Lumerical Hosts Online Photonic Inverse Design Challenges
Lumerical, a developer of photonic design and simulation tools, announces the Lumerical Inverse Design Challenge and the Lumerical Best-in-Class Challenge hosted during Lumerical’s Online Tech Day on March 13.
Throughout Lumerical’s Online Tech Day featuring over a dozen ecosystem partners, Lumerical will host two unique design challenges online, accessible globally. Throughout the day photonic designers can participate by simply specifying through a web interface the design targets and constraints for a 3 dB power splitter design workflow. These parameters drive a photonic inverse design optimization using the Lumerical FDTD photonic simulator running on Amazon Elastic Compute Cloud (EC2). In real time, each challenger’s entry will be created utilizing the power of Lumerical’s photonic Inverse design simulated with the industry leading FDTD nanophotonic simulator powered by the high-performance computing capability of Amazon Web Services (AWS). The Lumerical Design Challenge Channel online leader board will continuously update the standings throughout the day, highlighting the top designs. By email, Lumerical will provide each participant their resultant layout file along with their simulation results and design files for their future use. The leading designer will be provided space on an upcoming LIGENTEC InP MPW to fabricate the design of their choosing. Results will be announced at the conclusion of Lumerical Online Tech Day at 7:30 pm PDT. A selection of leading designs will be chosen for fabrication.
Over the course of Lumerical Online Tech Day, Lumerical will attempt to create a photonic design that exceeds the best published Figures of Merit for a functionally equivalent design. Similar world class devices can take months to design using traditional design methods. Lumerical plans to harness the power of photonic inverse design (PID) with Lumerical’s FDTD photonic solver distributed across an array of instances of Amazon EC2. This combination of world class software and the great computing power easily available through AWS makes it possible for Lumerical to attempt a world class design in a single day.
Lumerical will design and optimize an “O-band/C-band wavelength demultiplexer”. The device accepts input from a single, broadband waveguide and transmits the wavelengths between 1260nm and 1360nm into the upper output waveguide while wavelengths between 1500nm and 1600nm are transmitted into the lower waveguide. Lumerical will target a footprint of 2.8um x 2.8um.
Lumerical CTO James Pond said, “Because of the strength of designs that we have seen from photonic inverse design, I am confident that we can deliver on this challenge. Without photonic inverse design and AWS’ power of distributed high-performance computing, it would be very challenging to even attempt a world class design in such a short amount of time. I think it will be close, but I also think that participants at Lumerical Online Tech Day will see an impressive feat of photonic design."
The Lumerical Design Challenge Channel will continuously update a graphical scoreboard detailing design progress. Photonic designers are welcome to drop in online to monitor Lumerical’s progress and to discuss photonic inverse design at the Lumerical Ecosystem Experts Channel.
“The addition of inverse design methods to Lumerical’s products has been a game changer for our photonics design efforts – they have enabled better performance and much faster optimization of grating couplers and other complex photonic components, than with traditional sweeps and particle swarm optimizations,” said Michael Moebius, Senior Member of the Technical Staff at Draper.
Coupled with Lumerical’s industry proven FDTD solver, photonic inverse design enables designers to efficiently explore and optimize designs with hundreds or more design parameters. In contrast, traditional design approaches are limited to a small number of established device designs, exploring much smaller parameter spaces, typically fewer than ten parameters.
The Python based open source PID implementation is packaged together with Lumerical FDTD for ease of deployment. Alternatively, the original source code is freely available on GitHub. To help new users get started with PID, examples are included in Lumerical’s application gallery to speed designers’ time to implementation. More information on Lumerical’s Automation API, required to run PID, can be found at https://www.lumerical.com/products/aapi.
Look for more details on the design challenge and Lumerical’s Online Tech Day at https://www.lumerical.com/learn/event/techday-2020/
