[borkt]

The fact that our existing computational models are in agreement with the experimental behavior leads me to believe there is no new physics involved here. That being said this raises the question as to whether this class of material is known and characterized within our national research labs but has been kept classified, and if it was unknown it raises the question of why experimentation is what discovered it and why weren't we able to harness the computational model to identify it theoretically and develop a method to produce it practically prior to this. My money says there are people in this world who have long been aware of this, and its either already in use in a classified manner or it's interesting but there are other materials which are more practical in all use cases.

[dotnet00]

I think one factor that's overlooked with these simulations is that there are a ton of parameters to these which really hinder automated searching.

It's easy (relatively) to verify the results from a real world test since you know the physical parameters and can tweak the others based on intuition, where if the result matches the real world you can consider it valid, but if it doesn't you can have to check all sorts of things to be sure that it isn't a glitch due to some parameter not being reasonable.

That makes searching for materials really hard because you either need an absurd amount of computational power to be able to set the simulation parameters so high as to not worry about their effects or you get tons of false positives simply because the computer can't as easily tune those parameters to ensure it produces correct results.

As my PhD advisor has often put it regarding my own simulation work, if the simulations were that capable of modeling reality, there would be no need for billion dollar facilities to perform tests irl, you'd just spend all that on building many supercomputers.

[jacquesm]

> why we [we]ren't we able to harness the computational model to identify it theoretically and develop a method to produce it practically prior to this.

Because the computational requirements are off the scale in the most literal sense. The search space is so large that you won't be able to come up with an improvement in efficiency for your search unless you guide it very carefully with experimentally obtained results and that's exactly what these people were doing as far as I understand it. You mix up a batch of stuff, test it for gross properties, do crystallography and then use the information from that to do some numerical simulations to check if your assumptions and observations hold up.

I don't think we had this compound before.