Jun Yan peered from a hood covering all but his eyes. Visitors to this clean room, where game-changing research in semiconductor manufacturing occurs, must slip into a "bunny suit" and pull on latex gloves to guard against introducing contaminants – dust, skin cells, anything that could foul the infinitesimal circuitry of a microchip.
Yan, a post-doctoral researcher at the University of Arizona, flipped a switch and a turntable began to spin in what looked like a fish tank. Yan had built the setup by modifying donated equipment used for a completely different process. "That's research," he said "We do a lot of crazy things."
Sitting on the turntable was a post card-size object studded with circuits and trailing wires. Yan's creation, fashioned in another part of the lab, is a sensor that can answer a question that is vital to the semiconductor industry: When is clean, clean enough? When can rinsing to eliminate impurities stop?
The implications for the industry, which measures water savings in terms of the flow of Niagara Falls, and for society are formidable.
Yan's work represents one element of the research conducted by the Engineering Research Center for Environmentally Benign Semiconductor Manufacturing. The center, founded in 1996 by the National Science Foundation and the Semiconductor Research Corp., a consortium of companies and university research groups, is headed by Farhang Shadman, University of Arizona Regents professor of chemical and environmental engineering. Sematech, another industry research consortium, is a sponsor. Along with the UA, where the center is based, other founding universities are MIT, Berkeley and Stanford. Eleven other universities, including Arizona State University, are members. Corporate core members include Intel, IBM, Advanced Micro Devices, Samsung and Applied Materials.
"Why semiconductors?" Shadman asked at a colloquium in mid-October. He cited Moore's Law, named for Gordon E. Moore, a co-founder of Intel who suggested in 1965 that the number of transistors per chip would double every two years. Semiconductor circuitry is now on the nano scale, far closer in size to an atom than the width of a human hair. Intel's latest chip is expected to hold 2 billion transistors.
In 2005, the Semiconductor Industry Association estimated, the industry made more than 90 million transistors for every man, woman and child on earth; by 2010, it said, the number should be 1 billion transistors. In 2008, the industry employed more than 216,000 people in the U.S. and made $249 billion in worldwide sales.
Along with those benchmarks came staggering demands for energy and water and imposing environmental challenges. Shadman and other leaders in the field recognized that the industry had to embrace sustainability practices in order to continue to fulfill all that Moore had envisioned.
"Sustainability is like peace," Shadman wrote in the October colloquium presentation. "It enjoys a universal appeal and full agreement. The disagreement is only in what it means and how to implement it."
The center's work led to advancements in reducing water and energy consumption. In 2000, the Semiconductor Research Corp. gave Shadman its Landmark Innovation Award for his advanced research in semiconductor technology, which it said had improved the environmental safety of semiconductor manufacturing processes and saved the industry millions of dollars.
As semiconductors have gotten smaller and smaller, the need to eliminate impurities has grown. The lab found ways to extract toxic substances from discharge and recharge the groundwater. Intel estimates that over the past 10 years its 700-acre Ocotillo campus in Chandler has saved more than 34 billion gallons of water, the equivalent, by its estimation, of all the water flowing over Niagara Falls in 11 hours. The plant, recognized by the Environmental Protection Agency as one of six winners of its 2007 Water Efficiency Leader Award, was designed to use 75 percent less water than the industry standard.
Shadman estimates that semiconductor manufacturing plants now use about 50 billion gallons of ultra pure water a year.
Ultra pure water, Yan explained, is water that is treated with reverse osmosis, ion exchange and UV degassing to reduce bacteria to one part per cubic meter. The process requires a lot of energy – all the more reason to reduce water use.
There are more than 200 rinse steps to eliminate contaminants in semiconductor manufacturing, Yan said, "like doing your laundry."
How much water? "In the Phoenix area, there were 12 semiconductor plants in the 1990s," Yan said. "What they were using equals the consumption of 2 million people." Because of the nano scale of today's semiconductor chips, it's very difficult to tell if all the impurities have been removed. "So we over-rinsed," he said. "Be on the safe side – who cares?" But then the impact became pronounced – many of the plants are in arid areas of the West. "Our duty is to bring their water usage down," Yan said.
The Electro-Chemical Residue Sensor, which can track contamination by measuring electrical impedance, makes over-rinsing unnecessary. The sensor works in a batch rinse with a group of wafers in an immersion bath, or can be used to develop the right rinse recipe for an individual wafer. "Our preliminary results predict 50 percent to 65 percent water reduction," Yan said.
In another lab, Yun Zhuang, another post-doctoral researcher, works on another thrust of the semiconductor center: finding efficiencies in wafer polishing to save water and chemicals.
Silicon wafers, record album-like discs of 200 mm or 300 mm (12 inches), are built up in layers until fabrication is complete and the wafers are sliced into chips. The process is exacting.
"You have to remove excessive material from the surface. Just like building," said Zhuang, who works with Ara Philipossian, professor of chemical and environmental engineering. "A skyscraper – every floor has to be flat first, otherwise you can't build them up. Our lab is to make each floor flat." If the layers are not flat within the nanometer level, lithography, the optical printing of circuits, will fail.
The process is called chemical mechanical polishing. Chemicals are used to soften the surface, then mechanical force is used to remove it.
Another group at the center researches additive wafer fabrication, a process that saves materials and energy and lowers costs. Additive fabrication entails placing materials only in places where they are needed – the opposite of the subtractive processing, in which materials are built up in solid blocks or layers and then scraped away.
See Li Lie, a grad student in chemical engineering who works with Anthony J. Muscat, associate professor of chemical and environmental engineering, also researches gas-phase cleaning and surface modification.
Gas phase processing is more environmental benign, said Lie, because less chemical is used. "I don't think we can really make a process completely additive, but at least we can eliminate some steps and minimize the consumption and the waste," she said.
Her group is also studying atomic layer deposition, which deposits an oxide film layer by layer "We can control the growth at Angstrom levels," she said. "We're trying to understand the fundamentals of this so that we can engineer and control the surfaces."
Lie is also working with supercritical carbon dioxide, which, at high pressure and high temperature, is neither gas nor liquid. "It can enter through very fine structure, can create very fine porous structure," she said. It has very low surface tension, so it's helpful in cleaning very small devices and can replace harmful chemicals in deposition and cleaning stages. The process, developed by Muscat, earned him recognition from Scientific American in 2003 as one of the
50 people in the nation who contributed the most that year to the advancement of technology in science, engineering, commerce and public policy.
"To make nano-manufacturing sustainable and green," said Shadman, "we need much more than improvements in individual processes and materials; we need change in manufacturing approach."