Best Student Paper Award Session

This session presents the inaugural ISC Best Student Research Paper Award, won by Niklas Barr, jointly affiliated with Volkswagen Group Innovation and the University of Heidelberg, for work on energy efficiency of analog photonic processors; the 1,000 euro prize was sponsored by Jeff Hammond. Motivated by automotive AI workloads, where onboard inference could otherwise consume a large share of a vehicle's energy and cloud offload fails latency and connectivity requirements, the paper asks whether literature claims of four orders of magnitude energy advantage for photonic computing are realistic. Barr surveys four photonic computing approaches, Mach-Zehnder interferometers, crossbar arrays, micro-ring resonators, and diffractive elements, then builds a full-system energy-per-operation model that includes digital-analog and analog-digital conversion, memory access, optical energy governed by noise-equivalent power and losses, and weight programming energy, rather than counting only optical energy as many papers do. Key findings: efficiency improves with matrix size until memory bandwidth limits are hit; low-loss optics are essential, since losses scaling with matrix size eliminate the advantage; and weight reconfiguration costs must be amortized through wavelength multiplexing and kernel reuse such as convolutions. Using realistic parameters from existing components, the analysis concludes photonics can beat electronics by roughly one order of magnitude, not four. The talk closes with a new crossbar array design using parallel photodiode summation to reduce loss scaling from linear to square root, fabricated and under test.