[Pw_forum] Fwd: Noncollinear magnetization and spin orbit
Xiao Shi
xiaos at Princeton.EDU
Mon Mar 3 20:08:34 CET 2014
Hi Giovani
I've already tested those convergence.
For ecut, I tested the convergence on bulk modulus on bulk crystal which is cloes to 27.8GPa from 40/500 Ryd to 30/300 Ryd. Where the suggested ecut in the pseudopotential file is 31/442 (maybe not exactly 442 but close).
For k point, I've tested that 6*6*1 on a 1*1 cell convergence which is the same grid with 3*3*1 on a 2*2 supercell which is what I used
For degauss, I've tested 0.05 to 0.002 all of them won't work
Btw, I've done the calculation of Se2 dimer in the same lattice where it predict the correct spin and bond length and also converge on energy
On Sun, 2 Mar 2014 18:17:03 Giovani Faccin wrote:
>Dear Xiao,
>
>I'm didn't catch the entire discussion thread, so first of all, sorry if
>this might be redundant. Anyway, looking at your input file, three things
>came to mind:
>
>1 - Small cutoff radius (30 Ry / 300 Ry)
>2 - Small number of k-points. (3 3 1)
>3 - Perhaps, a too large degauss value (0.05 Ry).
>
>So, just in case, I should ask: are those values well converged for all
>those pseudopotentials you are using?
>
>If unsure, one simple way to test it is to evaluate dimers of each material
>(O2, Se2 etc.) and run scf tests varying those 3 parameters, to see how the
>total energy behaves. Once you find a setup that works on all the elements
>simultaneously, then you can come back to the full system, using this setup
>as a starting point for further tests. As a hint, for some magnetic
>phenomena a precision of at least 10^-2 Ry in total energy convergence can
>be relevant.
>
>Regards,
Best
Xiao
PS: the original thread is below:
---------
Dear PW users
I'm doing calculations on Bi2Se3 system and get some problem with the magnetization of this system.
As Bi2Se3 is a topological insulator, I need large spin orbit coupling effect which means I need to do noncollinear calculation.
After adding O2 molecule on the surface, I did spin polarization calculation where the energy doesn't converge and the system goes into anti-ferromagnetic like shown below:
total energy = -5045.36386241 Ry
Harris-Foulkes estimate = -5045.36391863 Ry
estimated scf accuracy < 0.00001751 Ry
total magnetization = -0.00 0.00 2.00 Bohr mag/cell
absolute magnetization = 3.42 Bohr mag/cell
If I continue for more steps, it will grows into very large anti-ferromagnetic like 20 Bohr mag/cell for absolute magnetization which is unphysical. The O2 I placed is far away from the surface and there shouldn't be any spin polarization for pristine surface in reality. Also the total energy doesn't change monotonically.
The input file is listed below:
This is a supercell system of Bi2Se3 with 2*2 in the hcp lattice and there are 15 layers of atom for the slab with vacuum larger than 20ans (O2 is at least 10ans away from the slab)
&system
ibrav=4,
celldm(1)=15.992,
celldm(3)=6.555356882,
nat=62,
ntyp=3,
ecutwfc=30.0,
ecutrho=300.0,
occupations='smearing',
degauss=0.002,
noncolin=.true.,
lspinorb=.true.,
starting_magnetization(1)=-0.001,
starting_magnetization(2)=0.001,
starting_magnetization(3)=0.5,
london=.true.
starting_spin_angle=.true.
nosym=.true.
/
&electrons
mixing_beta=0.2
ATOMIC_SPECIES
Bi 208.98 Bi.rel-pbe-dn-rrkjus.UPF
Se 78.96 Se.rel-pbe-n-rrkjus_psl.0.2.UPF
O 16.00 O.rel-pbe-n-rrkjus_psl.0.1.UPF
ATOMIC_POSITIONS {crystal}
O 0.582808125 0.518307602 0.030071968
O 0.438796581 0.511316952 0.033700923
Se 0.333363441 0.666726883 0.270506612
Se 0.333363312 0.166681597 0.270506668
Se 0.833409077 0.166681597 0.270506668
Se 0.833408961 0.666727129 0.270506668
Bi 4E-16 3E-16 0.298539365
Bi 8.34671E-08 0.500045262 0.298539398
Bi 0.500045295 -2.00018E-07 0.298539398
Bi 0.500045179 0.500045262 0.298539398
Se 0.666727117 0.333363441 0.332477455
Se 0.166681746 0.333363493 0.332477478
Se 0.166681629 0.833408578 0.332477478
Se 0.666726948 0.833408578 0.332477478
Bi 0.333363498 0.166681691 0.366247219
Bi 0.833408983 0.166681691 0.366247219
Bi 0.833408867 0.666726943 0.366247219
Bi 0.333363441 0.666726883 0.366247266
Se 3.44477E-08 0.500045237 0.394371017
Se 0.50004532 -1.51013E-07 0.394371017
Se 0.500045204 0.500045237 0.394371017
Se 5E-16 3E-16 0.394371148
Se 0.166681762 0.333363525 0.438001919
Se 0.166681646 0.833408563 0.438001919
Se 0.666726916 0.833408563 0.438001919
Se 0.666727117 0.333363441 0.438001924
Bi 0.333363644 0.166681764 0.466101291
Bi 0.83340891 0.166681764 0.466101291
Bi 0.833408794 0.666726797 0.466101291
Bi 0.333363441 0.666726883 0.466101294
Se 5E-16 4E-16 0.5
Se -1.1658E-07 0.500045162 0.5
Se 0.500045395 5E-16 0.5
Se 0.500045279 0.500045162 0.5
Bi 0.666727117 0.333363441 0.533898706
Bi 0.166681764 0.333363527 0.533898709
Bi 0.166681648 0.83340856 0.533898709
Bi 0.666726914 0.83340856 0.533898709
Se 0.333363441 0.666726883 0.561998076
Se 0.333363642 0.166681762 0.561998081
Se 0.833408912 0.166681762 0.561998081
Se 0.833408796 0.666726799 0.561998081
Se 7E-16 5E-16 0.605628852
Se -2.67539E-07 0.500045087 0.605628983
Se 0.50004547 1.51013E-07 0.605628983
Se 0.500045354 0.500045087 0.605628983
Bi 0.666727117 0.333363441 0.633752734
Bi 0.166681691 0.333363381 0.633752781
Bi 0.166681575 0.833408634 0.633752781
Bi 0.66672706 0.833408634 0.633752781
Se 0.33336361 0.166681746 0.667522522
Se 0.833408928 0.166681746 0.667522522
Se 0.833408812 0.666726831 0.667522522
Se 0.333363441 0.666726883 0.667522545
Bi -3.16627E-07 0.500045062 0.701460602
Bi 0.500045495 2.00018E-07 0.701460602
Bi 0.500045379 0.500045062 0.701460602
Bi 7E-16 6E-16 0.701460635
Se 0.166681597 0.333363195 0.729493332
Se 0.166681481 0.833408728 0.729493332
Se 0.666727246 0.833408728 0.729493332
Se 0.666727117 0.333363441 0.729493388
K_POINTS {automatic}
3 3 1 0 0 0
I wonder has any meet this problem before and can give me some suggestion on it?
PS: I've tried increase degauss to 0.05Ryd and decrease mixing_beta to 0.05 which doesn't help
---------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20140303/05b57cad/attachment.html>
More information about the users
mailing list