Skip to content

Quick Start

Get started with a2c_ase in minutes.

Prerequisites

pip install mace-torch

uv pip install mace-torch

Five-Step Workflow

1. Generate Random Structure

from a2c_ase.utils import random_packed_structure
from pymatgen.core.composition import Composition
import numpy as np

comp = Composition("Si64")
cell = np.array([[11.1, 0.0, 0.0], [0.0, 11.1, 0.0], [0.0, 0.0, 11.1]])

atoms, log = random_packed_structure(comp, cell, auto_diameter=True)

2. Run Melt-Quench MD

from a2c_ase.runner import melt_quench_md

amorphous, md_log = melt_quench_md(
    atoms, calculator,
    T_high=2000.0, T_low=300.0,
    equi_steps=2500, cool_steps=2500, final_steps=2500
)

3. Extract Subcells

from a2c_ase.utils import extract_crystallizable_subcells

subcells = extract_crystallizable_subcells(
    amorphous, d_frac=0.2, n_min=2, n_max=8, cubic_only=True
)

4. Optimize Structures

from a2c_ase.runner import relax_unit_cell

for subcell in subcells:
    relaxed, logger = relax_unit_cell(subcell, calculator, fmax=0.01)
    # Analyze relaxed structure

5. Analyze Results

from pymatgen.analysis.structure_analyzer import SpacegroupAnalyzer

# Find lowest energy and determine space group
# Compare to reference structures

Complete Example

See the full working example: example/Si64.py

Next Steps

Tips

Memory

Large systems with many subcells need significant memory. Use larger d_frac or smaller n_max to reduce subcell count.