Maximizing performance in Cs2CuBiCl6 perovskite cells through machine learning-driven absorber layer parameter analysis

Nikhil Shrivastav,Jaya Madan,Rahul Pandey
DOI: https://doi.org/10.1016/j.matlet.2024.135929
IF: 3
2024-01-18
Materials Letters
Abstract:Conventional perovskite materials offer reasonable power conversion efficiency (PCE) but are plagued by stability issues. To address this, researchers have explored alternative materials, such as MA and FA-based lead or lead-free perovskites, known for their improved PCE and stability. In this study, we focused on a cesium (Cs)-based cell, Cs2CuBiCl6, which not only exhibits acceptable PCE but also demonstrates superior stability compared to traditional perovskite materials. To optimize the Cs2CuBiCl6-based cell, we conducted simulations varying parameters like thickness, bulk defect density (BDD), and doping. The optimized PCE for the Cs2CuBiCl6-based cell reached 10.30 %. We generated 1000 datasets through simulations, which serve as input for machine learning (ML) algorithms, including random forest (RF) and XGboost (XGB). The XGB model outperformed the RF model, achieving a higher R-squared value (R 2 : 99.97 %) and a lower mean squared error (MSE: 0.0014).
materials science, multidisciplinary,physics, applied
What problem does this paper attempt to address?