[This is Part 4 of an In-Depth Report on The Future of Nuclear Power.]

On Feb. 16, 2002, the nuclear power plant called Davis–Besse on the shores of Lake Erie near Toledo, Ohio, shut down. On inspection, a pineapple-size section on the 6.63-inch- (16.84-centimeter-) thick carbon steel lid that holds in the pressurized, fission-heated water in the site's sole reactor had been entirely eaten away by boric acid formed from a leak. The only thing standing between the escape of nuclear steam and a possible chain of events leading to a meltdown was an internal liner of stainless steel just three sixteenths of an inch (0.48 centimeter) thick that had slowly bent out about an eighth of an inch (0.32 centimeter) into the cavity due to the constant 2,200 pound-per–square-inch (155-kilogram-per-square-centimeter) pressure.

That cavity of as much as 30 square inches (194 square centimeters) had formed on the Unit 1 reactor head—a hole apparently missed during prior safety inspections in 2000 and 1998 by the U.S. Nuclear Regulatory Commission (NRC), the government agency charged with monitoring the nation's nuclear reactors. The inspector noted deposits of boric acid but underestimated the potential impact, despite more than three decades of issues in nuclear reactors involving boric acid corroding metal. The hole was only discovered when a replaced nozzle tipped over during the repair process in 2002.

Circular cracks had formed around several of these steel nozzles in the corroded lid that shield the control rods—20 12-inch- (30-centimeter-) long sticks made of a silver–indium–cadmium alloy that are used to dampen or shut down a nuclear reaction—rendering them vulnerable to simply popping off. The NRC first explicitly identified this problem in 1993 (and it had been occurring in similar plants since the 1980s). But it took until 2001 for the NRC to require operators of reactors prone to the potential problem to check for it, a process that had not been completed at Davis-Besse.

As a result of that gaping cavity, the NRC shut down Davis–Besse for two years (from 2002 to 2004) in an attempt to patch the nozzle that had been leaking boric acid as well as to address design flaws that had been in place since the plant started operation in 1977 and what the agency described as a poor safety culture. In other words, an unwillingness by employees of owner Akron, Ohio–based FirstEnergy Corp. to report safety issues. All told, the repairs totaled $600 million.

A subsequent review of the entire power plant by the NRC in the wake of the near-miss revealed that its emergency cooling system—a critical line of defense in the event of a meltdown—might have failed due to clogging resulting from "generic" flaws built into the plant prior to 1977.

"The reactor vessel head could have failed between two and 14 months further out, that would have been a major loss-of-coolant accident," says NRC spokesman Scott Burnell. "There would have been a very real possibility of damaging the [nuclear] core. In terms of coming up with probabilities, the staff determined that this event was one of the most serious situations since Three Mile Island," the Pennsylvania reactor that partially melted down in 1979. And other models put the incidence of an accident as little as two weeks away, due to the cracked and buckled stainless steel.

Davis–Besse is running again, generating 7.7 billion kilowatt-hours of electricity in 2007, with a new reactor head scavenged from an uncompleted nuclear power plant in Michigan. "The NRC has no concerns," Burnell says.

Davis–Besse and FirstEnergy are not alone. Since 1979, after Three Mile Island partially melted down, U.S. nuclear reactors have had to shut down for a year or more for repairs or other safety improvements 46 times, according to the Union of Concerned Scientists. For example, in February 2000 a steam generator tube abruptly ruptured at the Indian Point nuclear power plant in Buchanan, N.Y.—a potential problem first identified in 1997 that had not been fixed.

All told, only four incidents in the history of the nuclear power industry have been worse than the cavity at Davis–Besse, and two have been roughly equivalent, according to the NRC, such as a radioactive steam pipe burst at PSEG's Salem nuclear generating station in New Jersey.

Even as the U.S. considers building as many as 26 new reactors, 51 of the 104 currently operating have received clearance from the NRC to extend their generating lives by 20 or more years. And the remainder are either under review by the agency or expected to apply. The question: Are they safe?

Inside the bathtub
At issue in the failure at Davis–Besse is the alloy metal used to craft the nozzles—known as Inconel 600 or Alloy 600. The alloy of nickel, chromium and iron is resistant to corrosion generally—but slowly cracks when exposed to boric acid and stress.

But it isn't just reactor heads that are made from the stuff. The steam generators that transfer the heat from the solution of water and boron, which comes into contact with the nuclear reactor, to the water heated into steam that turns the turbines to produce electricity also employ the alloy. "The tubes in the steam generators were susceptible to cracking," says Ken Karwoski, senior level advisor for steam generators and material inspection at the NRC's Office of Nuclear Reactor Regulation. "It's a combination of the temperature and the water chemistry."

