RICHLAND, Wash. _ Rising from the desert sand outside Washington state's Tri-Cities area is a construction project unlike any in history.
The buildings going up in this gated-off stretch of tumbleweed are reinforced with as much steel as three Eiffel Towers. The concrete structures are being threaded with enough pipe to funnel water 200 miles to Seattle. Workers are installing giant melting machines, which will burn so hot they could turn gold to soup.
This is supposed to be the most sophisticated garbage disposal on Earth, the centerpiece of the Western world's costliest environmental cleanup.
It's supposed to rid the Hanford nuclear reservation of its deadliest poisons: 53 million gallons of radioactive waste, now buried in aging, leaky tanks.
But after nearly a quarter-century of, and cost estimates that have nearly tripled to $12.2 billion, builders still haven't resolved this project's most vexing technical and safety issues.
That leaves some worried the federal Department of Energy and, its contractors may build themselves into a corner and produce a plant that's dangerous or doesn't treat as much waste as expected. They fear construction will drag on and increase environmental risk, or cost billions more to get right.
This one-of-a-kind plant is supposed to turn the nuclear slop from Hanford's 177 underground waste tanks into glass. Today, the plant is more than half-constructed. The DOE and its contractors are making the case to Congress that design and planning are all but finished.
But parts of the plant still face risks of bursting into flames, exploding or triggering uncontrolled nuclear-chain reactions, according to project documents, interviews and formal critiques by scientists at other federal agencies.
The government's own tests show that equipment may fail or pipes may clog in areas of the plant so hot with nuclear waste that no human or machine could ever get in and make repairs.
In addition, each of Hanford's underground waste tanks holds a unique mix of hundreds of toxic compounds and radioactive isotopes. But crucial elements of the plant's testing and design were based on samples that don't reflect this unusual cocktail.
"We figured out how to put a man on the moon in 10 years using slide rules," said Walt Tamosaitis, a high-level Hanford engineer who said he was removed from the project last year after raising safety concerns. "We still can't seem to get this right."
To be sure, cleaning up Hanford's atomic mess after 40 years of bomb-making is devilishly tricky. The DOE and lead private contractor Bechtel National insist they know how to make the plant work. They say they have a strategy for solving remaining problems and point to an outside review by scientists appointed by DOE that suggests they are on track. They say if future tests show an approach is risky, they'll alter it.
The plant is scheduled to begin operating in 2019.
"From the management perspective, we believe we have a handle on the larger technical issues," said Delmar Noyes, the treatment project's deputy director for DOE.
But the project presents what some call a "wicked problem."
Slight design changes in one area can trigger a cascade of unforeseen issues elsewhere. And the plant is being built before contractors have figured out details of the final design.
That worries the Defense Nuclear Facilities Safety Board, a federal panel of nuclear scientists appointed by the White House to oversee Energy Department projects. The board makes recommendations and updates Congress about safety issues.
"I think it's fair to say the board is concerned about many aspects of this project," Chairman Peter Winokur said in an interview.
Certainly Hanford cleanup demands haste. The 586-square-mile Manhattan Project site began producing plutonium for atomic weapons in the 1940s. Production continued until Hanford's last reactor was shut down in 1987.
The rush to arms produced billions of gallons of waste. Some of the hottest was funneled into concrete and carbon-steel tanks that today are decades past their projected life span. Some have spontaneously heated up or burped explosive gases. Sixty-seven are suspected to have leaked. At least a million gallons of radioactive goo has spilled into the ground and is working its way to the Columbia River.
But year after year, cost overruns and delays dogged cleanup.
The government spent $197 million building processing plants and vaults to dispose of waste by mixing it with cement only to drop the idea. It bagged plans to melt and seal nuclear material in insulated boxes _ after spending $418 million. Contractors got bonuses for doing work after scheduled start dates, and for completing projects not needed for years.
As recently as 2009, auditors learned a Hanford contractor had spent $103,000 in taxpayer money buying frozen dinners for employees working overtime. An inspector figured out the practice was probably illegal but had been written into Hanford labor agreements since 1955.
But there has been progress, too. Gone are hundreds of millions of pounds of contaminated dirt and concrete, along with nearly one-third of Hanford's buildings. Gone, too, are basins that housed leaky pools filled with tens of thousands of deteriorating nuclear-fuel rods. The spent rods are now packed in steel and buried in concrete vaults.
Still, after more than 20 years of full-time cleanup, the site's most urgent problem remains: the tanks.
Hanford's tanks range in size from 55,000 gallons to 1 million gallons, and they were built between the 1940s and the 1980s. The oldest 149 have only a single shell, and much of the liquid from them has been pumped into newer double-shelled tanks. Remaining waste is a mix of sludge, cakey salts, gas and liquid. In all, the 177 tanks hold two-thirds of the country's high-level bomb-making wastes.
(EDITORS: BEGIN OPTIONAL TRIM)
Over the years, each has become a distinct and changing caustic brew. As tanks filled during the Cold War, workers piped contents from one to another. Some spilled during transfer. Some was dumped intentionally into the ground to make room for more. Chemicals and heavy metals were added to neutralize waste, or to separate reusable nuclear materials. Tank managers poured in sodium so acids wouldn't dissolve tank walls. Water was added and boiled off.
