Add `fat bands` to the `BandPdosWidget`. #624

t-reents · 2024-02-21T18:05:25Z

As discussed, I worked on the support for the "fat bands" plots and I managed to merge this into the existing BandPdosWidget. While working on this, I also started to refactor some of the existing methods, see the commit message for more details, mainly to split things up in a more compact way.

I have some further ideas, but I already wanted to share this version, to get further input. One thing that would be useful to address in my opinion, are different linestyles for the spin-up and spin-down projected bands, as the underlying red and blue linecolour is not always easily visible.

Important to note, the Project bands checkbox is only used for this first draft. This would be changed to something more properly formatted in the final version.

One additional question @AndresOrtegaGuerrero: is it really necessary to have these nested loops to iterate each segment separately? At the moment, I don't really see the reason, please correct me if I'm wrong. As mentioned in the commit message, since this version relies on BandsData._get_bandplot_data, one could also plot the bands in a single trace (need to check how much this would improve performance).

aiidalab-qe/src/aiidalab_qe/common/bandpdoswidget.py

Lines 262 to 266 in f64607c

    
           for band in paths: 
        
               if not spin_type: 
        
                   # Non-spin-polarized case 
        
                   for bands in band["values"]: 
        
                       bands_np = np.array(bands)

The current implementation works for non-spin-polarized and spin-polarized calculations. Moreover, the projected bands can be used with and without the PDOS (and all kind of groupings are supported as well).

The orbital/atom contributions in the legend are grouped if the PDOS is plotted as well, so that the chosen contributions can be hidden in both subplots at the same time.

This is the first commit to support `fat bands` in the `BandPdosWidget`, to visualize the contribution of the orbitals directly on top of the bandstructure. The commit contains the following main changes: * `_add_projection_traces` is added to plot the projections on top of the bandstructure. It follows the logic of the existing methods. * A checkbox is added to `BandPdosWidget` to enable the projected bands. To group the orbitals etc., the existing grouping features for the PDOS are used for the projected bands as well. In case PDOS and projected bands are plotted at the same time, the legend is grouped so that the data can be hidden in both subplots simultaneously. * `get_bands_projections_data` is added to extract the projections from a `ProjwfcBandsWorkChain` (see `aiida-wannier90-workflows`). Moreover, the new version relies on `_get_bandplot_data` to extract the bandstructure. In addition to the previously used `_exportcontent` in `export_bands_data`, this method also provides the bands along the full k-path and not only the separate segments. This allows to plot all bands and projections (for each orbital group) in a single trace which should improve the performance. Important to note, the data of the previous method is still accessible. * `_prepare_projections_to_plot` is used to convert the projections into a certain format (see the comments in the implementation for details). This involves some numpy operations to address the previous point of plotting the groups in a single trace. * `_projections_curated_options` is refactored quite a bit. The method supports now the extraction and curation of the PDOS or the projections. Moreover, the different aspects are divided into separate functions to facilitate the maintenance

src/aiidalab_qe/common/bandpdoswidget.py

AndresOrtegaGuerrero · 2024-02-21T20:43:43Z

src/aiidalab_qe/common/bandpdoswidget.py

+    a band needs to be concatenated with its mirror image, first.
+    """
+    projected_bands = []
+    for spin in [0, 1]:


maybe try to use the naming from aiidalab_qe , as collinear

also try to see if this logic is consistent for SOC scenario ?

Not sure, if I fully get your point here. The 0 and 1 are the indicators for the different spin-channels that are used in the BandsData. _get_bandplot_data method. Therefore, I didn't use something like "up" and "down".

But good point, I'll check how these list would look like in case of SOC.

src/aiidalab_qe/common/bandpdoswidget.py

AndresOrtegaGuerrero · 2024-02-21T20:55:39Z

@t-reents thank you ! nice work I had a few comments. First It would be nice to test the logic for all conditions , selecting atoms , orbitals ... and so on , also it is necessary to test a SOC simulation to determine if the logic is preserved. If the plot is only bands with projection then the widgets for displaying the projection should be displayed if there is a projection and hidden if not. Regarding the nested loops, I agree it would be the best to do one single trace , however the way the data in the dictionary was given by traces. Could you please suggest a solution where is only one trace, and double check that you still can select points from the bands (plot)

AndresOrtegaGuerrero · 2024-02-21T20:58:05Z

src/aiidalab_qe/common/bandpdoswidget.py

@@ -397,6 +431,7 @@ class BandPdosWidget(ipw.VBox):
        Select the style of plotting the projected density of states.
        </div>"""
    )
+    projected_bands_width = 0.5


Maybe add this to the SETTINGS dictionary

Makes sense!

one question any particular reason for this number ? Maybe try to test a system with a good amount of atoms and/or kinds , so you see the appearance onced you have plenty of bands

This was just to have a first starting point, to check that everything works.

