Hacker News Reader

Comment by bluepanda928752 | Hacker News Reader

The only fusion project which has a chance of producing excess heat at Q>2 within a decade is the MIT SPARC, using proven plasma physics and scaling the size/cost down dramatically with high-field HTS magnets (think ITER but sooner and >10x cheaper). Why is this definite solution to climate apocalypse being developed within the framework of MIT startup accelerator instead of Manhattan Project while most of the publicity goes to unproven designs which are orders of magnitude from being anywhere close to Q>1 is beyond me

MIT SPARC overview presentation/recent progress: https://www.youtube.com/watch?v=h8uYNhevRtk

Journal of Plasma Physics issue with several papers about it: https://www.cambridge.org/core/journals/journal-of-plasma-ph...

View on HN

Definite solution? It's a definite non-solution, even if the plasma physics is more nailed down. The ARC reactor (fully scaled up SPARC with tritium breeding blanket) would have a power density 40x worse than a PWR primary reactor vessel, and supplying the world's primary energy demand with them would require 100x more beryllium than the USGS estimated resource (not reserve) of that element.

Not sure why lower power density might be a show-stopper here. Beryllium angle is interesting, haven't thought about that

The cost of the reactor will be proportional to its size, so the cost/power will be inversely proportional to the power density. Lawrence Lidsky (who was also at MIT) (and also a similar argument from Pfirsch and Schmitter in Germany) famously pointed out back in the 1980s that DT fusion reactors will inherently have terrible power density compared to fission reactors, and this will render them noncompetitive. Despite putative rebuttals at the time, nothing we've seen since contradicts their devastating argument.

http://orcutt.net/weblog/wp-content/uploads/2015/08/The-Trou...

https://ui.adsabs.harvard.edu/abs/1987oepn.book.....P/abstra...

Note that Helion wants to go with D-3He, and use direct conversion for at least some of the energy recovery. This might be the only hope for making fusion compete. But of course you need 3He; making it with DD fusion requires even more aggressive plasma physics.

ARC at least isn't quite as absurd as a tokamak the size of ITER, which has a gross fusion power density another order of magnitude lower.

I think high-field HTS tokamaks projected costs are in a reasonable range despite the raw power density being lower. There could be additional savings related to radioactive waste processing since fusion should generate less and also containment/security for similar reasons

Helion and other new fusion projects are, surely, interesting, however tokamaks are so much more ready and, with high-field magnets, likely economical

I don't believe cost projections for fusion. If you look at them, they're filled with assumptions that aren't supported by much of anything(*), but magically make the technology just competitive. As the competition has improved, the assumptions have gotten more desperate. They're less "this is what the technology will cost" and more "this is the least ridiculous set of assumptions we could find that would let our technology not be dead."

If you apply the same level of assumptions to, say, light water fission reactors, I'm sure you'd get cost estimates vastly lower than what they actually cost in practice.

(*) For example, one paper assuming the efficiency of converting thermal energy to power in the fusion reactor is 60%, a level that combined cycle power plants achieve by expanding combustion gas that starts at a temperature that would soften or even melt the turbine blades.

I've seen some claims that even a magically costless heat producing device connected to a steam turbine won't be competitive with solar PV so I'd guess Fusion would also fail that test if it is basically being used to generate heat (note, they're claiming they have some new tech that avoids this problem, but we don't know if actually does or not)

The real question is whether it'd be competitive with solar PV on cloudy winter days with enough battery to get through windless nights. If so, there's probably a place on the grid for it.

(In any case, Helion is planning direct energy conversion, without a heat cycle.)

At that point you're competing against energy storage, not renewables. No way you can afford to run any such power system just a few days a year.

It could all pencil out if you had some massive incentive for production on those specific days and some legally mandated quota for other times, but at that point you'd probably be beaten by a few factories saying they'll shut down for a few days if you gave them the same cash and/or renewables producers using the money to add storage.

Practically, it needs to run full out constantly and compete with an average of other power producer prices to be viable.

Energy storage is not all that cheap.

There's a reason I mentioned "cloudy winter days." The cheapest way to address those is probably to overbuild PV. That overcapacity probably won't be used on bright summer days, raising the capital cost of PV across the board.

I don't think you're going to find many factories that can be economically shut down for entire seasons. None of this is an issue right now because we use natural gas for backup, but we need to stop doing that.

Hydrogen burned in combustion turbines would likely be cheaper. For this use case minimizing capital cost is all important; the cost of the hydrogen itself much less so.

The cheapest way to generate power on cloudy winter days is wind power, not overbuilt solar. Though, yes, overbuilding wind and solar is generally more cost effective than most other alternatives and the overbuilt wind and solar will both contribute power even when not working at their seasonal peak and provide abundant cheap power for storage when overproducing near their peaks.

