[gmx-developers] Re: Force calculation
cxing at ucdavis.edu
Tue Oct 17 01:57:56 CEST 2006
Plus, with the first case r=z running EM (steep)
the result gives potential energy reaches xxx (a reasonable number) and max
force = inf on atom xxxx, norm of force = xxx ( a huge number).
while with force in z component only (set fx, fy to 0, actually) running EM
potential energy, max force and norm of force are all reasonable.
On 10/16/06, Chenyue Xing <cxing at ucdavis.edu> wrote:
> Dear developers,
> For non-bonded interactions, I know there's force-calculation carried out
> in nb_kernel. Is there any other place in the code involving a calculation
> of interactions? For example, the neighborlist setting part?
> I noticed that in the nb_kernel code, it first computes the distances
> between 2 atoms, then uses that distance r to look-up for energy and force
> values from VFtab array. Then it computes the force components fx, fy, fz.
> I want to test a new interaction type with z-dependence only. So I first
> set r =z instead of r=sqrt(x^2+y^2+z^2),but this won't work. But if I keep r
> as the original, then eliminate the fx, fy components to have a force in z
> direction only, the program works fine.
> As what I really expect is the first r=z case, I wondered if it's b/c of
> the system not equilibrated (I saw a sudden increase in distance which led
> to a huge nnn so that VFtab[nnn] out of bound). So I tried to change the
> nslist update freq to 1 and the program stuck somewhere, doing nothing. If I
> use a nslist update freq >= 2, it'll give a huge distance after say 4 steps
> and segmentation faulted.
> I wonder if gromacs computes the force when setting up the neighborlist
> for efficiency. If anyone can help me with the force calculation info, it'll
> be of great help for me to track down the error.
> Sorry for the long and messy email and looking forward to the replies!!
> Any response is highly appreciated!!
> Thank you very much for your time.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the gromacs.org_gmx-developers