An Efficient Workflow for Generation of Conformational Ensembles of Density Functional Theory Quality: Dimers of Polycyclic (Hetero-)Aromatics

Several composite density functional theory (DFT) methods, namely HF-3c, B97- 3c, PBEh-3c, r2SCAN-3c, and ωB97X-3c, were tested for accuracy and efficiency in computing binding energies of seven low-lying conformers of the pyrene homodimer, both in the gas phase and in toluene solution. The most promising method was B97-3c, with a Mean Absolute Deviation (MAD) for binding energies of 0.5 kJ/mol, relative to ωB97X-V/def2-TZVP results. Thus, B97-3c was used in a multi-tiered approach for generating conformational ensembles for a series of homodimers. The workflow involves six steps: (i) generate an initial ensemble, using the conformer-rotamer ensemble sam- pling tool (CREST), and its underlying GFN2-xTB method; (ii) reoptimize each mem- ber of the ensemble using B97-3c; (iii) discard duplicates and high-energy conformers; (iv) reoptimize the remaining conformers using ωB97X-D4/def2-SVP; (v) if needed, discard any high energy or duplicate conformers; (vi) compute vibrational frequencies using ωB97X-D4/def2-SVP and final single point energies using ωB97X-V/def2- QZVPP. The six-step workflow allows the generation of large DFT-quality ensembles efficiently, as demonstrated on the known pyrene dimer ensemble, and then applied to the homodimers of eight small polycyclic (hetero-)aromatic molecules related to asphaltenes: anthracene, phenanthrene, fluorenone, dibenzofuran, dibenzothiophene, dibenzothiophene oxide, N -methylcarbazole, and benzo[h]quinoline. The refined ensembles enabled an analysis of trends in dimerization structures and energies for these monomers, revealing a strong dependence for binding energy upon the magnitude of dipole cancellation.