PyQMC

PyQMC is a Python package for real-space quantum Monte Carlo (QMC) electronic structure calculations, designed to interoperate closely with PySCF.

Full documentation is available at pyqmc.readthedocs.io.

Features

• Variational Monte Carlo (VMC) and Diffusion Monte Carlo (DMC)
• Wavefunction optimization via stochastic reconfiguration
• Ensemble wave function optimization for excited state computation
• Trial wavefunctions: Slater-Jastrow, multi-determinant (CASSCF/selected CI), geminal, and three-body Jastrow
• Observables: energy, one- and two-body density matrices, extensible to your problem.
• Periodic boundary conditions with twist averaging and supercell support
• Parallel execution via MPI (mpi4py) or Dask
• GPU and JAX backends for high-performance evaluation
• HDF5-based checkpointing for restartable workflows

Installation

pip install pyqmc

to get the latest development version,

pip install git+https://github.com/WagnerGroup/pyqmc.git

Requirements: Python >= 3.10, PySCF >= 2.8, SciPy, h5py, pandas.

Quick Start

The high-level recipes interface handles a complete optimize -> VMC -> DMC workflow:

import pyscf
import pyqmc.recipes

# 1. Run a DFT/HF calculation with PySCF and save a checkpoint
mol = pyscf.gto.M(atom="He 0. 0. 0.", basis="ccECP_cc-pVDZ", ecp="ccecp", unit="bohr")
mf = pyscf.scf.RHF(mol)
mf.chkfile = "he_dft.hdf5"
mf.kernel()

# 2. Optimize the Slater-Jastrow wavefunction
pyqmc.recipes.OPTIMIZE("he_dft.hdf5", "he_sj.hdf5", slater_kws={"optimize_orbitals": True})

# 3. Run VMC
pyqmc.recipes.VMC("he_dft.hdf5", "he_sj_vmc.hdf5", load_parameters="he_sj.hdf5", nblocks=40)

# 4. Run DMC
pyqmc.recipes.DMC("he_dft.hdf5", "he_sj_dmc.hdf5", load_parameters="he_sj.hdf5",
                  nblocks=4000, tstep=0.02)

Results are saved as HDF5 files and can be read back with pyqmc.recipes.read_mc_output.

Parallel Execution

PyQMC supports parallelism via MPI using mpi4py.futures or any other futures object such as dask or concurrent:

import mpi4py.futures
import pyqmc.recipes

if __name__ == "__main___":
    npartitions = 4
    with mpi4py.futures.MPIPoolExecutor(max_workers=npartitions) as client:
        pyqmc.recipes.OPTIMIZE("he_dft.hdf5", "he_sj.hdf5", client=client, npartitions=npartitions)

Run with: mpiexec -n 5 python -m mpi4py.futures script.py

Package Structure

Citation

If you use PyQMC in your research, please cite the relevant papers listed in the documentation.

License

MIT License. Copyright (c) 2019-2026 The PyQMC Developers.