### Speakers

### Description

Polymorphism of elements or compounds means that these substances can exist in more than

one crystalline form. Due to the different mutual arrangement of molecules in these

polymorphs, both their static electron energies and their dynamic degrees of freedom vary.

As a result, different polymorphs of the same substance exhibit different physical properties.

Pharmaceutically active ingredients (APIs) often exhibit this polymorphism, rendering their

development significantly expensive and time consuming. At the same time, the

development of computational and predictive methods, together with the increased

availability of computational capacity, enables to predict crystal properties in silico,

bypassing part of the tedious experiments.

Quantum-chemical calculations for the solid state tend to use electron density functional

theory (DFT), which provide a relatively fair accuracy. Still, these DFT methods can be too

costly for complex molecular crystals, justifying the attempt to find some compromise

between the required accuracy and computational effort. Semi-empirical methods based on

the density functional tight binding (DFTB) could be this way, or at least could be helpful

for particular calculations. The lower computational demands of these methods are due to

the introduction of certain reference densities or pair-specific parameterizations. An

important step to predict the thermodynamically preferred polymorphic form is to know the

contributions of both the vibrational states and electron energy. For a sensible assessment of

the relative thermodynamic stability of the polymorphs, it is important too to take into

account the anharmonicity of vibrations in crystals. This is partially captured using the quasiharmonic approximation, which takes the energy of vibration with respect to the volume of

crystal.

A benchmark study, comparing the performance of DFT and DFTB calculations for

molecular crystals, is presented for several enantiopure or racemic crystals of APIs. DFT

calculations employ the PBE-D3/PAW level of theory implemented in the code VASP,

whereas the DFTB3 methods with the D4 dispersion correction method implemented in the

DFTB+ code is used as a DFTB representative. Both methods are used to calculate optimum

unit-cell geometries, their electron energies and the energy response to the variation of the

unit-cell volume. Phonon frequencies and densities of the phonon states are calculated in the

PHONOPY program at the both levels of theory. The aim of this procedure is to find an

efficient combination of DFTB and DFT methods for the individual computational steps

within the quasi-harmonic framework. Such a computational approach, validated through

reference experimental and DFT data, will enable to efficiently rank polymorphs also for

larger molecules in terms of their Gibbs energies at finite temperatures.