[QE-users] Time-reversal broken in DFT+U calculation without starting_magnetization

Dominik Gresch (Experis AG) a-dogres at microsoft.com
Wed Jan 13 18:28:43 CET 2021


Dear QE community,

I am investigating a system with noncollinear spin and +U correction, but without starting_magnetization. Due to the following sentence in the pw.x input description, I would expect this calculation to exhibit time-reversal symmetry:

In the spin-orbit case starting with zero starting_magnetization on all atoms imposes time reversal symmetry. The magnetization is never calculated and kept zero (the internal variable domag is .FALSE.).

However, the states at k=Gamma are not Kramers pairs (doubly degenerate), indicating that time-reversal symmetry is broken.

To illustrate the behavior on a simplified example, I have modified the Fe calculation from the test suite as follows [1]:

  *   Remove the starting_magnetization and related inputs
  *   Create a 2x2x2 supercell and randomly displace the atoms to break all spatial symmetries (using ASE's rattle method with stddev=0.01 Angstrom) [2]
  *   Reduce the hubbard_U and hubbard_J by a factor of 2, to (presumably?) tune the system out of the ferromagnetic regime [3]

The resulting input files can be found in the attachments, or pasted below. With +U correction [with_u.in], the energies at Gamma [4] are split by ~0.06meV on average, and up to ~0.46 meV. Without +U [without_u.in], the average split is ~3.6e-08 meV, and the maximum is ~3.6e-07 meV.

I would appreciate any input on the following questions:

  *   Does QE take time-reversal symmetry into account explicitly (by symmetrizing the relevant quantities), or implicitly (by selection of symmetry-inequivalent k-points)?
  *   If the latter, is there any way to explicitly enforce time-reversal symmetry?
  *   Is there anything about +U correction in particular that could break time-reversal symmetry (on a technical level, not because it tends to produce magnetic states)?
  *   Or is there anything about this combination of attributes that means this calculation doesn't make sense?

Thank you and best regards,
Dominik

---

Dr. Dominik Gresch

Simulation Engineer
Microsoft Quantum


[1] This isn't expected to be a physically correct calculation.
[2] This is to get rid of any degeneracies enforced by symmetries other than time-reversal.
[3] Without this change the calculation didn't converge without starting_magnetization. This wasn't a problem for the system I am trying to study (which I unfortunately cannot share), but maybe it is somehow related to the reason why time-reversal is broken?
[4] Taking into account only occupied states, since the unoccupied ones are converged to lower accuracy.

Version details: PWSCF v6.6, compiled with GCC v9.2.0, FFTW v3.3.8, MVAPICH2 v2.3.4, Netlib-SCALAPACK v2.1.0, OPENBLAS v0.3.10, run on Zen2 architecture

[with_u.in]
&CONTROL
  calculation = 'scf'
  disk_io = 'low'
  outdir = './out/'
  prefix = 'aiida'
  pseudo_dir = './pseudo/'
  verbosity = 'low'
/
&SYSTEM
  degauss =   1.0000000000d-02
  ecutrho = 280
  ecutwfc = 35
  hubbard_j(1,1) =   8.7500000000d-01
  hubbard_j(2,1) =   0.0000000000d+00
  hubbard_u(1) =   1.1000000000d+00
  ibrav = 0
  lda_plus_u = .true.
  lda_plus_u_kind = 1
  lspinorb = .true.
  nat = 8
  nbnd = 142
  noncolin = .true.
  ntyp = 1
  occupations = 'smearing'
  smearing = 'mv'
/
&ELECTRONS
  conv_thr =   1.0000000000d-08
/
ATOMIC_SPECIES
Fe     55.845 Fe.rel-pbe-spn-rrkjus_psl.0.2.1.UPF
ATOMIC_POSITIONS angstrom
Fe           0.0049671415      -0.0013826430       0.0064768854
Fe           1.4502302986       1.4326584663      -1.4373413696
Fe           1.4507921282      -1.4273256527       1.4303052561
Fe           2.8754256004      -0.0046341769      -0.0046572975
Fe          -1.4325803773       1.4158671976       1.4177508217
Fe          -0.0056228753       2.8598716888       0.0031424733
Fe          -0.0090802408      -0.0141230370       2.8846564877
Fe           1.4327422370       1.4356752820       1.4207525181
K_POINTS automatic
4 4 4 0 0 0
CELL_PARAMETERS angstrom
    -2.8700000000       2.8700000000       2.8700000000
      2.8700000000      -2.8700000000       2.8700000000
      2.8700000000       2.8700000000      -2.8700000000

[without_u.in]
&CONTROL
  calculation = 'scf'
  disk_io = 'low'
  outdir = './out/'
  prefix = 'aiida'
  pseudo_dir = './pseudo/'
  verbosity = 'low'
/
&SYSTEM
  degauss =   1.0000000000d-02
  ecutrho = 280
  ecutwfc = 35
  ibrav = 0
  lspinorb = .true.
  nat = 8
  nbnd = 142
  noncolin = .true.
  ntyp = 1
  occupations = 'smearing'
  smearing = 'mv'
/
&ELECTRONS
  conv_thr =   1.0000000000d-08
/
ATOMIC_SPECIES
Fe     55.845 Fe.rel-pbe-spn-rrkjus_psl.0.2.1.UPF
ATOMIC_POSITIONS angstrom
Fe           0.0049671415      -0.0013826430       0.0064768854
Fe           1.4502302986       1.4326584663      -1.4373413696
Fe           1.4507921282      -1.4273256527       1.4303052561
Fe           2.8754256004      -0.0046341769      -0.0046572975
Fe          -1.4325803773       1.4158671976       1.4177508217
Fe          -0.0056228753       2.8598716888       0.0031424733
Fe          -0.0090802408      -0.0141230370       2.8846564877
Fe           1.4327422370       1.4356752820       1.4207525181
K_POINTS automatic
4 4 4 0 0 0
CELL_PARAMETERS angstrom
     -2.8700000000       2.8700000000       2.8700000000
      2.8700000000      -2.8700000000       2.8700000000
      2.8700000000       2.8700000000      -2.8700000000
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20210113/b93b8546/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: without_u.in
Type: application/octet-stream
Size: 1223 bytes
Desc: without_u.in
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20210113/b93b8546/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: with_u.in
Type: application/octet-stream
Size: 1379 bytes
Desc: with_u.in
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20210113/b93b8546/attachment-0001.obj>


More information about the users mailing list