I was just reading about stellarators in the book "A Piece of the Sun" [1] about the history of fusion. I find it fascinating (and a bit worrying) they are still basically attempting to optimize the same two approaches that were developed in the 50s/60s, tokamak and stellarator.
I respect the scientists who continue to work on this. Many people have spent their whole careers on it and have died without making significant progress. It's inherently challenging because it's super expensive and working at super high temperatures. But the potential of the end goal might be enough to keep people working on it perpetually.
Fusion is simple — practically primitive — whereas plasma physics is insanely-mind-numbingly difficult.
To sustain a fusion process, plasma physics have to be tamed. That's the hard part. The Wendelstein 7-X, which the article is about, can't even do fusion, it's entirely about plasma physics.
(You can make a kind of shitty version of this in your garage with two microwaves, some thick wire and a healthy dose of personal risk: https://en.wikipedia.org/wiki/Fusor )
There's the Dense Plasma Focus, one version of which is being developed by Lawrenceville Plasma Physics. They're cool dudes, I once donated $10 to help them get a new Beryllium cathode about three years ago, they still send me periodic and detailed updates on their progress. You can read about their approach here, complete with radical 90's theme and pretty good video: http://lppfusion.com/fusion-power/dpf-device/
Sandia National Laboratories is using a massive Marx generator [1] to power the Z-Machine, a Z-pinch that uses powerful magnetic fields generated by high current to squeeze matter in the middle of a conductor: http://www.sandia.gov/z-machine/research/fusion.html
So there's a few different ideas out there. However, if you really want to know why no significant progress has been made[2] you need only look at this graph:
Note that it has a citation at the bottom, you should follow it and read the article to see that this isn't a bullshit internet chart. In the US at least, fusion has been consistently funded at a rate that the US government /knows/ is too low to see significant progress. You're welcome to make your own guesses on why this is the case.
> (You can make a kind of shitty version of this in your garage with two microwaves, some thick wire and a healthy dose of personal risk: https://en.wikipedia.org/wiki/Fusor )
This reminded me of that guy who tried splitting atoms in his apartment in Sweden in 2011.
Turns out last year in 2016 the police came back again because someone told them off he had Uranium. They shut down the whole neighbourhood and didn't let him come home for a full day while they searched it with a bomb squad and radiation team:
People have different risk tolerances for work and for home. Whether it invokes the bogeyman of radiation or nuclear reactions can be irrelevant. If some guy in an apartment next to mine is making a flamethrower inside, I might be a bit peeved. An accident might have consequences that might go far beyond their own property and well-being.
Does Uranium in the manner he was using it constitute a threat? I don't know, but it sounds like not. Does that mean the person reporting it, or the police responding knew that?
Trying to build a reactor in your home without asking for permission first does make further accusations about it significantly more plausible. He should probably have invited SSM over when he got his uranium glas, or at least told them by mail. But mostly he should have found some physics dude at technical museum or high school to play with instead!
I've also been following EMC2 and polywell ever since Bussards google talk. Last I heard Dr Park and team had verified whiffle ball confinement (i.e. That what Bussards hand waved was really happening) and were seeking funding for a large machine. Does anyone have more info on polywell?
>would it be easier to create a fusion reactor if we made it bigger?
Absolutely-but only when you approach mass scales on par with small suns. So not in a practical sense.
To elaborate, plasma in stellar bodies is essentially contained by gravity, as opposed to complex electroomagnetic fields that have to be painstakingly engineered and tend to be unstable.
A small black hole would probably glow very bright by itself, so it could be a useful energy source in orbit (for a while, until it crashes into Earth).
Odds are, it wouldn't crash "into" Earth. Rather, it would more likely crash "through" Earth.
Difference seems subtle, until you find out that a black hole (one made by us at least) is soooo tiny that it would miss most of the atoms in the Earth, falling _between_ them.
Apparently, you need lots of mass to make a black hole that is both sustainable and big enough to affect anything around it.
No ITER style reactors also get much easer when you scale them. Which is why ITER is larger than JET etc.
It's as simple as heat loss is on the surface, energy it released in the volume. You can somewhat get around this by increasing density, as eventually you need to slow down the reaction because your walls can only take so much heat and radiation. But, it's a lot more stable at lower density which means larger is vastly easer to control.
PS: We could have built an impractical, but working fusion reactor 30 years ago. But, the total cost would have been incredible and we only have so much tritium anyway.
