The last thing the night shift wrote was one line in the control-room log. “Pumped reactor water to contaminated water tank until reactor water level recorder came on scale. Indicates +5 ft. Replacing plugs, thimbles, etc., to all rods.” The line is not timed. It was the evening of Tuesday, 3 January 1961, at a small Army reactor on the lava plain of eastern Idaho, forty-one miles from Idaho Falls, with the temperature on its way to seventeen below zero. The three men who had written it went back up the stairs to the top of the reactor.

At 9:01 p.m. an automatic alarm from the reactor building reached the fire stations and the security headquarters of the National Reactor Testing Station. Nine minutes later a fire engine was at the locked gate of the SL-1 compound. The firemen saw no fire and no one; in the support building they found three lunch pails on a table. At the foot of the covered stairway to the reactor floor their survey meter went to 25 roentgens an hour, the top of its scale, and they went back out into the cold.

The men were on the reactor floor. Two were dead. The third died at 11:14 p.m. in an ambulance stopped at a highway junction in the desert. They were the first people a reactor had killed in the United States. What had killed them was a sheet of cadmium in a metal jacket, one of five, that a written procedure had allowed to be lifted four inches, and that a hand had lifted about twenty. The number was found months later, in a hot cell forty miles away, as a scratch on a steel rod.

Who ran it

The experiment was the Army’s. From 1953 the Distant Early Warning line, a chain of radar stations across the Arctic, had been watching for Soviet bombers, and every station ran on diesel oil flown or shipped in at a cost the Army wanted to stop paying. On 27 September 1955 the Department of Defense asked the Atomic Energy Commission for a nuclear plant that could be packaged in a few pieces, carried by cargo aircraft, put together by soldiers in a few hours, and left to run for three years on one loading of fuel.

The Commission’s Division of Reactor Development gave the design to Argonne National Laboratory outside Chicago, which had built the BORAX boiling-water reactors at the Idaho station and knew what they would do. The site was the National Reactor Testing Station, nine hundred square miles of sagebrush where a dozen reactors already ran. Construction began in July 1957. The reactor went critical on 11 August 1958 and made its first electricity on 24 October, and on 5 February 1959 Argonne handed it to the Army’s operating contractor, Combustion Engineering, Inc., of Windsor, Connecticut, under a cost-plus-fixed-fee contract. The Army named it Stationary Low-Power Reactor Number One, SL-1, to set it apart from the family of portable and mobile plants it had in mind.

Combustion Engineering kept a project manager at the site half the time, with a technical staff of about eight. The reactor itself was run by soldiers, sailors and airmen. The Army trained its operators at Fort Belvoir in Virginia and sent them to Idaho in cadres; the Navy, interested in the plant for its own purposes, sent Seabees. The Commission’s Idaho Operations Office administered the contract, and the Army Reactors Branch at the Commission’s headquarters had a staff role it described, before the investigating board, as “planning, observing, advising, appraising and recommending,” with no authority to give orders.

The crew that came on at four o’clock on the afternoon of 3 January 1961 was three men. Richard C. Legg was a Navy construction electrician first class, twenty-six, from Roscommon, Michigan. John A. Byrnes was an Army specialist fifth class, twenty-two, from Utica, New York. Both were certified reactor operators. Richard L. McKinley, an Army specialist fourth class of twenty-seven from Kenton, Ohio, was a trainee, due for certification in February. Byrnes and McKinley lived in Idaho Falls with their wives. The reports written afterwards do not name them. The board’s report calls them “the shift supervisor (a qualified chief operator), his operator-mechanic assistant (a qualified operator), and a trainee,” and the contractor’s report numbers them one, two and three by where they were found.

What they were trying to find out

The Army’s question was whether a reactor could be run like a diesel generator. Not whether a boiling-water reactor would make steam, but whether one could be built small enough to fly, started by enlisted men with a few months’ training, and kept by them at the far end of the world with no physicist in the building. SL-1 was the test of that, and the arrangement in Idaho was made to match. By September 1959 the shift crews had been cut to two men, a chief operator and an operator, both military. On the night shifts there was no contractor supervisor at all. The contractor’s part-time project manager wrote to the Commission on 29 November 1960 “requesting written confirmation of the oral agreement that CEI shift supervisors were not required for routine supervision of plant operation during the night shifts.” The letter had not been answered when the reactor blew up, but the board found that the Commission’s offices in Idaho and Washington “clearly believed that addition of night supervisors when only routine work was involved would defeat a part of the purpose of operating the reactor under the existing arrangement, i.e., to obtain plant operating experience with only military personnel.”

