Importance-Reweighted Fock-Space Variational Monte Carlo

Abstract:Fock-space variational Monte Carlo (FS-VMC) evaluates variational quantities by stochastic sampling over discrete many-body configurations. For ab initio electronic Hamiltonians, Born distributions can differ markedly between systems, and a Markov chain that mixes well need not yield low-variance estimators. We introduce importance-reweighted FS-VMC (IR-FS-VMC), which leaves the variational objective unchanged while redesigning its Monte Carlo evaluation. An auxiliary distribution and Hamiltonian-guided proposal define the Markov chain, while an evaluation Markov kernel defines an analytically tractable evaluation distribution. Self-normalized importance reweighting then recovers the Born-distribution expectations entering the energy, gradient, and stochastic reconfiguration (SR) matrix. Controlled comparisons on equilibrium H2O and Fe2S2 separate Markov-chain acceptance from local-energy fluctuations and optimization behavior. Using one production protocol, calculations for H2O dissociation, 36-site hydrogen lattices, and Fe2S2 and Fe4S4 active spaces yield accurate variational energies across different Born distributions and electronic-correlation regimes. Together, these results show that the production protocol can support robust FS-VMC optimization across the electronic-structure regimes studied here.

Chemical Physics (physics.chem-ph); Strongly Correlated Electrons (cond-mat.str-el); Quantum Physics (quant-ph)

arXiv:2608.22925 [physics.chem-ph]

(or arXiv:2608.22925v1 [physics.chem-ph] for this version)

https://doi.org/10.48550/arXiv.2608.22925 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.