G0W0 with PPA tutorial (Si)

[neodiplomat]

Dear YAMBO users and developers,

I am a new YAMBO user. I have tried to reproduce the G0W0 with PPA calculation of Si posted in the tutorial but my calculated DFT and QP gaps are different from the result in the tutorial package (in REFERENCE directory). The highest occupied band and lowest unoccupied band of my calculations are listed as follows.

B=4 Eo= 0.00 E= 0.75 E-Eo= 0.75 Z=0.80 nlXC=-12.90452 lXC=-11.42404 So= 2.41981
B=4 Eo= -0.74 E= -0.07 E-Eo= 0.67 Z=0.80 nlXC=-12.85139 lXC=-11.20135 So= 2.48975
B=4 Eo= -1.17 E= -0.58 E-Eo= 0.59 Z=0.79 nlXC=-13.07348 lXC=-11.16260 So= 2.65722
B=4 Eo= -1.82 E= -1.27 E-Eo= 0.55 Z=0.79 nlXC=-12.85590 lXC=-10.85330 So= 2.70675
B=4 Eo= -2.14 E= -1.65 E-Eo= 0.50 Z=0.78 nlXC=-13.04367 lXC=-10.84231 So= 2.83785
B=4 Eo= -1.29 E= -0.70 E-Eo= 0.59 Z=0.79 nlXC=-12.93063 lXC=-11.07818 So= 2.59831
B=4 Eo= -2.78 E= -2.36 E-Eo= 0.43 Z=0.78 nlXC=-13.21613 lXC=-10.71670 So= 3.04940
B=4 Eo= -3.73 E= -3.33 E-Eo= 0.40 Z=0.77 nlXC=-12.68737 lXC=-10.16253 So= 3.04144

B=5 Eo= 2.85 E= 4.53 E-Eo= 1.68 Z=0.81 nlXC= -5.60625 lXC=-10.11817 So= -2.43262
B=5 Eo= 2.30 E= 3.96 E-Eo= 1.66 Z=0.81 nlXC= -5.69949 lXC=-10.17888 So= -2.43532
B=5 Eo= 1.82 E= 3.43 E-Eo= 1.61 Z=0.81 nlXC= -5.80806 lXC=-10.16837 So= -2.38428
B=5 Eo= 1.45 E= 2.96 E-Eo= 1.51 Z=0.82 nlXC=-5.369830 lXC=-9.514294 So=-2.290576
B=5 Eo= 1.87 E= 3.42 E-Eo= 1.54 Z=0.82 nlXC=-5.064820 lXC=-9.384430 So=-2.430699
B=5 Eo= 2.79 E= 4.48 E-Eo= 1.69 Z=0.81 nlXC=-5.101730 lXC=-9.719190 So=-2.546006
B=5 Eo= 1.07 E= 2.52 E-Eo= 1.45 Z=0.82 nlXC=-5.024210 lXC=-9.153251 So=-2.361035
B=5 Eo= 4.57 E= 6.31 E-Eo= 1.74 Z=0.81 nlXC=-4.939329 lXC=-9.701880 So=-2.611636

Therefore, the DFT gap is 1.07-0.00 = 1.07 eV and the QP gap is 2.52-0.75 = 1.77 eV (is this right?). Whereas the REFERENCE results are

B=4 Eo= 0.00 E= 0.65 E-Eo= 0.65 Z=0.85 nlXC=-12.65043 lXC=-11.32916 So= 2.08036
B=4 Eo= -0.79 E= -0.21 E-Eo= 0.57 Z=0.85 nlXC=-12.61055 lXC=-11.10288 So= 2.18307
B=4 Eo= -1.23 E= -0.79 E-Eo= 0.44 Z=0.84 nlXC=-12.86428 lXC=-11.07633 So= 2.31265
B=4 Eo= -1.95 E= -1.52 E-Eo= 0.42 Z=0.84 nlXC=-12.66393 lXC=-10.75810 So= 2.41224
B=4 Eo= -2.27 E= -1.92 E-Eo= 0.35 Z=0.83 nlXC=-12.85983 lXC=-10.75168 So= 2.52215
B=4 Eo= -1.36 E= -0.91 E-Eo= 0.45 Z=0.84 nlXC=-12.70945 lXC=-10.98330 So= 2.26579
B=4 Eo= -2.93 E= -2.69 E-Eo= 0.24 Z=0.83 nlXC=-13.05775 lXC=-10.63289 So= 2.71766
B=4 Eo= -3.99 E= -3.76 E-Eo= 0.23 Z=0.82 nlXC=-12.52495 lXC=-10.05507 So= 2.74951

B=5 Eo= 2.58 E= 4.44 E-Eo= 1.86 Z=0.85 nlXC= -5.67689 lXC=-10.10042 So= -2.23001
B=5 Eo= 2.08 E= 3.84 E-Eo= 1.76 Z=0.85 nlXC= -5.72960 lXC=-10.16719 So= -2.36291
B=5 Eo= 1.61 E= 3.24 E-Eo= 1.63 Z=0.85 nlXC= -5.84143 lXC=-10.17337 So= -2.40812
B=5 Eo= 1.08 E= 2.60 E-Eo= 1.52 Z=0.85 nlXC=-5.452768 lXC=-9.509406 So=-2.261612
B=5 Eo= 1.48 E= 3.00 E-Eo= 1.52 Z=0.85 nlXC=-5.110384 lXC=-9.355169 So=-2.447804
B=5 Eo= 2.46 E= 4.22 E-Eo= 1.76 Z=0.85 nlXC=-5.120384 lXC=-9.668496 So=-2.473471
B=5 Eo= 0.64 E= 2.01 E-Eo= 1.37 Z=0.85 nlXC=-5.113478 lXC=-9.154862 So=-2.421409
B=5 Eo= 4.25 E= 6.02 E-Eo= 1.76 Z=0.84 nlXC=-5.003822 lXC=-9.718242 So=-2.616991