Not sure, how you think about it (I'm also not a 100% sure, what I think about it, as the layout might get too complicated with too many options), but one could also add a dropdown/slider to choose the maximum linewidth, since the perfect value might differ for different applications?

@t-reents Since the idea is that the widget it has a general behaviour , is important you check with a medium and big system , so you can see how is the appearance, and if you need to maybe add this as a extra widget to control this width

src/aiidalab_qe/common/bandpdoswidget.py

t-reents · 2024-02-21T21:05:04Z

Thanks for your comments @AndresOrtegaGuerrero. I tested all kind of selections and they seem to work. I was also thinking about a SOC calculation but don't have experience with those calculations and no data right now. Do you have a WorkChain to share as an archive, so that I can use it for the development?

Yes, I'll provide a "one trace" solution for the bands tomorrow, and I'll keep in mind that each band/k-point can still be selected.

If the plot is only bands with projection then the widgets for displaying the projection should be displayed if there is a projection and hidden if not

Exactly, I haven't done this yet, but the idea was that the grouping options are alway displayed in case PDOS or band projections are available

AndresOrtegaGuerrero · 2024-02-21T21:11:27Z

For SOC , you can install this plugin to your aiidalab_qe (https://github.com/AndresOrtegaGuerrero/aiidalab-qe-soc/tree/main) , and run a calculation , however for the projections , we need to do a modification , you can check in (https://github.com/AndresOrtegaGuerrero/aiidalab-qe-soc/blob/main/aiidalab_qe_soc/socworkchain.py) , the special settings are in (https://github.com/AndresOrtegaGuerrero/aiidalab-qe-soc/blob/main/aiidalab_qe_soc/workchain.py) , one single thing is to replace the BandsWorkChain for the one you use for the projection

t-reents · 2024-02-21T21:22:24Z

For SOC , you can install this plugin to your aiidalab_qe (https://github.com/AndresOrtegaGuerrero/aiidalab-qe-soc/tree/main) , and run a calculation , however for the projections , we need to do a modification , you can check in (https://github.com/AndresOrtegaGuerrero/aiidalab-qe-soc/blob/main/aiidalab_qe_soc/socworkchain.py) , the special settings are in (https://github.com/AndresOrtegaGuerrero/aiidalab-qe-soc/blob/main/aiidalab_qe_soc/workchain.py) , one single thing is to replace the BandsWorkChain for the one you use for the projection

Thanks, I'll give it a try. Probably I'll just do it via AiiDA-core as I'm used to it

AndresOrtegaGuerrero · 2024-02-21T21:22:30Z

@t-reents Currently i am also working with a modification to the widget here #512 , the new feature is to allow the app to use tot_magnetization. When you use tot_magnetization , you now have two fermi energies in the BandsWorkChain and also in PdosWorkChain. I would suggest you run your workchain adding tot_magnetization , and verify your logic when the data doesnt have "fermi_energy".

src/aiidalab_qe/common/bandpdoswidget.py

Different methods are refactored in this commit: * New functions to create the figures, add the different traces and to generate the final plot. Those replace the previous individual `_create..._plot` methods, to make things more compact * The methods to generate the axes were also refactored to avoid code duplication. * The bands are plotted in a single trace now, improving the performance of the widget

src/aiidalab_qe/common/bandpdoswidget.py

If `tot_magnetization` is specified, the results contain a fermi energy for each spin channel. The general support was added in aiidalab#512 and this commit introduces the required changes to the widget.

AndresOrtegaGuerrero · 2024-04-30T17:55:15Z

src/aiidalab_qe/common/bandpdoswidget.py

+            (False, 0): self.SETTINGS["bands_linecolor"],
+        }
+        fermi_energy_mapping = {
+            (False, 0): self.fermi_energy.get("fermi_energy_up", None),


what happens when there is only one fermi_energy?

The key would be (True, 0) or (True, 1) in that case but those are not present in the mapping. Therefore, the value for fermi_energy would be returned as expected:

aiidalab-qe/src/aiidalab_qe/common/bandpdoswidget.py

Lines 267 to 270 in 25edba6

fermi_energy = fermi_energy_mapping.get(

("fermi_energy" in self.fermi_energy, spin),

self.fermi_energy.get("fermi_energy"),

)

src/aiidalab_qe/common/bandpdoswidget.py

AndresOrtegaGuerrero

LGTM!, We did different tests and check the logic and test the qe app! It also improves the performance

AndresOrtegaGuerrero · 2024-05-02T09:52:09Z

This PR solved #626

AndresOrtegaGuerrero reviewed Feb 21, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Outdated Show resolved Hide resolved

AndresOrtegaGuerrero reviewed Feb 21, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Show resolved Hide resolved

AndresOrtegaGuerrero reviewed Feb 21, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Outdated Show resolved Hide resolved

AndresOrtegaGuerrero reviewed Feb 21, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Outdated Show resolved Hide resolved

AndresOrtegaGuerrero reviewed Feb 21, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Show resolved Hide resolved

t-reents added 2 commits February 22, 2024 14:29

Small improvements of layout and documentation.

