The data-driven computational methodology developed by our group combines quantum chemistry with machine learning (ML) to surpass size limitations of accurate but computationally demanding methods such as coupled-cluster (CC). Solving the CC equations is a computational demanding process that exhibits poor scaling with system size. Our approach is based on the prediction of the converged coupled-cluster singles-and-doubles (CCSD) amplitudes and thus, the coupled-cluster wave function by utilizing ML and electronic structure properties inherent to the MP2 level. Features are collected from quantum chemical data, such as orbital energies, one-electron Hamiltonian, Coulomb, and exchange terms. The data-driven CCSD (DDCCSD) is not an alchemical method because the actual iterative coupled-cluster equations are solved. Our results show that remarkable speedups in solving the CCSD equations are achieved, especially when the correct physics are encoded and used for training ML models. Recently, we developed an alternative molecular representation that explores the short-range behavior of electron correlation and includes the perturbative triples (T). The overall model was found to be highly transferable and size-extensive, necessitating very few training instances to approach chemical accuracy of a broad spectrum of organic molecules. The efficiency and transferability of the novel representation was demonstrated on a series of linear hydrocarbons, the potential energy surface of the water dimer, and on the GDB-9 database. For the GDB-9 database, we found that data from only 140 randomly selected molecules are adequate to achieve chemical accuracy (1 kcal/mol) for more than 133,000 organic molecules. Recent progress with respect to the effective selection of amplitude subspaces for increased transferability on larger molecules systems will be discussed. Finally, we have extended this approach to other quantum chemical methods, such as the complete active space second-order perturbation theory (CASPT2).

Recent developments of the data-driven quantum chemical methods

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.