This work turns the idea of amortizing the cost of solving the Schrödinger equation over many molecules into a practical approach in quantum chemistry with Orbformer, a transferable wavefunction model pretrained on 22,000 equilibrium and dissociating structures that can be fine-tuned on unseen molecules reaching an accuracy–cost ratio rivalling classical multireference methods.
Abstract
Reliable description of bond breaking remains a major challenge for quantum chemistry due to the multireference character of the electronic structure in dissociating species. Multireference methods in particular suffer from large computational cost, which under the normal paradigm has to be paid anew for each system at a full price, ignoring commonalities in electronic structure across molecules. Quantum Monte Carlo with deep neural networks uniquely offers to exploit such commonalities by pretraining transferable wavefunction models, but all such attempts were so far limited in scope. Here, we bring this paradigm to fruition with Orbformer, a transferable wavefunction model pretrained on 22,000 equilibrium and dissociating structures that can be fine-tuned on unseen molecules reaching an accuracy–cost ratio rivalling classical multireference methods. On established benchmarks as well as more challenging bond dissociations and Diels–Alder reactions, Orbformer is the only method that consistently converges to chemical accuracy (1 kcal/mol). This work turns the idea of amortizing the cost of solving the Schrödinger equation over many molecules into a practical approach in quantum chemistry.
New methods and workflows to overcome the challenges inherent to automating unrestricted coupled cluster calculations are developed and a transferable MLIP for gas-phase reactions, trained on unrestricted CCSD(T) data is developed.
Alice E. A. Allen, Rui Li, Sakib Matin et al.· Journal of Chemical Theory a...· 0 citations
Accurate modeling of bond breaking remains a central challenge for reduced density matrix functional theory (RDMFT). Although some modern functionals can yield reasonably accurate dissociation energies, they often fail to reproduce key properties of the dissociated fragments, such as a vanishing fragment population covariance (also known as the delocalization index) and the correct total spin angular momentum of each fragment (local spin). In this study, we revisit the construction of natural orbital functionals by correcting the cumulant contribution produced by the PNOF5 functional. Our method enforces known contributions of the cumulant to local spin fragments and the delocalization index at the dissociation limit. We obtain the closest cumulant consistent with these physically motivated constraints and subsequently purify the corresponding one- and two-electron reduced density matrices by imposing the standard P, Q, and GN-representability conditions. The resulting functional yields improved behavior in strongly correlated regimes. Benchmarking on the dissociation of the singlet states of N2, NO+, O2, S2, and CO shows that, in the dissociation regime, the energies computed from the updated cumulant exactly reproduce the complete active space self-consistent field energies. We further analyze the limitations of the approach and identify scenarios in which the current approach performs poorly. This study provides a pathway for systematically improving natural orbital functionals to achieve reliable bond-breaking calculations within RDMFT.
Valerii Chuiko, Paul W. Ayers, E. Matito· Journal of Chemical Physics· 0 citations
Accurate modeling of bond breaking remains a central challenge for reduced density matrix functional theory (RDMFT). Although some modern functionals can yield reasonably accurate dissociation energies, they often fail to reproduce key properties of the dissociated fragments, such as a vanishing fragment population covariance (also known as the delocalization index) and the correct total spin angular momentum of each fragment (local spin). In this work, we revisit the construction of natural orbital functionals by correcting the cumulant contribution produced by the PNOF5 functional. Our method enforces known contributions of the cumulant to local spin fragments and the delocalization index at the dissociation limit. We obtain the closest cumulant consistent with these physically motivated constraints and subsequently purify the corresponding one- and two-electron reduced density matrices by imposing the standard $P, Q, \text{and } G$ $N$-representability conditions. The resulting functional yields improved behavior in strongly correlated regimes. Benchmarking on the dissociation of the singlet states of \ce{N2}, \ce{NO+}, \ce{O2}, \ce{S2}, and \ce{CO} shows that in the dissociation regime the energies computed from the updated cumulant exactly reproduce the complete active space self-consistent field (CASSCF) energies. We further analyze the limitations of the approach and identify scenarios in which the current approach performs poorly. This work provides a pathway for systematically improving natural orbital functionals to achieve reliable bond-breaking calculations within RDMFT.
Valerii Chuiko, Paul W Ayers, E. Matito· 0 citations
This work employs a dual-coordinate approach where both solutes are explicitly present but do not interact with one another, and utilizes a harmonic ″anchor″ restraint to a central atom on each molecule to enforce spatial overlap without modifying the internal intramolecular dynamics of either solute.
Anna Katharina Picha, S. Boresch· Journal of Chemical Theory a...· 1 citation
A new small and highly accurate Gaussian basis set has been developed for the ab initio description of $^2$Li Rydberg excited states up to $n = 7$ and S, P, D, F, and G angular momentum symmetry, and the appropriate optimization protocol is presented. The obtained Rydberg excitation energies are compared with results from other highly accurate approaches. At the EOM-CCSD level of theory the new basis exhibits higher than $10^{-2}$ eV accuracy of the excitation energies and the ionization potential and provides superior results than an analogous universal Gaussian basis set, while utilizing even smaller number of basis functions. Plots of the 1-dimensional Rydberg-orbital cuts reveal a regular nodal structure along the logarithmic scale of the atomic radius reaching tens of angstrom far from the nucleus. The presented Rydberg basis set generation methodology is an important step towards routine ab initio Rydberg-state related investigations of more complex systems, such as large atoms and polyatomic molecules.
J. Šmydke, Benjamín Andreides, Simona Dubcová· 0 citations
We introduce quantum Monte Carlo (QMC)-trained multi-state machine-learned (ML) force fields for nonadiabatic excited-state dynamics, targeting photochemical processes in which the electronic character changes along the reaction path and a consistent correlated description is required. In this framework, variational Monte Carlo wave functions combine compact selected configuration-interaction expansions with a Jastrow factor that explicitly accounts for dynamical correlation, while neural networks convert the stochastic QMC data into smooth potential energy surfaces for large surface-hopping ensembles. We apply this approach to azomethane, a demanding test case involving torsional relaxation through conical-intersection regions and C--N bond dissociation on the hot ground state. Benchmark calculations support the accuracy of the QMC reference data and show robust force convergence across isomerization and dissociation geometries. The QMC-trained dynamics preserves the expected photoisomerization mechanism, strongly reduces the excessive C--N breaking obtained with complete active space self-consistent field, and predicts a small but non-negligible prompt dissociation component after internal conversion, with a timescale consistent with femtosecond-resolved mass-spectrometry experiments. These results establish QMC-ML as a practical route to nonadiabatic photochemical dynamics with accurate wave-function reference data.
Alfonso Annarelli, Emiel Slootman, C. Filippi· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