Inside that question sat a second one, about the physics, and it had an answer from the first week. A reactor meant to run three years without refuelling has to be loaded with far more uranium than it needs to go critical, and the surplus has to be held down. Argonne held it down two ways: with five control rods of cadmium, and with strips of boron, a burnable poison, tack-welded to the outside of the fuel assemblies so that the boron would burn away over the years at about the rate the fuel did.

The hazards report Argonne wrote for the reactor was reviewed in February 1958 by the Commission’s Hazards Evaluation Branch and by its Advisory Committee on Reactor Safeguards, and both approved operation at 3 megawatts of heat. The report estimated that the reactivity of the core would rise over its life by at most about 0.8 percent, which it translated into “one or two inches in the position of the central control rod.” The critical experiments of August 1958, before any boron strips had been fitted, found something else about that rod. With only sixteen fuel assemblies in place and the four outer rods fully inserted, the reactor went critical with the central rod alone withdrawn 14 to 14.5 inches. In the full core of forty, with the boron in place, the margin was wider, but the fact stayed what it was. The board put it in one sentence: “The reactor could have been made critical by withdrawal of the central control rod only, anytime after startup of the reactor.” After February 1958 nobody outside the project was asked to look at the reactor again. “No requests for independent review were made after initial operation.”

What they built

The building was a steel cylinder forty-eight feet high and about thirty-eight feet across. It was not a containment; the Commission judged the plant too small and too remote to need one. Inside, the reactor vessel, four and a half feet across and fourteen and a half feet high, stood in gravel on a concrete pad on the lava, with the operating floor at the level of its head, a fan floor above carrying the condensers, and a covered stairway up the outside. Forty assemblies of enriched uranium-aluminium plates, with 14 kilograms of uranium-235 among them, sat in a core twenty-six inches high and thirty-one across. The water that cooled them boiled, and the steam went straight to a turbine.

The core had channels for nine control rods, five cross-shaped and four T-shaped, because Argonne had left room for a bigger core if the Army ever wanted one. Only the five cross-shaped rods were fitted. Each was a sheet of cadmium, the absorber, thirty-four inches long, clad in aluminium-nickel alloy and running in a thin-walled shroud between the fuel assemblies. They were numbered 1, 3, 5, 7 and 9, and rod 9 was at the centre.

Each rod hung from an extension rod and a toothed rack that rose through a nozzle in the vessel head, into a shield plug and then a housing where a pinion gear, driven by an electric motor through a clutch, moved it up and down. The motors were slow by design: the outer rods could be driven at 2.85 inches a minute and the central rod at 1.85. Rod 9 was also the only rod with a switch of its own; the other four shared one. Scrammed, a rod fell home by its own weight, and was expected to fall its first ten inches in a second.

Connecting a rod to its drive was done by hand. At the top of the rack, above the shield plug, a washer and a slotted nut had to be fitted that would stop the rod when it fell, which could only be done with the rack lifted clear of its housing. The contractor’s written procedure for taking the drives apart gave the lift as a number. Step nine: secure the lifting tool to the top of the rack “and raise rod not more than 4 inches. Secure ‘C’ clamp to rack at the top of spring housing.” To put the drives together again the same steps ran in reverse: lift, clamp, fit the washer and nut, lift a little to free the clamp, and lower. The board, reading the procedure afterwards, noticed what was not in it. The control rod “was not to be raised more than four inches. The reason for this limit was not given in the procedure.”

Under the vessel head, four inches was a safe distance. The rod did not reach its critical position until it had been raised some sixteen inches, and the numbers in the different reports, sixteen and seventeen and twenty, depend on where one starts counting, since the disconnected rod rested about four inches below the indicator’s zero. What every report agrees on is that a man raising the rack to fit a washer had about a foot of travel in hand before anything would happen, and that nothing in the procedure told him so.

