The excess Gibbs-energy
of a two-component liquid molecular mixture
is modeled based on discrete clusters of molecules. These clusters
preserve the three-dimensional geometric information about local molecule
neighborhoods that inform the interaction energies of the clusters.
In terms of a discrete Markov-chain, the clusters are used to hypothetically
construct the mixture using sequential insertion steps. Each insertion
step and, therefore, cluster is assigned a probability of occurring
in an equilibrium system that is determined via the constrained minimization
of the Helmholtz free energy. For this, informational Shannon entropy
based on these probabilities is used synonymously with thermodynamic
entropy. A first approach for coupling the model to real molecules
is introduced in the form of a molecular sampling algorithm, which
utilizes a force-field approach to determine the energetic interactions
within a cluster. An exemplary application to four mixtures shows
promising results regarding the description of a variety of excess
Gibbs-energy curves, including the ability to distinguish between
structural isomers.