"Deviation too large for solvent" in NEB calculation


Click here for full thread
Clicked A Few Times
I would first check the pdb structures for your NEB beads for any obvious problems. To do this just run prepare command over numbered rst files in your perm directory

...
prepare
read ./perm/sys_neb002.rst
write sys_neb002.pdb
end
task prepare

Depending on what you see, you could run a few constant energy dynamics steps on just solvent part for the offending bead(s). Once you fix these problems you could "restart" your NEB calculations using these new rst files.

Regarding the solvent configuration here is how the code works. For new NEB calculation:
1. The solvent configuration is taken from the common rst file specified by the "system" directive in the MD block
(in your case react_neb.rst)
2. The QM region geometries for end points are extracted from the two rst files specified by set qmmm:neb_path_limits directive ( in your case react_opt.rst prod_ref.rst). Note that none of the solvent information is used from these files.
3. The interpolation is done to generate the intermediate QM region structures
4. In the first gradient calculation for the i-th bead the code check for the presence of rst file named sys_neb00i.rst in the perm directory. If it is not there (as typical for new NEB calculation) this rst file is created by taking common rst file (see step 1) and replacing its QM region with that of the i-th bead. Using the resulting rst file, the solvent optimization is performed around the fixed QM region, QM gradient is calculated, and the updated rst file is saved to the perm directory to be used in the next round of NEB calculation. And so on.

So the initial solvent configuration can be affected in two ways - either through common rst file (which will apply the same solvent configuration to all beads) and by depositing numbered rst files into the perm directory prior to calculation to affect individual beads. So for example, you could decide that reactant solvent configuration will work for first 5 beads, but that the product solvent configuration would be more appropriate to use for last 5 beads in you calculation. To accomplish this you copy react_opt.rst into common rst file, react_neb.rst, in the top run directory, and copy prod_ref.rst into sys_neb006.rst, ...., sys_neb010.rst in the perm directory.

Marat