Hi,
I am doing the dimer calculation. The procedure I followed is as below:
a. get the reactant and product by optimization.
b. generate the MODECAR
c. do the DIM calculation from reactant
But at the beginning, after about 20 cycles, the DIMCAR file is as follows:
1 0.77095 33.40430 -136.67848 22.13310 26.13658
1 0.77095 10.20762 -136.67848 5.86752 17.20170
1 0.77095 6.00792 -136.67848 3.12628 16.07192
1 0.77095 4.85855 -136.67848 1.45531 6.55196
2 1.34462 5.64206 -136.65534 1.20328 4.83476
2 1.34462 2.47612 -136.65534 0.76601 2.77666
2 1.34462 1.82875 -136.65534 0.60435 2.66151
2 1.34462 1.71894 -136.65534 0.54579 1.66426
I thank this is not true! Can you give me some advice?
about dimer calculation
Moderator: moderators
There are a couple of different ways to use the dimer method. One is to start from a guess at the saddle, and converge upon it. A second is to start from a minimum and find different possible saddles that lead to unknown final states.
In you description, it sounds like you are mixing these two approaches. If you know an initial and final state, I recommend starting with a NEB calculation. Do a few iterations, until the forces drop (roughly) below 1 eV/Ang. If you are happy with the reaction mechanism, you can continue the run (using the climbing-image) and find the saddle.
If you want to reduce the number of processors being used, you can take your roughly converged NEB and start a dimer run from the maximum along the band, to find the true saddle. We have the neb2dim.pl script to take your NEB and generate a suitible dimer run. If you start with this approach, and get comfortable with the dimer calculation, only then should you consider searching for saddles around a minimum.
[Note: I don't see anything wrong with your output.]
In you description, it sounds like you are mixing these two approaches. If you know an initial and final state, I recommend starting with a NEB calculation. Do a few iterations, until the forces drop (roughly) below 1 eV/Ang. If you are happy with the reaction mechanism, you can continue the run (using the climbing-image) and find the saddle.
If you want to reduce the number of processors being used, you can take your roughly converged NEB and start a dimer run from the maximum along the band, to find the true saddle. We have the neb2dim.pl script to take your NEB and generate a suitible dimer run. If you start with this approach, and get comfortable with the dimer calculation, only then should you consider searching for saddles around a minimum.
[Note: I don't see anything wrong with your output.]