Addressing the laser integration challenge in silicon photonics.

(Source: OpenLight) 

OpenLight, a newly launched, independent company formed by investments from Synopsys and Juniper, announced yesterday the world’s first open silicon photonics platform with integrated lasers. The California–based company seeks to provide chip manufacturers with a means to create photonic integrated circuits (PICs) that offer the highest performance possible. Applications will include datacom, telecom, and LiDAR markets, to name a few, all while operating at low power.

With a recent exponential increase in the use of artificial–intelligence and machine–learning technologies, silicon photonics has seen a recent surge. Chipmakers are now setting their sights on PICs thanks to their innate ability to address the growing bandwidth demands of high–level applications.

Yet as those bandwidth demands increase in size and complexity and laser integration becomes more costly, chipmakers are at somewhat of an impasse.

“It’s all about scale,” said OpenLight chief operating officer Thomas Mader. “When you make a very big, complex chip, if you don’t have a laser integrated in, you have to couple it from the outside. Whether that’s a separate package, or whether they try to solder it and align it, optical alignment is hard. If you do it once, it’s hard. And if you try to do it four times on a single product or eight times, it becomes progressively harder, and that means yield, that means cost, that means power lost.”

This is where OpenLight believes its use of integrated lasers sets it apart from alternative open silicon photonics solutions already on the market.

“There are other open silicon photonics platforms out there that have process design kits, but by adding the laser, that’s a pretty complex thing,” said Daniel Sparacin, vice president of Business Development and Strategy at OpenLight. “It actually adds a lot of complexity to our process design kit [PDK] because we need to worry about internal reflections, noise — things that other process design kits don’t really deal with because they don’t have to. So we’re working with EDA, we’re setting up a whole ecosystem now to enable this, and we’re showing our value both by seeing it from customers and by also enabling new things.”

The company’s PDK, which has passed qualification and reliability tests on Tower’s PH18DA process, consists of integrated lasers, optical amplifiers, modulators, and photodetectors that chipmakers can utilize while designing their own PICs.

“One thing that has been stubbornly missing from silicon photonics is the laser,” said Mader. “[With] our PDK, you can plop down a laser when you’re designing your chip, you can plop an optical amplifier, we have an indium phosphide–based modulator, and we have photodetectors. So we have all the active sort of elements, many of which — like the laser and amplifier — are completely unique.”

One of the key components that makes laser integration at scale possible, Mader explains, is indium phosphide. By processing indium phosphide directly onto the silicon photonics wafer, it allows chipmakers to achieve scalability, cost advantages, power benefits, and a level of reliability previously unattainable with traditional silicon photonic technologies.

“We have an indium phosphide modulator that easily does 200G per wavelength, and we believe we have a nice edge there on silicon–only modulators,” Mader said. “Also, power; ultimately, our modulator is very low–loss, and our laser is very low–loss — the laser only has a few percent loss getting into the silicon. Because we have such low loss between the elements, [and] such low–loss elements, we start to pull away in power efficiency.”

Regarding reliability, Mader explained that by emitting directly into silicon, they can avoid certain failure modes.

“Normally, the discrete lasers — they’re like a hunk of indium phosphide that has a structure on it and the edges are a critical part of that laser,” said Mader. “If that edge gets a little defect, it’s one of the ways they fail. We have no edge. We’re actually bonded to silicon, and we emit into silicon, and then we’re [hermetically sealed] on the top, so there are certain failure modes that simply aren’t there.”

In addition to its PDK, OpenLight will also offer select manufacturers the option to utilize a 400G–DR4 and 800G–DR8 PIC designs with 2km reach to speed up their time to market. 400G–FR4 and 2x400G–FR4 PIC designs are also in the process of being designed.

“We are offering, in select cases, some designs to get customers to market faster,” Mader said. “We have a 400G and 800G PIC design that we’re using as a shortcut.”

The company also expects to tape–out the first open multi–project wafer (MPW) shuttle to further lower manufacturing costs. The MPW shuttle will run on the PH18DA process.

As of today, OpenLight has approximately 40 employees and has obtained over 200 patents. The company is production–ready, with its first customer tape–outs expected in summer 2022.