> I find it fascinating (and a bit worrying) they are still basically attempting to optimize the same two approaches that were developed in the 50s/60s, tokamak and stellarator.
This would be a bit like worrying that aerospace engineers are still optimizing wings. Like wings, tokomaks/stellarators are the best tools for the job, as dictated by the underlying physics of the systems in question. Basically, donut-shaped magnetically-confined plasmas "leak" into themselves, rather than out into the world.
> Many people have spent their whole careers on it and have died without making significant progress.
This is a popular view that is entirely wrong. The figure of merit, which is the triple product of confinement time, density, and temperature, outpaced Moore's law right up until a gain of about 0.95, when the magnetic technology of the day caused the necessary size increase to put ITER in the realm of international cooperation. ITER absolutely will produce more power than it consumes. The plasma physics are just that well understood. The problem is size.
MIT have recently worked out that a new breed of superconducting magnets can more than double the available field strength, resulting in a 16-fold reduction in reactor size (due to a 4th power gain in confinement strength as a function of field strength). They hope to achieve a gain of about 2 with a university-scale reactor before ITER, designed with the magnets available at the time, is complete.
Most, if not all, of the above comes from https://www.youtube.com/watch?v=L0KuAx1COEk. ARC and SPARC, the MIT reactor concepts, are the most exciting thing I've heard about in...gosh, I guess my whole life. They could pull it off. If they do, we could save the planet with fusion-powered CO2 scrubbing. We could avert disaster.
You know, I had a sad thought about fusion-powered CO_2 and that was once it will be built, the energy is likely to be used not for CO_2 scrubbing but instead to mine Bitcoin.
That's interesting, I wish I ended the book with that type of optimism. It was a bit disheartening to hear about the many failed starts and the slow down of progress.
But that said it wasn't totally lacking optimism. It was more of the hard reality of how challenging the problem will be to solve.
I still follow the development of it and there is definitely incremental progress.
Thanks for think about MIT's work, that's interesting.
Wings work for what we want them to do, and have since they were first tried.
Stellarators don't.
Progress has been minimal.
The more so if one recognises the distinction between Texas Sharpshooting -- drawing targets around the holes we've made with various technical methods -- and the actual initial intent.
This particular problem has turned out to be vastly more difficult than anticipated or advertised. I've watched what little progress has been made over more than 40 years.
What held aircraft back was prime movers, fuel, and materials.
Once petrol-fueled engines and aluminium were available, the Wright beothers were flying within a few years of the Ford Model T. Monocoque airframes followed another few years later. The most perfect aircraft ever, the DC-3, which remains in active commercial use, was developed within the next 3 decades, largely thanks to aluminium alloys. Aivionics and wing design were largely solved by then.
With the gas turbine, jet aircraft appeared, and Boeing is still effectively building and selling scaled-up versions of its 707 airframe (1957). The USAF will be flying the very same aircraft for over 80 years, the B-52. First flight 1952, last date af manufature: 1962.
What progress has been made in aircraft over the last 65 years is largely limited to improved controls (both linkages and avionics), novel materials, and more recently, improved modeling duringg design and development. Outside military applications, that still has exceptionaally limited commercial impacts.
See various sources, though Robert Gordon's The Rise and Fall of American Growth has a good overview.
The contrast, particularly with man-years of effort and billions in spending, to fusion, couldn't be starker.
> Many people have spent their whole careers on it and have died without making significant progress.
I wonder... if there are entire domains of technology that simply -cannot- be achieved without AI – machines building and improving themselves.
We have already seen problems that simply cannot be solved within the lifetime of humans or in terrestrial environments, like telescopes in space vs. those on ground, or crossing long distances in "generation spaceships.."
you seem to mix the time it takes to bring a solution to the problem, with the invention of the solution itself.
For me, AI is just another tool on top of others. You can't build an AI to, say, control a plasma without a thorough understanding of plasma physics (acquired without the use of AI)... So I'd say AI may be useful, but not in the sense that it will bring a breakthrough...
It's occurred to me that AI might actually represent a form of nonscientific capability. That is, the ability to achieve results but without the understanding or structured knowledge that we consider to be part of science.
That's not saying that AI will solve the problem, just what the situation might be should it do so.
I'm still trying to wrap my head around what it actually does itself.
I watched this video called "MIT's Pathway to Fusion Energy (IAP 2017) - Zach Hartwig" which I thought was interesting (and which I submitted here two days ago but went unnoticed as it's the faith of many interesting HN submissions).
