Page 1 of 1

Questions about NEB converge and Script

Posted: Wed Oct 25, 2006 9:33 am
by thshwy
Dear all,
I would like to ask some questions about the CI-NEB and some of scripts.
1. I start the parameter the EDIFFG=-0.05, EDIFF=0.0001, IBRION=3,POTIM=0.01 and the calculation is completed after 311 ionic steps. But I found the “total force” in OUTCAR is lager than 0.05 for some images. I want to know the convergent condition about the force is means the force in every degree of all images should be converged? If it is, why is my calculation ended while the total force is lager than 0.05?
2. About vfin.pl
I have used vfin.pl and there is no output. “usage: vfin.pl (output directory)”
Does the output directory here mean the directory of my calculation or the directory after run the vfin.pl?
Is the script used to transform the results of this calculation to the next calculation?
Should the results of my calculation be put the directory of vtstscripts and the OUTCAR and XDATCAR be compressed? I have tried but there is no change.
3. About nebbarrier.pl
I have used the nebbarrier.pl and get the neb.dat file, but the number in the second list (the first is 1,2,3…) is 0.0000. Does it mean the distance between images and why is it 0.000? I have found the distance between the images in OUTCAR. I have copied the results of calculation to the vtstscripts directory.
I wonder what is wrong with the use of script and if the makefile of VASP should be put the vtstscripts directory.
Thanks.

Posted: Wed Oct 25, 2006 10:59 am
by andri
I couldn't find your results but:

1. POTIM = 0.01 is rather small, usually 0.1 is more reasonable as a first guess. Depending on your system then some of the constrained atoms can have considerable forces, which are not apart of the force minimization criteria. The NEB routine has the some convergence criteria as VASP itself, which is the length of the force vector on each atom has to be less than the criteria.

2. The vfin.pl <dir> should be run in the same directory as the band (the parent directory), i.e. where the 00, 01, etc. subdirectories are located. After, the 00, 01, ... in the parent directory will contain POSCARs that correspond to the finial configuration, that is the CONTCARs in the newly created <dir>. The 00,01, ... directories in <dir> will contain the initial POSCAR, along with the final CONTCAR and compressed version of the OUTCAR, XDATCAR and potential WAVECAR and CHGCAR. Make sure to include the location of the vtstscripts/ in your $PATH environmental variables and run everything from there. Do not copy anything to your current folder.

3. That's odd. Is just the first number zero or the entire column? It's normal to have the first number to be 0.0 since the column contains the "coordinate" of the MEP, starting from the first images at 0.0.

Posted: Thu Oct 26, 2006 2:16 am
by thshwy
Thanks,Andri.
1.In my system, the unconstrained atoms have lager forces in y direction but the calculation is completed. Are the results credible? What is the means about ‘the length of the force vector on each atom has to be less than the criteria’? Is the criteria is the number of EDIFFG?
2.The entire column numbers are zero. What is the reason? Or it is wrong with the use of script?
3.I want to know if that the result is better for the closer distance of the images.
4.I do a new calculation after I copied the CONTCAR to the POSCAR in the last calculation, and set the EDIFFG=-0.03, POTIM=0.01, but the results have not converged after 400 ionic steps. The force for the unconstrained atom is larger (more than 1.0). I think the last calculation is not credible. Now I will restart the calculation and set POTIM=0.1. I wonder if that the distance between the images should be reduced (e.g. increase the number of images). The distance in last calculation is about 0.51 A.
Thanks!

Posted: Thu Oct 26, 2006 11:52 am
by andri
Can you put your calculation somewhere that we can get to it?

Posted: Fri Oct 27, 2006 12:06 pm
by andri
Ok, it all seems to be working fine to me. I include here the order which I execute the commands to pack and clean the results and prepare for a new one. Note that repeated usage of "pwd" and "which" is not necessary, I only include them here to show where I am and the location of the Perl script being used.

1) Check the final energy and forces.

/Users/andri(56)> pwd
/Users/andri/Desktop/tshol2-6
/Users/andri(57)> which nebef.pl
/usr/local/vtsttools/nebef.pl
/Users/andri(58)> nebef.pl
0 0.00669700 -75.76058700 0.00000000
1 0.03101400 -74.87974700 0.88084000
2 0.01656100 -74.29104600 1.46954100
3 0.01864200 -74.00159200 1.75899500
4 0.01251500 -73.67004800 2.09053900
5 0.01662800 -75.77922300 -0.01863600
6 0.04950800 -76.61377600 -0.85318900
7 0.00884900 -76.83711900 -1.07653200

... all forces (2nd columns) are less that 0.05 so it's on the right track.

2) Pack and clean the results and send them to a new directory (names "ok1")

/Users/andri(60)> pwd
/Users/andri/Desktop/tshol2-6
/Users/andri(61)> which vfin.pl
/usr/local/vtsttools/vfin.pl
/Users/andri(62)> vfin.pl ok1
****
Cleaning up after a neb run
****
cp: KPOINTS: No such file or directory
In directory: /Users/andri/Desktop/tshol2-6/00
In directory: /Users/andri/Desktop/tshol2-6/01
In directory: /Users/andri/Desktop/tshol2-6/02
In directory: /Users/andri/Desktop/tshol2-6/03
In directory: /Users/andri/Desktop/tshol2-6/04
In directory: /Users/andri/Desktop/tshol2-6/05
In directory: /Users/andri/Desktop/tshol2-6/06
In directory: /Users/andri/Desktop/tshol2-6/07

... it finished ok. Presumably you have the KPOINTS file it complains about. The CONTCARs in ok1/01, 02, ... are now POSCARs in 01, 02, ... so you are ready for a new run.