The reactor ran for two years, and the record of those years is in its logs; the companion piece “The rod log” goes through them. The boron strips went first. In 1959 they were found bowing between their tack welds. An inspection in August 1960 found large pieces of them missing from some assemblies, and the central assemblies so hard to pull that “removal caused plates to fall off and flaking of material. A considerable number of flakes were collected from the bottom of the vessel.” Since pulling the fuel seemed to be making it worse, the inspections were stopped and the reactor ran on. The lost boron showed up in the rods, which had to sit deeper in the core to hold the reactivity down, and on 11 November 1960 the contractor pushed strips of cadmium into two of the empty T-rod shrouds to buy some of the margin back. The board’s estimate was that the shutdown margin of the cold reactor, designed at 4 to 5 percent in reactivity, had been about 3.5 percent when the reactor was new and 2 to 3 percent when it was shut down for Christmas. The loss of boron “was well known within the Division of Reactor Development at AEC Headquarters, although it was not categorized as a serious condition in the reports transmitted to Headquarters.”

The rods went second. “From early operations onward,” the board wrote, “intermittent and increasing difficulty was encountered in the free movement of the control rods.” For the first year the trouble was in the seals, and everyone at the plant had a remedy for it. In the autumn of 1960 it changed shape. In the first twenty-two months of operation about 2.5 percent of the scrams and drop tests had gone wrong; in the last month before the shutdown, 13 percent did. On 19 December a pipe wrench was used to pull two rods past a point where they stuck. On 20 December the plant superintendent ordered every shift to exercise the rods, “until further notice.” On the morning of 23 December the reactor was shut down for the holiday, and the log recorded the rods dropped one at a time: “#1 no drop. #3 dropped 1/2” and stuck. #5 clean drop in 0.82 sec. #7 no drop. #9 clean drop in 0.81 sec.” The contractor’s managers told the board they had not been aware of any significant difficulty and had not read the entries. The trouble with the rods “appears not to have been known at Headquarters.” Rod 9 had the best record of the five: 130 scrams in six months and one hesitation. The operating procedures had never been finished to the Commission’s satisfaction; the rod-drive work was taught from a training procedure that listed the mechanical steps. One cadreman said he had lifted rods for the washer and nut about 300 times and never felt one stick.

Twenty inches

The reactor was shut down on 23 December 1960 and the crews came back on 27 December to work three shifts round the clock on maintenance. The one job planned for the core itself was to push forty-four flux-wire assemblies, aluminium tubes carrying tiny cobalt wires, down between the fuel plates to record the neutron flux when the reactor ran again. To reach the core, the drive mechanisms had to come off the rods and the plugs out of the head. The night shift of 2 to 3 January did that, and the day shift, military men with contractor supervision, put in the flux wires. When it left at four o’clock, the head was open at the centre, the mechanism and plug for rod 9 off altogether, the water just under the head, the tools and parts laid out on the operating floor, and the gate locked behind them. Start-up was set for 4 January.

The night order book for the shift of 3 January, written by the plant superintendent, is nine lines long, and the first three are the night’s work. “1. Perform a reactor pump down – procedure No. 54. 2. Reassemble control rods, install plugs, place shield blocks, leave top shield off. 3. Connect rod drive motors.” The lines after them ran through the start-up checks and the rod drop tests to “9. Accomplish a normal startup to 3 MW operation.” The crew did the first, pumping the water down until the recorder read five feet above the core, and logged it. Then they went to the reactor. No nuclear instruments were on. The control room was empty. The reactor was a tank of warm water with a core at the bottom and five rods fully in, and nothing about it was expected to change.

From the state of the head afterwards, the contractor and later General Electric worked out how far they had got. On rod 9 the plug was in, the lifting tool was screwed to the rack, and the washer and nut were on. The operation under way at the instant of the accident, General Electric concluded, “was to raise the central rod manually a sufficient amount to remove the C-clamp.” That is a lift of about two inches, by a man standing on the head with an eighteen-inch pipe for a handle and the heaviest of the five rods on the end of it. Two men were on the head, one over rod 9 and one astride the shield plug of rod 7. The third was between the shield blocks, probably with tools or parts.

Rod 9 came up past sixteen inches, and past twenty. At 16.7 inches, by General Electric’s estimate, the reactor was critical; at twenty it had about 2.4 percent of reactivity above that, far beyond the fraction that delayed neutrons supply, and the chain reaction no longer needed them. In about four thousandths of a second the power reached something between 16,000 and 19,000 megawatts, six thousand times the plant’s rating, and the only thing that stopped it was the water. The centres of the fuel plates in the middle of the core reached the boiling point of uranium-aluminium, about 2,060 degrees Celsius, and burst; steam formed around them faster than it could get away; and steam, being a poorer moderator than water, shut the reaction down. The nuclear part of the accident released about 133 megawatt-seconds of energy, give or take ten, which is what SL-1 made in forty-five seconds of ordinary running. About a fifth of the core melted through to its cladding.

