SemiGenie

An intelligent AI agent for semiconductor material discovery and design. SemiGenie uses advanced AI techniques to generate, evaluate, and refine semiconductor candidates based on your specifications.

Features

• Intelligent Candidate Generation: AI-powered generation of semiconductor material candidates
• Flexible Evaluation: Multiple evaluation methods including literature research and DFT integration
• Adaptive Learning: Automatically improves candidate selection through iterative refinement
• Customizable Targets: Configure target properties and success criteria to match your needs

Installation

1. Clone or download this repository

2. Install dependencies:

pip install -r requirements.txt

3. Set up your OpenAI API key:

export OPENAI_API_KEY="your-api-key-here"

Or create a .env file:

OPENAI_API_KEY=your-api-key-here

Usage

Basic Usage

Run the agent with default settings:

python semiconductor_agent.py

Custom Settings

Specify custom parameters:

python semiconductor_agent.py \
    --target-bandgap 2.0 \
    --target-hit-rate 0.85 \
    --max-iterations 15 \
    --candidates 50 \
    --method chatgpt_deep_research

Parameters

• --target-bandgap: Target bandgap in eV
• --target-hit-rate: Target success rate (0.0-1.0)
• --max-iterations: Maximum number of iterations
• --candidates: Number of candidates per iteration
• --method: Evaluation method: chatgpt_deep_research or dft
• --api-key: OpenAI API key (or set OPENAI_API_KEY env var)

Programmatic Usage

from semiconductor_agent import SemiconductorAgent

# Create agent
agent = SemiconductorAgent(
    target_bandgap=1.5,
    target_hit_rate=0.9,
    max_iterations=10,
    candidates_per_iteration=100,
    evaluation_method="chatgpt_deep_research"
)

# Run agent
results = agent.run()

# Access results
print(f"Final hit rate: {results['final_hit_rate']:.1%}")
print(f"Best candidates: {results['best_candidates']}")

Overview

SemiGenie employs an iterative approach to semiconductor discovery:

1. Generation: The agent generates candidate materials based on your specifications
2. Evaluation: Each candidate is evaluated against target properties
3. Refinement: The system learns from results and improves subsequent iterations
4. Results: Final candidates are ranked and saved for further analysis

Evaluation Methods

Literature Research (Default)

• Fast evaluation using AI-powered literature analysis
• No computational resources required
• Suitable for initial screening and exploration

DFT Integration

For computational evaluation, SemiGenie supports integration with:

1. Materials Project API:

pip install mp-api
export MATERIALS_PROJECT_API_KEY="your-api-key"

2. Quantum ESPRESSO: Requires installation and setup

3. VASP: Requires license and setup

Output

The agent provides:

• Real-time progress updates
• Iteration-by-iteration performance metrics
• Ranked list of best candidates
• Comprehensive results saved to JSON

Configuration

You can customize the agent using environment variables or the config.py file:

# config.py
target_bandgap = 1.5
target_hit_rate = 0.9
max_iterations = 10
candidates_per_iteration = 100
evaluation_method = "chatgpt_deep_research"

Requirements

• Python 3.8+
• OpenAI API key
• Internet connection (for API calls)

Optional:

• Materials Project API key (for Materials Project evaluation)
• Quantum ESPRESSO or VASP (for DFT calculations)

Notes

• Evaluation accuracy depends on available data and computational resources
• Results improve with more iterations as the system learns
• Some materials may have limited available data

License

MIT License

Contributing

Contributions welcome! Please feel free to submit a Pull Request.