It’s no lightweight matter. DARPA is putting about $35 million in total funding on the table in the hope that it will spur researchers to work around fundamental physical constraints and build much larger-scale photonic circuits that do more of the computing with light, not electronics.
A recent solicitation from the Department of Defense’s research arm – the Photonic Integrated Circuit Architectures for Scalable System Objectives (PICASSO) – aims to scale photonics beyond today’s narrow demonstrations. Citing the need for more capable photonic systems, including those relevant to artificial intelligence workloads, DARPA is calling on researchers to submit proposals showing how circuit-level design can overcome the fundamental limitations that constrain current photonic computing approaches, using today’s photonic components rather than waiting for new ones to be invented.
For those unfamiliar with photonic computing, it’s the use of light instead of electrons to process and transmit data. If that sounds far-fetched, it’s not, as DARPA points out – there are already photonic circuits around today, albeit in limited form.
Using light to process data signals has advantages that are ideal for heavy workloads like AI due to greater bandwidth, less latency, and improved energy efficiency. Unfortunately, as DARPA points out, “systems incorporating photonic circuits struggle to show significant system-level performance advantages over electronic systems.”
The current generation of photonic circuits is limited in depth, which restricts their ability to do much beyond single linear mathematical operations. Individual photonic circuits included in larger systems also have to convert optical signals to electronic ones in order to hand them off to other components, essentially eliminating the advantage of nanosecond latency due to the millisecond latency of electronic circuits, which DARPA notes is a 106 performance degradation.
So, what’s holding photonic circuits back aside from, according to DARPA, an industry focus on component-level research? Why, just physics of course.
“The primary limitation to further scaling of circuit size and functionality is rooted in the fundamental properties of signaling with light,” DARPA noted. Cool – easy fix, provided researchers can address the two fundamental technical challenges the research organization points out.
First there’s signal degradation. Unlike metal-oxide-semiconductor circuits, which regenerate signals and filter out noise, photonic circuits have a fundamental issue of optical attenuation and noise that can’t be amplified out, as any attempt to amplify optical signals also amplifies the noise.
Second, there’s the issue of spurious wave interference, which leads to scattering, coupling, mode leakage, back reflections, and unwanted resonance. “Over many components, control of these errors becomes unpredictable, especially when combined with manufacturing variability and thermal and environmental instabilities,” DARPA explained.
Overcoming those limitations, as mentioned above, has traditionally resulted in photonic circuits being interfaced with electronic ones, and DARPA doesn’t want that.
“Heavy usage of electronics prevents system-level gains in latency, efficiency, and bandwidth offered natively by photonics,” the agency said.
Here’s where PICASSO comes in: What DARPA wants from its chosen participants is, like resolving the limitations imposed by the physics of light, simple: Just make better circuits so those two technical challenges aren’t an issue anymore.
“Drawing inspiration from modern electronics, in which clever circuit design overcomes the limitations of individual transistors, the program will foster innovative circuit-level strategies to achieve unprecedented system performance and stability,” DARPA said. “PICASSO confronts these challenges by embracing a new paradigm: creating tomorrow’s photonic circuits using today’s components.”
DARPA isn’t just soliciting ideas, either. By the time PICASSO phase 1 wraps up, 18 months after it kicks off in July, DARPA wants a “demonstration of predictable performance of photonic circuits,” and by the end of phase 2 (an additional 18 month period), it expects “generalized circuit functionality” to be demonstrated.
It also expects all that within a total program budget of about $35 million, spread across multiple awards. Proposals are due by March 6 – good luck. ®