The water did the rest. The steam, at some 500 pounds per square inch, drove the seven-foot column of water above the core upward like a piston, and the water struck the underside of the bolted head at about 160 feet a second. The impact compressed it to about 10,000 pounds per square inch. The shield plugs went out of their nozzles at up to fifty feet a second, and the vessel itself, taking the momentum of the water, sheared its piping, rose about nine feet into the air, hit the fan floor, and fell back into its support cylinder. The whole thing, “from the beginning of control rod withdrawal to the time that the vessel fell back into its support cylinder, was between two and four seconds.” Less than 1 percent of the energy had gone into moving anything; the bolted lid and the two feet of air above the water had done the rest. In an open-topped tank “the most destructive effects would not have occurred at all.”

The two men on the head were killed where they stood. One was thrown against a shield block. The other was carried up with the plug of rod 7, which had about the mass of a man, and pinned by it to the underside of the fan floor thirteen feet above the head. The man between the shield blocks was thrown back against another block and struck his head. The water that came out of the open ports carried the core with it, and the operating floor, the walls and the ceiling were painted with fission products.

On the fan floor the heat alarm went off, and a mile away, at the gate house of the Army’s gas-cooled reactor, a radiation monitor alarmed “and remained erratic for several minutes.” The fire department was at the gate at about 9:10. At 9:35 two health physicists arrived with a meter that read to 500 roentgens an hour, and one of them went to the top of the stairs and looked. “Observed radiation levels of the order of 500 R per hour forced their quick withdrawal. They saw some evidence of damage but no bodies.” The men who would run the night were driving the forty-one miles from Idaho Falls. “At approximately 10:35 p.m. the Combustion Engineering supervisors for plant operations and health physics, wearing Scott Air Paks and carrying two 500 roentgen scale Jordan Radectors, entered the reactor operating floor for less than two minutes. They saw two men; one moving.” They came out, went back with three more men and a stretcher, and were out of the west door in three minutes with the one who was alive. He died in an ambulance at the junction of Highway 20 and the site road. “When the doctor examined the casualty at 11:14 p.m. he pronounced him dead.” Another party found the third man and did not try to bring him down. The five men of the first rescue had taken 23 to 27 roentgens each. The companion piece “Sixty-five seconds” follows the six days that came after, a minute at a time.

The dead man in the ambulance was reading 400 roentgens an hour at the head, and the station had nowhere to put him; at four in the morning, by car headlights, at twenty below, a team cut his frozen coveralls off him in a minute and seventeen seconds, and it made no difference to the readings. The second man was brought down on the evening of 4 January by soldiers in two-man relays with a minute each on the floor. The third was on the ceiling, above the open reactor, in a field of more than 500 roentgens an hour, where anything knocked loose might fall into the core. Volunteers from an Army chemical unit at Dugway Proving Ground rehearsed on a full-scale mock-up, rigged a net on the boom of a crane under the body, and freed it in relays. “The third body was removed from the building at 2:37 a.m. on January 9, 1961.” No two-man team was inside for more than sixty-five seconds.

All three bodies read 100 roentgens an hour at the surface and more than 500 at a foot. A team from Los Alamos did the autopsies with long-handled tools, cut away the flesh in which fragments of the core were embedded, and settled from the injuries where each man had been standing. The caskets were lined with lead and went by military aircraft, each with a health physicist, to the cemeteries the families chose. Richard McKinley was buried at Arlington National Cemetery on 25 January 1961. Of the several hundred people on the site in the first week, twenty-two took between three and twenty-seven roentgens to the whole body. The plan had allowed a hundred to save a life. No one was hospitalised.

What it looked like from inside

The Associated Press had the story on the wire by the morning of 4 January: “A new-style nuclear reactor blew up last night, killed three men and touched off a high level of deadly radiation in a building at one of America’s major atomic testing stations.” A Commission spokesman said it “probably was a chemical explosion.” The reactor had been shut down for eleven days, and to the men on the spot a nuclear excursion in a cold reactor was the less likely of two unlikely things.