There Zach gives a quick overview (well hour long) of the current state of fusion and what its future is going to be. He didn't talk too much about stellarators but I feel I have now a better overall understanding of fusion physics thanks to him.
>"Let me make it very clear: fusion is, if you simply look at the carbon dioxide goals we have, fusion would be too late to bring those carbon dioxide emissions down at the rate that we need. If renewables do the job, fusion could become part of a network of dispatchable power generation units. However, if renewables don’t do the job, fusion will be too late to prevent serious damage. In that scenario, we would find ourselves in a bad situation for a period of time that extends beyond when the first fusion reactors come on line."
This quote is so telling on the limits of technology. So many people think of fusion as a technological miracle that can save the world. Here we have a man telling the world that we're fucked if we just sit on our asses waiting for this miracle.
Several options: pump it back down the oil wells it came from, use fast-growing juvenile trees to "pump" it into house frames and libraries, use algae to "pump" it into useful plastics, or, if you have infinite money, bottle it and shoot it at Mars to thicken the atmosphere.
There's a similar pipeline for food; the people who think about such things say we need to increase our food supply by 85% in the next 50 years. It takes about 35 years for a new piece of agricultural technology to go from the lab to ubiquitous, so we are really looking to have those technological advancements in the next ~15 years.
Maybe the problem that needs to be solved is the amount of time that it takes to bring an invention from "proof of concept" to "widespread use". In fact, maybe there has been significant progress on that particular issue just in the past couple of decades. I see new technology adoption curves that outpace anything we used to see.
There was a previous article on HN about the WX-7. It's an incredible piece of mechanical engineering. It's more symmetrical than it looks at first.
The thing has the thermal management problem from hell. The magnet coils are superconductors and have to be maintained down near liquid helium temperatures. The plasma is at a few million degrees. These are inches apart.
3D printing might help when the next one is built. Rocket engines have some of the same problems - they're plumbing parts filled with cooling hollows. Those take way too many joints and welds to build. The Space-X Dragon spacecraft has 3D printed engines, which gets the part count down.
If we ever get room temperature superconductors, these things will be much simpler. If we ever got high temperature superconductors, so that the whole thing could run hot, like a boiler, that might make it economically feasible.
If they have to keep the temperature down near liquid helium they must be using (and need) type I superconductors. If we get room temperature superconductors any time in the near future it will almost certainly be type II.
Turns out I was wrong. These applications use niobium-titanium or niobium-tin, both of which are type II superconductors. They do have relatively low Tc -- 10-20 kelvins versus >80 kelvin for the best HTSC, and are cooled even lower to maximize the critical field and current (i.e. maximum they can withstand before losing superconductivity).
If you speak German, this podcast is fantastic: https://alternativlos.org/36/. The sheer number of technological challenges in a fusion reactor is just crazy. Besides plasma physics you have to invent welding techniques, materials and a lot of other stuff. Everything is on or over the limit of what we can do now.
I wonder if you could make it in the form of a tree, where you took the leaves and wrapped them around to join the roots, like an inside out donut.
So in the middle you'd have your conjoined mega stream, and then that would flow upward, be separated out into filaments which then curl around in all directions down to the roots where they are re-merged into the megastream.
Probably too much turbulence at the branches, but maybe they can use vibration and learning to find beneficial harmonics. I would hire sound engineers.
I just have an intuition that a single stream will never have adequate confinement and you'll need redundancy in the geometry to get good concentration.
66 comments
[ 3.9 ms ] story [ 142 ms ] threadI respect the scientists who continue to work on this. Many people have spent their whole careers on it and have died without making significant progress. It's inherently challenging because it's super expensive and working at super high temperatures. But the potential of the end goal might be enough to keep people working on it perpetually.
[1] https://www.amazon.com/gp/aw/d/1468304933/
To sustain a fusion process, plasma physics have to be tamed. That's the hard part. The Wendelstein 7-X, which the article is about, can't even do fusion, it's entirely about plasma physics.
Working on it perpetually, you say? If only we could somehow harness that energy...
