Mobilis: an open-source toolkit for charge transport in organic semiconductors, from crystal structures to temperaturedependent mobilities with dynamic disorder and active learning

Computing a charge mobility for an organic semiconductor involves a chain of steps: monomer site energies, intermolecular couplings, the thermal fluctuation of those couplings, and a transport model whose validity depends on the ratio of coupling to reorganization energy. These steps are usually carried out with separate scripts and different quantum-chemistry codes, which makes results hard to reproduce and models hard to compare. Mobilis is an open-source Python package (GPL-3.0) with a graphical interface and a command-line interface that runs the whole chain from a CIF file or an MD morphology. It prepares and reads Gaussian and ADF calculations (and xtb from the command line), and contains its own extended-Hückel and ZINDO couplings, a polarizable site-energy model, Marcus–Jortner hopping with kinetic Monte Carlo, and transient localization theory. For dynamic disorder it builds coupling surfaces over the six rigidbody intermolecular coordinates and samples them by full-crystal rigid-body molecular dynamics. A Gaussian-process active-learning loop chooses the DFT dimer calculations needed for these surfaces and stops when the posterior uncertainty of the coupling fluctuation σV falls below a target. For pentacene, 24–40 DFT dimers per pair reproduce σV within 4%, against 273 points for a quadratic fit. Chiral molecules in centrosymmetric crystals, such as mer-Alq3 and Ir(III) emitters, are treated by mirror-image fragments. The code is tested by 34 automated test scripts, including an exact-solution check of the kinetic Monte Carlo, and is archived on Zenodo (doi:10.5281/zenodo.23593488).