High-Entropy Alloy | Hacker News Reader

High-Entropy Alloy

en.wikipedia.org

156 pointsleonidasrup27 comments

The challenge with modelling HEAs is that they have very complex electronic structures, its very tempting for a newbie to throw an MLIP at the problem but in reality you have many complicated bonding arrangements that are not captured by these models, this is also compounded by the fact that you dont just have a slab with a bunch of itinerant electrons but you end up with covalent and even ionic-like bondings forming in the SRO substructures. Then theres spin treatment (which matters a lot), and also because the configuration space is combitatorially large you also have to do some high throughout studies with statistical interpretation since by definition theres no such thing as a representative unit cell in an HEA

How do I know? We have invented multiple via simulation and have them in the lab for synthesis now!

Fascinating! Where is this written up?

Seriously!

This is knowledge I have gained through experience building my company to study these things over the last couple of years. Theres a lot LOT more to this, if youre interested I would recommend reading all the papers you can find on HEA. Getting a subscription to Advanced Materials from Wiley, and then trying to simulate some of the materials yourself. Don't start with HEAs they are hard and you need a lot of computing power, start with simple systems like bulk copper, aluminum and iron. Then move to binary systems, ternary systems and increase the complexity of what youre modelling while always checking against experimental data. Learn about all of the shortcomings of the simulations and then ask yourself "ok how can I improve that", while you're learning this you're developing an intuition for all of the settings in the simulations (whether atomistic. Meso-scale, macroscale or other)

I have a 15 GPU cluster in my house just so I can study HEAs - but I understand thats out of budget for a hobbiest so that's why I recommend you start with simpler systems and slowly increase complexity.

You might see various datasets for HEA, HEA property prediction, and synthesis predictors, but cold hard truth of the matter is that the quantum interactions at the interatomic level are so complex, the configuration space youre searching is so massive, that no dataset is going to make a dent in it, so models are only really useful as VERY VERY VERY approximate screening tools (sometimes) - and thats not even talking about micro-scale phenomena and macroscale phenomena - which are enormous subjects on their own and just as important!

You must simulate all of these, you can't just do a Microsoft Mattergen that spits out an idealized crystal structure at 0 Kelvin, because in the real world, thats barely the first step.

Because with so many metals in high entropy alloys, you can tune it to whatever, and that's why it's currently being investigated for potential room temperature superconductors.

Have any commercially interesting alloys been found? This article seems to be all about research.

It depends what you mean by commercially interesting. There’s loads of interest in aerospace (for high temp corrosion resistant structural components) and catalysis but these alloys are pretty much across the board at a relatively low level of technical readiness. It’s developed enough that there’s significant industry R&D and not just academic and government research, but I don’t think there’s really wide-scale deployment yet of alloys with 4+ principal elements

Most high-entropy alloys contain expensive metals so the primary domain of interest has mostly been as a coating for other metals. Recently there has been some work on AlCrMnFeNi, which is the cheapest composition I've seen by far.

has anyone ever attempted to create a machine that would trial semi-random material compositions with minimal human involvement?

Yes, someone did. It's actually an active field right now, from high-throughput simulations to try to limit the search space, to additive manufacturing of thousands of samples, to semi-automated characterisation and some testing (e.g. corrosion; for things like mechanics it's more difficult). The idea is te be more efficient than semi-random.

ok, so quasi-random. Would it be more useful to have such a machine to feed data back to models to better the simulations where we could do more in less time and prune down to more likely candidates?

That’s the active research area GP mentioned. In startup land there are a few large outfits, Lila Sciences, Period Labs, Radical AI are all doing a mix of simulations, AI, and autonomous laboratory infrastructure specifically for materials science. (Lila does a lot of biotech but the have materials researchers too)

Also lots of interest and activity in this space in the national labs and academic research scene

It’s something we are trying to do. Currently, there is no machine that can do that. The problem is that the composition space is huge (5 elements out of a dozen, plus about twenty possibly useful minor additives, the combinatorial explosion sets in very quickly) without even taking into account processes and things like microstructure.

No model is sufficient: predictive physical models like DFT are impractical at the required scale (in term of simulation size and compositions to consider, as well as computational cost), and all the fancy regression machines are terrible when extrapolating. Which is too bad, because again the search space is huge and the regions we actually know and can use to train our models are tiny, which means that apart from proofs of concept, we are always extrapolating. And so we need experimental data in the unknown regions of that space to validate the models. It’s like trying to describe the Earth with only being able to see 1cm squares from random positions.

We are not just sitting waiting for CS people to solve everything with LLMs, these things are genuinely complex ;)

I did some work with HEAs, specifically Paliney-6 and Paliney-7, and was pretty blown away by two things:

1. Material properties

2. Cost.

I'd love to hear more, especially on #1.

And while you're at it, #2

I was trying to make electrical contacts that didn't dissolve within minutes when charged with a little bit of sweat on them.

Learned about HEAs (can't remember where), saw some fantastic material properties (rather high yield strengths, solid elastic modulus), and was pleasantly surprised at how much electrolytic corrosion they could resist.

Then one of our money people got a quote from the vendor to see how much it would cost to put it in our product, laughed, and said, "Back to the drawing board you go!" I think it was something like 20x more expensive than a copper alloy contact.

> Paliney® 6 is an age-hardenable palladium silver-based alloy ideally suited for demanding low current sliding electrical contact applications

I wonder if the Xenonite is a high-entropy alloy :-)

In the book Grace says that it's a mess of proteins and molecules that he gives up on trying to understand

>mess of proteins and molecules

Because proteins famously aren't molecules.

be nice

This is a legitimate, understandable way to discuss a mixture of abstract and specific things. This is a novel we are referring to, here. The intended audience is very, very broad.

> Because proteins famously aren't molecules.

Tertiary and quaternary protein structures are much more complex than molecules and have emergent properties.

People refer to stuff like proteins as “biologics” and to things we synthesize traditionally as “small molecules” so it does make sense

These things are interesting but for the most part quite dull and very industry-facing. Just mash together a bunch of random elements and see if it improves the thing you want to optimize.

Random sampling! Known by computer scientists everywhere to be the worst search strategy

Reply on news.ycombinator.com