[Pw_forum] GW for graphene

[Andrea Pedrielli]

Dear users,
we are doing some GW calculation with Quantum Espresso, version 6.1. Our
interest is to compute the GW correction to DFT for high energy states. We
are using the GWL code with C.pbe-mt_fhi.UPF pseudopotential. We are facing
some difficulties, in particular when we plot the GW correction to DFT we
find some spikes (see attached image) that seem to be very unphysical
(corrections of the order of 5 eV ) and depends in a sensitive manner from
the parameters used in the simulations, but the presence of which is a
constant largely independently from the unit cell, the number of bands etc.
In general we are taking as reference this paper https://arxiv.org/pdf/
0806.3365.pdf , in which the GW band structure presents differences with
DFT one smaller than 5 eV at the Gamma point.

I would ask you some questions, the first of them whether some of you see
whatever clear error in our input files (that I attach below).

1) What could be conservative parameters for the graphene cell? In
particular, in the various tutorials on GWL is reported that the number of
atoms (8) for the Silicon example is too small "use at least 64 atoms cells
for production calculations". We tried with a 8 and 18 atoms unit cell and
the spikes in the GW correction remain.  Is the cell dimension a
fundamental parameter, due to the Gamma point GW calculation, though it is
sampled with a fine k point grid?

2) From the working experience with the code seems to be some relations
between the  ecutwfc (kinetic energy cutoff  for wavefunctions)   and
numw_prod (length of the polarizability basis), is it true? In particular
with small ecutwfc we found problems in increasing  numw_prod

3) What could be a reasonable starting point for numw_prod? Until now we
played with this parameter around a value of 100-200, but could it be a too
small value? In the GWL manual is reported that numw_prod scales linearly
with respect to the dimensions of the system, but is there a typical
relation? In the examples it is set to 100 for both Silicon and CH4, but
for 72 atoms (that could be a largely conservative unit cell for graphene)
should be 500, 1000, 2000?


If someone has any idea on the way to go forward, is very welcome.

Thanks in advance,

Andrea Pedrielli

FBK-ECT* Trento
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 1-Gsingle_scf0.inp
Type: application/octet-stream
Size: 1481 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 2-Gsingle_head0.inp
Type: application/octet-stream
Size: 267 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment-0001.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 3-Gsingle_nscf0.inp
Type: application/octet-stream
Size: 1263 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment-0002.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 4-Gsingle_pw4gww0.inp
Type: application/octet-stream
Size: 819 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment-0003.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 5-Gsingle_gww0.inp
Type: application/octet-stream
Size: 1569 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment-0004.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Gsingle_scf0en-50-fhi-df2-c09-bands.dat1
Type: application/octet-stream
Size: 20087 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170704/03b64b7e/attachment-0005.obj>