computational-input

Use when the user asks to generate DFT input files (VASP, Quantum ESPRESSO, LAMMPS), optimize structures with ML potentials (MACE, CHGNet, M3GNet), compute energy, or set up any computational chemistr

SKILL.md

GitHub

Computational Input Generation

Quick Decision Guide

Task Tool
VASP input (INCAR/POSCAR/KPOINTS) catgo_vasp_generate
Quantum ESPRESSO pw.x input catgo_qe_generate
LAMMPS input + data file catgo_lammps_generate
Multi-stage LAMMPS simulation catgo_lammps_sequential
ML potential relaxation catgo_optimize
Single-point energy/forces catgo_energy
List available calculators catgo_calculators

VASP Input

catgo_vasp_generate

  • Calculation types: opt, scf, freq, bader, dos, ddec, elf
  • Key params: encut (default 450 eV), gga ("PE"=PBE), ispin (2=spin-polarized), ivdw (12=D3-BJ), kspacing, fixed_indices/fixed_z_below

Common patterns:

  • Bulk optimization: calculation_type="opt", isif=3 (relax cell+ions)
  • Slab optimization: calculation_type="opt", isif=2, fixed_z_below=Z
  • DOS: calculation_type="dos", dense k-mesh

Call catgo_vasp_calc_types to list all available types with defaults.

Quantum ESPRESSO Input

catgo_qe_generate

  • Calculation types: scf, relax, vc-relax, nscf, bands
  • Key params: ecutwfc (default 60 Ry), ecutrho (default 480 Ry), kspacing, occupations, smearing ("mv"=Marzari-Vanderbilt), nspin

Call catgo_qe_templates for recommended settings per calculation type.

LAMMPS Input

catgo_lammps_generate

  • Simulation types: minimize, nve, nvt, npt
  • Key params: pair_style, pair_coeff, potential_file, temperature, pressure

catgo_lammps_sequential — Multi-stage MD protocol:

{"stages": [
  {"name": "minimize", "simulation_type": "minimize"},
  {"name": "heat", "simulation_type": "nvt", "temperature": 300, "run_steps": 10000},
  {"name": "equilibrate", "simulation_type": "npt", "temperature": 300, "run_steps": 50000},
  {"name": "production", "simulation_type": "nvt", "temperature": 300, "run_steps": 100000}
]}

Call catgo_lammps_pair_styles for available force fields. Call catgo_lammps_validate before generating to check configuration.

ML Potential Optimization

catgo_optimize

Quick relaxation using ML interatomic potentials:

  • mace: Best accuracy for most systems
  • chgnet: Good for oxides
  • m3gnet: General purpose
  • emt: Fast, metals only (testing)

Params: fmax (default 0.05 eV/A), max_steps (200), relax_cell (True to relax lattice)

catgo_energy — Single-point energy + forces without optimization.

Workflow Recipes

ML Pre-Optimization then DFT

  1. catgo_optimize(calculator="mace", fmax=0.05) → 2. catgo_vasp_generate(calculation_type="opt")

VASP Slab Calculation

  1. Build slab → 2. catgo_vasp_generate(calculation_type="opt", isif=2, fixed_z_below=Z, encut=520)