About Cymer (old Forbes article but still interesting) JGH
If Moore's Law holds for the next few years, much of the credit will go to Cymer Inc.'s Bob Akins and Rick Sandstrom.
Laser dudes
By Josh McHugh
ROBERT AKINS checks his watch, frowns and looks around the Fortune Cookie Restaurant for our waitress. It's been ten minutes since he ordered. The 45-year-old cofounder and chief executive of Cymer Inc. is getting anxious.
Pale, gaunt, 6 foot 6, Akins looks as though the last thing he needs is to miss a meal. But two minutes and a soft-spoken apology later, Akins is gone, back to his office, leaving his chief financial officer, a reporter and a bewildered waitress in his wake.
Forgive his impatience. For ten years Akins and Cymer have quietly tinkered with lasers in dusty northern San Diego, subsisting on a meager diet of government assistance, investments by a handful of foreign companies and sporadic shots of venture capital. Now, suddenly, equipment suppliers to Intel, NEC and Motorola are breaking down Cymer's doors to get its ultraviolet lasers.
Orders for these lasers, which print tiny circuit patterns onto silicon to make microprocessors and memory chips, are up 350% over the past year. The company's sales have exploded from $18 million in 1995 to $65 million in 1996. Cymer sold stock to the public for the first time in September, at $9.50 a share. Three months later it sold more at 42, making it the year's hottest new issue. These days Cymer trades at 50, making Akins' 2% stake worth $12 million.
Behind this action is Moore's Law, which postulates?pretty accurately, over the past 30 years?that chipmakers can double the number of transistors on a microchip every 18 months. For example, Intel's Pentium, vintage 1993, had 3.2 million transistors; its Pentium Pro, released about two years later, uses 5.5 million transistors.
Moore's Law, Corollary 1: The length and width of a transistor must shrink 29% every 18 months. Corollary 2: Tall, gaunt physicists who know about ultraviolet light get rich.
We left out a few steps between Corollaries 1 and 2. Here they are:
In the chip factory, a round, 8-inch wafer of crystalline silicon is coated with light-sensitive chemicals and put into a machine called a stepper. The stepper sends pulses of light, typically a thousand times a second, through a stencil depicting the circuit pattern of the nascent chip, onto the wafer. The chemical coating on the wafer reacts where the light touches it. That pattern in turn determines where lines will be etched on the wafer in a later step.
But here's where the laws of optics become a stumbling block for Moore's Law: The circuit lines can generally only be as thin as the wavelength of the light that shines through that stencil. So to keep increasing the density of transistors on a chip, engineers must shrink those wavelengths.
Originally chipmakers used visible light. By the late 1980s, chipmakers turned to the shorter wavelengths of invisible ultraviolet light. Ultraviolet light emitted by hot mercury gas makes lines as thin as 0.35 micron, a micron being one 25,000th of an inch.
That's pretty thin. But Moore's Law now says it's time to get shorter wavelengths than mercury ultraviolet. It's time for a Cymer excimer laser (see box) . Those photolithography lasers aren't easy to make. Cymer is one of three companies in the world capable of making them, and last year it had more than 80% of the market.
It's a far cry from the summer of 1985, when Akins and fellow physicist Richard Sandstrom tossed Frisbees on the beach and slurped beers in the Belly Up, a Southern California bar near Del Mar. Under discussion: whether or not to open a Wendy's outlet.
Akins and Sandstrom both had Ph.D.s from the University of California at San Diego. Akins' was in optical information processing?a still-impractical form of computing in which light replaces electricity. Sandstrom's specialty was in the behavior of laser light as it travels through the atmosphere.
After receiving their degrees, they took jobs with San Diego defense contractor HLX Inc., where they worked on esoteric projects including laser-induced nuclear fusion, a satellite-to-submarine laser communications link and the ill-fated Strategic Defense Initiative. After a half-dozen years, both decided to quit.
But Wendy's? "When the technology starts to get overwhelming," deadpans Sandstrom, "you want to be a farmer."
"We wanted to spend the productive years of our lives doing something more aligned with the real world," says Akins. "It was 1985?fast food franchises were hot."
Fortunately for the semiconductor industry, the two founded Cymer Laser Technologies in 1986 and set about building lasers. They brought in another colleague from HLX, Uday Sengupta, to handle marketing.
"We didn't have the money to see our way through the development effort if we were going to sink it into having components made," Akins recalls. So, with a little help from their friends in the UCSD machine shop, he and Sandstrom built their own laser, working the lathes and milling machines.
"We'd go in on a Saturday morning at 9 a.m. and come out at 7 or 8 p.m. with one set of electrodes," says Akins. It was hard work, but the price was right. "When it came time to pay for the equipment and materials, the guy who ran the shop told us, 'I'll just keep this thing [the invoice] in the bottom drawer until you guys get the money to pay for it.'"
