Spinor projection

[Chengcheng-Xiao]

Problem

With --lsorbit and --spin one can plot the spin components along a specific direction (x, y, z). However, If one doesn't use --occL option, the negative values in the projection (corresponds to the spin "down" state) is by default discarded and the plot will miss almost half of the weights.

For example:

Plotted with (BCC Iron):

pyband -y -10 10 --occ '1' --lsorbit --spin z

---

Benefit of fixing this

The only benefit of not using the line collection plot for this use scenario that if we also want to set --spd, then zero weighted points will disappear in scatter plot while in the line collection plot they will show up the same as spin flipping points. i.e.:

Plotted with (BCC Iron):

pyband -y -10 10 --occ '1' --spd 'p'  --lsorbit --spin z --occL --occLC_cbar_vmin -1 --occLC_cbar_vmax 1

---

Proposed solution

Obviously this use scenario is better suited with --occL so the easiest way is to issue an error or select --occL automatically if --lsorbit and --spin option is used together.

So, to enable this feature, one could override the user selected occMarkerColor (and issue a warning) and set the positive values to red and negative values to blue while retaining the weight based size for points by taking the absolute value of the weight. As a side note, I think only use red and blue is enough since these plots already have too many information on them and I doubt people would want to show multiple contributions from atomic orbitals alongside spin informations....

To do this, change these lines in bandplot to:

                if whts is not None:
                    for ii in range(len(opts.occ)):
                        color = ['red' if weight > 0 else 'blue' for weight in whts[ii][Ispin, :, Iband]]
                        ax.scatter(kpath, bands[Ispin, :, Iband],
                                   color=color,#opts.occMarkerColor[ii],
                                   s=np.abs(whts[ii][Ispin, :, Iband]) *
                                   opts.occMarkerSize[ii],
                                   marker=opts.occMarker[ii], zorder=1, lw=0.0,
                                   alpha=0.5)

The proposed solution yields:

Plotted with (BCC Iron):

pyband -y -10 10 --occ '1' --lsorbit --spin z

And:

Plotted with (BCC Iron):

pyband -y -10 10 --occ '1' --spd 'p'  --lsorbit --spin z

If you think this is good enough, I'll submit a pull request and add this use scenario to the README file. 🤓