HZHeng Zhao

Heng Zhao

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Heng Zhao has a total of 21 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Accurate Density Functional Theory Forces for Charged Noncovalent Complexes

    by Vinicius Fontenelle, Heng Zhao, Stefan Vuckovic

    Force accuracy is essential for training machine-learned force fields based on density functional theory (DFT) data and for reliably modeling noncovalent interactions (NCIs). Despite this importance and the sensitivity of forces to the underlying density functional approximation, a gap remains in benchmarking DFT force accuracy for NCIs. Here, we assess the force performance of dispersion-enhanced DFT against CCSD(T) references for representative charged NCI dimers. While conventional dispersion-enhanced density functional approximations (DFAs) exhibit substantial force errors for charged systems, the recently proposed (r2SCAN+MBD)@HF, which combines r2SCAN, many-body dispersion, and Hartree-Fock densities in a synergistic manner, systematically reduces these errors [Zhao, H.; Lőrincz, B. D.; Henkes, T.; Berta, D.; Nagy, P. R.; Tkatchenko, A.; Vuckovic, S. Sci. Adv. 2026, 12, eadz8521]. The improvements in force accuracy of (r2SCAN+MBD)@HF over other DFT approaches are most pronounced for strongly charged dimers, extend to vibrational frequencies, and propagate to improved vibrational spectra from machine-learned force field-based molecular dynamics simulations. This work supports (r2SCAN+MBD)@HF as a promising and high-quality reference level for training machine-learned force fields for the studied charged NCI cases and motivates its broader application to more complex ionic environments.

  • Open Access

    Accurate density functional theory for noncovalent interactions in charged systems

    by Heng Zhao, Balázs D. Lőrincz, Tobias Henkes, Dénes Berta, Péter R. Nagy, Alexandre Tkatchenko, Stefan Vuckovic

    Accurately modeling noncovalent interactions (NCIs) involving charged systems remains an outstanding challenge in density functional theory (DFT), with implications across natural and life sciences, engineering, e.g., in biochemistry, catalysis, and materials science. For these interactions, the interplay between electrostatics, polarization, and dispersion leads to systematic errors of up to tens of kilocalories per mole in standard dispersion–enhanced DFT methods. We solve this problem by introducing (r 2 SCAN+MBD)@HF, a DFT method without empirically fitted parameters that combines the r 2 SCAN functional and many-body dispersion, both evaluated on Hartree-Fock densities. We show that the unique synergy of these three components enables balanced treatment of short- and long-range correlation, which is crucial for accurate description of NCIs involving charged systems. Evaluations on standard benchmarks show that (r 2 SCAN+MBD)@HF significantly improves accuracy for NCIs involving charged systems while maintaining robust performance for neutral systems. Further tests on the Metal Ion Protein Clusters dataset introduced here demonstrate its improved accuracy for metal-protein interactions, including cases where the metal ion is surrounded by negatively charged ligands for which standard DFT can completely break down. Given the ubiquity of such interactions, (r 2 SCAN+MBD)@HF is broadly applicable from biochemistry and materials science, including for generating high-quality data to train machine-learning force fields.


  • Open Access

    Accurate Density Functional Theory for Non-Covalent Interactions in Charged Systems

    by Heng Zhao, Balázs Lőrincz, Tobias Henkes, Dénes Berta, Péter Nagy, Alexandre Tkatchenko, Stefan Vuckovic

    Accurately modeling non-covalent interactions (NCIs) involving charged systems remains an outstanding challenge in Density Functional Theory (DFT), with implications across natural and life sciences, engineering, e.g., in biochemistry, catalysis, and materials science. For these interactions, the interplay between electrostatics, polarization, and dispersion leads to systematic errors of up to tens of kcal/mol in standard dispersion-enhanced DFT methods. We solve this problem by introducing (r2SCAN+MBD)@HF, a DFT method without empirically fitted parameters that combines the r2SCAN functional and many-body dispersion, both evaluated on Hartree-Fock densities. We show that the unique synergy of these three components enables balanced treatment of short- and long-range correlation, which is crucial for accurate description of NCIs involving charged systems. Evaluations on standard benchmarks and a new Metal Ion Protein Clusters dataset introduced here show that (r2SCAN+MBD)@HF significantly improves accuracy for NCIs involving charged systems while maintaining robust performance for neutral systems. Given the ubiquity of such interactions, (r2SCAN+MBD)@HF is broadly applicable from biochemistry and materials science, including for generating high-quality data to train machine-learning force fields.

  • Open Access

    Real-space machine learning of correlation density functionals

    by Elias Polak, Heng Zhao, Stefan Vuckovic

    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.


  • The total synthesis of (−)-spiroaspertrione A: A divinylcyclopropane rearrangement approach

    by Wenbo Huang, Lu Pan, Heng Zhao, Fabian Schneider, Tanja Gaich

    The rise of multidrug-resistant pathogens poses a major threat to global health, with methicillin-resistant Staphylococcus aureus (MRSA) among the most challenging. One promising approach to overcoming resistance is using small molecules that resensitize MRSA to existing drugs. Here, we report the enantioselective total synthesis of one such promising candidate, (−)-spiroaspertrione A, a complex meroterpenoid of the andiconin family. This natural product has long eluded synthesis because of its densely functionalized polycyclic backbone. Our route features a stereoselective Diels-Alder cycloaddition, followed by a key divinylcyclopropane rearrangement forming the spirobicyclo[3.2.2]nonane core, which proved to be reversible and was further investigated by density functional theory calculations. Strategic late-stage functionalization of the compact cage architecture enabled access to the natural product and provided evidence for a plausible biosynthetic relationship with (−)-aspermerodione.


