Three Mile Island:
two hours and 20 minutes to wreck a core

A small team headed off a nuclear disaster. Four of them look back and tell what happened.

On march 28, 1979, a small coolant leak in Reactor 2 of the Three Mile Island nuclear power plant near Harrisburg, Pa., led to a partial meltdown of the fuel assembly. It was the sort of accident that the designers and engineers working at the plant had not foreseen, and it took hours and days until the situation was under control.

As disasters go, this one was pretty minor. Thanks to the work of engineers responding to the accident, very little radioactive matter escaped into the environment; no deaths have ever been attributed to TMI. And although the accident effectively halted any plans for new reactors in the United States, those that remain in service have benefited from a renewed emphasis on safety and maintenance.

On April 28, 2004, as part of the 12th meeting of the International Conference on Nuclear Engineering, four of the principal figures from the crisis gathered to share their personal experiences. The following are excerpts of their discussion, edited for length and clarity.

Jack Devine, former engineer with General Public Utilities (owner of TMI) and a leading member of the emergency response team: It took a long time to regain full control of the reactor to compensate for systems that were damaged and unusable. It's fair to say there was a real siege mentality at the station. We felt we were being pushed in every direction possible and the unfairness of it all was getting to us. One weekend, after several straight weeks of working, I got in my car and I was headed [home].

As I'm driving along, I noticed in front of me a vehicle, which I immediately recognized as the enemy. You've all seen this vehicle. This is a Volkswagen bus. It's got flowers painted on it, and bumper stickers that say things like, 'Better Active Today Than Radioactive Tomorrow' and 'Split Wood Not Atoms.'

A few minutes later, I noticed that they had pulled up, and they were now behind me. So I sped up a little bit more and they sped up. And I slowed down and they didn't pass, and then it seemed as I looked through my rearview mirror that they were pointing at my car. I'm driving a company car, and it's got a big TMI parking sticker on the bumper. They're pointing at my car and I know they spotted me. They pulled up next to me and start gesturing. And I ignored them for a while and then finally—I'm not the world's most patient person—I started gesturing back.

All of a sudden the light bulb dawned: They're telling me that my left rear tire is going flat.

There are a couple of messages: One is [that] things are not always as they seem. Another message is that we're not always as far apart from the people who don't agree with us as we might think.

Bill Dornsythe, former state nuclear engineer, and representative on site of Pennsylvania's governor, Richard Thornburgh: At about seven o'clock in the morning, the utility realized they had a radiation problem, and they declared a site emergency. I called back to the plant after the emergency management agency called me and was talking with some technician who, [as] it became pretty obvious, had no idea what was going on. Probably nobody at that point really knew what was going on.

While I was talking to this technician at the plant, I heard in the background, 'Now evacuate the fuel handling auxiliary building!' And the guy goes, 'Gotta hang up now.' At that point I should have gotten a bus and left town, but I wasn't that smart.

Wednesday morning, because I had been on the horn with the plant, I was asked to go over and brief the lieutenant governor. He wanted to have a press conference and let everybody know what was going on. I asked to speak to somebody in charge at the plant and tell me what the status was. They knew they had some fuel damage. [But] they had absolutely no idea that the core had been that damaged and we were about to see huge releases of noble gases.

Based on that information, I went over to brief the lieutenant governor and we decided on a very brief press statement, which basically said there was an incident at TMI and things are under control and there had been no releases. Unfortunately, on the way over to the press conference I had made the mistake of calling back to our office and the folks at the office said, 'Now we do have releases occurring.' I didn't have time to tell the lieutenant governor [before] he made his statement.

Of course, the press started asking technical questions. Somebody grabbed me by the collar and threw me in front of the microphone. I had to contradict the lieutenant governor. He obviously wasn't real happy about that. And, I started using terms, real technical terms like containment, radiation. And the first question from the press was, 'Well, what does that mean?' I said, 'Oh, man, we have a real uphill battle here.'

The next day I went down to the plant. and on the way down, I tasted what was called the iodine taste in the air. It was unusually warm. And there's a pulp mill about 20 miles downriver from the island. And, essentially, folks were smelling sulfur from that pulp mill. They probably smelled it periodically for years, but their senses were so heightened that they could have sworn they were tasting radioactivity. But it was a true taste in the air.

Also I remember on that Thursday seeing, believe it or not, readings on a GM detector sitting in our window in our office in Harrisburg of three to four millirem per hour from the noble gases that were being released from the site. We're talking 10 miles—10 miles away from the plant.

William Lowe, founder of Pickard, Lowe and Associates, and former senior technical advisor to GPU: I had a growing and overriding conviction that somebody should be in the control room—some of us should be in the control room. That's where the information was. When Jack Herbein, the vice president for nuclear operations, left to call the governor's office, I followed and told him emphatically that the problem was stabilization, not recovery, and that several of us should go to the control room. He came back into the conference room and stated that position and asked for volunteers. And Tom Crimmins and I volunteered. Tom, because of his basic solid characteristics, and I, because I had made the suggestion.

