Overview
GAtor is a massively parallel, first-principles genetic algorithm (GA) for molecular crystal structure prediction. Written in Python, it interfaces with the FHI-aims code for local optimizations and energy evaluations using dispersion-inclusive DFT.
Theoretical Basis
- Genetic algorithm optimization
- First-principles DFT energy evaluation
- Dispersion-inclusive functionals (vdW-DF, TS, MBD)
- Evolutionary niching with machine learning clustering
- Crossover and mutation operators for molecular crystals
Key Capabilities
- Molecular crystal structure prediction
- Massively parallel execution
- Machine learning-based niching
- Multiple fitness evaluation schemes
- Flexible breeding operators
Sources: J. Chem. Theory Comput. 14, 2246 (2018)
Key Strengths
Molecular Crystals:
- Designed for organic/molecular systems
- Handles conformational flexibility
- Proper treatment of dispersion
Parallelization:
- Massively parallel execution
- Efficient HPC utilization
- Scales to large populations
ML Integration:
- Clustering for evolutionary niching
- Diversity maintenance
- Efficient exploration
Inputs & Outputs
- Input formats: Molecular geometry, GA parameters
- Output data types: Crystal structures, energies, population history
Interfaces & Ecosystem
- DFT: FHI-aims (primary interface)
- Dispersion: vdW-DF, TS, MBD methods
- Analysis: Structure comparison, clustering
Workflow and Usage
- Prepare molecular geometry
- Configure GA parameters
- Set up FHI-aims interface
- Run parallel GA search
- Analyze converged structures
Performance Characteristics
- Excellent parallel scaling
- Efficient for molecular crystals
- ML-assisted diversity maintenance
Computational Cost
- DFT-limited (FHI-aims calculations)
- Parallelization reduces wall time
- Population size affects cost
Best Practices
- Use appropriate dispersion correction
- Tune population size for system
- Enable niching for diversity
- Validate with experimental data
Limitations & Known Constraints
- Requires FHI-aims license
- Computationally expensive (DFT)
- Focused on molecular crystals
Application Areas
- Organic crystal structure prediction
- Pharmaceutical polymorph screening
- Molecular materials design
- Crystal engineering
Comparison with Other Codes
- vs USPEX: GAtor molecular-focused, USPEX more general
- vs Genarris: GAtor includes optimization, Genarris generation only
- vs MGAC: Similar approach, different implementation
- Unique strength: First-principles GA, ML niching, molecular crystals
Community and Support
- Academic development (CMU)
- Published methodology
- Available upon request
Verification & Sources
Primary sources:
- Publication: Curtis et al., J. Chem. Theory Comput. 14, 2246 (2018)
- arXiv: https://arxiv.org/abs/1802.08602
Secondary sources:
- FHI-aims documentation
- Molecular crystal CSP literature
Confidence: VERIFIED
Verification status: ✅ VERIFIED
- Documentation: AVAILABLE (paper)
- Source: ACADEMIC (request)
- Development: ACTIVE
- Applications: Molecular crystal structure prediction