Momentum Matrix elements

[manuel_engel1]

Unfortunately, I cannot reproduce your issue. When I run the calculation you have uploaded together with the new INCAR that you have posted, it gives me non-zero entries in the WAVEDERF file. What VASP version are you using to run these calculations?

As a side note, when you use ALGO=exact, there is no need to do run so many electronic minimization steps, so you don't have to specify NELMIN. It could only potentially be relevant for ALGO=normal or similar.

[kaynat_kalvi]

Hi, please update me. Thanks

[manuel_engel1]

Since I cannot reproduce your issue, I'm unfortunately unable to give you any further advice. If you tell me the VASP version that you are using, it might give new insights. You may also upload all the input files of the problematic VASP calculation as a zip archive so that I can verify the problem again. This will make absolutely sure that what I have on my side is identical to what you have on your side. However, with what you have posted so far, I cannot reproduce the issue that you see with all zeroes in the output.

[kaynat_kalvi]

Hi,
Thanks for your responses, i now have a WAVEDERF file, now I am calculating oscillator strength from VasP using (https://www.vasp.at/wiki/index.php/Plot ... ture_of_Si) and using the formula = 2 |P|^2 /(m_e (Ec-Ev)), here P^2 are the sum of squares of last 6 columns of WAVEDERF file, and comparing both the results.
For this comparison, I am confused,
1. the oscillator strength values I got in vasprun.xml file after BSE calculation, corresponds of particular eigenvalue, How do I know that which eigenvalue corresponds to which n,k?
for-example for Si I need the transition from to CB which is n=9 from the VBs which are n=5 6 7 8 (having SOC),at k=0 (Gamma), and I want to locate this in the oscillator strength values I am getting out of vasp. please help me understand this

[manuel_engel1]

I'm glad that you were able to obtain a WAVEDERF file.

1. the oscillator strength values I got in vasprun.xml file after BSE calculation, corresponds of particular eigenvalue, How do I know that which eigenvalue corresponds to which n,k?

The BSE eigenvalues do not correspond directly to Kohn-Sham eigenstates but rather to BSE eigenstates. Therefore, a direct comparison is not generally possible (unless you are in the independent-particle approximation).
The oscillator strengths of the BSE states are ordered together with the corresponding BSE eigenvalues in the XML file. However, there is no k-dependence as these are the oscillator strengths corresponding to some BSE excitations. If you are interested in k-dependence, you can look at the BSEFATBAND file that contains information about the transition energies underlying a particular BSE eigenstate. This is nicely explained in the tutorial that you have linked.

[kaynat_kalvi]

Hi thanks, so i can calculate the oscillator strength from the formula i mentioned earlier, but what are the units of matrix elements written out in WAVEDERF file? In WAVEDERF the last 6 columns corresponds to P_x_real , P_y_real, P_z_real, P_x_img, P_y_img, P_z_img, please comment on this order of P matrix? The unit of energy in vasp is eV and distance is Angstrom? what are the units of electron mass? and effective mass?

[manuel_engel1]

• The matrix elements are written in Angstrom (inverse of k, which is inverse Angstrom).

• The last 6 columns are P_x_real, P_x_imag, P_y_real, P_y_imag, P_z_real, P_z_imag.

• Yes, in VASP we use eV and Angstrom for energy and length, respectively. The rest of the internal units are almost all in atomic units, including mass. However, not all output is necessarily reported in these units.

[kaynat_kalvi]

Hi, If matrix elements are in Angstrom, and energy in eV, then how is oscillator strength dimensionless? shouldn't it be A^2/eV?

I got WAVEDERF file with non-zero matrix elements for both LPEAD = true and LPEAD = FALSE, and the values are different in both cases, why is that? and which one should I use? I think I need to use LPEAD = TRUE, because only then I will derivate the wavefunction w.r.t k, according to Vasp wiki.

Plus some strange behaviour happens, i ran a calculation with certain parameters and got some values for WAVEDERF file, but i ran the same calculation (deleting al files ) and now the wavederf is all zeros in last 6 columns. why is this happening

[manuel_engel1]

It is not possible to use LPEAD=True in this context. As I have explained in one of my earlier posts, you can use LPEAD=False and you will still obtain the derivatives of the orbitals with respect to the Bloch vector in the WAVEDERF file. Could you please point me to the section on the VASP wiki that you refer to? I would like to make sure that there is no mistake in the documentation.

The oscillator strength is indeed dimensionless. I have spoken to one of our BSE experts to confirm this. Maybe he can provide additional information on this subject.

[kaynat_kalvi]

Hi,
here, https://www.vasp.at/wiki/index.php/LPEAD, its written that LPEAD= TRUE will calculate the derivative of cell periodic part of the orbitals, and in this example, https://www.vasp.at/wiki/index.php/Impr ... c_function, the WAVEDER is calculated using LPEAD = TRUE, which is what I have been following.
Regarding, oscillator strength, they should be dimensionless but if the matrix elements are in Angstrom (as you mentioned previously) than I think the Oscillator strength can't be dimensionless, or If you can provide me the formula vasp is using to calculate oscillator strength, It can clear the air.

[pedro_melo]

Hello,

I think a bit more detailed information about what data is written and where can be useful here to identify the units.

• BSEFATBAND file: the 9th and 10th columns are the real and imaginary part of the component of the BSE eigenvector, $X_\lambda^{cvk}$ , for that specific triplet of k-point, valence band, and conduction band. These are dimensionless.

• In the data block "opticaltransitions" inside vasprun.xml what you have is the square of the sum for all possible transitions of the BSE eigenvector multiplied by the corresponding dipole moment for that transition, as specified here (see the element that is squared). For optical transitions, (i.e. q->0), these have the dimension of Angstrom2.

• In WAVEDER you will find the derivatives with respect to k, which as my colleague has explained, are in units of Angstrom.

Hope this helps.
Pedro

[kaynat_kalvi]

Hi
thanks, it helpful, On vasp wiki its written that "optical transitions" its dimensionless, that is why i was confused.
please update on using LPEAD = TRUE tag as well, I need to have the derivative of wavefunctions in WAVEDER file (calling them matrix elements), both LPEAD= TRUE and FALSE, give me different non-zero value of matrix elements, why is that?