A MarkTechPost tutorial examines FAIRChem v2 and the UMA universal machine-learning interatomic potential as a unified framework for atomistic simulation across molecular chemistry, catalysis, and inorganic materials. The workflow uses FAIRChem with the Atomic Simulation Environment, or ASE, to manage atomic structures, optimizers, constraints, thermodynamic calculations, trajectory analysis, and GPU acceleration when available.

The tutorial configures the software environment, authenticates with Hugging Face to access gated UMA model weights, and initializes task-specific calculators for the omol, oc20, and omat domains. It applies the same pretrained potential across several computational chemistry tasks, including single-point energy and force prediction, molecular geometry optimization, spin-state comparison, reaction-energy estimation, vibrational analysis, surface adsorption, crystal-cell relaxation, equation-of-state fitting, molecular dynamics, and potential-energy surface scanning. That breadth is relevant because atomistic modeling workflows often require switching between domain-specific methods, reference conventions, and model setups when moving from isolated molecules to surfaces or periodic solids.

The setup installs FAIRChem, ASE, visualization tools, and Hugging Face dependencies while keeping the process safe to rerun in Google Colab. After authentication through Hugging Face, the workflow automatically checks whether GPU acceleration is available. It then loads the UMA predictor and creates separate calculators for molecular, catalysis, and materials simulations. FAIRChem is available through the project’s GitHub repository at https://github.com/facebookresearch/fairchem.

For molecular calculations, the tutorial uses the UMA molecular calculator to evaluate the energy, atomic forces, and atomization energy of a water molecule. It defines isolated hydrogen and oxygen reference atoms with the appropriate spin multiplicities for the atomization-energy calculation. The water geometry is then distorted and relaxed with the LBFGS optimizer, after which the converged bond length, bond angle, and energy trajectory are analyzed.

The workflow also compares the singlet and triplet electronic states of methylene to calculate the spin-state energy gap. It relaxes methane, oxygen, carbon dioxide, and water, then combines their predicted energies to estimate the electronic reaction energy of methane combustion. In another molecular example, the tutorial performs a finite-difference vibrational analysis of relaxed water to obtain normal-mode frequencies and zero-point energy. These examples cover common validation-style checks for learned potentials: whether forces can drive stable optimization, whether relative energies can be combined consistently, and whether local curvature around a minimum can support vibrational calculations.

For catalysis, the tutorial constructs a periodic Cu(100) slab, places a carbon monoxide molecule at a bridge adsorption site, and constrains the lower copper layers. The adsorbate–surface system is relaxed with the OC20 calculator, while the clean slab and gas-phase carbon monoxide references are optimized separately. The tutorial then evaluates a pedagogical adsorption-energy cycle, while noting that OC20 uses a task-specific energy reference convention.

For materials modeling, the workflow relaxes both the atomic positions and the simulation cell of BCC iron using the materials-domain UMA calculator and a Frechet cell filter. It samples energies over a range of compressed and expanded volumes, fits a Birch–Murnaghan equation of state, and estimates the equilibrium volume and bulk modulus. Cell relaxation and equation-of-state fitting are standard solid-state simulation steps because they connect atomistic energies to structural and elastic properties rather than only to single geometries.

The tutorial further runs Langevin molecular dynamics for water at 300 K, tracking temperature, potential energy, and O–H bond length over time. It also scans the potential-energy surface of water by systematically stretching one O–H bond across a selected distance range. At each geometry, the molecular energy is evaluated, normalized relative to the minimum, and used to visualize the resulting dissociation profile.

The article concludes that the workflow demonstrates how FAIRChem v2 and UMA can provide a shared learned potential across chemically distinct domains without requiring a separate model for every task. Molecular calculations are used to evaluate energies, forces, atomization behavior, spin gaps, reaction energetics, vibrational modes, and bond-stretch profiles. The catalysis domain is used to relax CO on a Cu(100) surface and examine adsorption energetics, while the materials domain is used to relax BCC iron and estimate its bulk modulus from an equation-of-state fit.

By combining UMA inference with ASE structure builders, optimizers, filters, vibrational tools, and molecular-dynamics utilities, the tutorial presents a reusable basis for extending the workflow to larger molecules, catalytic interfaces, crystalline materials, metal-organic frameworks, molecular crystals, and higher-accuracy UMA model variants. The tutorial’s full code is available at https://github.com/MARKTECHPOST-AI-MEDIA-INC/AI-Agents-Projects-Tutorials/blob/main/ML%20Project%20Codes/fairchem_v2_uma_multidomain_atomistic_simulation_Marktechpost.ipynb.