The comments on my submission to the Astral Codex Ten book review contest were highly productive. These are some other interesting comments that did not fit into the previous two categories.
Originally Written: June 2022.
Prerequisites: Book Review of THE FUTURE OF FUSION ENERGY by Jason Parisi and Justin Ball (2019).
Confidence Level: Varies, but should be pretty high throughout.
Other Fusioneers
Jason Parisi (one of the authors of The Future of Fusion Energy):
Great article, thanks for spending the time reading the book and writing this review.
You mentioned that stellarators are your ‘favourite.’ I think that a huge advance in the field in the past couple of years has been on stellarator optimization. Here’s some preliminary reading:
https://terpconnect.umd.edu/~mattland/projects/6_optimization/
https://simsopt.readthedocs.io/en/latest/
https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.128.035001
Next frontier in stellarators is optimizing for turbulent transport, which I suspect will be much trickier.
Another fusion physicist here (just a novice, though.)
What I always was afraid of regarding ITER was an Apollo Program-style Phyrric victory. “Hooray! We’ve achieved the nigh-impossible! Go us! Obviously not cost-effective, though, so let’s basically shelve the whole field.” Anything that’s cost-effective is going to have to be massively simpler and cheaper than ITER – being just a reactor rather than a super-diagnosed experiment would help, of course – and I do think that the key innovations that come to the rescue probably are indeed going to come from outside the field of plasma physics proper.
Take heavier-than-air flight, for example. Why was it first invented when it was, around the turn of the 20th century? It’s not because the Wright Brothers were far beyond everybody else (their advances in stability and control notwithstanding); you’ll note that there were rival claimants to the title, some of whom did pick up support from major institutions. But they were all within a few years of the Wright Brothers, anyway – why’s that? The reason is that the technological landscape had simply advanced far enough at that point, to the point of engines with sufficient power-to-weight ratios becoming available.
Can you imagine trying to make do without? Achieving heavier-than-air flight for humans by pushing the field of aerodynamics to the limit to make a perfectly-optimized jumbo-jet-sized airframe that is, in the end, capable of briefly lifting a baby into the air because all the rest of its power needs to go into lifting its own weight? There’s a Phyrric victory for you.
High-performing superconductors may be, I feel, the equivalent of high-performing engines here. Fusion power is (if I’m not terribly mistaken) proportional to the strength of the magnetic field to the fourth power. Just like you could make a brick fly with a good enough engine, you may well achieve the plasma-physically “impossible” just by spamming more magnetic field strength at the problem.
Which is something I’m hoping for, myself: my interest in plasma-based nuclear power is because it’s the sort of power source needed for high-thrust, high-specific-impulse (and thus high-squared power!) space propulsion. Tokamaks aren’t the most rockety concept out there, I’d have to admit. So I’m looking forward to what we’re able to develop – a brighter, positive-sum vision of the future is something I think is worth working for.
Welcome to the field !
The comparison between high-performing superconductors and high-performing engines is interesting. Some scientific and technological advances are done by lots of people at close to the same time, while some are isolated events. I don’t know of any general way to tell which something will be.
I would like to know all the supporting technologies that’ve enabled the recent fusion boom. I saw superconductors in the article and I’ve also heard new magnets.
C.f. the old theory from the SSC comment section about how all technological advancement ultimately comes from materials science
The high temperature superconductors are the (electro)magnets. They’re the big game changer. ITER has a list of supporting systems if you want to see more relevant technology: https://www.iter.org/mach/supporting
The 2.5 biggest supporting techs are better superconducting magnets and better computer simulations.
The magnets are an enormous deal because scaling up a 90s-style tokamak design to where we’d expect to get Q > 5 requires some combination of a bigger reactor and stronger magnets. The scaling effects are much more favorable to stronger magnets than to a bigger reactor: ITER and SPARC are expected to have similar Q factors, with ITER being about 3.3x the linear dimensions (35x volume) of SPARC, while SPARC’s magnets are about twice as strong. The magnets need to be superconducting because otherwise, even very small losses of power to resistance in the magnet conductors would be a prohibitive tax on the energy balance of the reactor: you can make a research reactor with cryogenically-cooled copper electromagnets, but not a viable commercial reactor.
