An excellent paper was posted to the thorium-forum recently on the work being done at the Nuclear Research Institute at Rez in the Czech Republic. The paper was written by Jan Frybort and Radim Vocka and was entitled:

Neutronic Analysis of Two-Fluid Thorium Molten Salt Reactor

Regular readers of this blog know that I am a big fan of the two-fluid fluoride reactor. What is the two-fluid reactor? It’s the version of the fluoride reactor where the fuel salt (containing the fissile material, typically U-233) is kept separate from the blanket salt (containing the fertile material, typically thorium).

The one-fluid reactor, on the other hand, has the fissile and fertile materials mixed together in one salt.

Why do I like the two-fluid reactor so much? A couple of reasons. The first one is is that it offers some really exciting improvements in safety. It’s much more responsive the changes in temperature than a one-fluid reactor, because heating up the salt causes the fuel to expand out of the core more than the fertile material.

It’s a lot easier to process than the one-fluid reactor, because a whole class of fission products (the lanthanides) that behave a lot like thorium are kept separate from thorium.

And it’s a lot easier to scale up and down, because the blanket fluid acts like a neutron shield in each case.

So what’s the downside of the two-fluid reactor? There’s two big ones. The first is the problem of how to keep the two fluids separate in a way that doesn’t eat too many neutrons and will last for a good long while during regular operation without getting too messed up. The second is the tendency of the two-fluid reactor to have a problem if the blanket fluid changes–what’s called the “blanket void coefficient”.

I think both problems will ultimately be solvable, and I have some good ideas, but I don’t have the definitive answers yet.

But back to the paper from Rez…

In this paper, Dr. Frybort and Dr. Vocka look at the two-fluid reactor design that ORNL was working on back in 1967 with fresh eyes. They do this because they had been looking at one-fluid reactor designs and found them wanting in several categories, just as I have. They wanted a reactor with strong negative temperature coefficients, simple processing, and good performance, and the one-fluid reactor had serious issues in all categories. So they rewound the clock from 1974 to 1967 and looked at where the Oak Ridge Lab had gotten when they set the two-fluid reactor down.

Oak Ridge worked on two-fluid reactor design seriously for about two years, from early 1966 to late 1967. Their work is detailed in three semiannual progress reports (ORNL-4037, ORNL-4119, ORNL-4191) and one final report (ORNL-4528). All of these reports are available in the document repository linked from this site.

ORNL was specifically trying to define a “molten-salt breeder experiment” that they called the MSBE, that would have been a follow-on to the Molten-Salt Reactor Experiment (MSRE), that ran successfully from 1965 to 1969. In order to know what should go into an MSBE, they needed a pretty good idea of what the eventual “molten-salt breeder reactor” (MSBR) would look like. They needed to start with the end in mind, so they devoted themselves to trying to understand what a two-fluid MSBR looked like, and from there they would backtrack and try to define an experimental reactor that would prove its capabilities.

Defining an MSBE was a challenge. The MSRE was far less than that–it only had a single fluid and had no blanket where thorium would be converted into new fuel. So the first thing to do would be to define a new core geometry where both goals were being satisfied simultaneously. This is where ORNL ran into the first problem of the two-fluid reactor–the so-called “plumbing problem”.

They tried to fit blanket salt, fuel salt, and graphite elements in the core in a way that would work neutronically (it would achieve criticality) and in a way that would work from a fuel-cycle perspective (it would produce more fuel than it consumed). The second part required understanding the fuel and blanket processing system just as much as the reactor itself.

ORNL reactor designers experimented with a variety of core geometries. Most were hexagonal elements of graphite with recursive tubes of fuel flowing through them. The whole arrangement was then immersed in a bath of blanket salt. This arrangement was actually quite good at satisfying the first and second conditions. Very good in fact. It was also easy to process and had strongly negative temperature coefficients. All good. But it was not good holding up to the neutron flux. Data indicated that the graphite would shrink and then swell. As that happened, the relative volumes of fuel and blanket would change, and that would alter performance. Additionally, and this was never mentioned in any of the design documents, so well as I can discover–these designs suffered from the problem of the blanket void coefficient. If blanket salt drained out, core reactivity would go up–way up, and the reactor would respond by the core temperature increasing. Eventually it would reach the point where the freeze plug would melt and the core would drain out into the drain tank.

In the paper from Rez, the performance of this design was reevaluated using modern tools, and the authors verified that the two-fluid reactor had the advantages that ORNL scientists had anticipated. In part 4A, they found that the temperature coefficient was strongly negative and that the breeding factor was good. In part 5, they looked at changing the design to improve it, and found that by making the fuel channels bigger than the original design, they improved nearly all parameters. This is an important result, since it’s not often that you change a parameter and find improvements in nearly all outputs from that parameter.

In part 7 of the paper, they mention the second key issue with a two-fluid reactor–the problem of the blanket void coefficient. Since the original ORNL design had the problem, and since they modeled only parametric variations on that original design, it’s no surprise that the problem still shows up. It must be fixed, probably through a new design approach to the two-fluid reactor. I have some ideas, most all of them based around physical situations where a loss of blanket fluid leads to a loss of moderation. I anticipate that this could be done by floating moderator elements (graphite) in the blanket salt, so that as the level of the blanket salt falls, the moderation decreases more than the absorption decreases from the loss of blanket. These ideas definitely need more modeling, but I think they are essential
ly sound.

The scientists from Rez have done us all a great service by “blowing the dust” off the old ORNL design, figuring out how it works with modern tools, and figuring out a few ways to make it better. There is a lot more work to be done, but recent work like this is often the catalyst to getting more scientists and engineers working on a problem that has the potential to have such an important result.

Thank you Dr. Frybort and Dr. Vocka!