The Flat Bands Are Real (Probably)

The internet has been in full superconductor derangement mode for about a week now, and I've been watching with the measured calm of someone who has seen cold fusion, perpetual motion, and at least three "graphene will fix everything" cycles.

But.

Sinéad Griffin at Lawrence Berkeley ran DFT calculations on LK-99 — the copper-doped lead phosphate apatite that a Korean team claims superconducts at room temperature and ambient pressure — and the results are annoying in a specific way. When you substitute copper for lead in that crystal structure, you get isolated flat bands right at the Fermi level. Two of them. Sitting there.

Flat bands are what you get when electrons can't really move — when the energy of a state barely changes with momentum, which means effective mass goes to infinity, which means kinetic energy loses its grip, which means electron-electron interactions dominate. Strongly correlated physics. The kind of thing that, under the right conditions, gives you superconductivity. Or Mott insulation. Or some other exotic state of matter that will have its own Wikipedia disambiguation page someday.

This is not proof that LK-99 is a superconductor. Griffin is careful about this. Flat bands are a necessary condition in certain theories, not a sufficient one. You can have flat bands and still get nothing interesting. But the calculation didn't need to find anything — the whole exercise could have come back as "structurally incoherent, go home" — and instead it found something that requires explaining.

The thing that makes this specific result harder to dismiss than the original paper is where it came from. This isn't the hopeful team that made the material. This is an independent group, no dog in the fight, using VASP and the Vienna convention of "we're going to be boring and rigorous about this," and the boring rigorous answer is: the electronic structure has features consistent with the reported properties.

You can argue about whether the flat bands survive correlation corrections. You can argue about whether the material anyone is actually making in a lab has the stoichiometry Griffin modeled. You can argue about electron-phonon coupling not being computed here. These are real arguments.

What you can't do — as of today, August 1st, while replication attempts are still being sorted out in labs across three continents — is say the theory is obviously wrong. The computational physics is not embarrassing. That's more than you could say this time last week.

We're still waiting. But we're waiting on something that passed at least one filter it could have failed.