Abstract
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 (DFT) and second-order Møller–Plesset perturbation theory (MP2). While these are efficient and realistic, they fall short of the accuracy of the “gold standard” coupled-cluster method, especially with respect to reaction and isomerization barriers. We report a major step forward in applying a Δ-machine learning method to the challenging case of acetylacetone, whose MP2 barrier height for H-atom transfer is low by roughly 1.1 kcal/mol relative to the benchmark CCSD(T) barrier of 3.2 kcal/mol. From a database of 2151 local CCSD(T) energies and training with as few as 430 energies, we obtain a new PES with a barrier of 3.5 kcal/mol in agreement with the LCCSD(T) barrier of 3.5 kcal/mol and close to the benchmark value. Tunneling splittings due to H-atom transfer are calculated using this new PES, providing improved estimates over previous ones obtained using an MP2-based PES.
Highlights
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 (DFT) and second-order Møller−Plesset perturbation theory (MP2)
There has been dramatic progress in using regression methods from machine learning (ML) to develop potential energy surfaces (PESs) for systems with more than five atoms, based on fitting thousands of CCSD(T) energies.[1−4] the CCSD(T) method, because it scales as N7, where N is the system size, is too computationally demanding for PES fits of systems with more than 10 heavy atoms. (This number of atoms is rightly not considered a “large molecule” by many readers; it is used here as a computational boundary for the CCSD(T) method.) One 10atom PES using the method we are aware of is the formic acid dimer (HCOOH)[2,5] which contains 8 heavy atoms
MP2 and DFT accuracy compared to benchmark CCSD(T)
Summary
Information (SI), and because the distribution is permutationally symmetric, we have shown only the upper half by taking. We applied an approximate 1d approach to obtain the tunneling splittings This method has been described in detail previously.[32] It was used in our work on AcAc based on the MP2 PES (VLL).[13] Briefly, a 1d potential, denoted V(Qim), which is the minimum-energy path as a function of the imaginary-frequency mode (Qim) of the H-transfer saddle point, was obtained by optimizing all of the other coordinates at fixed Qim values using the VLL→CC PES except for the methyl rotors, which cannot be described using rectilinear normal coordinates.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
