Physics Today on the web
Physics Community
What Happened at Tokaimura?

Analysts are trying to figure out how workers ended up putting enough uranium in one tank to initiate a chain reaction.

On 30 September, as workers at a Japanese nuclear fuel processing plant in Tokaimura were adding enriched uranium to a precipitation tank, they saw a blue flash—signaling the onset of a nuclear chain reaction. Exactly how did this accident occur? It will be some time before we get an official report and learn about the accident’s impact on Japan’s nuclear power program. (See previous story.) However, as Physics Today goes to press, independent analysts have already gleaned enough information from preliminary accounts such as those posted on the Web by Japan’s Science and Technology Agency (STA), which licenses nuclear facilities1, to piece together a picture of how a solution containing enriched uranium became critical, what power levels were reached, and what releases of radiation ensued.

Cautions not heeded
The plant where the accident occurred is operated by JCO Co Ltd. Its main function is to convert uranium hexafluoride into uranium dioxide fuel for some of Japan’s commercial nuclear power plants. This uranium has been enriched to contain up to 5% of the fissile isotope, U-235. In addition, the JCO plant occasionally purifies uranium to be made into fuel for an experimental breeder reactor known as Joyo, which requires fuel enriched to 18.8% 235U. For these higher levels of enrichment, one has to be far more careful because of the higher probability of accumulating a critical mass—that is, amassing so much 235U that at least one neutron from each fission, on average, stimulates another fission. STA regulations place a mass limit of 2.4 kg on the amount of 18.8% enriched uranium that can be processed at one time at the JCO plant. Nevertheless, the workers there added a total of about 16 kg to the tank, causing a self-sustaining chain reaction.

Process flow for the conversion facility at the JCO plant. Blue lines indicate the procedure approved by Japan’s Science and Technology Agency. Uranium oxide and nitric acid are fed through a dissolving tank into a buffer tank. Orange lines indicate a company-initiated procedure not approved by STA, in which uranium oxide and nitric acid are added by bucket directly to the buffer tank. The red line indicates buckets dumped directly into a precipitation tank—a further deviation from licensed procedure. (Adapted from ref. 1.)

The purification procedure licensed by STA for the Joyo fuel is shown by the blue lines in the figure above. The workers feed uranium oxide (U3O8) in powder form into a dissolving tank, where it is mixed with nitric acid to produce uranyl nitrate, or UO2(NO3)2, which is then transferred to a buffer tank. From there, it is sent into the precipitation tank, where ammonia is added to form a solid product (with contaminants remaining in solution). Uranium oxide is extracted from that solid, and the process is repeated until the oxide becomes sufficiently pure. At that point, the uranyl nitrate in the buffer tank gets shipped to another facility, where uranium dioxide is prepared and made into Joyo fuel. On the day of the criticality accident, workers were running fuel through the last steps of this process, according to Thomas McLaughlin of Los Alamos National Laboratory, one of three nuclear experts sent by the US Department of Energy to learn about the accident. The JCO plant only needed to mix some high-purity enriched uranium oxide (U3O8) with nitric acid to form uranyl nitrate for shipping. During this operation, the workers deviated from the licensed procedure in three basic ways. First, to speed up the process, they mixed the oxide and nitric acid in 10-liter buckets rather than in the dissolving tank (in doing so, they followed the practice that JCO had written into its manual—without STA approval). Second, for convenience, they added the bucket contents to the precipitation tank rather than to the buffer tank. That was a key misstep, because the tall, narrow geometry of the buffer tank precludes criticality. Third, in filling the precipitation tank, the crew added seven buckets, or roughly seven times more uranium than permitted by the STA license. It was the seventh bucket that caused the mixture to go critical.

According to Shunsuke Kondo, a nuclear safety expert from the University of Tokyo, who has done an independent analysis of the accident, the crew assigned to process the Joyo fuel that day was under time pressure: The crew chief was anxious to complete the current batch before a new team of workers arrived. Furthermore, Kondo reports, the workers were apparently not aware of the mass limitations on the uranium to be added to the precipitation tank.

Reaching criticality
The exact critical mass for the 18.8% uranium mixture in the JCO precipitation tank is not known. In the Joyo reactor, the minimum critical mass for the solid 18.8% uranium fuel is about 46 kg. But the critical mass is greatly reduced when the fuel is in solution because light atoms such as hydrogen slow neutrons between fissions, making it more likely that they will be absorbed. The critical mass was further reduced at Tokaimura because a water jacket surrounding the precipitation tank reflected neutrons back toward the center of the reaction.

A secondary effect of the water jacket may have been to prolong the chain reaction. Per Peterson and Joonhong Ahn of the University of California, Berkeley, point out that, without the water jacket, the heat generated by the chain reaction and the dissociation of water into hydrogen and oxygen would have expanded the solution, decreasing its density and slowing its reaction rate. With the water jacket in place to remove the fission heat roughly as fast as it was generated, however, the solution may have been kept just above the critical density.