The instruments that said otherwise were in the men’s pockets. The brass screw in one man’s cigarette lighter, the buckle of another’s watchband and the gold ring on the third had been made radioactive, copper-64 and gold-198, which only neutrons do, and a metal fragment from one man’s clothing carried the fission product yttrium-91 in a proportion that gave the first estimate of the burst: about 50 megawatt-seconds. The Commission’s statement of 6 January said that “there are strong but not conclusive indications that a nuclear excursion was involved in the incident.” On 19 January it said the explosion had been an accidental runaway.

The general manager, Alvin Luedecke, appointed a board of investigation on 4 January under Curtis Nelson, the director of the Commission’s own Division of Inspection, and it met that evening in Idaho, alongside a Commissioner, the general manager, people from other Commission sites and other federal agencies, and the press. From 10 January to the end of April nobody went into the building at all. A crane with a lead-shielded cab put lights and cameras through the freight doors and over the open reactor, and on 22 February a light was dropped into the vessel and a film made through the ports, which showed “a decidedly chaotic condition.” The board worked from the logs, the night order book, the pictures and the witnesses.

Its report, dated 10 May 1961 and sent to Congress on 5 June, was careful. “At the time of this writing (May, 1961), there still remains substantial doubt concerning the initiating event.” The evidence “strongly indicates a nuclear incident of 50 megawatt-seconds, or more, which could credibly have been induced by rapid and extensive motion of the central control rod,” and “there is no evidence to show that the actions of the operators on duty were in any way different than those prescribed and which had been carried out without incident many times before.” To produce the excursion, the board reckoned, the central rod would have had to go twenty-two to twenty-four inches, the last six or eight of them “at a rate of approximately 24 inches per second.” “From the positions of the men after the incident and the injuries they suffered, we are unable to rule out the possibility that one, or possibly two, of them were engaged in lifting the central rod at the time of the explosion. At present, however, there is no direct evidence on this point.”

Nelson’s covering memorandum to Luedecke, printed with the report, went further. “We cannot say, however, with any certainty, what initiated the SL-1 explosion, and it is possible that we may never know.” But before the accident “the condition of the reactor core and the reactor control system had deteriorated to such an extent that a prudent operator would not have allowed operation of the reactor to continue without a thorough analysis and review.” He set out the hypothesis the board found least implausible, that the rod had stuck in its shroud and come free “while one or more operators were exerting a large upward force on it,” and listed what might have broken the chain: a procedure “containing clear warning and explanation of the possible hazard associated with lifting the rod (rather than only the mechanical steps contained in the training procedure)”; a supervisor on the night shift; an operator in the control room with the instruments on; an inspection routine that had not been stopped; an audit by someone outside the operating organisation. “The hypothesized incident could not have occurred if the amount of withdrawal or the rate of withdrawal of the central control rod had been positively limited by mechanical restraint or by operational procedure.” Responsibility he spread upward, from the contractor through the Idaho office and the Division of Reactor Development to his own division and the general manager. The three men he set apart. “(We specifically absolve the military cadre, as such, from any responsibility.)”

The Joint Committee on Atomic Energy had the report in front of it on 13 June 1961, and Commissioner Robert Wilson told it that “it is all too clear that the accident could and should have been prevented.” He gave the committee six deficiencies, and the first was “a ‘design deficiency’ which permitted the reactor to go supercritical (run away) when a single control rod was withdrawn.” The other five were Nelson’s list. “If we had taken all these precautions, this accident wouldn’t have happened,” Wilson said. “I have no doubt of this in my mind. I can’t prove it because we don’t know exactly what happened, but cumulatively these precautions would have prevented the accident.”

The proof, such as it was, came from the reactor, and the companion piece “Twenty-six and a half inches” follows it. In May 1961 General Electric, whose aircraft nuclear propulsion project at the north end of the station had just been cancelled, took over the recovery, and on 30 November 1961 the vessel was in its Hot Shop forty miles up the road, where everything could be done by manipulator and television. The piece of evidence that mattered had been lying across the top of the head since January: the shield plug assembly of rod 9, whose guide tube had collapsed on to its extension rod and seized it so hard that it took 5,020 pounds to move it in a testing machine. Cut open, the rod and the tube carried matching marks, and a plaster cast of the tube fitted the impression on the rod: “The matching is virtually perfect.” They showed the rod in its plug at the moment the tube collapsed, 26½ inches withdrawn, and then forced back to 2½ inches, where it was found. The core had jumped about seven inches with the vessel before the water hammer collapsed the tube, which accounted for the difference. “The central control rod, however, was determined beyond all reasonable doubt to be in a 20-inch withdrawn position.” It was the only one of the five not fully in.

