Hacker News Reader

It's not quite like that.

> An atomic bomb is a prompt fast supercritical system.

Yes, it is, but more precisely it is a fast hypercritical system. The difference is this: in a supercritical system the ratio of neutrons in a new generation vs the old generation (generally denoted by k) is higher than one. For example 1.001 qualifies as supercritical. In a hypercritical configuration, k is higher than 2. In nature things generally evolve in a continuous way. To get to 2 you need to first get to 1 and 1.001, for example. For a bomb, it is very important to get to 2 extremely quickly, so that the assembly does not start the chain reaction while k is 1.001, or 1.1. To do that, you use either a gun or high explosives, to quickly reassemble a subcritical configuration into a hypercritical one, before a stray neutron has a chance to start a barely supercritical reaction (also known as a "fizzle").

With a reactor, there's no way to get an atomic bomb effect, because you go through the point of k=1.001. When you get there, there are plenty of neutrons around that this results in a "fizzle". Heat is produced and the core dilates and the ratio becomes low again. This results in "prompt excursions", each one lasting less than one second, each one some sort of a "bomb fizzle". These excursions are annoying, but they are not Hiroshima.

> I will not be surprised if their approach ends up being unlicensable in the US because safety cannot be assured by analysis under accident conditions.

What you are saying is a Catch-22. NRC can't approve a new design because it doesn't know how it performs under accident conditions, but then you can't know how something performs under accident conditions if you don't build it, and you don't build it if NRC does not approve it.

The fact is, Russia has been operating the BN-800 sodium-cooled fast reactor for 10 years [1], and the smaller BN-600 for 45 years. So sodium-cooled fast reactors can work, and Russia likes them enough that it plans to build a bigger version, BN-1200. They don't explode. You could say "not yet", and sure thing, this means that the NRC needs to work hard to cover all the bases. But this is not impossible.

[1] https://en.wikipedia.org/wiki/BN-800_reactor

You are assuming that as the core starts to expand, it doesn't expand in a way that increases k (for example, suppose liquid sodium is violently expelled from collapsing coolant channels; this could increase k). Can this be assured? Because if not, the yield could become very large. Granted, this doesn't seem likely, but can you assure it's impossible, in any possible accident, and any possible geometry of fuel and voids?

> What you are saying is a Catch-22. NRC can't approve a new design because it doesn't know how it performs under accident conditions, but then you can't know how something performs under accident conditions if you don't build it, and you don't build it if NRC does not approve it.

Even if you build it, and even if there is an accident, you don't know how it behaves under accident conditions. You just know how it behaved in that particular accident.

It might be easier for developers if we could build thousands a fast reactors and let some of them explode to get some statistics on what works, but that's not how the NRC is arranged, and even that would not ensure all the catastrophic risk had been retired.

Here's some of the software Terrapower wrote in order to model the neutron economy of the Natrium reactor [1]. Such software has been written ever since the early days of the atomic era.

Nuclear engineers know how k behaves, they don't need to make assumptions. Here's a 101 on reactivity feedback effects by the NRC [2].

The US has built all sorts of nuclear reactors, sodium-cooled included. Even one that was sodium cooled and used molten (hot) plutonium as fuel [3].

The experience with prompt criticality excursions is (for the better or worse) extensive [4].

The NRC absolutely has the ability (and duty) to analyze all possible accident conditions for this reactor. This is the page it maintains for Natrium [5]. They have already received and reviewed thousands of documents for it. We would not be at the point that they allow the building to start if they think there's a slight chance the reactor would be "unlicenseable".

[1] https://terrapower.github.io/armi/

[2] https://www.nrc.gov/docs/ml1214/ml12142a130.pdf

[3] https://www.osti.gov/biblio/4206527

[4] https://en.wikipedia.org/wiki/Prompt_criticality#List_of_acc...

[5] https://www.nrc.gov/reactors/new-reactors/advanced/who-were-...

The NRC has no duty whatsoever to analyze accident conditions. The NRC requires the licensee to do the analysis; the NRC just has to evaluate whether they did it properly and sufficiently.

I do wonder how you think all possible accident conditions could be evaluated. All possible geometries of melted fuel? The possibilities are vast, indeed exponentially vast.

Edward Teller in his 1967 comment expressed skepticism that the consequences of a serious accident in a fast breeder could be evaluated (yes, I know Natrium is not quite this configuration, uranium having a considerably higher bare core critical mass, but it's similar):

"For the fast breeder to work in its steady-state breeding condition you probably need something like half a ton of plutonium. In order that it should work economically in a sufficiently big power-producing unit, it probably needs quite a bit more than one ton of plutonium. I do not like the hazard involved. I suggested that nuclear reactors are a blessing because they are clean. They are clean as long as they function as planned, but if they malfunction in a massive manner, which can happen in principle, they can release enough fission products to kill a tremendous number of people.

