Science Foundation Ireland: Intel collaboration leads to huge leaps in a tiny space
Science Foundation Ireland nurtures world class research in Irish universities
Intel Ireland silicon research director Leonard Hobbs: SFI “provided a focus for what kind of advanced research Ireland should be involved in and this included advanced microelectronics and ICT”.
Some of the technologies which will underpin the development of future generations of microchips are being researched here in Ireland as a result of collaborations between Intel and a number of Science Foundation Ireland (SFI) funded research centres and groups.
“We have had a very active engagement with Irish universities for the past 25 years but our research activity here really started taking off with the establishment of Science Foundation Ireland 10 years ago,” notes Intel Ireland silicon research director Leonard Hobbs.
According to Hobbs, the establishment of SFI dramatically changed Ireland’s standing as a location for world class research activity.
“It did two very major things. Firstly it provided a focus for what kind of advanced research Ireland should be involved in and this included advanced microelectronics and ICT. Secondly, it provided serious levels of funding which attracted world class researchers to come and conduct world class research here. It also provided the funding for leading Irish researchers to stay at home. We were able to take advantage of that to increase our own research activity and we are now involved with Crann and Amber in TCD, the Tyndall Institute in UCC, as well as with other projects in DCU, UCD and UL.”
Hobbs points out that this is Intel’s largest research commitment outside the US and that it is motivated by what is known in the industry as Moore’s Law. Gordon Moore was one of the co-founders of Intel and he predicted back in the 1960s that the number of transistors which would be placed on each square centimetre of a silicon chip would double every 18 to 24 months. That observation has proven remarkably accurate over the ensuing years and it is this rate of progress that has delivered the digital revolution of the past few decades.
The progress enabled by Moore’s Law has really been quite startling. “Howard Aiken, who designed IBM’s first computer back in 1947, said that at most there would only ever be six computers needed in America”, Hobbs notes. “Now we have an amazing array of devices using chips which are continually advancing and improving in terms of capability and complexity.”
War and Peace in one second
He talks about what would have happened with some other industries if Moore’s Law applied to them. “Intel’s first microprocessor, the Intel 4004 back in 1971, could process 92,000 instructions per second. The latest Intel i7 can process 92 billion instructions per second. If typing speeds had accelerated at that rate you would be able to type War and Peace in one second. In the automobile industry a Porsche 911 would cost €1.”
But this rate of progress comes with its own pressures. “This means that we constantly need new tools and innovations to be able to make the new chips.”
He explains that the microprocessor industry originally used highly sophisticated lithographic print processes to etch the circuits onto the silicon. The limits of that technology were reached about 10 years ago, however, and ever more complex means have to be found to achieve the necessary performance gains.
“We are working at incredibly small dimensions,” he says. “The measurements involved are around nine or 10 nanometres. A nanometre is 100,000 times smaller than the width of a human hair. At this scale we can fit one billion transistors on one square centimetre of silicon – about the size of a fingernail.
“This makes the manufacture of semiconductors the most advanced manufacturing process in the world. We are using the most advanced techniques and materials and all that requires very active research. The entire industry is constantly looking for the next big breakthrough.”
These breakthroughs needn’t be spectacular. For example, in 2003 a major advance was the introduction of strained silicon. In very simple terms this involved stretching the silicon used in the chip which altered its properties in such a way that less power was required. In 2007 silicon dioxide was replaced with a metal oxide, again resulting in performance improvements. Another breakthrough came in 2010 with the introduction of the three dimensional chip.
“The biggest strength Intel has is our ability to innovate and then integrate this with high grade manufacturing” says Hobbs. “The results from our Irish research activity have been really impressive, both in terms of helping to create an ecosystem which provides the graduates we need to recruit and in the contribution it makes to the innovations we need to manufacture future generations of microprocessors.”
While much of this collaborative research is, naturally, highly commercially sensitive, Hobbs mentions a few examples which illustrate its complex and advanced nature. “The research spans the entire gamut of the semiconductor manufacturing process. One part of that process is transferring the pattern. You can do that by taking a photograph or you can do it by self-assembling molecules. A bit like the way oil and water separate, you can get polymers to mix in a certain pattern.
“We are also doing some very fundamental work looking at copper. Copper is a hugely important metal in a chip and when it is very, very small it starts exhibiting different properties. We are working with TCD and Tyndall to look at this issue.”
Another piece of fundamental research is actually quite simple in concept. While the traditional understanding of a transistor is of a switch that is on or off at any given time, it turns out that a transistor is never fully off and “leaks” power.
The creation of less “leaky” transistors could result in significant power savings and performance improvements. “What we want to do is create an old style relay or mechanical switch which creates a physical gap. We are working with TCD on the creation of a mechanical switch at the nano scale.”
All of these pieces of research, and more, may some day find their way into a new generation of Intel chips. “It takes about 10 years from the implementation of a research project here to its integration into high volume manufacturing. We are starting to see some of our early work being adopted. We had very good relationships with the Irish university sector up to 10 years ago but SFI brought it to a different level. We are working on some very exciting and interesting projects and we really don’t know where they will lead. The one thing we do know in this area is that the best ideas are not the ones you said you were going to do at the start.”