2692c01

AndresOrtegaGuerrero reviewed Feb 23, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Outdated Show resolved Hide resolved

AndresOrtegaGuerrero reviewed Feb 23, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Outdated Show resolved Hide resolved

unkcpz mentioned this pull request Feb 29, 2024

Plotly bands plot widget #626

Closed

2 tasks

Fix download method to be compatible with projections.

d3b487a

t-reents force-pushed the feature/fat_bands_plot branch from 0b6d0c5 to d3b487a Compare March 5, 2024 11:14

t-reents and others added 3 commits April 30, 2024 16:11

Merge remote-tracking branch 'origin/main' into feature/fat_bands_plot

fe6c184

Add support for tot_magnetization

b1a26c8

If `tot_magnetization` is specified, the results contain a fermi energy for each spin channel. The general support was added in aiidalab#512 and this commit introduces the required changes to the widget.

Merge branch 'main' into feature/fat_bands_plot

25edba6

AndresOrtegaGuerrero reviewed Apr 30, 2024

View reviewed changes

Fix tests

fd1344c

AndresOrtegaGuerrero reviewed May 2, 2024

View reviewed changes

src/aiidalab_qe/common/bandpdoswidget.py Outdated Show resolved Hide resolved

AndresOrtegaGuerrero added 4 commits May 2, 2024 07:43

fix test plugin electronic_structure

9b4b174

fix logic for displaying pdos_options_out box

49a0dd8

fix logic for displaying pdos_options_out box

b244590

improving readability of logic

58b7432

AndresOrtegaGuerrero and others added 4 commits May 2, 2024 10:23

Merge branch 'main' into feature/fat_bands_plot

17458d1

fix typo

0359b63

typo on display

0b6a5ee

Fix logic to display pdos options

2918240

t-reents marked this pull request as ready for review May 2, 2024 09:20

AndresOrtegaGuerrero approved these changes May 2, 2024

View reviewed changes

AndresOrtegaGuerrero merged commit 4b03549 into aiidalab:main May 2, 2024
14 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add `fat bands` to the `BandPdosWidget`. #624

Add `fat bands` to the `BandPdosWidget`. #624

t-reents commented Feb 21, 2024

AndresOrtegaGuerrero Feb 21, 2024

AndresOrtegaGuerrero Feb 21, 2024

t-reents Feb 21, 2024

AndresOrtegaGuerrero commented Feb 21, 2024

AndresOrtegaGuerrero Feb 21, 2024

t-reents Feb 21, 2024

AndresOrtegaGuerrero Feb 21, 2024

t-reents Feb 21, 2024

AndresOrtegaGuerrero Feb 22, 2024

t-reents commented Feb 21, 2024

AndresOrtegaGuerrero commented Feb 21, 2024

t-reents commented Feb 21, 2024

AndresOrtegaGuerrero commented Feb 21, 2024

AndresOrtegaGuerrero Apr 30, 2024

t-reents Apr 30, 2024

AndresOrtegaGuerrero left a comment

AndresOrtegaGuerrero commented May 2, 2024

	for band in paths:
	if not spin_type:
	# Non-spin-polarized case
	for bands in band["values"]:
	bands_np = np.array(bands)

	fermi_energy = fermi_energy_mapping.get(
	("fermi_energy" in self.fermi_energy, spin),
	self.fermi_energy.get("fermi_energy"),
	)

Add fat bands to the BandPdosWidget. #624

Add fat bands to the BandPdosWidget. #624

Conversation

t-reents commented Feb 21, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AndresOrtegaGuerrero commented Feb 21, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

t-reents commented Feb 21, 2024

AndresOrtegaGuerrero commented Feb 21, 2024

t-reents commented Feb 21, 2024

AndresOrtegaGuerrero commented Feb 21, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AndresOrtegaGuerrero left a comment

Choose a reason for hiding this comment

AndresOrtegaGuerrero commented May 2, 2024

Add `fat bands` to the `BandPdosWidget`. #624

Add `fat bands` to the `BandPdosWidget`. #624