I am a PhD student in Physics, but I am no expert in condensed matter physics (I do research in computational NMR). Based on my experience with theorists and simulation researcher, I am a bit concerned with anchoring bias and the speed at which simulation papers are being published here. Then again, I don't know exactly what kind of research steps they follow to do their research.
There’s an old saying that “simulations are bound to succeed,” but also as an academic that has a working knowledge of statistical mechanics but is not a (condensed matter) physicist, from what I’ve read it seems folks are approaching this with healthy skepticism likely due to the drama around the development and release of the first paper.
The simulation paper folks are talking about used what appeared to be an existing DFT simulation package. Now, DFT is an approximate theory used to render computation tractable, but to my understanding it is a popular and mature method. I was actually kind of impressed that they were able to reproduce results from the LK paper in simulation so quickly. While it’s possible the speed led to a bug or error in the analysis, simulations often don’t just magically work and can take a decent amount of parameter tuning — especially if the system being simulated has something tricky or exotic going on. The fact that they were able to get what appears to be an accurate simulation working quickly that also justifies the low yield rates has made me more cautiously optimistic than anything
I don't know why anyone should be concerned about anything here, unless you want immediate confirmation.
It's a process. Scientists will try to replicate and try to simulate and try to reason theoretically. They are bound to make mistakes but all of this can be critiqued and iterated on.
Again, there's no problem unless you need immediate confirmation or you think chasing this idea is a waste of time.
They theoretically agree that the material's properties could be nice for making some kind of superconductor, but present no clear indication that it would/should be a room temperature superconductor. The critical temperature depends on stuff that can't be probed via such simulations (i.e., electron-electron interactions).
From my understanding (I am not a superconductor person, but in an adjacent field), having flat bands at the Fermi level is not that rare. Such features appear in other materials that are evidently not superconductors, room temperature or otherwise. So the conclusions are more along the lines of "maybe it wouldn't be totally crazy", rather than "omg, we predict this material has astounding properties".
> If doped such an electronic structure [LK-99] might support flat-band superconductivity or an correlation-enhanced electron-phonon mechanism, whereas a diamagnet without superconductivity appears to be rather at odds with our results.
Sounds like it, if prepared right it could be a super conductor and would NOT be a diamagnet that would display the properties we saw in those videos.
That’s not unsurprising when it comes to a lot of research. The bench scientists I know often struggle with yield rates and labs have special knowledge about their particular set up when it comes to getting stuff to work.
Even in robotics (my area), if you are watching a video of a robot doing something cool, there’s often a bunch of times they ran the same demo and it didn’t work for some (often largely irrelevant to the main idea) reason. I also remember, in an undergrad analog circuits class, we had to build an amplifier on a breadboard with certain performance specs (e.g., a fairly high cut off frequency, etc.). This ended up being fairly difficult due to the tolerances in the components to which we had access and breadboard parasitics. I recall getting a non-trivial performance boost by swapping out a dozen 2n2222’s until we found “a good one.” The gray beard professor laughed and said that’s an expected part of our practical education.
"Origin of correlated isolated flat bands in copper-substituted lead phosphate apatite" (2023) https://arxiv.org/abs/2307.16892 :
> Abstract: A recent report of room temperature superconductivity at ambient pressure in Cu-substituted apatite (`LK99') has invigorated interest in the understanding of what materials and mechanisms can allow for high-temperature superconductivity. Here I perform density functional theory calculations on Cu-substituted lead phosphate apatite, identifying correlated isolated flat bands at the Fermi level, a common signature of high transition temperatures in already established families of superconductors. I elucidate the origins of these isolated bands as arising from a structural distortion induced by the Cu ions and a chiral charge density wave from the Pb lone pairs. These results suggest that a minimal two-band model can encompass much of the low-energy physics in this system. Finally, I discuss the implications of my results on possible superconductivity in Cu-doped apatite
No, Lawrence Berkeley National Lab, which is Up the Hill from Berkeley. They're run by the same underlying organization but are distinct (yet overlapping in many ways). LBL evolved out of the UC Berkeley Rad Lab, run by Earnest O. Lawrence (same name as the current lab). They do non-classified research.
