Capturing the Effect of Exact Exchange in Thio-Michael Additions with a Machine-Learned Functional at Meta-GGA Cost
Abstract
Semilocal density-functional approximations fail to describe the formation of anionic enolates in thio-Michael additions, the bond-forming step of covalent-inhibitor warheads; range-separated hybrids, which reintroduce long-range exact exchange, are required. Across the published CCSD(T)validated reaction pathways of eight warheads, we show that Skala—a deep-learned semilocal functional containing no exact exchange—matches the range-separated hybrid reference for the activation barriers, enolate geometries and relaxed reactant structures, outperforming the global hybrid B3LYP at meta-GGA cost. The one signature it does not reproduce is the orbital one: its Koopmans ionization potential stays at the semilocal scale, because the ionization-potential theorem constrains the exchange potential, not the energy. Skala thus captures the effect of exact exchange in the total energy, from training data alone.
Related
- Molecular Implementation of the Machine-Learned Skala Exchange-Correlation Functional in CP2K through GauXC
- Native implementation of the machine-learned Skala exchange-correlation functional in CP2K: Unified one-centre reconstruction for molecular and condensed-phase calculations
- Octupole deformation in even-even Ra isotopes from covariant density functional theory with localized exchange terms in a three-dimensional lattice space
- ROSE: Deep Learning exchange-correlation functional from reduced density-matrix representations
- Machine-learned exchange-correlation functionals for molecules, solids, and reactive surfaces
BibTeX
@misc{williamson2026capturing,
title = {{Capturing the Effect of Exact Exchange in Thio-Michael Additions with a Machine-Learned Functional at Meta-GGA Cost}},
author = {Connor Williamson and Ross Amory and David M. Rogers and Katherine Inzani and Jonathan D. Hirst},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15009612/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009612/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Capturing the Effect of Exact Exchange in Thio-Michael Additions with a Machine-Learned Functional at Meta-GGA Cost
AU - Connor Williamson
AU - Ross Amory
AU - David M. Rogers
AU - Katherine Inzani
AU - Jonathan D. Hirst
PY - 2026
DA - 2026/09/29
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009612/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009612/v1
AB - Semilocal density-functional approximations fail to describe the formation of anionic enolates in thio-Michael additions, the bond-forming step of covalent-inhibitor warheads; range-separated hybrids, which reintroduce long-range exact exchange, are required. Across the published CCSD(T)validated reaction pathways of eight warheads, we show that Skala—a deep-learned semilocal functional containing no exact exchange—matches the range-separated hybrid reference for the activation barriers, enolate geometries and relaxed reactant structures, outperforming the global hybrid B3LYP at meta-GGA cost. The one signature it does not reproduce is the orbital one: its Koopmans ionization potential stays at the semilocal scale, because the ionization-potential theorem constrains the exchange potential, not the energy. Skala thus captures the effect of exact exchange in the total energy, from training data alone.
ER -