At 3:30 P.M., both air-radiation monitors in the reactor building indicated a sharp rise in activity. By 5:00 P.M., radioactive air was going up the exhaust stack and into the atmosphere, and the radiation level over coolant channel seven was extremely high, at 25 roentgens per hour. Something had obviously broken, but for some reason it did not occur to the experimenters that radioactive products that would cause such an indication are produced in the fuel, and the fuel is welded up tight in stainless steel tubes. If all the fuel is intact, then nothing radioactive should be leaking into a coolant channel, and certainly not through the gas-tight reactor vessel and into the air in the room. By 8:57 P.M. they had shut down the reactor. They fixed the problem of leaking radioisotopes by replacing the sodium-level indicator over channel seven with a shield plug, and they restarted the reactor at 4:40 a.m. the next day, July 13.
By 1:30 that afternoon, they noticed that the graphite temperature was not going down when they increased the coolant flow. They knew not why. At 5:28 P.M., they were running at 1.6 megawatts and commenced a controlled power increase. The power seemed to increase faster than one would expect up to 4.2 megawatts, but then the reactor went suddenly subcritical and the power was dropping away. They started pulling controls to bring it back, and by 6:21 P.M. they had managed to coax it to run at 3.0 megawatts.
Up to this point, the reactor had been recalcitrant and unusual, but now it went rogue. Power started to run away. Control-rod motion was quickly turned around, sending them back into the reactor to soak up neutrons and stop the power increase. Instead, the power rise speeded up. By 6:25 P.M., the power was on a 7.5-second period, or increasing by a factor of 2.7 every 7.5 seconds. The fission process was out of control, and a glance at the power meter showed 24 megawatts and climbing.
Deducing that if this trend continued, then in a few minutes the reactor would be a gurgling puddle in the floor, the operator palmed the scram button, throwing in all the controls in the reactor at once and bringing the errant fissions to a stop.
As the reactor cooled down, only one question came up: Why was there not an automatic scram when the reactor period, spiraling down out of control, passed 10 seconds? The period recorder, which leaves a blue-line-on-paper graph of period versus time for posterity, had a switch that was supposed to be tripped when the pen hit 10 seconds on the horizontal recorder scale. This switch was supposed to trigger an automatic scram.[143] Testing found that the switch would have worked, but only if the period had been falling slowly. The trip-cam was modified so that the switch would operate even when a scram was most needed, with the period falling rapidly.
Seeing nothing else to fix, the operating staff brought the reactor back to criticality at 7:55 P.M. and proceeded to increase the power. By 7:00 a.m. the next morning, July 14, they were running hot, straight, and normal at 4.0 megawatts. Two hours later, the radioactivity in the reactor building was reading at 14,000 counts per minute on the air monitors. Technicians put duct tape over places where fission products were found escaping. There was only one other automatic scram for the whole rest of the day, when workers setting up a test of the main primary sodium pump accidentally short-circuited something.
On July 15, it was seen as pointless to be trying to run the electrical generator while trying to test at high temperature, so the staff drained out the secondary coolant loop and switched to the air-blast heat exchanger. The next day, at 7:04 a.m., the SRE was made critical once more. On July 18, the motor-generator set, which was supposed to prevent power surges into the control room instruments, failed. The operators switched power to unstabilized house current and continued operation. At 2:10 a.m. on July 21, the reactor scrammed suddenly, having picked up another fast power rise. The scram was attributed to the unstabilized power, and the reactor was restarted 15 minutes later. One more scram at 9:45 a.m.
The next day, July 22, channel 55 was giving trouble. This assembly contained various experimental fuels, and the temperature was fluctuating in the 1,100 to 1,200°F range. There was only one automatic scram on July 23, probably just a fluke, but by 1:00 P.M. the temperature in channel 55 was up to an eyebrow-raising 1,465°F. Operation continued. In the early morning on July 24, eight hours were spent trying to dislodge some apparent debris stuck in the fuel channels by jiggling the assemblies. It was noted in passing that four of the fuel modules seemed jammed and stuck firmly in place. There were two annoying automatic scrams later that day.
At 11:20 a.m., the Sodium Reactor Experiment RUN 14 was terminated, and the long-suffering machinery was allowed to rest quietly while the experimenters poked around in the core with a television camera. To their surprise, they found that the core of a nuclear reactor that had been acting oddly for six weeks, subjected to overheating and temperature fluctuations, many automatic scrams, pump seal coolant failures, oxidizing sodium, radiation leakage, and a power runaway, was wrecked. Of the 43 sealed, stainless-steel fuel rods in the core, 13 had fallen apart, scattering loose fuel into the bottom of the reactor vessel. How the thing had managed to run at all under this condition was an amazement in itself. Attempts to remove the fuel rods came to an end when the contents of channel 12 became firmly jammed in the fuel handling cask. An investigation of the damage and its cause was started immediately.
The interim report, “SRE Fuel Element Damage,” was issued on November 15, 1959. It was found that tetralin had been leaking into the sodium coolant through the frozen sodium seals in the pumps. In the high-temperature environment of the active reactor core, the solvent had decomposed into a hard, black substance, which would tend to stick in the lower inlet nozzles of the graphite/fuel modules and prevent coolant from flowing. In the blocked coolant channels, the sodium vaporized, which had been thought unlikely, and denied coolant to the fuel. The stainless steel covering the cylindrical fuel slugs melted, and structural integrity of the fuel assemblies was lost. Naked uranium fuel, having fissioned for hundreds of megawatt-hours, was able to mix with the coolant. Gaseous fission products presumably escaped the sealed reactor vessel, probably through the same leakpoint that allowed the sodium to oxidize, and other fission waste dissolved in the coolant, making the primary loop radioactive.
The most puzzling part of these findings was: Why did the stainless steel melt instead of the metallic uranium? The 304 stainless used for fuel cladding melts at 2,642°F, and the temperature in the reactor was nowhere close to that. Uranium melts at 2,060°F, and yet the stainless steel melted away, leaving the uranium unclad and unsupported. The fuel assemblies were, however, operating at temperatures for which the reactor was not designed. Where fuel slugs within the tubes were leaning against the inner walls, uranium diffused into the stainless steel, making a new alloy, a stainless steel/uranium eutectic. This uranium/steel mixture melts at 1,340°F, so with the reactor core overheating, the fuel assemblies fell apart. This problem would have been hard to foresee.
Another puzzle was harder to figure out. Why did the reactor run away, increasing power on a short period? The power excursion was simulated using the AIREK generalized reactor-kinetics code running on an IBM 704 mainframe. If the parameters were tweaked hard enough, the simulation could even be forced to agree with the recorded data, but even a well-tuned digital simulation could not indicate why the transient had occurred. It was easier to explain the subcritical plunge than the short-period lift-off. Further calculations and physical experiments proved that it was the sodium void in the blocked coolant channels that caused the fission process to run wild. It was the same phenomenon that caused the fuel assemblies to melt. Graphite is a better neutron moderator than just about anything, and the fission process improves when there is no other substance, such as coolant, in the way.
143
At least I assume it was a scram actuation. In the SRE plans, this switch is listed as a “setback ” actuator. Does this indicate that it was meant only to limit the power rise rate and not stop it completely? If so, it would not have worked as planned. With power increasing exponentially, a scram is quite appropriate.