The entry says that the number which explained the accident was found as a scratch on a steel tube. This piece follows the people who found it, the General Electric Company’s SL-1 project at the north end of the Idaho station, from the day they took over the ruined building to the two reports of 1962 in which they said what they could prove and what they could not. The board of investigation had worked from logs and witnesses. General Electric had the reactor.

The Hot Shop

In March 1961 President Kennedy cancelled the aircraft nuclear propulsion project, and General Electric’s Idaho Test Station, at Test Area North forty miles up the road from SL-1, found itself with five hundred people, most of them due to be transferred or let go, and the best remote-handling facility in the country standing idle: the Hot Shop, a cell with walls thick enough for a reactor to be dismantled inside it by manipulators while the operators watched through leaded glass and television. The Idaho office had asked earlier whether the Hot Shop might be used for SL-1 and been told it was not available. After the cancellation it was. General Electric took over the recovery in May 1961 as what the Commission later numbered Phase 3, and its physicist Jay Kunze said long afterwards that the scientists thought of it as a choice assignment: BORAX and SPERT had been built to blow up and had declined to, and here was a reactor that had done it on its own.

The building first had to be cleared. The operating floor read 20 roentgens an hour inside the freight door and 100 over the reactor on 29 May 1961, with the fan floor above it as bad, and a ratio of beta to gamma radiation so high that the skin dose limited how long anyone could stay. A control point was set up in a building by the fence; everyone passed through it in and out, in protective clothing, with a health physicist timing the entry. Hundreds of General Electric’s people, and men from the other contractors and from Dugway, volunteered to take their quarter’s allowable dose in a few minutes of work. “To clean it up, people had to make short trips inside and do limited tasks within a couple of minutes and then get out,” Kunze recalled. “Even though you suited up, those couple of minutes would expose you to your quarterly dose of radiation, and you couldn’t go back in for three months.” The timekeepers stood halfway down the stairs and banged on metal when a man’s vacuuming was over. Big pieces came out by remote means, small debris by sweeping and vacuuming, and shielding went over the vessel head. Of the roughly 475 people who made 3,240 entries into the fenced area during Phase 3, fewer than 6 percent went over the exposure guide, the worst of them by 16 percent. About 80,000 cubic feet of contaminated material went into a burial ground beside the site, each item logged with its radiation level and date.

By November the fan floor had been dismantled and the upper sections of the silo removed. To learn whether the core could be moved without disturbing the evidence, a hole was drilled through the shielding, the gravel, the support cylinder and the bottom of the pressure vessel, and the underside of the core was examined and photographed through a boroscope. The planners had meant to cut the vessel’s piping with torches, in relays of welders, until someone noticed small dents in the fan floor that matched the vessel head and worked out that the water hammer had already lifted the vessel nine feet and sheared every pipe for them. On 29 November 1961 a sixty-ton crane lifted it out into a cask, and on 30 November it was in the Hot Shop. The head came off, with difficulty, because the flange and its studs had been bent by the pressure; the loose parts lying above the core were lifted out with the manipulators; and among them was the central control blade, “tightly bound in its shroud in a position corresponding to 20-inch withdrawal.” A boroscope through the bottom confirmed that the other four blades were fully in. In April 1962 the core, in a tank built for the purpose, was filled with water and found to be grossly subcritical, with an effective multiplication of about 0.6; it had given its one burst. Then the vessel was cut apart and the core taken to pieces, every piece to the laboratory.

Shield plug 9

The decisive piece had been in plain sight since January, lying across the top of the vessel head, where it had fallen back after hitting the fan floor: the shield plug assembly of rod 9, with its guide tube and the extension rod inside it. In May 1961 it was lifted out of the building and taken to the Radioactive Materials Laboratory, where at up to 25 roentgens an hour it had to be examined by periscope. The guide tube was collapsed in two places, near the top and near the bottom, and the lower collapse had seized the extension rod so hard that the manipulators could not pull it free. The plug was coated in a white crystalline salt, which proved to be boric acid; the extension rod wore a hard black deposit, which proved to be chromous acid, the product of long service in steam and not, as first thought, of a high-temperature reaction. The assembly was cut at three points to isolate the seized section as a piece four and three-quarter inches long, decontaminated, and examined by hand.

Everything was measured. The guide tube had stretched a quarter of an inch. The seized section, put in a testing machine, needed 5,020 pounds to start the rod moving through it and 3,600 to keep it sliding. The shield plug flange, seven-eighths of an inch of stainless steel, had been bent at two opposite edges, and the imprints matched an angle bracket in the fan floor on one side and the corner of a condenser on the other. Under the microscope the surfaces inside the seized section showed “the very important fact that the final travel of the rod was upward with respect to the shield plug after the seizing action started.”

