ABINIT Discussion Forums

Dear all,

I want to calculate the el. band structure of doped rutile. First, I optimized the structure and then copied the therelaxed coordinates in the input for the el. band calc. However, the output of the 1st dataset of the el. bands calc. gives different total energy (and pressure: around 62 GPa) from the output of the relaxation calc. by 1.7 Ha. Since the relevant input variables are the same in these two calcs, I'm not sure how this happens.

Can anyone help me with this problem?

Thank you in advance,

Yours,

Igor Lukacevic

PS. Inputs are attached.

Hello,

I'm not sure this is relevant but the xc is not exactly the same (ixc 1 and and ixc 3 respectively). Also sometimes, you can end up with something wrong when using getwfk and getxred at the same time (I do not remember if this bug was fixed, and if it was the case only with cell relaxation also, not only internal relaxation). One last thing I noticed is dilatmx which is 1.2 in your relaxation and not defined in you el band struct file. This enlarges your energy cutoff so if your ecut is not converged, this might change your results (but I guess your ecut is converged, is it ? otherwise check it also)

David

Dear David,

thank you for your time.

david.waroquiers wrote:I'm not sure this is relevant but the xc is not exactly the same (ixc 1 and and ixc 3 respectively).

Yes, ixc's are different, but the pseudos are the same (I just forgot to change the ixc value in the el. band input).

david.waroquiers wrote:Also sometimes, you can end up with something wrong when using getwfk and getxred at the same time (I do not remember if this bug was fixed, and if it was the case only with cell relaxation also, not only internal relaxation).

I didn't know about this bug, because I never encountered this kind of problem before. I mean, I always get the same pressure, when I transfer the relaxed parameters to another calculation.

david.waroquiers wrote:One last thing I noticed is dilatmx which is 1.2 in your relaxation and not defined in you el band struct file. This enlarges your energy cutoff so if your ecut is not converged, this might change your results (but I guess your ecut is converged, is it ? otherwise check it also)

Yes, ecut is converged. At least that much, that this difference is just too large. But I'm not sure if I understand how dilatmx changes effectively the ecut value. Would you be so kind to explain it to me?

Cheers,
Igor

Hello,

The effective ecut is increased by dilatmx², so if you set 1.2, it means that you are using an effective ecut of 1.44 x 50 = 77 Ha. That's why I asked if your ecut is converged Another point, try by setting mqgrid (number of point in reciprocal space used to represent the pseudopotential) to a large value instead of the default one reported by abinit (test a value 10 times bigger). If this is changing things, let me know, I have to correct a few things in the estimation of mqgrid.

David

Dear David,

Ah, yes, yes, yes (shame on me). Total energy for ecuts 50 and 100 Ha differs by 0.008 Ha and cell parameters by 0.001 Bohr.

I'll try your mqgrid option.

Igor

Dear David,

I tried both of the options (dilatmx just for make sure that the variables are converged). However, nothing changes. I set mqgrid to 30010 (default is 3001). Again, total energy and pressure differ significantly.

Do you have any other ideas?

Yours,
Igor L.