Beyond CMOS: exploring new roads and putting them to work
At the same time, they look at new application domains – such as healthcare – and see how IC technology might be adapted to these new uses. Marc Heyns, fellow at the European nanotech powerhouse imec, sketches the R&D environment and sees a need for people with open, multidisciplinary minds.
We need new ideas, and we need many
For a scientist and technologist like myself these are exciting times. There are so many ideas floating around that we want to test and try out; see if we can put them to use and if we can build a technology from them.
Of course, we need these ideas now. Because scaling has become so much more challenging; because we would like to put CMOS to work in new domains; and because we already have to think about the period beyond CMOS.
And we need many ideas. Because only a few of them will survive the hard question: can we make them into a technology? I always say to my students: “nano is the easy part of ‘nanotechnology’, technology is so much harder”.
Ideas may seem so elegant and easy on paper and even in the lab, but that is just the start of it. Take carbon nanotubes. It’s relatively easy nowadays to demonstrate a single CNT device in the lab. But scaling this to a technology where you need a billion devices all neatly arranged and without defects, that is a different business. And that is what we’re doing at imec; that is our business.
Taking an idea to technology is usually a long path. Using high-mobility materials in transistors is another example. It has taken us already 12 years from idea to where we are today. That is a typical timespan, not many ideas come to fruition faster.
Along that path we need multidisciplinary people, more than ever. To build something new – be it technology, systems or applications – we now need large teams with widely varying specializations. But it is essential that these people understand each other: that a technologist knows what a designer does, that an application specialist knows what a system engineer does. And that all have an intimate knowledge of what is possible in a given technology.
Extending CMOS and going beyond
The upcoming ‘Beyond CMOS’ course introduces a number of concepts and techniques that are currently under development. It is an ideal way for PhD researchers and for engineers to look beyond their own specialization and get to know what may become part of their job in the future.
The concepts that are closest to becoming reality are those that stretch the current path of CMOS scaling as far as possible.
One bag of tricks has to do with materials. We have improved the electrical characteristics of our transistors by stressing the materials. When a few technology generations ago the power leakage and supply voltage relative to the transistor dimension became too high, we had to introduce high-k/metal gates. And next on the agenda are high-mobility materials, such as germanium and III/V materials.
The second road is to work on the devices themselves. To look for ways to obtain more control over the channel. That is why FinFET devices were introduced, also after more than a decade of work. From this we can go to horizontal nanowires. Turn these 90 degrees up, and you have vertical nanowire devices, so that is another road into the future.
How far can we stretch these efforts? Looking at what we have today, the 5nm node will already be extremely challenging. Structures get so small that you can really start to count the number of atoms. Using the current density of dopants in the channel of a transistor e.g., you come at a point where the channel has discrete dopant levels, only either one or two dopant atoms or none at all. So we’re then looking at something like 100% variability!
And so we have to start thinking beyond CMOS. Novel materials come into the picture, and we’ve started looking at how we can make electronics with 2D materials such as graphene or metal dichalcogenides such as molybdenum disulfide.
The fundamental problem however is the energy problem, the power density that is escalating. Using new materials, however exotic, to build what are essentially classical field-effect MOS devices is not going to solve the problem in the long term. So that is why we also have to look at alternative routes. Novel device concepts, e.g. steep subthreshold devices such as TunnelFETs allow to reduce the supply voltage considerably. And more exploratory devices that are no longer based on charges, such as spin-based electronics or nano-electromechanical devices.
In parallel, we also work on ways to add functions to chips. We look for ways to have them interact with the environment and create all sorts of sensors. Examples are image sensors or sensors that detect or identify biochemical molecules. Instead of making interconnections for electrical signals, we make pathways for light or fluids. The ‘Beyond CMOS’ course will also introduce some of these concepts, such as silicon photonics or the nano-bio interface.
Testing ideas in a time-box
Looking at all the ideas that come from academia and research institutes, it’s difficult to predict those that will survive and make it into technology. At imec, we’re thoroughly testing each idea in a time-box. Putting a number of specialists on it for a predetermined time. Seeing if and how we can put it to use. Seeing if we should build an activity around it, or if we have to discard it.
Wave computing e.g., which makes uses of the wave properties of electrons instead of charge or spin, allegedly uses very little energy. But what does this mean if you have to build a complete system? A system with machinery to generate waves, reading them out, and storing the result. Maybe you will need more energy instead of less.
Making that analysis is one of our strengths at imec. We can put together specialists from so many domains and see what the consequences are for design, engineering, fabrication. … Truly benchmarking a technology at the system level. That’s where we can make the difference.
Will the next IC technology still resemble CMOS? And how fast will it replace conventional technology? We don’t know yet. But if we look back in history, the transistor technology also took its time to replace the vacuum tube. At first, only specialized applications adopted the new technology, and new and old technology shared the market for a long time.
I expect that we will see something similar. CMOS technology will be pushed to its limits and will be good enough for most applications for a long time coming. In parallel, we will learn how to design in the new technology, leading to the first breakthrough applications. Over time, this will allow the new technologies to further evolve and widen their application space.
Visit imec at www.imec.be
About the author:
Marc Heyns obtained his PhD in Applied Sciences in 1986 from the KU Leuven (Belgium). As program director of imec’s Explore program he was responsible for exploratory research on novel materials and devices for ultimate CMOS technology and novel memory concepts. He was awarded the imec fellowship in 2001 and became professor at the KU Leuven at the department of materials engineering in 2005. Marc Heyns has authored or co-authored more than 350 publications in scientific peer-reviewed journals, more than 700 contributions at scientific conferences, and holds 30 patents.