ABINIT Discussion Forums

Hello fellow Abinit users!

My question here relates to the nwchan command in anaddb.

Does having a larger channel width increase the accuracy of my calculations or is the goal to have the smallest channel widths (which is what I assume).

So for instance if I am looking at the thermal displacement matrix out of anaddb (B matrix). Should I be looking at the matrix computed with the highest or lowest channel width if I want to get the most accurate results.

Thank you in advance!

Hello - this is indeed not all that clear in the doc:

there are several convergences being done simultaneously, with respect to q-grid density, for nwchan different widths of the bins in frequency (wide bins make a coarse bar graph DOS, and converge faster in q-grid). All of the widths from nwchan down to 1cm-1 are done for each q-grid, and convergence stops once all of the bin size DOSs are converged between 2 successive q-grids (total number of grids = ngrids).

What really should determine your precision is ultimately the max grid density given by ng2qpt. I'm not sure that nwchan is very useful, and it doesn't let you, e.g., go more precise than 1cm-1 bins...

Thank you mverstra, I really appreciate your output!

I am still a little confused about the channel widths though. Since each channel width gives a different answer for the B-matrix, and they all seem to have converged for the same ng2qpt which one is the most reliable?

Thank you so much for you reply!

If you get your B matrix from the DOS, the smallest channel size should give you the most accurate answer, and things should be converging wrt channel size as well. Ideally you don't want to depend on the DOS but just on the ng2qpt - I don't know how you get the B-matrix