Error with HSE06 and ALGO=Damped
Posted: Sun Jul 12, 2015 4:26 am
I am using HSE06 to calculate Si band structure with ALGO=Damped. I use a primitive cell with two atoms with a0=5.43A. BZ-sampling uses the 6x6x6 Gamma centered scheme. The following is the INCAR setting after the usual PBE SCF computation:
SYSTEM=Diamond-Si
ISTART=1
ENCUT=500
EDIFF=1.0E-7
NPAR=24
ICHARG=0
ISMEAR=0;SIGMA=0.01
GGA=PE
NSW=0
IBRION=-1
PREC=Accurate
PRECFOCK=Accurate
RWIGS=1.312
#LREAL=Auto
LWAVE=T
LCHARG=T
LHFCALC=.TRUE.;LFSCREEN=0.2;AEXX=0.25
ALGO=Damped; TIME=0.4; LDIAG=.TRUE.
NBANDS=24
NELMIN=10
I found the energy levels at the GAMMA point are different for cases with a nonzero weight and a zero weight:
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.4629630E-02
1 -8.253518
2 5.135292
3 5.135292
4 5.135292
5 9.147447
6 9.147447
7 9.147447
8 10.158161
9 14.637926
10 14.652046
11 14.652046
12 18.198122
13 18.198122
14 18.198122
15 22.105391
16 30.649416
17 30.649416
18 30.649416
19 31.949594
20 31.949594
21 33.511650
22 33.511658
23 33.511658
24 36.681558
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.0000000E+00
1 -8.253517
2 5.641836
3 5.641842
4 5.641837
5 9.147466
6 9.147446
7 9.147449
8 10.158160
9 14.652544
10 14.637926
11 14.652139
12 18.200481
13 18.198138
14 18.198637
15 22.107929
16 30.908348
17 30.925680
18 31.248338
19 32.203957
20 32.325096
21 33.690107
22 34.997818
23 37.239843
24 43.657886
If I changed the ALGO=Normal for HSE06 computation:
INCAR:
SYSTEM=Diamond-Si
ISTART=1
ENCUT=500
EDIFF=1.0E-7
NPAR=24
ICHARG=0
ISMEAR=0;SIGMA=0.01
GGA=PE
NSW=0
IBRION=-1
PREC=Accurate
PRECFOCK=Accurate
RWIGS=1.312
#LREAL=Auto
LWAVE=T
LCHARG=T
LHFCALC=.TRUE.;LFSCREEN=0.2;AEXX=0.25
ALGO=Normal; LDIAG=.TRUE.
NBANDS=24
NELMIN=10
I found the almost same values for the two the energy levels at the GAMMA point for cases with a nonzero weight and a zero weight:
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.4629630E-02
1 -8.253622
2 5.135155
3 5.135158
4 5.135161
5 9.147400
6 9.147401
7 9.147402
8 10.158043
9 14.637642
10 14.651982
11 14.651982
12 18.198022
13 18.198026
14 18.198031
15 22.105039
16 30.649262
17 30.649269
18 30.649275
19 31.949496
20 31.949496
21 33.511471
22 33.511473
23 33.511475
24 36.681518
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.0000000E+00
1 -8.253622
2 5.135155
3 5.135158
4 5.135161
5 9.147400
6 9.147401
7 9.147402
8 10.158043
9 14.637642
10 14.651982
11 14.651982
12 18.198022
13 18.198026
14 18.198031
15 22.105039
16 30.649262
17 30.649269
18 30.649275
19 31.949496
20 31.949496
21 33.511471
22 33.511473
23 33.511475
24 36.681519
0.0000000E+00 0.2500000E-01 0.2500000E-01 0.0000000E+00
1 -9.308461
2 4.012665
3 4.038898
4 4.038904
5 9.115732
6 9.191436
7 9.191438
8 10.213760
9 14.548280
10 14.664824
11 14.690839
12 18.218068
13 18.218076
14 18.264482
15 22.193258
16 30.472006
17 30.472017
18 30.662434
19 31.790132
20 31.843957
21 33.584013
22 33.625005
23 33.625009
24 36.770328
The energy levels at other points in the BZ only show slightly changes using ALGO=Damped and ALGO=Normal.
So two questions rises: What is the correct Si indirect bandgap in HSE06 using VASP. Using ALGO=Normal I got a value of 1.82 eV, which is very different
from J . Paier's reported value (1.3eV using HSE06). Using ALGO-Damped, and the valence top at GAMMA point 5.64 eV, I got 1.3 eV for the value of the indrect bandgap. However, it is clearly seen the 5.64 eV valence top (case of zero weight) is not consistent with the value 5.135 eV ( case of nonzero weight).
I put the questions in the region of Physical questions, but no one relys. I hope the admin can give me an explanation.
