Design strategies for heterojunction silicon/CsSnBr3 lead-free tandem solar cells using machine learning and SCAPS-1D
Velpuri Leeladevi, Piyush Kuchhal, Debasis De, Neeraj Kumar Shukla, Piyush Dua
Abstract
The integration of machine learning with SCAPS-1D simulations has led to significant advancements in optimizing tandem solar cells, particularly in perovskite/silicon architectures. This study focuses on enhancing CsSnBr 3 -based perovskite solar cells as the top cell in a tandem configuration, aiming to achieve efficient current matching and maximize power conversion efficiency. A dataset of 29,831 unique device configurations was generated, analyzing critical parameters such as bandgap alignment, dielectric permittivity, doping density, and charge transport properties using 10 different machine learning models. The optimized structure FTO/WS 2 /CsSnBr 3 /Cu 2 O/ITO/a-Si(n)/a-Si(i)/c-Si(p)/a-Si(i)/a-Si(p) demonstrated optimal short circuit current matching at 21.29 mA/cm², and open circuit voltage of 2.06 V achieving an impressive tandem solar cell efficiency of 36.44 %. This study also investigates how doping density impacts device performance, focusing on the delicate balance between carrier transport and recombination. The results underscore the importance of carrier behavior and interface optimization in maximizing the efficiency of perovskite/silicon tandem solar cells. By employing a data-driven approach, the research presents strategies for improving solar cell performance beyond 30 % efficiency, contributing to the development of affordable, scalable photovoltaic technologies for advancing future renewable energy solutions.