The world's desire for faster and smaller semiconductor products with greater functionality has indeed led to smaller and more powerful integrated circuits. The problem is that they use thinner interconnect lines (or copper wires) between transistors on chips and between chips. These thinner lines impair the speed of processing, offsetting the increased speed from the smaller transistors.
The problem is known, in industry parlance, as the interconnect dilemma. And if it is not resolved, it will strangle growth in the IT industry before the decade is out.
The Defense Advanced Research Projects Agency ( Darpa), the US agency responsible for the development of new technology for the military, deems the problem so serious it has allocated priority research funding to organisations such as the Massachusetts Institute of Technology and the Georgia Institute of Technology.
And market leaders in chip manufacture, such as Intel, ST Microelectronics and NEC, all have comprehensive research initiatives focused on the same outcome.
The solution may lie closer to home.
These industry giants are keeping a watchful eye on the work being done at the Carl & Emily Fuchs Institute of Micro-electronics (Cefim) at the University of Pretoria. A research team at the university is close to a technology breakthrough which could solve the interconnect dilemma.
The research group, working under Cefim director Monuko du Plessis, has developed a novel optical technology, known as injection-enhanced silicon in avalanche, or InSiAva, that has the potential to solve the chip-to-chip and on-chip interconnect problem facing the computing industry.
The technology replaces the copper interconnects with optical interconnects, which send signals by means of photons, not electrons.
It could be a US$250bn solution. "The research group that solves this problem will develop a technology applicable to all integrated circuits on silicon chips," says Du Plessis.
It is envisaged that the first applications will be in products that rely on a high switching speed. These include video applications, for instance in the entertainment and security industry, and imaging, such as medical imaging or geographic information systems.
Chip-to-chip optical interconnects could send data a thousand times faster. This potential advantage creates a huge market for all computing applications, making it a truly "billion unit" application.
These opportunities were not lost on the University of Pretoria (UP), which is noted among SA universities for its commitment to research, innovation and the commercialisation of intellectual property. Nor were they lost on other investors.
The SAip Fund, SA's only early-stage venture fund, managed by Triumph Venture Capital, has just agreed with UP to invest R15m over 30 months to fund the next leg of the InSiAva research.
The research was initially funded by the Carl & Emily Fuchs Foundation, which established Cefim at UP in the early 1980 s. Cefim and the university invested an estimated R45m in the project between 1992 and 2004.
A subsequent phase of the project was funded jointly by the Innovation Fund and UP to the value of R3,6m, with UP investing additional funds into commercialisation activities.
In the computer industry the medium for long-distance communication and data transfer has changed in recent years from copper wire to fibre-optic cables. This has enabled higher speeds and efficiency in data transfer.
However, within the computer itself, chip-to-chip and on-chip interconnectivity is still based on a copper medium, with negative results for speed and efficiency of data transfer, as mentioned above.
"For as long as the solutions are based on electronic interconnectivity, the interconnect dilemma will remain unresolved," explains Du Plessis. "The solution requires a technology paradigm shift."
Du Plessis started researching this problem in the early 1990 s. "We shared the globally held belief that it would be resolved through a switch from copper to optics, in the same way as it was in long-distance communication and data transfer." This was affirmed in the August 2002 issue of the scientific journal of the International Institute of Electrical & Electronic Engineers (IEEE), called IEEE Spectrum, which projected that the introduction of short-range optical data transfer would follow the trend outlined below:
- 2002: Computer to computer
- 2004-2007: Board to board
- 2007-2012: Chip-to-chip optical interconnects
- 2013-: On-chip optical interconnects
The first two projections have occurred as predicted. Global focus is now on the latter two milestones, which represent a technological challenge for the computing industry and offer an opportunity of immense value for whoever can provide solutions first.
Though the InSiAva research is focused on reaching the final two milestones, other technical solutions are also being explored by UP's global competitors. These include microwave interconnects and technologies that hope to use potential advances in superconductors and cooled conductors.