The rest of the mechanism was found and matched piece by piece. A thirty-two-and-a-half-inch length of the gear rack, with its washer and nut still on it, was recovered from the fan floor; the broken end of its stud was found inside the control rod handling tool, which was embedded in the fan floor above nozzles 5 and 6; a galled thread on the stud, which turned out to be old, served to prove the parts belonged to rod 9. The connecting rod lay on the vessel head with its roll pins sheared. A broken C-clamp, open to the diameter of the rack, was found on the fan floor; had it been on the rack, the rack would have been raised about two and a half inches. So the state of rod 9’s mechanism at the instant of the accident was known: plug in the nozzle, washer and nut on, handling tool screwed to the rack, C-clamp probably in place, “and the operation being carried out at the instant of the accident was to raise the central rod manually a sufficient amount to remove the C-clamp.”

The slide marks settled the distance. Down the extension rod ran longitudinal scratches in the black deposit, three of them spaced about 120 degrees apart, which at first were put down to the manipulator hands; but they ran up inside the guide tube where no manipulator could have reached. At the end of the most prominent of them, twenty-one and eleven-sixteenths inches down from the top of the guide tube, was an impression in the rod, and a plaster cast of the inside of the tube where it had collapsed matched it: “The matching is virtually perfect.” In the positions in which the parts were found, the two impressions were 24 inches apart, give or take an eighth. Put back together, “the extension rod and plug at the time the guide tube collapsed, were in a relative position corresponding to 26½ ± ¼ inches withdrawn. In the ensuing events of the incident the extension rod was forced through the collapsed guide tube to a final, securely clamped position corresponding to 2½ inches withdrawn.” The rod had been found almost home because the water hammer had driven it home.

Twenty-six and a half inches was the rod’s position in its plug when the tube collapsed, and the collapse came after the core had been lifted some seven inches with the jumping vessel. General Electric’s reconstruction, which the Commission’s briefing film drew as an animation, ran: the rod withdrawn about twenty inches by hand; the vessel’s jump carrying the core and the rod up another seven relative to the plug; the water hammer collapsing the tube and seizing the rod at about twenty-seven; the plug and rod accelerated out of the nozzle as one unit; the handling tool striking the ceiling and snapping the rack’s threaded end, which stayed in the tool; the plug falling back on to the head and the connecting rod snapping. “The central control rod, however, was determined beyond all reasonable doubt to be in a 20-inch withdrawn position.” At 16.7 inches, from the limited critical experiments that had ever been done with the central rod alone, the reactor was critical. At twenty it had 2.4 percent of reactivity in hand, give or take 0.3.

The tank and the oscilloscope

The board had asked in May 1961 whether a man could have moved the rod far enough and fast enough. General Electric answered by trying. A spare central control blade was attached to a control-rod and shield-plug mechanism and hung in a pool of water, without its shroud, so that the blade weighed in a man’s hands what it had weighed on 3 January, and the subjects used “the actual ‘incident’ handling tool,” the eighteen-inch pipe with a hex nut welded to its end. The movement always began from the scram position or with the rod held up on a C-clamp about three inches above it. Each man pulled alone unless the table says otherwise, and the position of the rod against time was recorded on an oscilloscope.

The table is short. Told to use maximum effort with the C-clamp in place, the subjects averaged 49 thousandths of a second from the critical position at 16.7 inches to twenty inches; with no clamp and the rod resting on its spring, 46; two men pulling together, 45, and 375 thousandths from fully inserted to twenty inches; a rod held stuck and quickly released, 56. With the tool gripped at its middle, 77, and one subject could not get the rod past twenty-seven inches; gripped at the top, 94, and another could not get it past twenty-two. “A ‘casual’ but complete withdrawal of the rod averaged 75 msec from the critical position to the 20-inch withdrawn position, about 60% longer than the time for maximum effort withdrawal.” Two other kinds of pull were simulated and plotted: a quick lift not followed through, which sent the rod to about twenty inches and let it fall back, and a quick lift arrested as soon as the man realised he should not have made it, “not to the occurrence of the nuclear excursion, but only to his realization that he should not have pulled the rod.”

Then the physics was run on an analogue computer. The reactor’s kinetics were set up with ramp withdrawals of various speeds from the critical position to twenty inches, three groups of delayed neutrons, a starting power of 300 to 400 watts, feedback from the expanding plates, and a shut-off that cut the excursion when the energy reached 130 megawatt-seconds. The casual pull, 75 thousandths, brought the rod to twenty inches about 30 thousandths of a second before the peak of the power; had the pull not stopped at twenty inches, it would have reached 21.2 by the peak, and 2.9 percent, which was outside the error of the reactivity estimate. A ramp of 100 thousandths brought the rod to twenty inches 13 thousandths before the peak, which was within the error. A ramp of 125 thousandths “found the rod just reaching the 20-inch position at peak power, and therefore is the best fit.” “The ‘maximum effort’ rate is obviously much too fast. Had it been employed, the rod would have gone far past the 20-inch position.” The free-fall and arrested-motion curves, properly chosen, fitted as well as the slow ramp. “In fact, a variety of ways (such as gripping the handling tool near the top, pulling of a stuck rod, etc.) exist in which the rod could be withdrawn manually to the observed 20-inch position in the time interval indicated by the analysis of the reactor kinetics. There seems to be no way of choosing between these various ways, since all of them can result in the same final” position.

