Overview
CrySPR (Crystal Structure Pre-Relaxation and PRediction) is a Python interface for implementing crystal structure pre-relaxation and prediction using machine-learning interatomic potentials (ML-IAPs).
Theoretical Basis
- Structure generation (PyXtal)
- ML interatomic potentials (M3GNet, CHGNet, MACE)
- Structure pre-relaxation
- Random search / PSO
- Energy ranking
Key Capabilities
- ML-IAP pre-relaxation
- PyXtal structure generation
- Multiple ML potentials supported
- Random search and PSO
- Fast screening
Sources: ChemRxiv (2024)
Key Strengths
ML Integration:
- M3GNet, CHGNet, MACE
- Fast relaxation
- DFT-level accuracy
Workflow:
- Automated pipeline
- PyXtal integration
- Multiple search methods
Flexibility:
- Multiple ML-IAPs
- Customizable workflow
- Python-based
Inputs & Outputs
- Input formats: Chemical composition, space group (optional)
- Output data types: Relaxed structures, energies
Interfaces & Ecosystem
- Generation: PyXtal
- ML Potentials: M3GNet, CHGNet, MACE
- ASE: Calculator interface
Workflow and Usage
- Define composition
- Generate structures (PyXtal)
- Pre-relax with ML-IAP
- Run global search (RS/PSO)
- Rank and validate
Performance Characteristics
- Fast ML relaxation
- Efficient screening
- Good pre-relaxation
Computational Cost
- ML relaxation: fast
- Search: moderate
- DFT validation: expensive
Best Practices
- Choose appropriate ML-IAP
- Use pre-relaxation
- Validate top structures
- Compare multiple potentials
Limitations & Known Constraints
- ML-IAP accuracy limits
- Training data dependent
- Requires validation
Application Areas
- Crystal structure prediction
- Pre-relaxation screening
- Materials discovery
- ML-accelerated CSP
Comparison with Other Codes
- vs PyMCSP: Similar purpose
- vs CrySPY: CrySPR pre-relaxation focused
- Unique strength: ML-IAP pre-relaxation interface
Community and Support
- Open-source (GitHub)
- Recent development
- ChemRxiv preprint
Verification & Sources
Primary sources:
- GitHub: https://github.com/Tosykie/CrySPR
- ChemRxiv: https://chemrxiv.org/engage/chemrxiv/article-details/66b308a501103d79c5fd9b91
Confidence: VERIFIED
Verification status: ✅ VERIFIED
- Website: ACTIVE (GitHub)
- Source: OPEN
- Development: RECENT
- Applications: ML-accelerated CSP