracy converges rather slowly ~ 1/sqrt(CPU time)
3. If all you care about is macroscopic properties, it does not make sense
to use a method that is primarily concerned with microscopic properties.
Given all this, I think you would be better off taking your presumably
accurate pure-component CEOS functions and estimating the necessary binary
mixing parameters, and then using these to describe the multicomponent
mixtures. Don't they teach that in chemical engineering schools anymore?
:-).
Finally, always check your results against any available experimental data.
Good luck!
Aidan