Phononic thermal conductivity from OpenMX

[LauraCaputo]

Dear all,

I am quite new to this topic and I am trying to learn how to properly do calculations on the Figure of Merit of some periodic models.

In particular, due to their periodicity and the zero voltage applied, I am doing Equilibrium calculations as reported here.

After calculating the Transmittance, Seeback coefficient and electronic thermal conductivity, I would like to conclude my study by including also the phononic thermal conductivity. Reading from the manual, something is not quite clear to me: in my calculations, I include two electrodes and a central region, that are equal due to the periodicity. When moving to the calculation of phononic properties, should I still consider the 'full' model with two electrodes? Only the central region is enough?

Thank you!

[ttadano]

The setup described in that manual page is for simulating conductance using NEGF. ALAMODE does not support simulating phonon conductance based on NEGF; it only supports calculating phononic conductivity in the diffusive regime. You cannot mix electronic conductance (device shape dependent) and thermal conductivity (bulk property) for evaluating the figure of merit.

[ttadano]

Yes. I usually use Boltztrap or Perturbo to compute carrier conductivity and Seebeck coefficient and use ALAMODE to compute lattice thermal conductivity.

[LauraCaputo]

Ok, thanks a lot.
One final question: I have read that BoltzTraP calculated the electronic part of the thermal conductivity in function of tau (Ke/tau).
When comparing the values with the ones of ALAMODE, should one multiply the electronic part with a certain value of tau or is this directly summable with Kp calculated by Alamode?

[ttadano]

You will have to estimate Ke, instead of Ke/tau, by assuming a reasonable value for electron tau. There are many papers going beyond this level of approximation (estimate tau using more realistic model), so please check them.

[LauraCaputo]

Ok, thanks!

[LauraCaputo]

Sorry again for writing you here. I have been digging in the Perturbo documentation, since you mentioned it.
I see that in this case, one could actually calculate Ke instead of Ke/tau and one shouldn't approximate on a fixed constant value of tau. This would make it easier to interface directly with Alamode for the calculation of KT. Is it right?