(443c) Bridging the Gap between Static Lattice Energy and Full Free Energy Differences of Organic Crystal Polymorphs When Predicting Crystal Energy Landscape

To reduce the risk of unknown polymorphs showing up in late stage development, crystal structure predictions (CSPs) have been employed to determine possible forms. Due to the fact that crystalline structures of small organic molecules can vary in properties, fields such as pharmaceutical formulation, organic electronics processing and explosives preservation have interest in fast and efficient methods to predict favorable structures. A failure to know all potential solid forms has led to recalls, reformulation, or a missed opportunity to utilize a material to its fullest extent. CSPs can be used to determine the solid energy landscape a priori by exhaustively generating solid forms and ranking them based on their relative energies.

Ranking of crystals in a CSP have historically relied on the lattice energy, but more recently free energy techniques have been introduced to help capture crystal reranking due to temperature changes. Thermodynamic ranking of the crystal structures allows one to determine the likely crystal structures to find experimentally. A CSP can generate 50-200 crystal structures within 2.5 kcal/mol of the global minimum so accurate ranking is critical when it’s likely that only a handful will actually crystallize. Lattice energy calculations have been the most widely used technique to rank crystal structures, but polymorph stability is known widely to be temperature dependent and therefore entropic effects cannot be neglected. Phonon calculations have been integrated to determine harmonic free energies for CSPs and, while not published, we are aware of free energy calculations using full molecular dynamics (MD) in industrial settings.

The required level of accuracy in free energy calculations for CSPs is unknown, which is why we have developed a whole host of methods in varying accuracy to compute the solid-solid free energy differences. MD is the gold standard, allowing us to compute a full thermodynamic cycle between two crystals for a given potential, but is fairly expensive to conduct on upwards of 200 crystals. On the opposite side of the spectrum are lattice dynamics techniques that use the static lattice phonons of the crystal to determine the entropic contributions to the free energy. Lastly, we have developed an intermediate technique to include the full dynamic behavior of molecular dynamics with the speed of compute the crystal phonons to come up with an approximation to the free energy that introduces error of 0.3 kcal/mol. I will discuss the specifics of these methods and compare their relative speed, accuracy, and direct implementation into a CSP to help screen for likely solid forms of molecules of interest.