What was ruled out

Two other mechanisms had been proposed in the first months, and General Electric disposed of both. The first was a chemical explosive in or under the core. Explosives experts from the Stanford Research Institute examined the core and reported, in an appendix printed in full, that chemical explosives were not the agent that had lifted the rod. The second was hydrogen, made by the radiolytic decomposition of water during eleven days of shutdown, collecting under the head and exploding. The literature made it credible that enough hydrogen could have formed. But an explosion in the space between the water and the head would have pressed on all five loose shield plugs alike, and the nuts and washers were known to have been on all five racks, so that ejecting a plug meant lifting its blade. “It is inconceivable that only the central control blade, the heaviest of the five, would have been lifted by the hydrogen explosion. Since the other blades were found in their normal scram positions after the incident, it is immediately apparent that the initiating mechanism was not a hydrogen explosion or any other explosion in the space between the water and the vessel head.”

The rest of the physical evidence fell into place behind the rod. Scale models of the vessel fired at Aberdeen Proving Ground with charges of Pentolite reproduced the bulges in the vessel wall and the distortion of the head flange; a half-pound charge in a quarter-scale model did the damage, which scaled to about thirty-two pounds in the real vessel. The neutron activation found on the operating floor and in the men’s possessions, far more than could have come through seven feet of water, was explained by delayed neutrons from the fuel while the vessel was in the air and from fuel thrown out of it. The flux wires that the day shift had installed on 3 January, forty-four of them fresh and unexposed, were recovered and read, and gave the energy of the burst region by region: 80 megawatt-seconds in the sixteen central assemblies and 53 in the outer twenty-four, 133 in all, with a standard deviation of about ten. The melting in the plates, read under the microscope, gave the period, about four thousandths of a second. The missing uranium, about 2 kilograms from the centre of the core, and the uranium recovered from the debris, 1,177 grams, gave the fraction of the fission products that had left the vessel, which was somewhere between 3 and 10 percent, and less than half of one percent of the iodine was ever found on the desert.

The sentence that was not written

The final report of the recovery is dated 27 July 1962 and runs to more than three hundred pages. Its conclusion on the cause is one paragraph. “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.” Its conclusion on everything else is one sentence, set apart in the analysis. “No attempt has been made to suggest possible motivations for pulling the rod 20 inches instead of 2.” The supplementary report of November 1962 added the refined energy and the peak power and said the same thing more plainly: “The motivation for withdrawing the rod is not known, but the analysis indicates that the action was not a maximum effort attempt to withdraw the rod completely.”

The Los Alamos reviewers who summarised the accident in 2000 wrote that the best evidence “suggests that the central rod was manually pulled out as rapidly as the operator was able to do so.” General Electric’s analogue computer had said the opposite: a maximum effort would have sent the rod far past twenty inches, and the pull that fitted was slower than casual. The two statements have stood side by side in the record for twenty-five years. The one made by the people with the rod in front of them is the slower one.

Kunze, the General Electric physicist, remembered one more experiment, which did not go into the report. “In King Arthur fashion,” the site’s historian wrote from his account, men of lesser, similar and greater strength than the crew tried to lift the rod on the mock-up, and most managed without difficulty. The scientists also asked the cadremen whether they had known the reactor would go critical if the central rod were removed. “Of course! We often talked about what we would do if we were at a radar station and the Russians came. We’d yank it out.”

Sources

• General Electric Company, Idaho Test Station, Final Report of SL-1 Recovery Operation, IDO-19311, 27 July 1962: sections I, II and III, and IV-1, for the cleanup, the Hot Shop, the shield plug examination, the slide marks and the 26½ inches, the C-clamp, Table III-10 and the analogue computer, the Stanford Research Institute’s appendix, the hydrogen argument, the Aberdeen tests, the activation, and the two sentences on cause and motive. inldigitallibrary.inl.gov
• General Electric Company, Additional Analysis of the SL-1 Excursion, IDO-19313, 21 November 1962, section II, for the 133 megawatt-seconds, the period, the peak power, the fuel inventory and the iodine. inldigitallibrary.inl.gov
• Thomas P. McLaughlin and others, A Review of Criticality Accidents, 2000 Revision, LA-13638, 2000, page 97, for “as rapidly as the operator was able.” osti.gov
• Susan M. Stacy, Proving the Principle, US Department of Energy, 2000, chapter 15, for Jay Kunze’s recollections of the cancellation, the cleanup, the dents in the fan floor, the lifting trials and the cadremen’s answer, all quoted at second hand from her interview. inl.gov
• US Atomic Energy Commission, Idaho Operations Office, The SL-1 Accident: Briefing Film Report, 1962, for the animated reconstruction, read through the Internet Archive’s machine transcript and not quoted. archive.org