[badman2001]

the dft results are definitely interesting, but i note that with the caveat that my background is in experimental condensed matter physics for materials like this and not theoretical, my understanding is that the dominant feature of the conclusions (the flat bands) is a necessary but not sufficient condition for superconductivity in the way the authors describe.

again, in experimental condensed matter physics it's acceptable to do a fine experiment and then throw in a half-baked "theoretical underpinning" to appease reviewers, so I wouldn't be surprised if the superconductivity turns out to be totally unrelated to the mechanism proposed in the paper. i would really like to see some more robust characterization work(biased because this is my background), hopefully some of the labs doing the replication studies can take a look at the juicy stuff

[adw]

I know some DFT (*), but very little superconductivity, but I read through Sinéad Griffin's preprint and there was nothing in there which looked weird from a methodological perspective – and the methods (and software) she is using are extremely well-established and well-categorized.

(*) it was, like, two decades ago, but I've got a first-author PRB paper so I wouldn't trust me compared to an active researcher but I'm not _entirely_ clueless

[montecarl]

The thing that stands out to me is that the DFT simulations show that the flat bands only occur in a particular crystal structure of the material and it is not the most stable state (at least according to the simulation). This would explain the synthetic challenges involved. These simulations are not perfect, but they can be VERY useful when guided by experiment and when they correlate strongly it is a good sign that you have a mechanistic explanation of the phenomenon.

[adw]

The way I'd characterize it is that they're usually _directionally_ (and mechanistically) correct. On something as sensitive as a band gap the error bars are larger; if the DFT simulations said "yeah, no way this has band density at the Fermi level" I'd regard it as strong evidence against the LK-99 claims, but the fact this is in the ballpark is – to be clear, pretty weak – evidence in favor.

[yreg]

naive question: if we can simulate this, can't we brute force other superconductors?

[adw]

Not really. Think of this as being analogous to public key/private key crypto; it's easy to check a signature (simulate a structure), it's extremely hard to forge one (design a new material ab initio).

[fzzzy]

Yes, but the search space is mind boggling.

[chevman]

How so? Just a function of all the different atoms and sub-atomic particles interacting?

[montecarl]

I'll link to another reply of mine: https://news.ycombinator.com/item?id=36966474

The "just a function of" is a non-trivial problem. Given a single set of elements you want to "try out" results in a huge global optimization problem to find the set of stable structures (low energy). When aided by experimental data it becomes a tractable problem.

[foven]

Something that I find hard to understand is why there is superconductivity without cooper pairs; granted my understanding is related to more traditional superconductors and I'm not really very knowledgeable about the cutting edge high-Tc stuff.