Open-Sourcing Our Reactor Design, and the Future of Transatomic | Hacker News Reader

Open-Sourcing Our Reactor Design, and the Future of Transatomic

transatomicpower.com

105 pointsareoform26 comments

I have no idea if molten salt reactors are as awesome as this website represents them to be, but I thought the following point was interesting:

"light water reactors can use only about 4% of their available energy."

They're comparing with 1960s reactors designs, so maybe more modern water reactors are more efficient. But, it's still amazing to me that there's that much more energy that we could potentially harness from the fuel.

It's because light water reactors use thermal (i.e. slowed down) neutrons, which can only fission U235, and that's 0.7% of natural uranium. We enrich the fuel to at least 2% U235 and that's mainly what the reactor can use.

The reactor also gets about a third of its energy by fissioning plutonium, which is formed when U238 absorbs a neutron.

Without reprocessing, a light water reactor can't even fission all the U235, because some of the waste products of fission reactions are strong neutron absorbers. (If you're in France you can reprocess the waste to remove the fission products.)

Molten salt reactors remove most of the fission products continuously. Some MSRs, like Terrestrial Energy's, are still thermal reactors using uranium, so they're about as efficient as France.

Some MSR designs use fast neutrons, which are able to fission more plutonium along with other transuranics and U238. Others use thorium fuel, which can be completely fissioned by thermal neutrons. (First thorium absorbs a neutron and becomes U233, then that fissions. Heavier non-fissile isotopes aren't created in the first place.)

Currently the fuel is very inexpensive - a tiny fraction of the total cost of the electricity produced. Most of the cost is the initial capital, and most of the ongoing cost is the (highly specialized) labor. A new reactor design that saves fuel is not really what the industry needs.

According to Wikipedia current uranium reserves will last for 135 years. It would be 13 years if we increase energy production in nuclear plants 10 times to replace coal and gas.

So even if fuel is inexpensive now, it can become expensive quite quickly.

https://en.m.wikipedia.org/wiki/Peak_uranium

I believe Peak Oil was supposed to happen a dozen times now; we keep finding more when it becomes worthwhile to do additional exploration, and extraction technology keeps improving. Maybe uranium is different, but I would believe it when I see it.

As per wikipedia: "Peak oil is the theorized point in time when the maximum rate of extraction of petroleum is reached". You can still discover new sources of petroleum after "Peak Oil", but the idea is that you will never find enough to increase the maximum extraction rate.

What we see as consumers is conflated by the fact that there are self-imposed limits on extraction by OPEC designed to raise prices. Due to various political situations, those limits have been slowly removed. When you see predictions of "Peak Oil", often they are based on existing production and not maximum production (and hence are nonsense).

I'm not aware of any literature that attempts to speculate on current maximum production capacity. Probably it exists, but I certainly can't find it. Whether we have hit "Peak Oil" already, or whether it will come in the future, I don't know. A lot of that kind of stuff is politically very sensitive and oil companies/OPEC are understandably reluctant to be straight forward with the data.

Finite resources eventually run out. When you don't know how much you have, it's tempting to assume that because you are finding more, you will continue to find more. There is no real guarantee of that. It's a risky strategy.

EROI has been driven down with every single year. One day it will take more energy to extract it than the energy we receive from it but we won't run out.

As with all of these 'Peak' this simply wrong. I don't know how many more times we have to go threw this cycle before we can stop with these peaks.

What about the ongoing cost of waste and eventual cost of decommissioning? Are those affected by quantity of fuel used (or, more specifically, by Transatomic's design)?

When a reactor is built in the US, the operator will start paying into an account for waste disposal. This is included in the cost of energy production. Since high level waste quantities are pretty low for nuclear reactors (About 27 tons a year)[1] disposal of fuel isn't the largest issue.

Decommissioning on the other hand is remarkably expensive because of secondary nuclear waste like contaminated reactor vessels and concrete make up 99% of the nuclear waste.

