1 December 2025 in Nature Communications
Machine learning (ML) plays a pivotal role in extending the reach of quantum chemistry methods for simulating both molecules and materials. However, leveraging ML to overcome the limitations of human-designed density functional approximations (DFAs), the primary workhorse for quantum simulations, remains a major challenge due to their severely limited transferability to unseen chemical systems. Here, we demonstrate how transferability is achieved using real-space ML, where energies are learned point by point in space through energy densities. Central to our real-space learning strategy is the derivation and implementation of correlation energy densities from regularized perturbation theory. This enables two key advances toward constructing highly transferable DFAs, grounded in the Møller-Plesset adiabatic connection framework, for correlation energies defined with respect to the Hartree-Fock reference. First, we introduce the Local Energy Loss, whose data efficiency (expanding each system’s single energy into thousands of data points) dramatically enhances transferability when combined with a physically informed ML model. Second, we formulate a real-space, machine-learned, and regularized extension of Spin-Component-Scaled second-order Møller-Plesset perturbation theory, yielding transferable DFAs that effectively mitigate the self-interaction errors common to traditional DFAs.
24 January 2022 in The Journal of Chemical Physics
A new non-decomposable approximation of the non-additive kinetic energy potential (NAKEP) is constructed starting from the same exact property in the limit (ρA->0 and ∫ρB=2) as introduced in [Lastra et. al, J. Phys. Chem. 129, (2008) 074107]. In order to cover the complete function space for exponentially decaying densities, the kernel of a differential operator Dγ[ρ] is introduced and analyzed in dependence of γ. The conclusive choice of γ=1 assures that the solution functions span the complete space of molecular electron densities. As a result, the new approximant preserves the desired feature of the older approximation, which is the reciprocal singularity if the electron density decays exponentially, and eliminates artificial shallow wells (holes), which were responsible for an artificial ``charge leak'. Numerical considerations using the standard validation procedure introduced in [Wesolowski and Weber, Chem. Phys. Letters 248 (1996)71], demonstrate the numerical performance of the developed approximation, which increases the range of applicability of semilocal functionals.
2 May 2023 in The Journal of Chemical Physics
In subsystem density functional theory (DFT), the bottom-up strategy to approximate the multivariable functional of the non-additive kinetic energy (NAKE) makes it possible to impose exact properties on the corresponding NAKE potential (NAKEP). Such a construction might lead to a non-symmetric and non-homogeneous functional, which excludes the use of such approximations for the evaluation of the total energy. We propose a general formalism to construct a symmetric version based on a perturbation theory approach of the energy expression for the asymmetric part. This strategy is then applied to construct a symmetrized NAKE corresponding to the NAKEP developed recently [Polak et al., J. Chem. Phys. 156, 044103 (2022)], making it possible to evaluate consistently the energy. These functionals were used to evaluate the interaction energy in several model intermolecular complexes using the formal framework of subsystem DFT. The new symmetrized energy expression shows a superior qualitative performance over common decomposable models.
21 May 2025 in Journal of Chemical Theory and Computation
17 November 2026 in ChemRxiv
Second-order Møller–Plesset (MP2) correlation energy densities can serve as ingredients for machine learning (ML) of density functionals that improve upon MP2-based double hybrids. However, limitations of existing implementations have restricted the evaluation of these energy densities and of ML-based functionals employing them to small systems. Here, we develop a GPU-accelerated implementation for evaluating MP2 correlation energy densities and their regularized and spin-resolved variants by combining a density fitting scheme that preserves high pointwise accuracy with memory-efficient tensor contractions. We demonstrate the use of the resulting energy densities in ML-based density functional approximations (DFAs) through the previously developed MLS2 model [Nat. Commun. 16, 11306 (2025)] and limMLS2, a variant introduced here that enforces a high density limit constraint to MLS2. The GPU-accelerated evaluation of MP2 energy densities enables routine application of these ML-based DFAs to systems described by more than 5000 atomic orbitals, with predictions remaining robust to the density fitting approximations and grid sizes tested. Overall, by making MP2 correlation energy densities accessible for larger systems, our work broadens their use in the development and application of highly accurate ML-based DFAs.