Towards a fully differentiable digital twin for solar cells

Advances in materials and device engineering, supported by numerical simulations, have accelerated performance optimization and reduced development costs of emerging PV. Many simulation frameworks only focus on individual sub-components of solar cells, limiting their ability to optimize the total energy yield (EY) of complex PV architectures. Therefore, we present a holistic, differentiable digital twin that calculates the EY from key input parameters. Our workflow integrates morphological modeling, optical simulations, electrical calculations, and device modeling. The differentiable design of the digital twin enables the systematic study of EY dependencies, and the framework identifies optimal design parameters to maximize the EY using a gradient-based approach.