Page 1 of 2
Increase of accuracy
Posted: Sun Apr 03, 2016 9:40 am
by Ali
Dear All,
This is my first contact with ab-initio calculations so thank you in advance for your understanding and support.
I use EPW-4 and latest QE.
First I calculated the Eliashberg function for Pb as in example (I haven't modified the input files).
Everything completed successfully but the obtained results are not very accurate as compared with the experiment. Please see figure below.
In the next steep I want to increase the accuracy.
In phonons folder I modified the ph.in file. I increased the nq1, nq2 and nq3 values from 6 to 8.
As a results I reached 29 dyn and dvscf files.
Then, in epw folder I calculated scf.in and nscf.in without modyfication and in epw.in I modified the section with nq1, nq2 and nq3 (I increased values from 6 to 8):
Code: Select all
--
&inputepw
prefix = 'pb',
amass(1) = 207.2
outdir = './'
elph = .true.
kmaps = .false.
epbwrite = .true.
epbread = .false.
epwwrite = .true.
epwread = .false.
nbndsub = 4
nbndskip = 5
wannierize = .true.
num_iter = 300
dis_win_max = 21
dis_win_min = -3
dis_froz_min= -3
dis_froz_max= 13.5
proj(1) = 'Pb:sp3'
wdata(1) = 'bands_plot = .true.'
wdata(2) = 'begin kpoint_path'
wdata(3) = 'G 0.00 0.00 0.00 X 0.00 0.50 0.50'
wdata(4) = 'X 0.00 0.50 0.50 W 0.25 0.50 0.75'
wdata(5) = 'W 0.25 0.50 0.75 L 0.50 0.50 0.50'
wdata(6) = 'L 0.50 0.50 0.50 K 0.375 0.375 0.75'
wdata(7) = 'K 0.375 0.375 0.75 G 0.00 0.00 0.00'
wdata(8) = 'G 0.00 0.00 0.00 L 0.50 0.50 0.50'
wdata(9) = 'end kpoint_path'
wdata(10) = 'bands_plot_format = gnuplot'
iverbosity = 0
elinterp = .true.
phinterp = .true.
tshuffle2 = .true.
tphases = .false.
elecselfen = .false.
phonselfen = .true.
parallel_k = .true.
parallel_q = .false.
fsthick = 6 ! eV
eptemp = 0.075 ! K
degaussw = 0.05 ! eV
a2f = .true.
dvscf_dir = '../phonons/save'
nkf1 = 20
nkf2 = 20
nkf3 = 20
nqf1 = 20
nqf2 = 20
nqf3 = 20
nk1 = 6
nk2 = 6
nk3 = 6
nq1 = 8
nq2 = 8
nq3 = 8
/
16 cartesian
0.000000000000 0.000000000000 0.000000000000 0.0092593
-0.166666666667 0.166666666667 -0.166666666667 0.0740741
-0.333333333333 0.333333333333 -0.333333333333 0.0740741
0.500000000000 -0.500000000000 0.500000000000 0.0370370
0.000000000000 0.333333333333 0.000000000000 0.0555556
-0.166666666667 0.500000000000 -0.166666666667 0.2222222
0.666666666667 -0.333333333333 0.666666666667 0.2222222
0.500000000000 -0.166666666667 0.500000000000 0.2222222
0.333333333333 0.000000000000 0.333333333333 0.1111111
0.000000000000 0.666666666667 0.000000000000 0.0555556
0.833333333333 -0.166666666667 0.833333333333 0.2222222
0.666666666667 0.000000000000 0.666666666667 0.1111111
0.000000000000 -1.000000000000 0.000000000000 0.0277778
0.666666666667 -0.333333333333 1.000000000000 0.2222222
0.500000000000 -0.166666666667 0.833333333333 0.2222222
-0.333333333333 -1.000000000000 0.000000000000 0.1111111
And I received the following error:
Code: Select all
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
task # 0
from readmat_shuffle2 : error # 1
wrong qpoint
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Please tell me what I do wrong?