Computer models have advanced in two significant respects since the 90s. One is that they’re informed by a quarter century or so worth of experimental data and theoretical analysis. The other is that a modern computer is orders of magnitude more powerful than a computer of equivalent cost was in the 90s. With a good enough simulation model running on good enough hardware, you can do design studies for prototype reactors much faster with a much higher degree of confidence than you could in the 90s, allowing designs to iterate faster and requiring fewer iterations of actually having to build full-scale hardware to validate your theories.
I’ve been working on fusion for the last year under an Emergent Ventures grant, and working alongside Schmidt Futures and Adam Marblestone to identify philanthropic opportunities in the space. That doesn’t by any means make me an expert, but it’s enough that I’m willing to share my opinion.
I think every prediction on this list is about an order of magnitude too optimistic, with the exception of CFS and ITER. I would also clarify that while Q>5 in steady state would be a momentous achievement, it’s still puts us a far ways off from fusion energy that is economically or climate relevant. For example, I am fairly optimistic that ITER will “work”, but it doesn’t actually provide us with a path to commercial fusion energy.
Having said that, I think fusion is absolutely worth pursuing, and that in fact, we’re severely underinvesting at the margin even with my revised predictions. Concretely, for the industry as a whole, I would give us a 20% chance of having fusion energy on the grid by 2035, and a 35% chance by 2040.
I’m happy to chat more with anyone here who’s working on fusion, or just anyone who’s interested:
Thank you for coming and sharing a different perspective !
1. After Commonwealth and ITER, who do you think are the most likely candidates to get Q > 5 before 2040? Do you have more hope for other startups or for government run DEMOs?
2. You think that it is less likely for fusion energy to be on the grid by 2035 or 2040 than I do. Is that because you think that my time frame is too optimistic or because you think that Commonwealth will be unable to solve the problems?
3. I know very little about the philanthropic side to fusion funding. My impression was that most of the money is either government or venture capital. Which of the players have been pursuing philanthropy?
1. I like Zap. There are also a couple pre-launch startups I’m funding and fairly optimistic about. I’ve talked to people smarter than me who like ICF.
2. I would describe myself as super optimistic about CFS, and would have been happy to invest in their latest round. But all startups naturally carry risk, and I just don’t feel like we have that many real shots on goal. For what it’s worth, I think the distribution is kind of bimodal, in the sense that if we don’t get fusion in the next few decades, we might lose the window of opportunity.
3. Not that many! Simons is one of the major players
Malcolm Handley has done a bit, as has Schmidt Futures.
Bill Gates is invested in fusion through Breakthrough Energy, and Jeff Bezos is invested, but I’m not sure if either has done much philanthropic giving in the space.
The issue is that energy is fundamentally a market, and fusion is a highly path dependent technology such that subsidizing reactors that aren’t going to be economically relevant on their own footing is not a great option.
I am personally excited about philanthropic funding for fusion “tools”. Things like material testing/development and tritium systems which benefit a variety of companies and reactor approaches.
1. [Discussed here.]
2. I could see the case for it to be bimodal. If we don’t figure things out in the next few decades, it would mean that the problem is much harder than we expect. I also don’t think that there will be nearly as much support if this generation of experiments fails.
3. I was aware of Simons, but hadn’t really categorized it as philanthropic funding for fusion. I had not heard of Strong Atomics. Thank you for the resource !
Philanthropy for fusion tools makes a lot of sense. Ideally, you’d try to remove a challenge that almost everyone will face, and make it available to everyone. Materials and tritium breeding are both good options. Diagnostics and remote maintenance would also fit in this category.
I wonder if SPARC will rent out time for other groups to test their materials, etc, with a 14 MeV neutron source.
Predictions & Markets
If you assume the success of the different projects is independent, your expectation of fusion by 2035 should be >98% and >99.75% by 2040 . So 80% and 90% seems off. Do you think the projects’ success is that much correlated?
I don’t think that you should be >99.75% confident about anything humanity will do by 2040 because X-risk by 2040 is probably above 0.25%.
Some problems for getting fusion would only affect one player, but others would affect multiple players at once – although probably not everyone. If the venture capital dries up, then that affects all the startups but not the government players. If ITER can’t get disruptions under control, then that affects all the tokamaks, but not the stellarators or other designs (although SPARC would probably have figured this out before ITER). We should probably count the 11 partially independent projects in the table as about 3 fully independent projects.
I also threw in a bit of a “maybe this person is crazy” factor, even for myself. The predictions for the particular players are all inside view. The headline predictions attempt to be outside view.
