Optimizing charge transport and bandgap alignment of ETL and HTL in Tin-based perovskite solar cells via machine learning
Perovskite solar cells (PSCs) are highly promising alternatives to silicon solar cells. They offer exceptional costeffectiveness and achieve high power conversion efficiency (PCE) for lead-based PSCs. However, the toxicity is a major concern for lead-based PSCs. As a result, tin-based PSCs are explo...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/44856/ http://umpir.ump.edu.my/id/eprint/44856/1/Optimizing%20charge%20transport%20and%20bandgap%20alignment%20of%20ETL%20and%20HTL.pdf |
| Summary: | Perovskite solar cells (PSCs) are highly promising alternatives to silicon solar cells. They offer exceptional costeffectiveness and achieve high power conversion efficiency (PCE) for lead-based PSCs. However, the toxicity is a major concern for lead-based PSCs. As a result, tin-based PSCs are explored as viable alternatives for lead-based PSCs. Despite this, the PCE of the tin-based PSCs is still very far from the lead-based PCE. Proper alignment of the absorber material with the electron transport layer (ETL) and hole transport layer (HTL) band gap is crucial to achieve higher PCE. In this study, 8 different ETLs and 8 different HTLs, a total of 64 combinations, were used to analyze the highest-performing structure. Four different machine-learning models were utilized in this study, such as K-Nearest Neighbours (KNN), Logistic Regression (LR), Artificial Neural Network (ANN), and Support Vector Regressor (SVR). To train the ML models, 900 data points were generated by adjusting different parameters such as absorber layer thickness, HTL thickness, defect densities of the absorber layer, and ETL defect density. The Fluorine-doped Tin Oxide (FTO)/Zinc oxide (ZnO)/Formamidinium tin triiodide (FASnIā)/Copper oxide (Cu2O) PSC structure achieved the highest PCE of 26.93 %. Among other ML models, the ANN performed well and maintained an accuracy parameter, Coefficient of Determination (R2 ), of 0.98. The study simplifies designing perfectly aligned Tin-based PSC for better photovoltaic performance. |
|---|