3) Extract the results with "nebresults.pl"

/Users/andri(66)> pwd
/Users/andri/Desktop/tshol2-6/ok1
/Users/andri(67)> which nebresults.pl
/usr/local/vtsttools/nebresults.pl
/Users/andri(68)> nebresults.pl

Unziping the OUTCARs ... done
Do nebbarrier.pl ; nebspline.pl
Do nebef.pl
Do nebmovie.pl
Do nebconverge.pl
Zipping the OUTCARs again ... done

Forces and Energy:
0 0.00669700 -75.76058700 0.00000000
1 0.03101400 -74.87974700 0.88084000
2 0.01656100 -74.29104600 1.46954100
3 0.01864200 -74.00159200 1.75899500
4 0.01251500 -73.67004800 2.09053900
5 0.01662800 -75.77922300 -0.01863600
6 0.04950800 -76.61377600 -0.85318900
7 0.00884900 -76.83711900 -1.07653200

Extrema 1 found at image 0.000000 with energy: 0.000000
Extrema 2 found at image 3.992948 with energy: 2.090569
Extrema 3 found at image 7.000000 with energy: -1.076532

The results are in mep.eps, movie.xyz, neb.dat (which looks good to me),spline.dat and exts.dat. Note that for the movie.xyz to be properly format then the *first* line in all the POSCARs (and CONTCARs) *must* contain the element symbols in the same order as the POTCAR is constructed. The <typeset.pl> (http://theory.cm.utexas.edu/vtsttools/scripts/) can be used to change the file header.

Also, I would replace all of

SPRING = 5
SPRING2= 5
ISPRING= 1
SPOWER = 1
EFIRST = -75.765075
ELAST = -76.836976

with just SPRING = -5.0, see http://theory.cm.utexas.edu/vtsttools/neb/

Posted: Mon Oct 30, 2006 1:34 pm
by thshwy
Hello, andri.
Thank you for your detailed reply! I tried to run the script for the whole day! The neb.dat is correct. But it is also failed when I run the vfin.pl. The output directory is empty. I don’t know how to set the location of the vtstscripts/ in my $PATH environmental variables. I found the nebmovie.pl, nebspline.pl and nebconverage.pl can not execute when I run them. The information is:
sh: gnuplot: not found
I run the script in the Linux system. Is it related to the system? We used the FORTRAN 77 language. The compile for VASP have done by others before. I found the dynmat.o neb.o dimmer.o lanczos.o in the VASP makefile. Is the failure is related to these? Or Should I recompile again? They haven’t used your script in our team. Thanks!

Posted: Mon Oct 30, 2006 4:22 pm
by andri
It never hurts to recompile but this problem shouldn't be related to your VASP executables, unless they are so old that our scripts misinterpreted the output from out routines. But I seriously doubt that.

Depending on the location (mine is /usr/local/vtstscripts) of your vtstscripts folder add

tcsh shell -> add to .tcshrc : setenv PATH /usr/local/vtstscripts:$PATH

bash shell : -> add to .bashrc : PATH=/usr/local/vtstscripts:$PATH
export PATH

and then log out and in again.

Posted: Mon Nov 06, 2006 8:59 am
by thshwy
Hello, andri,
I am sorry for response to your reply so later. Because the Linux system has not worked till this morning. Thank you very much! I have set $PATH and run the script this morning. There are also some wrong when I run the vfin.pl and nebspline.pl. The information is:
SGI% vfin.pl ok1
****
Cleaning up after a neb run
****
/xia/ths/vtstscripts/nebclean.sh[37]: $(ls: not found
-e In directory: /xia/ths/n-n/tshol2-6
../ok1/$(ls not found.
../ok1/$(ls/OUTCAR: No such file or directory
XDATCAR - No such file or directory
../ok1/$(ls not found.
../ok1/$(ls/XDATCAR: No such file or directory
CONTCAR - No such file or directory
/xia/ths/vtstscripts/nebclean.sh[37]: cd: bad option(s)
-e In directory: /xia/ths/n-n
../ok1/-d not found.
../ok1/-d/OUTCAR: No such file or directory
XDATCAR - No such file or directory
../ok1/-d not found.
CONTCAR - No such file or directory
../ok1/-d/XDATCAR: No such file or directory
/xia/ths/vtstscripts/nebclean.sh[37]: [0-9][0-9]): not found
-e In directory: /raid5/xia/ths
../ok1/[0-9][0-9]) not found.
../ok1/[0-9][0-9])/OUTCAR: No such file or directory
XDATCAR - No such file or directory
../ok1/[0-9][0-9])/XDATCAR: No such file or directory
../ok1/[0-9][0-9]) not found.
CONTCAR - No such file or director
SGI% nebspline.pl
sh: gnuplot: not found
In ok1 directory, there are
$(ls -d INCAR KPOINTS [0-9][0-9])
I don’t know this reason. Should the perl scripts be modified in my system? And how should I modify it? Thank you!