Debugging odd transition states in OPTIM

From CUC3
Jump to navigation Jump to search

Sometimes in OPTIM high energy transition states will be encountered. There are several strategies for understand why these are occurring.

1. OPTIM and PATHSAMPLE may be able to fine lower energy connections if the calculations are allowed to continue. This can very inefficient. One way of preventing high energy transition states from occurring is to explicitly prevent transition states of a certain energy or above from being allowed. This is done by using the keyword MAXTSENERGY in both PATHSAMPLE and OPTIM.

2. If letting the calculation resolve the problem or MAXTSENERGY does not work looking at the interpolation is necessary. In order to do so, new OPTIM keywords, DEBUG, DUMPNEBEOS, and DUMPNEBXYZ need to be added to print the NEB interpolated energy and coordinates in the files neb.xyz.31267.<n>, and neb.EofS.<n>.31267, where <n> is the number of NEB interpolations.

DEBUG 
DUMPNEBEOS 300
DUMPNEBXYZ 300

3. For the AMH, for highly anharmonic paths setting very small steps, and visualising the output has been helpful with the use of a combination of the keywords below.

USEDIAG 2
MAXBFGS 0.1
PUSHOFF 0.01
PATH 1000 0.00000000001
BFGSTS 1000 1 10 0.005 100
MAXSTEP 0.05
MAXMAX 0.1
MINMAX 1.0D-4
TRAD 0.1
CONVERGE 0.1 2.0D-6

4. Another set of parameters to explore are the below a larger number of NEWNEB images and cycles can help resolve an interpolation that may have intervening minima between the end points.

NEWNEB 500 10000 0.001
DEBUG
DUMPNEBEOS 1
DUMPNEBXYZ 1
BFGSTS 2500 2 30 0.0002

5. PATHSAMPLE can be used to create OPTIM jobs for each NEB candidate instead of chosend based on distance. To do so odata.connect must have:

DIJKSTRA  INDEX

And pathdata must have:

DIJINITCONT INDEX

6. Another strategy is to use BHINTERP to use basin-hopping to generate more physically reasonable interpolations between the end points. To do so you to use the keywords BHINTERP and BHDEBUG in OPTIM.

BHDEBUG
BHINTERP 5.0 -350.0 100 0.001 1.0 0.4 0.5 0.05 0.8

--mp466 13:48, 23 June 2009 (BST)