Best,
Ali
Re: Increase of accuracy
Posted: Sun Apr 03, 2016 12:33 pm
by sponce
Dear Ali,
Please also update your q-grid in the EPW input file.
Code: Select all
16 cartesian
0.000000000000 0.000000000000 0.000000000000 0.0092593
-0.166666666667 0.166666666667 -0.166666666667 0.0740741
-0.333333333333 0.333333333333 -0.333333333333 0.0740741
0.500000000000 -0.500000000000 0.500000000000 0.0370370
0.000000000000 0.333333333333 0.000000000000 0.0555556
-0.166666666667 0.500000000000 -0.166666666667 0.2222222
0.666666666667 -0.333333333333 0.666666666667 0.2222222
0.500000000000 -0.166666666667 0.500000000000 0.2222222
0.333333333333 0.000000000000 0.333333333333 0.1111111
0.000000000000 0.666666666667 0.000000000000 0.0555556
0.833333333333 -0.166666666667 0.833333333333 0.2222222
0.666666666667 0.000000000000 0.666666666667 0.1111111
0.000000000000 -1.000000000000 0.000000000000 0.0277778
0.666666666667 -0.333333333333 1.000000000000 0.2222222
0.500000000000 -0.166666666667 0.833333333333 0.2222222
-0.333333333333 -1.000000000000 0.000000000000 0.1111111
I guess there should be 29 points that are different from those.
Best,
Samuel
Re: Increase of accuracy
Posted: Sun Apr 03, 2016 2:27 pm
by Ali
Dear Samuel,
thank you for your quick response. Please just tell me how obtain the last column?
Best,
Ali
Re: Increase of accuracy
Posted: Sun Apr 03, 2016 4:08 pm
by roxana
Hi Ali,
You can simply run a scf calculations with your desired q-mesh and the coordinate and weight of each points is listed after "number of k points" line.
As an alternative you can use kpoints.f90 program that is provided in QE/PW/tools.
Best,
Roxana
Re: Increase of accuracy
Posted: Sun Apr 03, 2016 10:44 pm
by Arnie
Dear Roxana,
I used kpoints.f90 program in the following way
Code: Select all
bravais lattice >> 2
filout [mesh_k] >> 29
mesh: n1 n2 n3 >> 8 8 8
mesh: k1 k2 k3 (0 no shift, 1 shifted) >> 0 0 0
write all k? [f] >> f
# of k-points == 29 of 512
and as a result a I have:
Code: Select all
0.0000000 0.0000000 0.0000000 1.00
-0.1250000 -0.1250000 0.1250000 8.00
-0.2500000 -0.2500000 0.2500000 8.00
-0.3750000 -0.3750000 0.3750000 8.00
-0.5000000 -0.5000000 0.5000000 4.00
0.0000000 0.0000000 0.2500000 6.00
-0.1250000 -0.1250000 0.3750000 24.00
-0.2500000 -0.2500000 0.5000000 24.00
-0.3750000 -0.3750000 0.6250000 24.00
-0.5000000 -0.5000000 0.7500000 24.00
-0.6250000 -0.6250000 0.8750000 24.00
-0.7500000 -0.7500000 1.0000000 12.00
0.0000000 0.0000000 0.5000000 6.00
-0.1250000 -0.1250000 0.6250000 24.00
-0.2500000 -0.2500000 0.7500000 24.00
-0.3750000 -0.3750000 0.8750000 24.00
-0.5000000 -0.5000000 1.0000000 12.00
0.0000000 0.0000000 0.7500000 6.00
-0.1250000 -0.1250000 0.8750000 24.00
-0.2500000 -0.2500000 1.0000000 12.00
0.0000000 0.0000000 1.0000000 3.00
-0.2500000 0.0000000 0.5000000 24.00
-0.3750000 -0.1250000 0.6250000 48.00
-0.5000000 -0.2500000 0.7500000 24.00
-0.2500000 0.0000000 0.7500000 24.00
-0.3750000 -0.1250000 0.8750000 48.00
-0.5000000 -0.2500000 1.0000000 24.00
-0.2500000 0.0000000 1.0000000 12.00
-0.5000000 0.0000000 1.0000000 6.00
When I run epw calculations I have error:
Code: Select all
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
task # 0
from createkmap : error # 1
q-vec not commensurate
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Here you can see all my input files:
Code: Select all
https://drive.google.com/open?id=0B1BjLap8K2-oeU5PRlBzWVRWWVU
I will be very grateful if you can help me with this problem.
