[CP2K:4460] neb : question about the number of scf optimizations for each band step for each replica

Andrea Floris an.f... at gmail.com
Mon May 20 19:52:23 CEST 2013


On Mon, May 20, 2013 at 4:08 PM, Teodoro Laino <teodor... at gmail.com>wrote:

> DIIS performs a LS - what you see is the result of the line search.
> Use NO_LS<http://manual.cp2k.org/trunk/CP2K_INPUT/MOTION/BAND/OPTIMIZE_BAND/DIIS.html#list_NO_LS> if
> you don't like that.
>
>


What are the consequences of using NO_LS in a NEB calculation? Using NO_LS
would result in a single optimization of the image
(per band step) and  (maybe separate issue) no electronic optimization at
all for "fixed" images?

I am just running a test, with 3-5 replica only.  I see quite a strong
dependence on the K_SPRING constant if I vary it from 0.05 to 0.02
(the barrier decreases of about 0.05eV and the path is different).
Is is correct to expect a much reduced variation of the barrier with
K_SPRING, if the number of
images is increased, e.g. to 10-15? Which distance is typically OK between
adiacent images, 0.5 A?
Thank you for your help
Best
Andrea





> Moreover, I have no idea if this is a production run or simply a test, but
> you want to keep the distance between your replica much smaller than 3.0+
> Angstrom.
> Hope this helps.
> Teo
>
>
On May 20, 2013, at 1:14 PM, Andrea Floris <an.f... at gmail.com> wrote:
>
> Dear Teo,
>
> I attach the input and output and all relevant files.
>
> My problem is in the BAND* files , whose number is indeed equal to the
> number of replica (three in my case).
> If you in one of these files (prefix is neba) do
>
> grep -i  'Band  Step  Nr.'  neba-BAND1.out
>
> you have as output something like
> ...
>  ** Band  Step  Nr. :   13
>    **
>  ** Band  Step  Nr. :   13
>    **
>  ** Band  Step  Nr. :   13
>    **
>  ** Band  Step  Nr. :   13
>    **
>  ** Band  Step  Nr. :   13
>    **
>  ** Band  Step  Nr. :   14
>    **
>  ** Band  Step  Nr. :   14
>    **
>  ** Band  Step  Nr. :   14
>    **
> ...
>
> so, e.g. 5 optimizations (but geometrical or purely electronic scf???) in
> the band  step 13, and 3 in the band step 14...
> That's the point I don't understand...
>
>
> Moreover, I also do not understand why this is done *also* for the end
> points, when you choose to not optimize
> the initial and final replica...For example in my case here replica 1 was
> fixed.  In fact in the
>
> neba-1.ener
>
> file, the energy of replica 1 is fixed, i.e. it has always the same energy
> through the band optimization steps. But in the file  neba-BAND1.out many
> total energies are found...
> one for each scf optimization.
>
> Thanks in advance for you answer,
>
> Best
> Andrea
>
>
>
>
>
>
>
>
>
>
>
>
> On Mon, May 20, 2013 at 9:54 AM, Teodoro Laino <teodor... at gmail.com>wrote:
>
>> >>
>> >> 1. why  more than one scf optimization is performed? (every
>> optimization being achieved with ~40 scf steps)
>> > it depends on how many processors you are using and how many processors
>> per replica you requested. On the top of that, the number of scf
>> optimisations depends also on the type of optimiser you are using.
>> > Again without an input is almost useless to cover all possible things
>> you may see.
>> Correction: forget about the parallel setup - you should always have a
>> number of BAND* files equal to the number of frames of your NEB.
>> So not sure at all what you are seeing… (probably it is only the output
>> of the OUTER_SCF cycle)
>>
>> In any case, without input..
>>
>> >
>> >> 2. Are these purely electronic optimizations, or there is a
>> geometrical minimization involved?
>> > the geometry optimisation steps are not shown in the *BAND* output
>> >
>> > Regards
>>
>> --
>> You received this message because you are subscribed to a topic in the
>> Google Groups "cp2k" group.
>> To unsubscribe from this topic, visit
>> https://groups.google.com/d/topic/cp2k/_icaeFBVnZ4/unsubscribe?hl=en.
>> To unsubscribe from this group and all its topics, send an email to
>> cp2k+uns... at googlegroups.com.
>> To post to this group, send email to cp... at googlegroups.com.
>> Visit this group at http://groups.google.com/group/cp2k?hl=en.
>> For more options, visit https://groups.google.com/groups/opt_out.
>>
>>
>>
>
>
> --
>  Dr Andrea Floris
>  Research Associate
>  King's College London
>  Strand, London WC2R 2LS
>  United Kingdom
>  Phone: +44 (0) 207 848 2064
>  Fax    : +44 (0) 207 848 2420
>  Location:  Strand Building, 4th floor, Room 4.02
>  Emails: andrea... at kcl.ac.uk, an.f... at gmail.com
>
> --
> You received this message because you are subscribed to the Google Groups
> "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to cp2k+uns... at googlegroups.com.
>
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at http://groups.google.com/group/cp2k?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
> <neb_test_3replicas.tgz>
>
>
>  --
> You received this message because you are subscribed to a topic in the
> Google Groups "cp2k" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/cp2k/_icaeFBVnZ4/unsubscribe?hl=en.
> To unsubscribe from this group and all its topics, send an email to
> cp2k+uns... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at http://groups.google.com/group/cp2k?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>



-- 
 Dr Andrea Floris
 Research Associate
 King's College London
 Strand, London WC2R 2LS
 United Kingdom
 Phone: +44 (0) 207 848 2064
 Fax    : +44 (0) 207 848 2420
 Location:  Strand Building, 4th floor, Room 4.02
 Emails: andrea... at kcl.ac.uk, an.f... at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20130520/16dc62e3/attachment.html>


More information about the CP2K-user mailing list