Tutorial: Inverse design in nano-photonics via deep learning

Peter Wiecha

Artificial intelligence and in particular deep learning (DL) has proven in recent years to provide powerful numerical methods, with rapidly increasing applications in various fields of scientific research. DL is for instance highly promising for inverse design tasks which cannot be solved with analytical or direct approaches and hence currently require very cost-intensive iterative solvers. In particular for nano-photonics inverse design various deep learning methods have been proposed and benchmarked in the recent literature. In this tutorial, I will give an overview of deep learning for inverse design in nano-photonics and will critically discuss recent developments.

In 2012 I got my physics degree from the Technical University of Munich, and subsequently obtained a PhD in nano-physics from the Université Paul Sabatier in Toulouse (2016). After postdocs at CEMES-CNRS (2017-18) and at the University of Southampton (2018-20), I joined the LAAS-CNRS in Toulouse in 2020 as a CNRS researcher (CRCN). I am interested in phenomena occurring due to the interaction of light with nanostructures, in developing numerical simulation methods and since a few years also in deep learning based techniques for nano-optics.