Best,
Ali
Re: Increase of accuracy
Posted: Sun Apr 03, 2016 10:59 pm
by roxana
Dear Ali,
The number is column for needs to be divided by 512 (the total number of points).
To check this use my first suggestion and run a scf calculations.
I would also recommend that for questions related to QE you consult the QE tutorials and forum where you can find answers to many questions/problems.
Best,
Roxana
Re: Increase of accuracy
Posted: Mon Apr 04, 2016 11:42 am
by sponce
Dear Ali,
The last column is the weight of the q-point.
It is not the issue here.
You should provide the exact same list of q-points as during the phonon calculations you did (8x8x8).
Look at the output of your ph calculation, the q-point list should be written.
Also, in the phonon folder, don't forget to run the python script pp.py again as you need to put the dvscf and dyn files in the save folder.
If everything match, it should work.
Best,
Samuel
Re: Increase of accuracy
Posted: Mon Apr 04, 2016 10:19 pm
by Ali
Dear Samuel and Roxana,
thank you for your help and sugestions.
Now, I have good determined q-grid in the EPW input file but still I have the error:
Code: Select all
from createkmap : error # 1
q-vec not commensurate
I think that this may be connected with nscf.in file and section K_POINTS crystal.
I know that I need 512 k-points but I am not sure how to determine them. I try use kpoints.f90 (I try reproduce results for 216 points (6x6x6) like in your example) but I obtain wrong results (inconsistent with your results).
Please tell me how you calculate K_POINTS crystal in nscf.in.
And one more question. Are these parameters are good:
Code: Select all
nkf1 = 20
nkf2 = 20
nkf3 = 20
nqf1 = 20
nqf2 = 20
nqf3 = 20
nk1 = 6
nk2 = 6
nk3 = 6
nq1 = 8
nq2 = 8
nq3 = 8
or maybe I should increase nk1, nk2, nk3 to 8?
Thank you very much,
Ali
Re: Increase of accuracy
Posted: Tue Apr 05, 2016 8:01 am
by carla.verdi
Hi Ali,
First of all the k and q grids need to be commensurate, i.e. the k grid needs to be the same or a multiple of the q grid: therefore, if the q grid you are using is an 8x8x8, yes you should increase nk1, nk2 and nk3 to 8.
In the k_points section in the nscf file you will need to provide the 512 k points in crystal coordinates which is just a uniform mesh within [0:1).
Best
Carla
Re: Increase of accuracy
Posted: Sat May 28, 2016 7:01 pm
by Vahid
Dear EPW users,
I am running EPW-4 on TiO2 following the steps outlined in PRL115,176401 (2015). I get the error "q-vec not commensurate". According to the above posts, the k-grid should be the same or a multiple of the q-grid. However, in the PRL paper, the DFPT run uses a 4x4x4 q-grid while the BZ was sampled using a 6x6x6 k-grid. These grids are not commensurate. My guess is that the PRL paper used a different coarse k-grid but I may be missing something else here. I would appreciate any suggestions you may have.
Thank you,
Vahid
Vahid Askarpour
Department of Physics and Atmospheric Science
Dalhousie University
Halifax, NS Canada