Coupled Cluster con MoLe: Molecular Orbital Learning for Neural Wavefunctions

Density functional theory (DFT) is the most widely used method for calculating molecular properties; however, its accuracy is often insufficient for quantitative predictions. Coupled cluster (CC) theory is the most successful method for achieving accuracy beyond DFT and predicting properties that closely align with experiment. It is known as the ``gold standard'' of quantum chemistry. Unfortunately, the high computational cost of CC limits its widespread applicability. In this work, we present the Molecular Orbital Learning Model (MoLe), an equivariant machine learning model that directly predicts CC's core mathematical objects, the excitation amplitudes, from the mean-field Hartree-Fock molecular orbitals as inputs. We test various aspects of our model and demonstrate its very high data efficiency and remarkable out-of-distribution generalization to larger molecules and off-equilibrium geometries, despite being trained only on small equilibrium geometries. Finally, we also examine its ability to reduce the number of cycles required to converge CC calculations. MoLe can set the foundations for high-accuracy wavefunction-based ML architectures to accelerate molecular design and complement force-field approaches.

The best quantum chemistry methods can predict molecular behavior with high accuracy, but they are often too expensive to use routinely. Cheaper methods are faster, but they can miss details that matter for designing medicines, materials, and chemical processes. - We use molecular orbitals: a cheap description of where electrons are likely to be in a molecule. - MoLe learns from these orbitals to predict the key quantities used by coupled cluster theory, one of the most reliable but costly methods in quantum chemistry. - Because molecular orbitals already contain useful chemical information, MoLe can make accurate predictions even for larger molecules and shapes it did not see during training; this is the kind of generalizability needed for practical molecular design. In short: molecular orbitals are cheap, useful, and already available, and learning from them can make accurate molecular prediction faster and more practical!