Dear Pandaaka,
NumForce is restartable, so usually you can start it several times and it simply continues at the point it stopped before.
In some cases however, the check if a specific point has completely been done or not does fail. In those cases one can check the results by looking at the directory
numforce/KraftWerk. In there you will find several files for each point (p001, p002, ...):
- p001.coord - the coordinates of the distorted structure. Never remove the .coord files, otherwise a restart is not possible!
- p001.log - the output file of the energy and gradient run.
- p001.dipole - the dipole of p001.coord, is a result of the energy step.
- p001.grad - the default gradient file of Turbomole which contains also the energy, gradient norm, coordinates and gradients
For each
.coord file there must be corresponding
.log,
.dipole and
.grad files.
If one of those files is empty, especially if the
.dipole and
.grad file are not complete,
remove the .log, .dipole and .grad files and restart NumForce. This one step will be re-done in the next NumForce run.
Keeping this in mind it should not be a big issue to use a queuing system with short queues only. By the way: Perhaps you can use several CPUs at the same time... if you have PBS as queuing system which exports a proper $PBS_NODEFILE, NumForce will automatically use several CPUs/nodes.
Regards,
Uwe
