[Pw_forum] NEB Convergence Problem
Vic Bermudez
victor.bermudez at nrl.navy.mil
Mon Nov 7 00:33:01 CET 2011
Greetings All,
I'm struggling with a NEB calculation, and I wonder if someone can advise
me. I first converged the path using NEB without CI. This took many
iterations (810 to be exact), but it got there in the end. I then restarted
with the optimized path and CI set to 'auto'. As you can see from the
attached output, the convergence is extremely slow. In this case I know what
the "right" answer is because I'm reproducing results in the literature. The
forward activation energy should be about 1.65 eV, and at the current rate
at which the activation energy is increasing it will take many thousands of
iterations to get there. As a further comment, I originally had a problem
with erratic behavior (activation energies and NEB errors) because the SCF
convergence criterion was set too low, at the default of 10^-6. Increasing
to 10^-7 fixed that, and I've now increased it 10^-9, but that doesn't
improve the NEB behavior. Does anyone have an idea what the problem might be
?
Also on the subject of NEB, I have a few further questions:
(1) The default values of k-min and k-max are 0.1. However, the program
prints out suggested values of about 0.6, as can be seen in the output. How
important is it to use the suggested values rather than the defaults ? I've
tried varying k-min and k-max from 0.1 to 0.6 but with no effect on the rate
of convergence
(2) What are the pros and cons of using "broyden2" vs. "broyden" as the
opt_scheme. I've tried "broyden", "broyden2" and "quick-min", and none seems
to improve the rate of convergence. Is there any documentation describing
what "broyden2" actually does ?
(3) What is the meaning of the "ds" parameter, and what factors should one
consider in choosing the correct value ?
(4) What is the meaning of the "error" that NEB prints for each image at
each iteration ? Is this a measure of the force in a direction orthogonal to
the path ?
Thanks in advance for whatever help can be provided !
_____________________ NEB Output _________________________
Program NEB v.4.3.2 starts on 5Nov2011 at 5: 1: 0
This program is part of the open-source Quantum ESPRESSO suite
for quantum simulation of materials; please cite
"P. Giannozzi et al., J. Phys.:Condens. Matter 21 395502 (2009);
URL http://www.quantum-espresso.org",
in publications or presentations arising from this work. More details
at
http://www.quantum-espresso.org/wiki/index.php/Citing_Quantum-ESPRESSO
Parallel version (MPI), running on 64 processors
R & G space division: proc/pool = 64
reading file 'pwscf.path'
string_method = neb
restart_mode = restart
opt_scheme = quick-min
num_of_images = 15
nstep_path = 4000
CI_scheme = auto
first_last_opt = T
use_freezing = T
ds = 1.0000 a.u.
k_max = 0.2000 a.u.
k_min = 0.2000 a.u.
suggested k_max = 0.6169 a.u.
suggested k_min = 0.6169 a.u.
path_thr = 0.0500 eV / A
------------------------------ iteration
810 ------------------------------
tcpu = 1.2 self-consistency for image 1
tcpu = 558.0 self-consistency for image 2
tcpu = 1121.5 self-consistency for image 3
tcpu = 1648.4 self-consistency for image 4
tcpu = 2200.4 self-consistency for image 5
tcpu = 2758.9 self-consistency for image 6
tcpu = 3371.4 self-consistency for image 7
tcpu = 4029.8 self-consistency for image 8
tcpu = 4737.3 self-consistency for image 9
tcpu = 5287.6 self-consistency for image 10
tcpu = 5838.6 self-consistency for image 11
tcpu = 6379.0 self-consistency for image 12
tcpu = 6924.0 self-consistency for image 13
tcpu = 7472.4 self-consistency for image 14
tcpu = 8028.7 self-consistency for image 15
activation energy (->) = 1.223842 eV
activation energy (<-) = 1.107219 eV
image energy (eV) error (eV/A) frozen
1 -7731.9672919 0.034178 T
2 -7732.0316501 0.042825 T
3 -7732.2443631 0.048145 T
4 -7732.1803814 0.048816 T
5 -7731.1579038 0.046879 T
6 -7731.2953554 0.048145 T
7 -7731.1814917 0.045566 T
8 -7730.7434498 0.078176 F
9 -7731.4285103 0.049557 T
10 -7731.9250436 0.048245 T
11 -7731.9095723 0.040402 T
12 -7731.9057520 0.048858 T
13 -7731.9594729 0.049773 T
14 -7731.9442331 0.049314 T
15 -7731.8506693 0.043831 T
climbing image = 8
path length = 35.365 bohr
inter-image distance = 2.526 bohr
--
--
--
--
------------------------------ iteration 879 ------------------------------
tcpu = 107190.5 self-consistency for image 3
tcpu = 107381.2 self-consistency for image 4
tcpu = 107567.5 self-consistency for image 5
tcpu = 107836.9 self-consistency for image 6
tcpu = 108077.7 self-consistency for image 7
tcpu = 108322.3 self-consistency for image 8
tcpu = 108536.7 self-consistency for image 10
activation energy (->) = 1.242763 eV
activation energy (<-) = 1.126140 eV
image energy (eV) error (eV/A) frozen
1 -7731.9672919 0.034178 T
2 -7732.0362516 0.061682 F
3 -7732.2539800 0.064196 F
4 -7732.1603886 0.073400 F
5 -7730.8743402 0.058702 F
6 -7731.3502431 0.082296 F
7 -7731.1086585 0.039573 T
8 -7730.7245292 0.069207 F
9 -7731.5495175 0.058737 F
10 -7731.9427494 0.042724 T
11 -7731.9100248 0.035621 T
12 -7731.9057520 0.042466 T
13 -7731.9594729 0.049773 T
14 -7731.9442331 0.049314 T
15 -7731.8506693 0.043831 T
climbing image = 8
path length = 35.706 bohr
inter-image distance = 2.550 bohr
____________________________________________________________
Victor M. Bermudez
Code 6876
U.S. Naval Research Laboratory
4555 Overlook Ave., S.W.
Washington, DC 20375-5347
Phone: 202-767-6728
FAX: 202-767-1165
E-mail: victor.bermudez at nrl.navy.mil
More information about the users
mailing list