"We employ a process that's been around for a long time in microelectronics - chemical-vapour deposition," explained Hoenlein. "We also use a catalyst patterned by a lithographic method employed in microelectronics for a number of years. Putting these two processes together enables us to grow nanotubes at specific sites with specific characteristics: that was the great achievement."
Infineon says that its technique can lay down a uniform coating over the whole of a 6 to 8 inch diameter wafer. Key to the whole process is the detailed catalyst design - get that right and you'll grow nanotubes. What's more, the system produces an almost 99% yield of pure nanotubes, whereas other methods can create a large amount of soot. "The method we apply gives us a high yield of perfect carbon nanotubes," said Hoenlein.
The catalyst itself consists of "basically iron, nickel and cobalt, and mixtures of them," according to Kreupl. "We cannot go into details - it's one of the centres of our work."
To grow multiwalled nanotubes, the scientists use a mixture of hydrogen and acetylene as the source gases, whereas for single-walled nanotubes, the parent gases are hydrogen and methane.
From the point of view of integrating the process into a microelectronics environment, the lower the deposition temperature at which the researchers can produce high-quality nanotubes, the better. Indeed, the Infineon team is currently investigating decreasing that temperature by coupling energy into the system from a plasma.
The scientists also want to grow nanotubes onto substrates other than silicon. So far, apart from silicon, they've deposited the tubes onto tantalum, tantalum nitride, molybdenum and polysilicon surfaces. They're investigating a large number of metals but say that not all of them are promising. "Growing nanotubes on metallic surfaces is very important," explained Hoenlein. "But the catalyst normally reacts very strongly with the surface and this will result in different growth behaviour."
Using the tube
So just what is the point of all this hard work? "We have basically two applications in mind," explained Hoenlein. "One is what the other groups are doing - carbon nanotube transistors - and the other aims to improve the conductors in a chip. This means that the nanotubes would be applied in the first case in their semiconducting form and in the second case in their metallic configuration."
Hoenlein, Kreupl and colleagues are "largely engaged" on this second application at the moment. The main advantage is that carbon nanotubes can carry a much higher current than ordinary metallic interconnects - up to 1000 times as much as copper, for example.
"This is very interesting for the interconnect itself and also for the so-called vias that are interconnects between two metallic layers in a chip," said Hoenlein. "These are very prone to deterioration by electromigration - material transport due to high current densities. It's a preferred spot for the breakdown of the chip." So replacing a metal via with one made of carbon nanotubes could be a considerable advantage.
As yet, the Infineon team hasn't investigated the lifetime of such a via in detail - they've simply made electrical assessments. But Hoenlein says that other groups have shown that carbon nanotube conductors can have much longer lifetimes than ordinary copper lines.
And what about the transistor application? Since carbon nanotubes are about 1-2 nm wide, placing them between two contacts - a source and a drain - and applying a gate contact results in an extremely small transistor. That could help increase chip densities. And since nanotubes can carry high currents, it may also be the case that carbon-nanotube transistors are superior to very small silicon transistors.
Time for business?
That said, Hoenlein reckons it's difficult to put a date on commercialization. "We are researchers and we cannot currently foresee when we will have the transistors in a commercial application," he said. "But as far as our vias are concerned, we think that we can offer an integration process to our business groups within 3 to 5 years from now."
Part of the reason that it's easier to forecast the arrival of the commercial nanotube via is that vias require a supply of carbon nanotubes with a much lower reproducibility than transistors do. Today's production techniques can't make a sample of nanotubes that's entirely metallic or entirely semiconducting: there's always a proportion (around 30%, say) of the other type present.
"If you use multiwalled nanotubes in the vias, you can of course bypass the semiconducting ones and you do not need a 100% yield of metallic tubes," explained Hoenlein. "Whereas in the transistor application, we have to have a 100% yield of semiconducting single-wall nanotubes with identical characteristics. This is much harder to achieve and has yet to be realized by any group in the world."
• Wolfgang Hoenlein is senior director and Franz Kreupl is senior staff expert in Infineon's corporate research, nano processes group.