Hacker News Reader

Comment by svnt | Hacker News Reader

A dismissal coached by a theoretical physicist (Bernevig [0], the others appear to be students or postdocs) whose job it is to predict superconductors.

At least two reactions come to mind:

1) If his models could predict useful room temperature superconductors, it seems we would have found one already.

2) If his models don’t agree with LK-99 being a superconductor, and it ends up being one, he probably has a lot of revision to do.

I’m not an expert, but doesn’t it seem likely that finding a room temperature superconductor at this point would involve effects or combinations of effects that we don’t have a good theoretical framework for yet?

[0] https://phy.princeton.edu/people/bogdan-bernevig

View on HN

That's a pretty condescending take on what's evidently a paintakingly put-together theoretical+experimental study, from one of the best condensed matter groups in the world.

Believe it or not, researchers more than any other people love the possibility of previously-undiscovered exotic phenomena; these guys are not coming into this with the agenda of enforcing orthodoxy and stifling a potential discovery. But materials still ought to be consistent with what we know about the laws of physics, and it's totally valid to check this via theory and experiment.

To use an example from a different domain, finance, it is completely fair to be skeptical of models and assume there's huge things they may miss. In that case, we have seen models predict things like "this scenario should only happen once in a billion years" - and then occur a few times on consecutive trading days. "Don't overtrust the model" is practically an object lesson at this point, assuming the 'model' here is actually a model and not a literal restatement of physical laws.

> assuming the 'model' here is actually a model and not a literal restatement of physical laws.

I'm not sure there is a difference. Imhu, what we call "physical laws" are always models. Like, Newton's theory of gravity, which predicts that gravity is inverse to the square of the distance. Then Einstein came along and made a better model with relativity theory. Which cannot be the end of the story, because it is incompatible with quantum mechanics...

Well physicists try to make models that explain everything, so once you do get to the end you have the law. In fundamental physics we haven't gotten there but there's probably examples in other areas where we basically don't think it's gonna change.

In finance nobody thinks they have anything approaching a law, just a regularity that somewhat predicts some phenomenon well enough to make some bucks. There's no suggestion that all the effects are captured.

> so once you do get to the end you have the law

There is no end. Laws are just very very well performing models. But every 'law' breaks down if you violate the assumptions of the model: make things too big/small/hot/cold etc.

> There is no end.

You get to the end of whatever evidence you can gather. If you can explain that, you can't (and don't need) more refinements.

I'd say that the true "laws of physics" are things like conservation of mass-energy, CPT/Lorentz invariance, the principle of least action, and such. There aren't many of them, and they are inputs that constrain the models. The Standard Model, for instance, is designed to respect Lorentz invariance, because as far as we can tell, Lorentz invariance is a True And Unchanging Forever Law Of Reality.

As far as we can tell!

> If his models could predict useful room temperature superconductors, it seems we would have found one already.

This sort of model works like "put these atoms in this configuration, find the favourable energy states, look to see if we see signatures that might indicate superconductivity"

They are not "given the whole infinite number of chemicals and configurations of atoms, tell me which ones are superconducting"

> 1) If his models could predict useful room temperature superconductors, it seems we would have found one already.

How do you arrive at this?

Then they go on to say

> I’m not an expert, but..

I’m confused how you think it doesn’t follow? Maybe our conception of model is at differing levels?

I’m saying that if their models worked in a predictive way, since they have been established for so long and the value of confirming them is so high, they would have been confirmed by now.

Think by analogy with NP. Finding the solution (the right configuration of matter) is quite difficult compared to analyzing one that's been put in front of you.

Yes, that is fair. However, I was proposing that it followed we would have found one if we could predict it based on the theory, in effect deriving the matter architecture necessary.

> I’m saying that if their models worked in a predictive way, since they have been established for so long and the value of confirming them is so high, they would have been confirmed by now.

So, just run all possible materials through this hypothetical superconductor predictor model, and any that do exist would be found?

Oh, I see the disconnect. In my imaginary high temperature superconductor predictor, it tells you the confluence of factors required to produce zero resistance, and based on that sketch you go find the combinations of atoms that will stay in roughly that configuration and test them.

I understand that we are not there yet and these models only currently exist in piecemeal/can be used in the negative.

My higher-level intended meaning is that since we have so far been fairly unable to theoretically “find” a room temperature superconductor, if one exists it may rely on an effect for which we don’t have a theoretical model, which means that ruling it out based on existing theoretical models sort of just confirms that we don’t understand how it works, if it works.

Is it that easy? There would be some tractable number of arrangements this would spit out, and from there you can work backwards and find all materials that could form these arrangements, and try to synthesize them and see what works? Might be easier said than done.

I'm reminded of the story of two economists walking down the street. One of them says there's a $20 bill on the ground. The other dismisses the observation, asserting that if there was a $20 bill on the ground, someone would've picked it up already.

> 1) If his models could predict useful room temperature superconductors, it seems we would have found one already.

No such models exist.

From what I understand of the state of modeling here, it's like verifying primes. It's relatively easy to verify whether a specific number is prime or not, but very difficult to generate a new one. Likewise if you know the specific structure of a material, we can model / characterize it but the search space is so large that it's beyond our capabilities to find new candidates.

> It's relatively easy to verify whether a specific number is prime or not

Only the "not" part is easy

> If his models could predict useful room temperature superconductors, it seems we would have found one already.

A model that can check a given solution is not one that can quickly check the entire solution space. For a pure-CS analogy, this is the very definition of P vs NP.

Reply on news.ycombinator.com