Quantum Machine Learning Approach for Studying Atmospheric Cluster Formation

Quantum chemical (QC) calculations can yield direct insight into an atmospheric cluster formation mechanism and rates. However, such are extremely computationally demanding as more than millions of configurations might exist need to be computed. We present efficient approach produce high quality QC data sets for applications in studies how train accurate quantum machine learning model on the generated data. Using two-component sulfuric acid─water system a proof concept, we demonstrate that kernel ridge regression with Δ-learning trained accurately predict binding energies equilibrium mean absolute errors below 0.5 kcal mol–1. Additionally, enlarge training set nonequilibrium show possibility predicting new structures clusters several molecules larger those set. Applying leads drastic reduction number relevant explicitly evaluated by methods. The presented is directly transferable arbitrary composition will lead faster exploration configurational space systems.