(599c) Towards Digital Design of Crystals: Predicting Absolute Chemical Potentials for Solid, Solution and Gas Phases

Authors: 
Khanna, V., University of California, Santa Barbara
Doherty, M. F., University of California
Peters, B., University of Illinois Urbana-Champaign
Currently, more than 90% of small molecule drugs are delivered in crystalline form. Solubility of these crystalline molecules plays a crucial role in drug design as it directly affects the bioavailability of a drug. Also, the solubility as predicted by the force field is crucial in all efforts to predict nucleation and growth rates. The thermodynamic definition of solubility is, the concentration of the solute in the solvent which makes its (solute) chemical potential equal in both the phases (solid & solution).

Solubility prediction using atomistic simulations is a challenging task. The solubility prediction of NaCl took nearly a decade of research efforts. The solubility prediction of polyatomic molecules is an ongoing effort and we have developed a new approach that employs independent predictions of absolute chemical potentials in the solid and solution phases without having a common starting reference system. For the solid phase we apply the Einstein crystal method in its unaltered form although we have extended it differently from some of the previous works to compute the free energy of the molecular crystal (solid). In order to compute the absolute chemical potential in the solution phase, we have developed a new gas phase reference system to compute the absolute chemical potential of an isolated polyatomic gas molecule. To which we add the solvation free energy of the molecule at a given solute concentration to give the absolute chemical potential in the solution phase.

We leverage and demonstrate the computational tools developed so far to predict the solid-vapor equilibrium curve for succinic acid (our model compound), i.e., predict its sublimation vapor pressure as a function of temperature. We find excellent agreement between our predicted results and experiments in the literature. Thus, having tested our methodology we are currently applying it to the solubility problem.