Because Fermi 1 was designed as a plutonium breeder, liquid sodium was chosen as the coolant. A fast breeder core was much smaller than a thermal reactor of similar power, so the coolant had to be more efficient than water, and to hit the activation cross section resonance in uranium-238, the breeding material, the fast neutrons from fission had to endure a minimum loss of speed. Sodium was the logical choice, and two sodium loops in series were used, similar to the SRE design. It doubled the complexity of the cooling system, but it ensured that no radioactive coolant could contaminate the turbo-generator.
The use of sodium made things difficult and complicated. Refueling, for example, was comparatively simple in a water-cooled reactor. To refuel a water-cooled reactor you shut it down, and after it cools a while, unbolt the vessel head, crane it off, and lay it aside. Flood the floor with water, which acts as a coolant and a radiation shield. You pick up the worn-out uranium in the core, one bundle at a time using the overhead crane, and trolley it over to the spent-fuel pool. Carefully lower new fuel bundles into the core, drain the water off the floor, and replace the steel dome atop the reactor vessel. You’re done.
There is no way to do this with a sodium-cooled system. For one thing, the sodium gets radioactive running in the neutron-rich environment of fission, and it would take a week of down-time just to let it decay to a level of marginal safety. You cannot let it cool down, because in this state the sodium is solid metal, and you could not pull the fuel out of it. It cannot be exposed to air, and everything done must be in an inert-gas atmosphere. The entire building would have to be pumped down and back-filled with argon. Also, sodium, liquid or solid, is absolutely opaque. You cannot see through it, so, unlike looking down from the refueling crane through the water to see what you are doing, you would have to be able to refuel the reactor blindfolded. If anything weird happened in a water-cooled reactor core, you could conceivably stick a periscope down in it and have a look. Not so with sodium coolant. Fermi 1 was to be an electrical power-production reactor, running at 200 megawatts. Although in later terms this would be a very small reactor, in 1961 it was a clear challenge to design a machine of this size running in sodium.
The problems of dealing with refueling a sodium-cooled fast reactor were solved by enclosing the works in a big, gas-tight stainless steel cylinder, about two stories tall. In the bottom of the cylinder, off to the side, was an open-topped stainless tank containing the U-235 reactor core, surrounded by the U-238 breeding blanket.[146] Liquid sodium in the primary cooling loop was pumped into the bottom of the reactor tank through a 14-inch pipe. The sodium flowed up, through small passages between fuel-rods, taking the heat away from the fissioning fuel. The thick, hot fluid was removed by a 30-inch pipe at the top and sent to a steam generator for running the electrical turbo-generator. A “hold-down plate” pressed down on the top of the fuel to keep it from being blown out of the can by the blast of liquid sodium coming up from the bottom. Under normal operation, the liquid sodium in the structure was 34 feet deep. Any open space left was filled with inert argon gas.
An ingenious, complex mechanism was used to refuel the reactor and extract and replenish the irradiated U-238 rods in the breeding blanket. To one side of the reactor tank and inside the stainless vessel was another round can, this one containing a revolving turntable loaded with fuel and blanket rod-assemblies. It worked like the cylinder in a revolving pistol or the carousel in a CD changer, indexing around and stopping automatically with a rod assembly in position to be extracted or inserted. An exit tube led vertically to an airlock on the top floor of the reactor building, where a refueling car ran on rails and was able to reach down into the reactor vessel, through the air lock, and remove or replace rod assemblies in the rotor. The car, gamma-ray shielded by 17,500 pounds of depleted uranium, was driven back and forth on the rails by a human operator, under absolutely no danger from radiation in the reactor, the coolant, or in fuel being inserted or removed.
