A probable mistake of WenAlloy featurizer

Hi Matminer team!

I am currently using the WenAlloy featurizer to generate features for a high entropy alloy, as outlined in the reference. However, I have noticed that the feature named mixing enthalpy is different from the result that I have calculated manually. Specifically, I have noticed that all of the negative signs are missing from the WenAlloy-generated results, although the numerical values themselves are the same. As such, I believe that the difference between my calculations and the results generated by WenAlloy is due to a mistake of missing the minus sign. Could you please check the generation process and let me know if there is any feedback you can provide? I would greatly appreciate it.

Hi @Carlos2023 , thank you for noticing! If you happen to have a code fix in mind, would you consider making a pull request to the matminer repo?

Hi @ardunn, I’ve pinpointed the issue and believe it warrants a code fix. The source of the discrepancy appears to be the ‘valence_electrons’ column in the Miedema.csv file provided by MatMiner 0.8.0. Specifically, the valence electron counts for B, C, N, Al, Si, and P are currently listed as 13, 14, and 15 respectively, when they should be 3, 4, and 5. I look forward to the next update of Matminer!

Hi @Carlos, thanks for pointing this out! I didn’t have time to update and change the values with updated unittests at the moment, but I added it to the matminer issue tracker! See here: WenAlloy wrong valence electron counts · Issue #909 · hackingmaterials/matminer · GitHub