Obtain data for TDOS, PDOS and LDOS at once

ZZL

Hi, I'm new to RESCU Package and I am curious about the electron density of state calculations.

The question is: can we obtain Total DOS, Partial DOS and Local DOS data of any atomic system with one single calculation?

Take Silicon for example, when we did PDOS calculation, we also got TDOS data, theoretically it is because the PDOS is "partial" DOS, simply sums them up gets us to the total DOS.

But why do we need to specify energy when doing LDOS calculation? I noticed that with different ldosEnergy, it rendered different nrgvec arrays. Output data of PDOS calculation, i.e., *_pdos.mat, contains nrgvec values corresponding to ldosEnergy=0 only.

ZZL

It seems that for VASP calculations, set LORBIT=11 can render PDOS and LDOS data simultaneously

vincentm

The LDOS can be obtained at the same time using dos.ldos = true. For example,

info.calculationType = 'dos'
info.savepath = 'results/c2_pdos'
dos.range = [-0.70,0.20]
dos.projL = [0,1]
dos.projM = [-1,0,1]
kpoint.gridn = [45,45,1]
kpoint.shift = [0,0,0]
kpoint.sampling = 'tetrahedron'
LCAO.status = 1
rho.in = 'results/c2_scf'
dos.ldos = true

Since pdos is also activated by the keyword dos.projL, the projected local DOS is also computed. The definitions for each PDOS is explained here. The LDOS and PLDOS are found in the HDF5 file which can be parsed as follows

h5disp('results/c2_pdos.h5')

You'll find them in /dos/pldosVal.

The LDOS in RESCU is defined as in this equation. It's a field defined on a real space grid. Simply put, there is an infinitesimal real space charge at each energy. Some researchers are looking for how electronic charge looks at certain energies. If you sum over space, you get the DOS. If you sum over a certain energy range (mentioned in the link), you get the real space density. One can thus get the LDOS for a certain energy or energy interval.

Reading the VASP documentation on LORBIT, it appears that it is not computing the LDOS like we mean. LORBIT=11 computes the DOS and the site projected wave function character of each orbital. This information can be obtained computing the Mulliken population matrix as instructed here.

Thank you for your question.

ZZL

Thank you very much for your detailed answer.

I understand and agree with your description, but there is still one minor question.

Can we get the spatial charge distribution data like CHGCAR in VASP so that we can visualize the electronic structure?

In Eq.9.3.1, the LDOS is computed by summing over the energy range e1 to e2, if we do not specify the energy range, it seems the LDOS data is thus the charge density at point r. Hence it is equivalent to the CHGCAR data of VASP scf calculation?

vincentm

ZZL Can we get the spatial charge distribution data like CHGCAR in VASP so that we can visualize the electronic structure?

Yes. If you look in say h5disp('results/c2_scf.h5'), you'll find a /rho group and the group state density is found in /rho/rhoOut (some more info here).

In Eq.9.3.1, the LDOS is computed by summing over the energy range e1 to e2, if we do not specify the energy range, it seems the LDOS data is thus the charge density at point r. Hence it is equivalent to the CHGCAR data of VASP scf calculation?

There are a couple if's for that to be true. dos.ldosERange should be specified, and the value should go from negative enough to the Fermi energy. By default, the LDOS is calculated at the Fermi energy, thus giving the charge distribution of states around the Fermi energy. This is why in this example the LDOS looks like p-z orbitals, the dominant character of states near the apex of the Dirac cone.

I note you can get some information from the command line as follows (in MATLAB it will show hyperlinks)

>> rescu --parameter ? ldos


**********************************************************************
There are more than one parameters match the inquiry
**********************************************************************


----------------------------------------------------------------------
Density of states calculator
----------------------------------------------------------------------
    dos.ldos

    dos.ldosEnergy

    dos.ldosEnergyi

    dos.ldosEnergyk

    dos.ldosERange

    dos.ldosShiftEf

ZZL

Thank you for your reply.

It will be better if the RESCU team provide scripts to process the data in .h5 files, for example, to convert LDOS data to CHG or plot PLDOS by selecting atoms and orbitals, see here for examples of VASPKIT. At least make the array value in .h5 file easier to comprehend.

I believe those scripts can help researchers a lot.

vincentm

I forgot to mention, if the system you are studying is not too large, RESCU will save the data to files such as

results/BandStructure.txt
results/DensityOfStates.txt
results/Density.txt
results/LocalDensityOfStates.txt
results/Potential_vdh.txt
results/Potential_vna.txt
results/Potential_vxc.txt

This is because text files are quite demanding in disk resources. The PLDOS is missing, but we'll add it in the next release. Let us know if you prefer working with those. Otherwise, we could look into a CHG converter.

ZZL

Thank you very much. I appreciate your work and look forward to the new release.

Sure, the .txt file is generally large, but if we can process and plot figures on remote HPC, the file size may be just ok.

More importantly, Is there any data structure description of the arrays in .h5 file so that users can understand, for example, what the first row second column value means?

Best regards.

vincentm

ZZL More importantly, Is there any data structure description of the arrays in .h5 file so that users can understand, for example, what the first row second column value means?

You will find some guidelines on grid data here. As for say, PDOS, the meaning of each column is summarize in dos.pdosdef as explained here. That also holds for PLDOS. Let us know if you would like more information on a quantity in particular.