Integrated Photonic Neural Networks with Arbitrary Capabilities

As data generation and transfer rates have grown rapidly over the last decades, integrated photonic systems have become the key technology enabling modern communication systems.

In order to sustain future economic and societal growth, continued development of arbitrarily complex and multi-functional integrated photonic systems is therefore imperative.

Traditional design of these systems relies on determining device geometries using analytical electromagnetics, after which various parameters are optimized.

In contrast, the flexibility for more complicated optical functionality is currently only possible with “blank-slate” optimization routines.

In these algorithms, the device structure is determined by searches through thousands of degrees of freedom, which is computationally prohibitive when targeting arbitrarily complex functionality with larger devices.

To this end, this project will develop an artificial intelligence-based, universal photonic neural network architecture and its optimization framework to enable and experimentally demonstrate arbitrary photonic capabilities on-chip.

For the first time, this novel approach will allow solutions for designer-specified operations including arbitrary combinations of wavelength and polarization-specific transfer functions.

Resulting devices will be fabricated and characterized to demonstrate previously elusive on-chip functionality, and for rapid adoption and widespread use.

Customer needs in communications and sensing applications will be specifically targeted through an industrial secondment, and a structured innovation management/commercialization plan.

This framework and its industrial use represent a vast leap towards universal integrated photonic design for advancing European capability and economic drivers through innovation in future optical systems.

As such, the fellowship will transform my career towards future leadership at the intersection of academic research and industrial innovation.