[ase-users] energies are lost when switch on remove_rotation_and_translation in NEB

[Ask Hjorth Larsen]

Am Do., 23. Jan. 2020 um 19:17 Uhr schrieb Michael Joseph Waters
<michael.j.waters at northwestern.edu>:
>
> Hi Ask,
>
> I posted a related issue on the gitlab page about this. I've been wondering what the downsides of having the option to rotate the cached forces (and other vector properties) from a calculator? I know that externally applied fields would not be rotated, but perhaps we should think about a means storing them in the long run?

We use a 'dumb' trajectory file to store the information.  The file
does not know that it represents NEB images.  If instead we save an
intelligent object (one which knows that it's NEB) with helper
methods, we can provide more context and help users see why these
things happen and how to fix it.

The finite-difference error cancellation on 3D grids would not work
between rotated systems.  Many codes use 3D grids at least for XC.
Initial/final image energies are important when calculating the NEB
energy barrier.  Any error affecting those differently than the rest
of the images is likely to change the energy barrier.  So
computational accuracy can suffer dangerously if we don't recalculate
energies from rotated/translated systems.

It would be better to have a mechanism which ensures that
initial/final energies are recalculated after
remove_rotation_and_translation.

Best regards
Ask

>
> Best,
> -Mike
> ________________________________
> From: ase-users-bounces at listserv.fysik.dtu.dk <ase-users-bounces at listserv.fysik.dtu.dk> on behalf of Ask Hjorth Larsen via ase-users <ase-users at listserv.fysik.dtu.dk>
> Sent: Thursday, January 23, 2020 11:40 AM
> To: GENG SUN <gengsun at ucla.edu>
> Cc: ase-users <ase-users at listserv.fysik.dtu.dk>
> Subject: Re: [ase-users] energies are lost when switch on remove_rotation_and_translation in NEB
>
> Dear Geng,
>
> Am Do., 16. Jan. 2020 um 19:20 Uhr schrieb GENG SUN via ase-users
> <ase-users at listserv.fysik.dtu.dk>:
> >
> > Hello ASE users,
> >
> > I found that when I switch on remove_rotation_and_translation=True in NEB module of ASE, the trajectory generated by the optimizers (BFGS for example) will not contain any energies and forces calculated. This will cause some difficulties to monitor the process of NEB calculations.
> >
> > Is there a simple way to fix that?
>
> The rotation and translation invalidates the previously calculated
> energies for all images.
>
> The energies won't be plotted unless *all* images have energies.  The
> energies are calculated for intermediate images during NEB, but not
> initial and final images, since those don't move.
>
> Therefore: Most likely you just need to recalculate them manually for
> initial and final image and they should all show up.
>
> Best regards
> Ask
>
> >
> > Thank you very much in advance.
> >
> > Best,
> >
> > Geng
> > _______________________________________________
> > ase-users mailing list
> > ase-users at listserv.fysik.dtu.dk
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__listserv.fysik.dtu.dk_mailman_listinfo_ase-2Dusers&d=DwICAg&c=yHlS04HhBraes5BQ9ueu5zKhE7rtNXt_d012z2PA6ws&r=IWIXV02QNefgPQmc-Ovk6evJ6KEOZjyl5g4atrVIlcJ4WAvRgH8dNpuAIeIrVHn4&m=zKPDLyNwQ0F5TURL2bpCIjobJoMKXfDMZGswq_0cNCs&s=a9Bz4FW_DJ81xJYy6_q7y1hYRLObPaGmgljM24L0X7Y&e=
>
> _______________________________________________
> ase-users mailing list
> ase-users at listserv.fysik.dtu.dk
> https://urldefense.proofpoint.com/v2/url?u=https-3A__listserv.fysik.dtu.dk_mailman_listinfo_ase-2Dusers&d=DwICAg&c=yHlS04HhBraes5BQ9ueu5zKhE7rtNXt_d012z2PA6ws&r=IWIXV02QNefgPQmc-Ovk6evJ6KEOZjyl5g4atrVIlcJ4WAvRgH8dNpuAIeIrVHn4&m=zKPDLyNwQ0F5TURL2bpCIjobJoMKXfDMZGswq_0cNCs&s=a9Bz4FW_DJ81xJYy6_q7y1hYRLObPaGmgljM24L0X7Y&e=