https://www.sciencealert.com/ai-designs-radical-magnet-free-of-rare-earth-metals-in-just-3-months
We urgently need to move away from fossil fuels, but electric vehicles and other green technology can put their own pressures on the environment. That pressure could be eased with a new magnet design, free from rare-earth metals, that was built with AI in just three months.
Rare-earth metals are essential components in modern-day gadgets and electric tech – including cars, wind turbines, and solar panels – but getting them out of the ground costs a lot in terms of money, energy, and environmental impact.
As a result, technology that doesn't use these metals can help us transition towards a greener future more quickly. Enter UK company Materials Nexus, which has used its bespoke AI platform to create MagNex, a permanent magnet requiring no rare-earth metals.
This isn't the first such magnet to be developed, but discovering these materials typically requires a lot of trial and error and can take decades. The use of AI sped everything up by approximately 200 times – in just three months, it had designed, synthesized, and tested the new magnet.
The AI works by analysing over 100 million compositions of possible rare-earth-free magnets, weighing up not only the potential performance but also supply chain security, cost to manufacture, and environmental issues.
"AI-powered materials design will impact not only magnetics but also the entire field of materials science," says physicist Jonathan Bean, the CEO of Materials Nexus.
"We have now identified a scalable method for designing new materials for all kinds of industrial needs."
Materials Nexus partnered up with a team from the Henry Royce Institute at the University of Sheffield in the UK to produce the magnet, and it's thought that similar techniques could be used to develop other devices and components free of rare-earth magnets.
According to the makers of MagNex, compared with conventional magnets, the material costs are 20 percent what they would otherwise be, and there's also a 70 percent reduction in material carbon emissions.
In the electric vehicle industry alone, the demand for rare-earth magnets is expected to be ten times the current level by 2030, according to Materials Nexus – which indicates just how important these alternative materials have the potential to be.
As well as using AI to make manufacturing processes more efficient, researchers are busy trying to find ways to collect rare-earth materials in more sustainable ways. Breakthroughs like this should speed up the move away from fossil fuels and CO2 emissions.
[...] "The next generation of materials, unlocked through the power of AI, is highly promising for research, industry, and our planet."
(Score: 2, Insightful) by anubi on Friday June 28 2024, @09:42AM (4 children)
If there was one thing simulators showed me is although all the results may appear impressive, the ultimate proof is can I make one?
I remember AlNiCo was about the best for a long time, then Neodymium blew it away. There is no telling what else is out there that's even better.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 4, Interesting) by c0lo on Friday June 28 2024, @10:32AM (3 children)
Seems like they have already made it [materialsnexus.com] (and I doubt that you could make anything beyond magnetite magnets w/o spending a lot on equipment for ceramics/chemistry and time to set everything working).
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 3, Interesting) by HiThere on Friday June 28 2024, @01:47PM (2 children)
But at least the summary said nothing about the required operating conditions or the power/weight of the magnet (or, come to that, the power/volume).
It's *very* interesting, but I can't tell whether it's important.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 3, Informative) by VLM on Friday June 28 2024, @04:48PM (1 child)
MGOe is the figure of merit and it's "easy and cheap" to make 1960s ceramics with a number around 1 to 3, and pretty hard to make rare earth magnets more than the upper 40s or lower 50s. Non engineers will often call the MGO "the grade" or sell it as a model number LOL.
In certain restricted shapes and sizes for an enormous pile of money you can buy up to mid 50s MGO, COTS, shipped to your door today. That's the best you can do in the mid 2020s.
Note that's at a particular geometrical L/D ratio and if you invent a material that requires a weird physical L/D ratio it might not be commercially viable.
Also it's at a peak of induction vs field strength. Obviously at extremes of field strength the induction is zero. So you can game the system again by optimizing for unusable maxima.
It's like the concept of the peak production point of a solar panel; you'll never exceed that, in fact you'll generally never get close, but its useful for comparisons between similar materials unless you game the system extensively; Like AI are designed to do.
There's some handwavy physics / chemistry I don't entirely understand that even if you went hard sci fi and made magic atoms out of only the correct electron orbitals then you could only make a grade 150 magnet using the feromagnetic atoms our universe has, even if all problems could be magically handwaved away. And supposedly there are lab examples of a lucky grain of sand or whatever that once theoretically implied 130+ but I donno about all that.
Some things like optics can be made up against the physical formula boundaries of the world, diffraction limited optics etc. IC electronics getting close to physical universe levels. Magnets have a good ways to go but not really all that far. In theory as we understand basic physics and chemistry even if you crashed a magic-level UFO and found permanent magnets in the debris the magnets could only be ten times better than 1960s mags and maybe three times better than 2020s rare earth mags and frankly that's optimistic. Based upon physical limitations of this universe. You'd need entirely new physics to describe a magnet ten times better than COTS rare earth magnets.
