QSE Quantum Seminar: Efficient device optimization through GPU-accelerated simulations and automatic differentiation

Please join us for the QSE Center Quantum Seminar with Dr Momchil Minkov from Flexcompute Inc., who will give the talk "Efficient device optimization through GPU-accelerated simulations and automatic differentiation" on Thursday October 10. 
Location: BS 270.

Pizzas will be available before the seminar at 12:00. All PhDs, postdocs, students, and PIs are welcome to join us.

TITLE: Efficient device optimization through GPU-accelerated simulations and automatic differentiation

ABTRACT: We discuss the advantages of hardware acceleration of the finite-difference time-domain method of solving Maxwell's equations and show examples of photonic inverse design enhanced by GPU-accelerated simulations and automatic differentiation. Inverse design generally refers to the use of efficient optimization methods to explore a large space of tunable parameters and create devices that vastly outperform conventional designs. This approach has been popular in academia in the last few decades, but has also recently gained traction in industry applications due to demonstrations of devices that are significantly smaller and/or more efficient than their traditional counterparts. GPU-accelerated inverse design is poised to play a major role in the future generations of devices for e.g. telecommunications, and will likely also prove crucial both for photonic quantum computing and for enabling technologies in other quantum computing paradigms, like photonic interconnects, due to the fact that quantum photonic devices have extremely tight requirements on performance and acceptable losses.

BIO: Momchil Minkov is the Director of the Tidy3D solver at Flexcompute Inc., leading the development of the company’s high-performance, hardware-accelerated FDTD solver. Previously he completed his PhD at EPFL, Switzerland, and a postdoctoral fellowship at Stanford University, where he worked on various topics in computational electromagnetism and optimization of photonic devices.