There's the polywell, my personal favorite, which uses a powerful electric field to confine the fuel: https://en.wikipedia.org/wiki/Polywell
(You can make a kind of shitty version of this in your garage with two microwaves, some thick wire and a healthy dose of personal risk: https://en.wikipedia.org/wiki/Fusor )
There's the Dense Plasma Focus, one version of which is being developed by Lawrenceville Plasma Physics. They're cool dudes, I once donated $10 to help them get a new Beryllium cathode about three years ago, they still send me periodic and detailed updates on their progress. You can read about their approach here, complete with radical 90's theme and pretty good video: http://lppfusion.com/fusion-power/dpf-device/
Sandia National Laboratories is using a massive Marx generator [1] to power the Z-Machine, a Z-pinch that uses powerful magnetic fields generated by high current to squeeze matter in the middle of a conductor: http://www.sandia.gov/z-machine/research/fusion.html
This approach is known as ICF, and is also being pursued by the National Ignition Facility, who use lasers instead: https://lasers.llnl.gov/science/icf/how-icf-works
So there's a few different ideas out there. However, if you really want to know why no significant progress has been made[2] you need only look at this graph:
https://imgur.com/sjH5r
Note that it has a citation at the bottom, you should follow it and read the article to see that this isn't a bullshit internet chart. In the US at least, fusion has been consistently funded at a rate that the US government /knows/ is too low to see significant progress. You're welcome to make your own guesses on why this is the case.
[1] My second favorite thing in HV equipment after the explosively pumped flux compression generator: https://en.wikipedia.org/wiki/Marx_generator https://en.wikipedia.org/wiki/Explosively_pumped_flux_compre...
[2] I'm not sure this is really the case. I think that the fusion research community has made admirable progress under trying circumstances.
This reminded me of that guy who tried splitting atoms in his apartment in Sweden in 2011.
Turns out last year in 2016 the police came back again because someone told them off he had Uranium. They shut down the whole neighbourhood and didn't let him come home for a full day while they searched it with a bomb squad and radiation team:
https://richardsreactor.blogspot.ca/2016/04/20160401-and-aga...
Risky stuff to do in this age of over reaction.
People have different risk tolerances for work and for home. Whether it invokes the bogeyman of radiation or nuclear reactions can be irrelevant. If some guy in an apartment next to mine is making a flamethrower inside, I might be a bit peeved. An accident might have consequences that might go far beyond their own property and well-being.
Does Uranium in the manner he was using it constitute a threat? I don't know, but it sounds like not. Does that mean the person reporting it, or the police responding knew that?
In fact, the sun generates less energy per volume than a compost. It's all about the scaling of the surface area vs. the volume.
Though the sun can keep it up far longer than a compost can...
> It's all about the scaling of the surface area vs. the volume.
If I understand that statement correctly, would it be easier to create a fusion reactor if we made it bigger?
Absolutely-but only when you approach mass scales on par with small suns. So not in a practical sense.
To elaborate, plasma in stellar bodies is essentially contained by gravity, as opposed to complex electroomagnetic fields that have to be painstakingly engineered and tend to be unstable.
And if it did, it would largely continue to orbit through the Earth, or what bits of it remained following passage.
That said, I don't suspect the approach would be particularly viable.
Difference seems subtle, until you find out that a black hole (one made by us at least) is soooo tiny that it would miss most of the atoms in the Earth, falling _between_ them.
Apparently, you need lots of mass to make a black hole that is both sustainable and big enough to affect anything around it.
It's as simple as heat loss is on the surface, energy it released in the volume. You can somewhat get around this by increasing density, as eventually you need to slow down the reaction because your walls can only take so much heat and radiation. But, it's a lot more stable at lower density which means larger is vastly easer to control.
PS: We could have built an impractical, but working fusion reactor 30 years ago. But, the total cost would have been incredible and we only have so much tritium anyway.
This would be a bit like worrying that aerospace engineers are still optimizing wings. Like wings, tokomaks/stellarators are the best tools for the job, as dictated by the underlying physics of the systems in question. Basically, donut-shaped magnetically-confined plasmas "leak" into themselves, rather than out into the world.
> Many people have spent their whole careers on it and have died without making significant progress.
This is a popular view that is entirely wrong. The figure of merit, which is the triple product of confinement time, density, and temperature, outpaced Moore's law right up until a gain of about 0.95, when the magnetic technology of the day caused the necessary size increase to put ITER in the realm of international cooperation. ITER absolutely will produce more power than it consumes. The plasma physics are just that well understood. The problem is size.
MIT have recently worked out that a new breed of superconducting magnets can more than double the available field strength, resulting in a 16-fold reduction in reactor size (due to a 4th power gain in confinement strength as a function of field strength). They hope to achieve a gain of about 2 with a university-scale reactor before ITER, designed with the magnets available at the time, is complete.
