Hello,
I am a happy ELK user of a few years and I find myself stepping outside the realm of the thoroughly tested calculation protocols. Specifically, I want to test various functionals from libxc in pursuit of a realistic band structure for various oxides and then calculate magnetic effects in these materials.
I recently happened across the metaGGA functionals and thought this might be an appropriate platform to test them on. However, to calculate the effects that I am interested in, it would seem that inclusion of SOC is necessary to obtain realistic results. My most recent compilation of ELK (3.0.4), downloaded last week, gives me the following message:
Error(init0): meta-GGA is not valid for non-collinear magnetism
Does anyone have any information about the extent to which the code has been developed to utilize the libxc functions? For instance, if I recompile and force the code to accept my libxc choices, am I likely to break something or just get nonsense results? Is there an easy or at least reasonable workaround for this? I am hoping that any solution to this problem could likewise be implemented for other libxc functionals as necessary with the appropriate checking of results.
Regards,
Nik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
First let me point out the that when you switch on spin-orbit the system becomes non-collinear by definition. Most meta-GGA functionals are designed for collinear case only. So in trying to use a collinear functional in a non-collinear case you are making an error.
However, if you still want to do this here is what you do: first converge your calculation without spin-orbit (i.e. collinear with spins in z-direction). Then start with tasks =1 and set cmagz=.T.
cmagz=.T. forces your system to have moment pointing in z-direction (i.e. forces collinearity) despite spin-orbit.
Best
Sangeeta
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Evidently I need to look into the theory of the meta-GGA functionals a little more. However, in the intervening time, I simply commented out ELK's objections, recompiled and got what I would interpret to be a meaningful result (at least from the gap, need to check on the exact band structure). I still am unsure how to interpret this result since you tell me that the functional itself cannot accommodate non-collinearity.
Thank you for your restart suggestion. I thought this might be an option but was unsure exactly how to have it set up so that the code would not give me errors.
Thanks again,
Nik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
I am a happy ELK user of a few years and I find myself stepping outside the realm of the thoroughly tested calculation protocols. Specifically, I want to test various functionals from libxc in pursuit of a realistic band structure for various oxides and then calculate magnetic effects in these materials.
I recently happened across the metaGGA functionals and thought this might be an appropriate platform to test them on. However, to calculate the effects that I am interested in, it would seem that inclusion of SOC is necessary to obtain realistic results. My most recent compilation of ELK (3.0.4), downloaded last week, gives me the following message:
Error(init0): meta-GGA is not valid for non-collinear magnetism
Does anyone have any information about the extent to which the code has been developed to utilize the libxc functions? For instance, if I recompile and force the code to accept my libxc choices, am I likely to break something or just get nonsense results? Is there an easy or at least reasonable workaround for this? I am hoping that any solution to this problem could likewise be implemented for other libxc functionals as necessary with the appropriate checking of results.
Regards,
Nik
Hi Nik,
First let me point out the that when you switch on spin-orbit the system becomes non-collinear by definition. Most meta-GGA functionals are designed for collinear case only. So in trying to use a collinear functional in a non-collinear case you are making an error.
However, if you still want to do this here is what you do: first converge your calculation without spin-orbit (i.e. collinear with spins in z-direction). Then start with tasks =1 and set cmagz=.T.
cmagz=.T. forces your system to have moment pointing in z-direction (i.e. forces collinearity) despite spin-orbit.
Best
Sangeeta
Evidently I need to look into the theory of the meta-GGA functionals a little more. However, in the intervening time, I simply commented out ELK's objections, recompiled and got what I would interpret to be a meaningful result (at least from the gap, need to check on the exact band structure). I still am unsure how to interpret this result since you tell me that the functional itself cannot accommodate non-collinearity.
Thank you for your restart suggestion. I thought this might be an option but was unsure exactly how to have it set up so that the code would not give me errors.
Thanks again,
Nik