force constants for umbrella sampling

[Ali]

Dear CP2K users,

This isn't really a *technical* question about how to do umbrella
sampling simulations in CP2K but more an issue of choosing the
appropriate force constant for umbrella sampling simulations (for the
harmonic restraint). My understanding is that for umbrella sampling to
work well, one needs good overlap in the distributions and the
probability distributions from the simulation should be close to where
the restraining potential is put. I seem to have both these criterion
in my runs but I am finding it very hard to match points at the
boundaries as there is a lot of curvature at the ends of the unbiased
potential extracted from the runs.

However I discovered 2 things that I'd like some input on:

* I have been using NVE instead of NVT for the umbrella sampling.
Technically the equations for umbrella sampling work only with NVT or
NPT. Can this be a possible source of the problem?

* Should the force constant of the biasing potential 'K' be similar to
the force constant of -kT log (P(r)) ? Or should it work even if the
force constants are very dissimilar?

Any ideas/help would be appreciated.

Ali