...But, if you put together two tons of plutonium in a breeder, one tenth of one percent of this material could become critical.

I have listened to hundreds of analyses of what course a nuclear accident can take. Although I believe it is possible to analyze the immediate consequences of an accident, I do not believe it is possible to analyze and foresee the secondary consequences. In an accident involving a plutonium reactor, a couple of tons of plutonium can melt. I don't think anybody can foresee where one or two or five percent of this plutonium will find itself and how it will get mixed with some other material. A small fraction of the original charge can become a great hazard."

So your line of argument goes something like this: "->Do you concede that there are disaster scenarios that you didn't think of? ->Yes, I'm not God to know everything, I concede that. ->Well, QED, a fast reactor is unlicenseable." Case closed, right?

But the problem with this argument is that it can be applied to anything. In particular, if you really follow this argument, we should shut down all the current operating reactors, because, you know, there might be disaster scenarios that we didn't consider.

I hear you, I hear you. This argument only applies specifically to fast reactors, because this is what Dr. Teller said in 1967. But why did Teller say that?

In 1961 a nuclear reactor that the US Army commissioned with the stated intent to be virtually idiot proof, went boom. It became prompt critical. What happened was one of the most (if not the most) life-beats-fiction events in the history of nuclear reactors. It appears one of the operators decided to commit suicide by pulling a control rod as fast as he could [1]. Of course, since the guy died in the following half a second, we'll never know his true motives.

Another thing is that Teller was profoundly anti-communist. In 1957 another mind-bending accident happened in the Soviet Union, [2]. It had to do with plutonium, it went something like this: the Soviet Union was producing lots of plutonium to build bombs. Plutonium is extracted by chemical separation from spent uranium. After the separation, various steps are taken to increase the concentration, and eventually to purify the plutonium metal, but some chemical tailings that contained plutonium in low concentration remained. They were dumped in large underground tank. The solution contained ammonium nitrate, but it was mostly water. Water is a good moderator, so there's a good chance the solution would from time to time undergo prompt criticality excursions. Nothing major, no explosion happened. Yet. But because of the periodic heating, the water evaporated, and in time the ammonium nitrate became more concentrated, and it became a huge amount of (chemical) explosive. At some point this ammonium nitrate decided to go boom, like it happened in Beirut in 2020. The explosion was totally non-nuclear, but it put in the atmosphere a huge dirt cloud full of plutonium and other radioactive nasty stuff. The Soviet Union kept this a secret for many, many decades. But some people in the West learned some tidbits about this, and Teller, obsessed as he was with the Soviet Union, might have been one of the guys in the know. Plus, Teller (the father of the hydrogen bomb) had access to the highest level of top secret information in the US in regards to nuclear things. If anyone in the US knew anything about the Kyshtym, then Teller knew that too.

But why did Teller say that about fast reactors and not thermal reactors? Because that was the context. He was asked to testify if the US should build a sodium-cooled fast reactor. And he said it's a bad idea. If you know anything about Teller, is he said everything with a supreme confidence. Everything was very black and white for him, and he always knew with utmost certainty what was black and what was white. He thought for example that we should use hydrogen bombs to dig canals. Or that we should put anti-ballistic-missile thingies orbit that would zap the incoming missiles with a beam of X-rays produced by a nuclear bomb explosion [3]. This is not a joke.

Ok. Let's leave Dr. Teller aside for now. Are sodium-cooled reactors too risky to build. Well, patently not, considering that Russia is operating 2 right now and plans to build more. But let's say you are not happy with this. Will there be a point where you would concede that the experience the world has with building and operating such reactors is sufficient for the NRC to start granting such licenses? If not, what makes fast reactors any different that pressurized water reactors?

And by the way, the Natrium reactor is not designed to burn plutonium. It is not designed to be a breeder reactor. And the issue about prompt neutrons is actually less of a problem for such a reactor: U-238 has the highest ratio of delayed neutrons to prompt neutrons of any fissionable isotope and for fast reactors a larger proportion of fission events involve U-238 compared to thermal reactors [4].

[1] https://en.wikipedia.org/wiki/SL-1

[2] https://en.wikipedia.org/wiki/Kyshtym_disaster

[3] https://en.wikipedia.org/wiki/Project_Excalibur

[4] https://www.nuclear-power.com/nuclear-power/fission/prompt-n...

reddit had a nice list of the pros and cons: https://www.reddit.com/r/NuclearPower/comments/17k0wcc/natri...

I understand the risks around sodium, but the "passive natural circulation cooling" I don't understand. Is it more feasible with this design and why?

" Pros:

    high temperature means we can use process-heat which is a much more efficient use of heat.

    fast spectrum neutrons means we can burn importantly troublesome parts of nuclear waste.

    fast spectrum is also better for breeding new fuel, significantly increasing how much energy we can extract from uranium/thorium.

    passive natural circulation cooling is much more feasible.