In the 1970s and 1980s there were several such tubes that actually ruptured, and nuclear operators have employed patches to keep damaged steam generators in service (although this cuts down on the efficiency in generating power). But many have chosen to completely replace their steam generators since 1989 at a cost of as much as $600 million.

As early as the late 1950s there was some suggestion that this metal would crack under pressure but "the decision was made to go with this material," Karwoski says. "The perspective was that it should last but it didn't." And, as of today, there are still 15 nuclear power plants, including Davis–Besse, employing their old steam generators made from the alloy.

The U.S. fleet of 104 nuclear reactors—most built in the 1960s and 1970s—produced 806.5 billion kilowatt-hours of electricity in 2007, a record, and ran almost 92 percent of the time.

But metal fatigue, embrittled concrete (and even rotting wood in the case of Vermont Yankee) have all plagued the nation's aging fleet. Counterintuitively, though, the bulk of problems with these reactors occurred during their first decade of operation—a fact that bodes ill for the next round of nuclear power plant construction. "There will be deficiencies or defects that occur as you start new plants up, even though new plants are simpler and have fewer components," says Adrian Heymer, senior director for strategic programs at the Nuclear Energy Institute, an industry group. "There's always minor adjustments you have to make to the equipment."

This is a phenomenon engineers call the bathtub curve: In the early days of a piece of machinery's operation, as the kinks are being worked out, problems are numerous, but then it declines into a period of smooth operation. The question is: How long will that period be in the case of nuclear reactors? Because on the far side of the bathtub curve is a climb in the number of problems.

According to the NRC's analysis of so-called "accident sequence precursors," or events that could lead to a nuclear accident, U.S. nuclear reactors averaged seven such events per year between 1993 and 2004—three per year at boiling-water reactors and 11 at pressurized water reactors, most of which could be attributed to the loss of auxiliary, nonreactor-generated power required to run the safety equipment—an increase in such precursors that was statistically significant even when such loss-of-power events were not counted.

Human error
It may not be problems with the nuclear reactors themselves, however, that pose the greatest safety concern. Of all the failures, human error among the 50 to 55 reactor operators who staff a nuclear power plant control room in shifts looms largest, particularly in the case of the most notorious nuclear accident in U.S. history: Three Mile Island (TMI).

On March 28, 1979, Unit 2 at the TMI nuclear power plant near Middletown, Pa., suffered a partial meltdown—an overheating of the nuclear fuel rods caused by radioactive decay in the fission by-products. If unchecked, a meltdown can send superheated fuel through the steel and concrete that surrounds it, damaging or destroying the reactor and releasing extreme levels of radiation into the environment.

At about 4 A.M. local time, the main pumps feeding cooling water into Unit 2 failed and, due to confusion amidst the klaxon of alarms and flashing warning lights, the men operating the reactor made the situation worse when they mistakenly thought there was too much water in the core and shut off emergency pumps, thereby reducing further the amount of coolant reaching the reactor. Within three hours the nuclear fuel had boiled itself dry and then burst through its zirconium cladding before beginning to melt inside the reactor. "If they kept their hands in their pockets, everything goes much better," notes Gary Callaway, a former reactor operator at the Palo Verde and Indian Point nuclear plants in Arizona and New York State, respectively, and now an NRC trainer.

Before the problem was fixed, one half of the core had melted, though it never breached the walls of steel and concrete designed to contain it and only a small amount of vented radioactive steam, other gases and water was released—enough to give a dose of an average of eight millirems to people living within 10 miles (16 kilometers) of the plant, or the same as the dose from a chest x-ray, according to the American Nuclear Society (ANS), an industry group.

Nevertheless, the TMI accident essentially brought to a halt the first round of nuclear power plant construction in the U.S. And such operator error has been behind many of the worst disasters in the global nuclear power industry, including Chernobyl, in Ukraine (then part of the U.S.S.R.), which involved a reactor that on April 26, 1986, exploded through its containment after a series of mistakes by Soviet operators during an experiment. "You cannot regulate against stupidity," says John Ricci, manager of specialized technical training at the NRC. "As far as I know we've never killed anyone with a nuclear power plant."

Chernobyl was directly responsible for at least 56 deaths and as many as 4,000 more, according to the World Health Organization, though other estimates vary, and spread radioactive material as far as the U.K. Three Mile Island has never been conclusively linked to any deaths or health effects, though some individuals may have received radiation doses of as high as 100 millirems. An average American is exposed to an average of 360 millirems per year from natural sources of radiation, such as cosmic rays from space.

And nuclear power plants have been operating in the U.S. for 50 years without exposing workers or residents in surrounding areas to excessive radiation. "Radiation is mundane, it's a weak carcinogen," says Rod Reed, a senior health physicist at the NRC. "It leads to very mundane changes, not three-eyed fish."

