Abstract
Building on the success of Quantum Monte Carlo techniques such as diffusion Monte Carlo, alternative stochastic approaches to solve electronic structure problems have emerged over the past decade. The full configuration interaction quantum Monte Carlo (FCIQMC) method allows one to systematically approach the exact solution of such problems, for cases where very high accuracy is desired. The introduction of FCIQMC has subsequently led to the development of coupled cluster Monte Carlo (CCMC) and density matrix quantum Monte Carlo (DMQMC), allowing stochastic sampling of the coupled cluster wave function and the exact thermal density matrix, respectively. In this Article, we describe the HANDE-QMC code, an open-source implementation of FCIQMC, CCMC and DMQMC, including initiator and semistochastic adaptations. We describe our code and demonstrate its use on three example systems; a molecule (nitric oxide), a model solid (the uniform electron gas), and a real solid (diamond). An illustrative tutorial is also included.
Introduction
In quantum chemistry, the "Holy Grail" is Full Configuration Interaction (FCI)—the exact solution to the electronic structure problem within a given basis set. However, FCI suffers from the "curse of dimensionality": the number of possible configurations (determinants) grows exponentially with the number of electrons, making it impossible to solve for anything beyond tiny molecules.
HANDE-QMC is a powerful, open-source software package designed to solve the electronic Schrödinger equation using stochastic (Monte Carlo) methods. By discretizing wavefunctions into "walkers," it enables researchers to achieve near-exact "Full CI" quality results for molecules and solids that are far too large for traditional deterministic methods. This framework bridges the gap between high-accuracy quantum chemistry and scalable computational physics.
Methods
The HANDE-QMC package implements three primary stochastic algorithms that share a common "spawn-death-annihilate" core:
- Full Configuration Interaction Quantum Monte Carlo (FCIQMC): Represents the wavefunction as a collection of discrete "walkers" on Slater determinants. These walkers evolve in imaginary time, stochastically "spawning" onto connected determinants or "dying" based on Hamiltonian matrix elements.
- Coupled Cluster Monte Carlo (CCMC): Stochastically samples the Coupled Cluster expansion, allowing for high-order treatments (like CCSDTQ) without the prohibitive memory costs of deterministic CC.
- Density Matrix Quantum Monte Carlo (DMQMC): Samples the N-particle density matrix directly, enabling the calculation of exact thermal properties at finite temperatures.
Key innovations include the Initiator Approximation (i-FCIQMC), which suppresses the sign problem by restricting spawning to previously unoccupied determinants, and Semi-stochastic Propagation, which combines deterministic treatment of critical determinants with stochastic treatment of the rest to reduce noise.
Results and Benchmarks
The authors validate HANDE-QMC across three distinct regimes: molecules (Nitric Oxide), model solids (Uniform Electron Gas), and real solids (Periodic Diamond). The code interfaces with standard packages like Psi4, PySCF, and Q-Chem to generate the initial integrals.
In the Nitric Oxide (NO) molecule test (cc-pVDZ basis), i-FCIQMC achieved an accuracy of $\sim 0.1$ m$E_h$ using only 8 million walkers—a tiny fraction of the $10^{12}$ total determinants in the Hilbert space. As shown in the figure above, the stochastic CCSDTQ-MC implementation matched i-FCIQMC results almost perfectly, demonstrating that stochastic sampling is a viable path for very high-order Coupled Cluster theory across a range of bond lengths ($R$).
Key Insights
- Stochastic Sparsity: Even in a space of a trillion determinants, the actual wavefunction is sparse. Walkers naturally find and occupy only the most important regions of the Hilbert space.
- Unified Framework: By recognizing that FCIQMC, CCMC, and DMQMC all rely on similar population dynamics, the authors created a modular code where improvements in one method immediately benefit the others.
- Statistical Rigor: The project includes
pyhande, a Python library for "reblocking analysis," which automates the rigorous statistical analysis required for autocorrelated Monte Carlo data.
Conclusion
While HANDE-QMC is highly capable, future work aims to improve interoperability via standardized data formats (like FCIDUMP) and automate simulation parameter selection (like time-step and walker population) for non-experts. The HANDE-QMC project stands as a robust, transparent tool for the next generation of stochastic quantum chemistry research, utilizing modern HPC environments efficiently.
Summary updated