A Digital Twin to Overcome Long-Time Challenges in Photovoltaics

The recent successes of emerging photovoltaics are largely driven by innovations in material science. However, closing the gap to commercialization still requires significant innovation to match contradicting requirements such as performance, longevity and recyclability. In this contribution, we suggest a layout of a Digital Twin for PV materials able to provide the necessary acceleration of innovation.<br/>The rationale of the Digital Twin is to bridge the gap between first principles calculations, currently not able to predict the crucial solid state properties for emerging PV materials, and high throughput experimentation, currently not able to provide datasets of the necessary scale needed to train generative artificial intelligence (AI) models.<br/>The crucial aspect of our layout is featurization, that is, identifying and retaining only the relevant and non-redundant information present in a dataset. This allows designing fast proxy experiments and surrogate models across all relevant scales. We will learn more from faster and simpler experiments, and the resulting massive (but cheaply acquired) dataset will allow building better approximate models for solid state structure from chemical structure, closing the gap for molecular inverse design, that is, from a set of desired properties all the way back to molecular structure.<br/>Our proposed layout of the Digital Twin combines machine learning approaches, as performed in materials acceleration platforms (MAPs), with physical models and digital twin concepts used in engineering. This layout will allow using high-throughput (HT) experimentation in MAPs to improve the parametrization of quantum chemical and solid-state models. In turn, the improved and generalized models can be used to obtain the crucial structural parameters. HT experimentation will thus yield a detailed understanding of generally valid structure-property relationships. Most importantly, and different from black-box AI approaches, our physics-aware approach can benefit from human abstract thinking, increasing the chance for breakthrough innovation.<br/>Building a Digital Twin in PV materials is agile. We show that some of the necessary building blocks are already available, doing useful work today. Finally, we identify promising approaches for the open challenges such as fast scale-bridging surrogate models and large scale optimization under uncertainty.