Then the physicists did what the board could not. They hung a spare central rod from a real drive mechanism in a pool of water, gave several men the actual handling tool from the accident, and timed them with an oscilloscope. Pulling as hard as they could, the men moved the rod from the critical position to twenty inches in about 45 to 56 thousandths of a second; a “casual” but complete pull took 75. An analogue computer then tried these rates against the known energy of the burst, and the pull that fitted best was slower than any of them, a steady ramp of 125 thousandths of a second with the rod just reaching twenty inches at the peak of the power. A maximum effort “is obviously much too fast. Had it been employed, the rod would have gone far past the 20-inch position.” A rapid pull not followed through fitted; so did a pull that someone tried to stop. “There seems to be no way of choosing between these various ways.” Explosives were ruled out by experts who examined the core; a hydrogen explosion under the head was ruled out because it would have lifted all five rods. “Though positive proof is lacking, all information indicates that the excursion was caused by the manual withdrawal, with less than maximum effort, of the central control rod during the assembly of the rod-drive mechanism.” The final report of July 1962 has one sentence about the rest. “No attempt has been made to suggest possible motivations for pulling the rod 20 inches instead of 2.”

The Idaho office had decided on the morning of 4 January to film everything, and in the first of its three films the firemen, health physicists and cadremen who had answered the alarm played themselves. The Commission’s briefing film went out to the industry in 1962 with an animated water hammer and the same two findings: the rod, and no reason.

What happened

The site was cleared. About 475 people made 3,240 entries into the fenced area in the thirteen months of General Electric’s work, and fewer than 6 percent of them went over the quarterly guide. Some 80,000 cubic feet of contaminated equipment and material were buried in pits and a trench 1,600 feet from the compound, and in 1962 the silo itself was cut up and buried with them, behind a fence with warning signs that was, in the Department of Energy’s own words, “the only monument to the reactor.” On 22 June 1962 the ground was handed back to the Commission for other uses. Less than half of one percent of the core’s iodine-131 was ever found on the desert; the building that was not a containment had held almost everything.

The Commission had, within days of the accident, asked the operators of all forty-seven licensed reactors in the country about their shutdown procedures and control systems, and modified some of the licences. Wilson’s first point became a rule of design, and it still is one. The general design criteria that the Commission’s successor applies to every power reactor in the United States require that the control system hold the core within its limits “with appropriate margin for malfunctions such as stuck rods,” and that the design withstand “any single malfunction of the reactivity control systems, such as accidental withdrawal” of a rod. Every reactor licensed since has had to show on paper what one rod, pulled alone, would do.

The physics had one more thing to say, and the Commission arranged to hear it. In November 1962, at the SPERT facility on the same station, a core of plates like SL-1’s was driven on to a period of 3.2 thousandths of a second on purpose, with the roof off the building and everyone at a distance. The nuclear burst did what the theory said and shut itself down; about fifteen thousandths of a second later the molten fuel gave its heat to the water and the core was blown apart. That was the sequence at SL-1, now with instruments on it.

The Army’s programme went on. Its plant on the Greenland ice cap was shut down while its rods and procedures were reviewed, and started again. A month after the accident the Army’s mobile reactor, ML-1, arrived in Idaho on a semi-trailer; it ran, between leaks and failed welds, for 664 hours in all before it was shut down for good in May 1964. The Army cancelled its reactor programme in 1965, in the budgets of the Vietnam war, for the sake of first cost. The DEW Line kept burning oil.

The three men had been absolved by the board, and they have been argued over ever since. The reports leave a space where a reason would go, and in sixty-five years it has been filled with stories: a rod that stuck and let go, a joke, a quarrel, a man who meant it. The site’s own historian, writing for the Department of Energy, put the question as it stands: “did one of the cadremen deliberately withdraw the control rod, and if so, why? Or did the control rod stick, causing over-exertion and a sudden release? All of the science at the NRTS was unequal to this most perplexing question.” One of General Electric’s physicists remembered asking the cadremen whether they had known the reactor would go critical if the central rod were pulled. “Of course!” he was told. “We often talked about what we would do if we were at a radar station and the Russians came. We’d yank it out.”

