Hello Adriano,

For the calculation of the orbital decomposed DOS you need to add the switch

orbcomp="T"
 

to the banddos input like described here
https://www.flapw.de/MaX-5.1/documentati...omposed-charges

If you're still getting the same error or you already did this could you please provide the input file you used? If you are having other trouble with the plotting in
masci-tools you can also post your issues in the masci-tools github repository (e.g. as a comment to this issue https://github.com/JuDFTteam/masci-tools/issues/75).
The plotting of other kinds of DOS is not tested well at the moment so issues could appear.

Best,

Henning

Dear Kirill

The "tetra" mode needs no special handling in the inpgen. You just generate a grid, which includes the gamma point. The tetra option in the inpgen is just a leftover, from an earlier version of this implementation. Then in the inp.xml you make the modifications exactly as you describe, setting the mode to "tetra" and selecting the kpointList you generated.

Yes this mode works in self-consistent and force theorem modes .

Best,

Henning

Hello,

The 'tria' mode is not generating a regular mesh for tetrahedron integration. It generates the given number of kpoints in the IBZ in a non-uniform distribution that is then divided into tetrahedra. This mode is not really meant for accurate SCF calculations but for generating a nice density of states without having to use many kpoints and has been in this state for a long time.

Additionally in the Max4 version especially due to a bug the total number of kpoints, 1296 in your case, is generated in the irreducible(!!) brillouin zone. This is why the runtime skyrockets when you move to larger kpoint sets. You can use the tag <kpointCount count="1000"/> for example to circumvent this bug.

In versions following MaX4 a new mode 'tetra' was introduced which will do the tetrahedron decomposition like you expect it.

I hope this is helpful.

Henning

@Gregor At the moment this routine uses a rather naive approach, the color is based on which of the specified weights is largest at each eigenvalue. I implemented this mainly because someone approached me about doing such a plot for the orbital decomposition. So for example distinguishing px,py and pz by color. I'm sure there is a better way to decide the dominant character, but as a first step I was happy with it.

Hi,

Just to add to this discussion. I now implemented a plotting routine for bandstructures, which is close to what you are looking for I think. If I understand you correctly you want a bandstructure plot, where each band is colored according to the dominant band character (has the largest weight) at that point. The following code block shows this for highlighting the characters of the first atom. You'll need the newest available version of the masci-tools package for this

from masci_tools.io.parsers.hdf5 import HDF5Reader
from masci_tools.io.parsers.hdf5.recipes import FleurBands
 
from masci_tools.vis.fleur import plot_fleur_bands_characterize
 
filepath='banddos.hdf'
 
with HDF5Reader(filepath) as h5reader:
   data, attributes = h5reader.read(recipe=FleurBands)
 
# We need to define the weights we want to highlight and their associated color
weightNames = ["MT:1s","MT:1p","MT:1d","MT:1f"]
weightColors = ["blue","green","red","yellow"]
 
# Plot the bandstructure and save to a file bandstructure.png
ax = plot_fleur_bands_characterize(data, attributes, weightName, weightColors, limits={'y': (-10, 5)}, show=False, save_plots=True)
 
 

Note that this routine is by no means finished (for example points where none of the selected characters are dominant can be colored inconsitently) and feedback is welcome

Henning

Hello,

The latest release had some bugfixes for the Noco setup added shortly before releasing it. Unfortunately, I only noticed that some additional adjustment for the LDA+U part was needed after these fixes after the release was done.

Could you try out the newest version from the develop branch to see if you get a better convergence behaviour there?

Let me know if you have further questions or problems,

Henning Janßen