Professor Eric Larkins is head of the Photonic and Radio Frequency Engineering group at the University of Nottingham, as well as technical manager for the COPERNICUS Project, a European funded initiative looking to apply photonic crystal technology to develop high speed compact demultiplexing receivers. He said: "When you talk about electrons, you think of processors, memories and so on. When you think about photonics, it's about the transmission of information. Electrons do a good job when it comes to information processing and storage, but light is good at transport; it's faster and uses less power. The problem is bringing the two together."
One question you could ask is why do photonics and electronics need to come together? Prof Larkins pointed to processor speed. "For the last decade, processor speed has been stuck at 4GHz because they get too hot otherwise – and much of the heat is generated by driving information across the chip."
In his view, pushing electrons over relatively long distances in short periods of time is a problem. "If you double the data rate, you have to change the voltage in half the time. That doubles current and increases the power by a factor of four. It's not that we can't run things faster, but it's a matter of heat and how far you can send information in one clock cycle."
The solution, Prof Larkins believes, is photonics. "Photonic interconnects have a huge advantage and much lower power consumption. More transistors can be connected within one bit period," he said. Photonic-electronic convergence, which began with the connection of servers with optical active cables, is continuing with the development of optical backplane and optical pcb technology. Ultimately, nanophotonic interconnects will be used to send information across chips.
But there is a problem; how do you bring photons into an environment designed to operate using only electrons? Photoelectric components are available, but are significantly larger than the devices into which they might be integrated. "A photodiode measures 100µm, while a laser diode might measure 200µm. Yet transistors are much smaller than 1µm, so you can't connect the two domains using conventional photonics technology; the only structures you can use are photonic crystals or ring resonators with photonic wires."
COPERNICUS – which stands for Compact OTDM/WDM Optical Receivers based on Photonic Crystal Integrated Circuits – is looking to miniaturise and integrate optical functionality. The project, coordinated by Thales Research and Technology, involves University of Nottingham, CNRS, Technical University of Denmark, University of Ferrara, u2t Photonics and Thales Aerospace . One of its goals is to develop new nanophotonic devices, including all optical gates, filters, demultiplexers and photodetectors. Towards the end of the three year project, these devices will be combined to demonstrate monolithically integrated 100Gbit/s wavelength division multiplexing and optical time division multiplexing receivers.
Photonic crystals are nanostructured materials whose sub wavelength features control light in the same way that the atoms in a crystal control a semiconductor's electronic structure. Project coordinator Alfredo de Rossi from Thales Research and Technology, said photonic crystals enable 'unprecedented control' of the confinement of the light and allow the miniaturisation of key optical functions, such as filtering and waveguiding. "Photonic crystals enable high quality cavities with volume less than a cubic wavelength – 10^-18m^3 – thereby enhancing the nonlinear optical response and reducing the switching energy dramatically.
"We believe our approach has all the hallmarks of a highly disruptive technology, with the potential to place Europe at the forefront of photonics."
But it's not the first European project to explore photonic crystal integrated circuits: one trail blazer was the FUNFOX project, which concluded in 2007. "That project used multimode waveguides," Prof Larkins noted. "COPERNICUS is developing single mode devices, which eliminates modal dispersion and allows the realisation of narrow bandwidth filters and small optical waveguides."
He explained some of the work being undertaken. "We're trying to do real integration and developing some key technologies which will have wider application. For example, we have developed a 35GHz bandwidth detector, as well as three port all optical gates with good performance." That latter achievement is important as it represents the optical equivalent of the transistor.
"We've also got the necessary passive components," he continued, "including optical splitters and combiners; things for routing signals round chips and for efficient coupling of signals on and off chip."
Devices such as all-optical gates rely on the strong light-matter interactions arising from the large and rapid nonlinear optical response of III-V semiconductors (such as GaAs) and the strong resonant field enhancement in photonic crystals.
Currently, the project has developed photonic crystals which can switch in less than 5ps while consuming less than 100fJ (see fig 1). The switching energy-delay product is said to be two orders of magnitude less than rival technologies.
One of the important outcomes of COPERNICUS will be a portfolio of devices that work and the tools necessary to implement them. But, for the future, there is the question of just how photonics devices might be integrated into the domain of silicon electronics (see fig 2). "How will you mount these devices?" Prof Larkins mused. "If you mount them on a silicon wafer and the temperature rises by 50°C, there will be thermal issues – silicon and III-V photonic devices have the same problem."
In the short term, applications for deliverables from the COPERNICUS project will focus around telecoms and datacoms. High speed integrated optical receivers are needed in optical active cables for datacom applications in internet server farms and for passenger entertainment and avionics systems in aircraft. "Telecoms networks are moving towards fibre to the home," Prof Larkins pointed out. "As that happens, there will be the need for very low cost optical transceivers for pcs – low profile, high performance devices with a low price tag."
The logical next step is to bring this high speed optical data right to the processor. "Optical backplanes are moving into high performance computing," said Prof Larkins, "and optical pcbs are getting closer." Beyond that, however, Prof Larkins believes photonic interconnects on a processor could allow pc clock speeds to rise to 40GHz. "When this happens, the clock speed is comparable to that of a metropolitan telecom network. What happens when your pc can send and receive information at the same rate as the optical backbone? How will the networks supply this bandwidth? What will future network architectures look like?"
There are many potential applications, but Prof Larkins describes one as 'quite exciting'. "When you have strong optical fields, you can use them to generate nonlinear optical effects to investigate biomedical processes at the molecular and cellular level. If you can develop and integrate this sensor technology, there are many applications in health care and diagnostics," he concluded.