We arrived in the control room at about 10 p.m. on the 29th. And, by then, we were 42 hours into the accident sequence. Then things began to move fast. About an hour after we were up there, the operators lost control of pressurizer levels.

About 11 p.m., a young engineer assigned to collect data approached me and he asked, "Have you seen this?" And he held out a containment building pressure recorder chart showing a spike of about 28 psi at 13:50 on March 28. My impression on the spot was we should give him a medal. I [asked] him for temperature traces, and they confirmed what we were seeing here. At about the same time, at least half of the containment temperature measurements had shown a spike.

I asked him to get Xeroxed copies and went back to the shift supervisor's office where Tom Crimmins was with several others, and I told him that there had been a global hydrogen ignition in containment. That there was a hydrogen bubble in the primary system. That we had to measure it and that we had a fighting chance to get it out because hydrogen, "diffuses like a shot." It's a small molecule. The difficulty of controlling pressurizer levels could be accounted for by the presence of a compressible gas, namely hydrogen, in the primary system. The only source of enough hydrogen for global ignition in containment had to be from a reaction with fuel cladding at very high temperatures.

Harold Denton (left) and President Carter toured the Three Mile Island control room shortly after the accident. The president remained personally involved throughout the crisis.

I knew from personal experience that under high stress, one tends to lock in on a perception of reality that may be wrong. So Crimmins and I forced ourselves to take a time out to review the evidence and test out the hypothesis, the spike, bubble, and the implied fuel damage. When the evidence seemed solid, I called GPU management about 11:30 p.m. and asked that the best person available be sent to help us. That man, Jim Moore, arrived forthwith. The three of us sat around in the shift supervisor's office trying to figure out how to measure the bubble volume. Finally, after what seemed a long, long silence but was probably only a few minutes, Jim Moore said, "Boyle's Law ought to work." And I recall saying almost before he finished, "and the pressurizer can be our piston."

Boyle's Law states that other things being equal, the volume of a perfect gas is inversely proportional to the absolute pressure. I asked the TMI superintendent to increase primary pressure by a hundred psi and measure the change in liquid levels in the pressurizer, while recording pressures and temperatures. Operations said they had data like that from about 13 hours earlier.

Jim Moore and I then made calculations of the bubble size, using the pressure and volume change information. We had a bubble size of 1,568 cubic feet at 875 psi. I started to calculate how much cladding in the core must have burned to produce enough hydrogen for there to be a global ignition of hydrogen in the containment. The first estimate I got was that 200 percent of the cladding had burned—that was obviously not right. When I finally got back to look at the numbers that I had scratched out on a yellow pad, it immediately was obvious there were two errors. The first one was that one pound-mole of zirconium reacting with water makes two of hydrogen, not one. The other error was that I had picked up somewhere an idea that the density of zirconium was 200 pounds per cubic foot, when I knew it was 400. When we did correct both of those, it got down to an estimate that 50 percent of the zirconium cladding had been burned. That still sounded quite a bit much too high, but I had enough confidence in it to conclude that the core was probably destroyed.

Harold Denton, former deputy director of the Nuclear Regulatory Commission: We got the president on the phone. And I remember three things, sort of instructions I got. One was that he would make all of the resources of the federal government available to bring it to a safe conclusion. The second instruction was always tell the truth. And, third, was to keep him fully informed at all times of anything going on. He wanted to be called at a 7:45 every morning and at 3:45 every afternoon and any time anything significant happened.

I didn't realize at the time what a micro manager he was. But if you read the biographies of President Carter, he got into the details of anything that he was interested in. And this was certainly an example of that.

The NRC made two big mistakes. and we certainly did contribute to the stress of the citizens up there, and I was part of both of them to some extent. The first one was when the report came in on Friday of a reading of 1,200 millirem per hour. We didn't know where it was occurring. We didn't know how far away from the plant it was occurring. There was no way to authenticate that number, but yet we knew the EPA guideline for beginning evacuation was a total of five rems for members of the public. So in the absence of any other information we recommended the state be called and recommend evacuation. It took about an hour before we heard back from the state, and they said, "We've got people on the ground here, and there's no radiation levels like that at all anywhere." And it turned out that it was measured right over the stack during a venting of a waste gas. That was all attributable, in my view, to a communications problem, an inability to communicate between all the people who were making measurements and trying to analyze them.

The second one was over the hydrogen bubble. I briefed the governor on Saturday night and told him that I didn't think the bubble was an immediate problem. At that time we said, 'Number one, we don't think there's any hydrogen there. Number two, there was not a source of ignition. Number three, it ought to reach flammable limit before explosive limits. And, number four, we could always vent the thing if we had to.' I went to bed Saturday night. I had already told The White House it was safe for the president to come up on Sunday.

When I went to work Sunday morning, I found out that Bethesda had been working awfully hard and in overdrive. They had called all the national labs. They had tried to poll everybody with any knowledge of what might be in the bubble, and whether or not it could be explosive. And they had convinced themselves that it was a real hazard and, in fact, the commission was considering recommending—this is on the Sunday—a partial precautionary evacuation even while the president was flying up.

