This story is over 5 years old.

Design Advance Could Mean Commercial Light-Based Processors Within a Few Years

But challenges remain.
Image: Pixabay

A team of engineers from MIT, UC Berkeley, and the University of Colorado have overcome a major barrier to light-based computers, offering a means to move beyond the increasingly limited world of electricity-based computing without moving too far beyond the technologies used to implement that world. That is, we can now build photonic chips with existing materials and techniques. This is potentially very big news.


The advance, what the researchers call a "'zero-change' approach to the integration of photonics," is described in the current issue of Nature.

Our electronics nowadays are based on, well, electricity. Information is moved around a computing system in the form of high and low voltages—basically as thundering waterfalls of electrons. As processors get smaller and smaller, they're increasingly running up against the limitations of this scheme. For one thing, it turns out that at the smallest of scales, electrons start to do wacky things and become increasingly unpredictable and unmanageable. This is just the nature of the quantum world.

In addition, electronic information processing suffers from the limitations of what's known as parasitic resistance. That is, as it moves across a conductive material, electricity always experiences some amount of resistance inherent to that conductor. Thinner, longer wires mean lower bit rates and so the more things shrink, the more resistance matters.

As the paper notes, "data transport across short electrical wires is limited by both bandwidth and power density, which creates a performance bottleneck for semiconductor microchips in modern computer systems—from mobile phones to large-scale data centres."

What comes after electrons? That would be photons—particles of light.

An optoelectronic system is basically just a scheme in which computing hardware is able to handle information in both optical (photons) and electronic forms. To send information relatively long distances within a computer, we might imagine some electrons modulating a tiny laser beam. The beam, unencumbered by electrical resistance, transmits the message much faster than electrons moving along a wire, and, at the other end, photons in turn modulate some electrons which interface with the distant component.


The tricky part has been designing the actual hardware for this, or hardware that's reasonably inexpensive and practical enough to make optoelectronics feasible.

The group behind the current study managed to integrate 70 million transistors and 850 photonic components on a single chip capable of handling all of the logic, memory, and interconnection functions required by a computer processor. Crucially, it relies on a 45 nm manufacturing process, which is by now a mainstream scale for integrated circuits.

Image: Sun et al

"The chip was fabricated using a commercial high-performance 45-nm complementary metal–oxide semiconductor (CMOS) silicon-on-insulator (SOI) process," the researchers explain. "No changes to the foundry process were necessary to accommodate photonics and all optical devices were designed to comply with the native process-manufacturing rules. This 'zero-change' integration enables high-performance transistors on the same chip as optics, reuse of all existing designs in the process, compatibility with electronics design tools, and manufacturing in an existing high-volume foundry."

At the center of the new optoelectronic system are tiny components called micro-ring resonators. These are basically microscale optical versions of whispering galleries, in which very slight sounds (sound waves) are able to travel across a large circular space by doing laps along the room's curved wall. Instead of needing a prohibitively large microphone or photodetector, the initially very weak sound or light waves build in intensity as they race around the ring.

As a result, it's possible to integrate very tiny photodetectors on a chip. Rather than making the detector big enough to register a very weak signal, a resonating ring is used to amplify the signal such that it can be seen using a tiny detector.

Chan Sun, the lead author behind the new paper, already has a startup formed to commercialize the technology: Ayar Labs. It could be ready for market within "a couple of years," he told IEEE Spectrum.

Others are somewhat less optimistic. In a separate commentary for Nature, Laurent Vivien, a photonics researcher at the Institute of Fundamental Electronics in Paris, notes that "challenges remain before their zero-change approach can be used for the commercial production of such circuits. First, the on-chip optical communication rate of 2.5 gigabits per second is relatively slow compared with the rate achievable by state-of-the-art silicon photonics systems. An increase in the bandwidth of both the optical modulators and the detectors in the team's SOC would increase the performance of the memory-to-processor link."

"Second," Vivien continues, "a multiwavelength optical circuit may be needed in the future to resolve the interconnect bottleneck. Moreover, much larger numbers of photonic devices and functionalities, including switches, filters and delay lines with low power consumption, will one day become necessary to address the future requirements of computing systems."