I had assumed we were both taking that as the baseline alternative, since every country in the world is basically building that out right now, hence my suggestion that demand response would be a better choice than any plausible nuclear option to cover any gaps in that provision due to unseasonal weather which is both less windy and more cloudy than predicted.

But we appear to be starting from radically different assumptions about what power grids will look like (and already look like today)

Yes, we have a lot of wind power, but that also is backed by fossil.

The assumption that "e a fusion plant had to be ten times as big and therefore likely ten times as costly — as a present-day fission plant to produce the same amount of power" is unreasonable.

Fusion has a lower power density of the reactor itself than a fission reactor, but in terms of size fission reactors are extremely tiny. Nuclear power plants are big because you have a massive containment building around the reactor, and infrastructure both for handling radioactive materials and generating power. Fusion plants don't need the giant containment building and the various additional facilities would be nearly identical for the same level of power production.

Further, costs are not a simple function of size - things much bigger than fission reactors can be built for much cheaper; the problem is that the combination of safety regulations, delays due to public pushback, a lack of standardization, and the loss of a skilled workforce have skyrocketed the price of fission reactors far beyond what the simple engineering considerations would predict. Fusion does not carry fission's stigma, so it should suffer less from excessive regulations and NIMBYism, and engineers can take lessons learned from the history of fission reactors to design fusion reactors that are easily replicated.

If a fusion plant is just like a fission plant, but replaces the fission reactor with a fusion reactor, then it is entirely reasonable to compare the cost of the reactors.

That a fission reactor itself is a small part of the cost of a fission plant doesn't mean the same must be true of a fusion plant. And indeed, if you look at the cost of conceptual DT fusion power plants the reactor is a significant part of the total cost of the plant.

You are right that cost is not JUST a function of size. It's also a function of how exotic the materials are and how intricate the device is. By those metrics, fusion will do even worse. A fission reactor is a rather simple thing, in comparison.

Fusion's costs will be further increased by reliability concerns. The part of a nuclear plant that's too radioactive for hands-on maintenance must be extremely reliable. In a fission power plant, this part is rather small and simple. Multiple fuel rods in a fission plant can leak without necessarily shutting down the plant; a single leak of coolant into the vacuum vessel of a fusion plant will likely prevent it from operating.

Fusion power plants will almost certainly need containment buildings. The cryogens of ITER, for example, would (if fully vaporized) present a larger pressure x volume load than the steam from a fission reactor meltdown. Containing this gas is not cheap. In any case, tritium must be kept from leaking, which will imply expensive hermetically sealed buildings and seals (tritium will permeate through polymer seals.) Tritium will be everywhere inside the fusion reactor building. The tritium that cycles through a 1 GW(e) DT fusion reactor in 1 year is enough to contaminate two months of the flow of the Mississippi River above the legal limit for drinking water. Even small levels of leakage will be extremely vexing.

For ARC specifically, the magnets are shielded by titanium hydride. This material will fully decompose to titanium and hydrogen at the temperature of the molten salt, so it must be assumed that in a serious accident it will all decompose.

It's not unreasonable to compare them, but that comparrison must be made in context: we're saying that something that makes up a very tiny part of the cost will be more expensive, while something that makes up a huge part of the cost will be dramatically less expensive.

Instead of a $100 million reactor, you're looking at $1 billion in reactor spending, but instead of a $4 billion dollar plant that this reactor goes into, you're looking at a $2 Billion plant.

A fusion plant would be comparable to a very expensive fission reactor in a world where people weren't afraid of fission plants, but in that world a fission plant would be dirt cheap. In the real world, fission is way more expensive than the engineering challenges would imply. It's not the materials or the containment that is expensive, it's having all of your assets sit idle for years on end while yet another environmental impact study is conducted.

Also, some of your assumptions are unreasonable. For example the reason you need a containment building around a nuclear reactor is that you can't just vent to atmosphere, because the water contains large amounts of tritium. It's perfectly fine to just vent helium to atmosphere in case of an emergency as it's not radioactive. While a fusion reactor would use a lot of tritium over time, at any given moment the amount present is rather miniscule, it is being actively generated on site specifically and if anything the major technical issue is not having enough. A reactor the size of ITER would have approximately 0.6 g of Tritium in the reactor at any given time, losing all of that to atmosphere would be equivalent to approximately 2% of the annual tritium release from The Hague Nuclear Reprocessing plant. Decomposition of titanium hydride at the temperature of the molten salt is slow, while obviously undesirable, there is no danger in the magnets decomposing, the real issue is quenching, which is one of the few genuine safety concerns of a fusion reactor.

