(392b) Prediction of Solid State Phase Diagrams Using Multistate Reweighting and Jacobian Mapping | AIChE

(392b) Prediction of Solid State Phase Diagrams Using Multistate Reweighting and Jacobian Mapping


Schieber, N. - Presenter, University of Colorado Boulder
Shirts, M., University of Colorado Boulder
Dybeck, E., University of Virginia
Abraham, N., The University of Colorado Boulder
In materials with multiple metastable crystalline packings (polymorphs) the specific polymorphic phase affects many properties of the material. These include optical properties of dyes, charge transport in semiconductors, detonation properties of energetic materials, and bioavailability in small molecule organic pharmaceuticals. For this reason, effective prediction of the solid state phase diagrams of small molecules is important in the efficient development of materials.

In this study, we present a new approach for the predicting solid phase diagrams using the calculation of the relative Gibbs free energy of polymorphs at a range of temperatures and pressures. Sampling is required in each polymorph at a range of temperatures and pressures. Multistate reweighting is used to calculate free energy differences between temperature and pressure points within a polymorph. These results are then combined with a reference Gibbs free energy obtained with a pseudo-supercritical path to obtain a surface representing the Gibbs free energy differences between polymorphs at each point. Determining the line of intersection of these surfaces, which can be done by several different adaptive methods. gives the coexistence lines in the phase diagram and the most stable in each region is the lowest free energy structure. With this method, the uncertainty in each coexistence point returned by the method can be efficiently and reliably estimated.

The computational scaling of multistate reweighting with system size is a common concern. In addition to developing this method, the size dependence of the uncertainty in the phase diagram has been determined, and thus the scaling of the method with increasing system sizes. This method scales slightly better than N, where N is the number of molecules in the system. To increase the efficiency by decreasing the number of sampled states required, we have also used configuration mapping, and demonstrated the large increase in efficiency this provides for the Lennard Jones system. Configuration mapping allows direct energy reevaluation of a trajectory in at other temperatures and pressures in order to increase phase space overlap. Increasing this overlap allows increased spacing of temperatures and pressures, significantly increasing the computational efficiency. We have shown the feasibility of these approaches by producing the solid phase diagrams of two different systems, benzene and Lennard-Jones spheres.