[SOLVED] PAW carbon surface energy calculation diverges

Dear everyone,

I am trying to compute surface energy of carbon diamond structure (111) using PAW pseudopotential taken from http://www.pseudo-dojo.org/ (Type PAW (JTH v1.0), XC PBE). Everything is ok until I have 18 layers of C(111) or more with vacuum (does not matter how many angstroms of vacuum). Every computation is then oscillating and diverges. The same issue is also for WC(0001) in hexagonal structure. Everything works fine for both structures with NC SR (ONCVPSP v0.4.1), XC PBE pseudopotencial. Input and output files are attached.

I found this issue where was told that abinit is not ready for !! DFTP !! and GGA PAW pseudopotentials and there is a problem even with LDA when structure is 2D. Can this be the same problem?

Thank you

dia.in (2.62 KB)
dia.out (84.9 KB)

Hi Zanto,

your post should be in the ground state section, could you move it there? It is not related to response functions.

Your problem is probably related to preconditioning : try changing dielec and diemix, or iprcel 45 to get a better dielectric preconditioner.

According to the literature, there may be carbon suspended bond on the diamond (111) surface, which needs surface hydrotreatment. In fact, I encountered the same problem when calculating, and finally solved it in the way described in the literature.

@jkend could you provide references and some explanation of what you did?

Dear Mverstra:

When studying the slab model of diamond (111) surface, the structural optimization at the initial stage of calculation, whether it is 2 * 2 or 3 * 3 model, or 3 to 17 layers, can not ensure the stability of the model structure during calculation. After running for a period of time, it will be optimized to graphite structure. I tried to change iovmon = 2 \ 22 \ 7 according to the parameter description, and the diemac parameter to 1000 can not solve the problem well. Later, I read a document about the first principle calculation of diamonds, which adopted the method of surface hydrogenation. After I tried, the calculation converged quickly. You can leave your email and I’ll send it to you. Because there is no DOI.

I see! Ok, indeed H termination is a classic in any surface calculations, also for semiconductors. If the valence is mixed you sometimes need pseudo hydrogens with a charge of 1.5 or 1.25 depends on the valence of each atom. For pure C each bond is just 1 electron so real H is what you need.

Dear Mverstra:
You are right! In fact, after diamond cutting, the surface C presents an unsaturated state, and the purpose of structural optimization is to better find the structure with the lowest energy, which means more stable, so the graphite layer will appear.