ERROR: Missing bands in file eqp_co.dat

593 views
Skip to first unread message

sma...@mtu.edu

unread,
Sep 12, 2018, 1:44:19 PM9/12/18
to BerkeleyGW Help
Hello,

I am trying to get GW-bandstrcuture on MnO bulk system using interpolation code. I did the tutorial with Si  and was able to repeat for GaAs. I have successfully generated and epsilon and sigma for this system. But at the interpolation step, I am  getting this error : ERROR: Missing bands in file eqp_co.dat

I have eqp_co.dat that contains 1 to 40 bands.  I am using WFN_co from the same wave function that I used to get sigma (i.e WFN_inner=WFN_co). I have asked to get 100 bands (this is just a test) but asked only 1-40 bands for vxc.dat My WFN_fi is along the band-path -- I tried few cases with maximum number of bands in WFN_fi to be 40, 70, 100 etc. None of them worked.  What am I missing here ?  The error says missing bands in eqp_co.dat, --does it mean that I have less number of bands in WFN_co than WFN_fi ? 
Could you please explain this error ?

Please let me know if you need any more information. 

I would appreciate your help. 
Thanks ! 

--Subhasish

Zhenglu Li

unread,
Sep 12, 2018, 1:47:33 PM9/12/18
to sma...@mtu.edu, BerkeleyGW Help
Hi Subhasish,

When you do interpolation, you can only interpolate to a number of bands that is smaller than what you have directly calculated in sigma with the coarse grid.  So you should ask for a number < 40.  Also, the smaller number you asked, the better interpolation you get, as you are using more bands to interpolate fewer bands.

Best,
Zhenglu

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.
To view this discussion on the web visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/980ca1e4-3ce4-4842-ba4b-c45d56a8c80b%40berkeleygw.org.

Subhasish Mandal

unread,
Sep 12, 2018, 3:02:23 PM9/12/18
to Zhenglu Li, BerkeleyGW Help
Hi Zhenglu,

Thanks for your quick reply. This was a great help. In this case, if I use 38 bands, it still has the same error. If I use a very small number, say 30 ( I have 24 bands filled), it seems to work. I have an additional question in this regards: Do I always have to ask from 1st band to n-th band for sigma (or vxc.dat) in order it to use eqp_co.dat for interpolation? Or I can ask only a set of bands say 20-27 band, for example ? 
The "number_cond_bands_coarse" or "number_val_bands_fine" I give in the input file for interpolation code --does it count from the Fermi level ?  Do I have to specify anywhere the Fermi level since I will be plotting bands in EF +/- X eV ?  Or does the code already take care of that ? 

Thanks again! 

--Subhasish



To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

Zhenglu Li

unread,
Sep 12, 2018, 3:10:09 PM9/12/18
to sma...@mtu.edu, BerkeleyGW Help
Hi Subhasish,

I'm not sure about 38 bands, it could be due to degeneracy in your case, so the smaller number works.

For vxc.dat, no you don't have to calculate from band 1.  You just need to specify the bands of interest, i.e. that you asked sigma to calculate.  So yes, you can ask only 20-27 bands.

number_cond/val_bands_coarse/fine are counted from the number of bands in your DFT calculation.  So band 1 is the lowest energy band you get from your DFT (depending on pseudopotential, of course).  You don't have to specify Fermi level, the code determines that automatically.  But the code does not use smearing, so if you want an accurate Fermi level, you can specify Fermi level.

Best,
Zhenglu

To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

Subhasish Mandal

unread,
Sep 12, 2018, 4:08:54 PM9/12/18
to Zhenglu Li, BerkeleyGW Help
Hi Zhenglu,

Thanks again ! 

For spin-polirzed system, the bandstructure.dat contains bands for both up and down spin.  Does the plot_inteqp.py program take care of spin ? i.e how doe it plot bands for spin1 and spin2 ? I do not see any spin entry in the python code.  The one I have got for spin-polarized calculation looks weird, even for DFT bands.  I think it is related to the spin. 
Where can I find a plot tool that would include spin ? Or is there any flag in interpolation code to interpolate spin-wise so that the bandstructure.dat would be printed on the desired spin ? 

Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

Zhenglu Li

unread,
Sep 12, 2018, 4:22:16 PM9/12/18
to sma...@mtu.edu, BerkeleyGW Help
Hi Subhasish,

In principle, the spin-polarized case is treated as two subsets of bands.  I'm not very familiar with plot_inteqp.py, so maybe other people can give some suggestions.  But one needs to be careful when using spin-polarizations, because epsilon code assumes time-reversal symmetry.

Best,
Zhenglu


Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

Subhasish Mandal

unread,
Sep 12, 2018, 4:45:44 PM9/12/18
to Zhenglu Li, BerkeleyGW Help

Hi Zhenglu,

Thanks for your reply.  
the plot_inteqp.py could be easily modified with spin entry. 
I do not give any spin info in epsilon code --is it correct ? I give spin_index_max 2 only in sigma code and I get vxc.dat, eqp_co.dat for both the spin. Does it sound correct ? 
Is there any example of GW band structure calculations for spin-polarized system ? It would be really helpful. 

Thanks ! 
--Subhasish


Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

Zhenglu Li

unread,
Sep 12, 2018, 4:53:25 PM9/12/18
to sma...@mtu.edu, BerkeleyGW Help
Hi Subhasish,

That is great that you can modify the script.  The time-reversal symmetry in intrinsically assumed in epsilon, so in principle any spin-polarization cases (FM, AFM, etc.) would be incorrect.  However, it's up to how much the magnetic structure will modify the dielectric screening (in some cases, the effects might be small).  Currently, I think most current GW codes apply the same assumption (but I'm not very sure about this).  Full epsilon without using time-reversal symmetry is part of an ongoing project within BerkeleyGW group.

Hope it helps!
Zhenglu

Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.
To view this discussion on the web visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/CAP0F6N-hoiiLm%3Dv0vye5WhOrR4To3hn1cn_P-fFMb6%3D8-WLn7w%40mail.gmail.com.

Subhasish Mandal

unread,
Sep 12, 2018, 7:22:13 PM9/12/18
to Zhenglu Li, BerkeleyGW Help
Hello Zhenglu,

Thanks a lot for your clarification. I am a little confused. In the manual there is a reference ( https://journals.aps.org/prb/pdf/10.1103/PhysRevB.66.035102 ) with spin-polarized calculation. Is this feature not available presently in the software ? 

Thanks ! 

--Subhasish   

Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.
 To view this discussion on the web visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/CAOn4XLjcxL0rsjor7DV9MNif%2Bek3xu7xuZfP2hiB6TAhvU5iBg%40mail.gmail.com.



--

Zhenglu Li

unread,
Sep 12, 2018, 7:45:08 PM9/12/18
to sma...@mtu.edu, BerkeleyGW Help
Hi Subhasish,

I am not sure how the specific calculations were performed at that time.  I'd like to leave the question open to see if other experts in the group have comments.  The current code supports spin-polarization feature.  But I would consider it as an extra approximation in it.  The difference should be negligible, though. 

Best,
Zhenglu

Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.



--

David Strubbe

unread,
Sep 12, 2018, 9:04:49 PM9/12/18
to Zhenglu Li, sma...@mtu.edu, BerkeleyGW Help
Hello Zhenglu,

I'm not sure what you are getting at. The only two caveats about spin-polarized systems to be aware of that I know are:
a) To use the real version of the code, you must have time-reversal symmetry, which is often not the case in spin-polarized systems because they have a net magnetization. But if you use complex wavefunctions, there is no problem.
b) The Epsilon code considers only the dielectric susceptibility and not the magnetic susceptibility. This is potentially an issue for both spin-polarized and non-spin-polarized systems, though arguably the magnetic susceptibility may be more important in spin-polarized systems. Usually ignoring the magnetic susceptibility is an excellent approximation, and I think very few calculations have been done exploring its impact.

