Informatics-Guided Computational Design of Polyaromatic Hydrocarbons Suitable for Use as Hole Transport Materials

We present an informatics-guided design of polyaromatic hydrocarbon (PAH) molecules that are selected to satisfy necessary conditions to be usable as hole transport materials in optoelectronic applications: HOMO energy level, electron transport blocking ability, and hole transport ability. We generated a large number of PAHs that were complementary to the existing COMPAS database. The reorganization energies were machine-learned (ML) as a function of molecular composition and structure with an accuracy on the order of 0.01 eV. HOMO and LUMO energy levels, and the HOMO-LUMO gap were machine-learned as corrections to values computed with xTB with an accuracy of on the order of 0.05 eV. Target values were computed by DFT for a small subset of data, and new DFT calculations were added in cycles, selected by the criteria of low reorganization energy and sufficient HOMO-LUMO gap values predicted by the ML model. For most promising molecules, hole transport properties were estimated using Marcus theory. Molecules identified with the help of the ML-guided selection and DFT calculations that can be expected to be good hole transport materials specifically for perovskite solar cells (as an example application) were further used as a basis for ML-inspired manual design of additional structures with improved properties. In this way, PAH molecules possessing simultaneously reorganization energies as low as 0.02 eV, HOMO-LUMO gaps of higher than 2.7 eV, and HOMO levels near -5.4 eV (valence band maximum of methylammonium lead iodide) were identified.