跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
其他技能

chem-solution-md

Set up and run molecular dynamics simulations of molecules in explicit solvent boxes using Packmol for box construction and MLIPs for dynamics.

learningmatter-mit · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add learningmatter-mit/AtomisticSkills --skill chem-solution-md

技能说明


name: chem-solution-md description: Set up and run molecular dynamics simulations of molecules in explicit solvent boxes using Packmol for box construction and MLIPs for dynamics. category: [chemistry]

Solution-Phase Molecular Dynamics

Goal

Set up and run molecular dynamics (MD) simulations of molecules in explicit solvent. This skill covers three stages: (1) building a solvation box with Packmol, (2) running NPT/NVT MD using MLIPs, and (3) analyzing the trajectory for radial distribution functions (RDFs), coordination numbers, density convergence, and mean-square displacement (MSD).

[!IMPORTANT] This skill bridges gas-phase chem-* skills and condensed-phase mat-* skills by providing workflows for solvation dynamics, liquid structure characterization, and dissolution studies.

1. Prerequisites

  • Packmol binary must be installed and on PATH in the base-agent environment.
  • RDKit must be available in the base-agent environment (for SMILES → 3D geometry).
  • An MLIP backend must be available via MCP tools (MACE, MatGL, or FairChem).

2. MLIP Selection

Refer to the foundation-potentials skill for model selection.

[!NOTE]

  • Organic solvents: Use MACE-MH-1 with omol head, or UMA with omol task.
  • Aqueous inorganic systems: Use MACE-MH-1 with omat_pbe head, or MatGL/CHGNet.
  • Mixed organic-inorganic: Use UMA which handles both.

3. Workflow

Step 1: Build Solvation Box

Use the box-building script to create a solvated system with Packmol:

# Env: base-agent
# Pure solvent box (64 water molecules)
python .agents/skills/chem-solution-md/scripts/build_solvation_box.py \
    --solvent water \
    --num_solvent 64 \
    --output_dir research/my_folder/solvation_box

# Solute in solvent (NaCl in 64 water molecules)
python .agents/skills/chem-solution-md/scripts/build_solvation_box.py \
    --solute_smiles "[Na+].[Cl-]" \
    --solvent water \
    --num_solvent 64 \
    --output_dir research/my_folder/solvation_box

Key Parameters:

ArgumentDescription
--solventPre-defined solvent name (see available solvents below)
--solvent_smilesSMILES string for custom solvent
--solvent_filePath to solvent structure file
--solute_smilesSMILES string for solute (optional)
--solute_filePath to solute structure file (optional)
--num_solventNumber of solvent molecules (default: 64)
--box_sizeCubic box side in Å (auto-calculated from density if omitted)
--toleranceMinimum inter-molecular distance in Å (default: 2.0)
--output_dirOutput directory

Available pre-defined solvents: water, methanol, ethanol, acetonitrile, dmso, dmf, thf, toluene, acetone, dichloromethane, chloroform, hexane

Output files:

  • solvated_box.cif — Periodic structure for MD
  • solvated_box.xyz — Non-periodic XYZ for visualization
  • box_metadata.json — Box size, atom counts, solute indices

Step 2: Run MD with MLIP

Use MCP run_md tools for NPT equilibration followed by NVT production.

NPT Equilibration (stabilize density):

mcp_mace_load_model(
    model_name="MACE-MH-1",
    task_name="omol"
)
mcp_mace_run_md(
    structure_data="research/my_folder/solvation_box/solvated_box.cif",
    temperature=300,
    ensemble="npt",
    pressure=1.01325,        # 1 atm in bar
    steps=5000,              # 2.5 ps at 0.5 fs timestep
    timestep=0.5,            # 0.5 fs for systems with water (fast O-H vibrations)
    log_interval=10,
    monitor=True,
    monitor_type=["explosion", "volume"],
    output_dir="research/my_folder/npt_equilibration"
)

NVT Production (use the equilibrated structure):

mcp_mace_run_md(
    structure_data="research/my_folder/npt_equilibration/final_structure.cif",
    temperature=300,
    ensemble="nvt",
    steps=20000,             # 10 ps at 0.5 fs timestep
    timestep=0.5,
    log_interval=10,
    monitor=True,
    monitor_type="explosion",
    output_dir="research/my_folder/nvt_production"
)

Step 3: Analyze Trajectory

Run the analysis script on the production trajectory:

# Env: base-agent
python .agents/skills/chem-solution-md/scripts/analyze_solution_md.py \
    --trajectory research/my_folder/nvt_production/trajectory.traj \
    --rdf_pairs "Na-O,Cl-O,O-O" \
    --msd_elements "Na,Cl" \
    --log_interval_fs 5.0 \
    --output_dir research/my_folder/analysis

Key Parameters:

ArgumentDescription
--trajectoryPath to ASE .traj trajectory file
--rdf_pairsComma-separated element pairs for RDF, e.g. "Na-O,Cl-O"
--rmaxMaximum RDF distance in Å (default: 8.0)
--start_frameFirst frame to include in analysis (default: 0)
--strideFrame stride (default: 1)
--log_interval_fsTime between frames in fs (default: 10.0)
--msd_elementsComma-separated elements for MSD (optional)

Output files:

  • solution_analysis.json — Full results (RDF data, coordination numbers, density, MSD)
  • rdf_plots.png — RDF plots for each element pair
  • density_convergence.png — Density vs. time
  • msd_plot.png — MSD for specified elements (if requested)

4. Examples

Pure Water Box

# Env: base-agent
python .agents/skills/chem-solution-md/scripts/build_solvation_box.py \
    --solvent water --num_solvent 64 \
    --output_dir .agents/skills/chem-solution-md/examples/pure_water

Expected: 192 atoms (64 × 3), box ~12.4 Å

NaCl in Water

# Env: base-agent
python .agents/skills/chem-solution-md/scripts/build_solvation_box.py \
    --solute_smiles "[Na+].[Cl-]" --solvent water --num_solvent 64 \
    --output_dir .agents/skills/chem-solution-md/examples/NaCl_in_water

After MD + analysis, expected RDF peak positions:

  • Na–O first peak: ~2.4 Å
  • Cl–O first peak: ~3.2 Å
  • Na coordination number: ~5–6

5. Constraints

  • Timestep:
阅读完整来源 ↗