JOHN WILLIAMSON SILICONPHOTONICS
PROGRAMMABLE, COMMON ARCHITECTURE SILICON PHOTONICS CHIP DEVELOPED
BUT FUTURE’S BRIGHT Notwithstanding theses sorts of challenges, the future for silicon photonics certainly looks bright. As data rates continue to grow, Intel expects to see the integration of optics with networking silicon, either in the form of high-density switches, network interface controllers or FPGAs with optical I/O. “We see the drive to smaller and smaller form factors with lower power per bit, higher bandwidth densities, and lower cost per bit,” says Blum. “Silicon photonics is a market of big promises - especially in data centres and high-performance computing, huge markets that will dwarf all other silicon photonics applications in the near future,” predicts Dr. Mounier. “Also, silicon photonics can be seen as an “enabling technology” for other applications, for example sensors, life science, quantum computing, telecommunications, consumer and automotive.” As described by Dr. Mounier, silicon photonics is a mix of several technical blocks - optical, but also integrated circuits for processing, MEMS for packaging, copper pillars and so on, and it involves several core components. These are: l Lasers at the heart of any optical device. Today’s lasers use Indium Phosphide to produce coherent infrared laser light. l Photons are then modulated to break the light into optical pulses. l Optical waveguides and other interconnections are necessary to move pulses from one place to another. l Multiplexers/demultiplexers are used to separate and combine different wavelengths. l Finally, detectors convert an optical signal into electric signals. SILICON PHOTONICS: THE NUTS AND BOLTS
Researchers from the Silicon Photonics Group at the Optoelectronics Research Centre (ORC), University of Southampton, and from the Institute of Telecommunications and Multimedia Applications (iTEAM) at the Universitat Politècnica de València, have developed what they say is the first photonic integrated chip that enables multiple functionalities by employing a single common architecture. The chip’s behaviour is similar to that seen in electronic microprocessors, where a common hardware is programmed to perform a desired operation. The team’s results have been published in the journal Nature Communications. By programming the internal connections of a single chip architecture, the chip can be configured to perform different functionalities, and could be used in any field susceptible to the requirement of optical or radio frequency signal processing. This includes, for instance, self-driving cars, mobile communications, quantum computing, distributed sensors, sensing monitoring, the Internet of Things, defence, avionics and surveillance systems. The chip architecture relies on a honeycomb waveguide mesh implemented by properly cascading tunable light couplers. The independent basic coupler configuration allows the definition of flexible interconnection schemes as well as the definition of different optical signal processing circuits. The chip was designed by both teams, fabricated in the Southampton Nanofabrication Centre by the members of the Silicon Photonics Group, and characterised by the València team. The main advantage of this approach is that the physical hardware architecture is manufactured independently from the targeted functionality to be performed, which reduces design cost, fabrication and testing iterations. Once designed and tested, the chip enables the configuration of more than 100 photonic signal processing circuits, of which around 30 configurations have been demonstrated by the team in Nature Communications, resulting in the highest number reported to date. “This represents a paradigm shift in the field of integrated photonics, from application specific photonic integrated circuits to generic purpose and programmable devices, in the same way as the success experienced by the electronic field in the 1980s,” said Professor José Capmany, lead researcher of the València Group.
Standardisation may be another issue. According to Kannan a major restraining factor for the growth of silicon photonics is that standards for developing photonic integrated circuits are yet be universalised. “This has led to an unclear state in the supply side of the value chain as independent original equipment manufacturers are unsure whether their product would be interoperable,” he says. This also may have an impact on investment decisions regarding mass manufacturing capabilities.
use industry-standard form factors that don’t necessarily take advantage of all the benefits that come with using silicon photonics. More specifically, Dr. Mounier instances laser sources and packaging. He says that since silicon cannot have a laser effect because of its indirect bandgap, the laser cannot be monolithically built on the silicon die. So, there are different solutions to laser integration. “After years of R&D, Intel succeeded doing wafer-level integration of the laser by bonding an Indium Phosphide (InP) chip (rather) than doing post-processing, so alignment is not so crucial as it will be performed by lithography,” he observes. “The second solution is to flip-chip the laser source, but it is a complex process due to low throughput and high alignment accuracy.” When dealing with optics, packaging accounts for a major share of the final component cost, because of alignment issues and the need to integrate different chips in the same package. Dr. Mounier calculates that in semiconductors, silicon accounts for 80% of the cost while packaging is 20%. In optics, it is the opposite as packaging can be as high as 80% of the final cost. “So solutions are currently developed to reduce cost by increasing assembly throughput at high accuracy,” he explains.
Silicon photonics wafer fabrication at the ORC, University of Southampton, UK.
www.opticalconnectionsnews.com
29
ISSUE 11 | Q4 2017
Made with FlippingBook - Online Brochure Maker