I do not know if I understand your question. Nevertheless, if you want to change the composition of the mixture I guess you should change the value of the chemical potential of the components. If your adsorbed mixture is in equilibrium with a bulk phase  (reservoir), you will need to know the chemical potentials of the components in that reservoir.

Dear Noe,

Thanks for your feedback and sorry that my question was not clear. Yes, the chemical potential of each component should be known prior to GCMC simulations. In theory, one can estimate the chemical potential of a single component i in a mixture using:

Miu_i   = Miu*_i (T,P) + RTlnx_i (assume ideal mixture)

where Miu*_i(T,P) and x_i are the chemical potential of the pure component i and mole fraction of that component. However, it is not a trivial task in practice. One has to perform a separate NPT simulation to estimate Miu_i in a mixture. The problem is that Widom insertion method has been reported to be not very reliable in estimating Miu_i in a mixure. 

I myself would like to use GCMC for my work but the estimation of chemical potential for each component in a mixture is a real pain. Unless I receive useful advice from our experts for estimating Miu_i, I am leaning toward the Gibbs ensemble. 

Khoa.