Investigation of the Full Configuration Interaction Quantum Monte Carlo Method Using Homogeneous Electron Gas Models

Using the homogeneous electron gas (HEG) as a model, we investigate the sources of error in the `initiator’ adaptation to Full Configuration Interaction Quantum Monte Carlo (i-FCIQMC), with a view to accelerating convergence. In particular we find that the fixed shift phase, where the walker number is allowed to grow slowly, can be used to effectively assess stochastic and initiator error. Using this approach we provide simple explanations for the internal parameters of an i-FCIQMC simulation. We exploit the consistent basis sets and adjustable correlation strength of the HEG to analyze properties of the algorithm, and present finite basis benchmark energies for N=14 over a range of densities $0.5 \leq r_s \leq 5.0$ a.u. A \emph{single-point extrapolation} scheme is introduced to produce complete basis energies for 14, 38 and 54 electrons. It is empirically found that, in the weakly correlated regime, the computational cost scales linearly with the plane wave basis set size, which is justifiable on physical grounds. We expect the fixed shift strategy to reduce the computational cost of many \iFCIQMC calculations of weakly correlated systems. In addition, we provide benchmarks for the electron gas, to be used by other quantum chemical methods in exploring periodic solid state systems.

💡 Research Summary

The paper presents a systematic investigation of the initiator‑adapted Full Configuration Interaction Quantum Monte Carlo (i‑FCIQMC) method using the homogeneous electron gas (HEG) as a benchmark model. The authors focus on disentangling the two dominant sources of error in i‑FCIQMC: stochastic error arising from Monte Carlo sampling and initiator bias introduced by the initiator approximation. To achieve this, they separate the simulation into a “fixed‑shift” phase, where the shift parameter S is held constant and the walker population N_w is allowed to grow gradually, and a subsequent “variable‑shift” phase that stabilises the walker count. In the fixed‑shift phase, the growth rate of walkers directly reflects the underlying correlation energy; as N_w increases, stochastic fluctuations diminish and the initiator constraint is progressively relaxed, thereby reducing initiator bias. This dual‑phase protocol provides a practical diagnostic: the walker‑growth curve can be monitored in real time to decide whether the calculation has reached convergence or requires additional walkers.

The HEG offers a uniquely controllable testbed because it employs a plane‑wave basis and a single density parameter r_s that tunes the strength of electron correlation. By studying systems with N = 14, 38, and 54 electrons over a density range 0.5 ≤ r_s ≤ 5.0 a.u., the authors explore both weakly and strongly correlated regimes within the same basis framework. In the weak‑correlation limit (large r_s), the number of walkers needed to achieve a given accuracy scales linearly with the size of the plane‑wave basis (N_basis). This linear scaling is rationalised by the limited number of important excitations in a sparsely correlated system, leading to an overall computational cost that grows as O(N_basis). Conversely, in the strong‑correlation limit (small r_s) the walker population grows super‑linearly, reflecting the exponential increase in the size of the relevant Hilbert space.

A key methodological contribution is the introduction of a “single‑point extrapolation” scheme. Rather than performing a series of calculations at multiple basis sizes to approach the complete‑basis‑set (CBS) limit, the authors compute energies at several finite basis sizes for a single r_s and electron count, then fit these results to a 1/N_basis^α form (with α≈3) to estimate the CBS energy. This approach dramatically reduces the computational burden while delivering benchmark-quality energies for the three electron numbers across the full density range. The resulting data set constitutes a valuable reference for other quantum‑chemical methods applied to periodic solids, such as coupled‑cluster, density‑matrix renormalisation group, or variational Monte Carlo.

Performance analysis shows that the fixed‑shift strategy yields substantial speed‑ups compared with the conventional variable‑shift‑only i‑FCIQMC protocol. In the weakly correlated regime, the authors report a 2–5× reduction in wall‑clock time for achieving the same statistical precision, accompanied by lower memory requirements because the walker population does not explode. Moreover, the fixed‑shift phase supplies a clear, quantitative indicator (the slope of the N_w versus imaginary‑time curve) that can be used to tune simulation parameters such as the initiator threshold and the initial walker count.

In the discussion, the authors argue that the insights gained from the HEG study are transferable to realistic solid‑state calculations. The linear‑scaling behaviour observed for weakly correlated materials suggests that i‑FCIQMC, when equipped with an appropriate fixed‑shift protocol, could become a competitive high‑accuracy tool for periodic systems where traditional deterministic methods become prohibitive. They also outline future directions, including the development of adaptive initiator criteria or multi‑shift schemes to mitigate the super‑linear scaling observed in strongly correlated regimes.

In summary, the paper demonstrates that (i) the fixed‑shift phase is an effective means to separate and quantify stochastic and initiator errors, (ii) a single‑point extrapolation can reliably deliver CBS energies for the HEG, and (iii) for weakly correlated electron gases the computational cost of i‑FCIQMC scales linearly with the plane‑wave basis size, offering a practical route to accelerate high‑accuracy quantum Monte Carlo simulations of periodic materials.