So I had about 30 minutes to meet with the staff before the president's helicopter landed. And my deputy was there, and he was strongly of the view that it was impossible to have an explosion in the bubble. He met some of the other staff who had arrived. So when I briefed President Carter, I tried to give him both views on the matter—and that we could handle the problem.

One of my highest anxiety-raising issues was when we took him through the control room and went back to turn in our dosimeters. He read his self-reading pocket dosimeter, and it was reading fairly high—almost off scale, like, you know, 90 millirem. And I knew that mine was reading zero. His wife's was reading very high, equally high as his. Governor Thornburg's was reading zero, and it took a few minutes before we found out that the company was just giving the dosimeters out and writing down what they read each time and subtracting the difference. Trying to convince the presidential party that they really had not gotten 85 or 95 millirem was one of my biggest crises.

Jack Devine: Nuclear safety is not a hypothetical concept. It is not an abstract concept. It is real. At TMI, a small break loss of coolant, misdiagnosed and not corrected . . . wrecked a reactor core. Similarly, as a corollary to that, nuclear safety is not a matter of regulatory compliance. Obviously, we have to comply with regulations, but that's not the heart of the matter. And I should tell you I find myself in that trap all the time . . . thinking in terms of regulations and guidelines and rules and specifics, etc. You do calculations; you do numbers. The decay heat numbers are very big. It doesn't seem real that we wrecked a core in two hours and 20 minutes. And that's what this is all about.

Very bad accidents may unfold and, in fact, are likely to unfold as very insidious, seemingly manageable combinations of events. That's what happened here. I personally think that a large break guillotine rupture, which we designed for, will not happen. We designed for seismic events, which are more serious earthquakes than anybody's ever seen. The TMI accident was a case in point in which the kind of things we all see—an upset in the secondary plant leading to a reactor trip, etc.—somehow managed to go into a spiral of circumstances that we never expected.

Because when we're dealing with a pair of minor problems, we can find ourselves assuming we can deal with them because they will take the direction we always expect, leading a fixation on a presumed answer. It corrupts any potential for an objective assessment.

Engineers dangled a television camera to take these pictures of the damaged core. Some 50 percent of the zirconium cladding has burned away.

This accident can happen. I found myself driving to TMI ... with a little bit of information, finding it almost hard to believe that this was even going on because we really looked at nuclear safety as something which we're required to do, not as a threat to the survival of this planet.

The control room operations crew at TMI—this wasn't a Gilligan's Island thing. This wasn't the back shift of the worst possible folks. I know those operators. They were well-trained. They met requirements.

But they were trained to deal with a different set of circumstances, and they couldn't figure it out. And they looked at the pressurizer level and it was going up. And they learned from many years at a solid plant there's a big problem. So they stopped putting water into the plant, which was desperately needed. And what they were seeing was water level in the pressurizer, not water level in the plant. Their fixation was that was an indication of the condition of the whole reactor. And they couldn't break out of that and understand what was happening.

A year and a half later, we collectively decided to take a timeout from the plans to disassemble the reactor to look inside the reactor and find out what the condition of the fuel actually was. You'd think that we would know exactly what the condition of the fuel was. And Bill Lowe talked about how he had calculated 50 percent of the zirconium had been oxidized at that point. The thinking at that time, with all the knowledge that we had, was that we were gonna find a damaged core—meaning that it was a core that looks like a core, but there had been local melting, there had been surface damage.

There were a lot of people who felt we shouldn't waste our time looking because when we poked this camera into the hole at the top of the core, we won't be able to see anything because we'll be simply at the top of the fuel. We'll see the top of the fuel assembly. But we went ahead with this very crude experiment, sending people to the top of the reactor vessel and dangling a camera into the core.

We're 10 feet above the core or we're eight feet above the core or we're six feet above the core. We're at the core level and we can't see anything at all. It's black. And the naysayers in the room started staying, we knew this would happen. You can't see anything. So we're lowering it further. And he lowers the camera and now we're one foot below the top of the core elevation. We're two feet below the top of the core elevation. Nothing. Three feet, four feet, five feet. This is a 12-foot-deep core. After five feet, all of a sudden something started to come into the screen. And we find ourselves looking at something that is—looks kind of like my gravel driveway. This is what the core was. Five feet of it is gone, and what's there is junk.

We did not believe, you know, in our heart of hearts that that was the condition of that core. The numbers couldn't be right. And they flavored our thinking. We were designing tools to pick out whole fuel assemblies and there were no whole fuel assemblies.

There's a completely cogent view that this accident was the best thing that ever happened. Because if you look at every performance parameter, safety and operations and radiological, etc., in commercial nuclear in the United States from the time of the accident till now you find unbelievable improvement. Capacity factors were in the neighborhood of high 50 percent, early 60s in that time, which is a really pretty crummy performance.

They're now over 90 percent. That's like increasing the numbers of plants out there by 50 percent. It's a phenomenal increase. Radiological exposure levels are a few percent of what they were in the 1970s. So it's the best thing that ever happened. It was a wakeup call and a learning experience for all of us.



Return to Index