Cymer had roughly $250,000 in unpaid tabs to the university by the end of 1986. "That was our principal seed financing," Akins chuckles. The company was able to coax funds out of the government's defense research budget, as well as technical and financial help from Sematech, the Austin-based semiconductor consortium. Akins and Sandstrom both took out second mortgages on their homes to keep Cymer afloat.
In 1988 Cymer repaid UCSD when it received its first real outside investment, from venture capitalist Richard Abraham. A researcher and factory manager at Fairchild Semiconductor, Motorola and Texas Instruments, Abraham knew the semiconductor industry well, and liked what he saw in the fledgling company. But his money came with strings attached: Cymer had to focus entirely on semiconductor applications for its lasers.
If Cymer can make enough lasers, it should be a half-billion-dollar company in four years. If not, it will have squandered one of the great opportunities of all time.
At the time, chipmakers assumed they would jump all the way from mercury light to X rays, which have smaller wavelengths than visible or ultraviolet light. Abraham figured there would be a stopping point along the way for deep ultraviolet. He was right. X rays have so far proved too costly and difficult to use in chipmaking.
In 1988 a delegation of scientists and executives from Canon, a stepper manufacturer, paid Akins and Sandstrom a visit. Soon after that, a group from Nikon, another Japanese stepper-maker and a fierce competitor of Nikon's, dropped by Cymer's modest laboratory. Nikon and Canon teamed up to buy 6% of Cymer.
It looked as though Cymer might have its big break around 1990. Light from mercury bulbs, with a wavelength of 0.48 micron, was expected to be useless for drawing circuits thinner than 0.5 micron. But by coating wafers with new high-resolution photoreactive chemicals, chipmakers have been able to draw lines thinner than the light's wavelength, down to the 0.35 micron features on today's cutting-edge chips.
Luckily for Cymer, two years ago chip and stepper companies realized that that technique was about to hit a wall. Thus began the switch from mercury light to excimer laser light. Cymer's krypton fluoride laser delivers a pure 0.25 micron beam.
The semiconductor industry began to slump in late 1995. But paradoxically the slump increased demand for Cymer's next-generation technology. Chipmakers decided that the only way to shore up their shrinking profit margins was to move ahead. "Moore's Law is more of an economic imperative than anything else," says Akins. "If the industry wasn't in the doldrums, this wouldn't be happening."
Cymer expected stepper manufacturers to follow an established buying pattern, ordering just a few of the lasers for testing, then buying large numbers nine months later. Didn't happen. Demand from chipmakers forced the stepper companies right into high-volume buying soon after Cymer launched its $450,000 model.
With the $80 million raised in its two public offerings, Cymer has built a 137,000-square-foot factory in San Diego and is forming a laser-making partnership with Japan's Seiko Instruments. But whether Cymer can keep up with frenzied demand for its lasers and keep its customers happy remains to be seen.
"We're experiencing the fastest growth of any semiconductor equipment company?ever," Akins says. Far from bragging, he's feeling the full pressure of the $132 billion semiconductor industry bearing down on him. "If we don't perform, the whole industry gets slowed down," he says. "We're a bottleneck."
If Cymer can make enough lasers and keep the lion's share of the market to itself, as it has so far, it should be a half-billion-dollar company in four years. If it can't keep up with demand, and if the industry goes somewhere else for lasers, Cymer will have squandered one of the great opportunities of all time. Either way, the waitresses will have to be fast if they want to serve Bob Akins.
A beam is born
CYMER'S ultraviolet excimer laser beams are generated inside a 2 1/2-foot long aluminum chamber. Tanks near the chamber pump in a mixture of gasses, mostly krypton and fluorine. A power unit on top of the chamber delivers a 12,000-volt charge across two electrodes inside.
The charge, which lasts for 75 nanoseconds (billionths of a second), excites the krypton atoms enough to get them to couple with the fluorine atoms. A krypton fluorine molecule thus formed is known as an excited dimer, hence the term "excimer."
When the charge stops, the molecules break apart, releasing a burst of heat and deep-ultraviolet light. The light shoots out of a window at one end of the chamber and through a series of mirrors and strips of aluminum, which purifies the beam. The beam enters the stepper machine, then passes through a stencil and emblazons circuitry onto the silicon wafer.
Why is Cymer's laser so hard to make? It has to run for months at a time while handling 1,000 jolts of electricity per second. The chamber's electrodes have to be shaped and spaced perfectly to create the right "plasma" of excited molecules. The apparatus assures most of the wavelengths in the beam hover around 0.25 micron.
That slice of ultraviolet is suitable for etching lines just 0.25 micron wide on a silicon chip, making possible another leap forward in the density of transistors on a chip.
The next state of the art in chip manufacture calls for lines 0.18 micron wide. Cymer physicist Richard Sandstrom is hard at work on an argon fluoride laser that would draw lines at 0.17 micron, just barely small enough.
?J.McH.
|
