Making Electronics-Photonics Integration a Reality

Researchers at University of California, Santa Barbara push the limits with GF Fotonix™

by Gary Dagastine 

In previous posts, we’ve profiled some of the collaborations GlobalFoundries (GF) conducts with leading academic researchers through its University Partnership Program (UPP). These collaborations are important because they lead to proof points and reference designs for new applications of GF’s differentiated technologies. They complement the work of GF’s internal research and development teams. 

The recently introduced GF Fotonix™ platform benefits from academic research in many ways, as we’ll learn later in this blog post. GF Fotonix integrates high-performance CMOS, radio frequency (RF) and photonic components monolithically on the same chip and has the potential to disrupt many existing technologies. 

With GF Fotonix, customers can create innovative electro-optical solutions while taking advantage of the scale and efficiency of GF’s 300mm silicon manufacturing processes, enabling innovative packaging solutions, and the design tools needed for first-pass success. 

New electro-optical solutions are urgently needed because data volumes are increasing exponentially, driven by society’s digital transformation. Electronics-based solutions alone can’t keep up with the needs for more capacity, speed and energy efficiency in data centers and in the overall communications infrastructure. Photonic, or light-based, solutions offer many advantages, but getting electrons and photons to work in a highly integrated, reliable and cost-effective fashion has been difficult. 

No longer. “GF Fotonix is a very large step in systems integration, because for the first time millimeter-wave (mmWave) circuits and digital systems are integrated with photonic building blocks on the same piece of silicon. This opens up many new system and product alternatives, initially to develop next-generation optical interconnects for data centers, eventually leading to the disaggregation of the datacenter,” said Ted Letavic, GF Corporate Fellow, who leads GF’s Silicon Photonics Technology Solutions team and provides innovation and technical leadership to GF Labs. 

“As you can imagine, getting 300 GHz-class control electronics and photonic elements to play well together has been a difficult challenge,” he said. GF’s integration efforts have been aided by the real-world testing and benchmarking work performed by Professors James Buckwalter and Clint Schow at the University of California, Santa Barbara (UCSB), among others. 

Free-Form Design in Silicon 

Beyond its usefulness for next-generation optical interconnects, Letavic said that GF Fotonix incorporates novel photonic elements which will lead to the design of new photonic solutions directly in silicon. 

“A different set of our university partners is exploring the use of mathematically driven design techniques to increase photonic performance and to open up entirely new applications,” Letavic said. “These techniques are variously called inverse design, optical transform design or subwavelength design. Essentially, you decide how light needs to travel in your photonic system to perform a needed function, then you create a series of mathematical expressions – similar to back propagation in neural networks – which solve for a set of physical coordinates which are mapped to a physical shape in silicon that delivers the desired performance.” 

GF and its university partners have made good progress in proving the accuracy of structures built using this approach. “In fact, we’ve gotten to the point where we’re beginning to optimize our process design kits (PDKs) and other design tools to implement these ‘inverse’ techniques into our design flow.” 

“It’s really an entirely new way to think about how to design things in silicon, challenging traditional photonic design at its core,” Letavic said. “It moves us toward what we’re calling free-form design, where you can design any structures that you desire and as long as it complies with our ground rules, we can build it. It’s a game-changer that will open up our customers’ creative floodgates, and it’s cost-effective, too, because we don’t have to change our foundry flow.”  

GF is working with still other university partners to explore entirely new photonic frontiers, such as photonic computing, photonic quantum computing, and biomedical applications. “These research teams are conducting studies to answer blue-sky questions like, ‘What are the elements needed to enable photonic quantum computing in silicon?’ ‘What would a COVID-19 sensor look like if it were photonic-based? We are collaborating to understand how to use GF technologies to address these new frontiers,” he said. 

Longstanding Academic Partners 

The University of California, Santa Barbara is a hotbed of research in RF and mixed-signal technologies, and nobody better exemplifies that focus than two of GF’s partners there, Professors James Buckwalter and Clint Schow. Each has a long-standing relationship with GF, and they are partners on certain GF-related projects as well. 

