by Hank Hogan
The world's semiconductor manufacturers are taking waste stream management to heart as new reclaim technologies begin to make environmental and economic sense
Often the result of government mandate, contamination control of airborne and waterborne waste is starting to make environmental and economic sense to the world's semiconductor manufacturers. As a matter of fact, waste streams are being turned into revenue streams in some instances.
Consider, for example, Texas Instruments Inc. (TI; Dallas, TX). TI's fabs chew through a lot of high-grade, expensive chemicals, and after use most of these are washed away by various deionized (DI) water rinses. TI recently installed systems to recapture chemicals such as sulfuric acid from the wastewater exiting the cleanroom. After extracting acids and other compounds from the water, TI sells the reclaimed chemicals to other industry users.
“We buy chemicals at a semiconductor grade, but when we get through using them they meet the specifications as a technical grade chemical,” says Laurie Lehmberg, environment and energy manager for TI's facilities worldwide.
It's a win-win-lose situation.
TI reduces its environmental impact and recovers some money. Those buying the reclaimed chemicals get a competitive deal. The only losers are the chemical companies, because a customer, TI, is now a competitor. But Lehmberg thinks that even chemical companies are ultimately winners in this end-of-the-line contamination control.
“Deep down in their heart, I think they can appreciate that it's the right thing to do,” Lehmberg says.
Goals and barriers
TI isn't alone in its quest to reclaim and reuse chemicals and water. Industry players such as the Intel Corp. (Santa Clara, CA), National Semiconductor Corp. (Santa Clara, CA) and LSI Logic Corp. (Milpitas, CA) are currently reclaiming and reusing or considering moving in that direction.
The consensus is that semiconductor cleanrooms in the Pacific Rim are far ahead of their US counterparts in regard to water reclaim and reuse.
Bill McClain, vice president of global engineering of microelectronics systems at ultrapure water systems-maker Ionics Inc. (Watertown, MA), says some Asian requirements for recovery and reuse are as high as 85 percent. This is largely due to government mandate, driven mainly by the capacity of the public infrastructure for both water and wastewater treatment.
Europe is also thought to be somewhat further along than the US in various pollution and contaminant abatement areas. But the feeling is that reclaim, reuse and abatement will become increasingly popular.
Sematech's (Austin, TX) 2001 International Technology Roadmap for Semiconductors shows the water recycle rate in state-of-the-art processes climbing from 60 percent to 70 percent by 2005.
At the same time, chemical use per area per process mask step is supposed to decline five percent per year. Finally, the net water feed over the same span is supposed to drop from 5.9 liters per square centimeter of processed silicon to 3.5 liters. The only way to achieve all of these goals, industry analysts and observers think, is by reclaim and reuse of the waste stream.
That may be the goal, but industry observers say there are hurdles to overcome first.
Some of these mainly involve economics and psychology. This shows up particularly in the case of water, because ultrapure water is expensive to produce and comes into direct contact with silicon during processing.
This water can be reclaimed to the point where it's actually of higher quality than water coming into the treatment plant. Indeed, there are those who say that such water, when piped back into the input stream, can stabilize and improve the output from the ultrapure water plant.
Proponents of this cite studies and case histories, but that doesn't sway everyone.
“It's kind of an uncontrolled situation when you have reclaimed water,” says Richard Banks, corporate environmental manager for National Semiconductor. “If the water has high organic materials in it, that can really damage your ultrapure water process. Then, if it gets into your semiconductor process itself, it can wreak havoc on your products. So there's been a lot of reluctance.”
“Why would you throw away water that is better quality when it leaves your plant than when it comes into your facility?” argues reuse proponent Roger Bonisolli, vice president of technology at design and construction focused Fluor Corp. (Aliso Viejo, CA). “The problem is a psychological barrier, not a technological barrier, in the reuse of treated process effluent in chip manufacturing.”
Divide and conquer
On the technical side, many of the challenges have been overcome. The key to reclaim/recycle/reuse is to design with this method in mind.
In practical terms, this means that the wastewater stream has to be separated according to what is in that particular stream. There may be several liquid waste streams coming out of a tool-one consisting of DI water, one destined for waste treatment and one that could be used as part of a reclaim/recycle stream.
These different outputs may be combined into streams carrying solvents, corrosives, fluorides or heavy metals. The combinations can be large and the breakdown needs to be exact.
“We segregated a large number of streams within the facility, all the way down to sulfuric and concentrated phosphoric. So we actually have a total of 13 streams coming out of the fab proper,” says Mike Bennett, senior facilities engineer at LSI's Gresham, OR, plant.
Designing in and implementing such segregation is a tactic that TI follows as well. According to both companies, there are several reasons for this.
The water flow of these partial streams is less than that of the total and the offending chemicals exist in higher concentrations. That makes treatment and removal of the contaminant much easier.
As for the air exhaust stream, that too follows a similar scheme. Again, a lower overall air flow makes pollution abatement simpler. That's because the waste stream spends more time in the treatment tool. However, in this case, there's an additional reason to divide and conquer. The exhaust from a semiconductor tool is often laden with gases that are reactive; and those gases also frequently carry compounds that contain silicon. So combining gases can be dangerous.
