The problem of convergence at point k

[sunxl]

Dear all:

When I judge the convergence of k-mesh by calculating the spectrum, I use different k point densities from 12*12*1 to 24*24*1 for two-dimensional materials with lattice constant of 3.55 angms. For the first absorption peak, I can always see the tiny displacement of the absorption peak in energy, when k=12, the highest peak is 1.88eV,
k=12, ~1.88eV.
k=14, ~1.98eV.
k=16, ~2.06eV.
k=18, ~2.10eV.
k=20, ~2.16eV.
k=24, ~2.22eV.
Although the energy position of the first absorption peak is biased, it is consistent with the light absorption intensity. Does it look like we still haven't reached the point k of convergence? I appreciate any guidance and tutorial on that!

fig.png

Best,
sunxl

[Daniele Varsano]

Dear sunxl,

please sign your post with your name and affiliation, it is a rule for the forum! You can do this once for all by filling the signature in the user profile.
Yes, it seems that convergence is very slow. My advice is to increase the grid point uniformly in terms of k spacing, ie Nkx and Nky proportional to the unit cell lengths.

Best,
Daniele

[sunxl]

Dear Daniele:

Thank you for your advice. For my material lattice parameter of a=b=3.55 angstroms, do you mean that I should continue to increase k points until convergence?

[Daniele Varsano]

Dear sunxl,

in this case your sampling is uniform, and you will need to increase the sampling if you aim at a tighter convergence.
Best,
Daniele