Accelerating Cycles of Learning for Silicon Heterojunction Architectures: Experimental Design and Data-Driven Degradation Pathway Prediction

Advanced crystalline silicon photovoltaic (PV) cell architectures, e.g., heterojunctions, mitigate energy conversion losses present in traditional architectures. However, the use of new materials and processes introduces the potential for new failure modes. Our experimental design includes (1) applying different accelerated aging exposures with combined stressors including light, temperature, and acid to unencapsulated textured silicon heterojunction (SHJ) cells and measuring their electrical performance before and after exposure; (2) applying the same aging conditions to planar non-textured test structures (c-Si/aSi:H/ITO stacks), in order to conduct parallel materials analyses, including spectroscopic ellipsometry, X-ray photoelectron spectroscopy (XPS), and ToF-SIMS. Using data-driven network structural equation modeling, we will study the contributions of each stressor, deconvolute the potential multiplicative effects from combined stressors that cause changes in electrical performance of cells, and predict future degradation pathways. By further materials analysis and connecting specific performance degradation mechanisms of SHJ cells to the underlying root causes at the materials level, we aim to identify key design constraints to inform a more robust technology. The goal of this research is to establish the basis for a generalized approach to rapid screening of degradation mechanisms in unencapsulated cells.