[1] http://www.world-nuclear.org/information-library/nuclear-fue...

Site isn't working now, but their GitHub is here [1]. My understanding on this is that Transatomic started as two young MIT students/grads who did some calcs on a molten salt reactor, found that they could run on high-level nuclear waste in a thermal-neutron spectrum, got really excited, and started a company. They did TED talks and everything [2] and really became this awesome story about advanced nuclear, nuclear startups, private investment in nuclear doing exciting new things, and so on. They were heavily featured in a documentary about advanced nuclear [3] that you can get on Amazon, where we all got a look in their doors.

Then the trouble started. They (smartly!) asked MIT to do a design review and a few questions came up on their key claims. Turns out, there were some calculation or assumption mistakes that showed that it couldn't just run directly on nuclear waste (you traditionally need fast neutrons and/or aggressive reprocessing to do this, even in molten fuel) [4]. They adjusted their copy and moved forward with an adjusted design and new calcs. Eventually they decided to close up shop and open source their design effort as a soft closure.

It's highly regular for advanced nuclear efforts to adjust their design and strategy as they move forward. In fact, I'd go so far to say it happens in 100% of efforts. The only thing Transatomic did that might be a lesson learned is that they got really hyped up early on in their effort, before they shook it down with engineering reviews and whatnot.

As far as I can tell, no one is curating progress or roadmaps in the open source effort. I LOVE the idea of open-source reactor design. It gets around huge institutional barriers clogging up opportunities for collaboration and alignment in the nuclear space. Nuclear is traditionally very secretive, and all these startups promising the world in 2024 without collaboration is basically a doomed effort. The industry must team up, find overlapping tech development efforts and facilities, and work together to build a market. Once they build the market, they can worry about competing in it. Open source can also transcend traditional funding cycles, like when national labs work on a reactor for 2-4 years, make tons of exciting progress, and then drop it. If it's open, universities, international institutions, companies, etc. can all coordinate on it.

Doing open source reactor design would require a lean and highly-dedicated non-profit or foundation or something with a handful of nuclear design pros in it to do the coordination. These projects don't run themselves. It would also need lots of legal resources to navigate export control space. Everything Adm. Rickover said in 1953 about academic vs. practical reactors is incredibly accurate to this day [5], and using open-source philosophy to face these facts brutally may be essential for progress in nuclear.

If anyone has ideas on how to fund that kind of thing, let me know.

[1] https://github.com/transatomic/reactor/

[2] https://www.youtube.com/watch?v=AlRhcHNJed0

[3] https://www.newfiremovie.com/

[4] https://www.technologyreview.com/s/603731/nuclear-energy-sta...

[5] https://en.wikiquote.org/wiki/Hyman_G._Rickover#Paper_Reacto...

Your proposal would be a great fit for the Effective Altruism community. Climate change is an existential risk and if this organization has a 0.1% chance of working out. The utility function for that ROI versus failure would still work out I’m the favor of funding this idea.

But the problem will always be. How do we go from designs to an actual, legitimate organization that can play with nuclear materials in Dyson’s little red schoolhouse? https://blogs.scientificamerican.com/the-curious-wavefunctio...

Funding will always be contingent, in the minds of most readers, the trustworthiness/robustness of a rough path that goes from a brain trust to a prototype reactor. If we can solve that, we could probably get significant funding lined up by 2024

> and all these startups promising the world in 2024 without collaboration is basically a doomed effort.

Indeed, I always thought that for any company the idea is 5% of the success, everything else is the execution. If a company can use a perfectly working design and just run it (as in PowerPlant as a Service) then that's where the value is, both for them in financial terms and for the society in advancement terms. Commoditize the design of the core reactor and innovate on its different uses and the whole supply chain.

Hard to see how open source nuclear reactors would work. Open source works well for software because getting to the 80% solution is fun, and most people only need the 80% solution. An 80% nuclear reactor isn’t useful to anyone, and even getting to 80 percent is not that fun. Getting to 20% is a blast, but then it is time to do all of the tedious work.

