Materials Computation Environment

This container includes a full materials science computation environment. Use these tools for atomistic simulation tasks.

Available Computation Engines

Quantum ESPRESSO 7.5 (DFT)

• Binary: pw.x (also ph.x, pp.x, bands.x, dos.x, projwfc.x, etc.)

• Location: /opt/qe/bin/

• Use for: electronic structure, band gaps, density of states, phonons, elastic constants

• Run with MPI: mpirun --allow-run-as-root -np N pw.x < input.in > output.out

• Pseudopotentials download URL:

  Copy

  https://pseudopotentials.quantum-espresso.org/upf_files/<FILENAME>.UPF
  

  Common pseudopotentials (PAW PBE from pslibrary):

  • Si: Si.pbe-n-kjpaw_psl.1.0.0.UPF
  • Cu: Cu.pbe-dn-kjpaw_psl.1.0.0.UPF
  • Al: Al.pbe-n-kjpaw_psl.1.0.0.UPF
  • Ni: Ni.pbe-n-kjpaw_psl.1.0.0.UPF
  • O: O.pbe-n-kjpaw_psl.1.0.0.UPF
  • C: C.pbe-n-kjpaw_psl.1.0.0.UPF
  • H: H.pbe-kjpaw_psl.1.0.0.UPF
  • N: N.pbe-n-kjpaw_psl.1.0.0.UPF
  • Fe: Fe.pbe-spn-kjpaw_psl.1.0.0.UPF
  • Ti: Ti.pbe-spn-kjpaw_psl.1.0.0.UPF
  • Zn: Zn.pbe-dn-kjpaw_psl.1.0.0.UPF
  • Ba: Ba.pbe-spn-kjpaw_psl.1.0.0.UPF
  • Li: Li.pbe-s-kjpaw_psl.1.0.0.UPF
  • Na: Na.pbe-spn-kjpaw_psl.1.0.0.UPF
  • K: K.pbe-spn-kjpaw_psl.1.0.0.UPF

  Example download:

  bashCopy

  wget -q https://pseudopotentials.quantum-espresso.org/upf_files/Si.pbe-n-kjpaw_psl.1.0.0.UPF -O Si.UPF
  

  Alternative sources: SSSP or PseudoDojo

LAMMPS (Molecular Dynamics)

• Binary: lmp
• Use for: MD simulations, thermal properties, diffusion, mechanical properties
• Supports OpenKIM potentials (pre-installed)
• Run: lmp -in input.lammps
• EAM potentials: use pair_style eam/alloy with potentials from OpenKIM or download from NIST
• For water: use SPC/E or TIP3P model with pair_style lj/cut/coul/long

RASPA3 (Monte Carlo)

• Binary: raspa3
• Use for: gas adsorption in porous materials (MOFs, zeolites), adsorption isotherms, Henry constants
• Run: cd /path/to/simulation && raspa3
• Input format: JSON files (simulation.json, force_field.json, molecule JSON files)
• Official examples: /usr/share/raspa3/examples/ (basic MC, adsorption, breakthrough, etc.)
• IMPORTANT: Always copy an official example as starting point and modify it:
  bashCopy

  cp -r /usr/share/raspa3/examples/basic/1_mc_methane_in_box /tmp/my_sim
  cd /tmp/my_sim && raspa3
  

Python Materials Science Stack

Pre-installed in base conda environment:

• pymatgen: Crystal structure manipulation, phase diagrams, electronic structure analysis
• ASE (Atomic Simulation Environment): Atoms objects, calculators, optimization, MD
• mp-api: Materials Project API access (needs API key)
• MACE-torch: Universal machine learning interatomic potential
• spglib: Space group analysis
• torch: PyTorch (CPU version)
• numpy / scipy / matplotlib: Scientific computing and visualization

Conda/pip available:

The agent can install additional packages as needed:

# Create isolated environment for specific tasks
conda create -n myenv python=3.11 -y
conda activate myenv

# Install additional ML potentials
pip install chgnet sevenn

# Install workflow managers
pip install fireworks jobflow atomate2

Common Workflows

DFT Calculation (QE)

1. Prepare structure (use pymatgen to read CIF/POSCAR and generate QE input)
2. Download pseudopotentials
3. Run SCF calculation: pw.x < scf.in > scf.out
4. Post-process: bands, DOS, charge density, etc.

MD Simulation (LAMMPS or ASE+MACE)

1. Prepare structure and force field
2. Set up LAMMPS input or ASE calculator
3. Run simulation
4. Analyze trajectory: RDF, MSD, thermal conductivity, etc.

MLIP Calculation (ASE + MACE)

from ase.io import read
from mace.calculators import mace_mp
calc = mace_mp(model="medium", device="cpu")
atoms = read("structure.cif")
atoms.calc = calc
energy = atoms.get_potential_energy()
forces = atoms.get_forces()

VASP (via vasp-remote)

If the user has configured VASP access (run /add-vasp), use vasp-remote to execute VASP calculations:

1. Generate INCAR, POSCAR, KPOINTS, POTCAR in the current directory
2. Run: vasp-remote run (submits to cluster or runs locally, waits for results)
3. Parse vasprun.xml with pymatgen

Check if VASP is available: vasp-remote config 2>/dev/null. If not configured, fall back to Method A (MACE) or Method B (QE).

Monte Carlo (RASPA3)

1. Prepare framework structure (CIF)
2. Configure simulation input (guest molecules, temperature, pressure)
3. Run RASPA3
4. Analyze adsorption isotherm

Skill Reference Index

IMPORTANT: Before performing any materials science computation, check the relevant skill guide below. Each skill provides complete, runnable code for three approaches: ASE+MACE (fast screening), QE DFT (accurate), and VASP (future external). Read the SKILL.md in the skill directory for step-by-step instructions.

Skills are located at ~/.claude/skills/<group>/<sub-skill>/SKILL.md.

Structure & Symmetry

Electronic Structure

Mechanical & Thermal Properties

Bonding & Charge Analysis

Optical, Magnetic & Transport

Ferroelectric & Piezoelectric

Catalysis & Defects

2D Materials & Semiconductors

Monte Carlo & Phase Diagrams

Code Interfaces & MLIP

Other

How to Use a Skill

1. Identify the relevant skill from the index above
2. Read the SKILL.md: cat ~/.claude/skills/<group>/<sub-skill>/SKILL.md
3. Choose method: Method A (ASE+MACE, fast) or Method B (QE DFT, accurate) or Method C (VASP)
4. Follow the step-by-step code in the SKILL.md — all code is complete and runnable
5. Check the "Common Issues" table at the bottom if anything goes wrong