March 1998

Go for the Glow

By David Graham

Astronomers at San Diego State University (SDSU) have made a name as builders of electronic light detectors for many of the world's great telescopes-among them the Keck telescope in Hawaii and the Hale on California's Palomar Mountain. Now these astronomers have aimed their technology at a new target: computer chips. Mounted on microscopes rather than telescopes, the detectors can find flaws in computer chips more easily-and potentially more cheaply-than existing methods.

In an industry where small technological improvements can make a big difference in profit margins, these detectors, which are sensitive to radiation in the infrared part of the spectrum, could have a significant impact.

"We think it should be quite useful," says Robert Leach, an SDSU astronomer who helped pioneer electronic imaging devices for telescopes in the 1980s with SDSU engineer Frank Beale. Recognizing the range of potential applications, Frank Low, president of Infrared Laboratories of Tucson, Ariz., began working with SDSU scientists in the spring of 1996 to develop infrared detectors for use in chip manufacturing. The infrared emission microscope they developed, known by the acronym IREM 1, went on the market last fall.

IREM 1 is descended from an infrared detector Low's company built that is now flying on the Hubble Space Telescope. As the microscope passes over the surface of a computer chip, any infrared radiation (heat) emitted by the chip collects in the 65,000 wells, or pixels, of a silicon wafer mounted at the end of the microscope. Sensors in each well measure the amount of light collected; in a matter of seconds, the information from these pixels combines to generate an image on a computer monitor.

The devices could solve a nettlesome problem for computer manufacturers, the efficient testing of new chips before they reach consumers. Even the tiniest flaw in a computer chip-perhaps a fleck of dust built into the circuitry or a place where insulation has eroded-can make electrical current jump between transistors, leaking heat and undercutting performance, Low explains.

To find such flaws, manufacturers put their chips through a series of tests, including running current through them and scrutinizing them with detectors of light. But since flawed chips give off more heat than light, the leakage of heat is much easier to detect. "In fact, the effect is fairly dramatic-the thing lights up," says Leach.

He suggests that chip manufacturers could use IREM 1 to scan for chips that are obviously flawed, saving more complicated and expensive tests for chips that pass this crude "first cut," and argues that the new device might lead to improvements in chip design by identifying recurring flaws in specific microcomponents on the chip.

Such improvements are "absolutely critical to the competitiveness of these companies," says Jeff Weir, a spokesman for the Semiconductor Industry Association, the principal trade association for American chip manufacturers. "Anything that makes it quicker and easier to find a faulty chip is important. Things that can expedite production are money makers."

According to Low, the inventors have already sold IREMs to two companies, one of which bought several of the devices. The first to buy them, one of the world's largest chipmakers, is testing the device in its production process. (Low declined to name the companies, citing confidentiality contracts.)

Infrared Labs is now planning a second-generation array that will be dramatically faster than the the IREM 1. With 16 times as many pixels-1.04 million compared with the IREM 1's 65,000-it will enable the device to view 16 times as much chip space at once.