JARVIS-Tools

Official Resources

• Homepage: https://pages.nist.gov/jarvis/
• GitHub: https://github.com/usnistgov/jarvis
• Documentation: https://jarvis-tools.readthedocs.io/
• PyPI: https://pypi.org/project/jarvis-tools/
• Web Portal: https://jarvis.nist.gov/
• License: NIST Software License

Overview

JARVIS-Tools (Joint Automated Repository for Various Integrated Simulations) is a comprehensive open-source Python package for atomistic data-driven materials design developed at NIST. It provides tools for setting up calculations, post-processing, analysis, visualization, and machine learning across multiple simulation codes. The package is tightly integrated with the JARVIS databases containing >100,000 materials.

Scientific domain: High-throughput materials science, DFT post-processing, machine learning, database-driven research Target user community: Materials scientists, computational researchers, ML practitioners in materials science

Theoretical Background

JARVIS-Tools interfaces with multiple levels of theory:

• DFT calculations (VASP, QE, Wien2k)
• Tight-binding (Wannier90, WTBH)
• Classical MD (LAMMPS)
• Machine learning potentials
• Property predictions from descriptors

Capabilities (CRITICAL)

• Workflow Automation: VASP, QE, Wien2k, WTBH, Wannier90, LAMMPS
• Band Structure: Electronic band analysis and plotting
• DOS/PDOS: Density of states processing
• Database Access: JARVIS-DFT (>75,000 materials), JARVIS-FF
• Machine Learning: ALIGNN, CGCNN, descriptors
• Structure Analysis: Symmetry, defects, surfaces
• Property Prediction: Formation energy, band gap, elastic constants
• Visualization: Structure and property plotting

Key Strengths

Multi-Code Support:

• VASP (comprehensive)
• Quantum ESPRESSO
• Wien2k
• BoltzTraP
• Wannier90
• LAMMPS
• GPAW
• WTBH (Wannier tight-binding)

Database Integration:

• JARVIS-DFT: 75,000+ 3D materials
• JARVIS-2D: 2D materials database
• JARVIS-FF: Force field database
• JARVIS-ML: ML model repository
• REST API access

Machine Learning:

• ALIGNN (graph neural network)
• CGCNN integration
• Descriptor generation
• Pre-trained models
• Property prediction

High-Throughput:

• Automated workflow generation
• Error handling
• Job management
• Result parsing

Inputs & Outputs

• Input formats:

  • POSCAR, CIF, XYZ structures
  • VASP input/output files
  • QE input/output files
  • Database queries (materials ID)

• Output data types:

  • Band structures, DOS
  • Elastic tensors
  • Optical properties
  • ML predictions
  • JSON/CSV exports

Interfaces & Ecosystem

• Python integration:

  • NumPy, Pandas for data
  • Matplotlib for visualization
  • PyTorch for ML models
  • ASE compatibility

• Framework compatibility:

  • Jupyter notebooks
  • REST API
  • Command-line tools
  • Web interface

Installation

pip install jarvis-tools

With ML dependencies:

pip install jarvis-tools[ai]

Usage Examples

from jarvis.core.atoms import Atoms
from jarvis.db.figshare import data
from jarvis.tasks.vasp.vasp import VaspJob

# Load structure from database
dft_3d = data("dft_3d")
atoms = Atoms.from_dict(dft_3d[0]["atoms"])

# Get band structure
from jarvis.io.vasp.outputs import Vasprun
vrun = Vasprun("vasprun.xml")
bands = vrun.get_bandstructure()

# ML prediction with ALIGNN
from jarvis.core.atoms import Atoms
from alignn.pretrained import get_figshare_model
model = get_figshare_model("jv_formation_energy_peratom_alignn")
prediction = model.predict(atoms)

Performance Characteristics

• Speed: Efficient file parsing
• Scalability: High-throughput ready
• ML: GPU-accelerated models
• Database: Fast API queries

Limitations & Known Constraints

• Learning curve: Many features require exploration
• Documentation: Extensive but distributed
• VASP-centric: Best support for VASP
• Dependencies: ML features need PyTorch

Comparison with Other Tools

• vs pymatgen: JARVIS has ML focus, database integration
• vs ASE: JARVIS specialized for high-throughput, ML
• vs atomate: Different workflow philosophy
• Unique strength: NIST databases, ALIGNN ML, comprehensive ecosystem

Application Areas

• High-throughput materials screening
• Machine learning for materials
• Database-driven discovery
• Property prediction
• Workflow automation
• 2D materials research
• Defect calculations

Best Practices

• Use database for initial screening
• Leverage pre-trained ML models
• Validate ML predictions with DFT
• Use workflows for systematic studies
• Cite JARVIS papers appropriately

Community and Support

• GitHub issue tracker
• NIST support
• YouTube tutorials
• Active development
• Regular database updates

Verification & Sources

Primary sources:

1. Official website: https://pages.nist.gov/jarvis/
2. GitHub: https://github.com/usnistgov/jarvis
3. K. Choudhary et al., npj Comput. Mater. 6, 173 (2020)
4. JARVIS-DFT database papers

Confidence: VERIFIED - NIST-developed, peer-reviewed publications

Verification status: ✅ VERIFIED

• Official homepage: ACCESSIBLE
• Documentation: COMPREHENSIVE
• Source code: OPEN (GitHub)
• Developer: NIST (K. Choudhary et al.)
• Academic citations: >500 citations
• Active development: Regular releases, database updates