Classification of bands does not work well with PAW pseudopotentials
I observed that some of bands are not properly classified in some cases depending on pseudopotentials I use. All calculations are done with QE 6.7MaX with fixes !1300 (merged) and !1336 (merged). All the inputs and outputs are attached at the end.
There are two cases at least which I'm aware, Mn3Ir and Mn3Sn. For example, for Mn3Ir, I found four bands (66-69th) are not classified well at Gamma point if I use PAW pseudopotentials
ATOMIC_SPECIES
Ir 192.22 Ir.rel-pbe-n-kjpaw_psl.0.2.3.UPF
Mn1 54.938 Mn.rel-pbe-spn-kjpaw_psl.0.3.1.UPF
Mn2 54.938 Mn.rel-pbe-spn-kjpaw_psl.0.3.1.UPF
Mn3 54.938 Mn.rel-pbe-spn-kjpaw_psl.0.3.1.UPF
that can be obtained from QE website. The output of bands.x run at Gamma point is
**************************************************************************
xk=( 0.00000, 0.00000, 0.00000 )
the magnetic double point group is C_i (-1) [C_i (-1) ]
using the double point group C_i (-1)
there are 4 classes and 2 irreducible representations
the character table:
E -E i -i
G_2+ 1.00 -1.00 1.00 -1.00
G_2- 1.00 -1.00 -1.00 1.00
the symmetry operations in each class and the name of the first element:
E 1
-E -1
i 2
-i -2
Band symmetry, C_i (-1) [C_i (-1) ] magnetic double point group,
using C_i (-1)
e( 1 - 1) = -62.83955 eV 1 --> G_2+
e( 2 - 2) = -62.83913 eV 1 --> G_2+
...
e( 64 - 64) = 20.12733 eV 1 --> G_2+
e( 65 - 65) = 20.22847 eV 1 --> G_2+
e( 66 - 66) = 20.41537 eV 1 --> ?
e( 67 - 67) = 20.42213 eV 1 --> ?
e( 68 - 68) = 20.42292 eV 1 --> ?
e( 69 - 69) = 20.42455 eV 1 --> ?
e( 70 - 70) = 20.43817 eV 1 --> G_2-
e( 71 - 71) = 20.44618 eV 1 --> G_2-
e( 72 - 72) = 20.64627 eV 1 --> G_2+
e( 73 - 73) = 27.21129 eV 1 --> G_2+
e( 74 - 74) = 27.21622 eV 1 --> G_2+
e( 75 - 75) = 29.51427 eV 1 --> G_2+
e( 76 - 76) = 29.62934 eV 1 --> G_2+
e( 77 - 77) = 29.63355 eV 1 --> G_2+
e( 78 - 78) = 29.69724 eV 1 --> G_2+
e( 79 - 79) = 30.31378 eV 1 --> G_2-
e( 80 - 80) = 30.31571 eV 1 --> G_2-
**************************************************************************
On the other hand if I use ultrasoft pseudopotentials,
ATOMIC_SPECIES
Ir 192.22 Ir.rel-pbe-n-rrkjus_psl.0.2.3.UPF
Mn1 54.938 Mn.rel-pbe-spn-rrkjus_psl.0.3.1.UPF
Mn2 54.938 Mn.rel-pbe-spn-rrkjus_psl.0.3.1.UPF
Mn3 54.938 Mn.rel-pbe-spn-rrkjus_psl.0.3.1.UPF
that can also be obtained from QE website, I obtained well classified bands.
**************************************************************************
xk=( 0.00000, 0.00000, 0.00000 )
the magnetic double point group is C_i (-1) [C_i (-1) ]
using the double point group C_i (-1)
there are 4 classes and 2 irreducible representations
the character table:
E -E i -i
G_2+ 1.00 -1.00 1.00 -1.00
G_2- 1.00 -1.00 -1.00 1.00
the symmetry operations in each class and the name of the first element:
E 1
-E -1
i 2
-i -2
Band symmetry, C_i (-1) [C_i (-1) ] magnetic double point group,
using C_i (-1)
e( 1 - 1) = -62.93866 eV 1 --> G_2+
e( 2 - 2) = -62.93827 eV 1 --> G_2+
...
e( 64 - 64) = 20.17319 eV 1 --> G_2+
e( 65 - 65) = 20.27052 eV 1 --> G_2+
e( 66 - 66) = 20.40455 eV 1 --> G_2-
e( 67 - 67) = 20.41111 eV 1 --> G_2-
e( 68 - 68) = 20.42745 eV 1 --> G_2-
e( 69 - 69) = 20.43579 eV 1 --> G_2-
e( 70 - 70) = 20.49786 eV 1 --> G_2+
e( 71 - 71) = 20.49911 eV 1 --> G_2+
e( 72 - 72) = 20.72830 eV 1 --> G_2+
e( 73 - 73) = 27.21503 eV 1 --> G_2+
e( 74 - 74) = 27.22035 eV 1 --> G_2+
e( 75 - 75) = 29.50170 eV 1 --> G_2+
e( 76 - 76) = 29.62672 eV 1 --> G_2+
e( 77 - 77) = 29.62982 eV 1 --> G_2+
e( 78 - 78) = 29.70341 eV 1 --> G_2+
e( 79 - 79) = 30.25430 eV 1 --> G_2-
e( 80 - 80) = 30.25637 eV 1 --> G_2-
**************************************************************************
I also tested with optimized norm-conserving Vanderbilt pseudopotentials that can be obtained from here and got well classified results. I tested with oncv pseudopotentials based on ABINIT database.
I guess there seems a problem with PAW pseudopotentials because the problem only occurs with them. Maybe this is due to my wrong choice of options. So can anyone give an advice how can I obtain well classified bands with PAW pseudopotentials? Or is this a real bug?
I attach inputs and outputs for Mn3Ir.
Mn3Ir | kjpaw | rrkjus | oncv |
---|---|---|---|
scf | scf.in, scf.out | scf.in, scf.out | scf.in, scf.out |
nscf | nscf.in, nscf.out | nscf.in, nscf.out | nscf.in, nscf.out |
bands | bands.in, bands.out | bands.in, bands.out | bands.in, bands.out |
The same problem but a lot more unclassified bands occurs with Mn3Sn. The problem occurs only with PAW pseudopotentials. I also attach inputs and outputs of them.
Mn3Sn | kjpaw | rrkjus | oncv |
---|---|---|---|
scf | scf.in, scf.out | scf.in, scf.out | scf.in, scf.out |
nscf | nscf.in, nscf.out | nscf.in, nscf.out | nscf.in, nscf.out |
bands | bands.in, bands.out | bands.in, bands.out | bands.in, bands.out |
Thanks in advance.
Minkyu Park
University of Ulsan, Republic of Korea