Synthesis Recommendation

Goal

To provide experimentally validated synthesis routes for target inorganic materials by querying Materials Project's text-mined database of synthesis recipes extracted from scientific literature. This skill returns precursor materials, synthesis procedures, reaction equations, and DOI references to published papers.

Instructions

1. Query Synthesis Recipes

Search for synthesis recipes for a target material using the Materials Project API:

# Env: base-agent
python .agents/skills/mat-synthesis-recommendation/scripts/recommend_synthesis.py "LiFePO4" --limit 10 --output synthesis_recipes.json

Parameters:

• formula: Target material formula (e.g., "LiFePO4", "Li2CO3", "NMC811")
• --limit: Maximum number of recipes to display (default: 10)
• --output: Optional JSON file to save results
• --type: Filter by synthesis type (e.g., "solid-state", "hydrothermal", "sol-gel")
• --min-temp: Minimum synthesis temperature in °C
• --max-temp: Maximum synthesis temperature in °C

Output: Recipes are automatically ranked by:

1. Simplicity: Fewer precursors preferred
2. Temperature: Lower synthesis temperatures preferred
3. Synthesis type: Common methods (solid-state, hydrothermal) ranked higher

2. Interpret Results

Each recipe contains:

• Target material: Normalized chemical formula
• Precursors: Starting materials/reagents
• Synthesis type: Method category (solid-state, hydrothermal, sol-gel, etc.)
• Procedure: Step-by-step synthesis description from the paper
• Reaction equation: Balanced chemical equation (when available)
• DOI: Link to the source publication for full experimental details

3. Validate Synthesis Feasibility (Optional)

Cross-check the recommended precursors with other skills:

# Check if target material is thermodynamically stable
# See: ../mat-stability/SKILL.md
python .agents/skills/mat-stability/scripts/calculate_stability.py target.cif --output stability_analysis.json

# Calculate formation energy to verify synthesizability
# Energy above hull (E_hull) < 0.1 eV/atom indicates likely synthesizability

Examples

Example 1: Basic Query for LiFePO4

# Env: base-agent
python .agents/skills/mat-synthesis-recommendation/scripts/recommend_synthesis.py "LiFePO4" --limit 5

Expected output:

• 5 synthesis recipes ranked by simplicity
• Common precursors: Li₂CO₃, FeC₂O₄, NH₄H₂PO₄
• Typical methods: solid-state reaction, hydrothermal synthesis
• DOI links to papers in J. Electrochem. Soc., Chem. Mater., etc.

Example 2: Filter by Synthesis Type

# Env: base-agent
# Query only hydrothermal synthesis routes for LiCoO2
python .agents/skills/mat-synthesis-recommendation/scripts/recommend_synthesis.py "LiCoO2" --type hydrothermal --limit 10 --output LiCoO2_hydrothermal.json

Example 3: Temperature-Constrained Search

# Env: base-agent
# Find low-temperature synthesis routes (< 600°C) for Li2CO3
python .agents/skills/mat-synthesis-recommendation/scripts/recommend_synthesis.py "Li2CO3" --max-temp 600 --limit 10

Constraints

• API Key Required: Requires Materials Project API key via MP_API_KEY environment variable or ~/.atomistic_skills.yaml configuration
• Conda Environment: This skill requires the base-agent environment (includes mp-api, pymatgen)
• Coverage Limitations: Not all materials have synthesis recipes in the database
  • Database contains ~55,000 recipes for common inorganic materials
  • Coverage is best for battery materials, ceramics, and metal oxides
  • Organic materials and MOFs have limited coverage
• Text-Mining Accuracy: Recipes are automatically extracted from literature using NLP
  • Precursors and procedures are generally accurate but should be verified against the source DOI
  • Temperature values may be missing or incomplete in some entries
• Data Freshness: The text-mined database is periodically updated but may not include the most recent publications

Data Source

The synthesis recipes are extracted from scientific literature using natural language processing by the Materials Project team. The underlying datasets include:

• Solid-state synthesis: 19,488+ recipes from the CederGroup text-mined database
• Solution-based synthesis: 35,675+ recipes for solution, hydrothermal, and sol-gel methods
• NLP pipeline: Transformer-based models for paragraph classification, named entity recognition, and synthesis action extraction

References:

• Materials Project Synthesis Explorer: https://materialsproject.org/synthesis
• Text-mined synthesis datasets: CederGroup GitHub
• Kim et al., "A Database of Synthesis Recipes for Inorganic Materials," Sci. Data (2017)

Related Skills

• mat-stability: Verify thermodynamic stability before attempting synthesis
• mat-intercalation-voltage: Calculate electrochemical properties of synthesized cathode materials

---

Author: Bowen Deng Contact: GitHub @learningmatter-mit