input of graphite

[zhengjian4210]

Dear Nakib,

I hope this message finds you well.

I am calculating the transport coefficients of graphite (undoped). But I have some issues with the input parameters.

1. I use the Fermi level calculated by epw as enref and chempot. However, due to the overlap of graphite's valence and conduction bands, the setting of the chemical potential inevitably leads to doping. Is my understanding correct?
2. Since graphite is a semimetal, should I enable metallic=.t. ?
3. The calculated charge conductivity (1e7 S/m) seems to be an order of magnitude higher than the experimental results (2.6e6 S/m) (DOI: https://doi.org/10.1103/PhysRev.95.22).
   drag_el_sigma_tot.txt
   input.txt
   output.log

Looking forward to your reply. Thanks.

Best regards,
Jian Zheng

[nakib]

Dear Jian Zheng,

Thank you for your questions.

1. If you have overlapping valence and conduction bands, you inevitably have both p- and n-type doping.
2. Depends on the context. If you would like to have control over the doping (by shifting the chemical potential), you should use the default metallic = .f.. This way, elphbolt will output the electron and hole carrier concentrations for you.
3. It is difficult for me to comment on the difference you are seeing between your calculation and the measurements. But maybe the following questions will help you figure it out by yourself. How good are the samples in the experiment? Are there defects that are not included in your calculation? Are you including all the bands that take part in transport? You mentioned that there are overlapping valence and conduction bands, and yet you have included only the conduction bands in your transport (as is evident from the fact that the hole concentration came out to be zero). Shouldn't you also include bands 7 and 8? Are your k- and q-meshes converged? By the way, I suggest using the tetrahedron method for 3d systems. It looks like the coupled electron-phonon BTEs iterator is unstable, as can be seen from the failure to enforce the Kelvin-Onsager relation.

I hope that was helpful.

Best,
N

[nakib]

Ok, so it looks like in the new method the phonon thermal conductivity is giving physically meaningful numbers and the sign of sigmaS stopped fluctuating wildly. Have you tried a finer set of k and q meshes? I would also increase the maximum number of iterations.

[zhengjian4210]

Okay, I will give it a try.

[zhengjian4210]

I increased the k-grid to three times the q-grid and also raised the number of iterations, but it seems the situation has gotten worse. I will continue to adjust the grid and try.
output.txt

[nakib]

It looks like the phonon system takes a lot of iterations to converge. Have you checked in the literature whether a solution to the phonon BTE in graphite exists? The BTE is not guaranteed to have a solution. If the zone centered ZA modes are carrying most of the heat, I can see why such a system would be difficult to converge. A practical advice would be to include boundary and impurity scattering.

You also might want to check what happens when you increase the q-mesh. Both k and q mesh densities are convergence parameters.

[zhengjian4210]

The phonon Boltzmann equation for graphite has already been solved. The final output of the phonon thermal conductivity results is fairly reasonable, but I am not sure about the electrical conductivity and electronic thermal conductivity. And the zone-centered ZA modes in graphite carry very little heat. I am still not familiar with incorporating boundary and impurity scattering. I will gradually consider them afterward.