ase-geometry-optimization-md

name	ase_geometry_optimization_md
description	Running molecular geometry optimization and molecular dynamics (MD) simulations using ASE and MLIPs.

Geometry Optimization & Molecular Dynamics Simulations

Use this skill when running relaxation calculations (to find minimum energy structures) or Molecular Dynamics (MD) trajectories using Atomic Simulation Environment (ASE) and a Machine Learning Interatomic Potential (MLIP).

Conceptual Framework

Geometry Optimization (Structural Relaxation)

Finding the ground-state conformation requires minimizing the total potential energy of the system with respect to atomic coordinates.

• L-BFGS Algorithm: A quasi-Newton method that uses gradients (forces) to iteratively update coordinates, converging to a configuration where forces are close to zero: $$F_i = -\nabla_{R_i} E \approx 0$$
• Max Force Threshold ($f_{max}$): The relaxation continues until the maximum atomic force component falls below a specified convergence criterion (typically 0.01 to 0.05 eV/Å).

Molecular Dynamics (MD)

MD simulates the physical movements of atoms over time by solving Newton's equations of motion: $$F = m \cdot a$$

• Integration Timestep: Typically 0.5 to 1.0 femtoseconds (fs) to capture high-frequency molecular vibrations (such as C-H stretches).
• Thermodynamic Ensembles:
  • NVT (Constant Number, Volume, Temperature): Uses a thermostat (e.g., Langevin or Nose-Hoover) to maintain target temperature by coupling the system to a virtual heat bath.
  • NPT (Constant Number, Pressure, Temperature): Uses a barostat to maintain pressure alongside temperature.

Execution

Run the script ase_md_simulation.py to perform optimization and Langevin dynamics:

python scripts/ase_md_simulation.py --temp 300 --steps 100

This script relaxes a cluster of atoms and runs NVT dynamics, saving a .traj history.