Transferred momenta

[zchen]

Hi all,
Recently I am studying the graphene optical spectrum. I have 13 transferred momenta. After running BSE I got 13 out put files. Can I just use the first one to plot the spectrum? I checked two dimension example of yambo. I found QpntsRXd is 1|1 . In another words example only calculate one transferred momenta( the first one and ignore other transferred momenta) and use it to plot the spectrum. Another question is how to involve the GW calculation into the BSE calculation?

Thanks

Zhifan Chen
Dept. of Physics
Clark Atlanta University
Atlanta GA 30314
USA

[Daniele Varsano]

Dear Zhifan,
when calculating the OPTICAL spectrum, you need the q-->0 limit of the
epsilon, this is the reason you need the QpntsRXd is 1|1.
You can find here a little
bit of theory, and I suggest you to study the references indicated in the Documentation.
If you want to study the electron energy loss spectra (EELS) you will have a look to the
finite q vectors.

5n order to include the GW correction of the QP energies, you have different
option:

use the -V qp flag when creating your BSE input file. Then you can either read the QP database
using the variable KfnQPdb
KfnQPdb= "E < ./SAVE/ndb.QP" # [EXTQP BSK BSS] Database

or, in the case the scissor operator is reasonable:
% KfnQP_E
0.000000 | 1.000000 | 1.000000 | # [EXTQP BSK BSS] E parameters (c/v)
%

the first number is the scissor value, while the other two are used if ou want also apply a stretching
to the valence and conduction bands.

Hope it helps,

Cheers,

Daniele

[zchen]

Dr. Daniele:
Thanks


Zhifan Chen
Dept. of Physics
Clark Atlanta University
USA

[zchen]

Dr. Daniele:
In the YAMBO example of two dimension Si(111)2x1 surface there are two group of k points. Is the second group of k points specially for calculating the band structure? When I used k points 6x6x1 for both data sets for my graphene there is no problem. When I used the 6x6x1 k points for the data set 1 and used special k point for band structure calculation for data set 2 there is no problem for Abinit calculation. However, yambo -i -V 1 calculation give the error message : Stop signal received while in [03] Transferred momenta grid [RL indx] 2 equivalent points in the rlu grid found. Can I use 6x6x1 for the second group of k point? Should we calculate all the 13 transferred momenta then choose q goes to 0 (or choose the data where q is smallest)? Is the q goes to 0 always the first group data like the example?

Thanks

Zhifan Chen
Dept of Physics
Clark Atlanta University
Atlanta GA 30314
USA

[andrea marini]

When I used the 6x6x1 k points for the data set 1 and used special k point for band structure calculation for data set 2 there is no problem for Abinit calculation. However, yambo -i -V 1 calculation give the error message : Stop signal received while in [03] Transferred momenta grid [RL indx] 2 equivalent points in the rlu grid found.

Please, post the PW input file you used and the version/revision of the Yambo code you are using. I would like to reproduce your error message first.

[zchen]

Dr. Marini:
We are running Abinit 6.0.3 and yambo-3.2.1-r.448. here is files I used in Abinit calculation. Second data set of k point is obtained by choose the path in BZ for the band structrue calculation.

Thanks

iprcel 45
ndtset 2
kptopt1 1
ngkpt1 6 6 1
shiftk1 0.5 0.5 1.0
kptopt2 0
nkpt2 25
kpt2
0.0000000000 0.0000000000 0.0000000000
0.0000000000 0.0555555556 0.0000000000
0.0000000000 0.1111111111 0.0000000000
0.0000000000 0.1666666667 0.0000000000
0.0000000000 0.2222222222 0.0000000000
0.0000000000 0.2777777778 0.0000000000
0.0000000000 0.3333333333 0.0000000000
0.0000000000 0.3888888889 0.0000000000
0.0000000000 0.4444444444 0.0000000000
0.0000000000 0.5000000000 0.0000000000
-0.0666666667 0.5333333333 0.0000000000
-0.1333333333 0.5666666667 0.0000000000
-0.2000000000 0.6000000000 0.0000000000
-0.2666666667 0.6333333333 0.0000000000
-0.3333333333 0.6666666667 0.0000000000
-0.3030303030 0.6060606061 0.0000000000
-0.2727272727 0.5454545455 0.0000000000
-0.2424242424 0.4848484848 0.0000000000
-0.2121212121 0.4242424242 0.0000000000
-0.1818181818 0.3636363636 0.0000000000
-0.1515151515 0.3030303030 0.0000000000
-0.1212121212 0.2424242424 0.0000000000
-0.0909090909 0.1818181818 0.0000000000
-0.0606060606 0.1212121212 0.0000000000
-0.0303030303 0.0606060606 0.0000000000
# acell in A 2.459 2.459 10.6436, below in bohr
acell
4.646836419553192 4.646836419553192 20.11348842846364
rprim 1.0 0.0 0.0
-0.5 0.8660254037844 0.0
0.0 0. 1.