Cons:

    fast spectrum is a little more complicated to control.

    fast reactors require high enrichment.

    inspection of the plant is very difficult with liquid metal.

    high temperature liquid metal doesn't play nicely with metal pipes.

    sodium burns in air and is explosive with water.

    we simply do not have nearly as much experience with sodium as we do water and that really cannot be understated.
"

I suppose that "passive natural circulation cooling" means that plain convection of the coolant(s) is sufficient to cool the reactor, without involving pumps which could fail. Convection can't fail as long as there is coolant and no significant obstacles.

Russia has been operating a sodium-cooled fast reactor for 45 years.

Based on your comment it sounds unreasonable to select this design. It must have some reason to exist?

Fast reactors do have some attractive features. They have better neutron economy and work better with plutonium. They can achieve breeding ratios comfortably above 1 with the U-Pu system. They produce less actinide waste since the chance of neutron capture not causing fission is lower, and can more effectively destroy actinide waste. Sodium-cooled fast reactors will operate at higher temperature than LWRs, enabling the salt thermal storage scheme they propose to use.

In large reactors, these features have not been enough to compensate for the disadvantages and sodium-cooled reactors have not been successful, coming in more expensive than light water reactors for a given power output. France, which had been developing fast reactors, has recently mothballed the effort with no plan to restart before 2050.

Russia has two of them, one 800MW and one 600MW, and plan to build a 1.2GW version.

Even the Russians admit their LWRs are cheaper.

The main advantages of using a fast reactor are less nuclear waste and more energy for a given fuel input.

Even burning some of the "nuclear waste" which is just nuclear fuel than needs refining.

It can make more fuel than it uses. It needs enriched uranium at startup but then can convert regular uranium and/or thorium to usable fuel.

> If the fuel itself rearranges to become more compact, say by melting and flowing, the reactivity could increase.

Wouldn't you just design the shape of the reactor so that if it got too hot for any reason, the shape it would melt into would result in a less compact geometry that would slow down rather than speed up the reaction?

How do you do that in a way that's amenable to conclusive analytic demonstration? For example, how do you prevent melted fuel from flowing into the cooling channels that go through the core?

About the only approach I'd be comfortable with would be dissolving the fuel in molten salt (probably chloride salt). This is not Natrium's approach.

Melting of fuel is not a theoretical problem -- it has actually happened at two fast reactors in the US (EBR-1 and Fermi-1, the latter the reactor in the hyperbolically titled book "We Almost Lost Detroit"). No explosions occurred, but it's very troubling the fuel melted at all. The NRC will surely insist on analysis of the consequences of partial fuel melting accidents.

To begin with you might start with a geometry which by design is already close to maximally compact, so that a geometry change would tend to go in one direction.

> For example, how do you prevent melted fuel from flowing into the cooling channels that go through the core?

Expect that to happen and use a geometry that doesn't cause the reaction to increase in speed if it does, e.g. because fuel flowing into the cooling channels would make the fuel less rather than more compact.

Slightly hesitant to jump in since pfdietz definitely knows more about this than I do... but...

Cooling typically means things like maximizing surface area, minimizing the thickness of the object being cool, etc.

Maximum neutron density presumably happens in a sphere, which is coincidentally the shape that minimizes surface area.

The whole point of a nuclear reactor is that it heats up, and you can convert that heat into useful fuel. Presumably that means you need to carry quite a lot of heat away per volume. Presumably that means putting the fuel into a spherical shape really doesn't work that well.

To cool something, you need a material which is a decent conductor of heat. The reaction materials are mostly uranium, plutonium and thorium, which are metals. They conduct heat pretty well all on their own.

Metal fuels also melt at considerably lower temperature than oxide fuels.

Uranium oxide melts at 2865 C; uranium metal at 1132 C, plutonium metal at just 639 C. In contact with iron, plutonium forms a eutectic with a melting point of just 410 C, below the melting point of zinc. There was a crazy reactor at Los Alamos, LAMPRE, that used molten eutectic Pu-Fe in tantalum tubes as the fuel.

1132 C shouldn't be an infeasibly low temperature when the reactor is only expected to heat the coolant to 300-350 C. Meanwhile thorium metal is 1750 C.

Or use uranium carbide. High melting point with still decent thermal conductivity.

Why would fuel melting be possible? The way I'd show it is by having the increased Doppler broadening and thermal conductivity and lots of headroom make that sort of accident impossible.

Presumably, those operating EBR-1 and Fermi-1 (and SRE in California, which also melted down) didn't think those would melt either. The issue is showing by analysis such an occurrence is not possible. It's not incumbent on anyone to show it is possible, it's incumbent on the prospective licensee to show it isn't.