  • Open Access

    Real-space machine learning of correlation density functionals

    by Elias Polak, Heng Zhao, Stefan Vuckovic

    Machine learning (ML) plays a pivotal role in extending the reach of quantum chemistry methods for both molecules and materials. However, leveraging ML to improve upon human-designed density functional approximations (DFAs), the primary workhorse for quantum simulations, remains challenging due to severely limited transferability to unseen chemical systems. Here we address this challenge through real-space ML, where energies are learned point by point via correlation energy densities per particle obtained from regularized perturbation theory. We pursue two crucial strategies that enable the construction of 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 datapoints) dramatically enhances transferability when combined with a physically informed ML construction. Second, we construct a real-space, machine-learned, and regularized extension of Spin-Component-Scaled second-order Møller–Plesset perturbation theory, opening new avenues for developing transferable DFAs particularly suited for overcoming self-interaction errors common to traditional DFAs.

  • Møller–Plesset Adiabatic Connection Theory for Diverse Noncovalent Interactions

    by Etienne Palos, Heng Zhao, Kimberly J. Daas, Eduardo Fabiano, Francesco Paesani, Stefan Vuckovic

    Møller-Plesset adiabatic connection (MPAC) theory provides a powerful framework for constructing approximations to wave function-based correlation energy, enabling modeling of noncovalent interactions (NCIs) with near-CCSD(T) accuracy. We show that approximate MPAC functionals consistently outperform MP2 and dispersion-corrected DFT (DFT+DISP) across diverse systems, including charged and charge-transfer complexes. MPAC functionals operate holistically at the electronic level, require no heuristic dispersion corrections, and achieve near-chemical accuracy even for abnormal NCIs, cases where DFT+DISP errors exceed those of DFT. To further improve MPAC for abnormal cases without compromising overall performance, we introduce MPAC25, a simple two-parameter functional treating neutral and charged NCIs equally, as demonstrated on DES15K benchmarks. Overall, MPAC functionals effectively describe a wide range of NCIs, including those beyond the reach of other methods, representing a significant step toward predictive simulations of molecular interactions in complex environments and motivating further MPAC developments.


  • Open Access

    Møller-Plesset Adiabatic Connection Theory for Diverse Noncovalent Interactions

    by Etienne Palos, Heng Zhao, Kimberly J Daas, Eduardo Fabiano, Francesco Paesani, Stefan Vuckovic

    Møller–Plesset adiabatic connection (MPAC) theory provides a powerful framework for constructing approximations to wavefunction-based correlation energy, enabling modeling of non- covalent interactions (NCIs) with near-CCSD(T) accuracy. We show that approximate MPAC functionals consistently outperform MP2 and dispersion-corrected DFT (DFT+DISP) across diverse systems, including charged and charge-transfer complexes. MPAC functionals operate holistically at the electronic level, require no heuristic dispersion corrections, and achieve near-chemical accuracy even for abnormal NCIs, cases where DFT+DISP errors exceed those of DFT. To further improve MPAC for abnormal cases without compromising overall performance, we introduce MPAC25, a simple two-parameter functional treating neutral and charged NCIs equally, as demonstrated on DES15K benchmarks. Overall, MPAC functionals effectively describe a wide range of NCIs, including those beyond the reach of other methods, representing a significant step toward predictive simulations of molecular interactions in complex environments and motivating further MPAC developments.

  • Open Access

    Exact Møller-Plesset Adiabatic Connection Correlation Energy Densities

    by Kimberly Jennifer Daas, Heng Zhao, Elias Polak, Stefan Vuckovic

    The Møller–Plesset adiabatic connection (MPAC) provides a powerful tool for developing density functional theory (DFT)-like approximations that map Hartree–Fock densities to the wavefunction-based correlation energy, thereby leveraging both wavefunction and DFT concepts for electronic structure approximations. A key object in this context is the correlation energy density, which represents the local (pointwise) contribution to the total correlation energy. While well-studied in DFT, it remains largely unexplored in the wavefunction framework. Here, we introduce a rigorous formulation of the wavefunction-based correlation energy density within MPAC, implement it via full configuration interaction calculations, and analyze its behavior and physically meaningful contributions for representative small (di)atomic systems. We define this quantity by employing a general gauge strategy, from which the conventional DFT correlation energy density gauge also arises. We then discuss the resulting commonalities and differences between correlation energy densities in the DFT and wavefunction frameworks and derive the small-interaction (MP2) limit of the latter in terms of Hartree–Fock orbitals. Finally, we show how these newly introduced energy densities can serve as new approximation targets in both machine-learning-assisted and traditional electronic structure methods for mapping HF-density-based features to correlation energy within the wavefunction framework.

  • Open Access

    Deep Mind 21 functional does not extrapolate to transition metal chemistry

    by Heng Zhao, Tim Gould, Stefan Vuckovic

    The development of density functional approximations stands at a crossroads: while machine-learned functionals show potential to surpass their human-designed counterparts, their extrapolation to unseen chemistry lags behind. Here we assess how well the recent Deep Mind 21 (DM21) machine-learned functional [Science, 2021, 374, 1385–1389], trained on main-group chemistry, extrapolates to transition metal chemistry (TMC). We show that DM21 demonstrates comparable or occasionally superior accuracy to B3LYP for TMC, but consistently struggles with achieving self-consistent field convergence for TMC molecules. We also compare main-group and TMC machine-learning DM21 features to shed light on DM21's challenges in TMC. We finally propose strategies to overcome limitations in the extrapolative capabilities of machine-learned functionals in TMC.