ntypat 1 # There are 1 types of atom
znucl 6 # The keyword "zatnum" refers to the atomic number of the
natom 4
typat 4*1
xangst
0.000000000 0.000000000 6.810864761
0.000000000 0.000000000 3.832735239
1.229500000 0.709852156 6.816634076
0.000000000 1.419704312 3.826965924
ecut 24.0 # Maximal kinetic energy cut-off, in Hartree
# enunit 2
ixc 1
prtden1 1
nstep1 150
toldfe1 1.0d-6 # Will stop when, twice in a row, the difference
# between two consecutive evaluations of total energy
# differ by less than toldfe (in Hartree)
diemac 5.0 # Although this is not mandatory, it is worth to
# precondition the SCF cycle. The model dielectric
# function used as the standard preconditioner
# is described in the "dielng" input variable section.
# Here, we follow the prescription for bulk diamond.
diemix 0.5
iscf2 -2 # non self-consistency, read previous density file
mkmem 0
mffmem 0
nband2 50
getden2 -1
prtvol 2
tolwfr2 1.0d-9 # it is not important as later there is a diago
# cannot use this nbndbuf2 10
nstep2 100
npwkss2 0 # planewave of kohn-sham state to build Hamiltonian
nbandkss2 50
kssform2 3
istwfk2 20*1



files file is given below

./tgw.in
tgw.out
tgwi
tgwo
tgw
../6c.pspnc

[andrea marini]

Second data set of k point is obtained by choose the path in BZ for the band structrue calculation.

With the posted input file abinit writes to the KSS file (which is read from yambo) the Kohn-Sham levels of the second list of k-points. However Yambo can calculate QP corrections ONLY if the input k-grid is uniform or if contains a uniform subgroup of k-points.

So, the question is, is the second list of k-point uniform ? If this is not the case this explains why Yambo complains.

To calculate QP corrections on a list of non uniform points there a more lenghtly and complicate procedure. Can't you use a regular grid and extract the corrections to other points by interpolation ?

[zchen]

Dr. Marini:
Thanks. The question is back to the original one. Is it necessary to use the special k-point (for band structure calculation) for sencond set of k point if we want to calculate GW and BSE? Are we always choosing the first q data to plot spectrum if we have 13 q points from say 6x6x1 mesh? I think GW calculation need special k point for calculating band structure. Then GW can do the correction to the band structure. What about spectrum?
In my input data the first part of second k set is uniform. About second part I cannot decide. How can we tell easyly it is unform or not?

Zhifan Chen
Dept. of Physics
Clark Atlanta Univ.
Atlanta, GA 30314
USA

[Daniele Varsano]

Dear Zhifan,
if you want to calculate a QP bandstrucutre, as you can see in the documentation,
for each point of your band structure, in order to calculate the QP correction, you
have to perform an integral over the Brillouin Zone. In order to do that you need
an uniform k-point sampling. So, if you want your GW QP band structure along a given path,
you can choose it, provided that the whole grid is uniform. In summary, you need an uniform grid,
and from that you can calculate the GW correction for a given set of k/band points (eventually along the path you need).

Are we always choosing the first q data to plot spectrum if we have 13 q points from say 6x6x1 mesh?

Sorry, but I cannot understand here what you mean.

What about spectrum?

What do you want to know, which spectrum, and in which approximation? Please try to be more explicit.

Cheers,
Daniele

[andrea marini]

Dear Zchen, in your post you pose many different questions, not exactly connected. The general point has been perfectly explained by Daniele. ANY Yambo run needs to find a uniform grid of k-points. To be more precise and answer to your question

About second part I cannot decide. How can we tell easyly it is unform or not?

if you simply do a 3D plot of you points you should see a net without holes. In reciprocal lattice units the k-points should lay on a prefect grid. So, for example, if you pick up points along Gamma-X these points will not be uniform.

The question is back to the original one. Is it necessary to use the special k-point (for band structure calculation) for sencond set of k point if we want to calculate GW and BSE?

Again, if you use a uniform grid (a N1xN2xN3 in abinit), you can perform GW calculations on any k-point of the grid and BSE calculations only at the gamma point.

Are we always choosing the first q data to plot spectrum if we have 13 q points from say 6x6x1 mesh?

If you are referring to the BSE calculation the answer is yes. The first q-point is always Gamma and you can calculate absorption only at gamma.

I think GW calculation need special k point for calculating band structure. Then GW can do the correction to the band structure.

What do you mean as "special" k-point ? If you are thinking to, e.g., the Gamma-X direction then you cannot (easily) make Yambo to calcualte QP corrections on this line. However if you use uniform grids in general (depending on the shift you apply) some high-symmetry points fall in the grid. If I remember properly, for example, in the case of a diamond structure a 4x4x4 grid shifted in the origin includes Gamma, X and maybe L.

What about spectrum?

If you thinking to the absorption spectrum see above and read carefully Daniele's post.