Integrated Photonics

Experts at the Table, part 1: Are we there yet? Where is the demand coming from and which types of product will drive innovation?

popularity

Semiconductor Engineering sat down to discuss the status of integrated photonics with Twan Korthorst, CEO for PhoeniX Software; Gilles Lamant, distinguished engineer for Cadence; Bill De Vries, director of marketing for Lumerical Solutions; and Brett Attaway, director of EPDA solutions at AIM Photonics, SUNY Polytechnic Institute. What follows are excerpts of that conversation.


Fig. 1: Photonics panel. Photo by Brian Bailey

SE: For readers who are not actively working in this area, can you provide a brief summary of what Integrated Photonics means to you.

Korthorst: Integrated photonics has been around for a long time. It is mainly applied today for long-haul telecommunications systems based on Indium Phosphide. In recent years, lots more R&D and product development has been put into integrated photonics solutions for shorter distances — data center communications, where links are created using silicon as a base material. With the promise of being able to create integrated photonics circuits on a silicon platform, the interest in this technology has been growing steadily. From a maturity standpoint, there are many applications served by integrated photonics in long-haul telecommunications systems, and what we see coming is real products for the data center and beyond.

Lamant: As an EDA provider, we started to hear from several major customers about two or three years ago. They were asking us to help them, saying ‘We can’t do it anymore and we need to move from lab research into production and have it work with our normal design teams’. It reached the point where it is being asked for by the design and design management teams because it needs to be integrated, and it was hurting. The big companies providing network equipment needed to get those things out, and they had a roadmap for the three-to-four year time range, which would get us to today. The demand is real and it is going into more areas. The original application was communications but now I am starting to see people doing LiDAR, people doing sensors, and lab on a chip. It started from a tool perspective being very driven by communication. But in the last year there has been a broadening of the base, and it goes with the broadening of the technology—not just silicon and integrated silicon, but a much broader scope of integration with different components.

De Vries: From our perspective it has been one of the bigger growth areas, where we see traditional photonics companies expanding their capabilities. They have done a number of things associated with photonics, and we have started to see integrated photonics becoming an emerging technology. We have seen that transition in the past two or three years. It has transitioned from the ‘build a better mousetrap’ academic research, where they are trying to find how many different ways they can build a modulator or photo detector, to ‘what are the end applications that can be served with this?’ Optical communications for data center interconnect is the big new application where there is promise, and a number of large-scale organizations are acquiring technology to be able to commercialize it. It is really only the beginning of that, and there are a lot of places where it will extend farther—deeper integration, where we will see architectures for changing the way semiconductor devices are put together, the connectivity between them, processing in memory, board-level optical interconnect to alleviate bandwidth and power constraints. All of that will be served by photonics and it is in the emerging phase. It was driven by long-haul optical communication, which is high-margin. Companies developed leading-edge technology for that space, and that enabled innovation. That is now being leveraged as you get into more margin-squeezed applications.

Attaway: I am part of the AIM Photonics Institute, which is trying to standardize a wafer-build technology that can be used for multiple purposes. So we are not the ones who are pushing the state-of-the-art with regards to process technology, or ‘what could I do if I adjusted these layers on a process?’ We rely on the past 15 years of processing technology that is already in place to standardize a recipe. The challenge for us is how to grow the user base and be able to design something using these technologies. Similarly in the other direction, trying to understand the problems of the world, such as data comms, or sensing, and help them understand how better solutions can be met with this technology. As far as the industry is concerned, the data comms realm is already shipping high volumes of silicon photonics. When you start to look at the needs of the data centers, you see the demand for smaller and smaller connections where photonics could be applied, and those are interesting areas that we are on the verge of moving into. These are lower-margin, smaller but higher-volume applications. The things that will enable more of this will be electro-optical design automation.

Korthorst: For a long time, state of the art in this technology meant development of the best components – state-of-the-art laser or modulator or photo detector or filters. What really changed is bringing those bits and pieces together and not as individual diced components but in a circuit. Bringing together components and hooking them up into a full circuit, either completely photonic or electro-photonic. That is the additional trend for the past couple of years.

