Comment by 317070 | Hacker News Reader

Comment by 317070 | Hacker News Reader

Parent Full thread

Fusion happens when hot enough hydrogen ions (=hydrogen atoms with the electron stripped so only the nucleus is left) are running into each other often enough.

So you need a high enough density of those that stay hot enough for long enough (the triple product, vertical axis) at a high enough temperature (horizontal axis) for enough atoms to fuse.

At the boundary, enough energy will start being released from fusion reactions to keep the reaction going. at Q=1 you have theoretical break-even. Somewhere between Q=5 and Q=10 you could extract more heat than you put in. The heat produced by the reactions should be enough to sustain the reaction. At Q=inf it just keeps burning like a fancy camp fire. Just throw in fuel, remove ashes and enjoy the heat. Commercial fusion can only start being viable probably somewhere between Q=10 and Q=20.

Do note that the projected future in OP's gif should contain way more attempts from various companies to reach that limit. Just mentioning sparc is a bit selective.

Commercial fusion will not be competitively viable at any Q. Even unlimited free heat is not valuable if it costs too much to make power from it. A volcano, earthquake, or hurricane releases many TWh of energy, but not usefully so, for reasons.

Once you start to get enough neutron kinetic energy out, those neutrons have to be captured and their kinetic energy degraded to heat in thousands of tons of molten lithium pumped in big pipes snaking around inside your magnets, which you must then use to boil water and drive a turbine.

The cost of operating such a plant would be at least 10x the equivalent fission plant. But fission is already not competitive, and gets less so every day.

Part of operation would need to be purifying grams of tritium, daily, out of those thousands of tons of radioactive molten lithium, for the fuel needed to continue operating. That might not be possible. No one is even trying it yet.

The power density of your D-T plasma would be much less than fissioning uranium (Th, Pu, etc.), which means you need a great deal of it, and a huge plant. If superconducting magnets squeeze it smaller, you have the problem that all this neutron flux has to come through a wall of limited area, destroying it in short order. Magnets so strong would need thousands of tons of steel to hold them in place, which would weaken rapidly under heavy hot neutron bombardment.

So, the D-T fusion chased is a series of fascinatingly hard technical problems, but there is no plausible prospect of ever getting any commercially valuable power out. And, they are not even safer than a regular nuke; the thousand tons of molten radioactive lithium is hugely inflammable, even explosive, and the smoke turns into radioactive drain opener on contact with anything damp, e.g. lungs.

Are you suggesting that all those people doing fusion research don’t know about this, or that they make false promises to the public to fund their research?

At least many know and do not care. Many deceive themselves, as often seen here. Others don't think about it, considering it Somebody Else's Problem. Promises of commercial power are naked lies they will never be called to account for.

Probably a mix of both, and other stuff. For an example of other stuff, look at the inertial confinement fusion reactors. Those are only plausibly useful for fusion weapons research (as fusion-based nuclear weapons work on the same principle, but are somewhat harder to study up close).

However, that's unfashionable to talk about, so they are touted as a potential source of energy. Given that a single shot costs many tens of millions of dollars (because of the incredibly precisely machined piece of gold, called a target, that gets heated up by lasers until it emits X rays that cause the fuel pellet to implode in a perfectly symmetrical fashion), and is over in nanoseconds. So, a continously operating plant would literally spend millions of dollars per minute even with optimistic improvements in the cost of the targets - not exactly a promising new "free energy" generation technology.

Effectively harnessing volcanos and earthquakes really should be a milestone on our path to Kardashev status. Elon, if you are reading this, we need drone swarms that can eat and store explosive geoenergy release.

> A volcano, earthquake, or hurricane releases many TWh of energy, but not usefully so, for reasons.

/s

> The cost of operating such a plant would be at least 10x the equivalent fission plant. But fission is already not competitive, and gets less so every day.

People keep repeating this (fission is not competitive) like it's a law of nature, but the reason they are not competitive are entirely human-made.

They don't repeat it "like a law of nature" but as an empirical description of our technological capacity. If there is a large advancement of construction productivity that may make nuclear more cost effective, it would be good to know about, but nobody who claims this is possible seems to have some sort of methodological description of this advancement. Similarly, those that claim that it's just regulations getting in the way of nuclear don't seem to have an answer for why Flamanville in France has issues, or have the specific change of regulations that will alleviate the construction issues that we regularly see in the Western Hemisphere.

