[Pw_forum] NEB restart from scratch
Paolo Giannozzi
paolo.giannozzi at uniud.it
Tue Nov 26 21:18:09 CET 2013
Not sure I understand correctly what the problem
is but: if you start from scratch, with the same
cutoff and k-points and other parameters, using
the final positions of a previous neb run, and
get very large forces (especially those around
images 11 !!), the simplest explanation is that
the final positions are not what you think they
are. It is rather easy to make a mistake in
setting up a neb path.
P.
On Sat, 2013-11-23 at 19:43 +0100, paolo paoli100 wrote:
> Dear all pw users,
>
>
>
> I performed neb calculations (espresso-5.0.2) with a 25 Ry cutoff and
> a 6 6 1 k-points grid.
>
> After reaching convergence, with the error on each image lower than
> the selected threshold, I tried to refine my calculations by
> increasing the cutoff up to 40 Ry, and by using more k-points (8 8 1
> grid).
>
> Converged neb (25 Ry)
>
> ------------------------------ iteration 39
> ------------------------------
>
>
>
> cpu = 0 tcpu = 23906.3 self-consistency for image 2
>
> cpu = 0 tcpu = 23978.2 self-consistency for image 4
>
> cpu = 0 tcpu = 24086.9 self-consistency for image 6
>
> cpu = 0 tcpu = 24200.3 self-consistency for image 8
>
> cpu = 0 tcpu = 24308.2 self-consistency for image 10
>
>
>
> activation energy (->) = 0.894658 eV
>
> activation energy (<-) = 0.836243 eV
>
>
>
> image energy (eV) error (eV/A) frozen
>
>
>
> 1 -19891.2503277 0.071116 T
>
> 2 -19891.2158612 0.049589 F
>
> 3 -19891.1523863 0.188484 F
>
> 4 -19891.0150938 0.162087 F
>
> 5 -19890.9987459 0.191508 F
>
> 6 -19890.8817497 0.106483 F
>
> 7 -19890.6985072 0.123973 F
>
> 8 -19890.4818044 0.093420 F
>
> 9 -19890.3556701 0.086773 F
>
> 10 -19890.5472740 0.113798 F
>
> 11 -19891.1235293 0.166418 F
>
> 12 -19891.1919133 0.070926 T
>
>
>
> climbing image = 9
>
>
>
> path length = 10.335 bohr
>
> inter-image distance = 0.940 bohr
>
>
>
> ---------------------------------------------------------------------------
>
>
>
> neb: convergence achieved in 39 iterations
>
>
>
> NEB : 6h39m CPU 6h46m WALL
>
>
>
>
>
> This run was terminated on: 7:53:22 3Jul2013
>
>
>
> =------------------------------------------------------------------------------=
>
> JOB DONE.
>
> =------------------------------------------------------------------------------=
>
>
>
>
>
> So, I created a new input file with the new values of these parameters
> (cutoff and k-points grid), and I copied the atomic coordinates from
> the .crd file, of the converged path with 25 Ry and 6 6 1 grid. Then,
> I ran neb.x with the restart mode: “from_scratch”. Unfortunately the
> first iteration of the new neb simulation, with 40 Ry, showed very
> large forces.
>
>
>
> ------------------------------ iteration 1
> ------------------------------
>
>
>
> cpu = 0 tcpu = 6.6 self-consistency for image 1
>
> cpu = 0 tcpu = 281.5 self-consistency for image 2
>
> cpu = 0 tcpu = 559.0 self-consistency for image 4
>
> cpu = 0 tcpu = 832.1 self-consistency for image 6
>
> cpu = 0 tcpu = 1092.8 self-consistency for image 8
>
> cpu = 0 tcpu = 1365.2 self-consistency for image 10
>
>
>
> activation energy (->) = 38.645703 eV
>
> activation energy (<-) = 38.552298 eV
>
>
>
> image energy (eV) error (eV/A) frozen
>
>
>
> 1 -19892.5055371 0.008676 T
>
> 2 -19892.4796891 0.244186 F
>
> 3 -19892.4470814 0.232055 F
>
> 4 -19892.3936725 0.309786 F
>
> 5 -19892.2598441 0.174595 F
>
> 6 -19892.2326900 0.376662 F
>
> 7 -19892.1508214 0.237737 F
>
> 8 -19891.9033607 0.629380 F
>
> 9 -19891.5413520 1.554399 F
>
> 10 -19890.5368565 23.555285 F
>
> 11 -19853.8598346 268.470304 F
>
> 12 -19892.4121325 0.021811 T
>
>
>
> climbing image = 11
>
>
>
> path length = 11.793 bohr
>
> inter-image distance = 1.072 bohr
>
>
>
> I don’t know why so huge forces appeared since my initial path is the
> converged path with a 25 Ry cutoff, I think the cutoff may affect
> forces, but what about the restart mode “from_scratch”?
>
>
>
> So I even tried to run the same inputfile with a 25 Ry cutoff and 6 6
> 1 k-points grid. Since the initial path is the converged path with 25
> Ry and 6 6 1 k-points (copy and paste from .crd file) , my neb
> calculation should converge after one iteration, instead I have large
> forces:
>
>
>
> ------------------------------ iteration 1
> ------------------------------
>
>
>
> cpu = 0 tcpu = 3.5 self-consistency for image 1
>
> cpu = 0 tcpu = 268.4 self-consistency for image 2
>
> cpu = 0 tcpu = 522.8 self-consistency for image 4
>
> cpu = 0 tcpu = 773.8 self-consistency for image 6
>
> cpu = 0 tcpu = 1020.3 self-consistency for image 8
>
> cpu = 0 tcpu = 1270.4 self-consistency for image 10
>
>
>
> activation energy (->) = 38.753850 eV
>
> activation energy (<-) = 38.695436 eV
>
>
>
> image energy (eV) error (eV/A) frozen
>
>
>
> 1 -19891.2503277 0.071116 T
>
> 2 -19891.2304040 0.241051 F
>
> 3 -19891.2049038 0.226864 F
>
> 4 -19891.1588213 0.318751 F
>
> 5 -19891.0267566 0.158294 F
>
> 6 -19891.0032580 0.382541 F
>
> 7 -19890.9283308 0.232692 F
>
> 8 -19890.6852427 0.655155 F
>
> 9 -19890.3234875 1.547988 F
>
> 10 -19889.3355302 23.552337 F
>
> 11 -19852.4964773 269.371205 F
>
> 12 -19891.1919133 0.070926 T
>
>
>
> climbing image = 11
>
>
>
> path length = 11.793 bohr
>
> inter-image distance = 1.072 bohr
>
>
>
>
>
> Could anyone tell me how QE computes forces at iteration number 1, if
> the restart mode is from scratch?
>
> How can I deal with this problem?
>
> Thanks to all,
>
> Luca Dietz
>
> Ph.D. student in Chemical Engineering
>
> Politecnico di Milano, Italy
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
--
Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
More information about the users
mailing list