Energy above Hull and Stability

Materials Project
1

Dear all,

Given a material with a nonzero energy above the convex Hull, but exhibiting:

Energetic stability, as confirmed by formation energy minimization with respect to the lattice constant (i.e., negative formation energy);

Dynamic stability, verified through phonon analysis (no negative or imaginary phonon modes);

Mechanical stability, demonstrated by elastic constants satisfying the Born stability criteria; and

Thermal stability, shown via ab initio molecular dynamics (AIMD) simulations (no significant structural fluctuations over extended time periods and elevated temperatures)

can we still consider this material to be stable, despite its nonzero energy above the Hull?

Any insights would be much appreciated.

2

It would be accurate to call this compound metastable with respect to thermodynamic decomposition

3

Thank you for your reply @mattmcdermott . A follow-up question, please: how can we test for metastability of a compound using DFT and ab initio calculations?

I use both Quantum ESPRESSO and CASTEP. Is the energy above the convex hull the only indicator of metastability? If not, what other criteria or methods can be used?

Additionally, how can I calculate the energy above the convex hull using my own settings — specifically, with a chosen exchange-correlation functional and pseudopotentials? I assume that the energy above the hull is sensitive to the computational parameters and settings used.

Any guidance on these matters would be highly appreciated.

4

Hey @Ballmer, these are really more research / background questions than something specific to Materials Project data/software. Take a look at, e.g. this paper to get you started.

Is the energy above the convex hull the only indicator of metastability? If not, what other criteria or methods can be used?

You could look at, e.g., dynamic stability via phonon spectra

I assume that the energy above the hull is sensitive to the computational parameters and settings used.

In short: yes but as long as you compute well-converged energy differences (formation energies for example) using the same pseudopotentials, you can compare across codes/pseudopotentials. The absolute total energy in a pseudopotential calculation has no physical meaning, but energy differences do

1 Like