Buckwalter, an IEEE Fellow, researches high-frequency devices at the intersection of RF, mmWave and photonics, for front-end interfaces, signal processing and mmWave communications at 140/220 GHz. Among the technologies he has used for many years are GF’s FD-SOI (GF FDX™) and SiGe processes. “I’ve graduated more than 30 Masters and Ph.D. students over my career, all of whom have used GF’s technologies, and they take that knowledge with them to their employers,” he said. 

Schow, meanwhile, has more of a photonics background. An IEEE and OSA Fellow, he worked at IBM twice, where he gained deep familiarity with technologies that are now part of GF’s portfolio, such as the 90WG silicon photonics platform and SiGe BiCMOS. He also worked for a tunable laser startup in Santa Barbara, which gave him the desire to teach at UCSB. “I’ve been a believer in GF’s technologies forever,” he said. 

Greater than the sum of its parts 

Buckwalter said that in their research, they don’t look at photonic devices and electronic circuits as two separate things, but rather as elements in a toolbox that can be put together in various ways to solve problems, and GF Fotonix helps them advance that approach. 

“We’ve been driving toward this co-design where we break apart the photonic elements and incorporate them, for example, as small segments inside of electronic amplifiers,” he said. “We don’t treat them as separate entities, we just break everything apart, mix them up and put them back together again. Ultimately, I think in five or 10 years this hybridization of photonic and electronic elements is how design will be done in the high-frequency realm. It’s bringing two things together to get something that’s greater than the sum of the parts.” 

Schow said GF’s design manuals are a great help in this regard. “They’ve always been fantastic, and help us so much in figuring out what the ground rules are so that we can explore within those guidelines. They’ve helped us build things like custom heaters, which might sound like low-tech devices but are key to tuning circuits, as well as custom phase-shifters to optimize performance. From the standpoint of understanding the physical design space as we pursue hybrid designs, it’s been a great experience,” he said. 

Low-power coherent optical communications 

A good example of their hybrid approach is a project the two are conducting which involves broadband waveform generation and detection for the use of coherent optics in data centers. Coherent optics technology is seen as a way to dramatically increase the amount of data light can carry through a fiberoptic cable in a data center, by modulating (i.e., changing) the amplitude and phase of the light, and by transmitting it across two different polarizations.

It involves digital signal processing at both the transmitter and receiver, and the researchers are using GF Fotonix for this work. “While coherent optics is already used in long-haul data transmission, that requires a lot of energy. What we’re trying to do is to make it low-powered enough to be used in short-reach communications links in data centers, to improve networking architectures,” Schow said. 

“But building complete coherent link subsystems is really difficult. On the electronics side you need very efficient high-swing drivers integrated with photonic modulators to generate signaling waveforms, and then on the receiver side, you need a full photonic hybrid to separate the waveforms and to interface with high speed electronics,” he said. “So, there are very highly integrated circuits on both sides; very highly integrated photonic integrated circuits in the middle which must be designed to work with the receiver/transmitter subsystems; and they must also function as the full link which brings everything together. 

“This forces us as faculty advisors, and also our students, to have a broad view. I think it produces students who are well-equipped to go out there and make a big difference in the industry, and of course, all of them will be familiar with the GF technologies that make it possible,” Schow said. 

Looking forward 

Buckwalter said that silicon photonics integrated with mmWave electronics will play an increasingly important role in many areas going forward. One is the relentless march of mmWave wireless communications to frequencies of 200 GHz-300 GHz and above. 

“There’s a gulf between the upper reaches of the mmWave spectrum and the wavelengths of infrared and visible light. Silicon photonics is going to play a very important role here because the losses become so great at these frequencies that there will need to be some combination of photonics and electronics as a low-loss way to send information,” he said.

“And while there are going to be a lot of interesting applications for mmWave-enabled photonics, it will work the other way around, too, with photonics enabling mmWave. For example, for the cellular networks of the future we’re going to need to show higher and higher dynamic range over extremely wide bandwidths, and electronics alone can’t ever meet those challenges,” Buckwalter said. 

Schow said the integration of electronics and photonics is inevitable, whether it is a hybrid type of integration where things are positioned close to one another and densely packaged together, or a monolithic approach. 

“The incumbent way of doing things never wants to die. But I think we’re seeing that play out right now,” he said.