Utilizing Machine Learning and Diode Physics to Investigate the Effects of Stoichiometry on Photovoltaic Performance...

Area of Science:

• Materials Science
• Renewable Energy
• Photovoltaics

Background:

• Organic-inorganic metal halide perovskite solar cells offer excellent solution processability.
• Achieving uniformly crystalline perovskite films often requires complex deposition techniques.
• Optimizing fabrication conditions is crucial for enhancing photovoltaic performance.

Purpose of the Study:

• To develop a streamlined method for optimizing perovskite solar cell performance.
• To elucidate the relationship between experimental variables and device characteristics.
• To reduce the reliance on extensive trial-and-error in device fabrication.

Main Methods:

• Harmonized Shockley diode-based numerical analysis with machine learning (Gaussian process regression).
• Extracted photovoltaic parameters and predicted power conversion efficiencies using the Shockley diode equation.
• Trained machine learning models on current-voltage curves to identify optimal fabrication settings.

Main Results:

• Successfully extracted key photovoltaic parameters and predicted device efficiencies.
• Identified optimal experimental conditions for enhanced perovskite solar cell performance.
• Demonstrated a significant reduction in the need for exhaustive experimental testing.

Conclusions:

• The combined approach of numerical analysis and machine learning effectively optimizes perovskite solar cells.
• This methodology accelerates the development and fine-tuning of next-generation solar cell technologies.
• Provides a deeper understanding of device physics and charge recombination mechanisms.