Exchange Monte Carlo for continuous-space Path Integral Monte Carlo simulation

We present a novel Exchange Monte Carlo (EMC) method designed for application in continuous-space Path Integral Monte Carlo (PIMC) simulations at finite temperature. Traditional PIMC methods for bosonic systems suffer from long autocorrelation times, particularly when measuring observables affected by particle permutations, such as the winding number. To address this issue, we introduce an exchange update scheme that facilitates replica transitions between different interaction regimes, significantly accelerating Monte Carlo dynamics-especially for global observables sensitive to permutation effects. Furthermore, we incorporate Stochastic Potential Switching (SPS) to efficiently decompose interactions, substantially enhancing computational efficiency for long-range interatomic pair potentials such as the Lennard-Jones and Aziz potentials.

💡 Research Summary

The paper introduces a novel Exchange Monte Carlo (EMC) framework specifically designed for continuous‑space Path Integral Monte Carlo (PIMC) simulations of bosonic systems at finite temperature. Traditional PIMC, even when combined with the worm algorithm, suffers from extremely long autocorrelation times for observables that depend on particle permutations, such as the winding number, because local updates struggle to generate long permutation cycles. Moreover, the computational cost of evaluating long‑range pair potentials (e.g., Lennard‑Jones, Aziz) scales as O(N²), limiting the size of feasible simulations.

To overcome these bottlenecks, the authors propose three tightly integrated components:

1. Interaction‑based Exchange Monte Carlo – Instead of the conventional temperature ladder, auxiliary replicas are created that differ in their interaction strength or in the discretization of imaginary‑time slices. Periodic exchange moves between a low‑temperature “physical” replica and a high‑interaction‑strength replica allow the physical system to inherit the rapid local dynamics of the auxiliary replica, thereby facilitating transitions between different winding‑number sectors. The exchange acceptance probability follows the standard Metropolis‑Hastings form, but the ladder is constructed in the space of interaction parameters rather than temperature, which is crucial because high temperature suppresses permutation cycles.

2. Hamiltonian Monte Carlo with No‑U‑Turn Sampler (NUTS) – For local updates of bead positions, the authors replace random‑walk Metropolis steps with Hamiltonian dynamics. NUTS automatically determines the number of leapfrog steps required to avoid a U‑turn, eliminating the need for manual tuning of trajectory length. This yields proposals with high acceptance rates even in the presence of the smooth, WCA‑type softened potentials introduced by the next component.

3. Stochastic Potential Switching (SPS) – The original pair potential U(r) is stochastically decomposed into a short‑range, smooth reference potential ˜U(r) (chosen as a WCA‑type repulsive core) and a complementary long‑range part ¯U(r). A switching probability S(r)=exp