(Score: 2) by HiThere on Friday June 28 2024, @08:19PM
Sometimes you just need to look at the equations in a different way. ... But that's not really the question here. The question here is how good is this new "doesn't use rare earths" magnet. (I'd also prefer that it not require a liquid helium bath.)
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 2) by looorg on Friday June 28 2024, @10:23AM (11 children)
They couldn't come up with a better name then MagNex? Did they ask the AI for that one to cause a simple online search for it reveals there are already several products called Magnex -- seems to be a lot of supplements and other "health" products.
They didn't even secure any of the domain names yet for this new super-magnet ...
(Score: 2) by c0lo on Friday June 28 2024, @10:37AM (10 children)
What do you think is wrong with that name?
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 5, Touché) by Snospar on Friday June 28 2024, @11:03AM (1 child)
Obviously it should be "MagNexAI". We add AI to everything now right?
Huge thanks to all the Soylent volunteers without whom this community (and this post) would not be possible.
(Score: 0) by Anonymous Coward on Friday June 28 2024, @04:06PM
OK, you register the "MagNexAI" domain. I'll register some variations like AMagNexI and we'll auction them off to the magnet companies...
(Score: 4, Informative) by looorg on Friday June 28 2024, @11:13AM (1 child)
It's not that it's a bad name per se but that it's already taken by other products.
(Score: 2) by c0lo on Friday June 28 2024, @12:01PM
Thanks, I wasn't aware - never stumbled on one, it turns a lot of them on a search.
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 3, Funny) by DannyB on Friday June 28 2024, @01:55PM (2 children)
The name needs to firmly stick to your brain and be very difficult to pry apart from it once attached.
People who can't distinguish between etymology and entomology bug me in ways I cannot put into words.
(Score: 2) by Freeman on Friday June 28 2024, @02:30PM
Should have just called it one of the expletives that we all know. That would sure stick.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 3, Funny) by c0lo on Friday June 28 2024, @04:47PM
Say what you want but I don't think "amyloid beta plaque" is good for a new magnet name.
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 2, Insightful) by Type44Q on Friday June 28 2024, @04:01PM
It's a perfect name for anti-static wipes.
(Score: 3, Funny) by DannyB on Friday June 28 2024, @04:51PM (1 child)
MagNex sounds like something you use on your neck. Maybe to get rid of maggots?
People who can't distinguish between etymology and entomology bug me in ways I cannot put into words.
(Score: 2) by c0lo on Friday June 28 2024, @05:14PM
1. how very anglish of you
2. I don't see the logic in a "the name of the product must not create confusion about the nature of the product or its intended use" argument. After all, we don't go to the pharmacy asking for 1,2,3,4-tetrahydro-β-carboline
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 3, Insightful) by progo on Friday June 28 2024, @01:22PM (3 children)
Invented by a non-sentient AI -- that means no patent monopoly, right?
(Score: 2) by HiThere on Friday June 28 2024, @01:50PM
Interesting point. That may well mean it's only patentable in some countries. And currently that doesn't appear to include the US.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 5, Insightful) by DannyB on Friday June 28 2024, @02:00PM
If I used any other computerized or software based tool to assist me in creating an invention, there wouldn't be any doubt that I were the inventor.
The AI is just a tool. What is missing from AI is the "I" part of it. It just does what it is supposed to do. It isn't conscious. It doesn't think. It doesn't plan, plot or ponder. It doesn't create. It is extremely good at recognizing or predicting patterns based on trained data. Computers are very fast at trying out lots of possibilities.
How is this not just human inventors using a tool?
People who can't distinguish between etymology and entomology bug me in ways I cannot put into words.
(Score: 2) by VLM on Friday June 28 2024, @04:53PM
Classic parallel construction. The AI says number 7 in step 13 is likely 1.23456789 and the scale-up process engineers figure out 1.25 works better IRL so you patent the latter. You'd want to patent what actually works in mass production anyway.
(Score: 3, Insightful) by gnuman on Friday June 28 2024, @09:18PM
This guy is absolutely correct because material science is more of an art than science. Quantum mechanics can't simulate these things because the solution space is far too large. And since AI is good at pattern recognition, maybe it can stare at crystals and other solids and find patterns where humans see nothing but noise.
So if you feed it 1000s of examples of magnets, maybe it can find that gem by pure iteration. The field is ripe for solutions. The only question is whether these solutions can be easily manufactured.