problem in ypp -s b

[fatimazahra]

Dear all
I ran ypp -s b for graphene, but, I found bad results, In the attached file you find input and output.
I have another question, is this resltats (ypp) with GW approximation or just GGA without GW correction? (I used GGA aprt in Quantum espresso)
thanks

Untitled Folder.tar.gz

[Daniele Varsano]

Dear Fatima,
can you tell most explicitly what you mean for "bad results" ?
From the attached files: at the end of the output there is a mirror of the input file, which is different from the input you posted.

Code: Select all

# .-Input file : ypp.in
# | electrons                    # [R] Electrons (and holes)
# | bnds                         # [R] Bands
# | cooIn= "rlu"                 # Points coordinates (in) cc/rlu/iku/alat
# | % BKpts
# |  0.00     | 0.00     | 0.00     |        # Bands circuit
# | %
# | BANDS_steps= 30              # Number of divisions
# | % BANDS_range
# |   4 |  5 |                   # Bands Range
# | %

please check it.
About the inclusion of the GW corrections, as reported in the documentation, see here,

In order to include previously calculated Quasi-particle correction just type ypp -s b -V qp and then modify the line GfnQPdb= "none" in GfnQPdb= "E < SAVE/ndb.QP" where ndb.QP is the database containing the quasi-particle corrections.

Best,
Daniele

[fatimazahra]

Dr Daniele
I ploted o.bands-sp1 , I found this courbe

k-point.png

This is input that I used

#
# ooooo oooo .. ooo ooo ooooooooo. .oooo.
# `88. .8" .88. `88. .88 `88" `Y8b dP" `Yb
# `88. .8" .8"88. 888b d"88 88 888 88 88
# `88.8" .8" `88. 8 Y88. .P 88 88oooo888" 88 88
# `88" .88ooo888. 8 `888" 88 88 `88b 88 88
# 88 .8" `88. 8 Y 88 88 .88P `8b d8"
# o88o88o 888o8 88 o88bood8P" `Ybod8P"
#
# GPL Version 3.3.0 Revision 1887
# http://www.yambo-code.org
#
bnds # [R] Bands
electrons # [R] Electrons (and holes)
cooIn= "rlu" # Points coordinates (in) cc/rlu/iku/alat
% BKpts
0.00 | 0.00 | 0.00 | # Bands circuit
0.33333333333| 0.333333333 | 0.00 | # Bands circuit
0.00 | 0.50 | 0.00 | # Bands circuit
0.00 | 0.00 | 0.00 | # Bands circuit
%
BANDS_steps=30 # Number of divisions
% BANDS_range
4| 5 | # Bands Range
%


For ypp-s b -V qp, I already tried, but I found the same results that I have already found by ypp-s b
Thank you very much for your help.

[Daniele Varsano]

Dear Fatima,
ypp is performing an interpolation so it depends on your initial k-point grid sampling.
Your plot is not much different from the one reported in the documentation, beside the zero gap region.
That point most probably need to be calculated, and if you miss it in your initial grid I doubt that you can get it.
Check your ground state and try to include high symmetry points in your sampling.

About the GW corrections: did you include this line in your input?

Code: Select all

GfnQPdb= "E < SAVE/ndb.QP"

please note that this valid if you compiled Yambo with the netcdf, otherwise it will be db.QP, check it in your SAVE folder.
A last suggestion: check the convergences of your calculations, I do not know if your calcualtion it is converged with respect k-point, bands in the screening and in the green function, but surely, it is not with respect the block size in the response function, as in the input you posted you have:

Code: Select all

NGsBlkXp= 1 

and clearly your system need more than that.

Best,

Daniele

[fatimazahra]

Dr Danielle
Hello
I read in yambo forum that NGsBlkXp~=1/10*EXXRLvcs,
I ran the yambo -g n -p p, I found that EXXRLvcs= 27451 , should I change this value ?
Thank you for your help and patience.

[Daniele Varsano]

Dear Fatima,

I read in yambo forum that NGsBlkXp~=1/10*EXXRLvcs,

Surely NGsBlkXp is always lower than EXXRLvcs, but how many vector to include it is really system dependent and has to be checked, there is not a strict rule.

I ran the yambo -g n -p p, I found that EXXRLvcs= 27451 , should I change this value ?

By default the code put the maximum number of Gvectors for the Exchange summation, or the last value used in previous calculations. If you do not have memory problems you can keep it to the maximum, otherwise you can reduce it always checking you are not loosing too much in accuracy.

Best,
Daniele

[fatimazahra]

Dr Daniele
how can i know the best value of NGsBlkXp ?
because when I changed NGsBlkXp, E-E0 value also change.
Thanks

[Daniele Varsano]

Sure,
this is a variable that has to be converged, and it is system dependent.
As in all convergence test, you need to raise it until you can, and look when the quantity of interest (E-E0) does not change anymore in according to the accuracy you need. So it is needed to perform different runs increasing the NGsBlkXp value.
Best,
Daniele

[fatimazahra]

Thank you very much for all your help