Metal azides on the convex hull

I’ve noticed that there are several azides in Materials Project which are predicted to be on the convex hull, even though experimentally they decompose explosively. Examples include CaN6 and NaN3, with CaN6 having an experimental decomposition energy of -483 meV/atom (A. D. Yoffe, Proc. Royal Soc. A 208 (1951)) while the calculated value is +478 meV/atom.

  1. Do you know why the discrepancy is so large? Of course, the experimental decomposition happens around 400-500 K, but that doesn’t seem like it would explain such a large difference. The formation energy for Ca3N2 is almost exactly in agreement with experiment, it’s just the azide which is so far off target.
  2. Do you have any recommendations for how I could calculate a more accurate value that is still somewhat compatible with MP parameters, in order to draw a more accurate convex hull?

Thank you

Hi Steven, here’s no answer but just hoping that you’ll get an (admin) reply to your interesting question soon. I have recently asked a question that might be related - concerning the anion correction for N-compounds. Let’s see!

Hello @Steven_Hartman and @Cambrium, thanks for reporting this. We are aware of the issue and investigating, and will report back when we have a good solution for how to resolve this. Thanks for your patience while we look into it.

Hi Steven,

When there are no Hubbard +U corrections, DFT-GGA systematically overbinds peroxides, disulfides, and pernitrides, in other words, the family of complex ions that are isoelectronic to (O2)^2-. Because azides exhibit similar bonding characteristics to pernitrides and diazenides, this DFT overstabilization could plausibly be in effect here as well.

You can find some discussion on this in the supplemental information of my paper in Chemistry of Materials, In that work, I simply removed all azides from hull calculations.

There is a subtlety to your statement that the experimental decomposition energy is -483 meV/atom: If the reaction occurs in air, then the reaction is CaN6 + 0.5 O2 --> CaO + 3 N2, rather than CaN6 --> Ca(metal) + 3 N2. The first reaction will be more exothermic, driven by the formation of CaO, whereas the second reaction is the one that is considered in hull calculations.

Wenhao Sun

Hi Wenhao,

Thanks for the informative reply. I think I will follow your procedure, and exclude axides from the convex hull. The -483 meV/atom was obtained in an argon atmosphere, so it’s likely that the DFT-GGA overbinding is the cause of the discrepancy.