Did you take all of that into account to estimate your 80% and 90% or were they more gut estimates?
The estimates for particular experiments were gut estimates. But I did take all of that into account for the headline estimates. I thought it was likely that someone would challenge me to a bet on the headline estimates and so made them as defensible as I could.
Since this book review makes some explicit forecasts and this community is in general interested in forecasting: there are a bunch of questions on Metaculus pertaining to fusion. One of them pertains to “fusion ignition” (https://www.metaculus.com/questions/3727/when-will-a-fusion-reactor-reach-ignition/) . Unfortunately, there appears to be no agreement as to what fusion ignition means exactly (cf. my comment in the discussion there). Hence IMO the resolution criteria of the question are not ideal. I have tried my best to bring up some of the issues but I am a total lay person in this area.
It would be great if someone with actual expertise could weigh in there and also on the Wikipedia article (https://en.wikipedia.org/wiki/Fusion_ignition) on fusion ignition so that we can get better forecasts and knowledge dissemination on this topic.
The Metaculus question is poorly written. It gives two different definitions of ‘ignition’:
“If Q increases past this point, increasing self-heating eventually removes the need for external heating. At this point the reaction becomes self-sustaining, a condition called ignition. Ignition corresponds to infinite Q, and is generally regarded as highly desirable for practical reactor designs.”
“This question will resolve on the date when a nuclear fusion reactor has sustained a reaction which produces more energy and heat than the external energy delivered to the system”
In the first definition, external heating is unnecessary and can be turned off. So external heat = 0 and Q = infinity. In the second definition, external heating is less than energy produced. So Q = 1. The first one is ‘ignition’, as defined by the magnetic confinement community. The second one is ‘breakeven’. These are unlikely to occur at the same time.
The source of the problem is that inertial confinement fusion and magnetic confinement fusion use ‘ignition’ differently. The magnetic confinement community first started using the term for fusion. The plasma is assumed to be in steady state and is able to sustain itself with no external heating. Inertial confinement is never in steady state, so this definition doesn’t make sense. Instead, their definition is that the amount of energy currently being produced by fusion in the plasma is greater than the amount of energy which is currently leaving the plasma. This means that the temperature of the plasma is increasing as the result of fusion reactions. This is not steady state, and the fusion will quickly burn through the fuel pellet.
To make this more clear, let’s make a comparison to steam engines vs internal combustion engines. Steam engines have a fire going in steady state. You need to start the fire by adding heat. It only makes sense to say that a steam engine has reached ignition once it doesn’t require any more external heat. It is then self-sustaining. Internal combustion engines have a fire only in small pulses. Each pulse requires external heat from the spark plugs. It is never self-sustaining. Instead, what happens is that combustion rapidly increases the temperature of the gas and burns through all of the fuel.
I don’t think that ignition is the right goal for this question. Ignition probably isn’t desirable for a magnetically confined fusion power plant. Even if we could reach ignition, we would probably run the plant at Q~30 because it gives us more control over the plasma.
It’s better to avoid the term ‘ignition’ if you want to compare multiple very different approaches to fusion. Instead, define what you’re talking about precisely in terms of Q. I chose Q>5 and steady state or 1 shot/sec as my criterion.
(Feel free to use this on Metaculus or Wikipedia.)
I have a manifold prediction market on the likelihood 2% of the grid is fusion by 2050. It’s pretty low (33%). https://manifold.markets/J/will-fusion-provide-2-of-us-electri I’ll write some criticisms of this post in the morning.
Current state of the art is q=1 with exorbitantly expensive DT fuel. Economic breakeven would require q>4 using DD fuel instead, and DD requires ~5x higher temperatures. Radiative losses are usually proportional to the fourth power of temperature, and there are other additional ways for electrons to lose energy at extreme temperatures. A design that yields Q=1 at temp T would probably be like Q=0.001 at temp 5T. So they need to get three and a half orders of magnitude better.
Interesting. Thank you for making the market !
By calling DT fuel exorbitantly expensive, I’m guessing that means that you don’t think that the tritium breeding blanket will work. Tritium is very expensive, but lithium is not. You breed it using the neutrons from the fusion reaction itself. Each fusion reaction consumes one tritium and produces one neutron. Each breeding reaction consumes one neutron and produces one tritium. You can compensate for any losses with a neutron multiplier like beryllium.
