Scholarlink

<p>This research introduces a sophisticated computational methodology that combines DFT, SCAPS-1D simulations, and machine learning to enhance the development of lead-free Sr<sub>3</sub>BiBr<sub>3</sub> perovskite solar cells (PSCs). DFT simulations indicate that Sr<sub>3</sub>BiBr<sub>3</sub> possesses a direct bandgap of 1.44&#xa0;eV, elevated absorption coefficients, and remarkable stability, making it an excellent choice for solar energy applications. SCAPS-1D simulations were utilized to evaluate device performance by examining different electron transport layers (ETLs), including WS<sub>2</sub>, C<sub>60</sub>, SnS<sub>2</sub>, and IGZO, as well as hole transport layers (HTLs) such as CuI, CFTS, and Cu<sub>2</sub>O. Among the configurations evaluated, the pairing of WS<sub>2</sub> as ETL and Cu<sub>2</sub>O as HTL attained the maximum power conversion efficiency (PCE) of 30.18%, while the configurations utilizing CuI and CFTS exhibited PCEs of 27.44% and 23.52%, respectively. Additionally, three machine learning models, Random Forest (RF), Gradient Boosting (GB), and Decision Tree Regressor (DTR), were used to forecast the optical performance of PSCs based on 10,989 SCAPS-1D simulated datasets. The models were trained on 80% and tested on 20% of critical PSC parameters, with prediction accuracy evaluated using error measures such as RMSE, MSE, MAPE, and <i>R</i><sup>2</sup>. Of the three, RF attained the highest accuracy (RMSE = 0.0779, <i>R</i><sup>2</sup> = 0.9973), surpassing both GB and DTR. SHAP analysis indicated that defect density, interface defects, and acceptor density were the predominant factors affecting PCE. The RF model exhibited significant predictive accuracy, great generalization, and efficient feature importance assessment, establishing it as the most dependable approach for projecting PSC efficiency.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advanced design and optoelectronic evaluation of Sr3BiBr3-based perovskite solar cells: insights into transport layers via simulation and machine learning

  • 期刊论文

  • 发表日期：

    2025年12月6日

查看全文
  • Bipul Chandra Biswas,
  • Asadul Islam Shimul,
  • Indrojit Paul,
  • Karim Kriaa,
  • Mohamed Benghanem,
  • S. AlFaify,
  • Md. Azizur Rahman,
  • Noureddine Elboughdiri

摘要

This research introduces a sophisticated computational methodology that combines DFT, SCAPS-1D simulations, and machine learning to enhance the development of lead-free Sr3BiBr3 perovskite solar cells (PSCs). DFT simulations indicate that Sr3BiBr3 possesses a direct bandgap of 1.44 eV, elevated absorption coefficients, and remarkable stability, making it an excellent choice for solar energy applications. SCAPS-1D simulations were utilized to evaluate device performance by examining different electron transport layers (ETLs), including WS2, C60, SnS2, and IGZO, as well as hole transport layers (HTLs) such as CuI, CFTS, and Cu2O. Among the configurations evaluated, the pairing of WS2 as ETL and Cu2O as HTL attained the maximum power conversion efficiency (PCE) of 30.18%, while the configurations utilizing CuI and CFTS exhibited PCEs of 27.44% and 23.52%, respectively. Additionally, three machine learning models, Random Forest (RF), Gradient Boosting (GB), and Decision Tree Regressor (DTR), were used to forecast the optical performance of PSCs based on 10,989 SCAPS-1D simulated datasets. The models were trained on 80% and tested on 20% of critical PSC parameters, with prediction accuracy evaluated using error measures such as RMSE, MSE, MAPE, and R2. Of the three, RF attained the highest accuracy (RMSE = 0.0779, R2 = 0.9973), surpassing both GB and DTR. SHAP analysis indicated that defect density, interface defects, and acceptor density were the predominant factors affecting PCE. The RF model exhibited significant predictive accuracy, great generalization, and efficient feature importance assessment, establishing it as the most dependable approach for projecting PSC efficiency.