Skip to content

Crystalyse Documentation

Welcome to Crystalyse - a computational materials design platform that accelerates materials exploration through AI-powered analysis and validation.

Overview

Crystalyse is a computational materials design platform that combines large language models with rigorous computational chemistry tools. Built on the OpenAI Agents framework and integrated with advanced materials science tools via the Model Context Protocol (MCP), it provides researchers with a dual-mode system for exploring chemical space and analysing materials properties.

Key Features: Crystalyse bridges the gap between AI creativity and scientific rigour, enabling researchers to go from materials concepts to validated computational analysis in under 2 minutes, significantly accelerating traditional design workflows.

Core Capabilities

Dual-Mode Analysis System

  • Creative Mode: Fast exploration using Chemeleon crystal structure prediction and MACE energy calculations (~50 seconds)
  • Rigorous Mode: Complete validation pipeline with SMACT screening, Chemeleon structures, MACE energies, and comprehensive analysis (2-5 minutes)

Materials Analysis Pipeline

  1. Query Processing: Natural language materials requirements and specifications
  2. Composition Analysis: SMACT-validated chemical compositions and feasibility screening
  3. Structure Generation: Chemeleon crystal structure prediction with multiple candidates
  4. Energy Evaluation: MACE force field calculations for formation energies and stability
  5. Visualisation: Interactive 3D structures and comprehensive analysis plots

Dual Interface Options

  • Unified CLI: Single-command interface with /mode and /agent switching capabilities
  • Command-line Tools: Direct access via crystalyse discover, crystalyse chat
  • Session Management: Persistent conversation history and context across multi-day projects
  • Interactive Mode: Real-time mode switching between creative and rigorous analysis

Documentation Structure

Getting Started

Core Concepts

Chemistry Tools

How-To Guides

API Reference

Key Features

Advanced Materials Design

  • Significant Speed: From 6-18 months to 2-5 minutes per material design
  • Dual Validation: AI creativity + computational rigor
  • Complete Pipeline: Composition → Structure → Energy → Recommendations
  • High Accuracy: 89.8/100 capability score with rigorous validation

Advanced AI Integration

  • OpenAI Agents Framework: Production-ready agent architecture
  • o4-mini Support: Ultra-high rate limits (10M TPM, 1B TPD) for creative mode
  • o3/gpt-4o: Balanced performance for rigorous validation
  • Anti-Hallucination: 100% computational honesty with response validation

Professional Tool Integration

  • SMACT Validation: Semiconducting Materials from Analogy and Chemical Theory
  • Chemeleon CSP: State-of-the-art crystal structure prediction
  • MACE Energy: Machine learning force fields for energy calculations
  • MCP Protocol: Seamless tool integration with persistent connections

Research-Grade Features

  • Session Persistence: SQLite-like conversation management
  • Memory Systems: Discovery caching and pattern recognition
  • Interactive CLI: Rich terminal interface with progress tracking
  • Cross-Platform: Windows, macOS, Linux support

Applications

Energy Materials

  • Battery cathodes and anodes (Li-ion, Na-ion, solid-state)
  • Solid electrolytes and ion conductors
  • Photovoltaic semiconductors and perovskites
  • Thermoelectric materials

Electronic Materials

  • Ferroelectric and multiferroic materials
  • Magnetic materials and spintronics
  • Semiconductor devices and memory materials
  • Superconductors and quantum materials

Catalysis and Environment

  • COâ‚‚ reduction catalysts
  • Water splitting and hydrogen production
  • Chemical synthesis catalysts
  • Environmental remediation materials

Structural Materials

  • High-entropy alloys
  • Advanced ceramics and composites
  • Lightweight structural materials
  • Wear-resistant coatings

Scientific Integrity

Crystalyse maintains the highest standards of computational honesty:

  • 100% Traceability: Every numerical result traces to actual tool calls
  • Zero Fabrication: No estimated or fabricated energies, structures, or properties
  • Complete Transparency: Clear distinction between AI reasoning and computational validation
  • Validation Pipeline: Response validation system prevents hallucinations

Performance Characteristics

Execution Times

Operation Creative Mode Rigorous Mode
Simple Query ~80 seconds 2-5 minutes
Complex Discovery 2-3 minutes 5-10 minutes
Batch Analysis 5-10 minutes 15-30 minutes

Validation Accuracy

  • SMACT Validation: >95% agreement with experimental feasibility
  • Structure Prediction: High-quality crystal structures with CIF output
  • Energy Calculations: ML force field accuracy with uncertainty quantification
  • Discovery Pipeline: End-to-end validation from composition to properties

Prerequisites

  • Python 3.11 or higher
  • OpenAI API key (preferably OpenAI MDG for high rate limits)
  • 8GB RAM recommended (4GB minimum)
  • Internet connection for tool downloads and API calls

Next Steps

  1. Follow the Quickstart Guide to begin using Crystalyse
  2. Read the CLI Usage Guide to master the interface
  3. Explore Analysis Modes to understand the discovery workflow
  4. Check the SMACT Integration documentation for detailed capabilities
  5. Review the Python API for programmatic usage

Support and Community

Crystalyse is actively developed and welcomes community engagement: - Issues: Report bugs and request features on GitHub - Documentation: Comprehensive guides and API reference - Examples: Practical usage examples and tutorials - Updates: Regular improvements and new features

Acknowledgments

Crystalyse builds upon exceptional open-source tools: - SMACT: Semiconducting Materials from Analogy and Chemical Theory - Chemeleon: Crystal structure prediction and analysis - MACE: Machine learning ACE force fields - OpenAI Agents SDK: Production-ready agent framework - Model Context Protocol: Seamless tool integration

Ready to accelerate your materials design? Start with the Quickstart Guide to begin using computational materials science tools.