Property-specific ACPs
ACPs can be designed for use with a specific method/basis set combination to target very accurate results for a specific property. ACPs of this type can be used for those interested in the properties themselves, or by those developing machine learning models to predict those properties and require an accurate, high-throughput method to generate fitting data.
Water Cluster Binding Energies
Error in the predicted binding energy of water clusters, i.e. (H2O)n → n H2O, obtained with and without ACPs. The ACPs uniformly predict water binding energies to better than 0.05 kcal/mol/water molecule.
Holmes, J. D.; Otero-de-la-Roza, A.; DiLabio, G. A. J. Chem. Theory Comput. 2017, 13, 4205–4215.
Bond Dissociation Energies
ACPs fit to (RO)CBS-QB3 bond dissociation energies to improve results obtained with ωB97XD/pcseg-1.
Prasad, V. K.; Otero-de-la-Roza, A.; DiLabio, G. A. Mach. Learn.: Sci. Technol. 2024, 5, 015035
https://iopscience.iop.org/article/10.1088/2632-2153/ad27e1/pdf
Data from: https://www.nature.com/articles/s41597-021-01088-2
Reaction Energies and Barrier Heights
ACPs fit to DLPNO-CCSD(T)/CBS quality barrier heights and reaction energies to improve results obtained with ωB97XD/pcseg-1.
Prasad, V. K.; Otero-de-la-Roza, A.; DiLabio, G. A. Mach. Learn.: Sci. Technol. 2024, 5, 015035
https://iopscience.iop.org/article/10.1088/2632-2153/ad27e1/pdf
Data from: https://pubs.acs.org/doi/abs/10.1021/acs.jctc.1c00694
Potential Energy Surfaces
ACPs can be used in a Δ-DFT type approach, to elevate the energies predicted by an inexpensive DFT method to generate energies close to complete basis set CCSD(T). The data below show the results for the HFCO molecule: The ACP method gives a mean absolute error in the PES of 5.8 cm-1 and required only 100 data for fitting.
A general protocol for the development of ACPs to predict the PES of molecules of arbitrary size is given here: https://chemrxiv.org/engage/chemrxiv/article-details/67f1a35981d2151a022fb8a6
Ashani, M. N.; Huang, Q; Brown, A.; Aerts, A.; Otero-de-la-Roza, A.; DiLabio, G. A. J. Phys. Chem. A 2023, 127, 8015–8024.