In fact, a typical coal-fired power plant exposes local residents to as many as 18 millirems of radiation yearly, whereas a nuclear power plant emits less than six millirems per annum, according to researchers at Oak Ridge National Laboratory.

Reed adds: "Radiation should be respected, not feared."

Ultimately, the safety of a nuclear power plant depends on the operation of its various safety systems, from cooling water to the control rods that modify or halt the nuclear reaction. Such shutdowns are known as "scrams". (Their name derives from the safety control rods that were moved by ropes on the first primitive reactor. To prevent a runaway reaction, an ax wielder stood ready to literally chop the rope, which would drop the control rods and stop nuclear fission if a meltdown was imminent—so literally safety control rod ax man.) Scrams can occur for everything from a loss of coolant to a loss of auxiliary power.

And there's a new problem that has appeared in recent years with surging demand for new parts: counterfeits. The NRC has found fake and possibly faulty valves and breakers—those not actually verified to stand up to the rigors of a nuclear power plant—at two facilities, according to a notice in 2008. "These recent examples have not found their way into the safety systems of our nuclear plants, but I'm sure you can appreciate the seriousness of the situation if that were to happen," NRC commissioner Peter Lyons said in a speech to the ANS last June. "The volume of the flow of parts and components will surge with new plant construction, and we all must remain vigilant."

"The nuclear industry in general is an industry where it's in our best interest that everybody builds safe and reliable plants," said Steve Winn, executive vice president of strategy, environment and nuclear development at Princeton, N.J.–based utility NRG Energy, in an interview with Scientific American in 2007. "All of us are only as good as the weakest member of the group."

Statistical risk
The goal for nuclear power plant operators and builders is to reduce the risk of a serious accident to less than one in 100,000 or more years, according to the NRC. "There is no such thing as zero risk," the commission's Ricci says. "We cannot guarantee perfect protection."

But there has already been a core damage event in the U.S. nuclear industry—TMI Unit 2—"so we've already blown the goal," says Rick DeVercelly, a former operator at the Vermont Yankee and James A. Fitzpatrick (near Oswego, N.Y.) nuclear plants, and now a trainer at the NRC.

And, even in the case of Three Mile Island, a catastrophe was averted, although nearby residents were evacuated and the damaged Unit 2 waits to be dismantled. More recently, the problems at Davis–Besse resulted in one independent contractor receiving a jail sentence and FirstEnergy paying $33.5 million in civil and criminal penalties—and no accident occurred. "The agency is committed to making sure that nothing like that ever happens again," says the NRC's Burnell.

But aging nuclear power plants may make that goal impossible. In 2006 the NRC granted Vermont Yankee permission to increase its operating temperature in order to generate more power. In 2007 a sagging bracket in the wooden cooling tower at the Vermont facility caused a cooling pipe to burst, and last year, for example, even a new bracket sagged enough to cause a pipe leak.

And human error led the operators of Vermont Yankee to misplace segments from two spent fuel rods, a chronic problem with nuclear material. Plutonium from the Manhattan Project ended up in a safe at the nuclear dump in Hanford, Wash., and a U.S. Department of Energy survey in 1996 of all highly enriched uranium produced in the country revealed that 3.2 tons of it is missing. Most of the world's radiation-related fatal accidents occur because of such misplaced or mishandled radioactive material, such as that in Goiania, Brazil, in which two junkyard workers died of radiation poisoning.

At the same time, Vermont Yankee provides as much as one third of that state's electricity at a cost of 5 cents per kilowatt-hour, well below any other generating source, and it provides 220 jobs to the region. It has operated relatively safely for 35 years and the NRC is currently considering whether to extend its license to operate for another two decades, to 2032.

"It is the oldest plant still in operation and has had a myriad of problems from collapsing cooling towers to losing a spent fuel rod," says Oona Adams, a union organizer who grew up in Guilford, Vt., near the plant. "As a child, we had a radio supplied by Vermont Yankee that broadcast weather and also served as an early warning device. Early warning is not a good phrase for kids and it's one we heard in school, at home and when the sirens blast out a test anthem every first Saturday of the month."

Similar extensions have been granted to 51 nuclear power plants since 2000 and 19 more are pending, including Oyster Creek in New Jersey. "We don't believe that nuclear power is a religious thing, it's a business thing," says Craig Nesbit, vice president of communications for Exelon, which owns 17 nuclear power plants, including Oyster Creek. "A lot of what happens with the future of nuclear in this country depends on how well we continue to run these plants. We have a pretty good track record for nuclear power plant operation in the U.S. and that has to continue. We have to have continued public and political support for nuclear in this country. We hold the key to that in our own hands."