The Problem with Nuclear Fusion

Humans are in a forever quest to find the most efficient way to boil water that spins something

"Technology is [almost] always overestimated in the short term, and underestimated in the long."

As a fusion researcher, titles like this are pretty frustrating since we have very little control over the public image of fusion, and impressions that don't click on the video get a negative impression. However, this video itself is pretty solid, i'm halfway through and it's been all good and accurate points. The key is we need a societal push to produce effective fusion, ideally similar to space-race level support. otherwise it will continue to take a long time. People like to say it's always been 20 years away, but fail to mention that that's primarily because the funding for fusion research has been slashed over and over again ever since we stopped competing with the soviet union over it as a technology. If you don't fund something, it stops moving as quickly.

The fact that we can all learn about this stuff whenever we like is absolutely incredible. Thank you so much for sharing this content with the world!

Damn, this is really impressive timing. This video clearly took a long time to make, and you managed to release it right on the heels of the major fusion announcement. Props man.

Great video. I’ve worked on magnetic confinement nuclear fusion at a National lab. One of the big problems is that everything is a massive partial derivative. Let’s say you have a hundred million particles inside the reactor. Each one has 7 dimensions of conserved motion. So traditionally to simulate it, you need either an extremely watered down equation that can make dozens of simplifying assumptions or a supercomputer that can take days to do a short simulation. Each particle can massively change their trajectory in just 10 microseconds, meaning if you want a perfect simulation for like 5 minutes, you need to run (100,000,000 * 5 * 60 * 10^5)^7 calculations which is probably some number bigger than the amount of atoms on earth. The good thing is that machine learning and GPUs are perfect for this kind of problem. So it’s gradually becoming the backbone of confinement modeling so the particles won’t fly out unexpectedly or the machine suddenly crash. In fact confinements not much of an issue anymore I’d say. Like I was able to take a 42 hour long simulation, rewrite it in CUDA, and have it run the same thing in less than 20 minutes. And on the experimental side, people have been able to run their devices for almost 10 minutes now which is extremely stable. And on top of all that, MIT recently made a high termperature superconductor that can generate massive magnetic fields so the machines can be scaled down to like 1/10th the size and still produce more power than ITER. While it’s still experimental, it’s basically in its last stages. The only things to worry about now is elmo. The next stage will be finding the right materials to make the walls out of that will preserve the reaction and dissipate the energy over long runs. Tungsten has been generally used as the newer material but there’s still room for massive improvement. People have been experimenting with depositing an atom layer thick lithium gas on tungsten or using Liquid Metal walls that act kind of like a waterfall + cpu cooler inside the tokamak. The tritium breeding issue is still on the horizon. Theoretically it shouldn’t be too hard to carry out. Just get a slab of lithium metal and put it right behind the tokamak or whatever device. But that’s a issue for another day. Edit: also don’t forget that’s only tokamak fusion. There’s also like 20 other types of fusion techniques, some of which are very promising. ICF (kind of a budget laser-powered nuclear bomb) is experimentally like 70% (now >100%!) of the way to net energy. This is a massive jump considering how it was 5% just 3 years ago. Other ones are field reverse configuration with folding plasma whirlpools in the z direction, magnetic mirrors, z pinch (like a nuclear lightning bolt), and whatever that doesn’t need tritium and won’t produce neutrons so that the energy can be directly harvested with a magnet. Edit: from Joesph Li I do research currently at a fusion research institute. Some things worth noting: You're right that lots of folks are looking to ML and big data but traditional MHD and kinetic simulations are still the mainstay. Whole volume gyrokinetic simulation is widely used and actively researched to better understand transport as a whole. Neutral and wall interactions, and especially precise treatment of edge regions is mostly done via traditional methods rather than ML. Next on fusion techniques. You're correct that there are many different methods for containing a plasma, but the region the device landscape looks as it does today is because magnetic mirrors, Z-pinch, and other open designs have serious inherent confinement issues. As for ICF, the NIF actually reported a break even last summer on fusion gain in 2021 (not sure if published yet). One major issue the NIF faces is that ICF fusion requires several laser shots per second to be economical, but as of now I believe the rate of discharge is closer to once every several hours. Lastly, I want to comment on smaller devices and different fusion fuels. MITs technology definitely holds a lot of promise, but one should temper expectations on output in smaller devices. Neutrons don't interact strongly and slips through solid material given enough energy (like those resulting for D-T reactions), so any D-T fuelled device necessitates fairly thick blankets and walls to efficiently extract energy. Moreover, one must still shield the outside of the vacuum vessel from neutron irradiation lest we risk the creating of radioisotopes outside the device from neutron bombardment. The obvious solution of using different fuel isn't so simple when one considers the much greater activation energy of those reactions. We can't even break even on D-T, never mind considering continuous operation on D-D or D-He3. Progress is certainly being made but science takes time. And fusion is too expensive to get wrong. It's best to not get too hasty with our expectations. Edit edit: just now, NIF achieved laser Q > 1 which is amazing for science. But it’s still a ways away from economic Q > 1 (where you can actually use it to produce power). But definitely a milestone and not number fudging anymore.

