[gpaw-users] Error in response calculation

Thomas Olsen tolsen at fysik.dtu.dk
Thu Aug 28 16:40:45 CEST 2014 

Hi Glenn

It is too bad that the response code does not work for spin-polarized calculations if you really want to go beyond RPA.

But since you are looking at optical absorption spectra I am not sure it will matter at all. Have you tried comparing RPA and ALDA_X for the spin-paired case? Very often the difference cannot be distinguished with the naked eye, because the RPA kernel  will "kill"  any local kernel in the long wavelength limit.

/Thomas
________________________________
Fra: Glenn Jones [gjones at matthey.com]
Sendt: 28. august 2014 13:04
Til: Thomas Olsen
Emne: RE: [gpaw-users] Error in response calculation

Hi Thomas,

Thanks, yes your right it is a bulk calculation, so there is no vacuum.

Using: xc='RPA'

Has led the calculation to converge without error.

There is an assertion error for ALDA, but when I looked in the code I saw this error wasn’t raised for ALDA_X, hence I was trying to use the ALDA_X

Best regards,

Glenn


From: Thomas Olsen [mailto:tolsen at fysik.dtu.dk]
Sent: 28 August 2014 09:55 AM
To: Glenn Jones; Falco Hüser; gpaw-users at listserv.fysik.dtu.dk
Subject: SV: [gpaw-users] Error in response calculation

Hi Glenn

First of all. I don't suppose you have any vacuum in your system? The ALDA kernel diverges at low densities so the plane wave implementation of simple xc kernels cannot be used for absorption spectra. But since you write bulk oxides somewhere (and it works for spin-paired systems) I guess this is not the case.

Second. I am actually not sure the implementation is geared for spin-polarized calculations at present. For non-vanishing xc-kernels the response function acquires a non-trivial spin-structure and one can no longer work the the spin-summed response function as the case in RPA. I think there used to be an assertion error for spin-polarized calculations with xc-kernels. But of course, this should rather be implemented properly...

Try using xc='RPA' to see if this is the problem.

BR Thomas
________________________________
Fra: gpaw-users-bounces at listserv.fysik.dtu.dk<mailto:gpaw-users-bounces at listserv.fysik.dtu.dk> [gpaw-users-bounces at listserv.fysik.dtu.dk] på vegne af Glenn Jones [gjones at matthey.com]
Sendt: 26. august 2014 16:33
Til: Falco Hüser; gpaw-users at listserv.fysik.dtu.dk<mailto:gpaw-users at listserv.fysik.dtu.dk>
Emne: Re: [gpaw-users] Error in response calculation
Hi Falco,

Thanks for the suggestions, however the problem is persisting. A quick summary of things I have tried:


1)      Starting from the example Si script – this ran fine

2)      Using the Si script as the basis for materials I am interested in, ran fine for spin unpolarised (realise the parameters aren’t optimised – but it is a starting point)

3)      Switched to a spin polarised version of one of the materials from (2) and the error started to come up.

4)      Tried a different material, spin polarised, get the same error – just to rule out corrupt .gpw file


>From your suggestions:


1)      I have raised the ecut 100 eV and 340eV – 340eV is very slow but the 100eV has converged -  error persists (still waiting on the 340)

2)      Reducing number of cores (16 is how many a node has so my default submission script) , any 16, 8, and 4 all give the same error with 100 eV cut off.

Naively It seems to me it is related to the presence of spin seeing as the calculations run OK without spin,

Best regards,

Glenn

From: Falco Hüser [mailto:falco at chem.ku.dk]
Sent: 26 August 2014 10:17 AM
To: gpaw-users at listserv.fysik.dtu.dk<mailto:gpaw-users at listserv.fysik.dtu.dk>; Glenn Jones
Subject: Re: [gpaw-users] Error in response calculation

sorry, I overlooked that you are using k-point parallelization, so 16 cores would be fine.
But try increasing ecut. For very unevenly shaped unit cells, a low cutoff can result in a too small number of G vectors.
On 08/26/2014 10:06 AM, Falco Hüser wrote:
Hi Glenn,

can you try using less CPUs? You are running on 16 cores, which is absolutely not necessary, and there are only 5 G vectors (plane waves). The code uses a mixed frequency/G vector parallelization, so maybe some cores just get an empty matrix, which cannot be diagonalized.
Anyway, your plane wave energy cutoff seems very low (12 eV), increase it for a larger number of plane waves.

Falco
On 08/26/2014 09:54 AM, Glenn Jones wrote:
Hi,

I am starting to use the response module in GPAW for calculating absorption spectra of TM oxides.

Non-spin polarised calculations worked fine, but I am now running into some issues with spin polarised calculations, that aren’t so well documented:

I have pre optimised my bulk system which is read in from out.gpw (which I am unable to send by email because it is too big).  However on running the following script I get a ‘Singular matrix error’ – see attached .txt file for the full output:

from ase.io import read
import numpy as np
from gpaw import GPAW, FermiDirac, Mixer
from ase import Atom, Atoms
from ase.visualize import view
from gpaw.mixer import Mixer
from gpaw.response.df import DF
from gpaw.mpi import world
ABS = 1


if ABS:
    df = DF(calc='../out.gpw',
            q=np.array([0.0, 0.0, 0.00001]),
            w=np.linspace(0,10,200),
            eta=0.05,
            ecut=12,
            optical_limit=True,
            xc='ALDA_X'
            #kcommsize=4
            )

    df.get_absorption_spectrum()
____________________________________________________________



Any clues as to how to get around this would be appreciated,

Thanks,

Glenn

If the reader of this email is not the intended recipient(s), please be advised that any dissemination, distribution or copying of this information is strictly prohibited. Johnson Matthey PLC has its main place of business at 5th Floor, 25 Farringdon Street, London (0207269 8400).
Johnson Matthey Public Limited Company Registered Office: 5th Floor, 25 Farringdon Street, London EC4A 4AB.Registered in England No 33774
Whilst Johnson Matthey aims to keep its network free from viruses you should note that we are unable to scan certain emails, particularly if any part is encrypted or password-protected, and accordingly you are strongly advised to check this email and any attachments for viruses. The company shall NOT ACCEPT any liability with regard to computer viruses transferred by way of email.
Please note that your communication may be monitored in accordance with Johnson Matthey internal policy documentation.



_______________________________________________

gpaw-users mailing list

gpaw-users at listserv.fysik.dtu.dk<mailto:gpaw-users at listserv.fysik.dtu.dk>

https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users




_______________________________________________

gpaw-users mailing list

gpaw-users at listserv.fysik.dtu.dk<mailto:gpaw-users at listserv.fysik.dtu.dk>

https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users


This message has been scanned for viruses by MailControl<http://www.websense.com/>, a service from BlackSpider Technologies<http://www.websense.com/>.
Click here<https://www.mailcontrol.com/sr/+wESOp4ozsjGX2PQPOmvUv8Fcr8RSh1B4lQQtj!0PYl8HsK97Uw3StXoH36MkswjvO2IMtgU0s6K0Rphlb7DTA==> to report this email as spam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20140828/f2789bb3/attachment-0001.html>

More information about the gpaw-users mailing list