Breaking the Coupled Cluster Barrier for Machine-Learned Potentials of Large Molecules: The Case of 15-Atom Acetylacetone

Machine-learned potential energy surfaces (PESs) for molecules with more than 10 atoms are typically forced to use lower-level electronic structure methods such as density functional theory and second-order Moller-Plesset perturbation (MP2). While these efficient realistic, they fall short of the accuracy ``gold standard'' coupled-cluster method, especially respect reaction isomerization barriers. We report a major step forward in applying $\Delta$-machine learning method challenging case acetylacetone, whose MP2 barrier height H-atom transfer is low by roughly 1.5 kcal/mol relative benchmark CCSD(T) 3.2 kcal/mol. From database 2151 local energies, training few 430 we obtain new PES 3.49 agreement LCCSD(T) one 3.54 close value. Tunneling splittings due calculated using this PES, providing improved estimates over previous ones obtained an MP2-based PES.