DARPA seeks to scale photonic computing with new program

Photonic chips promise massive speed and efficiency advantages over electronics, but those benefits evaporate in real-world systems because today's photonic circuits are too limited in scale and complexity.

That forces constant conversion back to electronics that adds milliseconds of delay to systems measured in nanoseconds.

The Defense Advanced Research Projects Agency’s new PICASSO program aims to break through that bottleneck by developing fundamentally new circuit architectures that can perform far more processing in the optical domain, DARPA said in a Thursday notice to release the solicitation.

DARPA plans to award several Other Transaction Authority contracts worth up to $35 million each to develop photonic circuits that can handle more computing than the limited current photonic circuits.

Today’s photonic circuits typically are limited to between 10 and 20 components that perform isolated functions. While these circuits can process data in nanoseconds, systems must continuously convert optical signals to electrical signals for most processing.

That adds milliseconds of latency and wipes out any advantages photonic circuits bring in terms of speed, bandwidth and energy efficiency.

"Systems incorporating photonic circuits struggle to show significant system-level performance advantages over electronic systems," the solicitation states. "This is due to the photonic circuits being typically confined to narrow, isolated functions due to the limited scaling of photonic circuits."

DARPA is looking for circuit-level improvements rather than improve individual components. The agency wants teams to create "perfect photonic circuits with imperfect components.”

The program has two 18-month phases. Phase one will focus developing circuit-level strategies to preserve optical signal integrity and suppress problems caused by reflections, signal leakage and crosstalk across photonic circuits.

Phase two will focus on generalizing those approaches to achieve distinct circuit functionality.

Proposals must address both phases. DARPA will make a down-select between phase one and phase two.

DARPA has a strong preference for U.S.-based manufacturing of the photonic circuits. Any use of non-U.S. manufacturing must include an explanation on why an equivalent domestic capability is not available.

All photonic circuits developed under the program will be delivered to a government-designated design repository. DARPA requires at minimum Government Purpose Rights to intellectual property developed under the program.

Proposals are due March 6.