Development of a low-cost and quantitative materials analysis platform by physics-based machine learning
This project has delivered a new materials analysis platform that surpasses conventional techniques in revealing the underlying physics of novel semiconductors for advanced photovoltaics. This is key to overcoming the challenges of commercialization of next-generation technologies, such as the identify the mechanism for short lifetime of perovskite-based solar cells.
The developed platform has been demonstrated in a analysis of TRPL measurements of perovskite, which enables the extraction of up to eight material parameters. Based on this technique, we have identified the thermal degradation mechanism of perovskite using the new platform, so that the possible resolution can be explored. This platform will retain the fundamental physics insights during material analysis, e.g., how each parameter affects the performance or stability of perovskites. Thus, it complements existing quantitative methods based on simplified models and recently emerging “black-box” ML methods which may provide predictive capabilities, but often at the expense of physical insight.
This platform has been packaged as software - AiNU - that is available to everyone, presenting researchers with an advanced tool for materials innovation. Beyond this project, it can be applied to any experiment that can be quantitatively modelled with a validated simulation.
Funding
This project is supported by ACAP (funded by ARENA).
People
Dr Hualin Zhan
