For years, the people who make semiconductors have been trying to unlock their industry’s next big technological leap. A fledgling company in Seattle thinks it may have an answer—and right now, it’s crammed into the basement of a University of Washington research lab.
The startup, called Zplasma, has a patented method for producing the high-powered light used to turn pieces of silicon into the powerful microprocessors that make modern computing possible. And, the company says, its method is more stable and more powerful than the technologies being used today.
Those claims are sure to face plenty of skepticism. Many other upstarts have made grand projections in this arena, only to disappear without a word, CEO Henry Berg says.
“In many ways, I feel like we’re crashing a party,” Berg says. “We’ve got an industry that’s been going down one road, and we’re saying, `Hey, you know that road you said was a dead end years ago? We think it’s not a dead end with this technology.’
“That’s an uncomfortable position to be in.”
Microprocessors are the basic building blocks of computing, harnessing pulses of electricity to perform millions of mathematical operations each second. They’re made by etching a bunch of electric circuits on a wafer of a carrier material, usually silicon.
Making them requires immensely complicated machines, and high-powered light is at the heart of it. The machines that etch wafers, called steppers, rely on far-out sounding contraptions to produce this powerful light. Shine it through a kind of stencil onto the surface of a silicon wafer, and it creates an intricate pattern of tiny channels that can then be turned into electrical circuits.
The microprocessor industry is constantly working to make its products denser to keep up with Moore’s Law, the famous prediction that says the number of transistors on an integrated circuit will double about every two years.
To keep up with Moore’s Law, the channels that are etched into the wafers have to keep shrinking in width, allowing more circuits to be crammed onto a chip. And that’s where chipmakers run into a big problem: the lasers used in today’s chip manufacturing produce light at a wavelength of 193 nanometers, too fat to keep making smaller patterns for too much longer.
Here’s how IEEE Spectrum put it: “Trying to use today’s ultraviolet lasers to print the next generation of circuits would be like trying to trace a fine line with a preschooler’s crayon.”
The industry has come up with some creative workarounds to get finer results out of the current lasers, but that bag of tricks is almost empty. So it’s settled on a new standard for the next generation of microprocessor production: extreme ultraviolet or EUV light, which has a much smaller wavelength of 13.5 nanometers. But so far, nobody’s been able to build a machine that can spit out a reliable stream of EUV light that’s also bright enough to make microprocessors at a profitable rate.
Getting EUV light in the first place takes some pretty intricate physics. The methods typically used today start with tiny droplets of tin that are zapped by a laser. That vaporizes the tin into a plasma, which emits the prized EUV light. An array of mirrors can then direct that light toward a silicon wafer.
This can be a messy process, with tin fragments along with electrons and ions that come flying off the plasma, possibly contaminating the delicate mirrors. Critically, this method of making EUV also doesn’t yet produce light bright enough to get up to production speed.
Commercial manufacturers need to produce one etched silicon wafer every minute to break even, Berg says, which will require light of about 100 watts—and manufacturers would actually like to double that to start earning a real profit.
ASML, one of the two major manufacturers of microprocessor machines, is the furthest ahead in testing existing sources of EUV light in its products, Berg says. But those sources, mostly supplied by San Diego, CA-based Cymer, aren’t up to the 100-watt level yet.
Cymer CEO Bob Akins recently joked that the pressures of getting brighter light out of his company’s EUV machines were taking a toll: “Two years ago I was six-feet-six. Now I’m only six-feet-four.”
Another method of producing EUV light has been largely left behind because it wasn’t hardy enough for microprocessor manufacturing. With that approach, a bolt of electricity is fired through a gas cloud, creating a plasma. Electromagnetic forces then compress the plasma down into something called a “pinch,” a volatile state that causes it to emit EUV light. The problem with this method is that the pinch can’t last very long, making it too undependable to be used in manufacturing.
That’s where Zplasma turns into the party-crasher. The company’s co-founders, UW professors Uri Shumlak and Brian Nelson, discovered a way to make these kinds of special pinches last longer—about 100 times longer.
In Nelson and Shumlak’s device, plasma flows through a long metal tube before being hit with electricity and turned into a light-emitting pinch. The key, however, is that some of the plasma is flowing faster before the electric shock, which changes the way the pinch forms.
The physics are pretty complicated, but Berg likens it to freeway traffic. “Think about if you have traffic merging, and you have different cars going at different speeds,” Berg says. “It will change the way the traffic behaves when it comes together.”
When it finally dies out, this pinch also falls apart in a more controlled manner, which means that it’s easier to control the debris that can otherwise contaminate delicate equipment. “What we’ve done is invent a pinch that’s stable, and no one thought that was possible,” Berg says.
And, crucially, Zplasma believes its technology is capable of producing much brighter light than competing light sources. While conventional EUV technology is still striving to produce light at 100 watts, the Zplasma device is designed to start out at 200 watts.
Turning this technology into a company, however, would take some more work—and money. That’s when Berg’s experience came in handy. The veteran technology executive, with stints at startups and Microsoft on his resume, had come to the UW for an entrepreneur-in-residence gig at the school’s Center for Commercialization. He was looking for promising technology to spin out of the university, and Shumlak and Nelson’s light source was right on the verge.
To get the technology out of the lab, they needed to show it would work on a much smaller scale. The apparatus Shumlak and Nelson had was about 10 feet long, and it created a relatively large pinch that didn’t emit light in the very small beam needed for commercial uses.
So Berg turned to the Washington Research Foundation, which gives grants to help universities and nonprofits get their technology into the market. Last fall, the foundation gave a gift to the UW to help turn Shumlak and Nelson’s research into an early commercial prototype.
“So now, gambling on success, we incorporated the company and negotiated a license with the university that transfers the exclusive rights to the patents and the prototype to the company,” Berg says. “And the prototype began producing light exactly as hoped for in February.”
Zplasma is now an open secret. Berg is out talking to leaders in the semiconductor industry in search of possible partnerships to develop the Zplasma technology even further.
If larger companies aren’t ready to bankroll the next phase of work, Berg hopes to get a list of milestones that manufacturers want to see before taking on the technology, which could serve as critical proof of interest for possible venture investors.
Young as it is, this company thinks it’s spent plenty of time in the lab. The Zplasma team is restless to get down to business.
“We have a technology that we think solves a tremendously important problem that the semiconductor industry is currently facing,” Berg says. “We don’t want to build a product as a technology demonstration. That’s not what this is all about. We want to see this used to make chips in high volume.”