What it taught

The reactor did what its own measurements of August 1958 said it would do. With the four outer rods fully in, the central rod by itself could make it critical, and the people who built and ran it knew that in the first week, wrote it down, and lived with it for two and a half years. A reactor that one rod can start is a reactor that one hand can start, and the whole safety of SL-1 came to rest on the distance a hand would choose to lift a rack. The procedure fixed that distance at four inches and said nothing about why. The design had given the men lifting it a foot of grace and had told them nothing about where it ended.

The Demon Core is the same story told with a screwdriver. At Los Alamos in 1946 a procedure rested on one steady hand, and when the hand failed it failed in the direction it was moving; the fix was not a more careful hand but an assembly built so that a slip opened the gap instead of closing it. SL-1’s fix was the same kind: after June 1961 no single rod, stuck out or pulled out, could do what rod 9 did, which is a rule about geometry and not about care. The reviewers who wrote that the accident “was probably independent of the poor condition of the core” were making a hard point: the lost boron and the sticking rods had shortened the lift that would do it, but the lift had always been there to make. The condition of the core was what the Commission and the contractor could be blamed for. The design was what killed.

The instrument had been speaking for months, in a language everyone at the plant understood. Two and a half percent of the drop tests had failed in twenty-two months, and 13 percent in the last one; “#1 no drop. #7 no drop” was in the log on the morning the reactor was shut down for Christmas. Flakes of the poison that held the core down had been scooped from the bottom of the vessel in August. Each of these was logged, and each was an old and familiar trouble with a known remedy, and Nelson’s memorandum saw what that did: a new fault arriving in the shape of an old one is the hardest kind to hear. The organisation had every reading and no one whose job it was to add them up, which is the item Nelson put at the top of his list.

The experiment on organisation gave its answer too. The Army wanted to know whether a plant could be run by its own cadre with no engineer in the building at night, and the arrangement was made so that it would be found out: no supervisor, no operator at the console, no instruments on. On the night the reactor spoke, it spoke to an empty control room. The board did not find that the men had done anything they had not been trained to do; it found that the test had been set up so that the first sign of trouble would be the heat alarm.

Then the complication, which the investigators were honest about. They could show, beyond reasonable doubt, where the rod was, that a man could have put it there, and that no explosion did. They could not show why, and they said so in every report, and the space they left has been filled by people with less evidence and more certainty. The one measurement that settled anything was made by men pulling a rod out of a tank of water with the tool from the accident while an oscilloscope watched, and its conclusion is the one that should outlive the stories: nobody had to do anything remarkable. What the record supports without argument is the log entry, the alarm at 9:01, the slide marks at 26½ inches, and a procedure that said four inches and gave no reason.

Check the story before you learn from it. The wire services of 5 and 9 January gave Byrnes as twenty-seven and McKinley as twenty-two, and the notice of McKinley’s burial on 25 January gave them the other way round, which every later account follows. The accident is timed at 9:30 in the first dispatches and at 9:01 by the alarm. The rod was pulled “as rapidly as the operator was able,” says the Los Alamos review of 2000, and “with less than maximum effort,” says the General Electric report it cites. The scratch says 26½ inches and the rod was withdrawn twenty, and both are right. None of it changes where the rod was.

