Literatures guide reaction-specific filtering of compositionally complex electrocatalysts

This repository reproduces the analysis figures and tables for the manuscript.

The trained Word2Vec model and material-system CSV inputs are already included in the expected folders, so the normal run starts from the downstream analysis.

1. Create The Environment

conda env create -f environment.yml
conda activate env_pareto_front

2. Check The Inputs

python scripts/check_repo.py

This verifies that the model, code, and material-system CSV files are present.

Required input locations:

01_word2vec_model/model/multi_source_papers_electrocatalyst.model
02_pareto_prediction/material_systems/MinDMaxC/
02_pareto_prediction/material_systems/MaxDMinC/
02_pareto_prediction/material_systems/Ni_Pd_Pt_Ru_subset/

3. Reproduce Figures And Tables

bash scripts/run_reproduction.sh

The script runs all workflow stages in order. You do not need to enter the numbered subdirectories manually.

4. Find The Outputs

All user-facing outputs are collected in:

results/

Main outputs:

results/tables/final_processed/
results/tables/unfiltered_reference/
results/tables/main_figure_checks/
results/figures/main/
results/figures/supplementary_plots/
results/figures/filter_diagnostics/

The final processed tables are the main numerical results:

results/tables/final_processed/current_density_statistics.csv
results/tables/final_processed/element_statistics.csv
results/tables/final_processed/iterative_isolated_edge_cluster_proof_summary.csv

License

See LICENSE.