Hacker News Reader

Comment by andy_ppp | Hacker News Reader

So their main advancement is stronger magnets with "rare-earth barium copper oxide (ReBCO) on metal tape" being wound into extremely tight loops (and is presumably extremely thin).

Will we enter into a tipping point of materials science that allows magnets strong enough and suddenly we get fusion and it becomes ever better as we make better superconducting magnets?

View on HN

Fusion seems like it will advance the fastest through (a) materials science unlocking better magnets and (b) simulation and physical experimentation with reactor designs, including highly experimental designs.

Computational modeling seems to be helping as well:

https://ai.googleblog.com/2017/07/so-there-i-was-firing-mega...

Disclaimer: I am not a plasma physicist.

That's pretty much what's happening here, fusion output increases with the fourth power of the magnetic field. Double the field, 16X the output. But if we take it much further, we'll reach a point where the limit is the structural strength of the reactor.

A few years ago I got to tour MIT's Alcator C-Mod, which had the most powerful field of any tokamak to date. A grad student showed us a metal tie rod, about a meter long, and said they'd calculated that two of them could hold down the Space Shuttle while it was trying to launch. To hold the reactor together while it was operating took 38 of those.

The peak on-coil magnetic field in the ARC design (from the 2014 paper) is 23 Tesla. The pressure of a 23 T magnetic field is 2100 bar, nearly twice the pressure at the bottom of the Marianas trench, and comparable to the chamber pressure of a handgun. 60% of the mass of their design was the stainless steel structure needed to support the magnets against the outward JxB forces.

Fusion's prospects would likely be helped some by plasma configurations with much higher beta than tokamaks. At least the experiments would be cheaper.

"Suddenly", no. They're targeting 2025 for the first tests if everything goes perfectly, 2027 for a first demo reactor, and we don't know how expensive a real reactor would be so we have no idea if this is the design that will one day unlock UNLIMITED POWER!

but on the scale of civilizations, yeah this could be it.

4 years in the fusion field is pretty "suddenly".

Why is that fusion reminds me of the James Webb space telescope .... https://xkcd.com/2014/

Fusion at this point is closer to general intelligence AI than to JWST.

JWST is a project level failure not a “field level” failure.

Fusion has been 10 years off for the past nearly 100 years.

Tokamaks date to the 50’s and the first patents for a fusion reactor were issued in the 1940’s https://worldwide.espacenet.com/publicationDetails/biblio?CC...

Uh, no. We are fairly sure that we can build a working fusion plant and we know with good accuracy what remaining engineering problems are. We have literally no idea what the path to AI even looks like.

Energy positive fusion? I’m not so sure that we do...

We already have working AI, not prototypes. Look at big corporations for example.

@dogma1138 said general AI, what we have in big corporations is specialist AI.

A general AI must, by definition and at minimum, be capable of doing any intellectual task currently done by humans; right now, they’re good at what they do, but are very limited in what they do, and slow to learn. For example, self driving cars can still only do limited environments and conditions, despite Tesla having over 18,000 human-professional-equivalent years of driving experience as of April last year, which they achieved shortly after they added “Traffic Light and Stop Sign Control” to their feature list — and that item is still listed as “(Beta)”.

It’s the same everywhere: GPT-3 is fantastic… but despite having “read” more than any human could in a lifetime, it still struggles with moderate arithmetic; Voice assistants do amazing things… but I’ve literally had easier times attempting to converse with pet dogs; Google Translate has made moving and travelling abroad much less stressful, and it knows more languages than I can name to a higher standard than I know a second language… but I’ve seen it hallucinate words in lawns, and it does make translation mistakes that even I can spot.

And even if they were perfected within their domains, none of these are generalists for all domains.

I'm not sure how much potential improvement you can expect there in the near future. There's a pretty large jump between the conventional superconductors and the new high-temperature superconductors used here, I'd suspect that it'll take a long time of incremental improvement to build stronger magnets with these new materials.

There's a lot of practical problems with building very high-field superconducting magnets. I'm also not sure how much you can gain from the thinnness of the material, conventional superconducting magnets have a lot of non-superconducting material in there as well to conduct heat so that the magnet isn't immediately destroyed on a quench.

> Will we enter into a tipping point of materials science that allows magnets strong enough and suddenly we get fusion and it becomes ever better as we make better superconducting magnets?

I think that's the idea. Iter is about as small as it could possibly be and still work given the magnet field strength they had designed around. With stronger magnets, we can make smaller reactors, which are cheaper to make and (if I understand correctly) have better power density. At some point it stops being practical to make it any smaller as the limits become "how thin can we make this shielding material?" or "how much heat energy can we remove by pumping fluids around?" And then once we've proven the concept and we've settled into an optimal size the engineering focus turns to "how cheaply can we manufacture this?" and "how can we reduce the total operating cost per megawatt hour?".

Iter and other big or old projects are going the safe route with magnets that need to be at 4 or so Kelvin. All these ‘hotter’ superconductors are relatively unproven (it’s a oversimplification for a broad category), especially their strength when made into the shapes that make the magnetic field and form part of the load bearing structure of the container. The winners would enter that tipping point.

It is worth pointing out that Iter started in 1985, I think before REBCO was discovered (and it was a little bit after that they acheived superconductivity at 77K, liquid nitrogen temp). Sturdy and flexible tape form from ion beam-assisted deposition I think was later but I can't find the date.

(edit: maybe in the lab in 1992? I'm not sure when it scaled production: https://en.wikipedia.org/wiki/Superconducting_wire#cite_note... )

Practical use of the new high-temperature superconductors in high field magnets is even more recent. NMR is probably the closest commercial application for this kind of magnet, and the first spectrometers using the new superconductors were sold last year.

The sandwich structure described in the article makes sense: separate the superconducting and the load-bearing functions into different layers.

the magnet singularity is coming

Quite possibly. The magnet as a key tool for harnessing the immense power of nuclear reactions is only just now coming to light, most recently when Dr. Indiana Jones survived a nuclear blast by hiding in a fridge covered by a variety of small magnets.

Reply on news.ycombinator.com