Path integral molecular dynamics for bosons

Barak Hirshberg, Valerio Rizzi, Michele ParrinelloProceedings of the National Academy of Sciences 2019 · DOI

TL;DR

This paper introduces a novel method for simulating systems of identical bosons using path integral molecular dynamics (PIMD). By developing a recursive approach to handle particle permutations, the authors overcome a significant computational bottleneck, enabling simulations of larger systems than previously possible. This advancement is crucial for understanding and designing quantum technologies that rely on bosonic systems.

What Problem Does This Paper Address?

Simulating quantum systems accurately is a major challenge in physics and chemistry. For identical particles like bosons, their quantum nature dictates that they are indistinguishable, meaning any arrangement of particles is physically equivalent to another if it's just a permutation of the original. This "exchange symmetry" is fundamental and must be accounted for in simulations.

Traditional PIMD methods often treat particles as distinguishable, which simplifies calculations but misses crucial quantum effects. To include exchange symmetry, one must consider all possible permutations of particles. For bosons, this leads to a combinatorial explosion: the number of permutations grows factorially with the number of particles (N!). This makes direct enumeration of all permutations computationally intractable for even moderately sized systems. While Monte Carlo methods have seen advancements in handling these permutations (like the worm algorithm), applying this to molecular dynamics has been significantly more challenging. Previous attempts using PIMD for bosons either required calculating permanents (which is equivalent to enumerating permutations) and were thus limited to very small systems, or were not fully developed. This paper aims to bridge this gap by providing an efficient PIMD method for bosons that scales favorably with system size.

Methods

The core innovation of this paper lies in a recursive method for evaluating the potential energy and forces in PIMD simulations of bosons. Instead of explicitly enumerating all N! permutations, the authors leverage a recurrence relation to compute the necessary terms. This relation effectively builds up the complex configurations of permuted ring polymers from simpler ones.

The standard PIMD framework represents quantum particles as "ring polymers," where each bead in the polymer corresponds to a discrete time slice of the particle's quantum mechanical path. For distinguishable particles, the partition function can be sampled by simulating these ring polymers classically. To account for indistinguishable bosons, the partition function must sum over all possible permutations of these ring polymers. The authors' recursive approach avoids the direct calculation of the permanent (which sums over all permutations) by cleverly structuring the calculation. This allows for the evaluation of thermal expectation values, including energy and forces, without the prohibitive cost of enumerating all permutations. The resulting algorithm scales cubically with the system size (N³), a significant improvement over factorial scaling.

Datasets

The paper does not introduce new datasets. Instead, it benchmarks its method against:

  • Analytical results: For non-interacting bosons, the method's results for ground-state energy and temperature dependence are compared to known analytical solutions.
  • Numerical diagonalization of the Hamiltonian: For small interacting systems, the authors compare their PIMD results to highly accurate results obtained by directly diagonalizing the many-body Hamiltonian. This is a standard benchmark for quantum many-body methods.
  • Previous PIMD simulations: The method is compared to prior PIMD work on bosons, highlighting the limitations of those approaches.

Key Insights and Ideas

  • Efficiently handling boson permutations: The most significant insight is the development of a recursive algorithm that bypasses the need to enumerate all N! permutations, making PIMD simulations of bosons computationally feasible for larger systems.
  • Cubic scaling is a game-changer: The N³ scaling of the new method is a substantial improvement over factorial scaling, opening doors to simulating systems that were previously out of reach for PIMD.
  • Exchange effects analysis: The paper demonstrates how the relative probabilities of different ring-polymer configurations can be used to analyze the importance of exchange effects, providing a new tool for understanding quantum statistics in PIMD.
  • Well-behaved estimators: The authors highlight that their derived estimators for thermal expectation values are well-behaved, unlike some estimators for distinguishable particles whose variance can grow with the number of beads.

Key Results

  • Accurate energies: The method accurately reproduces ground-state energies for up to 64 non-interacting bosons, showing excellent agreement (within ~0.5%) with analytical results.
  • Validation with interacting systems: For small interacting systems (N=2-4), the PIMD results for energy and density show very good agreement with exact diagonalization results, with mean absolute deviations around 1-3%.
  • Scalability demonstrated: The method was successfully applied to 32 interacting particles in a 2D trap, a system where enumerating all permutations is already prohibitive.
  • Temperature dependence analysis: The paper provides an analysis of how exchange effects become more pronounced at lower temperatures, as evidenced by the increasing probability of the longest ring-polymer configurations.

Limitations and Future Work

  • Computational cost for very large N: While cubic scaling is a significant improvement, it can still become computationally demanding for extremely large numbers of particles.
  • Fermion sign problem: The current work focuses on bosons. Extending this method to handle the infamous "sign problem" for fermions remains a significant challenge.
  • Further applications: The authors suggest that this method could be combined with other techniques, like ring-polymer molecular dynamics for chemical dynamics, to study more complex bosonic systems. They also express optimism about applying it to even larger systems than those tested.

Summary updated

Abstract

Trapped bosons exhibit fundamental physical phenomena and are at the core of emerging quantum technologies. We present a method for simulating bosons using path integral molecular dynamics. The main difficulty in performing such simulations is enumerating all ring-polymer configurations, which arise due to permutations of identical particles. We show that the potential and forces at each time step can be evaluated by using a recurrence relation which avoids enumerating all permutations, while providing the correct thermal expectation values. The resulting algorithm scales cubically with system size. The method is tested and applied to bosons in a 2-dimensional (2D) trap and agrees with analytical results and numerical diagonalization of the many-body Hamiltonian. An analysis of the role of exchange effects at different temperatures, through the relative probability of different ring-polymer configurations, is also presented.
BibTeX
@inproceedings{hirshberg2019path,
  author = {Barak Hirshberg and Valerio Rizzi and Michele Parrinello},
  title = {{Path integral molecular dynamics for bosons}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2019},
  doi = {10.1073/pnas.1913365116},
}
Figure