"If you look at the very early studies of chemical carcinogenesis," says Dr. Steingraber, author of "Living Downstream," "a lot of them were done by researchers who were industrial toxicologists, who might have originally worked for an oil company or something like that. They're right on the front lines ... When they have the courage and integrity to publish their findings, that's some of the best science that we have showing the relationship between chemical carcinogens and cancer."

Harrison's catalog of health risks is staggering. He lists potential exposures of workers to arsenic in the manufacture of gallium-arsenide wafers; to acid aerosols in the "wet etch" stage of chip lithography; and to toxic gases of arsine and boron in the operation of dopant implantation tools. He attests to cases of hydrofluoric acid burns during the cleaning of furnace tubes; of exposure to corrosive solvents in wet-stripping processes; and of untested photoactive compounds being sprayed by photoresist spinners. He warns of "catastrophic accidents" in the replacing of gas cylinders and the draining and refilling of wet chemical baths; of malfunctioning ventilation systems; and of widespread respiratory complaints among workers, including sinusitis, laryngitis and asthma. He documents mercury exposure from arc lamps; "relatively frequent" chemical fires at storage sinks; and solvent overflows in tool exhaust systems.

Harrison begins his article with an extensively diagrammed treatment of what remains the most worrisome -- and least acknowledged -- pathway of exposure in clean rooms: the vaporized mix of organic chemicals recirculated by the ventilation systems. A rule of thumb proposed by Harrison is that 90 percent of the air in a clean room is recirculated per hour, to minimize the introduction of contaminants that might degrade semiconductors or other advanced technological fabrications. He also shows how fumes can enter into circulation through "service cores," where vapors escape during equipment maintenance and where chemical spills are most likely to occur.

	Living Downstream By Sandra Steingraber Vintage Books 374 pages Buy it Imperial San Francisco By Gray Brechin University of California Press 428 pages Buy it Print story E-mail story

On top of that, recent evidence suggests that 15 percent of new fume hoods -- the local exhaust system for clean-room workstations -- fail to operate properly, potentially blowing toxic vapors back at the worker and into the clean-room environment.

"The ventilation conditions in clean rooms are very turbulent, and they cause a lot of problems," says Tom Smith of Exposure Control Technologies, a business that tests and evaluates laboratory ventilation systems. "Fume hoods [designed for the microelectronics industry], when we've tested them in clean rooms, generally only have a capture effectiveness about six inches above the work surface. If you get above that, or if you have a very volatile process, they just spill. And the clean-room airflow is so turbulent that it competes with these hoods, and the vapors escape from these hoods and infiltrate the return air system and are recirculated with the air handler."

And there are, without question, plenty of chemical vapors that can escape into the air system during the manufacture of a single computer chip, beginning with the pulling of a silicon crystal to the apotheosized "metallization" of the wafer -- the industry's term of art for deposition of electrical connections of aluminum on silicon. Figures based on a speech by a Texas Instruments fellow at the International Symposium on Semiconductor Manufacturing in September 1993 estimate that Intel's state-of-the-art chip fabrication plant in Rio Rancho, N.M., consumes, in a single year of manufacturing, 832 million cubic feet of bulk gases, 5.72 million cubic feet of hazardous gases and 5.2 million pounds of chemicals.

These figures, though prodigal, are deceptively simple, for they do not indicate the unprecedented spectrum of chemicals used in semiconductor manufacturing. In his 1992 article, Harrison prefaces a section titled "Selected Toxic Hazards" with the disclaimer: "An attempt to review the toxicology of all the thousands of chemicals in use at a typical fabrication plant is doomed to be superficial and of little value."

And the acceleration of the use of new techniques and new chemicals in new combinations in high-tech manufacturing makes safety evaluation harder all the time. "Professionals associated with this industry," wrote Harrison, "have invariably commented on the rapid pace of change in tools and materials, and on the fact that adequate toxicologic assessment of chemicals almost never precedes their introduction into manufacturing settings."

Harrison's frustration is echoed by Joseph LaDou, director of occupational and environmental health at the University of San Francisco. LaDou calls chip making "one of the most chemical-intensive industries ever conceived."

"The air-filtering systems do not alter chemicals except to dilute and recirculate them; and smocks and head gear do not protect workers from toxic exposures," LaDou wrote in 1984. He reiterates the point in an interview 17 years later. "Not only are you recycling the vaporized chemicals, but you're presumably allowing them to react with one another and introducing reactants into the air and recycling those as well."

"Most of our [health] regulations are predicated on workers being exposed to one chemical, maybe two or three -- but what do you do when they're exposed to a hundred?" LaDou asks. "What we have here is a brand-new work setting with an almost scientifically impossible question to answer -- how do you determine if a recirculated mix of chemicals is safe? -- and there is no magic formula."

"The problem with the spectrum of chemicals used in semiconductor manufacturing is that it could conceivably cause any cancer anywhere in the body," says LaDou. "When you find a cancer in a semiconductor worker, it's almost impossible to find a smoking gun."

- - - - - - - - - - - -

The Free Software Project
Read Andrew Leonard's book-in-progress on Linux and open source -- and post your comments.