Personally, I believe an advancement in construction productivity is quite possible, but that it will take an entirely new generation of bright, ambitious, and innovative individuals entering the field, as well as compensation to attract them from other fields that currently outpay construction by quite a bit. But that's a loooong and expensive change of economies.

Here is a peer-reviewed global survey of nuclear reactor costs: https://www.sciencedirect.com/science/article/pii/S030142151...

For a quick overview, skip down to figure 12, which shows a plot of overnight construction costs across time and countries. Note that the bulk of them are below $3000/kW, quite a bit less than the most expensive reactors in the US.

So here's an optimizer that finds the lowest overall grid cost, based on various assumptions you can modify: https://model.energy/

Pick a country. "Show advanced assumption settings." Check the box for "dispatchable technology 2." Set the overnight cost down to $3000. The lifetime of 25 years is unrealistic, given that the average age of US reactors is 40 years. Set that to a more accurate value, like 60 years. Finally, give it the same discount rate as everything else.

You'll find that in many countries, nuclear does quite well. In countries without good wind/solar resources, even a higher capital cost can result in a 100% nuclear grid.

That paper is about historical costs, which is not informative for current costs.

Saying 60 years is "more accurate" is kind of putting one's thumb on the scale, as is putting an overnight cost at $3000/kW. These sorts of overly rosy and unrealistically optimistic assumptions are how we get into such terrible estimates for nuclear.

Even though MIT has been one of the sources of inaccurate nuclear optimism in the past, I think that this is perhaps the best assessment of how to make nuclear competitive again:

https://energy.mit.edu/research/future-nuclear-power/

I still thinks it's somewhat overly optimistic, but it at least tried to address the real and deep issues within the industry instead of hand waving them away.

Until nuclear proponents both acknowledge the full reality of construction, in the 2020s, and the errors of the past, they will never be able to get to $10k/kW or even $5k/kW. A future at $3k/kW in the US is not worth even contemplating until we could reliably even hit $5k/kW, and a world with $5k/kW is currently not worth contemplating because we are so far away from that.

If it's been done before, then we don't need a "large advancement of construction productivity" to do it again. If, as the paper shows, other countries have managed lower costs recently, then we should look at what they're doing. If the average lifespan of our existing nuclear reactors is 40 years, then setting a lifespan of more than 40 years is indeed more accurate than setting it to 25 years.

We live in a different technological setting than we did in the past, and all sorts of skilled labor that was economical in the past is no longer economical. As parts of the economy become more productive, labor costs rise, and the less productive parts of the economy become more expensive in comparison.

We see this cost disease in all sorts of large construction projects, not just nuclear. Though nuclear seems to be particularly bad. And that's why we need breakthrough in productivity. Personally I think that such a construction productivity breakthrough would have its greatest impact on decarbonization through cheaper subway construction, high speed rail construction, and through increased urban density with cheaper big buildings. These type of construction projects have few realistic alternatives, whereas renewables plus storage are already a great way to power an energy system.

So if we can fix companies like Bechtel, we will see lots of improvement throughout our society far beyond nuclear construction.

Yet there's Japan, with low nuclear costs within the past couple decades. Were their skilled labor costs all that low?

High cost of large projects is certainly a general problem in the US, but it may not be due to labor costs alone. Here's a Bloomberg article that attributes it to "over-design, inefficient project management and misaligned politics." https://www.bloomberg.com/news/articles/2021-12-08/why-build...

I think it's entirely possible that in the US, where we have copious wind and solar resources and we've gotten bad at doing large projects economically, nuclear is not the best option. That says little about the rest of the world, though. And it's possible that small modular reactors would help even in the US, because while we suck at big infrastructure projects, we're not too bad at high-volume production in factories.

And that brings us back to fusion, because a lot of modern fusion designs would be more like factory-produced modular devices rather than large site-built infrastructure projects.

$3/W capex, even if it could be achieved, is well north of the current cost for renewables. It anyway fails to address extraction cost, for which there is no plausible means to match renewables' near zero cost.

