Home

UW startup holds key to the future of microprocessing

The Zplasma prototype will produce light with enough power to manufacture next-generation silicon microchips. Photo by Features

Moore’s Law states that the number of transistors on a microchip, a pattern of channels that can be turned into electrical circuits, doubles every two years, resulting in more powerful, efficient, and speedy computer parts.

The law rendered the incredible advances in computing possibilities during the past few decades to create better and brighter microprocessors, which enable computers to run the millions of mathematical operations a second needed to function.

However, for several years, the semiconductor industry has been working around an integral production problem: The ultraviolet light currently used to produce the circuits crammed onto chips is stuck at 193 nanometers in wavelength, which limits the possible size and number of etchings on a chip. In effect, Moore’s Law is in danger of being proven incorrect, and chip size could plateau, stagnating advances in computing power.

“The industry hit the wall in production 10 years ago,” said Henry Berg, CEO of Zplasma, a UW startup intent on continuing to prove Moore’s Law correct. “They’ve been working around it with creative and incredible engineering, but the problem is still there.”

Berg and two UW scientists think they have a solution to the problem. Uri Shumlak, professor of aeronautics and astronautics, and Brian Nelson, research associate professor of electrical engineering, were working on a project with plasma — hot, charged gases in which electrons are stripped from their nuclei resulting in high-energy material. After running the experiment with xenon gas, an astonishingly bright light resulted, indicating the product’s strength.

“The engineer looking in literally said ‘Wow, that was really bright,’” Nelson said. “Its [power and duration] … was really striking.”

After searching for an application for this bright light, Shumlak and Nelson landed upon chip manufacturing as a possible candidate. The microchip industry needed a light with a wavelength of 13.5 nanometers, less than one-fourteenth of the current minimum achievable size, as a future standard in manufacturing.

This extreme ultraviolet light (EUV) can be produced by plasmas, which are difficult to stabilize. At present, molten tin droplets are shot with lasers to produce plasma that sparks light. A multimillion-dollar optical machine then employs mirrors to reflect light onto a silicon wafer, etching patterns through a mask to layer the correct patterns. The silicon wafers are then chemically treated to complete their transformation into microprocessors.

“They lose a lot of efficiency and energy trying to control that instability,” Shumlak said. “The method we had produces plasma that is more stable and lasts longer.”

Henry Berg got involved in April 2011, when researchers began filing patents for the product. The light produced from the UW fusion reactor lasted up to 1,000 times longer than other beams of light, but the question for the new discovery ­­­is this: Could it be narrowed to the 13.5 nanometers needed to continue Moore’s Law?

“I did a lot of research on the commercialization,” Berg said. “We needed to make a viable prototype that could demonstrate everything the industry needed in order to get funding and get it to the market.”

Berg’s diverse background — he served in the Navy, then obtained a Ph.D. in computer science from Stanford, and most recently became an entrepreneur-in-residence at the UW — led to his enthusiastic response to Shumlak and Nelson’s concept and eventual entrepreneurial role in their work.

A grant from the Washington Research Foundation allowed the team to make specialized equipment to help provide a 200-watt commercial light source small enough to be compatible with industrial-scale optical machines. The first fully working prototype created the right light in February 2012.

“We have the right size, right dimensions, all the right characteristics,” Shumlak said. “We’re only missing appropriate power supply at this moment, which will come with enough funding to operate. Then the next step is to move into a commercial lab to [make the] product.”

Berg, Shumlak, and Nelson negotiated with the UW to obtain an exclusive license to the technology in order to form Zplasma in October 2011, with Berg taking on the role of CEO. The UW is the official patent holder, yet Zplasma retains the right to manufacture and implement the product.

“[Our role] depends on how the company develops,” Shumlak said. “Every step has been exciting, from the first generation of the light.”

Zplasma is in the process of raising $5 million in order to create an EUV light source that consumes 40 kilowatts of power, runs at 5,000 hertz, and delivers 200 watts of EUV light to a device used to create circuitry, Berg said.

“We need to partner with the right people now,” Berg said. “Not only people who have the money, but also people who can put the technology into industrial processes.”

Shumlak, Nelson, and Berg also plan to establish the device on a repetitive scale so that it can pulse the light thousands of times, making it industry-standard. That’s where the $5 million comes in: generating enough power to accomplish the repetition in addition to their other goals is very costly.

The team is trying to get a head start on production according to Berg, but by planning as comprehensively as he speculated, it may take two years for full function, and then another two years to hit the factory floor in the industry.

“We can and want to provide the key new element that produces the light [needed by the computer science industry],” Berg said. “That’s our goal.”

Reach reporter Garrett Black at science@dailyuw.com.

Twitter: @garrettjblack

Please read our Comment policy.