Dear Elk users,
I was noticing that running the same elk.in file with increased k-grids (say 20 20 20 up to 36 36 36 with steps of 4) the values of the electron DOS keep increasing but do not seem to converge, as I would expect instead, and I'm wondering why... is this normal?
Even if they don't differ very much one from the other, I don't see a true convergence:
for example the values I get for the DOS at Fermi level are
k-grid: 20 20 20: DOS(E_F)=23.35510204
24 24 24: 24.12538044
28 28 28: 24.14067055
32 32 32: 24.76787567
36 36 36: 24.96999314
Are there any particular reasons why I should not expect convergence? Or maybe this is telling me that I made a mistake?
Thank you in advance.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
this depends on your choice of swidth. Apparently, it is fairly low and you need an extremely fine k-point mesh to achieve convergence. Try to set it to 0.01 and you'll find convergence faster, but the value will be inaccurate.
Regards
Markus
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Markus,
thank you for your response.
Actually, I hadn't specified the value of swidth because (in fact I haven't understood very well how it works).
However, following your advice, I tried first swidth=0.005 Ha and obtained
20 20 20 22.85787172
24 24 24 23.92042288
28 28 28 24.07798834
32 32 32 24.56283569
36 36 36 24.58553026
then I tried swidth=0.01 Ha and I got
20 20 20 23.82580175
24 24 24 23.30755959
28 28 28 24.45116618
32 32 32 23.78273010
36 36 36 24.20508616
Unfortunately, the values seem to keep oscillating too much.
Would it make sense to increase swidth further? Or should I just increase a lot ngridk?
Thank you again.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Elk users,
I was noticing that running the same elk.in file with increased k-grids (say 20 20 20 up to 36 36 36 with steps of 4) the values of the electron DOS keep increasing but do not seem to converge, as I would expect instead, and I'm wondering why... is this normal?
Even if they don't differ very much one from the other, I don't see a true convergence:
for example the values I get for the DOS at Fermi level are
k-grid: 20 20 20: DOS(E_F)=23.35510204
24 24 24: 24.12538044
28 28 28: 24.14067055
32 32 32: 24.76787567
36 36 36: 24.96999314
Are there any particular reasons why I should not expect convergence? Or maybe this is telling me that I made a mistake?
Thank you in advance.
Dear peterp,
this depends on your choice of swidth. Apparently, it is fairly low and you need an extremely fine k-point mesh to achieve convergence. Try to set it to 0.01 and you'll find convergence faster, but the value will be inaccurate.
Regards
Markus
Dear Markus,
thank you for your response.
Actually, I hadn't specified the value of swidth because (in fact I haven't understood very well how it works).
However, following your advice, I tried first swidth=0.005 Ha and obtained
20 20 20 22.85787172
24 24 24 23.92042288
28 28 28 24.07798834
32 32 32 24.56283569
36 36 36 24.58553026
then I tried swidth=0.01 Ha and I got
20 20 20 23.82580175
24 24 24 23.30755959
28 28 28 24.45116618
32 32 32 23.78273010
36 36 36 24.20508616
Unfortunately, the values seem to keep oscillating too much.
Would it make sense to increase swidth further? Or should I just increase a lot ngridk?
Thank you again.