vibeflow-lammps

name	vibeflow-lammps
description	Use when preparing, running, debugging, post-processing, or extending VibeFlow/VibeCAE LAMMPS molecular dynamics and particle simulation workflows from 0-caseDict/caseDict, including LAMMPS input scripts, data files, potentials, fixes, computes, dumps, restarts, Python post-processing, and optional custom LAMMPS package or style development.

VibeFlow LAMMPS

This skill is responsible for using LAMMPS as VibeCAE's backend for molecular dynamics, particulate systems, material microscale simulation, and coarse-grained particle simulation.

• Usage mode: generate LAMMPS inputs, data, potential configuration, run scripts, and post-processing from 0-caseDict/caseDict.
• Development mode: develop reusable LAMMPS workflows, input templates, Python analysis tools, or custom styles/packages when explicitly needed.

When to Use

• The user task involves molecular dynamics, particulate systems, atomistic materials, soft matter, heat transport, diffusion, collision, deposition, or coarse-grained particle simulation.
• in.* input scripts, data.* structure files, potential file references, dump, compute, fix, or thermo_style need to be generated or repaired.
• MD results need to be converted into trajectories, statistics, radial distribution functions, mean-square displacement, stress-strain curves, thermal conductivity, diffusion coefficients, or similar quantities.
• LAMMPS results need to be weakly coupled with, or handed off as data to, continuum CAE workflows.

First Decide: Usage Mode or Development Mode

• If only the current case needs to be configured, run with LAMMPS, and post-processed: use usage mode.
• If Python data processors, template generators, batch parameter sweeps, coupling interfaces, or custom LAMMPS styles are needed: use development mode.
• If lmp, MPI, shared libraries, or package availability are abnormal: switch to vibeflow-toolchain first.
• If an initial particle distribution must be generated from CAD or mesh data: switch first to the relevant geometry or mesh skill.

What to Read First

• The physical system, materials, particle/atom species, boundaries, initial conditions, thermodynamic ensemble, run controls, and post-processing objectives in 0-caseDict/caseDict.
• references/lammps-playbook.md
• references/lammps-practical-tips.md: software knowledge base, key concepts, failure interpretation, and adapter design points.
• Existing potentials, structure files, or reference inputs under C-run/lammps/, D-post/, and Ref/.

Inputs

• 0-caseDict/caseDict
• LAMMPS data files, molecule files, potential files, and restart files
• User-provided material parameters, lattices, particle-size distributions, temperature/pressure conditions, loading paths, and time scales

Outputs

• C-run/lammps/case-*/in.case
• C-run/lammps/case-*/data.*
• C-run/lammps/case-*/potentials/
• C-run/lammps/case-*/Allrun and Allclean
• C-run/lammps/case-*/logs/
• C-run/lammps/case-*/dumps/
• D-post/lammps_postprocess.py
• Charts, statistics tables, converted trajectory artifacts, and result notes

Recommended Structure

C-run/lammps/
└── case-1/
    ├── in.case
    ├── data.system
    ├── potentials/
    ├── dumps/
    ├── restarts/
    ├── logs/
    ├── Allrun
    └── Allclean
D-post/
└── lammps_postprocess.py

Usage Mode Workflow

1. Map the unified physical description in caseDict to LAMMPS concepts: units, atom_style, boundary, region, group, pair_style, bond_style, fix, and compute.
2. Clarify the potential source and applicability; do not use a potential that merely looks similar but does not match the physical system.
3. Generate in.case and data files; before running, check units, time step, neighbor list, thermostat/barostat method, and output frequency.
4. Write Allrun, preferably selecting the LAMMPS executable through an environment variable instead of hard-coding a local path.
5. Run a short-step smoke test, confirming that the input parses, initializes, and produces no ERROR.
6. During the production run, record logs, dumps, restarts, and key thermo outputs.
7. Use D-post/lammps_postprocess.py to read logs or dumps and extract the physical quantities that the user actually cares about.
8. Write artifacts, assumptions, potential sources, and error status back to caseDict.

Development Mode Workflow

1. Define the development target: template generator, Python analyzer, coupling interface, custom fix/compute/pair style, or batch-processing tool.
2. Write development notes first, recording the LAMMPS version, package dependencies, input interface, output interface, and validation examples.
3. For custom styles, prefer an independent development directory and do not directly pollute the system LAMMPS source; modify upstream source only when the user explicitly asks for it.
4. Provide units, defaults, error messages, and a minimal example for every new input keyword.
5. Use at least one tiny system for a loading or run smoke test.

Core Rules

• The LAMMPS input script is an execution artifact; caseDict is the source of truth for case intent.
• Potential functions must record their source, version, applicable elements or particle types, unit system, and limitations.
• Time step, thermostat, barostat, constraints, walls, and contact models must be consistent with units.
• Do not judge correctness only by "it runs"; also check whether energy, temperature, pressure, particle count, and boundary conditions are reasonable.
• Large dump files should not be fully loaded into memory by default; post-processing should stream chunks or read only necessary fields.

Validation

• The LAMMPS executable is available.
• in.case can parse and initialize.
• The smoke test has no ERROR, no obvious NaN, and no lost particles.
• Thermo output contains the required monitored quantities.
• D-post can generate target charts from logs or dumps.

Fallback Rules

• Physical system, potential, or units are unclear: return to vibeflow-case-dict
• LAMMPS environment is unavailable: return to vibeflow-toolchain
• Post-processing charts are insufficient to support the conclusion: return to the post-processing step in this skill
• Continuum CFD/FEM is needed: hand off to the corresponding OpenFOAM, SU2, Code_Aster, Elmer, CalculiX, or similar skill