DARPA bankrolls optical computing project

As with all DARPA projects, the target metrics are ambitious. At the end of the three-phase programme, DARPA plans to demonstrate an optical processor capable of delaying light by 500 ns with a waveguide loss of less than 2 dB/µs – about 100 times lower than is possible with existing integrated photonic waveguides. The module must also be able to handle power levels of up to 100 mW and have a footprint of no more than 10 cm².

But DARPA has not put a timeline against these targets, and neither has it said how much funding is available. "We are asking the offerors to propose to us the length of each phase based on their approach and the effort required," Jan Walker, a DARPA spokesperson, told optics.org. "We have encouraged proposers to address all areas of interest in a comprehensive proposal."

The ability to delay an optical signal is important in a number of photonics systems, including buffers for all-optical routing networks and interferometers for sensing applications. Such systems currently exploit low-loss optical fibre, but fibre-based solutions are quite bulky and do not offer the precision needed to achieve more complex functionality – such as variable-delay lines and parallel processing of optical signals.

In contrast, photonic integrated circuits offer the scalability and precision needed for more complex applications, but large transmission losses currently limit their ability to achieve long optical delays. Today's best integrated photonic waveguides suffer losses of around 200 dB/µs – some three orders of magnitude greater than in optical fibre transmission – which means that a delay of 250 ns would incur a loss of 50 dB.

DARPA believes that three different technologies have the potential to offer a solution: hollow-core waveguides that guide the light through air; silica waveguides deposited by flame hydrolysis; and waveguides made from nanocomposite materials with a high refractive index. In each case, however, significant developments in design and fabrication would be needed to meet DARPA's performance targets, while improved optical coupling is also essential to maintain that performance at the module and system levels.

To achieve these goals, DARPA has subdivided the programme into three technical areas:

• Ultralow-loss waveguide design: Development of low-loss polarization-maintaining optical waveguides at telecommunication wavelengths. The waveguides must be capable of handling by smooth and abrupt changes in direction, and must be able to handle high optical powers.

• High-efficiency input/output coupling: Multilayer 3D waveguide structures, most likely exploiting passive optical vias in combination with several layers of planar waveguides, will be needed to achieve long delays within a compact footprint. Efficient layer-to-layer power transfer will be essential to minimize input-output losses, while robust mode-matching techniques will also be needed for many different device architectures.

• High-precision on-chip splitting/combining: Approaches will be needed to accurately fabricate waveguide devices with 50/50 on-chip splitting and combining power ratios. Techniques for trimming and tuning the coupling ratios and delay times will also be important.

The project has also been broken down into three phases, each with its own performance targets. Phase I will focus on reducing waveguide losses; Phase II will aim to reduce the size of the integrated waveguide and improve the input/output coupling; while Phase III will produce a technology demonstration as well as achieving the "stretch" performance goals highlighted above.

Organizations interested in bidding for awards under the IPhoD programme must submit a proposal abstract by 20 May, and then have until 15 July to file a full proposal. DARPA's solicitation announcement contains full details for prospective proposers.