Using a Novel Transfer Learning Method for Designing Thin Film Solar Cells with Enhanced Quantum Efficiencies

Area of Science:

• Engineering
• Materials Science
• Computational Science

Background:

• Surrogate models approximate complex functions for efficient optimization.
• Changing design specifications necessitate retraining surrogate models.
• Transfer learning offers a method to leverage prior knowledge for new models.

Purpose of the Study:

• To propose a novel transfer learning framework for design optimization.
• To enhance the accuracy and efficiency of surrogate-based optimization.
• To apply the framework to thin film multilayer solar cell design.

Main Methods:

• Developed a transfer learning approach to refit surrogate models.
• Applied the method to optimize the design of thin film multilayer solar cells.
• Focused on maximizing external quantum efficiency.

Main Results:

• Surrogate model accuracy improved 2-3 times with transfer learning.
• Required only half the training data compared to traditional methods.
• Achieved improved surrogate-based optimization with reduced computational time by transferring material knowledge.

Conclusions:

• Transfer learning significantly enhances surrogate model performance in design optimization.
• The proposed method is effective for complex engineering problems like solar cell design.
• This approach accelerates the optimization process and reduces resource requirements.