To refuel the core, the hold-down device was first lifted off the top of the core. The fuel-transfer rotor was turned by a motor until it clicked into place with an empty slot under the handling mechanism, which was a long, motor-driven arm hanging down next to the reactor. The arm then turned to the core, gently snatching a designated rod assembly, picking it up and out of the reactor, swinging over to the transfer rotor, and lowering it down into the open slot. The rotor then rotated clockwise and clicked to the next open slot. After the rotor became completely full, the refueling car would bring up the rod assemblies one at a time and put them into shielded storage. To put new fuel or breeding material in the reactor, the process was reversed, with the empty rotor first loaded with fresh fuel from the refueling car. The rotor would index around, presenting the new rod assemblies one at a time to the arm, which would pick them up, swing them through the core, and insert them. All this happened under liquid sodium. It was a totally blind operation, depending only on mechanical precision to find rod assemblies where they were supposed to be and transporting them with great accuracy. When a refueling operation was completed, the hold-down device would descend and cover the top of the core. To be able to automatically refuel this reactor from the comfort of a swivel chair by pushing a few buttons was a marvel to behold.[147]
Given the recent tribulations at the SRE, great attention was given to the design of the coolant pumps. Made of pure 304 stainless steel, one primary pump was the size of a Buick. It was a simple centrifugal impeller, driven by a 1,000-horsepower wound-rotor electric motor, with a vertically mounted drive shaft 18 feet long. It turned at 900 RPM to move 11,800 gallons of thick, heavy, 1,000°F liquid sodium per minute and lift it 360 feet. The motor speed was controlled using a simple 19th century invention, a liquid rheostat, as once used to dim the lights in theaters.
There were no liquid-cooled ball bearings to worry about, and no seal to keep sodium out of the pump motor. There was no tetralin seal coolant, of course. The main bearing at the impeller end was a simple metal sleeve, with the stainless shaft running loose inside it. It was a “hydrostatic bearing,” like the bearing on a grocery-cart wheel, but instead of having oily grease to keep it from squeaking, the lubricant was liquid sodium, supplied out of the fluid that the impeller was supposed to pump. The sodium was allowed to puddle up in the long shaft gallery, rising several feet above the impeller housing. Its depth was controlled by compressed argon gas introduced into the top of the pump, and this kept the sodium off the motor and out of contact with air.
As one last step before the reactor tank was assembled, the engineers decided to enhance a safety feature. A problem with fast reactors was that in the unlikely event of a core meltdown, the destroyed fuel matrix, becoming a blob of melted uranium in the bottom of the tank, could become supercritical. In this condition, it would continue making heat at an increasing rate and exacerbating a terrible situation. An ordinary thermal, water-cooled reactor, losing its symmetry and carefully planned geometric shape in a meltdown, would definitely go subcritical, shutting completely down. The thermal reactor depends greatly on its moderator, the water running between fuel rods, to make fission possible. The fast reactor does not.[148] To ensure a subcritical melt in a fast reactor, at the bottom of the tank is a cone-shaped “core spreader,” intended to make the liquefied fuel flow out into a shape that does not encourage fission, flat and neutron-leaky, like pancake dough in a skillet. Before the core tank was finished, the builders installed triangular sheets of zirconium on the stainless steel core spreader, just to make it more high-temperature resistant. They beat it into shape with hammers, making the sheet metal conform to the cone. This last-minute improvement was not noted on the prints approved of by the AEC.
146
A breeder reactor is supposed to run off fuel that it produces, which in this case was plutonium-239. Until it has actually made fuel and it has been extracted from the breeding blanket, the breeder usually starts off with a load of uranium-235. Unlike a water-cooled reactor, a fast breeder runs off pure fissile isotope, so the starter fuel must be highly enriched to simulate its eventual load of Pu-239.
147
When I was in school, fast breeder reactors were all the rage, and a wild story about the refueling machinery in the Fermi 1 was going around. It was said that a spin-off of the breeder reactor project was the AMF pin-spotting machine, now used in many bowling alleys. The method used for collecting and placing fuel assemblies in an automatic, utterly blind situation under liquid sodium was found useful for doing a similar job with bowling pins. I am sad to say that I can find no connection between bowling alley machinery and fast reactor refueling devices, much as I would like to. The patent for the AMF “candlepin” pin spotting machine was filed on April 14, 1964, which is close to when the Fermi 1 was designed, but it was based on the original AMF 82–10 pin spotter, first demonstrated in 1946.
148
An enduring question of fast reactor design is: “What is the sodium void coefficient?” If the sodium coolant leaks out of a fast reactor, does it go subcritical or supercritical? The correct answer if this question should come up on your Ph.D. qualifying exam is: “That depends.” There are many factors that come to bear, including the geometrical shape of the core.