There is also Lawrence Livermore National Lab, which is nearby, but in Livermore. They do classified research in addition to non-classified. I suppose it's one of the two places they simulate nuclear weapons... errr, run large scale multi-physics combustion codes for stockpile stewardship.
> I suppose it's one of the two places they simulate nuclear weapons... errr, run large scale multi-physics combustion codes for stockpile stewardship.
Back in the '90s my friend (jokingly) lamented that they wouldn't let him try to play Everquest on their computer.
yes, but if you reported 100 different substances reaching superconductivity, 100 theorists will publish papers demonstrating their model supports the observation... regardless of whether those substances actually reach superconductivity.
Sure it’s post hoc, but these are ab initio calculations, there’s not all that much wiggle room that you could make anything you like come out looking like a superconductor. Plus now I’ve seen three calculations with three different functionals and reasonable convergence criteria and they all find similar electronic structure
Well, then why spend all this time searching in a lab? Instead, run a bunch of simulations until the compounds with properties you desire just sort of pop out as the top hits in your simulations?
(I say this as somebody with a decade+ of experience running large ensembles of classical MD simulations, but not so much experience with inorganic DFT)
There are high throughput DFT projects (e.g., the materials project) but (1) the calculations are very computationally expensive and (2) the search space is really large. People are doing cool stuff with ML and generative models, but it’s a pretty open research area still
Also, at the end of the day, DFT is still an imperfect approximate model. Relative trends are generally more reliable than exact correspondence with experiment, and it can have system-specific systematic errors that are hard to account for in a high throughput setting
Crystal structure prediction by itself is a very hard problem. Just given the chemical formula and asking "what is the most stable structure" is a global optimization problem. And this is just for one single composition. It isn't an easy global optimization problem either. You not only have to determine the unit cell vectors, the positions of the atoms in the unit cell, but also the number of atoms in the unit cell needed to represent the structure! Just because you are even given "Pb9Cu(PO4)6O" as the formula finding the most stable structure, let alone the superconducting non-minimum energy state is a huge undertaking. Now expand that to all of chemical space and you can see that it is not that easy!
Edit: Also look at how long these (short pre-print) DFT articles are. These aren't simple calculations to interpret.
If you run a bunch of simulations to brute-force-find what you're looking for, you run into a problem of infinitely many things to simulate. It is better to use domain knowledge to eliminate things that don't work and narrow down things that would have high chance of working in theory to give some sort of directional guidance for your research.
That domain knowledge wouldn't have suggested exploring this space for superconductivity.
Turn the material science problem around: instead of looking for a substance that has a specific property, look at many substances until small amounts of any interesting property (young's modulus, etc) show up. By looking for "anything interesting" you are more likely to find something of interesting (ideally, several somethings). And then you also know a starting place to begin optimiziation.
(I'm not saying these things out of ignorance; this technique has worked well for me at times when I had exceptionally large amounts of CPU available to me, and it's also worked well in the drug industry, which has similar problems to material science.)
I think the first insight to pursue LK-99 by the researchers was from the deceased scholar from their graduate school (department chair I believe?). The material was already found in 1999, but they need to try different synthesis methods over 1,000 times for slightly different chemical compositions. I am not sure if there are simulation methods to do that, but it was definitely theoretical insight that first convinced their teacher to start, and the work made the pupils to believe in what they are pursing as far as apparent background stories are concerned.
Keep in mind that a theoretical model agreeing with possible superconductivity is a pretty weak signal. Even more a DFT one.
Real models for superconductivity take a lot of work to create, it's not something people do for an unverified material, and it's not something you get out of DFT. That paper's agreement is more on the lines of "yeah, all superconductors are grey, and this thing is grey, it can be."
But then, they talked about diamagnetism (graphite-like one, I imagine). Honestly, I have no idea how one could disprove (graphite-like) diamagnetism with that simulation, but disproving it is really good news.
I think that at least some parts of this material has superconductivity properties. That's why the videos shown have some very tiny speckles of the material showing the diamagnetism effects and just one anonymous video showing full levitation (type 2 superconductor property as I understood).