Most, if not all, of the above comes from https://www.youtube.com/watch?v=L0KuAx1COEk. ARC and SPARC, the MIT reactor concepts, are the most exciting thing I've heard about in...gosh, I guess my whole life. They could pull it off. If they do, we could save the planet with fusion-powered CO2 scrubbing. We could avert disaster.
But that said it wasn't totally lacking optimism. It was more of the hard reality of how challenging the problem will be to solve.
I still follow the development of it and there is definitely incremental progress.
Thanks for think about MIT's work, that's interesting.
Stellarators don't.
Progress has been minimal.
The more so if one recognises the distinction between Texas Sharpshooting -- drawing targets around the holes we've made with various technical methods -- and the actual initial intent.
This particular problem has turned out to be vastly more difficult than anticipated or advertised. I've watched what little progress has been made over more than 40 years.
Pretty sure people have been messing with wings for centuries longer than they have worked.
Once petrol-fueled engines and aluminium were available, the Wright beothers were flying within a few years of the Ford Model T. Monocoque airframes followed another few years later. The most perfect aircraft ever, the DC-3, which remains in active commercial use, was developed within the next 3 decades, largely thanks to aluminium alloys. Aivionics and wing design were largely solved by then.
With the gas turbine, jet aircraft appeared, and Boeing is still effectively building and selling scaled-up versions of its 707 airframe (1957). The USAF will be flying the very same aircraft for over 80 years, the B-52. First flight 1952, last date af manufature: 1962.
What progress has been made in aircraft over the last 65 years is largely limited to improved controls (both linkages and avionics), novel materials, and more recently, improved modeling duringg design and development. Outside military applications, that still has exceptionaally limited commercial impacts.
See various sources, though Robert Gordon's The Rise and Fall of American Growth has a good overview.
The contrast, particularly with man-years of effort and billions in spending, to fusion, couldn't be starker.
I wonder... if there are entire domains of technology that simply -cannot- be achieved without AI – machines building and improving themselves.
We have already seen problems that simply cannot be solved within the lifetime of humans or in terrestrial environments, like telescopes in space vs. those on ground, or crossing long distances in "generation spaceships.."
For me, AI is just another tool on top of others. You can't build an AI to, say, control a plasma without a thorough understanding of plasma physics (acquired without the use of AI)... So I'd say AI may be useful, but not in the sense that it will bring a breakthrough...
It's occurred to me that AI might actually represent a form of nonscientific capability. That is, the ability to achieve results but without the understanding or structured knowledge that we consider to be part of science.
That's not saying that AI will solve the problem, just what the situation might be should it do so.
I'm still trying to wrap my head around what it actually does itself.
https://www.google.de/search?q=wendelstein+7+x+magnete&tbm=i...
https://youtu.be/lyqt6u5_sHA
Also search HN for previous articles.
https://www.youtube.com/watch?v=L0KuAx1COEk
There Zach gives a quick overview (well hour long) of the current state of fusion and what its future is going to be. He didn't talk too much about stellarators but I feel I have now a better overall understanding of fusion physics thanks to him.
This quote is so telling on the limits of technology. So many people think of fusion as a technological miracle that can save the world. Here we have a man telling the world that we're fucked if we just sit on our asses waiting for this miracle.
The thing has the thermal management problem from hell. The magnet coils are superconductors and have to be maintained down near liquid helium temperatures. The plasma is at a few million degrees. These are inches apart.
3D printing might help when the next one is built. Rocket engines have some of the same problems - they're plumbing parts filled with cooling hollows. Those take way too many joints and welds to build. The Space-X Dragon spacecraft has 3D printed engines, which gets the part count down.
If we ever get room temperature superconductors, these things will be much simpler. If we ever got high temperature superconductors, so that the whole thing could run hot, like a boiler, that might make it economically feasible.
"Wendelstein 7-X - from concept to reality" (10m30s)
https://www.youtube.com/watch?v=lyqt6u5_sHA
So in the middle you'd have your conjoined mega stream, and then that would flow upward, be separated out into filaments which then curl around in all directions down to the roots where they are re-merged into the megastream.
Probably too much turbulence at the branches, but maybe they can use vibration and learning to find beneficial harmonics. I would hire sound engineers.
I just have an intuition that a single stream will never have adequate confinement and you'll need redundancy in the geometry to get good concentration.