Edit: I should have your Rickover link before posting. It captures my point in more detail, and with more pith.

How? China or India uses the opensource work and takes the design further. Then we license it from them like we do with existing designs.

It's not like you clone the plans off a Github repo and build an uncertified nuclear reactor in your basement, order nuclear fuel from Amazon and power up your house. Instead it'a bit like what happened to Android: it started as an OSS product but then Google added their services on top which make it actually useful.

Could the essential aspects of a reactor design be patented so that value can be captured despite innovation being created in the open?

Possibly, as long as participants can still be motivated to working within the open effort.

Frankly, effective reduction to practice is required 1000x more than traditional "innovation" in nuclear. Every bright mind in the 1950s was dreaming up whacky and incredible reactor designs, and everything I've ever thought of, from spinning vortex/whirlpool fuel to giant uranium gaseous pistons already has 50-page publications from back then sitting on OSTI. What's needed is to figure out how to get nuclear technology into a practical rapid iteration process where we can shake down good ideas vs. bad ideas, and put some rigor behind cost and schedule estimates.

Otherwise it's just a game of who can get investors/government most excited about their particular reactor concept, which, without actually understanding, adjusting, and improving upon what's happened in the past, will struggle mightily.

> What's needed is to figure out how to get nuclear technology into a practical rapid iteration process where we can shake down good ideas vs. bad ideas, and put some rigor behind cost and schedule estimates.

Sadly this is currently just not possible. You can not get regulatory approval for a prototype beyond a tiny research reactor (that is unsuited to test real production reactors). Thus nuclear companies are forced to directly produce to a production design, and to production regulation and not just that, but also prove operational regulation and so on.

This comes with incredible cost and it is very much possible that a nuclear reactor company builds one design and works with it for 20-30 years because doing any substantial redesign would require a completely new regulatory process.

Thankfully regulators in Canada and the US have been waking up to the fact that regulation as currently designed have totally crushed innovation and are trying to reform. Canada is leading in this and British, US and Canadian companies all have started to go threw the Canadian regulatory process because of this. But 40 years of fearmongering and regulation introduced right when nuclear scare was at its peak are deeply embedded both in the regulation and the culture of these agency.

They can be motivated by acquiring stake in the success of their research and development. But the only way I can picture that working is if there is way to capture the value that is compatible with developing it in the open, hence the question.

Publishing (public disclosure) starts a 1-year clock on patentability: https://en.wikipedia.org/wiki/Public_disclosure

In an effort that incrementally builds to something patentable, it would be difficult to find the perfect moment where the previous year's efforts amount to something meaningful and patentable.

That's fine as long as you can patent some core innovation and then develop the rest later. There is protectable IP to be shared between collaborators.

What's the best way to get in touch? I'm at username [-] google's email.

Thanks, I'll hit you up.

I helped run a TEDx event (I did event branding and ad designs) where Leslie Dewan gave a talk about this! Unfortunately it looks like the video from her talk at our event isn't up yet, but I will ask the team if they could have it posted.

Its cool that they open sourced their design. I thought it was very likely that they would fail, even more so then most nuclear startups. They made design choices that were questionable and their claim to run directly of nuclear waste seemed very surprising. They also seem to be way to optimistic about the regulatory and business aspects.

For those interested, Terrestrial Energy and Moltex Energy are somewhat more serious attempts to build molten salt reactors, both are quite different form Transatomic.

Terrestrial Energy and Moltex both are in the regulatory process in Canada (the only place a new reactor can realistically be approved) and Terrestrial Energy is the furthest along of any GenIV reactor in terms of commercial regulation in the Western world.

This talk by the main designer of Moltex is a pretty nice summation why they think it would work:

https://www.youtube.com/watch?v=TvXcoSdXYlk

Got excited about an open source reactor pattern

Reply on news.ycombinator.com