The take-away
- The energy consumption in data centres is growing by a factor of 1.7 per year. Photon-based circuits could replace microelectronic circuits, as they require 1,000 times less energy.
- The main challenge: integrating more components on photonic circuits. Right now, 3,000 components can fit in circuit, but microelectronics circuits will integrate billions of components.
Photonics has long been heralded as the next big thing in computing, a technology for the 21st century in which photons replace electrons to offer light-speed super energy-efficient data processing and myriad new devices that benefit society. Yet for the most part the dream of light-based chips has yet to make it out of the lab.
Ton Backx – serial entrepreneur and CEO of both the Institute for Photonic Integration and Photon Delta at Eindhoven University of Technology – believes we are now on the cusp of ushering in this new photonic era. And he is putting in place the fundamental research, technology and infrastructure needed to do so.
Technologist: What is photonics?
Ton Backx: What you are most likely familiar with are microelectronics and microelectronic circuits, out of which your computer systems are built – basically any product that you buy these days makes use of these microelectronic circuits. Photonics is similar to microelectronics. In photonics devices, instead of using electrons as the information carriers, you use photons, which are packages of light energy.
T. Why do we need an alternative to electronics?
TB. The problem we are facing is that the handling and processing of data, which we do in data centers, is still done by microelectronics, and microelectronics is not very energy efficient. In 2016, 5% of all electrical power generated was used for processing data in data centers and telecommunications centers – and that’s growing by a factor of about 1.7 per year. Also, with the Internet of Things, Netflix, Facebook, YouTube and other Internet activities, data centers are growing capacity by a factor of 2 per year.
Microelectronics is a limiting factor if we want to grow traffic at the same rate it has been growing for the past 20 years. And this means that we need to do something different very rapidly.
T. How is photonics better?
TB. Looking at the energy efficiency of the most advanced microelectronics chip, processing a bit of data consumes roughly a picojoule. When you do this for photon-based technologies, the energy budget is less than a femtojoule per bit – so a factor of a thousand less energy demand. The other advantage is that when you look at the frequencies you can cover with photons, they are a factor of about 1,000 – 10,000 higher than the spectrum you can cover with electrons. So you can go to much higher frequencies and you can do that far more energy efficiently.
T. What can you do with a photonic circuit that you can’t with microelectronics?
You can build a single-circuit LIDAR . LIDAR is extremely important for autonomous cars in the automotive industry. When you can provide the full functionality in a single circuit and develop that at a price level that complies with what the automotive industry needs, then basically you are in business.
You can build an optical coherence tomography (OCT) sensor system. OCT is the optical equivalent of ultrasound, but it can do 3D visualization of tissue at the cell level and even look at the metabolism within the cells. This technology is already used today but consists of a cabinet stuffed with all sorts of equipment. What we can do with photonic integrated circuit technology is reduce a cabinet of electronics to a single photonic circuit – and with higher accuracy and higher resolution.
Another application is sensor systems for aeroplanes; you build a single photonic integrated circuit and attach it to a bundle of optical fibers to build a nervous system, like ours in the human body that can continuously monitor load on the heavily loaded parts of the aeroplane. Development is already underway, and prototypes have already been implemented in Airbus and Boeing aircraft.
T. How do you build integrated circuits out of light?
TB. When you look at a microelectronics circuit, you have active components which are the transistors, and you have passive components; resistors and capacitors. With those three building blocks and the electrical wires connecting them, you can build any type of functionality.
When you look at photonic integration technology, you also have basic building blocks, which are an optical amplifier, a polarizer and a phase shifter. With those three components, together with a light guide wave, which is the equivalent of an electric wire, you can engineer any functionality to build any type of processing you can do with microelectronics. The only difference is that the information carrier is going to be a photon instead of an electron.
T. What is holding up photonic circuit deployment?
TB. If you look at photonic integrated circuits today, the complexity is comparable to the complexity the microelectronics industry was integrating into their circuits in the mid-1970s. That means we can integrate about 3,000 components onto a single circuit. It is not a very high integration density when you can integrate billions of components into a single microelectronics circuit. So, one of the major challenges we are facing is increasing photonic integration density.
T. Where will the first applications of photonic integrated circuits be?
TB. The leading market will be data and telecommunications. We have to think of communication speeds of 400 Gb per second across optical interconnects in 2019 and rapidly growing to 1–2 Tb per second in the coming years, which can only be realized with photonic integrated circuit technology.
The same applies when 4G is replaced by 5G. We need to replace current base stations with something much smaller and far less energy demanding that can handle much more traffic – which means we have to use optical technology. I expect that photonic integrated circuits will be the next generation technology that will be used to solve many of the challenges the world faces.
T. What kind of challenges?
TB. If you look at how we want to move towards preventative healthcare, we need to be able to detect illnesses, like cancer for example, at a very early stage and at low cost. The sensor systems that can do that – detect cancer, monitor your blood – are photonic integrated technology based. When you look at smart cities, one of the key things is the control of air pollution, and the sensors you need for that are the photonic integrated circuits that can provide that functionality at low cost. There will be many, many applications we can’t dream of today.
T. Where does silicon fit into this?
TB. Silicon is a very attractive material for photonic applications simply due to the fact that we have an infrastructure for processing it, with very extensive factories all over the world that can mass-produce silicon-based circuits. But from an application point of view, silicon is not attractive at all because it has the highest losses – the losses are at least a factor of 10 higher than in indium phosphide – and you can’t produce active components: you will always need another material connecting to these silicon-based photonic circuits to get light into the circuit.
The only problem with indium phosphide is that, as a semiconductor material, it is not as far developed as silicon is, so it requires further investment in basic infrastructure to provide indium phosphide materials and circuits based upon them at scale levels comparable to what you can do with silicon today.
T. What is your team at TU/e doing to drive forward photonic chip technology?
TB. We are working on fundamental research into how we structure our materials to physical properties in terms of photonic applications, and being able to synthesize these materials to build devices with them.
We are also formulating similar types of standards to those in the microelectronics industry. This involves fundamental research on standard building blocks – standard optical amplifier, phase shifter, polarizer – from which standard functional blocks can be made, and ultimately the circuits with the functionalities we are looking for. It also means creating the software libraries which enable people to design photonic integrated circuits.