Exciton wavefunction vs FFTGvecs
Moderators: Davide Sangalli, andrea marini, Daniele Varsano
-
- Posts: 287
- Joined: Fri Apr 09, 2010 12:30 pm
Exciton wavefunction vs FFTGvecs
Dear developers,
I have plotted the exciton wavefunction in real space using Ypp (version 5.2.0), and I found the default (22586 RL) and reduced (16113 RL) FFTGvecs give the very different results, as attached. Is it normal?
Thanks!
Shudong
I have plotted the exciton wavefunction in real space using Ypp (version 5.2.0), and I found the default (22586 RL) and reduced (16113 RL) FFTGvecs give the very different results, as attached. Is it normal?
Thanks!
Shudong
You do not have the required permissions to view the files attached to this post.
- Daniele Varsano
- Posts: 3868
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Exciton wavefunction vs FFTGvecs
Dear Shudong,
yes, I think it can happen, increasing the FFT grid you are also increasing the real-space grid. Yambo fix the hole in the real space grid in the closest point to the one indicated in the input, so in the two calculations the hole is placed in slightly different points
and this can be the reason of the encountered difference. The position of the hole is reported in the output.
Best,
Daniele
yes, I think it can happen, increasing the FFT grid you are also increasing the real-space grid. Yambo fix the hole in the real space grid in the closest point to the one indicated in the input, so in the two calculations the hole is placed in slightly different points
and this can be the reason of the encountered difference. The position of the hole is reported in the output.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
-
- Posts: 287
- Joined: Fri Apr 09, 2010 12:30 pm
Re: Exciton wavefunction vs FFTGvecs
Dear Daniele,
Thanks for your quick reply. I have checked the output files and the situations as you said. Since the real space distribution is totally different, do we need to converge the FFT vs. the wavefunction distribution? I mean, the reduced FFT can not capture the 'real' exciton features.
Thanks!
Merry Christmas and happy new year in advance!
Shudong
Thanks for your quick reply. I have checked the output files and the situations as you said. Since the real space distribution is totally different, do we need to converge the FFT vs. the wavefunction distribution? I mean, the reduced FFT can not capture the 'real' exciton features.
Thanks!
Merry Christmas and happy new year in advance!
Shudong
- Daniele Varsano
- Posts: 3868
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Exciton wavefunction vs FFTGvecs
Dear Shudong,
I think that the wfs distribution is so different because the hole is placed in different position.
Try to have an FFT grid fine enough to have the hole near the position you want to place it, possibly in a point presenting some
symmetry.
Best,
Daniele
I think that the wfs distribution is so different because the hole is placed in different position.
Try to have an FFT grid fine enough to have the hole near the position you want to place it, possibly in a point presenting some
symmetry.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
-
- Posts: 287
- Joined: Fri Apr 09, 2010 12:30 pm
Re: Exciton wavefunction vs FFTGvecs
Dear Daniele,
I have tested the FFT values vs. the exciton distributions, and I found that the reduced FFT grid gives the same positions as I set in the input of ypp.in. However, when increase the FFT, the position is far from the value I set in the input as below from the log:
...
Extended grid : 348 348 128
<03s> P2: Hole position in the DL cell : 0.00000 4.02759 39.41810 [c.c.] (I set in the ypp.in)
<03s> P2: position in the FFT grid : 0.00000 4.02760 39.17786 [c.c.] (reduced FFT from the default value, which is the same as the set value)
<03s> P2: translated position : 48.83200 88.60709 39.17786 [c.c.]
...
If I leave the FFT in the default, it is
...
Extended grid : 522 522 175
<05s> Hole position in the DL cell : 0.00000 4.02759 39.41810 [c.c.] (I set in the ypp.in)
<05s> position in the FFT grid : 0.38756 4.02759 39.61244 [c.c.] (different from the set value)
<05s> translated position : 49.21954 88.60709 39.61244 [c.c.]
...
Thanks!
Shudong
I have tested the FFT values vs. the exciton distributions, and I found that the reduced FFT grid gives the same positions as I set in the input of ypp.in. However, when increase the FFT, the position is far from the value I set in the input as below from the log:
...
Extended grid : 348 348 128
<03s> P2: Hole position in the DL cell : 0.00000 4.02759 39.41810 [c.c.] (I set in the ypp.in)
<03s> P2: position in the FFT grid : 0.00000 4.02760 39.17786 [c.c.] (reduced FFT from the default value, which is the same as the set value)
<03s> P2: translated position : 48.83200 88.60709 39.17786 [c.c.]
...
If I leave the FFT in the default, it is
...
Extended grid : 522 522 175
<05s> Hole position in the DL cell : 0.00000 4.02759 39.41810 [c.c.] (I set in the ypp.in)
<05s> position in the FFT grid : 0.38756 4.02759 39.61244 [c.c.] (different from the set value)
<05s> translated position : 49.21954 88.60709 39.61244 [c.c.]
...
Thanks!
Shudong
- Daniele Varsano
- Posts: 3868
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Exciton wavefunction vs FFTGvecs
Dear Shudong,
this is quite odd.
In the output file of the wfs you have the real space grid, can you check that the hole position in the larger grid case is indeed the point closer to the one in the input file?
To make this test set consider the unit cell only i.e. ncell (1,1,1).
Best,
Daniele
this is quite odd.
In the output file of the wfs you have the real space grid, can you check that the hole position in the larger grid case is indeed the point closer to the one in the input file?
To make this test set consider the unit cell only i.e. ncell (1,1,1).
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
-
- Posts: 287
- Joined: Fri Apr 09, 2010 12:30 pm
Re: Exciton wavefunction vs FFTGvecs
Dear Daniele,
Below is the log of 1x1x1 cell canculation:
...
<04s> P1: Extended grid : 18 18 175
<04s> P1: Hole position in the DL cell : 0.00000 4.02759 39.41810 [c.c.]
<04s> P1: position in the FFT grid : 0.00000 4.02759 39.61244 [c.c.] (default FFT)
<04s> P1: translated position : 0.00000 4.02759 39.61244 [c.c.]
...
in which the hole position is almost identical. And I tested the number of the cell vs. the hole position, and found when the cell is increased to 25x25x1, the hole position got the problem. The k-point I used in nscf.in is 36x36x1.
Best
Shudong
Below is the log of 1x1x1 cell canculation:
...
<04s> P1: Extended grid : 18 18 175
<04s> P1: Hole position in the DL cell : 0.00000 4.02759 39.41810 [c.c.]
<04s> P1: position in the FFT grid : 0.00000 4.02759 39.61244 [c.c.] (default FFT)
<04s> P1: translated position : 0.00000 4.02759 39.61244 [c.c.]
...
in which the hole position is almost identical. And I tested the number of the cell vs. the hole position, and found when the cell is increased to 25x25x1, the hole position got the problem. The k-point I used in nscf.in is 36x36x1.
Best
Shudong
- Daniele Varsano
- Posts: 3868
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Exciton wavefunction vs FFTGvecs
Dear Shudong,
that's weird, I would need an example to reproduce the problem and inspect it.
Can you provide a minimal example where the problem arise? Please note that it should be independent of the size of the kernel so you can
try with a kernel using just two bands, x only (so no screening) with very few G vectors.
Mnay thanks,
Daniele
that's weird, I would need an example to reproduce the problem and inspect it.
Can you provide a minimal example where the problem arise? Please note that it should be independent of the size of the kernel so you can
try with a kernel using just two bands, x only (so no screening) with very few G vectors.
Mnay thanks,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/