[ESPResSo] Metadynamics in ESPResSo

[Tristan Bereau]

Dear all,

I've implemented Metadynamics into Espresso. Metadynamics is a sampling scheme 
that allows to caculate the potential of mean force of a system as a function 
of a reaction coordinate from a canonical simulation. The algorithm relies on 
the history-dependent force that is applied to the system, in order to avoid 
values of the reaction coordinate that have been previously sampled. 
Ultimately, the biased system is able to self-diffuse along the reaction 
coordinate, providing an estimate of the potential of mean force (PMF).

Two reaction coordinates have been implemented: 
 - the distance between two particles (e.g. PMF of the end-to-end distance of a 
polymer)
 - the relative height of a particle with respect to another one (e.g. distance 
of a particle from an interface)
the code can be easily extended to other reaction coordinates.

The scheme is fast due to the implementation in C (rather than through the Tcl 
interface), and other tricks such as a discretization of the reaction 
coordinate into bins and the use of so-called Lucy's functions rather than 
gaussians (avoids calculating exponentials). 

Right now, the metadynamics sampling scheme does *not* support MPI. It would be 
a useful extension but will require me to learn more about particle 
communication between nodes.. Still, the input/output features allow to 
trivially parallelize the sampling scheme by starting independent "walkers" 
that exchange information on a regular basis.

Also, the code works seamlessly with VIRTUAL_SITES, such that one can work with 
groups of particles in order to define molecules or interfaces.

The code, as well as the user guide documentation, was updated to the CVS. 

Let me know if you have any questions or comments.

Best,
Tristan