Wait, the idea that the cost of a power plant structure is proportional to its size seems very remarkable to me even within a single technology like light water reactors or gas turbines. My understanding is that generally there's some most efficient size to a reactor or turbine or such because of non-linearities and if you want to increase the power generation of a plant you replicate these most efficiently sized structures rather than scaling them up.

The idea that you could apply the same linear scale to both fusion and fission reactors seems frankly incredible on the face of it. Do you have any details on why this should be so? I don't seem to have access to the second source you listed and the first just made this assertion without explanation. All this isn't to say that I'm sure fusion reactors would have to be less expensive per cubic meter than fission reactors, the opposite seems like it could be a possibility. It's just the idea that we should expect the price to be the same in both cases that I'm finding hard to swallow.

From Lidsky:

>Fusion will almost certainly have a lower power density than fission and therefore will require a larger plant to produce the same output. Suppose a fusion plant had to be ten times as big and therefore likely ten times as costly — as a present-day fission plant to produce the same amount of power.

Fission is currently not cost-competitive due to the expense of ensuring that fission reactors do not pose an unacceptable risk of radioactive contamination in their vicinity. However, fusion is not subject to this constraint, or anyway suffers from it much less. There are no long-lived radioactive byproducts, and judicious selection of the construction materials (already implemented) can ensure that neutron activation of the walls is not a problem either. Furthermore, the inherently unfavorable nature of fusion reactions mean that criticality accidents ('meltdowns'; a la Chernobyl) are not possible.

From Pfirsch and Schmitter:

>It is shown that the claims made therein for the economic prospects of pure fusion with tokamaks, when discussed on the basis of the present-day technology, do not stand up to critical examination.

The analysis in the fulltext relies on a variety of plasma parameters estimated based on technology available in 1987. I cannot immediately determine if it generalizes to designs using HTS, but the comments on pp 1473-4 about the achievable B field strengths and corresponding betas suggests that they do not. Cf. this paragraph:

>Another possibility is to use higher magnetic fields: 6 T instead of 5 T would increase fw to values between 1.3 and 2.0 MW/m2, which are still very low. The latter comes close to the value of 3 MW/m2 obtained in Sec. IV.A.l from thermal wall load constraints. Higher fields would, of course, again increase the cost.

Overall I don't think that these links provide nearly as strong an argument as you suggest they do.

Fission plants are expensive for a couple of reasons. One is that they need additional layers of heat exchangers. Another is that their parts must be very reliable, to reduce the probability of serious accidents.

Fusion reactors will also require layers of heat exchangers, to isolate the tritium. They will also require very reliable parts: not because of public safety, but because fusion reactors will have so many parts in the hot area where hands-on maintenance is impossible. And this reliability will be expensive, even though the requirement for it is more to avoid a financial meltdown rather than a physical one.

> The analysis in the fulltext relies on a variety of plasma parameters estimated based on technology available in 1987.

The point of these arguments is that beyond a certain power, the plasma parameters become irrelevant. The limit is imposed by what the first wall can withstand, not what the plasma can put out.

If you look at areal power densities of fusion reactor concepts, the older studies had HIGHER areal power densities. But those higher power densities were found to be unrealistic.

Lidsky concluded DT fusion reactors would be an order of magnitude worse (in volumetric power density) compared to fission reactors. In this, he was being too generous: ARC is 40x worse than a PWR; ITER is 400x worse (and DEMO almost as bad).

The arguments there were farseeing, and experience since then has buttressed them, not contradicted them.

>Fusion reactors will also require layers of heat exchangers, to isolate the tritium. They will also require very reliable parts: not because of public safety, but because fusion reactors will have so many parts in the hot area where hands-on maintenance is impossible. And this reliability will be expensive, even though the requirement for it is more to avoid a financial meltdown rather than a physical one.

All of this applies to fission. Radiation equipment is expensive, period. We pay four figures for a block of plastic. A very, very accurate piece of plastic. The components in a fission reactor are not easy to replace either; the cost of a fusion meltdown is the reactor, while the cost of a fission meltdown is the reactor + up to several square miles of the area around it, the latter being so large that we typically ignore the very expensive reactor cost!

But more simply, you're underestimating concrete. Fission facilities are critically dependent on the stuff, wall after wall, being the only material that can be assembled thick enough to guarantee the safety of radiation workers who sit in the plant all day. Lowering the intrinsic radiation burden reduces the use of concrete, which is one of the most expensive parts of nuclear plant construction:

https://www.forbes.com/sites/jeffmcmahon/2018/10/02/4-ways-t...