Record-keeping was poor or inaccurate. No one knows precisely what's in them.
"The waste was changing and chemically reacting all the time," said Dirk Dunning, with Oregon's Department of Energy.
(END OPTIONAL TRIM)
Visions of a plant that could melt this garbage into glass to be safely stored forever gained traction during the late 1980s. The process is used in Europe and South Carolina, but for substantially smaller quantities of waste. It's never been tried with such a toxic mix.
In the late 1990s, the DOE hired a British company to design and build the plant. The government sacked the contractor in 2000 after cost estimates rose to $15.2 billion. Energy officials then hired Bechtel, for $4.3 billion.
As designed, Hanford's plant is supposed to pump waste from the tanks into a series of holding vessels. From there, the radioactive elements will be separated and turned into two kinds of glass.
But Bechtel, too, soon faced criticism. It installed an important holding vessel only to discover later, and by accident, that the welds were faulty. It was forced to return flawed steel beams.
Congressional investigators with the Government Accountability Office complained repeatedly, as early as 2003, that the plant moved to construction before proving parts of the design would work. Costs shot back up to $12.2 billion. The project fell a decade behind schedule.
An Army Corps of Engineers report found little evidence Bechtel controlled costs. It reported Bechtel and DOE were overly optimistic about how fast and well they could finish their work. It revealed engineers were struggling to figure out technical problems _ how to protect against fires, keep explosive gases from building, or keep waste mixed up safely. The report was written in May 2005.
Five years later _ halfway through construction _ plant builders still wrestle with the same issues.
In the fall, a panel of scientists appointed by the DOE said contractors had finally found a reasonable "path forward" with remaining technical problems. Bechtel's "professionalism and effectiveness," the panel determined, would keep the project on track.
But even those scientists made clear that didn't mean the problems were actually solved. And Bechtel's approach still makes other scientists uneasy.
For example, one of the most confounding and dangerous issues is how to keep nuclear waste stirred. A consortium of university researchers and DOE's own Pacific Northwest National Laboratory said Bechtel's approach is based on overly simple tests that are too small to reflect real-world operations.
There are other concerns, too. Fast tracking this project has led to so many late design changes that DOE officials recently claimed the plant was too complex. So they began simplifying the design by scaling back safety measures.
For instance, because waste can generate dangerous gases, builders first tried designing and aligning piping systems in ways that prevented fires and mini-explosions. But now, pipes are being designed so that explosions are expected and allowed; plant operators will just have to keep them manageable.
An independent review team raised concerns about this approach. Other government scientists said Bechtel's confidence again seemed based on simplistic tests.
Allowing explosions also means radiation leaks are more likely. Contractors said that would be OK because leaks would quickly be discovered and cleaned up. But an increased possibility of leaks makes it that much more important to understand and limit exposure risks to workers and the public.
In the fall, scientists pointed out that the Energy Department had underpredicted by a factor of four just how far radiation could spread in an accident. The DOE conceded its numbers were wrong but insisted other calculations guaranteed the public's safety.
One of those sounding alarms is Tamosaitis, an engineer who helped identify many of the problems.
"I maintain that debating how many hydrogen explosions pipes can withstand is the wrong thing to be discussing," Tamosaitis said. "When you start adding up the marginal factors ... you're setting the stage for a major problem. You begin to see how an accident like BP's in the Gulf could happen."
Tamosaitis raised these issues with his bosses last summer but said he was immediately removed from work on the project. He since has filed whistle-blower claims against the contractors and DOE. None will comment on his case. A Labor Department investigation is under way.
The Defense Nuclear Facilities Safety Board also has questions about whether the plant will work fast enough. As is, the plant isn't expected to process all the waste until 2047, and operational costs could hit $100 billion. By then some of Hanford's underground tanks will be a century old.
And that schedule is doable only if everything works as planned. Things at Hanford often don't work as planned.
Some worry processing the waste may take longer, and the plant is being designed to work for only 40 years.
Recently, the Energy Department reignited a push for new technologies, and is working with a company to investigate alternative waste-treatment ideas. Officials have said the hope is to reduce the timetable for Hanford cleanup.
But critics like Tom Carpenter, with the activist group Hanford Challenge, see that as a tacit admission the waste-to-glass plant won't do as much as once intended.
One of the technologies being considered is steam reforming _ turning waste to gas using steam and converting it to crystals _ a process once rejected as inappropriate for Hanford.
"I think it shows how desperate they are," Carpenter said. "I think it makes clear they're worried it isn't going to work."
Noyes, with DOE, disputes that characterization. Others aren't sure what to think.
Ken Niles, who is head of Oregon's nuclear-cleanup program, said he fears some of the current issues are "significant."
But he doesn't see any option but pushing forward with the plant.
He acknowledges his position requires a leap of faith.
"I still believe the plant is going to work," Niles said. "But I believe that, in part, because it has to. It's got to work."