Advanced fission reactors are also incredible. Breeder Reactors that produce their own fuel are also possible. You essentially use something that doesn't slow down neutrons as much as water and add thick Uranium cladding to inner walls of the reactor. Preferably that uranium is deluded in a liquid so it can be refined easily. One of the issues is that we use about 1% of the available uranium and the rest is waste. But with a working Breeder reactor you could have so much more efficient designs that don't need a refuel in decades

I’d like a citation on how fission reactors are uneconomical. I googled it and yeah they’re expensive to start up but they make a lot of power for a long time after startup. Most of the reason for shutting them down recently is fear mongering and concern trolling about nuclear accidents.

Hi, I've been an engineer on one of the largest laser driven nuclear devices for a couple decades now and while I disagree that actinide contamination of a beryllium multiplier will be a major issue for future machines (several common chemical purification techniques could refine the Be such that it is a nonissue), I liked the video overall and thought it was a good layman's overview to the present state of MFE research problems. I can also offer your viewers a bit of "inside" rumor-mill information on fusion that I don't see being reported anywhere in the press yet that they may be interested in. Perhaps you heard of the record yield shot on the NIF laser in summer of last year that produced 1.3 megajoules of energy for an input of 1.9MJ laser light and that this was 25 times the previous highest record shot taken just a couple years earlier. Well they've been trying to recreate that magic shot for over a year and a half now without success....until 2 weeks ago. The rumor is a shot in the last week of November exceeded 2.5 megajoules in yield. The scientists are in the process of crossing their t's and dotting their i's before going to publication in the coming days and issuing a press release to make sure the result is real, but it's likely to be confirmed and is beyond any shadow of a doubt an unambiguous achievement of thermonuclear ignition and breakeven in the laboratory. This is a MAJOR breakthrough (and that's coming from someone who loathes the overuse of that word) that countless scientists and engineers have devoted their entire careers to attaining without seeing it happen over the last half-century, and now it is done. NIF is of course not a power reactor, just an experiment, and so this is the achievement of scientific breakeven rather than engineering breakeven. But keep in mind there is on the order of roughly 100 times more fuel in one of these capsules yet to be burned, and the laser driver can be increased from its current 1% efficiency to >40% efficient with the use of diode laser pumps and a crystalline lasing material. Even in the very unlikely event the fusion yield of these implosions isn't increased further, this is still a tremendous milestone that brings an entirely new ultra-bright neutron source "tool" for research into the laboratory. EDIT: looks like the cat's out of the bag much earlier than I thought it would be - article is up on the Financial Times titled "Fusion energy breakthrough by US scientists boosts clean power hopes"

Given the last 24 hours, I'd love to hear an update on opinions of where this is all going.

Of note: (And this is a common error in media) when you mention cooling of MRI machines, the clip is of a CT machine. This is likely a common error because CT machines are safe to film when not actively scanning, while MRI magnets are always active even when not in use.

Amazing video. Easy to understand for People not close to the topic and slightly detailed with some technical info. Congrats !!

Breeder blankets: being a nuclear engineer who has researched tritium breeder blankets, one material being considered is F-Be-Li molten salts. It gives you a combination of features: as a heat transfer fluid, Be for neutron multiplication, and the lithium to generate tritium. It would also maximize tritium generation if enriched to nearly 100 percent as Li-6. Looking forward to your documentary on Helion since I have been reading up on them.

Wendover and real engineering fighting each other like siblings always makes me chuckle

The thing is that unlike some other technologies were their was something missing we didn't know that made a massive leap forward, theirs doesn't appear to be anything as of yet like that with Fusion. We just have to do the hard work and make out plasma hotter and more stable till its hot enough and stable enough for it to be cost effective.

I'm not entirely sure, but my understanding is that He3 (for the most part) does not fuse with another H1 atom to become He4 (it does on occasion, but it's a miniscule amount of the He4 atoms that are produced during fusion in a star). The main branch of the proton proton chain which produces He4 is when 2 He3 atoms fuse together into He4, releasing two H1 atoms.

In spite of there being 100s of fusion videos on YouTube, you find the perfect balance between highly engaging presentation with juicy technical details and engineering considerations, like Tritium breeding and annual consumption. Well done, Real Engineering.

I ran a copper alloy foundry for a while, and several of my customers made me sign agreements that said that the facility did not work with ANY beryllium alloys (they are common in electrical contact gear made of copper alloys). I always thought it was because Be is insanely toxic, but in hind sight, there may have been radiological considerations after watching this episode. All of the customers asking for those "no Be" contracting were defense contractors, and it would make sense that they were sensitive to background radiation in their bearing components.

At this point, there is no debating the fact that this channel produces the absolute best video content on pure unadulterated engineering so I wont comment on that. I just love the fact that Wendover Productions and Real Engineering, my two favourite channels, are buddies and can engage in friendly "leg-pulling" on Youtube with complete understanding. Cheers guys!! I would love nothing more than the two of you co-creating a video or a series of videos together which have aspects of both engineering and business on some of the most pressing issues/problems. Merry Christmas and a happy new year World :) Much Love!!

Just had a nuclear module and this video basically covered the first introduction lesson, thanks for the quick and easy explanation.