Bayesian inference of material parameters in organic solar cells

This repository contains code for Bayesian inference of material parameters from JV curves in organic solar cells.

Note:
Due to licensing issues for the drift-diffusion simulator, the corresponding code is omitted.

For details, see the main paper https://doi.org/10.48550/arXiv.2506.13308: https://doi.org/10.1002/solr.202500648:

---

📁 Example Data

Large example datasets (>1GB) used in this project are hosted on Zenodo due to GitHub size limits.

Download data:
https://zenodo.org/record/15480770

Description:

• exp_data_example_d171nm_4illus_0p02-0p84V.h5: Raw illumination-dependent JV curves (active blend - T1:BTP-4F-12)
• example_d171nm_8_param_train_test.h5: Train and test dataset for building the NN model
• example_d171nm_8_paramscaler.joblib: Standardization tool for NN model input (material parameters)
• y1/example_d171nm_8_param_y1_trained_model.h5: Trained NN model for shifted JV curves
• y2/example_d171nm_8_param_y2_trained_model.h5: Trained NN model for short-circuit current densities

How to use:

1. Download the required files from Zenodo.
2. Extract them into the repository folder.
3. Use the pdf_analysis_example.ipynb notebook to infer parameters from experimental or NN-generated data and visualize the results.

---

Requirements

• Python 3.8+
• numpy, matplotlib, tensorflow, pymoo, ...

Install all dependencies using:

pip install -r requirements.txt