Building on 's work with the fine-tuned MatterGen model, we evaluated 400 candidate materials for superconductivity using the Tc classification model. Prior work on fine-tuning:
Using a 3DSC published superconductor dataset we fine-tuned MatterGen to enable critical temperature property conditioned generation of 'S.U.N' crystal structures.The 3DSC dataset was intentionally de
The finetune doesn't seem to have learned much about the magnitude of critical temperature, or what causes it. With a maximum found Tc of 17.5 K, this is significantly lower than what we find when exploiting known superconducting families. That said, it discovered whole new families of superconductors not seen before.
Even the best superconducting families we know of, like Ba-Ca-Cu-Hg-O have low Tc (and non-superconducting) materials in them. Knowing what families are worth searching is half the battle. We now have completely new families we can explore more of.
With luck, we may have found a family here that doesn't suffer from the same break down of superconductivity as the temperature nears room temperature.
Of the 400 samples generated,
82 had some level of superconductivity
Mostly binary and ternary systems, with a few quaternary
6 materials with Tc greater than 10 K
52 new chemical systems discovered
You can find the crystal structures generated here:
400 .cif files of candidate structures property condition generated by MatterGen where tc = 298.15K
And the evaluation results:
Evaluation results for the MatterGen fine-tuned model candidates, with new superconducting families labeled.
While no room temperature superconductors were found, we did find some new families we might find one in. Some promising ones worth exploring further:
Co-Nb
As-Mo-Ru
Bi-Pb-Pt
P-Os-Sr
Bi-Cd-La
Au-Hg-Zr
Ge-Pt-Zr
C-Mo-Sn
Al-Os-Zr
Na-Pt
Li-Nd-Si