SOC and U calculation (from wien mailing list)

Gavin Abo Tue, 04 Aug 2015 17:36:43 -0700

 

I don't know everything about DFT+U+SO calculations, but I will try to answer your questions. See below. 

Dear users I am intended to perform FM+U+SOC calculation for Ca2CoSiO7. The step that I understood to perform for such calculations are:

1. run simple SCF with LDA using spin polarized calculations
2. save and run initso_lapw

it asks about magnetization direction, Emax value, add RLO for NONE/ ALL/ CHOOSE elements, spin-polarized case, selection of new structure for SO, Number of kpoints.

3. Import .inorb and .indm from SRC_templates for LDA+U calculation and edit them by requirement.

4. Finally use command runsp_lapw -p -so -orb -cc 0.0001 -ec 0.0001 -NI.

My queries are:
a. Is it the right approach for the mentioned calculation?

Yes, the approach that you mentioned above looks fine.

b. In second step, can we change magnetization direction?

If step 1 came before it, then you can 'set' the magnetization direction in step 2. If you just completed step 4 and are going to back to step 2, I think you need to do a save then a restore (i.e., restore_lapw) of the step 1 calculation before you should 'change' the magnetization direction using initso_lapw. 

how much Emax should be increased (default value is 5 Ry)?

You should be able to increase it until the output parameter you are interested in has converged (i.e., do a convergence test with Emax) [ https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg02113.html ]. 

Which option is correct for adding RLO?

If you don't know when to add them, enter N for NONE for RLO:

https://www.mail-archive.com/wien%40zeus.theochem.tuwien.ac.at/msg08381.html
https://www.mail-archive.com/wien%40zeus.theochem.tuwien.ac.at/msg05958.html
https://www.mail-archive.com/wien%40zeus.theochem.tuwien.ac.at/msg11478.html

As it says in the Wien2k 14.2 usersguide on page 120 [ http://www.wien2k.at/reg_user/textbooks/usersguide.pdf ], RLOs are used for elements with semicore p-states: /We also provide an additional basisfunction, namely a ”relativistic-LO” (RLO) with a// //p1/2 radial wavefunction, which improves the basis and removes to a large degree the dependency// //of the results on EMAX and RMT (see Kuneˇs et al. 2001). It is particular /*/helpfull for heavier atoms/**/ /**/with semicore p-states/*/, but it must not be used for EFG calculations./ Ca, Co, Si, and O should be relatively 'light' elements in the periodic table [ https://en.wikipedia.org/wiki/Chemical_element#Atomic_numbers ]. 

do we need to increase kpoints (before it was 200 for simple scf)?

In the WIEN2k 14.2 usersguide (section 4.5.5 Spin-orbit interaction), it mentions that SO can reduce the symmetry depending on how the direction of magnetization is chosen. In the "Notes about spin-orbit (pdf)" file (novak_lecture_on_spinorbit.pdf) at http://www.wien2k.at/reg_user/textbooks/ (section 4.4), it mentions that the reduction in symmetry has the consequence that the irreducible wedge of the Brillouin zone must be enlarged. So generally yes, the number of k-points need to be increased for the SO calculation. How much should you increase the k-points? You should able to determine how many from a convergence test with k-points [ https://www.beilstein-journals.org/bjnano/content/supplementary/2190-4286-2-45-S1.pdf ]. 

c. Is there any way (like in mBJ we run init_mbj_lapw that imports inm_vresp files and makes changes) to import .inorb and .indm file by command line?

No, I'm not aware of a script like init_mbj_lapw for SO calculations.

You could do programming of your own script to cp and/or edit .inorb and indm files. I put such an example script called init_orb_lapw on my github page [ https://github.com/gsabo/WIEN2k-Patches/tree/master/14.2 ] (click WIEN2k-Patches and Download ZIP). You may try or modify init_orb_lapw at your own risk (it should give some automation to what you have described for step 3). 

d. Is the final command correct?

Yes, the step 4 command looks correct. However, removing the -NI in step 4 might even help you catch a problem; for example, I think if you forget to do a save_lapw, it will indicate that the broyd files exist from a previous calculation and give you a few seconds before it automatically removes and continues with the new calculation. Whereas with -NI, it might continue with previous broyd files and possibly crash the calculation. 

and if someone wants to use GGA+U then simply it is required to run simple SCF by GGA and then apply U or any other method?

Yes, you can simply run the SCF with GGA selected in case.in0 and then apply U. For example, you could probably even do that between steps 1 and 2: 

1a. save -d LDAwithoutSO
1b. Select GGA in case.in0
1c. runsp_lapw
1d. save -d GGAwithoutSO
1e. Create .inorb and .indm
1f. runsp_lapw -orb
1g. save -d GGAUwithoutSO
1h. Select LDA in case.in0

Reference: https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg12781.html