Are Elastic Properties of Molecular Crystals within Reach of Density Functional Theory? Accuracy, Robustness, and Reproducibility of Current Approaches

Solid form selection and design of
crystalline small molecule active
pharmaceutical ingredients (APIs) would benefit from computational
prediction and rationalization of the mechanical properties. Before
such practical applications can be considered, the robustness and
reproducibility of the computed properties with respect to the chosen
level of theory must be understood. In this work, elastic constants
of eight molecular crystals, with an emphasis on APIs, have been calculated
using dispersion-corrected density functional theory (DFT). The different
DFT methods considered do not, in general, consistently predict the
absolute magnitudes of the elastic moduli, which disagree by over
50% for some crystals. Relative properties such as elastic anisotropy
are more robust, mostly consistent between models, and in qualitative
agreement with experiment. Calculated anisotropies could also be rationalized
in terms of the structural features of the crystal. Overall, this
work reveals that DFT-computed elastic properties may not offer a
ground truth in absolute terms. Future applications of DFT in the
context of high-throughput material screening or training of machine
learning models will therefore require judicious selection of target
properties and evaluation metrics. Only after methodological limitations
are properly identified can more in-depth investigations be undertaken
to assess the feasibility of applying DFT in pharmaceutical development.