However, the InSiAva solution is several steps ahead of its competitors. Its most significant strategic advantage stems from the fact that most of the contenders in the race are not compatible with the silicon-based CMOS (complementary metal-oxide-semiconductor) standard on which the semi conductor industry is built.
"We have spoken to Intel and NEC and know firsthand that a breakthrough by any technology that is not silicon compatible will face significant industry resistance," says Du Plessis. "That's because it would make obsolete about US$500bn in irreversible investments that the semiconductor industry has made in silicon wafer fabrication plants."
Indeed, Intel was sufficiently interested in Cefim's work for Intel chairman Craig Barrett to visit UP last year.
The InSiAva research deliberately focused on fast-switching electronic devices that would emit light from silicon to facilitate optical data transportation on and between silicon chips.
The challenge is that silicon is inefficient at generating light from electrical current.
"While others were researching more exotic devices, we concentrated on the field of silicon light emission," says Du Plessis, who is an engineer by training and passion. This, he says, makes his research efforts far more pragmatic and industry-friendly than those of some other academic researchers. "We believed a solution had to be compatible with established integrated circuit technology."
As a result, UP's InSiAva technology is totally compatible with CMOS.
Though the research process is not yet complete, the potential is there, and commercialisation of this technology has been under way for almost two years. "Commercialisation is a process, not an event," says Gerrie Mostert, the consultant retained by the university to assist with the process.
To facilitate private-sector investment and commercialisation, UP established a company, InSiAva (Pty) Ltd, to which the global rights to the intellectual property (IP) are being assigned. The SAip Fund will hold shares in this company. "For us it was an opportunity to invest in a technology that has global mass market potential," says Triumph Venture Capital CEO Wellington Chadehumbe.
An international advisory board is being assembled under the leadership of Laurie Olivier, a UP alumnus and venture capitalist based in Atlanta, US. The board will advise the technical team and the investors on the development of the technology and assist with the identification and selection of options for commercialisation.
"Our network showed us that an international advisory board is a must-have for technology start-ups around the world," says Mostert. "It is a precondition should we want to raise money overseas."
The advisers have been chosen with potential investors in mind. "Venture investors will look at the advisory board, the people on it and what their credentials are. We have identified our preferred exit companies and we know whose opinions are important to them."
Working on the development of a technology with the potential to shape the future course of IT is an ambitious project. "It goes beyond technical brilliance, and relies on international standing, credibility and strategic contacts if it is to succeed," says Chadehumbe.
Credibility and visibility are two things Du Plessis and his team have achieved over the past 15 years. The technology has been peer reviewed by the international academic community through more than 50 refereed research papers, and published in noted scientific journals. In addition, two US patents have been granted. Du Plessis himself has become an international resource on the subject. For one, he is an IEEE distinguished lecturer. "This is not a trivial achievement; the IEEE is one of the largest professional societies in the world," says Mostert. "Monuko du Plessis' standing helps to get us appointments and open doors around the world."
In addition, Du Plessis and his team have established a research alliance with the Georgia Institute of Technology, an institute with vast resources, infrastructure and influence. It is also spearheading the Darpa research into the interconnect dilemma.
"Ultimately the two parties may pool their complementary IP," says Mostert. " But it is early days and much of this must still be negotiated with Georgia Tech and approved by SA regulatory authorities."
Of course UP is aware of the importance of partnerships, too. "If we go it alone for too long we could lose out completely. There are very big groups involved in the same quest. We need big partners."
But first there is one significant technical hurdle to overcome.
The silicon-based technology is not yet sufficiently efficient in converting electrical power to light, or electrons into photons. This is the focus of current development work. "We have achieved some efficiency gains but we need to increase efficiency by a factor of 10 or 20," says Du Plessis.
There's a big incentive. The leading company in the US to develop integrated photonic elements with electronic circuits on one chip is California-based Luxtera. But its technology does not include a light source. The company has expressed interest in the InSiAva technology as a potential light source.
If the team can get this right, Luxtera may be but one of many suitors.
