| name | high-throughput |
| description | High-Throughput Computational Workflows (8 sub-skills: batch-calculations, batch-screening, convergence-automation, materials-filtering, matpes-dual-static, phase-stability, property-prediction, screening-workflow) |
High-Throughput Computational Workflows
Sub-skills
| Skill | Path | Description |
|---|
| Screening Workflow | screening-workflow/SKILL.md | End-to-end materials screening: query Materials Project for candidates, batch MACE relaxation, filter by stability/band gap/elastic modulus, rank candidates, validate top hits with DFT |
| Batch Calculations | batch-calculations/SKILL.md | Run batch DFT calculations on multiple structures: input generation with pymatgen sets, parallel execution, result collection, error handling and restart logic for QE and VASP |
| Phase Stability | phase-stability/SKILL.md | Phase stability and convex hull analysis: compute formation energies, build convex hulls, determine energy above hull, phase diagram sections, integration with Materials Project reference data |
| Property Prediction | property-prediction/SKILL.md | Rapid MACE-based property prediction pipeline: lattice constants, bulk modulus, DOS-based band gap proxy, surface energy, vacancy formation energy, comparison with databases, correlation plots |
| Batch Screening | batch-screening/SKILL.md | High-throughput screening workflow: fetch candidates from Materials Project, rapid MACE screening, filter/rank, detailed QE for top hits |
| Materials Filtering | materials-filtering/SKILL.md | Query and filter the Materials Project database: chemical system search, property ranges, stability criteria, export structures and data |
| Convergence Automation | convergence-automation/SKILL.md | Automated DFT convergence testing: sweep ecutwfc and k-grids, find optimal parameters via energy-per-atom threshold (1 meV/atom), combined cutoff-then-kgrid workflow, batch mode for multiple materials, convergence plots and JSON reports. Inspired by pyiron_atomistics ConvEncutParallel/ConvKpointParallel. |
| MatPES Dual Static | matpes-dual-static/SKILL.md | Dual-functional PBE + r2SCAN static calculations: run PBE first, reuse wavefunction for efficient r2SCAN, optional PBE+U, collect energies at two levels of theory for ML training data or cross-validation |
General Pattern
- Source candidates -- query Materials Project or build structures programmatically.
- Fast screen -- use MACE-MP-0 to relax and evaluate properties in seconds per structure.
- Filter and rank -- apply thermodynamic / electronic / mechanical criteria.
- Validate -- run Quantum ESPRESSO (or VASP) DFT on the short-listed candidates.
- Report -- export CSV/JSON, generate plots, build property tables.
Method Decision Guide
What kind of high-throughput task?
Screen many candidates for a target property?
--> screening-workflow/ (full pipeline with MP query + MACE + DFT validation)
Run DFT on a known set of structures?
--> batch-calculations/ (input generation, parallel execution, error handling)
Assess thermodynamic stability of candidate phases?
--> phase-stability/ (convex hull, formation energies, decomposition analysis)
Rapidly predict multiple properties for a set of structures?
--> property-prediction/ (MACE-based lattice constants, bulk modulus, surface energy, etc.)
Just query/filter the Materials Project database?
--> materials-filtering/ (database queries, property ranges, export)
Need both PBE and r2SCAN energies efficiently (ML training data, cross-validation)?
--> matpes-dual-static/ (PBE wavefunction warm-starts r2SCAN, optional PBE+U)
Determine optimal ecutwfc / k-grid for a new material or pseudopotential?
--> convergence-automation/ (automated sweeps, threshold-based selection, batch mode)
Prerequisites
MP_API_KEY environment variable set (obtain from https://next-gen.materialsproject.org/api).
- Python packages:
mp-api, pymatgen, ase, mace-torch, numpy, pandas, matplotlib.
- For DFT validation: Quantum ESPRESSO 7.5 (
pw.x). VASP support via external access in future.
- Optional:
pip install fireworks jobflow atomate2 custodian for advanced workflow management.