By the time any new nukes could be brought online, capex and opex for renewables will have fallen far, far below even their most aspirational projections.

> advancement in construction productivity ... will take an entirely new generation of bright, ambitious, and innovative individuals

Lack of those is not the cause of construction cost overruns, so increasing their supply would not help. Cost overruns are a deliberate consequence of wholly-legal institutional corruption, which no one shows any appetite for rooting out. Ingesting exemplary individuals just generates disillusioned individuals out the other end.

So what you're telling me is that fusion is 20 years away? /s

"How about never? Is never good for you?"

> A volcano, earthquake, or hurricane releases many TWh of energy, but not usefully so, for reasons.

I don't quite follow your point here. Presumably the issue with earthquakes and hurricanes is that they aren't dispatchable and the work environment for operating a plant is quite hazardous. The issue with volcanoes (although I'd bet we already do use volcanoes for power) could very easily be as simple as that they tend to be in difficult to reach locations because of the plate tectonics (ie, mountains) & the hot parts take a lot of drilling to get to.

If there is energy, we can find a use for it.

We do not try to capture and bank energy from more or less reliably annual hurricanes because we are not fucking insane. They are rightly classed as natural disasters, not resources. Nukes will rightly come to be seen much the same way, a phenomenon to hope people elsewhere suffer, if anyone must.

That isn't an argument or a reason. The problem is they aren't reliably available. If hurricanes were an ongoing phenomenon we would harvest the energy - like we do for wind. Using volcanoes for power is absolutely on the cards. Apparently the hard part is finding the magma chambers [0].

The barriers are entirely practical in that these power sources don't scale well and aren't easy to get on demand. None of the problems there are similar to having a tokamak - it isn't like we are going to struggle to find the tokamak (it will be marked on the map with something like "tokamak here") and it will likely have a big on/off switch.

[0] https://www.science.org/content/article/could-volcanoes-powe...

We reliably get multiple hurricanes every year, each releasing far more energy than we could use. We do not, in fact, extract that energy.

We also never, ever drill into magma chambers. Drilling into a magma chamber is a good way to get a new, vitrified surface, several feet deep, for miles around what was once your drilling rig.

> Magnets so strong would need thousands of tons of steel to hold them in place, which would weaken rapidly under heavy hot neutron bombardment.

I'm just going to nitpick this part. In ARC, the magnets go outside the blanket, which is a molten salt, fully surrounding the inner wall. The salt has to absorb almost all the neutrons, to generate enough tritium. The steel outside of that shouldn't be getting many neutrons.

The ARC proposal argued for a cost per kWh of 26 cents, for a pilot plant not necessarily optimized to full commercial configuration. This requires a fairly revolutionary further advance for general commercial relevance but note that the current cost for ratepayers in Germany is 32 cents. A much cheaper or much stronger structural material than steel 216 would quickly change the cost as the strength determines the operating magnetic field which increases output to the fourth power. More interestingly there appears to be a similar revolution newly predicted. Based on a first principles simulation it was projected that density can be doubled over current limits in a reactor as powerful as ARC. I believe density translates directly to pressure which increases output to the second power - possibly quadrupling output over what was expected. Roughly speaking this should mean a cost of less than 10 cents for ARC if that pans out. Hinkley Point C is guaranteed a price of 14 cents, which is the world average for electricity. The doubled density prediction appears to bring a conventional tokamak in line with the projected costs of various alternative designs such as General Fusion's 8 cent projection. None of the alternatives have been tested at the scale of tokamaks but there are a few with the potential of a revolutionary price - Helion projects 1-6 cents based on reactor survivability, and don't use tritium. In fact they may be able to produce it fairly efficiently if they fall short of standalone electricity production. Fusion is generally a more expensive, more of a longshot competitor to SMRs, but with the very dramatic advantage of zero meltdown risk that makes it worth a shot

Cost to extract power does not scale with power density of plasma.

No optimistic projections for cost of extraction are better than idle speculation. Practicality of extracting diffuse 3H from a molten metal blanket, needed for fuel, is as previously noted purely speculative.

Renewables are driving cost per kWh well below any plausible, or even aspirational, figures.

