In this investigation, supervised machine learning (ML) was utilized to accurately predict the optimum bromine doping concentration in single-junction MASnI 3−x Br x devices. Data-driven optimizations were carried out on 42 000 unique devices built utilizing a solar cell capacitance simulator (SCAPS). The devices were investigated through variations of bromine doping %, bandgap, electron affinity, series resistance, back-contact metal, and acceptor concentrationparameters that were specifically chosen because of their tunable nature and ability to be modified through facile experimental fabrication techniques of the device. Five different algorithms were utilized to explore feature engineering. The first step before bromine doping within the device included validation studies of a pure tin-based system, MASnI 3 : a power conversion efficiency (PCE) of 6.71% was achieved, having close congruence with experimental data. ML analyses for optimal bromine doping resulted in the discovery of two devices with bromine concentrations of 22.43% (Br22) and 25.63% (Br25), with the latter being a more fine-tuned value obtained through extra rigorous analysis. To understand the total and relative impact of each feature on power conversion efficiency (PCE), Br22 and Br25 were analyzed with a state-of-the-art algorithm, namely, the SHapley Additive exPlanations (SHAP) algorithm. Focusing on the two discovered devices, further device optimizations were carried out utilizing SCAPS. Modulations of absorber thickness, bulk and interfacial defect density, and choice of electron transport layer (ETL) and hole transport layer (HTL) materials were tried. Device stability was analyzed through carrier lifetime studies. Following these optimization steps, Br22 and Br25 demonstrated final high PCE values of 20.72 and 17.37%, respectively. The ML-assisted quantitative analysis of the current work provides significant confidence for optimal bromine-doped tin-based devices to be considered as viable and competitive nontoxic alternatives to traditional technologies.
In this research,
solar cell capacitance simulator-one-dimensional
(SCAPS-1D) software was used to build and probe nontoxic Cs-based
perovskite solar devices and investigate modulations of key material
parameters on ultimate power conversion efficiency (PCE). The input
material parameters of the absorber Cs-perovskite layer were incrementally
changed, and with the various resulting combinations, 63,500 unique
devices were formed and probed to produce device PCE. Versatile and
well-established machine learning algorithms were thereafter utilized
to train, test, and evaluate the output dataset with a focused goal
to delineate and rank the input material parameters for their impact
on ultimate device performance and PCE. The most impactful parameters
were then tuned to showcase unique ranges that would ultimately lead
to higher device PCE values. As a validation step, the predicted results
were confirmed against SCAPS simulated results as well, highlighting
high accuracy and low error metrics. Further optimization of intrinsic
material parameters was conducted through modulation of absorber layer
thickness, back contact metal, and bulk defect concentration, resulting
in an improvement in the PCE of the device from 13.29 to 16.68%. Overall,
the results from this investigation provide much-needed insight and
guidance for researchers at large, and experimentalists in particular,
toward fabricating commercially viable nontoxic inorganic perovskite
alternatives for the burgeoning solar industry.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.