Skip to content

Hello-QM/catgo-LRG / .claude/skills/vasp-md

vasp-md

Verified

Ab initio molecular dynamics (AIMD) with VASP. NVT/NVE ensembles, temperature control, trajectory analysis.

VASP Ab Initio Molecular Dynamics (AIMD)

Run Born-Oppenheimer molecular dynamics with DFT forces at each step. Expensive but provides finite-temperature behavior, diffusion coefficients, and reaction dynamics.

When to Use

  1. Thermal stability — check if a structure is stable at operating temperature
  2. Diffusion — compute diffusion coefficients (e.g., Li-ion conductors)
  3. Reaction dynamics — observe bond breaking/forming at finite temperature
  4. Free energy sampling — metadynamics or thermodynamic integration
  5. Amorphous structures — melt-quench to generate amorphous phases

Basic NVT Molecular Dynamics

from catgo.workflow import Workflow
from catgo.workflow.builtins import geo_opt, md

wf = Workflow("AIMD at 600K")
struct = wf.add_task("structure_input", structure=structure_json)

# Optional: relax first
opt = wf.add_task(geo_opt, structure=struct.output.structure,
                  system_name="relax")

# NVT MD at 600 K
run = wf.add_task(md, structure=opt.output.structure,
                  IBRION=0,      # Molecular dynamics
                  NSW=5000,      # Number of MD steps
                  POTIM=1.0,     # Time step in fs
                  TEBEG=600,     # Starting temperature (K)
                  TEEND=600,     # Ending temperature (K)
                  SMASS=0,       # Nose-Hoover thermostat (NVT)
                  ISIF=2,        # Fix cell shape/volume
                  system_name="AIMD_600K")

wf.submit()

MCP Workflow

catgo_workflow_engine(action="create", params={"name": "AIMD 600K"})

catgo_workflow_engine(action="add_task", params={
  "workflow_id": "wf_xxx",
  "task_type": "structure_input",
  "structure": "<json>"
})

catgo_workflow_engine(action="add_task", params={
  "workflow_id": "wf_xxx",
  "task_type": "md",
  "software": "vasp",
  "structure": "{{t_001.output.structure}}",
  "NSW": 5000,
  "POTIM": 1.0,
  "TEBEG": 600,
  "TEEND": 600,
  "SMASS": 0,
  "system_name": "AIMD_600K"
})

catgo_workflow_engine(action="submit", params={"workflow_id": "wf_xxx"})

Key Parameters

ParameterDefaultPurpose
IBRION0Molecular dynamics mode
NSW1000Number of MD steps
POTIM1.0Time step in femtoseconds
TEBEG300Initial temperature (K)
TEEND300Final temperature (K). Set equal to TEBEG for isothermal
SMASS-1Thermostat: -1=NVE, 0=Nose-Hoover NVT, >0=Nose mass
ISIF2Fix cell (NVT). Use ISIF=3 for NPT (rare in AIMD)
NBLOCK1Write trajectory every NBLOCK steps

Thermostat Selection (SMASS)

SMASSEnsembleUse case
-1NVE (microcanonical)Energy conservation test, short dynamics
0NVT Nose-HooverStandard production MD at fixed T
1-3NVT with Nose massLarger SMASS = slower T coupling (less perturbation)
-3Langevin thermostatBetter T control for small systems

Recommendation: Use SMASS=0 (Nose-Hoover) for most production runs. Use SMASS=-1 (NVE) for energy conservation checks and very short equilibration diagnostics.

Temperature Ramp (Heating/Cooling)

To heat from 300 K to 1500 K (simulated annealing or melt-quench):

run = wf.add_task(md, structure=s,
                  NSW=10000,
                  POTIM=2.0,
                  TEBEG=300,     # Start at 300 K
                  TEEND=1500,    # Ramp to 1500 K
                  SMASS=0,
                  system_name="heating_ramp")

Time Step Selection (POTIM)

SystemRecommended POTIM (fs)Reason
Heavy elements (Pt, Au, Ru)2.0Heavy atoms, slow dynamics
Oxides (TiO2, RuO2)1.0-1.5O is light, moderate step needed
Light elements (H, Li)0.5-1.0Fast H vibrations need small step
Proton transfer0.5H requires fine time resolution

Rule of thumb: if the total energy drifts upward in NVE, reduce POTIM.

Performance Settings

AIMD is expensive. Optimize performance:

run = wf.add_task(md, structure=s,
                  NSW=5000,
                  POTIM=1.0,
                  TEBEG=600,
                  SMASS=0,
                  # Performance
                  ALGO="VeryFast",    # Fastest SCF convergence per step
                  NELM=60,            # Limit SCF steps (MD does not need tight SCF)
                  EDIFF=1e-4,         # Looser SCF for MD (still accurate forces)
                  LREAL="Auto",       # Real-space projection for speed
                  LWAVE=False,        # Do not write WAVECAR each step
                  NCORE=4,
                  system_name="AIMD")

EDIFF=1e-4 is acceptable for MD. Forces at 1e-4 SCF convergence are sufficiently accurate for MD trajectories. This saves 30-50% compute time vs 1e-5.

Supercell Size

AIMD requires large enough supercells to avoid finite-size effects:

  • Minimum: 64-100 atoms for bulk liquids/diffusion
  • Surfaces: use the slab supercell (typically 2x2 or 3x3 surface unit cell)
  • Small molecules on surface: existing slab supercell is usually fine

Build a supercell before MD:

catgo_structure(action="supercell", params={"scaling": [2, 2, 1]})

Output

The md task produces:

  • output.trajectory — atomic positions at each step (XDATCAR)
  • output.energy — energy vs time

Monitoring a Running MD

catgo_workflow_engine(action="status", params={"workflow_id": "wf_xxx"})
# Check NSW progress, temperature stability

catgo_analyze(action="convergence", params={"task_id": "t_md"})
# Energy vs step, temperature vs step

Troubleshooting

ProblemFix
Temperature explodesReduce POTIM, check initial structure for overlapping atoms
Energy drift in NVEReduce POTIM, tighten EDIFF to 1e-5
SCF not converging at each stepUse ALGO=VeryFast, increase NELM
Too slowReduce ENCUT (400 eV ok for MD), use LREAL=Auto, fewer k-points
Atoms evaporating from slabAdd more vacuum, or constrain bottom layers

Typical Simulation Lengths

PurposeNSWPOTIMTotal time
Quick stability check10001.01 ps
Equilibration50001.05 ps
Production (diffusion)20000-500001.0-2.020-100 ps
Melt-quench10000+2.020+ ps
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 197kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall