Nice article. I especially liked the history of fusion research. A couple points:
(1) As you mentioned, the energy density (energy per unit mass) of fusion is four times higher than nuclear fission. However, the power density (power per unit volume) of fusion is 1-2 orders of magnitude lower than fission. In other words, fusion is great if you want to build a thermonuclear warhead, but if you want to build a power plant, fusion's low power density is a significant drawback compared to fission. To explain in more detail: LWRs have a power density on the order of 100 MW/m^3 in the core, while D-T fusion has a (maximum) power density of 0.1 MW/m^3 times the pressure (in atmospheres) squared. ITER is expected to have a pressure of 2.6 atm (0.5 MW/m^3), while tokamak power plants would likely have a pressure of about 7 atm (5 MW/m^3). So, a D-T fusion reactor is expected to have somewhere in the ballpark of 20x lower power density than a LWR.
(2) The power density (at constant pressure) of deuterium-tritium (D-T) fusion is 2 orders of magnitude higher than the second-highest power density reaction, deuterium-helium 3 (D-3He). Other fuels (such as D-D or p-11B) are roughly another order of magnitude lower still than D-3He. Anyone attempting to build a fusion reactor with "alternative" (non-D-T) fuels will either need to create extraordinarily high pressures, build an extraordinarily large reactor, or produce extraordinarily little power. In my opinion, it's almost impossible to imagine reactions with such low power densities being used for power plants.
(3) The "success" of a fusion reactor is often discussed, as it mostly is here, in terms of achieving a triple product greater than the Lawson criterion (8 atm*s for DT). However, there are two other essential things a fusion reactor must do. First, it must be able to create tritium fuel inside the reactor. Second, it must not destroy itself. These are extremely difficult engineering problems. Part of what makes them so challenging is that, while possible solutions can be analyzed theoretically before a reactor is built, those solutions can only be tested inside a power-producing reactor. Thus, progress on these difficult problems must be made in series, not in parallel. Only once you've spent a few billion dollars on a power-producing fusion reactor can you actually test whether it "works". At that point, you might not produce as much tritium fuel as you had calculated that you would. Or, your reactor could destroy itself -- and it is quite likely to do so, either via neutron damage, heat flux damage, or in a disruption. In my opinion -- and as far as I know this is a widespread view among academics unaffiliated with fusion startups -- it is certainly possible that one of the fusion startups will "successfully build a" [power-producing] machine, but it is very unlikely that any of them will successfully build a fusion power plant. The opposing view -- and to the best of my knowledge, this is how Commonwealth Fusion Systems thinks about these challenges -- is that the public excitement created by the achievement of high-Q magnetic confinement fusion will lead to so much investment in fusion that we'll be able to afford to solve those difficult engineering challenges later on. Kicking the can down the road, if you will. In any case, I think a discussion of the engineering challenges associated with fusion power plants is essential.
(4) Using the development of the atomic bomb as an analogy: in 1940, the Frisch-Peierls memorandum performed some simple calculations suggesting that an atomic bomb was theoretically possible, and would require a few kilograms of U-235. We now have enough knowledge, at least for tokamaks (the reactor concept with by far the best performance), to calculate whether a fusion reactor can scale to a power plant. This is done in "Designing a tokamak fusion reactor—How does plasma physics fit in?" (J.P. Friedberg et al., 2015), which writes "Unfortunately, a tokamak reactor designed on the basis of standard engineering and nuclear physics constraints does not scale to a reactor." In my opinion, this is a fundamental difference between the development of nuclear fusion and nuclear fission or the atomic bomb: with fusion, our best calculations suggest that it will not work. Or more precisely, tokamaks will not work barring significant improvements in certain engineering constraints.
I think you've done a nice job of laying out the bull and bear cases for fusion. However, I think the bear case for fusion could be made much stronger with some relatively simple technical arguments. (I've thought about writing some of that up myself, now that I'm done with my PhD and have some time off.)
Aren't some of the engineering assumptions in the Friedberg paper out of date? They assume a maximum magnetic field strength of 13T, but Commonwealth has already built 20T magnets.
Yes, good points Josiah and Brian. I agree that if the 20T high-field magnet works as advertised, it might change the conclusions of the Freidberg paper. The ARC design, at 23T, is an example of what that might look like.
Nevertheless, even if 20T magnets can be built and operated at reactor scale (which I'm not 100% convinced they are) and SPARC were to generate Q >> 1 or even ignition, I think the smart money would be to remain skeptical about the high-field tokamak approach to fusion power. I've heard ARC described by someone who worked on the design as "running it like an absolute psycho". Dangerously close to instability limits, and almost certainly with too much heat for the first wall.
Like Josiah, I worry a lot about first wall heat loading. I worry a lot about neutron damage. I worry a lot about disruptions in tokamaks. Preventing a tokamak (or any DT fusion device, really) from destroying itself is a really hard problem.
And even with that, ARC design was the mass of a couple of WW2 destroyers, for (IIRC) 190 MW(e) (net). The volumetric power density was a factor of 40 worse than a PWR.
Disruptions may be solvable by a more recent invention: disruption mitigation coils. These are energized automatically by induction when the plasma current starts to decay, and cause the electrons to lose confinement, so they cannot accelerate to runaway energies.
Right. I'm saying that turning on a single 20T magnet in a lab, running it for 120 hours, and having it break in a quench test @ 75% current -- while undeniably a massive accomplishment and breakthrough that should be celebrated -- does not imply that 18 such coils can be built, assembled, and operated in a high neutron flux environment with sufficient reliability as to be a useful in a power-producing reactor. Possible in theory, yes. In practice? I don't pretend to know what the odds are, but I've seen enough failed experiments and false promises that I'll put them at less than 100%.
One thing worth noting is that boiling water reactors have power densities half that of PWRs, and gas cooled reactors as low as a tokamak at 2-4 MW/m3 according to my passing knowledge
Gas reactors don't seem to be considered as contenders for bulk electricity so that's not as persuasive as I'd like. However there are some very notable prospects for higher power densities. A tokamak mode has been achieved by GA's DIII-D reactor that has 60% more neutron output than standard H-mode with greater stability. Another really signal possibility is spin-polarized fueling which increases reactivity by 50% and power from increased self heating adding to that to total 80-90% more power
If spin-polarization can be done at the bulk levels of fusion fuels it would also be possible to reduce the tritium fraction at the cost of some of the extra power, radically increasing the burn up* and according to a recent paper by Parisi et al making tritium breeding ratio and start up inventory non issues
(*I presume as the tritium will have a much higher chance of connecting with a deuterium rather than bouncing off another tritium before leaving the plasma)
These advances would melt the walls for sure - but they may also make highly resistant liquid walls with higher sputtering reducing plasma purity affordable in terms of still having a power density of maybe even 10 MW/m3. Tokamak Energy with their spherical tokamak intend to have liquid walls which makes sense as spherical tokamaks are kind of underrated, my understanding being that they have 9x as much power per plasma volume. Albeit with a serious cost of having no heat removal and breeding blanket protecting the central solenoid
While this is still much less than a BWR, which doesn't have steam generators to pay for, the overhead of fission regulation makes a case that a liquid wall 10 MW/m3 tokamak could compete in the market potentially getting kid glove treatment from regulators. Currently physics demonstrator plants in the US are regulated under the class of medical radioactive material
Another more pedestrian factor with tritium breeding ratio is that a simple graphite reflector would increase the TBR by 20%, a dramatic effect with no other cost than activating small quantities of long half life Carbon 14. In practical terms this is not a problem for any country already handling long lived spent nuclear fuel
I've seen it argued a few times by people in the industry that the prevalence of PWR and BWRs is more a historical artefact of what drove most of the early power plant reactor construction, namely the US desire to advance the technology used in naval nuclear reactors and that it may have smothered fission in the crib due to the expensive engineering issues that PWRs and BWRs need solving that alternative fission reactors do not. Quite frankly PWRs and BWRs haven't been seen as actually economically viable reactors for a long while and when constructed are political energy security decisions not private capital allocation.
I agree with all these points. I think you captured the gist of Friedberg's paper well; it is worth also noting he is much more optimistic about feasibility when using high field magnets which is the approach taken by Commonwealth, and I wonder if the paper was written in partial advocacy for Commonwealth. One other constraint and top of my list of concerns is the first wall heat loading, which Friedberg neglects in that analysis.
Thank you for this outstanding article. I teach fusion physics on a university level, and this is the best historical overview I have ever read. Keep up the great work!
Great post. I believe some of the Q numbers are off when reporting records for JET and JT-60U. The record for magnetic confinement was achieved by JET with Q=0.67, and no tokamak has achieved Q>1 yet. This was partially why NIF breaking this milestone in 2022 was so momentous -- it was the first device ever to do so. However, inertial confinement does not scale to power plant conditions as well as magnetic confinement due to the small energy yields and laser efficiency.
Ah, it looks like those Q values for JET and JT-60 are "extrapolated Q" based on what they WOULD have achieved had deuterium-tritium been used. Will correct.
The challenges with fusion remind me of those hypothetical scenarios(*) for spaceflight where if gravity was X times stronger then chemical rockets would be _nearly_ impossible. Which raises the question, would people on such a planet ever actually develop a working rocket? What would it be like to be those frustrated engineers? With fusion, we are the people on that planet.
Re: wind and solar, I technically agree with you, but given incredible progress 'so far' could be a supermajority of energy production fairly easily, in particular if we get cleverer around storing excess energy in distributed household hot water tanks and electric cars, without even introducing large scale household batteries.
Yeah, multiple fusion insiders such as Nick Hawker of First Light Fusion have referenced "the last 20%", indicating that they believe that in a large continental grid that can balance wind and solar production lulls with transmission they will likely make up 80% of electricity
This is also a reason multiple fusion companies reference process heat as a market - the magnetic mirror company Realta specifically aim to build pure process heat machines as they could more easily custom size smaller than tokamaks
I've seen that argument a lot. The big problem with it is that if 80% of demand is being met by renewables, the residual 20% will be very unsteady, with long periods of zero residual demand. This residue is unsuited to being covered by a baseload source like fusion (or fission). Instead, it would be better covered by a low capital cost, highly dispatchable source, such as burning an e-fuel (fuel produced from renewable electricity, for example green hydrogen) in combustion turbines. Combustion turbines (or combined cycle plants) have very low capex per unit of power output, an order of magnitude cheaper than fission plants and likely similarly cheaper than fusion plants.
This makes a lot of sense to me, and I've myself gotten to a 75-80% figure for intermittent renewables without too much storage (and a small handful of gas fired plants that can be fired up quickly when needed, paid for via capacity contracts rather than by production).
That was a great write-up. Thanks for the information.
I think one factor driving the private investment in fusion is the extremely low interest rates from about 2010 through 2022. When private capital can make 0.5% per year (less than the slowest inflation) with government bonds, extremely speculative investments with potentially very large payouts start to look attractive, even if the payout might take decades.
There was also rising inequality in that a larger and larger fraction of the nation's wealth was concentrated at the top. Most of it just went to pushing up asset prices, but at least some of it has been going to support new technologies.
Thanks for your well-informed essay. I'm strongly on the side of the pessimists. I see no fundamental reason why fusion should be impossible, but find the available evidence compellingly negative: Inertial confinement is perhaps useful for weapon design, but it's hard to imagine it ever being a commercial energy source. Magnetic confinement is more plausible, but has proved to be incredibly difficult. The discipline is packed with enthusiasts, who have maintained a steady upbeat rhetoric now for many years; it seems to me that a realistic assessment needs to recognize this positive bias. I'd be shocked if a commercially-viable fusion power device generating > 10 MW were operating within the next 50 years.
Great article. The potential benefits of nuclear fusion are world changing, but the chances that it will never amount to anything cost effective are also just as high.
Readers may be interested in the milestones for viable nuclear fusion mentioned in this article by Commonwealth Fusion Systems:
“1) Can you produce plasmas that are stable enough to be a fuel for the fusion power machine, or is the plasma an unstable blip? The plasma must be fully ionized. Photos and data traces of the plasma provide evidence.
2) Can you heat the plasma to 10,000,000° C? This is 1keV, for electrons and ions, for those steeped in the science. There are established ways to measure this, but they’re tricky and need to be reviewed.
3) Is it a serious plasma, confined and dense enough that collisions necessary for fusion are starting to play a serious role (10^19 keV/m^3/s for those following along)? This is routinely measured with a metric called the triple product.
4) Can your plasma make actual fusion reactions that produce net energy gain? That threshold, denoted as Q>1, shows that, at the plasma level, a machine makes more heat than it took to heat it up. The plasma is an amplifier. Getting Q>1 shows a fusion machine is actually making meaningful fusion happen, not just reaching the right plasma conditions.
5) Can your fusion machine make enough power to sell the excess? This isn’t just at the level of the plasma, but rather considering the whole plant. Typically this is about electricity, and is called the “net electric” milestone. Importantly, this means the plant has to cover all its own needs and then some, it can’t just convert incoming electricity to “fusion-ified” electricity as a pass-through. This is measured in megawatts (MW).
6) Is your fusion power plant economical enough to eventually be a competitive option in any market in the world? This is about all the costs that go into building and operating the plant. This is typically measured in a levelized cost of electricity. If your power costs $50 per megawatt-hour (MWh), you win the market. If you can make $100/MWh you can at least enter it…
Right now there are many 1s, several 2s (mirrors, pinches, FRCs, tokamaks, stellarators, laser inertial, magneto-inertial fusion), four 3s (tokamaks, stellarators, laser inertial, MagLIF), one 4 (NIF at Lawrence Livermore National Laboratory), and no 5s or 6s.”
Well said. I don't know any where near enough enough to have an opinion. But thing I've always heard is that EVEN IF they resolved these problems they might still face severe tritium supply limitations which could prevent mass deployment
A DT reactor would breed its own tritium by neutron bombardment of lithium, tritium supply for a real fusion reactor is a non-issue.
The real killer is the materials science issues. We can't even effectively simulate the neutron flux from a fusion reactor, there's a whole other facility adjoining ITER, IFMIF (International Fusion Materials Irradiation Facility) to let us explore that regime, but it's even further behind schedule. Commercial fusion ain't happening.
Cool! Thanks pointing that out, thats interesting, I didn't know about that. I know very little about this stuff.
I just asked internet ai about it and it said the below are its challenges, are they about right?:
Breeding Ratio: The breeding ratio, which is the number of tritium atoms produced per neutron absorbed, needs to be greater than one for a fusion reactor to sustain itself. Achieving this ratio in practice can be complex due to inefficiencies in neutron capture and other engineering challenges. If the breeding ratio is not adequately high, the reactor may not generate enough tritium to sustain continuous operation.
Materials and Design: The materials used to construct the breeding blankets (where tritium is bred) must not only be capable of withstanding intense neutron bombardment but also efficient at absorbing neutrons to produce tritium. Materials that meet these requirements are still under development, and their long-term durability and effectiveness in a real reactor environment are not yet fully proven.
Neutron Economy: Efficient neutron management is crucial for effective tritium breeding. Neutrons must be slowed down and captured by lithium to breed tritium, but they can also be lost through various processes such as escape from the reactor or capture by other materials. Optimizing the neutron economy in a reactor design is a significant engineering challenge.
Technological and Engineering Hurdles: The engineering required to integrate a breeding blanket into a fusion reactor is non-trivial. It involves not only handling high-energy neutrons but also managing heat extraction, material integrity, and the chemical processing of lithium and bred tritium. These systems must operate reliably under extreme conditions.
Scale-Up and Commercial Viability: Even if tritium breeding is demonstrated to work effectively in experimental reactors like ITER, scaling this up to commercial reactors presents additional challenges. The consistency, reliability, and economic viability of tritium breeding at a commercial scale remain uncertain.
Even if they solved those problems, there are at least two other problems. If you look at the 2013 ARC paper on arxiv, two things pop out:
First, a single ARC machine used a good fraction of the world's annual production of beryllium to build it. Extrapolating to power the world, and using the USGS estimate of ultimate available Be resource (not reserve), these reactors might supply maybe 2% of current world primary energy demand.
Second, the power density of the reactor was lower than the power density of a PWR by a factor of 40. This is the power density using the volume of the entire fusion reactor vs. the entire PWR reactor pressure vessel. The ratio of power densities of the plasma vs. fission reactor core is not quite as large, but we'd be paying for the reactors, not just those cores.
At least ARC wasn't as bad as ITER, where the ratio is a factor of 400! Still probably bad enough to render the economics unworkable, though. The reactor of a PWR power plant is maybe 12% of the capex; multiply that by 40 and you've increased the capex of your power plant by a factor of more than 5.
I've heard that even lead could be an issue, if lead is used as the neutron multiplier (by (n,2n) reactions), although not to the extent of beryllium.
There's no current ability to do isotopic separation of lithium at the scale needed. A single ARC reactor would use Li-6 in quantity approaching that of the entire US hydrogen bomb program, and the technology used in that program is no longer usable due to concerns about mercury pollution. A new technology would have to be developed, demonstrated at scale, and built out.
Even the structural materials needed have supply constraints. For example, trace materials must be strictly controlled to keep activation within limits. Even traces of nitrogen will produce enough carbon-14 that the structural materials will no longer qualify as low level waste (after a shortish cooling off period).
Tritium is not essential to fusion reaction. There are many different combinations of elements and isotopes that are potentially viable. That is why I think research needs to focus more on elements that are widely available.
I don;t know about it. But what I've heard and read is that none the potential alternatives has gotten the level of tech maturity AND economic viability (adequate cost effective supply) required for mass scale roll out (neither has tritium but from what I've gathered its much closer). But I would agree that they should be researched
Yes, that is definitely true. The technology is still very early in its development phase. We still have not even gotten to the prototype phase, and we may never get there.
Another hedge against the bear case for fusion is that temperatures matter.
A host of industrial processes need heat that is impractical to generate with electrical resistance heating and is currently generated by burning (usually very dirty) stuff; fusion might not be entirely cost-competitive on electricity alone but co-generation near major industrial facilities may pencil out much better as it replaces expensive, dirty combustion fuels.
Taking that principle further, what realms of materials science and chemistry will become possible when temperatures a hundred or a thousand times as great of those which combustion can achieve are available for industrial applications?
There is nothing gathered the heat is easier to get into a process than heat from electrical resistance. You still need heat exchangers and working fluids.
Fission will easily get to the max temps before the transfer is the limit anyway.
Also, heat (even at high temperature) is extremely storable, so electric resistance heating can exploit electricity when prices are low. The storage of heat in firebrick is a technology going back more than a century, where it was extensively used in the iron and steel industry, in blast and open hearth furnaces.
Well layed out and interesting!. I don’t have an opinion on the matter, but there’s a fun book it and its history (link below) called “The Fairy Tale Of Nuclear Fusion” (written by L.J. Reinders). The following text is a segment from its conclusion segment:
“This book’s central thesis, if there is one, is that there is no chance whatsoever that in this century nuclear fusion will make a contribution of any significance to the carbon-free energy mix, implying that it will not play a role in the urgent decarbonization of energy production, in spite of the claims to the contrary made by the fusion community. No commercially viable nuclear fusion power station will be working in this century, and probably never will, quite a climbdown from its original boast of “too cheap to meter”, and in spite of all the “breakthroughs” in the time between. With every breakthrough the generation of energy by nuclear fusion, cheap or expensive, was delayed by a further year or two. In addition, the expected price of the energy went up, quite contrary to the ordinary behaviour of the price of a commodity when a breakthrough is achieved. It is clear that the emperor has no clothes and never had any. This conclusion stands, even if everything goes according to plan, which it never does (just think about the rather simple mishap of a magnet shorting out with the upgrade of the NSTX in 2016, a problem that now, four years later, has still not been sorted). The main conclusion of the book can be summarised as follows: “If ITER fulfils all its promises, meaning that it: shows that energy production by controlled nuclear fusion is in principle possible (i.e. that it succeeds in achieving Q = 10, without yet producing any net energy as more energy will be put in than comes out); shows that tritium production in a lithium or other blanket is possible; shows that this tritium can be collected in sufficient quantities; shows that disruptions and instabilities can be kept at bay; shows that the structure of the facility can withstand the onslaught of the neutrons (not such a problem at ITER, but a very real one for any follow-up device); then a still bigger DEMO device must first be built in which all this experience will have to be incorporated, resulting in some net energy production, plus a surplus amount of tritium (self-sufficiency of tritium is a must and nobody has so far shown in any convincing way that this is possible). The DEMO must also show that the materials used, some of them still to be developed, will not too easily be activated. Such a DEMO can be built at the earliest after 2040, will need a further twenty odd years to show its viability and resolve the problems and difficulties that will undoubtedly arise, before the construction of the first pilot power plant can be contemplated, constructed and set to work. Results from DEMO cannot be expected before the 2060s, after which this pilot plant will have to show that commercial energy production from fusion is indeed possible. This will take us at least into the 2080s, if not a considerable time later still, after which the construction of real commercial power stations can possibly start. As discussed in this book, hundreds, if not thousands, of such power stations are needed for nuclear fusion to make a sizable contribution to energy generation. The scenario in Chap. 19 requires 2760 fusion power plants to provide 30% of baseload electricity by the end of this century; so, a paltry 3% would already require 276 power plants! There is simply not enough time to build a sizable number of power plants in the at most two decades left for this. The experience laid out in this book shows beyond any reasonable doubt that all the remaining eighty years of this century will be far too short for all this to happen.”
I felt like I was flying over the history of nuclear fusion developments with sparkling clarity and appreciating the geniuses at work. Very informative and exhilarating. Thank you.
Maybe because we aren’t stars? Seriously, I don’t get all the hoopla over fusion. Nuclear power is way easier, and so way cheaper than fusion. If fusion was such a great choice for power generation, so would be nuclear and virtually all electric power would long ago have gone to nuclear.
Excellent work as always, thank you so much. I think the most import point is the last one: why does this method of boiling water beat the suite of {fission, advanced geothermal, cleaned-up (or not) natural gas, ...}?
I confess to being a bit mystified by your last paragraph. You have conclusively demonstrated that the road to fusion is absurdly expensive and uncertain compared to deploying fission, advanced geothermal, or (synthesized or cleaned-up) methane. That makes it an unreasonable bet.
Fusion was sold as generating energy using safe, cheap and abundant fuel without generating nuclear waste. That's true of H-H fusion as happens in the Sun.
Conversely, the D-T fusion reaction (almost) everyone works on uses two radioactive, expensive and extremely rare fuels to generate lots of low-level nuclear waste due to neutron activation. That's not what we signed up for! Even if it worked perfectly, a D-T fusion power reactor is not a politically or socially desirable product for the same reasons as fission power reactors. They are a bit better than fission but would still be a political and social dead-end, even if the engineering and economics worked.
The various aneutronic reactions like D-He3 or H-Li6 are so much more difficult than D-T that I think it's a waste of time to consider them at this point. And they have fuel issues of their own, even if they fix the nuclear waste problem.
The case for fusion just doesn't hold up once we get into the actual implementation details. We aren't getting what we were sold.
Deuterium is neither radioactive nor extremely rare. And the idea is that a DT reaction would breed its own tritium.
It's still not gonna work because of the insane materials issues required (nothing we know of can withstand the thousands of displacements per atom the intense neutron flux would cause), though.
With the small supply of tritium and it's short half-life would there even be enough to get commercial reactors started once they come on line? Or can tritium be 'extracted' from one reactor to use in another?
Tritium is presently manufactured in fission reactors. We can make enough to start up a fusion plant if we need to.
The general idea is that you blanket your fusion reactor in lithium. Neutrons escape the plasma, impact the lithium, and turn some of it to tritium. Periodically you tap off the tritium and now you can do what you want with it. Feed it back into that reactor, feed it to another reactor, whatever.
Nice article. I especially liked the history of fusion research. A couple points:
(1) As you mentioned, the energy density (energy per unit mass) of fusion is four times higher than nuclear fission. However, the power density (power per unit volume) of fusion is 1-2 orders of magnitude lower than fission. In other words, fusion is great if you want to build a thermonuclear warhead, but if you want to build a power plant, fusion's low power density is a significant drawback compared to fission. To explain in more detail: LWRs have a power density on the order of 100 MW/m^3 in the core, while D-T fusion has a (maximum) power density of 0.1 MW/m^3 times the pressure (in atmospheres) squared. ITER is expected to have a pressure of 2.6 atm (0.5 MW/m^3), while tokamak power plants would likely have a pressure of about 7 atm (5 MW/m^3). So, a D-T fusion reactor is expected to have somewhere in the ballpark of 20x lower power density than a LWR.
(2) The power density (at constant pressure) of deuterium-tritium (D-T) fusion is 2 orders of magnitude higher than the second-highest power density reaction, deuterium-helium 3 (D-3He). Other fuels (such as D-D or p-11B) are roughly another order of magnitude lower still than D-3He. Anyone attempting to build a fusion reactor with "alternative" (non-D-T) fuels will either need to create extraordinarily high pressures, build an extraordinarily large reactor, or produce extraordinarily little power. In my opinion, it's almost impossible to imagine reactions with such low power densities being used for power plants.
(3) The "success" of a fusion reactor is often discussed, as it mostly is here, in terms of achieving a triple product greater than the Lawson criterion (8 atm*s for DT). However, there are two other essential things a fusion reactor must do. First, it must be able to create tritium fuel inside the reactor. Second, it must not destroy itself. These are extremely difficult engineering problems. Part of what makes them so challenging is that, while possible solutions can be analyzed theoretically before a reactor is built, those solutions can only be tested inside a power-producing reactor. Thus, progress on these difficult problems must be made in series, not in parallel. Only once you've spent a few billion dollars on a power-producing fusion reactor can you actually test whether it "works". At that point, you might not produce as much tritium fuel as you had calculated that you would. Or, your reactor could destroy itself -- and it is quite likely to do so, either via neutron damage, heat flux damage, or in a disruption. In my opinion -- and as far as I know this is a widespread view among academics unaffiliated with fusion startups -- it is certainly possible that one of the fusion startups will "successfully build a" [power-producing] machine, but it is very unlikely that any of them will successfully build a fusion power plant. The opposing view -- and to the best of my knowledge, this is how Commonwealth Fusion Systems thinks about these challenges -- is that the public excitement created by the achievement of high-Q magnetic confinement fusion will lead to so much investment in fusion that we'll be able to afford to solve those difficult engineering challenges later on. Kicking the can down the road, if you will. In any case, I think a discussion of the engineering challenges associated with fusion power plants is essential.
(4) Using the development of the atomic bomb as an analogy: in 1940, the Frisch-Peierls memorandum performed some simple calculations suggesting that an atomic bomb was theoretically possible, and would require a few kilograms of U-235. We now have enough knowledge, at least for tokamaks (the reactor concept with by far the best performance), to calculate whether a fusion reactor can scale to a power plant. This is done in "Designing a tokamak fusion reactor—How does plasma physics fit in?" (J.P. Friedberg et al., 2015), which writes "Unfortunately, a tokamak reactor designed on the basis of standard engineering and nuclear physics constraints does not scale to a reactor." In my opinion, this is a fundamental difference between the development of nuclear fusion and nuclear fission or the atomic bomb: with fusion, our best calculations suggest that it will not work. Or more precisely, tokamaks will not work barring significant improvements in certain engineering constraints.
I think you've done a nice job of laying out the bull and bear cases for fusion. However, I think the bear case for fusion could be made much stronger with some relatively simple technical arguments. (I've thought about writing some of that up myself, now that I'm done with my PhD and have some time off.)
Aren't some of the engineering assumptions in the Friedberg paper out of date? They assume a maximum magnetic field strength of 13T, but Commonwealth has already built 20T magnets.
Yes, good points Josiah and Brian. I agree that if the 20T high-field magnet works as advertised, it might change the conclusions of the Freidberg paper. The ARC design, at 23T, is an example of what that might look like.
Nevertheless, even if 20T magnets can be built and operated at reactor scale (which I'm not 100% convinced they are) and SPARC were to generate Q >> 1 or even ignition, I think the smart money would be to remain skeptical about the high-field tokamak approach to fusion power. I've heard ARC described by someone who worked on the design as "running it like an absolute psycho". Dangerously close to instability limits, and almost certainly with too much heat for the first wall.
Like Josiah, I worry a lot about first wall heat loading. I worry a lot about neutron damage. I worry a lot about disruptions in tokamaks. Preventing a tokamak (or any DT fusion device, really) from destroying itself is a really hard problem.
And even with that, ARC design was the mass of a couple of WW2 destroyers, for (IIRC) 190 MW(e) (net). The volumetric power density was a factor of 40 worse than a PWR.
Disruptions may be solvable by a more recent invention: disruption mitigation coils. These are energized automatically by induction when the plasma current starts to decay, and cause the electrons to lose confinement, so they cannot accelerate to runaway energies.
Nick, they build the 20T magnet at (effectively) reactor scale 3 years ago:
https://news.mit.edu/2021/MIT-CFS-major-advance-toward-fusion-energy-0908
Right. I'm saying that turning on a single 20T magnet in a lab, running it for 120 hours, and having it break in a quench test @ 75% current -- while undeniably a massive accomplishment and breakthrough that should be celebrated -- does not imply that 18 such coils can be built, assembled, and operated in a high neutron flux environment with sufficient reliability as to be a useful in a power-producing reactor. Possible in theory, yes. In practice? I don't pretend to know what the odds are, but I've seen enough failed experiments and false promises that I'll put them at less than 100%.
The 20T on-coil measurement is roughly a 12T on-axis measurement. See A. Creely's "Overview of the SPARC Tokamak" paper.
One thing worth noting is that boiling water reactors have power densities half that of PWRs, and gas cooled reactors as low as a tokamak at 2-4 MW/m3 according to my passing knowledge
Gas reactors don't seem to be considered as contenders for bulk electricity so that's not as persuasive as I'd like. However there are some very notable prospects for higher power densities. A tokamak mode has been achieved by GA's DIII-D reactor that has 60% more neutron output than standard H-mode with greater stability. Another really signal possibility is spin-polarized fueling which increases reactivity by 50% and power from increased self heating adding to that to total 80-90% more power
If spin-polarization can be done at the bulk levels of fusion fuels it would also be possible to reduce the tritium fraction at the cost of some of the extra power, radically increasing the burn up* and according to a recent paper by Parisi et al making tritium breeding ratio and start up inventory non issues
(*I presume as the tritium will have a much higher chance of connecting with a deuterium rather than bouncing off another tritium before leaving the plasma)
These advances would melt the walls for sure - but they may also make highly resistant liquid walls with higher sputtering reducing plasma purity affordable in terms of still having a power density of maybe even 10 MW/m3. Tokamak Energy with their spherical tokamak intend to have liquid walls which makes sense as spherical tokamaks are kind of underrated, my understanding being that they have 9x as much power per plasma volume. Albeit with a serious cost of having no heat removal and breeding blanket protecting the central solenoid
While this is still much less than a BWR, which doesn't have steam generators to pay for, the overhead of fission regulation makes a case that a liquid wall 10 MW/m3 tokamak could compete in the market potentially getting kid glove treatment from regulators. Currently physics demonstrator plants in the US are regulated under the class of medical radioactive material
Another more pedestrian factor with tritium breeding ratio is that a simple graphite reflector would increase the TBR by 20%, a dramatic effect with no other cost than activating small quantities of long half life Carbon 14. In practical terms this is not a problem for any country already handling long lived spent nuclear fuel
I've seen it argued a few times by people in the industry that the prevalence of PWR and BWRs is more a historical artefact of what drove most of the early power plant reactor construction, namely the US desire to advance the technology used in naval nuclear reactors and that it may have smothered fission in the crib due to the expensive engineering issues that PWRs and BWRs need solving that alternative fission reactors do not. Quite frankly PWRs and BWRs haven't been seen as actually economically viable reactors for a long while and when constructed are political energy security decisions not private capital allocation.
I agree with all these points. I think you captured the gist of Friedberg's paper well; it is worth also noting he is much more optimistic about feasibility when using high field magnets which is the approach taken by Commonwealth, and I wonder if the paper was written in partial advocacy for Commonwealth. One other constraint and top of my list of concerns is the first wall heat loading, which Friedberg neglects in that analysis.
Thank you for this outstanding article. I teach fusion physics on a university level, and this is the best historical overview I have ever read. Keep up the great work!
Great post. I believe some of the Q numbers are off when reporting records for JET and JT-60U. The record for magnetic confinement was achieved by JET with Q=0.67, and no tokamak has achieved Q>1 yet. This was partially why NIF breaking this milestone in 2022 was so momentous -- it was the first device ever to do so. However, inertial confinement does not scale to power plant conditions as well as magnetic confinement due to the small energy yields and laser efficiency.
Ah, it looks like those Q values for JET and JT-60 are "extrapolated Q" based on what they WOULD have achieved had deuterium-tritium been used. Will correct.
The challenges with fusion remind me of those hypothetical scenarios(*) for spaceflight where if gravity was X times stronger then chemical rockets would be _nearly_ impossible. Which raises the question, would people on such a planet ever actually develop a working rocket? What would it be like to be those frustrated engineers? With fusion, we are the people on that planet.
* See eg https://space.stackexchange.com/questions/14383/how-much-bigger-could-earth-be-before-rockets-wouldnt-work/
I personally believe we should bet on nuclear fission power generation. It works, it's reliable and relatively safe.
Wind and solar is only going to get us so far.
Solar is growing enormously. Same with batteries. It's far more likely that will cover our needs than Fusion in a 10 year time frame.
Re: wind and solar, I technically agree with you, but given incredible progress 'so far' could be a supermajority of energy production fairly easily, in particular if we get cleverer around storing excess energy in distributed household hot water tanks and electric cars, without even introducing large scale household batteries.
Yeah, multiple fusion insiders such as Nick Hawker of First Light Fusion have referenced "the last 20%", indicating that they believe that in a large continental grid that can balance wind and solar production lulls with transmission they will likely make up 80% of electricity
This is also a reason multiple fusion companies reference process heat as a market - the magnetic mirror company Realta specifically aim to build pure process heat machines as they could more easily custom size smaller than tokamaks
I've seen that argument a lot. The big problem with it is that if 80% of demand is being met by renewables, the residual 20% will be very unsteady, with long periods of zero residual demand. This residue is unsuited to being covered by a baseload source like fusion (or fission). Instead, it would be better covered by a low capital cost, highly dispatchable source, such as burning an e-fuel (fuel produced from renewable electricity, for example green hydrogen) in combustion turbines. Combustion turbines (or combined cycle plants) have very low capex per unit of power output, an order of magnitude cheaper than fission plants and likely similarly cheaper than fusion plants.
This makes a lot of sense to me, and I've myself gotten to a 75-80% figure for intermittent renewables without too much storage (and a small handful of gas fired plants that can be fired up quickly when needed, paid for via capacity contracts rather than by production).
That was a great write-up. Thanks for the information.
I think one factor driving the private investment in fusion is the extremely low interest rates from about 2010 through 2022. When private capital can make 0.5% per year (less than the slowest inflation) with government bonds, extremely speculative investments with potentially very large payouts start to look attractive, even if the payout might take decades.
There was also rising inequality in that a larger and larger fraction of the nation's wealth was concentrated at the top. Most of it just went to pushing up asset prices, but at least some of it has been going to support new technologies.
Thanks for your well-informed essay. I'm strongly on the side of the pessimists. I see no fundamental reason why fusion should be impossible, but find the available evidence compellingly negative: Inertial confinement is perhaps useful for weapon design, but it's hard to imagine it ever being a commercial energy source. Magnetic confinement is more plausible, but has proved to be incredibly difficult. The discipline is packed with enthusiasts, who have maintained a steady upbeat rhetoric now for many years; it seems to me that a realistic assessment needs to recognize this positive bias. I'd be shocked if a commercially-viable fusion power device generating > 10 MW were operating within the next 50 years.
Great article. The potential benefits of nuclear fusion are world changing, but the chances that it will never amount to anything cost effective are also just as high.
Readers may be interested in the milestones for viable nuclear fusion mentioned in this article by Commonwealth Fusion Systems:
https://cfs.energy/news-and-media/building-trust-in-fusion-energy
The key passage is:
“1) Can you produce plasmas that are stable enough to be a fuel for the fusion power machine, or is the plasma an unstable blip? The plasma must be fully ionized. Photos and data traces of the plasma provide evidence.
2) Can you heat the plasma to 10,000,000° C? This is 1keV, for electrons and ions, for those steeped in the science. There are established ways to measure this, but they’re tricky and need to be reviewed.
3) Is it a serious plasma, confined and dense enough that collisions necessary for fusion are starting to play a serious role (10^19 keV/m^3/s for those following along)? This is routinely measured with a metric called the triple product.
4) Can your plasma make actual fusion reactions that produce net energy gain? That threshold, denoted as Q>1, shows that, at the plasma level, a machine makes more heat than it took to heat it up. The plasma is an amplifier. Getting Q>1 shows a fusion machine is actually making meaningful fusion happen, not just reaching the right plasma conditions.
5) Can your fusion machine make enough power to sell the excess? This isn’t just at the level of the plasma, but rather considering the whole plant. Typically this is about electricity, and is called the “net electric” milestone. Importantly, this means the plant has to cover all its own needs and then some, it can’t just convert incoming electricity to “fusion-ified” electricity as a pass-through. This is measured in megawatts (MW).
6) Is your fusion power plant economical enough to eventually be a competitive option in any market in the world? This is about all the costs that go into building and operating the plant. This is typically measured in a levelized cost of electricity. If your power costs $50 per megawatt-hour (MWh), you win the market. If you can make $100/MWh you can at least enter it…
Right now there are many 1s, several 2s (mirrors, pinches, FRCs, tokamaks, stellarators, laser inertial, magneto-inertial fusion), four 3s (tokamaks, stellarators, laser inertial, MagLIF), one 4 (NIF at Lawrence Livermore National Laboratory), and no 5s or 6s.”
Well said. I don't know any where near enough enough to have an opinion. But thing I've always heard is that EVEN IF they resolved these problems they might still face severe tritium supply limitations which could prevent mass deployment
A DT reactor would breed its own tritium by neutron bombardment of lithium, tritium supply for a real fusion reactor is a non-issue.
The real killer is the materials science issues. We can't even effectively simulate the neutron flux from a fusion reactor, there's a whole other facility adjoining ITER, IFMIF (International Fusion Materials Irradiation Facility) to let us explore that regime, but it's even further behind schedule. Commercial fusion ain't happening.
Cool! Thanks pointing that out, thats interesting, I didn't know about that. I know very little about this stuff.
I just asked internet ai about it and it said the below are its challenges, are they about right?:
Breeding Ratio: The breeding ratio, which is the number of tritium atoms produced per neutron absorbed, needs to be greater than one for a fusion reactor to sustain itself. Achieving this ratio in practice can be complex due to inefficiencies in neutron capture and other engineering challenges. If the breeding ratio is not adequately high, the reactor may not generate enough tritium to sustain continuous operation.
Materials and Design: The materials used to construct the breeding blankets (where tritium is bred) must not only be capable of withstanding intense neutron bombardment but also efficient at absorbing neutrons to produce tritium. Materials that meet these requirements are still under development, and their long-term durability and effectiveness in a real reactor environment are not yet fully proven.
Neutron Economy: Efficient neutron management is crucial for effective tritium breeding. Neutrons must be slowed down and captured by lithium to breed tritium, but they can also be lost through various processes such as escape from the reactor or capture by other materials. Optimizing the neutron economy in a reactor design is a significant engineering challenge.
Technological and Engineering Hurdles: The engineering required to integrate a breeding blanket into a fusion reactor is non-trivial. It involves not only handling high-energy neutrons but also managing heat extraction, material integrity, and the chemical processing of lithium and bred tritium. These systems must operate reliably under extreme conditions.
Scale-Up and Commercial Viability: Even if tritium breeding is demonstrated to work effectively in experimental reactors like ITER, scaling this up to commercial reactors presents additional challenges. The consistency, reliability, and economic viability of tritium breeding at a commercial scale remain uncertain.
Even if they solved those problems, there are at least two other problems. If you look at the 2013 ARC paper on arxiv, two things pop out:
First, a single ARC machine used a good fraction of the world's annual production of beryllium to build it. Extrapolating to power the world, and using the USGS estimate of ultimate available Be resource (not reserve), these reactors might supply maybe 2% of current world primary energy demand.
Second, the power density of the reactor was lower than the power density of a PWR by a factor of 40. This is the power density using the volume of the entire fusion reactor vs. the entire PWR reactor pressure vessel. The ratio of power densities of the plasma vs. fission reactor core is not quite as large, but we'd be paying for the reactors, not just those cores.
At least ARC wasn't as bad as ITER, where the ratio is a factor of 400! Still probably bad enough to render the economics unworkable, though. The reactor of a PWR power plant is maybe 12% of the capex; multiply that by 40 and you've increased the capex of your power plant by a factor of more than 5.
Interesting! Another angle. So there are more than one potential debilitating supply constraints, are there others besides beryllium and tritium?
I've heard that even lead could be an issue, if lead is used as the neutron multiplier (by (n,2n) reactions), although not to the extent of beryllium.
There's no current ability to do isotopic separation of lithium at the scale needed. A single ARC reactor would use Li-6 in quantity approaching that of the entire US hydrogen bomb program, and the technology used in that program is no longer usable due to concerns about mercury pollution. A new technology would have to be developed, demonstrated at scale, and built out.
Even the structural materials needed have supply constraints. For example, trace materials must be strictly controlled to keep activation within limits. Even traces of nitrogen will produce enough carbon-14 that the structural materials will no longer qualify as low level waste (after a shortish cooling off period).
Interesting. It sounds like, regarding it as a replacement for fission, it may be theres a chance its just trading one waste problem for another
Tritium is not essential to fusion reaction. There are many different combinations of elements and isotopes that are potentially viable. That is why I think research needs to focus more on elements that are widely available.
All of those other materials are incredibly harder to get working than DT.
I don;t know about it. But what I've heard and read is that none the potential alternatives has gotten the level of tech maturity AND economic viability (adequate cost effective supply) required for mass scale roll out (neither has tritium but from what I've gathered its much closer). But I would agree that they should be researched
Yes, that is definitely true. The technology is still very early in its development phase. We still have not even gotten to the prototype phase, and we may never get there.
Another hedge against the bear case for fusion is that temperatures matter.
A host of industrial processes need heat that is impractical to generate with electrical resistance heating and is currently generated by burning (usually very dirty) stuff; fusion might not be entirely cost-competitive on electricity alone but co-generation near major industrial facilities may pencil out much better as it replaces expensive, dirty combustion fuels.
Taking that principle further, what realms of materials science and chemistry will become possible when temperatures a hundred or a thousand times as great of those which combustion can achieve are available for industrial applications?
There is nothing gathered the heat is easier to get into a process than heat from electrical resistance. You still need heat exchangers and working fluids.
Fission will easily get to the max temps before the transfer is the limit anyway.
Also, heat (even at high temperature) is extremely storable, so electric resistance heating can exploit electricity when prices are low. The storage of heat in firebrick is a technology going back more than a century, where it was extensively used in the iron and steel industry, in blast and open hearth furnaces.
Well layed out and interesting!. I don’t have an opinion on the matter, but there’s a fun book it and its history (link below) called “The Fairy Tale Of Nuclear Fusion” (written by L.J. Reinders). The following text is a segment from its conclusion segment:
“This book’s central thesis, if there is one, is that there is no chance whatsoever that in this century nuclear fusion will make a contribution of any significance to the carbon-free energy mix, implying that it will not play a role in the urgent decarbonization of energy production, in spite of the claims to the contrary made by the fusion community. No commercially viable nuclear fusion power station will be working in this century, and probably never will, quite a climbdown from its original boast of “too cheap to meter”, and in spite of all the “breakthroughs” in the time between. With every breakthrough the generation of energy by nuclear fusion, cheap or expensive, was delayed by a further year or two. In addition, the expected price of the energy went up, quite contrary to the ordinary behaviour of the price of a commodity when a breakthrough is achieved. It is clear that the emperor has no clothes and never had any. This conclusion stands, even if everything goes according to plan, which it never does (just think about the rather simple mishap of a magnet shorting out with the upgrade of the NSTX in 2016, a problem that now, four years later, has still not been sorted). The main conclusion of the book can be summarised as follows: “If ITER fulfils all its promises, meaning that it: shows that energy production by controlled nuclear fusion is in principle possible (i.e. that it succeeds in achieving Q = 10, without yet producing any net energy as more energy will be put in than comes out); shows that tritium production in a lithium or other blanket is possible; shows that this tritium can be collected in sufficient quantities; shows that disruptions and instabilities can be kept at bay; shows that the structure of the facility can withstand the onslaught of the neutrons (not such a problem at ITER, but a very real one for any follow-up device); then a still bigger DEMO device must first be built in which all this experience will have to be incorporated, resulting in some net energy production, plus a surplus amount of tritium (self-sufficiency of tritium is a must and nobody has so far shown in any convincing way that this is possible). The DEMO must also show that the materials used, some of them still to be developed, will not too easily be activated. Such a DEMO can be built at the earliest after 2040, will need a further twenty odd years to show its viability and resolve the problems and difficulties that will undoubtedly arise, before the construction of the first pilot power plant can be contemplated, constructed and set to work. Results from DEMO cannot be expected before the 2060s, after which this pilot plant will have to show that commercial energy production from fusion is indeed possible. This will take us at least into the 2080s, if not a considerable time later still, after which the construction of real commercial power stations can possibly start. As discussed in this book, hundreds, if not thousands, of such power stations are needed for nuclear fusion to make a sizable contribution to energy generation. The scenario in Chap. 19 requires 2760 fusion power plants to provide 30% of baseload electricity by the end of this century; so, a paltry 3% would already require 276 power plants! There is simply not enough time to build a sizable number of power plants in the at most two decades left for this. The experience laid out in this book shows beyond any reasonable doubt that all the remaining eighty years of this century will be far too short for all this to happen.”
Link to book:
https://www.amazon.com/Fairy-Tale-Nuclear-Fusion/dp/3030643433
I felt like I was flying over the history of nuclear fusion developments with sparkling clarity and appreciating the geniuses at work. Very informative and exhilarating. Thank you.
Maybe because we aren’t stars? Seriously, I don’t get all the hoopla over fusion. Nuclear power is way easier, and so way cheaper than fusion. If fusion was such a great choice for power generation, so would be nuclear and virtually all electric power would long ago have gone to nuclear.
Excellent work as always, thank you so much. I think the most import point is the last one: why does this method of boiling water beat the suite of {fission, advanced geothermal, cleaned-up (or not) natural gas, ...}?
I confess to being a bit mystified by your last paragraph. You have conclusively demonstrated that the road to fusion is absurdly expensive and uncertain compared to deploying fission, advanced geothermal, or (synthesized or cleaned-up) methane. That makes it an unreasonable bet.
Fusion was sold as generating energy using safe, cheap and abundant fuel without generating nuclear waste. That's true of H-H fusion as happens in the Sun.
Conversely, the D-T fusion reaction (almost) everyone works on uses two radioactive, expensive and extremely rare fuels to generate lots of low-level nuclear waste due to neutron activation. That's not what we signed up for! Even if it worked perfectly, a D-T fusion power reactor is not a politically or socially desirable product for the same reasons as fission power reactors. They are a bit better than fission but would still be a political and social dead-end, even if the engineering and economics worked.
The various aneutronic reactions like D-He3 or H-Li6 are so much more difficult than D-T that I think it's a waste of time to consider them at this point. And they have fuel issues of their own, even if they fix the nuclear waste problem.
The case for fusion just doesn't hold up once we get into the actual implementation details. We aren't getting what we were sold.
It kinda reminds me of that California high-speed rail project…
Deuterium is neither radioactive nor extremely rare. And the idea is that a DT reaction would breed its own tritium.
It's still not gonna work because of the insane materials issues required (nothing we know of can withstand the thousands of displacements per atom the intense neutron flux would cause), though.
With the small supply of tritium and it's short half-life would there even be enough to get commercial reactors started once they come on line? Or can tritium be 'extracted' from one reactor to use in another?
Tritium is presently manufactured in fission reactors. We can make enough to start up a fusion plant if we need to.
The general idea is that you blanket your fusion reactor in lithium. Neutrons escape the plasma, impact the lithium, and turn some of it to tritium. Periodically you tap off the tritium and now you can do what you want with it. Feed it back into that reactor, feed it to another reactor, whatever.
Wonderful article. Thanks so much for the effort.
Fascinating read, thanks. While fusion is a scientist’s ideal, its utility disappears if you don’t care about carbon and nuclear waste.