| name | machine-learning-force-field |
| description | Concept skill for Machine Learning Force Fields (MLFFs). Describes what MLFFs are, the distinction between training and inference, and which tool skills to use. Load this before selecting a specific MLFF framework (DeePMD, MatterSim, etc.). |
| metadata | {"dependent_skills":["deepmd","mattersim","ase-deepmd"],"tags":["MLFF","machine-learning-potential","deep-potential","force-field","training"]} |
Machine Learning Force Field (MLFF)
A Machine Learning Force Field (MLFF) is a surrogate model trained on DFT reference data that predicts atomic energies and forces at a fraction of the computational cost. MLFFs enable large-scale and long-timescale molecular dynamics simulations that would be prohibitively expensive with DFT directly.
Workflow Phases
| Phase | Description |
|---|
| Pre-trained model | A general-purpose model trained on large, diverse datasets. Try this first. |
| Fine-tuning | Adapts a pre-trained model to a target system using domain-specific DFT data. |
| Training from scratch | Used in distillation: train a lightweight student model on teacher-labeled data. |
| Inference / MD | Deploy the trained model for structure relaxation or molecular dynamics. |
When to Use
- Run MD exploration to sample the configuration space of a new material.
- Relax generated structures before DFT validation.
- Screen large candidate sets efficiently before expensive DFT runs.
- Replace DFT in high-throughput workflows after validating accuracy.
Available Tool Skills
| Skill | Framework | Best For |
|---|
deepmd | DeePMD-kit | Training and fine-tuning DPA-1/DPA-2 models; PFD workflow |
mattersim | MatterSim | Pre-trained universal MLFF; structure relaxation and MD |
ase-deepmd | ASE + DeePMD | Running MD with a trained DeePMD model via ASE interface |
Load the appropriate tool skill for detailed instructions (e.g., load_skill("deepmd")).
Choosing a Framework
- Use
deepmd when you need to train or fine-tune a model on your own DFT dataset.
- Use
mattersim when you want a pre-trained universal model without retraining.
- Use
ase-deepmd when you need to run MD simulations with a trained DeePMD potential.