Misc
If by “get fusion” we take the definition in footnote 1 (somebody does an experiment that gets Q > 5), then yes, we will probably get fusion on something like the timeline the author describes. However:
1. Q = 5 is far from commercially viable. Q represents an energy gain in terms of the energy going into the plasma. It does not account for all the other energy needed to run the system. The most optimistic number I have seen is that Q > 20 is commercial fusion. And even that is optimistic because of point #2.
2. Neutronic fusion (basically everyone doing D-T fusion) relies on neutrons crashing into stuff and generating heat, which then boils water, which then drives a steam turbine. So one input into the electricity economics for fusion is the cost of the generating equipment, the steam turbine itself. If we look at the other modes of electricity production that also use steam turbines, and compare the cost of their steam, we will see that fusion is clearly worse than everyone else. Coal: we mine this thing from the ground and set it on fire and it gives us heat to make steam. Fission: we mine these magic rocks from the ground and hold them together and it gives us heat to make steam. Geothermal: we drill holes in the ground until we reach hot temperatures, circulate water, and it yields steam. Fusion: we build the most complex and finnicky machine ever known to man (read: very expensive) and then it produces heat which we can use to make steam. Neutronic fusion is an expensive way to make steam, so the economics are unlikely to work if it is actually competing on a level playing field with other ways of making steam.
3. The insanely complex and finnicky machine neutronic fusion uses to make heat will need to undergo maintenance, and perhaps more often than other machines, because it is constantly being bombarded with radioactive neutrons. When it needs to undergo maintenance, it will be radioactive because of all the radioactive neutrons that have bombarded it. So maintenance will be expensive.
4. The neutrons emitted by neutronic fusion can be used to breed plutonium and other fissile material, so the technology will have to be closely controlled, which is a further headwind on it ever becoming economic.
5. Fusion advocates have tried to get fusion to be perceived as “safe” because it cannot lead to a meltdown. While it can’t melt down like a fission reactor, fusion can suffer from a confinement failure (i.e., a big explosion). A confinement failure is likely to strew radioactive parts everywhere in its vicinity.
D-T fusion is indeed the closest technology to scientific breakeven (Q > 1) and maybe even to something like Q > 5 like the author suggests. But that does not mean it will ever be commercially viable. I believe that the scientists working on it know that it won’t ever be commercially viable, but they don’t want to say so because they want to continue to get funding either from governments or private investors to achieve the breakeven milestone.
Most of my criticisms of fusion are totally standard; they were noted by Lawrence Lidsky from MIT in 1983. http://orcutt.net/weblog/wp-content/uploads/2015/08/The-Trouble-With-Fusion_MIT_Tech_Review_1983.pdf
I have not looked at the details of how to make a fusion power plant commercially viable. I’m guessing that a lot of the answers are currently unknowable. This seems to be your main point, and I doubt that I will be able to respond to it satisfactorily. SPARC is about 1/10th the cost of ITER, so progress here is definitely being made.
Here are responses to your other points.
1. In terms of the triple product, $Q = \infty$ is only about twice as hard as $Q = 5$. Getting $Q > 5$ is most of the technical challenge.
2. There isn’t a level playing field. If there were, my guess is that fission would win. But it’s currently mired in a regulatory mess. I would love to see that change, but wouldn’t bet on it. Coal is cheap until you count all the people it kills. Most of what I know about geothermal comes from your blog post. Maybe it is a better option, but I would like to see both available.
3. The tritium breeding blanket should block most of the neutrons from getting to most of the components. But yes, the maintenance will be radioactive.
4. This is possible, but I don’t think it’s a big challenge, for several reasons. (a) Over half of the world’s energy is used by countries that already have nuclear weapons. I’m not as worried about e.g. China using commercial fusion reactors to make weapons if they can just get one from their stockpile. (b) Fusion reactors are much easier to inspect than fission reactors. For fission reactors, you need to keep track of how much fuel is where continually. Fusion reactors should have any uranium or plutonium there at all. Detecting if an element is present is much easier than keeping track of how much there is. (c) I don’t think that anti-proliferation inspections are a significant fraction of the cost of fission plants today. The easier inspections for fusion shouldn’t a significant fraction of the cost of them either.
