NEB calculation help: CO migration on Cu(111)

[fangyongxinxi]

Dear Sir,

I have a prolbem when I do neb calculation with atk_14_version.

the system is CO migration on Cu(111), and four files are listed in the attachment.

ini.txt, is the initial structure,
final.txt, is the final structure,
neb_input.txt, is the calculation file generated by ATK neb toolbox,
error.txt, is the error file.

there are seven images in the neb calculation, so I use "mpiexe -n 14 input.py > out.log" to do the calculation.
the calculation is stopped after a few minites.
I check the error message, but I can not solve the problem by myself.

could you have a check for the input file ?
and I want to know what's wrong with my input file.

Thanks!

[Julian Schneider]

In general, the reported error message means that you are using constraints on a part of the system but the positions of the constrained atoms are not exactly the same in all NEB-images. This can happen when you use the NEB creator with the IDPP or Halgren-Lipscomb method to create your initial NEB images.
However, in your system (as given in the file neb_input.py) everything looks ok, i.e. all constrained atoms have constant positions across all images. Accordingly, I could not reproduce your error message, running your file neb_input.py in ATK2014.2. Are you sure that this is the file that generated the error?

Another point, in ATK2014 there is no possibility to parallelize over the NEB images, so during the geometry optimization the images would be calculated sequentially, using 14 MPI processes for each DFT calculation. Parallelization over the NEB images will be available in ATK2015.

[fangyongxinxi]

In general, the reported error message means that you are using constraints on a part of the system but the positions of the constrained atoms are not exactly the same in all NEB-images. This can happen when you use the NEB creator with the IDPP or Halgren-Lipscomb method to create your initial NEB images.
However, in your system (as given in the file neb_input.py) everything looks ok, i.e. all constrained atoms have constant positions across all images. Accordingly, I could not reproduce your error message, running your file neb_input.py in ATK2014.2. Are you sure that this is the file that generated the error?

Another point, in ATK2014 there is no possibility to parallelize over the NEB images, so during the geometry optimization the images would be calculated sequentially, using 14 MPI processes for each DFT calculation. Parallelization over the NEB images will be available in ATK2015.

Thanks for your reply.
1.
I run the script like "mpiexe -n 14 atk14 input.py > out.log"
The script can run for about one hour, and the stopped. I am sure that is the error message.

why I use 14 processes, that is the tutorial said so. I do not know atk14 can not Parallelization over the NEB images.
2.
You finish the calculation with my input file ?
I suggest you set the "Max_force = 0.2" in the input file ( in order to finished the calculation quickly), and run the script.


Thanks, and look forward to you reply.

[Julian Schneider]

Can you tell me if any optimization steps had been performed, before the error message popped up? Could you maybe send me the full log file?
Also, if you remove the line

device_configuration.update()

in line 1293 of your script, then the error message should pop up immediately. Do you see this?

Regarding the parallelization:
As stated in the Note at the beginning of the NEB-tutorial, the image-parallelization works only in ATK2015, not in ATK2014.

[fangyongxinxi]

In general, the reported error message means that you are using constraints on a part of the system but the positions of the constrained atoms are not exactly the same in all NEB-images. This can happen when you use the NEB creator with the IDPP or Halgren-Lipscomb method to create your initial NEB images.
However, in your system (as given in the file neb_input.py) everything looks ok, i.e. all constrained atoms have constant positions across all images. Accordingly, I could not reproduce your error message, running your file neb_input.py in ATK2014.2. Are you sure that this is the file that generated the error?

Another point, in ATK2014 there is no possibility to parallelize over the NEB images, so during the geometry optimization the images would be calculated sequentially, using 14 MPI processes for each DFT calculation. Parallelization over the NEB images will be available in ATK2015.

"This can happen when you use the NEB creator with the IDPP or Halgren-Lipscomb method to create your initial NEB images. "

this is the reason.
When I create initial NEB images with linear method,
the calculation goes one well. No errors.


Thanks for your help.

[Julian Schneider]

Just as an information, if you want to use constraints together with the Image-dependent pair potential or the Halgren-Lipscomb method in ATK2014, you can tag the group of constrained atoms using the tag name 'constrained'. Then the NEB builder will automatically recognize these tagged atoms and keep them fixed.
In ATK2015, the NEB builder will have a proper constraints widget similar to OptimizeGeometry or MolecularDynamics.