Small is getting smaller


Dr John White explains the latest innovations in microchip manufacturing – and how UCD is involved.

IN THE SEMICONDUCTOR business, small is good. Consider the incredible shrinking size of integrated circuits in mobile phones, cameras and computers. To continue expanding the multi-billion dollar chip business, however, small has to get even smaller.

Just how they make integrated circuits so small is the focus of a recently funded Science Foundation Ireland research project in the UCD School of Physics, headed by Professor Gerry O’Sullivan.

In a process similar to etching a pattern on a lithographic print, light is used as a paint brush to etch transistors and integrated circuits on a silicon chip. The shorter the wavelength of the light, the thinner the etching lines, and therefore more transistors can be packaged on the chip.

What began with 1000 transistors in the 1103 memory chip of the 1969 Apollo 11 flight, now numbers over two billion for Intel’s latest Core I7 due to ship later this month.

Over the same period, the utilised wavelength of light has decreased significantly; from a yellow 578-nm line produced by a mercury discharge lamp, to an ultraviolet 193-nm line produced by an argon fluoride laser. Then the light is further reduced by optics to reach sub-micron sizes – yielding the resultant high-density, high-power chips.

“The UCD School of Physics has been doing fundamental research into light sources for over 25 years.”

But can the ‘light-shrinking’ process continue beyond the UV into what the industry calls ‘extreme ultraviolet’, or EUV? So far, the process has kept a torrential pace – dubbed Moore’s Law, after Intel’s co-founder Gordon Moore. He predicted in 1965 that the number of components would double every 18 months with corresponding reduced consumer cost. But can this be sustained?

At Intel’s Irish plant in Leixlip, Co. Kildare, plans for continued reduction include immersion lithography, where the light wavelength is shortened by passing it through water. After that, however, as Dr Tom McCormack, a former Intel engineer and now lecturer at UCD comments, “Lasers can’t be used, because of diffraction problems from the coherent light.”

Here industry meets university, where the Atomic, Molecular, and Plasma Spectroscopy group in the UCD School of Physics has been doing fundamental research into light sources for over 25 years – specifically, the study of atomic spectra induced by lasers.

This gives them an advantage in studying EUV, and thus they are well-placed to create technology, to use the next generation wavelength of 13.5 nm. At a recent workshop, Vivek Bakshi, founder and president of EUV Litho stated, “EUV source technology has many positive features that make it an attractive choice for high-volume microchip manufacturing.”

At UCD, the required EUV light will be created by using a laser to produce a plasma. As the excited atoms recombine, they give off light at specific wavelengths, one of which is in the coveted EUV. Computer simulations will also aid in the search for the ultimate 13.5-nm industry set-up.

Padraig Dunne, Associate Professor within the UCD Spectroscopy group, comments about the task ahead, “Imagine holding a welder’s torch one centimetre from a lens to focus the arc light, and then multiply that by five without trying to melt anything in the process. That’s the kind of power required.”

In fact, the light source power requirement is greater than the radiant output from parts of the sun – an incredible challenge, in order to keep up to the consumer’s ever-expanding need, for an ever-shrinking microchip.

Of course, eventually, small won’t be small enough as transistor sizes shrink towards atomic dimensions, but hopefully by then the next big thing will be ready—the realm of quantum computing.

Dr John White is a lecturer and researcher in the UCD School of Physics Atomic, Molecular and Plasma Physics group.