The only fusion technology that has any chance of competitive extraction cost is Helion's, which does not depend on a round trip through heat. But it depends on a regular supply of 3He, a decay product of 3H which must be produced via operation, filtered from the working plasma, and then held for its 12y decay half life. It is anyway not clear if its process can be achieved at all.

SMRs depend on the round trip through heat and incur all the other costs of traditional nukes, so offer no prospect of keeping up with plummeting cost of renewables.

Fission's fuel rod meltdown is not the only catastrophic failure mode available. Fukushima demonstrated hydrogen detonation. Breached containment of thousand-ton inflammable, caustic, radioactive molten metal neutron absorption blanket would be adequately catastrophic.

Would it be somewhat correct to say that the Sun (or generally stars of similar type) is approaching that Q=inf at this stage of its life?

Stars are at Q=inf, they don't have external heating. They are also on a weird place on this plot, considerably more to the top left side of the high Q's.

Had to have a second look at the chart after reading your explanation to understand the variables. Thank you!

That's a great explanation, thank you. What does the black curve that intersects some of the Q curves represent?

It's the theoretical breakeven event (Q=1) when you use high power lasers rather than the magnetic confinement of a tokamak. The NIF is doing research on this inertial confinement fusion, where they shoot a lot of power with lasers on a tiny pellet with hydrogen. So you get extreme densities and temperatures, but not a lot of confinement time, i.e. the ions cool down _really_ fast. That shifts the curve a little bit.

In my opinion, it is a bit misleading to add this to the plot, because for this laser-based fusion, the theoretical breakeven lies really far from the practical, commercial breakeven. But as shown there on the gif, the NIF did manage to cross the Q=1 for inertial confinement fusion last year!

NIF is nothing other than dressed-up weapons research. But it is not more deceptive than other methods.

Author here. The black curve represents the hot-spot ignition condition for a laser inertial confinement fusion (ICF) experiment (like the NIF). This means that during the short period of inertial confinement, the self-heating exceeds all losses in the hot spot leading to an increase in temperature due to self heating. It only applies to the black 'x' points.

The Q_sci^MCF contours correspond to scientific energy gain (ratio of fusion power to heating power crossing the vacuum vessel boundary) for a magnetic confinement experiment.

For ICF we can't draw simillar Q_sci^ICF contours because the total fusion energy released depends on the degree to which the ignited hot-spot propagates a burn in the surrounding cold fuel. And this depends on other variables like the symmetry of the implosion which are not captured in this plot.

If you're curious to read more about this check out Section III.F of the linked paper (pp.10-11).

So does eg Q=100 mean that it'll still go out on its own at some point? Ie is Q a measure of duration/total energy output for a given energy expenditure?

Yes, it means you will get out 100 times more energy than you put in, but at that point it does still go out if you don't keep supplying that external heat.

At Q=infinity, you get out as much energy as you want compared to what you put in initially, but it keeps burning (as long as you add fuel and remove ashes, that is). At that point you pay an initial start-up cost and can fuse as much as you want.

I see, thanks! So this doesn't take into account the hydrogen, just the external energy/heat you put in? I guess it must, otherwise Q=inf would be free energy.

Correct. The energy balance comes from the fuel mass that is converted to energy with E=mc^2.

Right, thanks for the explanation!

You still have to run the cryogenic system for the superconductors and the pittance of power to keep the confinement coil power supplies. Q=inf means ignition: an entirely self-heated plasma. The collisions of the helium ash on the un-burnt fuel are enough to bring the fuel above the flash point. In this way it is like a campfire.

To keep operating you need to extract heat, make power, and drive it from that power, or it goes out or blows radioactive burning molten lithium all over.

In practice, no such plant will ever exist, because there is no plausible prospect of the power being competitively priced.

In the same way there was no plausible prospect of launching cargo to orbit for <$1000/kg at the turn of the century?

No, in the sense that the costs for the competition, renewables + storage, are still falling at a reliably exponential rate, leaving everything else far, far behind. There is no reasonable prospect of catching up even ignoring capital cost and just comparing operating expense. The cost of maintaining steam turbines will not be coming down at all, never mind going to zero.

Reply on news.ycombinator.com