Session: 17-01-01: Research Posters

Paper Number: 146279

146279 - Joint Radiation and Recombination in Nanoporous Thin Film Solar Cells Using Physics Informed Deep Learning Approach Abstract. 

In this study, we address two different major challenges in thin film solar cell research. We introduce an advanced nanoporous silicon (Si) thin-film solar cell design, uniquely embedded with (CdSe)ZnS Quantum Dots (QDs) to harness plasmonic-like effects without necessitating additional metal-dielectric interfaces. This design leverages a complex network of nano-scale pores within the thin film, enhancing light absorption similarly to plasmonic effects, further boosted by excitonic resonances from QDs to amplify the localized electromagnetic field density. To accurately predict the solar cell's spectral response, we employ a Physics-Informed Deep Learning (PIDL) model, preceded by the generation of ground truth data through the resolution of Maxwell's equations within the design domain. This methodology integrates a charge carrier dynamics model, facilitating the evaluation of external quantum efficiency (EQE) and absorptivity by accounting for both the geometric and dynamic characteristics of Quantum Dots (QDs). Dynamic features of QDs are extracted through a dedicated electron dynamics study, which is then incorporated into our Physics-Informed Deep Learning (PIDL) model training.

Another challenge in thin-film solar cell research is the significant drawback of recombination effects. In this work, we delve into the quantification of surface recombination, which impacts conversion efficiency more than trap-assisted Shockley-Read-Hall, Augur recombination, and radiative recombination. Through analytical modeling, we develop a closed-form solution to evaluate surface recombination. We also analytically calculate the electron-hole generation rates using an optical transfer-matrix approach. This analysis allows for a comparative evaluation of the solar cell's quantum efficiency before and after accounting for surface recombination effects, showcasing the critical influence of this parameter on the overall performance.

To further refine our model and optimize the solar cell design, we implement a Physics-Informed Deep Learning (PIDL) model. This method focuses on enhancing radiative properties such as absorptivity and EQE while minimizing the effects of surface recombination. The PIDL model demonstrates exceptional predictive accuracy, with R^2 values exceeding 0.96 for absorptivity and 0.98 for EQE, underscoring the efficacy of our approach in optimizing nanoporous thin-film solar cell performance. In short, our study illustrates the feasibility of achieving high-performance thin-film solar cells without the dependency on metal-dielectric interfaces. The integration of nanoporous thin films with QDs, combined with the advanced PIDL model, paves the way for efficient, optimized solar cell designs, minimizing the reliance on extensive computational simulations, and facilitating quicker, more informed design modifications.
With an emphasis on minimizing recombination effects, this study advances lightweight nanoporous thin film solar cell design, aiming to decrease material usage and recombination rates, enhancing longevity and reducing overall lifecycle costs.

Presenting Author: Farhin Tabassum Stevens Institute of Technology

Presenting Author Biography: Farhin Tabassum is a graduate student at Energy, Control, and Optimization lab (ECOlab) at the Stevens Institute of Technology.

Authors:

Farhin Tabassum Stevens Institute of Technology