According to José Arnó, director of research and development at ATMI Inc. (Danbury, CT), a producer of point-of-use pollution abatement equipment for semiconductor tools, “There's a potential that either some bad reactions could happen or it could be that they create particles that could end up clogging the exhaust system.”
Water, water everywhere
With regard to the particulars of water reclaim and reuse, TI is widely cited as an industry leader. This is particularly true when it comes to the somewhat contentious issue of actually recycling water back into the fab itself.
Lehmberg says that TI reaps several benefits from its efforts in this regard. The quality of water going into the ultrapure water plant is higher and more consistent with this approach. Therefore the cost to operate the water plant is lower, the output is of better quality and the plant's uptime is greater.
We buy chemicals at a semiconductor grade, but when we get through using them they meet the specifications as a technical grade chemical.
-TI’s Laurie Lehmberg
TI recently won an award for innovative water engineering at its Dallas facility, a project where TI reconfigured its ultrapure plant, reducing the company's demand for water by several hundred gallons per minute and saving more than $550,000 a year.
In creating ultrapure water, reverse osmosis is used to filter out contaminants. The result is purified water on one side and contaminant-laden brine on the other. The brine is rejected and sent into sewage treatment, while the ultrapure water is used in the fab. In large reverse osmosis setups, there is typically a redundant, idle array. The TI innovation was to use that array to recover more ultrapure water from the brine before it is discarded.
TI is not the only semiconductor company reclaiming and recycling water. Intel is not as aggressive in its approach, but the company does reclaim water.
Terrence McManus is director of Intel's environmental, health and safety technologies. He notes that Intel uses fab rinse water for makeup water in cooling towers or for local landscaping.
Also, in its Chandler, AZ, factories, Intel takes wastewater from its semiconductor operations and sends it through a reverse osmosis treatment system. Intel built the system that is operated by the city of Chandler, an arrangement that satisfies an Arizona state requirement for zero ground water usage.
The plant, McManus says, “…produces an effluent that is drinking water quality. The city of Chandler reinjects that water into the ground water supply.”
For their part, ultrapure water treatment system vendors like Ionics have mixed feelings about the recycle and reclaim approach. On one hand, closing the loop can lead to better performance and happier customers. On the other hand, a closed cycle is more vulnerable to human error originating in the fab itself.
Gary Pitts, vice president of key accounts for Ionics' microelectronics center of excellence, notes the possibility that the wrong chemical could be dumped down the wrong drain.
“That's what the system has to allow for through divert valves and instrumentation and storage and post-treatment,” says Pitts. “The technology exists today to easily treat the recycle water for these mistakes, but we must be prepared for this occurrence.”
On a clear day
As for air abatement and treatment of exhaust, that can actually be helped by water reclaim. One of the most popular ways to reduce air contaminants to an acceptable level is through the use of a scrubber.
Many scrubbers use water to cleanse the air, and the water used does not have to be ultrapure. Hence, water reclaimed from the fab can be used to clean up the fab's exhaust.
According to ATMI's Arnó, the semiconductor industry is different than other industries when it comes to removing airborne contaminants and pollutants.
The cost of an abatement system ranges from $30,000 to $100,000 or more. For most industries that makes point-of-use methods too expensive to consider. Semiconductor tools, however, run upwards of one million dollars and semiconductor cleanroom airborne effluent can be toxic, flammable, highly reactive or all the above. Because of these factors, abatement is often done at the semiconductor tool itself.
There are three major technologies used for this chore. There's wet scrubbing, which employs water to get rid of acidic gases and some bases like ammonia. There's thermal oxidation, which uses an open flame or electrically heated furnace to burn the contaminant. And there's dry scrubbing, which sends exhaust through a canister filled with a material that chemically absorbs the gases.
For the most part, semiconductor exhaust is largely particle free-at least until abatement starts. Then, especially if the treatment is a thermal one, particles may be produced. Arnó notes the system must be engineered to handle large particle flow. Typically this is done using cyclonic separation.
As for the future, Arnó points to perfluor ocarbons (PFCs) as a concern. An example of this type of gas is carbon tetrafluoride (CF4). Used in semiconductor processing, CF4 has tens of thousands times the global warming potential of carbon dioxide. Perfluorocarbons are very high infrared absorbers and possess another characteristic that makes them particularly troublesome.
“On the abatement side, they're very hard to burn; they're very stable,” says Arnó.
One possibility is the use of a plasma. In this approach, electrons in the plasma shatter the molecule and break it apart. It's an efficient technique but it has yet to be deployed.
According to industry participants, there's interest in the effort to control greenhouse gases, but no overwhelming urgency. There is no government directive in the matter, although there is a voluntary agreement to reduce PFC levels by 2010.
A better approach might be to eliminate the source of the problem. In fact, research is underway to eliminate PFCs from semiconductor processing.
As this illustrates, the dynamic nature of semiconductor processing is both good and bad for effluent treatment. On the negative side, the constant change means that companies must deal with a constantly evolving waste stream. On the positive side, the rapid pace of change means that improvements to effluent treatment can be deployed as often as new technologies roll out.
As Intel's McManus notes, “We're almost on an every two year cycle. That means every two years I get to look at another one and do an improvement.”
Hank Hogan is a special correspondent to CleanRooms. He lives in Austin, Texas.