SE: It would appear that it used to be integrating things “onto” a chip and now we are talking about integrating them “into” a chip?

Korthorst: There have been several techniques, such as Integrating on a chip or integrating on a silicon optical bench, integrating them into a microassembly of bits and pieces. But today, it is really one traditional flow of having multi-layer, thin film fabrication, having printing of all of the devices in one process.

De Vries: We are starting to see the tipping point, which was historically, ‘Look at my modulator, look at how fast it goes or what I can do with it,’ to being able to place a half-dozen of these and look at the modulation applications that I can now achieve with it. Long haul is the easy one. There are all of the complex coherent modulation schemes and there are a number of organizations that are developing these complex multi-hundred Gigabit interfaces. Integrated photonics plays a role in that, not because the modulator is better but from a functional application perspective, it is one of two or three very effective ways to achieve power/density and all of the things required to make that application successful. So we are seeing that tipping point from building something for technology’s sake to, ‘I am building it to achieve a goal of an application with the commercial constraints that I have.’ Long haul was easy because it is high margin. There are a handful of people who can do it. It’s very specialized. Now it is going into data centers infrastructure where there are orders of magnitude more volume that can be shipped. That becomes very compelling if you can figure out the right way to do it.

Lamant: Let me give you an example of a new application for the usage of photonics. If you open the cabinet of our emulator, it is all fibers between the racks. This is not even a data center, but optical is there.

Attaway: You used the word integrate. That word, ‘integration’ of these functions deserves some definition because there are multiple kinds of integration. Consider monolithic integration or heterogeneous integration, where you could analyze the current state of things based on the kinds of integration that is happening. This is really driving the toolsets, driving the EDA, PDA toolsets. The key is the energy, speed, density factors coming together. Size, weight and power is another way to think about it. These are the things that are driving the interest, and this is what is pushing manufacturing—the assembly of these things, getting fibers to the right place, correctly, with high reliability. These are being driven by these factors coming together – integration.

SE: Is bandwidth or power the primary driver?

Korthorst: Both of those and form factor. If you look at a datacenter, you have a front plate. Through that front plate you need to connect fibers, and the size is limited. The amount of fibers that you can attach is limited. At the back of this fiber is a solution that limits bandwidth. So to get more bandwidth through the front plate, you want more bandwidth in the fibers. That is one driver. On the other hand, while increasing bandwidth, we don’t want to increase the energy consumption.

Attaway: Energy per bit has to drop.

Korthorst: So the main driver is bandwidth. If we solve that problem then you can distribute CPUs from memory, from storage. But the connections need very low latency. Again, optical is a better solution than copper. And if you compare the amount of distance you can travel in copper versus optical, optical wins by orders of magnitude. This is driving the innovation cycle.

Lamant: You can link it to another buzzword, which is machine learning and big data. When you look at what companies are doing, they have to connect the memory to the processor, and the only way to achieve the bandwidth they need to feed the processor is through an optical connection. The bandwidth is a big player there and is driven by the need for data.

Korthorst: CPU speeds have not been going for 10 years. We are all hungry for data and 4% of total energy consumption today is consumed by data centers. It is growing at 7% a year. If you make the calculation and don’t do anything about power, then 10 years from now…Facebook, Google and Amazon are asking the industry to deliver better solutions. As an industry, we try and deliver, as software vendors, we try and bring the design solutions into the hands of those companies to design their next generation of solutions.

Attaway: That is the here and now. That is happening now. It will continue to evolve and get better. Some other areas that are on the verge, that could be high volume very soon, will be the sensor arena—IoT sensors for everything from medical to environmental observations. There is plenty of room for innovation.

Related Stories
Integrated Photonics (Part 2)
What can be done to reduce costs and improve packaging options, and what makes sense in terms of fabrication technologies.
Silicon Photonics Comes Into Focus
Using light to move large quantities of data looks promising, but gaps remain and the adoption timeline will vary by application.
Photonics Moves Closer To Chip
Government, private funding ramps up as semiconductor industry looks for faster low-power solutions.
Focus Shifting To Photonics
Using light to move data will save power and improve performance; laser built into process technology overcomes huge hurdle