AI Has Just Solved One Of The Biggest Issues With Fusion Energy
AI Has Just Solved One Of The Biggest Issues With Fusion Energy
A new AI can tame the plasma dragon.
I have said this a million times by now, but nuclear fusion power could be one of the most revolutionary technologies ever. All the on-tap power and low carbon emissions of traditional nuclear power, except it’s even more fuel efficient, uses a super-abundant and practically renewable fuel, and produces no nuclear waste. It is literally the perfect energy source. However, for decades, we have struggled to unlock it and bring this sci-fi technology into reality. However, that might be about to change thanks to a new AI that can tame the erratic plasma beast that is thwarting our current efforts.
Before we dive in, let’s quickly recap what fusion is. Nuclear fusion happens when two atoms are forced together with so much energy that they overcome the repulsive force that keeps atoms separate. They then fuse into a new, larger atom. However, because of a weird quirk in physics, this atom is actually slightly lighter than the mass of the two it formed from. During fusion, this leftover mass is turned into energy and released through heat and radiation. If you remember, clever clogs Einstein figured out that a bit of mass is equivalent to a biblical amount of energy (E=MC²), so even though it’s only a tiny amount of mass being transformed here, it releases an insane amount of energy!
Hydrogen, being the lightest element, is also the easiest to fuse as it takes the least energy to overcome its repulsive force. But to give you an idea of how potent fusion is, fusing just 17 tonnes of hydrogen will emit enough energy to power the entire US for a year.
But there is a problem. It takes energy to initiate fusion. So the trick is building a reactor that can create fusion so efficiently that the energy it expends is less than the energy it can get from the fusion reactions. This is the crux of the nuclear fusion problem.
One of the most famous reactor designs aimed at solving this problem is the tokamak. These reactors have a doughnut-shaped chamber surrounded by extremely powerful superconducting electromagnets. Superheated hydrogen plasma is injected into the chamber, and because hydrogen plasma interacts strongly with electromagnetic forces, the magnets can heat and compress the plasma incredibly efficiently and initiate fusion.
Until very recently, this was by far the best fusion reactor design we had, as it would return around 30% of the energy put into it. As you can tell, that is far from a viable energy source. To increase this efficiency and get more energy out than we put into these things, we need to build bigger, more powerful tokamaks, which we are already doing (ITER), and solve one of the most significant flaws in tokamaks: plasma instability.
These plasmas combine three of the most complex parts of physics, fluid dynamics, electromagnetic dynamics, and quantum mechanics. This is far from a recipe for predictability and stability, so the plasma acts erratically, creating turbulence, which reduces the efficiency of the tokamak.
By far the worst and most common type of plasma instability is tearing instability. In these circumstances, magnetic fields that run in opposite directions, which are needed to heat and compress the plasma, break away from their straight path and form little eddy islands in the magnetic field. Scientists refer to these as “radially-localized magnetic island chains”, and they suck for fusion. They effectively stop electromagnetic energy from flowing into the plasma and instead channel it to create electromagnetic turbulence. Tearing instability events are incredibly common in modern tokamaks, and the energy they rob significantly impacts the overall efficiency.
But, if we can find a way to predict when and where tearing instability will happen in a tokamak and then actively stop it from happening, we could dramatically improve the overall efficiency in one fell swoop.
But these events happen in milliseconds, and because the physics behind them is so convoluted, we can’t predict them using sensors and pure maths.
So, researchers have turned to AI to try and control the plasma instability beast, and the results are astonishing.
This AI estimates the likelihood of a tearing instability 300 milliseconds in the future from already available diagnostic data from a tokamak, then dynamically manipulates the electromagnetic field in the reactor to keep the likelihood of a tearing instability below a threshold. This was tested in the DIII-D tokamak, the largest tokamak in the US, and was found to be highly effective, allowing the plasma to actively track the stable path even while in “H-mode”, which has a greater rate of fusion and higher probabilities of tearing instabilities.
In short, this AI may have just solved one of the most significant issues with tokamaks.
So, will this unlock nuclear fusion power and our utopian future? No. Well, not yet.
This AI has yet to be tested in a tokamak capable of reaching energy-net-gain levels of efficiency. In fact, it probably won’t for a while, as ITER, the first tokamak that could theoretically reach this point, is still years from completion. Moreover, just because this AI works in a smaller reactor doesn’t mean it can cut the mustard in a massive one. So, while this AI could theoretically be a giant leap forward, it still has a long way to go before it delivers any revolution.
Thanks for reading! Content like this doesn’t happen without your support. So, if you want to see more like this, don’t forget to Subscribe and follow me on BlueSky or X and help get the word out by hitting the share button below.
Sources: Nature, Energy.gov, IOP, Independent, Vice, Nature, GA, CEA