Meint Smit: Next Steps for InP-based Photonic Integration Technology
Indium Phosphide is a versatile semiconductor material that allows for integration of a broad range of components in a single chip. Basic building blocks include lasers, optical amplifiers, modulators, detectors and a variety of passive components, including wavelength demultiplexers, filters and couplers. InP-chips with record complexities of more than 1700 components integrated in a single chip have been reported. InP-based foundry processes offer low-cost access to mature integration processes with high performance and wafer-scale integration of InP-based photonic circuits with silicon electronics is emerging. Meint Smit will discuss what’s coming next and why this phase is important.
The Story so far
“We’ve analysed the success of microelectronics. Now we’re putting those lessons learned to work for integrated photonics” says Professor Meint Smit of Eindhoven University of Technology.
The development of silicon based microelectronics means today it costs only a few cents to design and develop per square millimetre of chip, as the technology is mature and highly standardized. In addition, its development costs are low because we have sophisticated software for the fast and accurate design of the chips.
Teams at Eindhoven University of Technology (TU/e), led by Professor Meint Smit, want the same for photonic devices. And now that tipping point has been reached.
“Around the start of this new century, we were all asking ourselves: when is photonics integration really going to take off in the same way as silicon micro-electronics?” explains Smit. “After all, in 1980 I had started saying it would kick start in the 1990. Ten years later, the tipping point had moved to the start of the new millennium. By that time, some critics started to say that the promise of photonics integration would always remain in the future.”
Understanding the success of microelectronics
“In Eindhoven, we looked at the world of microelectronics and discovered that things were arranged in a very different way. Up until the 1980’s you had a lot of different technologies being used. There were a whole range of transistor types and countless variations between them. But in the 1980’s things started converging and you started to see the development of generic processes.”
“Designs start to use the same basic building blocks of transistors, resistors and capacitors. Even some of today’s most complicated processors, with over 1 billion transistors, have less than 10 different components. So you are using the same set of building blocks to create all kinds of things.”
“We asked ourselves, how can this be applied to photonics? If you look to light, it has an amplitude, a phase and a polarisation. So if you make a component for manipulating the amplitude, another one for altering the phase, and one for changing polarisation, you could achieve a lot of things. Then you simply need a waveguide to connect them.”
“So we started developing a process with an optical amplifier for the amplitude, a phase modulator to change the phase, and a polarization converter.”
“There is a big advantage if everyone uses the same generic process. You can combine a lot of different designs on one chip. In a recent successful test case, we had 20 different designs on one wafer.”
There’s always more than one run
“The challenge in this business of integrated photonics has been the cost of making a chip prototype. If you develop a chip it always takes 2-3 runs before you get it right, because the chip never performs exactly as you expect.”
“Traditionally, one process run currently costs at least €200,000, if you do it on your own. So this is a huge barrier for small companies. But if 20 users can share the same wafer, then the costs drop to €10,000 each and that’s within the scope of many startups. You get 8 identical cells back from the foundry which you can test and measure. You can then iterate the design, so that the next run gives even better results.”
Open Collaboration is key to maintaining Europe’s Lead
Eindhoven Technical University (TU/e) also drives and actively contributes to the International JePPIX platform. This is an open collaboration between over 15 partners, most of them key-players in photonic integration, designed to connect researchers, PIC designers, and generic foundries to dramatically cut costs and speed up the time to market. TU/e have also set up Smart Photonics, to make its knowhow on generic Photonic IC fabrication available to users on a commercial basis.
“The role of JePPIX is to act as an independent broker between users, designers and foundry’s. We announce a production schedule, collect the designs and bring them together in one big mask set. That goes to the foundry that produces the wafers, dices them in separate chips and sends the chips to the various designers. So this is the way TU/e has introduced generic photonics manufacture into Europe.”
The race is on
“Clearly, others have recognized we’re on to something. In October 2014, US President Obama reserved US$ 220 million to set up something similar in the United States. Several leading universities and technology companies took part in the tender to create the “Integrated Photonics Institute for Manufacturing Innovation”. In late July 2015, we learned that the winning bid was the city of Rochester. Additional private funding means a total of around US$610 million has been set aside to fund the US program in photonics chip manufacture. It means the creation of many local jobs in that region of the US.”
“It turns out that the generic integration foundry infrastructure and logistics is quite complex, so being able to master this gives us a head start. Now we’ve reached a very important tipping point. And we are only at the beginning of what is possible. When we look at the world of semiconductors, in the 1980s no-one thought such generic processes could ever drive very fast 60 GHz radio frequency circuits. Personally, I think if you are willing to invest enough money in a single technology, then it will surpass most of the other technologies.”
Indium Phosphide Wins on Price and Performance
“Now, let’s suppose your chip works and you’re happy with its performance. Then scaling up to produce 100,000 pieces is no problem. You can order them immediately knowing that the specs will remain constant. This part can be done using a more traditional process. A large Photonics foundry like Oclaro can manufacture 10,000 wafers a year.”
“The other point is the chip cost. Many people think that InP is very expensive. In comparison with silicon fabs we are actually cheaper for low and medium volumes, while at the same time offering excellent performance. Of course, silicon photonics can make their chips cheaper if you buy millions.”
“But few small companies start with millions. They want to start with a few hundred or a few thousand. These volumes are much too small for a silicon foundry because the entry/set-up costs are very much higher than in a smaller InP fab. So if you look at the functionality per Euro, Indium Phosphide wins and we are very competitive in the lower and medium volumes.”
“The next step, and we have the first products lined up for that, is to do the whole thing again but then not on a substrate of Indium Phosphide, but on Silicon. But then we will add a very thin layer stack of Indium Phosphide.”
Riding the Seven League boots of Silicon
“The holy grail of this industry is to put lasers onto a piece of Silicon. Silicon itself is not suitable for this – you can’t generate light directly, whereas with Indium Phosphide you can generate light directly in the chip. We are developing methods to build full photonic circuits in a thin Indium Phosphide layer on top of a silicon wafer containing the driver, receiver and control electronics, so that we get very short and efficient connections between the two.”
“We think that Indium Phosphide will champion by applying the methodology of microelectronics to InP-based photonics and we believe our generic approach will show rapid scaling.”
About Meint Smit
Meint K. Smit graduated in Electrical Engineering in 1974 at the Delft University of Technology in the Netherlands and received a Ph.D. in Integrated Optics in 1991, both with honours. He invented the Arrayed Waveguide Grating, for which he received a LEOS Technical Achievement award in 1997. In 2000 he became leader of the Photonic Integration group at the COBRA Research Institute of TU Eindhoven. His current research interests are in InP-based Photonic Integration, including integration of InP circuitry on Silicon. Meint Smit is a LEOS Fellow. In 2012 he received an ERC Advanced Grant.
About Photonic Integration Group (PhI)
The Photonic Integration Group (PhI) within the Eindhoven University of Technology is involved in a number of national and international research projects and plays a leading role in InP-based photonic integration. PhI research offers ample opportunities for Master and PhD students to do research on advanced photonic ICs or technology and to get a job in a variety of smaller or larger high-tech companies that use Photonic ICs in their products or start your own business.