Sources

• US Atomic Energy Commission, SL-1 Accident: Atomic Energy Commission Investigation Board Report, with the General Manager’s letter of transmittal of 5 June 1961 and Curtis A. Nelson’s memorandum of 10 May 1961, printed for the Joint Committee on Atomic Energy, June 1961 (IDO-19302). The administration of the project, the night-shift agreement, the hazards review of 1958, the critical experiments of 1958, the boron and rod history, the night orders and the scram log of 23 December 1960, the sequence of events, the four-inch limit, the 16 inches and the 24 inches per second, the conclusions, and every quotation from the board and from Nelson. inldigitallibrary.inl.gov
• Combustion Engineering, Inc., SL-1 Reactor Accident on January 3, 1961: Interim Report, IDO-19300 (CEND-128), 15 May 1961. The crew organisation and training, the fuel loading, the sticking statistics (2.5 and 13 percent, 63 incidents, rod 9’s 130 scrams), the procedure’s step nine, the shutdown work of 27 December to 3 January, the night order book and the log entry of 3 January, the state of the head, the reconstruction of the men’s positions, and the Commission’s “Sequence of Events” with the times of the rescue, quoted there in full. inldigitallibrary.inl.gov
• General Electric Company, Idaho Test Station, Final Report of SL-1 Recovery Operation, IDO-19311, 27 July 1962. The recovery and the Hot Shop, the 20-inch position, the critical position of 16.7 inches, the 2.4 percent, the seized guide tube and the slide marks, the 26½ inches, the C-clamp, the mock-up withdrawal tests of Table III-10 and the analogue computer, the water hammer and the vessel’s nine feet, the two to four seconds, the exposures of the 475, and “20 inches instead of 2.” inldigitallibrary.inl.gov
• General Electric Company, Additional Analysis of the SL-1 Excursion: Final Report of Progress July through October 1962, IDO-19313, 21 November 1962. The 133 megawatt-seconds, the 4-millisecond period, the peak power of 16,000 to 19,000 megawatts, the melting and vaporisation, and the iodine on the desert. inldigitallibrary.inl.gov
• US Atomic Energy Commission, Idaho Operations Office, Health and Safety Division Annual Report, 1961, IDO-12021. The radiation levels of the bodies, the autopsies and the 3.7 roentgens, the lead in the caskets, the escorts, and the year’s exposure record. inldigitallibrary.inl.gov
• Thomas P. McLaughlin and others, A Review of Criticality Accidents, 2000 Revision, Los Alamos National Laboratory report LA-13638, May 2000, pages 97 to 98. The summary of the accident, the shutdown margins, “as rapidly as the operator was able,” “probably independent of the poor condition of the core,” and the SPERT destructive test of 5 November 1962. osti.gov
• US Atomic Energy Commission, Idaho Operations Office, The SL-1 Accident: Briefing Film Report, 1961 to 1962, forty minutes. Read here through the machine transcript of the copy on the Internet Archive; nothing is quoted from it. archive.org
• Susan M. Stacy, Proving the Principle: A History of the Idaho National Engineering and Environmental Laboratory, 1949 to 1999, US Department of Energy, Idaho Operations Office, DOE/ID-10799, 2000, chapters 15 and 16. The Army’s purpose, the building, the lunch pails, the first hours, George Voelz’s and John Horan’s recollections, the Dugway team, the caskets, the three films, the reactions of the press and the unions, Commissioner Wilson’s testimony, the survey of licensed reactors, the ML-1 and the end of the Army programme, the burial of the building, the 1981 poster, and the two sentences on the question the investigation could not answer. Jay Kunze’s recollection of the cadremen’s answer is quoted at second hand from chapter 15. Chapter 15, Chapter 16
• Associated Press dispatches in the Evening Star, Washington, 4 January 1961 (pages A-1 and A-6), 5 January (page 41), 7 January (page 5), 9 January (page 14), 20 January (page 35), 24 January (page 33) and 11 June 1961 (page 79), and the Nome Nugget, 4 and 9 January 1961: the first reports, “chemical explosion,” McCone’s promise, the statement of 6 January, the recovery of the third body, the names and ages, and the board’s report as reported. 4 January, 5 January, 7 January, 9 January, 11 June
• “Richard L. McKinley, 27, Victim of Atomic Radiation,” Evening Star, 25 January 1961, page 24, on the burial at Arlington and the men’s ages and homes. loc.gov
• William Hines, “Idaho Blast Avoidable, AEC Member Testifies,” Evening Star, 13 June 1961, page A-2. Commissioner Wilson’s six points and every quotation from him. loc.gov
• Title 10, Code of Federal Regulations, Part 50, Appendix A, General Design Criteria for Nuclear Power Plants, Criteria 26 and 27, for “appropriate margin for malfunctions such as stuck rods.” govinfo.gov
• The photographs are US Atomic Energy Commission site photographs of the National Reactor Testing Station: the building, from the Department of Energy’s archive on Flickr, which marks it a United States Government Work (HD.6B.007); and INEEL 61-9, 61-667 and 81-3966, as reproduced in the Department of Energy’s history above, via Wikimedia Commons (61-9, 61-667, 81-3966).