Minimal Input, Maximum Insight: Inferring Material Parameters From Solar Cell JV Curves Alone

Area of Science:

• Materials Science
• Device Physics
• Machine Learning

Background:

• Linking macroscopic device performance to microscopic properties is a major challenge in developing next-generation solar cells, sensors, and transistors.
• Current material development often relies on trial-and-error due to a lag in characterization techniques.
• New materials are proposed faster than their underlying properties are understood.

Purpose of the Study:

• To present two neural network-based methods for extracting key material parameters from optoelectronic devices.
• To enable rapid, low-cost characterization using minimal resources.
• To provide confidence assessment for predicted material parameters.

Main Methods:

• Developed two neural network-based methods requiring only a single light current-voltage (JV) curve measurement.
• Applied methods to organic solar cells (PM6:Y12 and PM6:BTP-eC9).
• Utilized a non-Gaussian likelihood distribution for parameter prediction, allowing confidence assessment.

Main Results:

• Successfully extracted key material parameters like charge carrier mobility and trap state density.
• Demonstrated the technique on freshly fabricated and degrading organic solar cells.
• Tracked the evolution of carrier lifetime, mobility, and shunt resistance in a degrading device.

Conclusions:

• The presented methods offer a rapid, low-cost approach for extracting essential material parameters from simple JV measurements.
• This technique can significantly accelerate the optimization of next-generation solar energy materials and devices.
• The approach is practical for laboratories with limited equipment.