By the way, the above results are computed with Si-PBE pseudopotential. I have also tried Si-GW pseudopotential, the trend is the similar.
Thanks a lot!
Paul
SYSTEM=Diamond-Si
ISTART=1
ENCUT=500
EDIFF=1.0E-7
NPAR=24
ICHARG=0
ISMEAR=0;SIGMA=0.01
GGA=PE
NSW=0
IBRION=-1
PREC=Accurate
PRECFOCK=Accurate
RWIGS=1.312
#LREAL=Auto
LWAVE=T
LCHARG=T
LHFCALC=.TRUE.;LFSCREEN=0.2;AEXX=0.25
ALGO=Damped; TIME=0.4; LDIAG=.TRUE.
NBANDS=24
NELMIN=10
I found the energy levels at the GAMMA point are different for cases with a nonzero weight and a zero weight:
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.4629630E-02
1 -8.253518
2 5.135292
3 5.135292
4 5.135292
5 9.147447
6 9.147447
7 9.147447
8 10.158161
9 14.637926
10 14.652046
11 14.652046
12 18.198122
13 18.198122
14 18.198122
15 22.105391
16 30.649416
17 30.649416
18 30.649416
19 31.949594
20 31.949594
21 33.511650
22 33.511658
23 33.511658
24 36.681558
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.0000000E+00
1 -8.253517
2 5.641836
3 5.641842
4 5.641837
5 9.147466
6 9.147446
7 9.147449
8 10.158160
9 14.652544
10 14.637926
11 14.652139
12 18.200481
13 18.198138
14 18.198637
15 22.107929
16 30.908348
17 30.925680
18 31.248338
19 32.203957
20 32.325096
21 33.690107
22 34.997818
23 37.239843
24 43.657886
If I changed the ALGO=Normal for HSE06 computation:
INCAR:
SYSTEM=Diamond-Si
ISTART=1
ENCUT=500
EDIFF=1.0E-7
NPAR=24
ICHARG=0
ISMEAR=0;SIGMA=0.01
GGA=PE
NSW=0
IBRION=-1
PREC=Accurate
PRECFOCK=Accurate
RWIGS=1.312
#LREAL=Auto
LWAVE=T
LCHARG=T
LHFCALC=.TRUE.;LFSCREEN=0.2;AEXX=0.25
ALGO=Normal; LDIAG=.TRUE.
NBANDS=24
NELMIN=10
I found the almost same values for the two the energy levels at the GAMMA point for cases with a nonzero weight and a zero weight:
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.4629630E-02
1 -8.253622
2 5.135155
3 5.135158
4 5.135161
5 9.147400
6 9.147401
7 9.147402
8 10.158043
9 14.637642
10 14.651982
11 14.651982
12 18.198022
13 18.198026
14 18.198031
15 22.105039
16 30.649262
17 30.649269
18 30.649275
19 31.949496
20 31.949496
21 33.511471
22 33.511473
23 33.511475
24 36.681518
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.0000000E+00
1 -8.253622
2 5.135155
3 5.135158
4 5.135161
5 9.147400
6 9.147401
7 9.147402
8 10.158043
9 14.637642
10 14.651982
11 14.651982
12 18.198022
13 18.198026
14 18.198031
15 22.105039
16 30.649262
17 30.649269
18 30.649275
19 31.949496
20 31.949496
21 33.511471
22 33.511473
23 33.511475
24 36.681519
0.0000000E+00 0.2500000E-01 0.2500000E-01 0.0000000E+00
1 -9.308461
2 4.012665
3 4.038898
4 4.038904
5 9.115732
6 9.191436
7 9.191438
8 10.213760
9 14.548280
10 14.664824
11 14.690839
12 18.218068
13 18.218076
14 18.264482
15 22.193258
16 30.472006
17 30.472017
18 30.662434
19 31.790132
20 31.843957
21 33.584013
22 33.625005
23 33.625009
24 36.770328
The energy levels at other points in the BZ only show slightly changes using ALGO=Damped and ALGO=Normal.
So two questions rises: What is the correct Si indirect bandgap in HSE06 using VASP. Using ALGO=Normal I got a value of 1.82 eV, which is very different
from J . Paier's reported value (1.3eV using HSE06). Using ALGO-Damped, and the valence top at GAMMA point 5.64 eV, I got 1.3 eV for the value of the indrect bandgap. However, it is clearly seen the 5.64 eV valence top (case of zero weight) is not consistent with the value 5.135 eV ( case of nonzero weight).
I put the questions in the region of Physical questions, but no one relys. I hope the admin can give me an explanation.
By the way, the above results are computed with Si-PBE pseudopotential. I have also tried Si-GW pseudopotential, the trend is the similar.
Thanks a lot!
Paul