Grand canonical generators model fluids with variable particles and conditions
Grand Canonical Generators
Machine Learning
Summary
Simulating systems where the number of particles can change is tricky. The authors introduce a new approach called Grand Canonical Generators (GCG) that can handle this by either simulating the whole system at once or breaking it down into smaller, easier parts. Their method works well on fluid and gas simulations, capturing how particles enter and leave the system under different conditions. This helps predict physical properties more accurately and efficiently.
What this means in practice
- •For chemical engineers: Model gas adsorption in porous materials under varying conditions using more accurate and flexible particle simulations.
- •For computational physicists: Generate particle configurations that adapt dynamically to chemical potential changes for simulations in physical chemistry and materials science.
Authors
Andreas Burger, Malte Franke, Luka Mucko, Kjell Jorner, Alan Aspuru-Guzik
Abstract
We introduce Grand Canonical Generators (GCG), a generative framework that extends Boltzmann generators to the grand canonical ensemble. We present two designs. The first conditions a variable-size generative model on the chemical potential, sampling particle number and configuration jointly. The second factorizes the grand canonical distribution into a particle-number distribution and the corresponding canonical Boltzmann density. This factorized formulation can use any existing Boltzmann generator for the canonical component, encodes the known linear chemical-potential dependence analytically, and yields a tractable likelihood that supports self-normalized importance sampling (SNIS). Empirically, GCG accurately reproduces grand canonical observables on a Lennard--Jones fluid and methane adsorption in a zeolite, demonstrating generalization across chemical potentials and correction via SNIS and grand canonical Monte Carlo.