BBC News
watch One-Minute World News
Last Updated: Wednesday, 26 February 2003, 10:12 GMT
Computer chips pushed to edge
By Mark Ward
BBC News Online technology correspondent in San Jose

The job of keeping up with the prediction by Intel founder Gordon Moore that computer power doubles every 18-24 months used to be quite straightforward.

Computer chip
Chips have been steadily falling in size
Computer chips are made by beaming light through a stencil, or mask, that casts an image of a processor circuit design on to a silicon wafer.

The size of the components on a chip dictates the speed at which it runs, so a quick route to making more powerful processors is to shrink the components.

Until the mid-1990s, Intel and other chip makers kept proving Gordon Moore right by exploiting advances in lasers and lenses to make smaller components and more powerful chips.

For the past five years, however, Intel has had to rely on very different ways of making faster chips because advances in lasers and lenses have not been keeping pace with the iron dictates of Moore's Law.

Shaping silicon

The lenses used to focus light on to silicon wafers have got as big as they are going to get and lasers that use light with very small wavelengths are too expensive to build and run.

Silicon wafer, Intel
Tricks to shape smaller features onto silicon wafers
Instead, Intel has turned to advances in the masks, or stencils, used to form the image on the silicon substrate that will eventually find its way into a desktop computer, server or cellphone.

Dr Barry Lieberman, engineering manager at Intel's mask operation, said he and his colleagues were tasked with finding ways to make masks that could shape ever smaller features on to silicon wafers.

"As we do not get on the wafer what we ask for on the mask, the trick in lithography is to ask for something different," he said.

One trick goes by the formidable name of optical proximity correction. It tries to ensure features on silicon wafers appear close to their ideal form by adding nodules or bumps to corners.

Light shone through a mask using this trick will produce fuzzy features on a wafer but it gives the resulting components more definition than they would have if the bumps were absent.

Dr Lieberman said masks used to project patterns on to wafers to make chips with components 90 nanometres (billionths of a metre) across had more than a trillion elements.

Because masks are the master stencil for production runs of chips they had to be free of all defects more than one micron (millionth of a metre) in size, said Dr Lieberman.

He said the ratio of defect sizes that must be found to the overall surface area of the mask was similar to spotting a single basketball in an area the size of California.

Keeping Moore going

Intel is now starting to work with a novel form of mask that uses extreme ultra-violet light to form components on wafers.

Gordon Moore
Moore predicted computer power doubling every 18-24 months
Using light of this wavelength has forced Intel to change the way that patterns are created on wafers.

Before now, light shining through a mask first passed through a silica substrate before reaching the silicon beneath.

However, extreme ultra-violet is absorbed by almost every substance so Intel is using molybdenum-silicon mirrors beneath a mask to project a chip design on to the wafer.

The molybdenum-silicon mirror is made up of many layers, each one of which reflects back some of the light shone on to it.

This technology can form components only 32 nanometres across and, said Dr Lieberman, will help Intel keep up with Moore's Law until at least 2007.

The BBC is not responsible for the content of external internet sites

Has China's housing bubble burst?
How the world's oldest clove tree defied an empire
Why Royal Ballet principal Sergei Polunin quit


Americas Africa Europe Middle East South Asia Asia Pacific