Judging from the levels of gamma and neutron radiation measured near the plant perimeter, the criticality excursion seems to have lasted about 20 hours; after that time, the radiation levels dropped below detection limits. The chain reaction was shut off by draining the cooling water out of the jacket and made safe by adding boric acid.

Radiation exposures
The greatest source of radiation exposure in a criticality accident is the flux of neutrons and gamma rays that emanates directly from the fissioning nuclei and rapidly decaying fission products. Such radiation is most harmful to individuals who are nearby and falls off as the square of the distance. A second contribution comes from the volatile fission products, such as isotopes of xenon, krypton, and iodine. At Tokaimura, these gases were vented (an exhaust fan was not shut down until 12 October), but concentrations of xenon and krypton are generally thought to have been below the regulatory limits. Levels of iodine-131 measured on 8 October were about twice those allowed by STA regulations. A third source of radiation is the activation nuclei, or nuclei made radioactive by the absorption of neutrons. Valerie Putman, who works on criticality safety issues at the Idaho National Engineering and Environmental Laboratory, told us that studies have found that the decay of activated atmospheric nitrogen can contribute up to half the total dose if people are not evacuated from in and around the site of the criticality.2

According to STA, the three workers in the room at the time the precipitation tank went critical received doses of 17, 10, and 3 sieverts (the doses were deduced from the levels of radioactive sodium in the victims’ bodies). (One sievert, which equals 100 rems, is a measure of the biological response to the absorbed radiation.) Doses of 10 and 17 Sv are above the levels normally considered fatal, but the two workers who received such high doses were still alive at press time, perhaps because they were treated with blood stem cell transplants. The worker who received a dose of 3 Sv did not require transfusions and is expected to recover fully. In addition to these three severe cases, there were 66 individuals—plant workers, firemen, and others who responded to the accident, and a few city residents—who were exposed to measurable levels of radiation. Most criticality accidents in the past haven’t involved exposures of private citizens, but the JCO facility is sited quite close to the surrounding town. Within the 350-meter radius evacuated immediately after the accident, there were 47 houses and 150 people.

Monitors placed at a number of sites outside the plant detected the radiation levels. At one of the closest monitoring sites, STA reported dose rates of 4.5 mSv/hr for neutrons and 0.50 mSv/hr for gamma rays about 11 hours after the onset of criticality. That gamma dose rate was about 1000 times higher than the normal background level.

The magnitude of the accident
Based on the observed neutron radiation levels, Hiroshi Sekimoto from the Nuclear Reactor Institute of the Tokyo Institute of Technology initially estimated that the chain reaction may have involved 1–8 X 1018 fissions, consistent with a steady-state power 0.7–4 kW. (The thermal output of a typical commercial power reactor is about 3000 MW.) Since then, using information about the fission products found in samples taken from the precipitation tank on 20 October, he has revised his estimate to 1.8–2.8 X 1018 total fissions. Peterson and Ahn have also made a preliminary estimate of the power level reached during criticality and hence the maximum radioactive releases, by making some assumptions about the heat balance in the tank. They have concluded that the chain reaction generated heat at a rate of 5–30 kW. At that power level, it may have produced 30 to 180 curies of xenon-133 and 10–60 curies of iodine-131. (The explosion at the 1986 Chernobyl nuclear power plant in Ukraine spewed out tens of millions of curies of these isotopes.) Thomas Cochran of the Natural Resources Defense Council has put the Tokaimura episode in perspective by examining 22 criticality events at US nuclear facilities other than reactors (all but one of which occurred before 1964). He found that the number of fissions generated by fairly similar accidents was in the range of 1017–1019. Assuming that the Tokaimura accident was in the same ballpark, Cochran estimates releases of 131I that overlap with those calculated by Peterson and Ahn. Cochran has concluded that the radiological impact on the public of the Tokaimura episode is not likely to be larger than that of the 1979 nuclear accident at the Three Mile Island nuclear power plant in Pennsylvania.

McLaughlin has been working over the past year to update a report on criticality accidents around the world by incorporating data now available on accidents in the former Soviet Union. Although he can’t yet say what happened at the JCO plant, he did refer us to the list of “lessons learned” from past accidents. He noted that what many of the accidents have had in common have been failures in communications and operator training, improper procedures, lack of fissile-material accountability, and new or unfamiliar operations. Judging by the standards in the US, McLaughlin said, it appears that, in the Tokaimura incident, regulatory agencies and plant managers were not diligent in following approved procedures.

The entire JCO plant, not just the purification operation, is now shut down, and STA has revoked JCO’s operating license for the plant. Various investigations by government agencies are under way.

--Barbara Goss Levi

1. Science and Technology Agency’s Web site (

2.  C. M. Hopper, B. Broadhead, An Updated Nuclear Criticality Safety Slide Rule, NUREG/CR-6504 Vol. 2; ORNL.TM-13322/V2. US Nuclear Regulatory Commission, Washington, D.C., April 1998.

Japan Rethinks its Nuclear Future

© 1999 American Institute of Physics

Current Contents Past Contents Buyer's Guide
Job Ads Upcoming Meetings About PT
Subscribe Contact Us PT Home
Advertising Information Web Watch