arXiv:2609.40112
Smart Machine Learning for Solving the Full Electron-Nuclear Schrödinger Equation with a Quantum-Chemically Inspired Wavefunction Ansatz
Yaolong Zhang, Bin Jiang, Hua Guo
Abstract
Directly solving the full electron-nuclear Schrödinger equation remains one of the grand challenges in quantum mechanics. Here, we propose a quantum-chemically motivated orbital-based neural network wavefunction ansatz with explicit many-body electron-electron correlation terms, with parameters depending on the entire electron-nuclear configuration, enabling an accurate description of strongly correlated and/or multi-reference electronic structures by a single Slater-determinant across a broad nuclear configuration space. To train this wavefunction, we further develop an efficient hybrid optimization strategy that combines the variational Monte Carlo technique with local-energy constraints, mitigating the intrinsic sampling bias of the former and considerably reducing the statistical error. These advances are embedded in a unified neural network framework, Schrödinger, enabling the determination of globally accurate potential energy surfaces with a single training for systems ranging from strongly correlated diatomic molecules, triatomic reactions containing conical intersections, to polyatomic multi-reference molecules. This work offers a practical machine-learning tool for solving the electron-nuclear Schrödinger equation and capturing full quantum effects beyond the Born-Oppenheimer approximation.
AI Summary
The paper addresses neural-network representations and optimization methods for molecular wavefunctions across nuclear configurations, with the stated aim of treating electron–nuclear quantum mechanics beyond the Born–Oppenheimer approximation. It proposes an orbital-based wavefunction ansatz with explicit many-body electron–electron correlation terms and parameters that depend on the complete electron–nuclear configuration. This dependence is intended to let a single Slater-determinant representation describe strongly correlated or multireference electronic structures over a broad range of nuclear geometries. The proposed training method combines variational Monte Carlo with local-energy constraints. The abstract reports that this hybrid optimization reduces statistical error and mitigates sampling bias in variational Monte Carlo. The ansatz and optimization are integrated into a neural-network framework called Schrödinger, which is reported to obtain globally accurate potential-energy surfaces from a single training run. Applications span strongly correlated diatomic molecules, triatomic reactions involving conical intersections, and polyatomic molecules with multireference character. Only the title and abstract are available. They provide no numerical accuracy, computational cost, system sizes, benchmark comparisons, or details of how the local-energy constraints are enforced. No explicit limitations are stated. In particular, the abstract does not establish how the reported potential-energy-surface calculations demonstrate a solution of the coupled electron–nuclear Schrödinger equation, or quantify the treatment of nonadiabatic effects beyond Born–Oppenheimer theory.
Written by gpt-6.1-sol from the abstract only. Automated: it can be wrong.
- Comments
- 6 fig, 2 table