5. No it’s not. Fission reactors meltdown because they contain months worth of fuel that can all burn at once. Fusion reactors would contain minutes or hours worth of fuel. There is not enough total energy in the reactor at any time to breach the building. The worst disruptions would melt part of the interior of the reactor, which would be expensive to fix. But it would not strew radioactive parts everywhere in its vicinity.
I know lots of people who are working on fusion who believe that it will become commercially viable. It’s possible that we’re wrong, but that is not the consensus of the community.
Eh. Great to see that at least the human race is at least kinda trying at the moment. But still, not a word on how we are actually supposed to extract reliable and safe energy from such a plasma, without the benefit of half a million kilometers of colder matter to shield us from its onslaught. Am I just supposed to fit some radiator tubes inbetween those super magnets? And I can somehow strike a balance between that fluid not becoming super radio active, and my reactor lining vaporizing / transmuting away, and contaminating my plasma?
I genuinely do not know if these are all trivial problems that have been solved 50 years ago; or if we are still so busy actually building plasmas, that we unironically have no idea how to get anything resembling economic utility out of them, once they are humming along. I fear the latter. I mean its easy to talk about breeder blankets and what not; but its another one to actually build one, and run a viable busines using one.
The tritium breeding blanket is what extracts the energy: https://www.iter.org/mach/TritiumBreeding
There are multiple proposed designs. I’ll describe one of them here, called the “Pebble Bed”.
Most of the energy comes out of the plasma as high energy neutrons. We want these neutrons to be absorbed by lithium in order to capture the neutrons and to make tritium. The neutrons’ energy will be deposited in the lithium.
The solid surface closest to the plasma is called the first wall. It’s job is mostly to survive and to keep high Z atoms from getting into the plasma.
Behind the first wall, there will be a layer a few feet thick filled with pebbles. These pebbles will be made of a neutron moderator, to slow down the neutrons, beryllium, which increases the number of neutrons so the blanket doesn’t need to be perfectly efficient, and lithium, which absorbs the neutrons. Almost all of the neutrons should be absorbed by this layer so they don’t damage anything behind it.
To keep the pebbles from getting too hot, a fluid flows through the pebble bed and to a heat exchanger. Helium gas has been proposed because it does not absorb neutrons and so does not become radioactive.
From the heat exchanger onwards, it’s a standard steam turbine used to generate electricity.
A few of the pebbles can be removed at a time and replaced by others, so we can extract the tritium to use as fuel.
This is just one design. ITER will be testing multiple Test Blanket Modules with different designs to see which is best at absorbing neutrons, producing tritium, and transporting the heat.
“The fusion reaction chain in the sun burns six protons (hydrogen nuclei) into helium-4, two protons, and two positrons over the course of five fusion reactions. What we do is simpler.”
Is this why human-made fusion reactors have so much higher energy density than the Sun? I used to think that fusion of course would be possible, because it’s happening constantly in the Sun. But with more research, I found that a glob of sunstuff is slightly warm and slightly glowy, and the only reason the sun is so hot and so bright is because it’s so big that the surface area to volume ratio means that that slight amount of heat and light builds up to an incredible degree. The exceptional thing about the Sun as a power plant is that it will keep on burning for billions of years.
Since any fusion power plants we build will be a lot smaller than the sun, and reasonable-sized globs of sunstuff would be pretty useless for power generation, it seems like we’d have to do something fundamentally different from what the Sun does to get useful results.
That’s most of it. We also use higher temperatures than the core of the sun.
There are important differences between what happens in a fusion reactor and in the core of the sun, but it is the most similar thing to compare to.
I’m a bit confused by the discussion of funding. The claim is made that funding for fusion research has been very low in the US, but lots of expensive experiments are described. I guess international funding is higher? I wish the article went into a bit more depth about funding levels.
International funding is higher. ITER’s total cost should be about €20B. Europe is responsible for 45%. The other countries involved, including the US, are responsible for 9% each.
The largest fusion experiments currently operating in the US were built in the 1980s or 1990s. Since the US rejoined ITER in 2003, the funding is almost entirely to maintain & use existing experiments or to build ITER. There have been no new expensive plasma experiments in the US in decades.
NIF gets its funding from nuclear stockpile management. It gets about $300M per year. Their main goals are not fusion.
The private companies get their funding from venture capital. Commonwealth has raised the most: about $2B so far, and they don’t seem to having too much trouble getting the funds to build SPARC.