David

Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.



--

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.
To view this discussion on the web visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/CAOn4XLjzKQ2Pq19n--0-bkjPyseVTDrbGnsZDmAf%3DfSUhT2BRw%40mail.gmail.com.



--
David Strubbe
Assistant Professor of Physics
University of California, Merced

Subhasish Mandal

unread,
Sep 12, 2018, 9:38:51 PM9/12/18
to David Strubbe, Zhenglu Li, BerkeleyGW Help
Hello David,

Thanks for the helpful information.  So if I use complex version (which obviously one would use) there would not be any problem in spin-polarized GW calculation -- is it right ?  In the epsilon code, there is no flag but in sigma code there is a flag for spin. If I use that flag I should be able to get GW-bandstrcuture for both the spins ?This is what I am doing. The only thing is the python plotting program after interpolation does not recognize the spin, which could be easily fixed.  If there is anything  I am missing here, please let me know.  If sharing my calculations/runs here would help, I could do that too.  

Thanks ! 

--Subhasish
 

  

Thanks !     

To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.



--

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.



--
David Strubbe
Assistant Professor of Physics
University of California, Merced



--
Subhasish Mandal
Postdoctoral Research Associate,
Dept. of Physics & Astronomy,
Rutgers University
Piscataway, NJ 08854

David Strubbe

unread,
Sep 12, 2018, 11:48:03 PM9/12/18
to Subhasish Mandal, BerkeleyGW Help
Hello Subhasish,

On consultation with Zhenglu, I revise my comments. Some more recent work, after the paper you mentioned, has suggested there is actually an approximation occurring in Epsilon when there is no time-reversal symmetry. We think it is probably not a problem, but work is ongoing to verify that.

Best,
David

Subhasish Mandal

unread,
Sep 13, 2018, 3:37:25 PM9/13/18
to David Strubbe, BerkeleyGW Help
Hello David,

Thanks for your info. Please keep us updated on this issue.  I think for anti-ferromagnetic it should be okay though I am not sure if you are summing or averaging for two spin contributions for the polarizability. In the log file of epsilon code, I do see : inverse epsilon RDyn/ADyn nspin= 2 or independent matrix elements of chi       spin index= 2 2; so it is considering spins.

On a separate note, when I try to get a GW-bandstrcuture on AFM system, (here MnO), I can't reproduce DFT bands using interpolation code. May be I am missing something.  
I have done a COHSEX calculation with only 100 bands just to check if the code works for spin-system. So my GW calculation is not appropriate. After I run interpolation code, bandstrcutre.dat file contains bands for both the spins (I have spin-polarized DFT WFN). I first separate out them and plot for GW and LDA bands.  But unfortunately  I do not get a proper DFT bands after I run the interpolation code on the same band-path.  Is there anything I am missing or doing wrong ? Though I know my GW bands are not converged and not accurate but should I not be able to get back my DFT bands after the interpolation step ? 

Here is my inteqp.inp:

number_val_bands_coarse 18

number_val_bands_fine 18

number_cond_bands_coarse 35

number_cond_bands_fine 35


use_symmetries_coarse_grid

no_symmetries_fine_grid

no_symmetries_shifted_grid



I have band4 to band56 in bandstructure.dat. Please let me know if you need more information.  

Thanks ! 
--Subhasish

 

Zhenglu Li

unread,
Sep 13, 2018, 7:10:08 PM9/13/18
to sma...@mtu.edu, David Strubbe, BerkeleyGW Help
Hi Subhasish,

You can try to reduce number_val/cond_bands_fine, such that it is smaller than number_val/cond_bands_coarse, to see if it works.

Best,
Zhenglu

To view this discussion on the web visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/CAP0F6N9zTO_2a8PJYGwGoe9XVgkohsohnW7-xXqvOBW8jjerSA%40mail.gmail.com.
Reply all
Reply to author
Forward
0 new messages