While some of this applies to fusion reactors, it doesn't seem appropriate to compare only the power-generating components of fusion vs. fission reactors while ignoring the safety components when the primary advantage of fusion is safety. Regardless, I've made a note to read more about it.

>Lidsky concluded DT fusion reactors would be an order of magnitude worse (in volumetric power density) compared to fission reactors. In this, he was being too generous: ARC is 40x worse than a PWR; ITER is 400x worse (and DEMO almost as bad).

If the real power densities are available, arguments about the theoretical power density are irrelevant. I can probably build a warehouse ten times the size of a nuclear reactor for a tenth the cost of said reactor. ARC's true power density -- or that of any other reactor -- must obviously be factored into any cost projections. The power density of a particular design is not usually something you need to read a paper about!

A fission plant can be extremely cheap in certain environments. For example, in the cloud tops of Venus, Saturn, Uranus, Neptune, Ganymede, Titan, or Triton, a fission reactor is as simple as a big fabric tube suspended from a balloon, with a naked atomic pile hanging near the bottom, and a wind turbine at the top. Radiation is absorbed by the air inside the (sufficiently broad) tube, which rises through the turbine at the top. You could dispense with the balloon if you constrict the exit aperture just right.

Oops, not Ganymede or Triton. Not enough atmosphere.

But the planets beyond Jupiter have surprisingly gentle gravity.

Could they replace the beryllium with lead? It multiplies neutrons the same way; last I saw, that's what General Fusion was planning to use.

Molten metal flowing past metal structures in their high magnetic field would be a non-starter, I think, due to induced currents and JxB forces.

Does the metal have to flow? Let it sit there and run cooling pipes through it. Every now and then turn off the fusion when you need to fire up the pumps and swap in new lead/lithium.

(Also, I'm dumb but beryllium is also a metal, how does it differ from lead in this respect?)

The ARC design immerses the vacuum vessel in a bath of molten salt (which is where the Be is, in lithium beryllium fluoride (FLiBe) salt). That salt is where the neutrons deposit their heat. Replacing the Be with lead means the heat is getting deposited in that lead (or, more likely, molten lead-lithium alloy).

Even though ARC uses salt, it would also have to worry about voltages induced by flow across magnetic field lines -- not because of currents, but because if the voltage becomes high enough it can induce electrochemical reactions, like production of elemental fluorine (or corrosion of metal where the fluorine would have been evolved.) I think they keep the velocity x coolant channel diameter low enough to avoid that, but it's still a consideration they have to address.

SPARC is great, but I'm under the impression that a gen 1 DT MCF reactor will almost certainly need to be a stellarator to work around the engineering challenges and pulsed nature of tokamaks. Optimization, HTS magnets, and clever coil winding enable them. Tokamaks are easier so SPARC should certainly be made to make its splash.

The real problem is that DoE's Office of Science is relatively reducing funding of Fusion Energy Science. It's barely enough to meet the US' ITER contribution. The very few existing projects are running on fumes. No one in the US is making a machine and hasn't been for over a decade.

https://www.energy.gov/sites/default/files/2021-05/doe-fy202...

People are making machines, it's just with private funding. This includes SPARC, which MIT spun off into Commonwealth Energy. As of a year ago they'd raised over $200M.

https://techcrunch.com/2020/05/26/with-84-million-in-new-cas...

The context is US publicly funded projects.

The reason I used this context is because fusion is not profitable, won't be for at least 30 years in the optimistic estimates, and may very well never be profitable. It is exactly the kind of thing that should be public works.

Maybe it should, but since the government is not doing it, we're lucky that investors disagree with your assessment. CE, Tokamak Energy, Tri Alpha, General Fusion, and Helion have all gotten substantial private funding. Tri Alpha was over $700M last I checked. One of Helion's investors is YCombinator.

It's a matter of perspective and wager. I wager that the public image cost of failed startups leads to a reduced likelihood that fusion will be properly funded in the next 100 years. Fusion already has a public image deficit to overcome.

One could be optimistic and say the few potential successful startups such as SPARC or potentially successful moonshots such as Helion will lead to more private investors and/or public funding, but it's a community betting its public image when it's already down. I don't have a safer alternative to suggest.

Not just a public-image deficit. The $billions already poured down that rathole would take decades for the first fusion plant to pay back, if it had to, before ever achieving the break-even that actually counts. Especially so, when running it only at night after cloudy days when the much cheaper wind, solar, and storage flag.

If there was a definite solution it would be funded unless all the scientists involved lack communication skills to raise capital. VCs spend millions on apps that say ‘yo’…

Reply on news.ycombinator.com