Genarris¶
A scalable, MPI-parallel Crystal Structure Prediction workflow for organic molecular crystals.
Genarris (gnrs) generates random molecular crystal structures and drives
CSP workflows from generation through energy evaluation, clustering,
and selection.
🏛️ Developed by the Noa Marom Group at Carnegie Mellon University.
📦 Installation
Set up Genarris with pip in minutes.
🚀 Quick Start
Run your first CSP workflow step by step.
📖 API Reference
Full module and class documentation.
💎 Case Studies
CSP results and applications.
Key Features¶
Generate random molecular crystal structures across all 230 space groups with configurable packing parameters and volume estimation.
Improve packing acceptance rates with symmetry-preserving rigid-press geometry optimization.
Evaluate energies with state-of-the-art MLIPs including UMA, MACE-OFF, and AIMNet2 with GPU acceleration.
Cluster structures using Affinity Propagation or K-Means with ACSF descriptors. Select representatives via center or energy-window strategies.
Scale across hundreds of cores with MPI-parallel execution. GPU worker/feeder pattern for efficient resource utilization.
Configure multi-step CSP pipelines via simple INI files. Extensible architecture with abstract base classes for custom implementations.
Supported Energy Calculators¶
🔧 Calculator |
🏷️ Type |
🖥️ GPU |
📝 Description |
|---|---|---|---|
MLIP |
✅ |
Universal Model for Atoms from Meta FAIR |
|
MLIP |
✅ |
MACE-OFF organic molecular crystals model |
|
MLIP |
✅ |
AIMNet2 neural network potential |
|
Semi-Empirical |
— |
Density Functional Tight Binding |
|
DFT |
— |
All-electron DFT code |
|
DFT |
— |
Plane-wave DFT code |
Citation¶
Please cite
Yang, Y., Tom, R., Wui, J. A., Moussa, J. E., & Marom, N. Genarris 3.0: Generating Close-Packed Molecular Crystal Structures with Rigid Press. Journal of Chemical Theory and Computation, 21, 11318–11332 (2025).
See the full Citation page for all related papers.