Re: [ESPResSo-devel] harmonic potential

[harmanda]

Hi All,

Just a small contribution this interesting discussion about harmonic  
potentials.

A) First of all harmonic potentials are used in the atomistic modeling  
to mimic the real chemical bonds. They are good approximation for such  
cases as long as we don't model extreme cases where chemical bonds can  
break (very rare if we consider that a chemical bond has a strength of  
about 1-2 EV). For such extreme cases these potentials are wrong.

B) Concerning the CG models, the case is different because such  
harmonic potentials are much softer than the atomistic ones. As  
Burkhard said, harmonic potentials are good only for model cases like  
a comparison with a Gaussian chain. But if we simulate such systems  
under non-equilibrium conditions (eg. shear flow) then bonds can be  
very long, causing problems in the parallelism. As Axel explained, the  
length of the "bond" can become larger than the length of the box in  
the cell lists. But this is completely unphysical, ie. the infinite  
extensibility of the "bonds" is an artificial wrong result of the  
model. This is also the main reason why the Rouse model gives  
completely wrong results in the non-equilibrium dynamics of polymer  
melts. There is a clear solution , which we all know, to this problem  
(at least for most cases): the use of the FENE bonds.

C) Now for the case that someone needs to model problems where bonds  
are forming and breaking during the simulation (I think there was such  
a problem in this list some time ago), then anyway to use harmonic  
potentials is not the best choice. Morse, or even Lennard-Jones, type  
of potentials are better.

From the practical point of view of course a simple solution is to  
use a maximum distance (or better as Markus proposed a maximum energy)  
as the cut-off of the bond. But both solutions are purely empirical  
and more important they cause discontinuities in potentials and in  
forces. This would have a strong impact on the results.

So I would propose to encourage people in the User Guide to not use  
harmonic potentials for CG simulations of problems like  
non-equilibrium flows or where you might have forming/breaking bonds.  
For the latter Morse type potentials going smoothly to zero (using for  
example a spline) are much more physical.

Any more ideas-suggestions?

Best,
Vagelis

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.