Restarting Energy Barrier Calculatons

[zacbullard]

Hello Everyone,

I have searched the forums and the internet for the answer to this question, to no avail. Please forgive me if I missed something obvious.

I have both completed and interrupted NEB energy barrier calculations that I would like to restart and continue. I would love to find out how to do so if possible, with an example of the input variables.

Here are the NEB variables used in the input file:

Code: Select all

 nimage 12                       # Number of points along the string
 imgmov 2                        # Selection of "String Method" algo
 ntimimage 50                    # Max. number of relaxation steps of the string
 tolimg 5.0d-5                   # Tol. criterion (will stop when average energy of cells < tolimg)
 dynimage 0 10*1 0               # Keep first and last images fixed
 fxcartfactor 1.0                # Time step for evolution  step of string metho
 prtvolimg 0                     # Printing volume (0=full, 1=intermediate, 2=minimal)

Thank you for your time.

[woffermans]

Dear ABINIT friends,

I encounter the same problem.

What is the procedure to continue an interrupted minimum energy path calculation?

I cannot find the info in the tutorials.

Can someone help me out?

[ebousquet]

Dear Willem,
I think the restartxf flag is what you are looking for.
Best wishes,
Eric

[woffermans]

The restartxf flag does indeed help ... but only partially.

If one analyses the previous calculation, then it appears that the calculation stopped at image 6 of time step 2:

Previous calculation: $grep STRING ../5M5V2x2_test2/pd.log
STRING METHOD - TIME STEP 1
STRING METHOD - CELL # 1/ 7
STRING METHOD - CELL # 2/ 7
STRING METHOD - CELL # 3/ 7
STRING METHOD - CELL # 4/ 7
STRING METHOD - CELL # 5/ 7
STRING METHOD - CELL # 6/ 7
STRING METHOD - CELL # 7/ 7
STRING METHOD:
STRING METHOD - TIME STEP 2
STRING METHOD - CELL # 2/ 7
STRING METHOD - CELL # 3/ 7
STRING METHOD - CELL # 4/ 7
STRING METHOD - CELL # 5/ 7
STRING METHOD - CELL # 6/ 7

If one analyses the restarted calculation, then it appears that the calculation starts at image 2 of time step 2.

Restarted calculation: $grep STRING pd.log
STRING METHOD:
STRING METHOD - TIME STEP 2
STRING METHOD - CELL # 2/ 7

Does time step 2 need to be recalculated for all images or can the calculation be restarted at image 6 of time step 2?

[woffermans]

Dear Eric and ABINIT friends,

A renewed restart of the STRING calculation showed that restartxf flag is not the only parameter to consider.

Once more, all files of the previous run were present. I did not change anything, before I resubmitted the job.

After the job was started, I got the following in my output file:

<snip>
...
================================================================================
STRING METHOD FOR 1 TIME STEP
================================================================================
</snip>

The end of the output file of the previous job showed:

<snip>
...
Moving images of the cell...


================================================================================
STRING METHOD - TIME STEP 3
================================================================================
</snip>

This is more frustrating, since it was only for a trivial reason that the previous job was stopped.

The parameter pawovlp did not do its work!