MLIP-MC: A Framework for Adsorption Simulations using Machine-Learned Interatomic Potentials

TL;DR

MLIP-MC is an open-source framework enabling GCMC simulations with universal machine-learned interatomic potentials, revealing current models' biases and the need for finetuning for accurate gas adsorption predictions.

Contribution

We introduce MLIP-MC, a framework for GCMC simulations using universal MLIPs, and benchmark several models for CO2 adsorption on MOFs, highlighting their limitations and data dependence.

Findings

Universal MLIPs show systematic biases in adsorption energetics.

Accuracy depends on training data, with MOF-adsorbate interactions being crucial.

Errors increase linearly with CO2 uptake, indicating compounded inaccuracies.

Abstract

Grand canonical Monte Carlo (GCMC) simulations are essential for screening metal-organic frameworks (MOFs) for gas adsorption, yet their accuracy is limited by underlying interatomic potentials. Universal machine-learned interatomic potentials (MLIPs), trained on diverse chemical datasets, promise zero-shot prediction without system-specific training. We introduce MLIP-MC, an open-source Python framework to conduct GCMC simulations with MLIPs, and use this framework to benchmark a series of universal models, including MACE-MP-0, ORB-v3, and fairchem ODAC, for CO2 adsorption on ZIF-8, ZIF-4, and Mg-MOF-74. All universal models exhibit systematic biases, consistently over- or underestimating adsorption energetics. Crucially, accuracy depends on training data composition: only models trained on MOF-adsorbate interactions achieve reasonable agreement with a density functional theory derived reference. Errors grow linearly with CO2 uptake, reflecting compounding inaccuracies in adsorbate-adsorbate interactions. Our results demonstrate that current universal MLIPs require finetuning for quantitative adsorption predictions and demonstrate the power of MLIP-MC to rapidly test models.