And the DFT and QP gap is 0.64 and 1.36 eV, respectively. I use the YAMBO Version 3.3.1 Revision 2126 but in the tutorial it is Version 3.3.1 Revision 2126. I find that with the version that i use I cannot do the option "-k hartree" so in my input file the line "Chimod= "Hartree" is missing. Could you please tell me what is the importance of this option and how does it affect the calculated results? Thank you very much

Best,
Suwit

[Daniele Varsano]

Dear Suwit,

I find that with the version that i use I cannot do the option "-k hartree" so in my input file the line "Chimod= "Hartree" is missing. Could you please tell me what is the importance of this option and how does it affect the calculated results?

The tutorial are meant for the 3.4 release, where some input variables changed (e.g. Chimod) with respect previous releases. Of course you should reproduce same results using previous release, by having equivalent inputs, but without seeing your input file I cannot tell you which is the problem. The Chimod=Hartree is needed to include local field effects in the dielectric matrix, and they do affect considerably the results. Anyway, form what I can see from you results, this is not the only discrepancy with the reference results as also the lXC is different and it is independent from the screening. Did you use the tutorial databases or you calculated the ground state by yourself? It looks me that the starting ground state is also different.

My suggestion is to use the 3.4 release. If you downloaded the code from the svn repository, you should have two directory (stable and devel). The devel directory contains the 3.4 release.

Hope it helps,

Daniele

[neodiplomat]

Dear Daniele,

It really helps me a lot. You are right that I calculated the ground state energy by myself since the YAMBO version that I am using cannot read the wavefunction files in the SAVE directory. Anyway, I used the input file from DFT directory to do the non-consistency SCF calculation. Then, I used the p2y script to convert to YAMBO input and followed the instruction in the tutorial. Now that the difference of KS gaps also indicates that the wavefunction from pwscf between my calculation and the tutorials' are different because YAMBO only calculate E (QP energy) not Eo (bare energy). Am I correct?

Thank you very much,
Suwit

[Daniele Varsano]

Derar Suwit,

Now that the difference of KS gaps also indicates that the wavefunction from pwscf between my calculation and the tutorials' are different because YAMBO only calculate E (QP energy) not Eo (bare energy). Am I correct?

Yes, right!

... since the YAMBO version that I am using cannot read the wavefunction files in the SAVE directory.

This is because, the tutorial database are generated using netcdf libraries, if you want to read them , you should compile yambo linking netcdf. You have first to compile the netcdf and then recompile yambo linking to them (configure -H, tells you how to link).

Best,
Daniele

[neodiplomat]

Dear Daniele,

Thank you very much. I will try what you suggest.

Best,
Suwit

[neodiplomat]

Dear Daniele,

I have a follow up question. I am wondering why my pwscf calculation gives the KS band gap very different from the output in the tutorial (1.07 vs 0.64 eV). This time I only compare the the output from pwscf calculation not from YAMBO. I used the same input file in the DFT directory. The pwscf version is also the same (v. 5.0.2). If you know by any chances, could you please tell me? I think this also significantly affects my calculated QP gap. Thank you very much for your kindly helps. Below is the input file.

&control
calculation='nscf',
prefix='Si',
outdir='./'
pseudo_dir= './',
wf_collect=.true.,
verbosity ='high',
/&end
&system
ibrav=2,
celldm(1)=10.183,
nat=2,
ntyp=1,
ecutwfc=15.0,
nbnd=50,
force_symmorphic= .true.,
/&end
&electrons
diago_thr_init = 1.0d-5
/&end
ATOMIC_SPECIES
Si 28.086 Si.vbc.UPF
ATOMIC_POSITIONS
Si 0.00 0.00 0.00
Si 0.25 0.25 0.25
K_POINTS (automatic)
4 4 4 0 0 0

Best,
Suwit

[Daniele Varsano]

Dear Suwit,
Without looking the input/output I cannot say much. Anyway please check if you are looking at the *same* k-points in your output and in the reference one. I do suspect that the difference is due to the k-point sampling: in one of the cases there is a shifted grid:

Code: Select all

K_POINTS (automatic)
 4 4 4 1 1 1

and you have:

Code: Select all

K_POINTS (automatic)
 4 4 4 0 0 0 

Cheers,

Daniele

[neodiplomat]

Dear Daniele,

Sorry for asking so many questions. For both output files of my calculation and tutorial's, I have checked the energy levels of highest occupied and lowest unoccupied bands for every kpoint. I find that my calculated KS gaps at every kpoint are approximately 0.25 eV larger than those of the tutorial's one. The energy levels of the highest occupied and lowest unoccupied bands of my calculation are generally lower than those of the tutorial's. However, the difference of the highest occupied bands are larger leading to differences in band gaps. I also attach input and output files for references. Thank you very much for your helps.

Best,
Suwit

[Daniele Varsano]

Dear Suwit,
this is a non self consistent calculations. Probably your density calculated in the scf cycle is different.
I suggest you to start from scratch and checking everything step by step.

Daniele