I am currently performing transport property calculations for PbTe, using mBJ + SOC. For the deformation potential calculations, I utilized a 13 x 13 x 13 Gamma-centered k-mesh to sample the Brillouin Zone. However, I observed that the calculated conduction band minimum (CBM) and valence band maximum (VBM) are located at (0.46, 0.46, 0.46), whereas they are expected to be at (0.5, 0.5, 0.5). Should this discrepancy be a concern for the accuracy of the deformation potential calculations? Should I use a even denser k-mesh during the calculation? If so, what would be an appropriate density?
Additionally, when using AMSET for these calculations, the output indicated the following band structure characteristics:
Input band structure information:
├── # bands: 36
├── # k-points: 2197
├── Fermi level: 5.243 eV
├── spin polarized: False
└── metallic: False
Band gap:
├── direct band gap: 0.468 eV
└── direct k-point: [0.46, 0.00, 0.00]
Valence band maximum:
├── energy: 5.046 eV
├── k-point: [0.46, 0.00, 0.00]
└── band indices: 19, 20
Conduction band minimum:
├── energy: 5.514 eV
├── k-point: [0.46, 0.00, 0.00]
└── band indices: 21, 22
I'm wondering why the positions of the CBM and VBM differ from the ones given by deformation potential calculation. (I used the same KPOINTS for deformation calculations and nscf calculation)
Best regards,
Zane
I would like to know if you have got solution for your query as we are also facing the same. Also should we use same uniform k-grid we used for wavefunction coefficient for elastic+deformation_potential+DFPT calculations? or a non-uniform can also be used for the latter 3? Thanks.
I still don’t know why the position of the CBM and VBM are different in the deformation potential log and amset log.
For your second question, you do not need the same k-mesh for elastic and DFPT calculations, but it is recommended to use the same k-mesh for deformation potential calculation.
Thanks for valuable opinion. You may have a look at this thread on the same issue in case you have missed.
Also, settings.yaml file needs single values of elastic constant and deformation potential. However, we have matrix of elastic constants and deformation potential. Can you please tell if inserting path in file works or any other way possible?
Elastic constant, deformation potentials, and dielectric constants can be expressed either in tensor form or as constants. For the deformation potential, you can simply set the tag to deformation.h5, allowing the code to automatically read the tensors. As for the elastic constant and dielectric constants, they can be extracted from the OUTCAR file of the VASP calculations.
Attached is an example of my settings.yaml file. settings.yaml (1.0 KB)
Thank you for the file and information. I have some more queries, I would be grateful if you could help.
I would like to know whether we should use tetrahedron method to generate uniform band structure for wavefunction coefficient?
Is AMSET now made compatible for 2D materials?
Also, what is the free_carrier_screening input and what calculations one should do for this input?
Should we use tetrahedron method to generate uniform band structure for wavefunction coefficient?
Yes, we should use tetrahedron method for the calculation of wavefunction and the deformation potential.
Is AMSET now made compatible for 2D materials?
I have never studied 2D systems, so I cannot answer this question.
What is the free_carrier_screening input and what calculations one should do for this input?
When the carrier concentration is high, free carriers can effectively screen the perturbations caused by polar optical phonons (POP). Turn on this feature at high carrier concentrations will significantly reduce the POP scattering rate. This was discussed briefly in Phys. Rev. Applied 11, 014058. While there is no clear threshold , I recommend turn on the screening only when the carrier concentration is below 1E18 cm-3.
Thanks for your informative replies. The issue with tetrahedron method arises when we increase k-mesh for nSCF and the grid becomes inconsistent with that of charge density and wavefunction from SCF run. In such case, can Gaussian smearing be an alternative?
As free carrier screening occurs when concentration is high, did you actually want to say “above 1E18 cm-3” instead of below?
For your first question, it’s standard procedure to increase the k-mesh for nscf calculations. As long as VASP successfully reads the WAVECAR and CHGCAR files from the previous scf calculation without generating any error messages, there’s nothing to worry about.
For your second question, I meant “below” in my previous comment. Here’s my understanding:
For semiconductors, carrier concentration is mainly provided by intrinsic point defects and dopants. At high carrier concentrations, high concentration of defects can significantly distort the lattice, leading to increased scattering of charge carriers. (sometimes band structure might not even be the same)
High carrier concentration also enhances electron-electron scattering.
I believe neither of these mechanisms is included in AMSET. Therefore, turning off the screening at high carrier concentrations would overestimate the POP scattering rate, compensating for these simplifications (qualitatively). In my experience, turning on screening consistently results in unrealistically high electrical conductivity (mobility) at high carrier concentrations. Because of this, I recommend only turn on screening when working with systems that have low carrier concentrations.