Dear amber experts and users,
I think I've discovered a bug in cpptraj that we should be aware of (Or
else I'm doing something wrong).
I have been troubleshooting some odd behavior from the cpptraj diffusion
output data - there would be large jumps in the MSD randomly, especially
accentuated by larger and more viscous solvent molecules (like ionic
liquids). The first thought was of course box-boundary error trying to
unwrap the periodic boundary. I dismissed this after verifying that when
molecules crossed one boundary the MSD remained intact. Later on I came
back to that thoery though and isolated that the problem is when one
molecule sitting on the box edge or corner happens to cross the boundary in
two different dimensions during the same frame change. In this scenario the
wrapping of the second periodic boundary is not accounted for and the MSD
jumps up by n * L^2/N (L= simulation box length. n= number of molecules in
the mask double wrapping that frame, and N = the number of molecules in the
mask). Although in large water-solvated systems there's enough molecules
and a small enough surface area/volume ratio that these jumps aren't
readily noticed they do still artificially increase the simulated diffusion
rates output by cpptraj.
Am I doing something wrong or is this a persistent problem throughout
cpptraj calculations? Is there a patch/fix for this? For the moment I'm
just using the *_x and *_y diffusion data because in my system there should
be no dimensional bias and it seems the z contains the vast majority of the
errors (presumably because the first wrap-around is accounted for
accurately leaving most often the z to receive the error).
This is my current theory anyway. Has anyone else observed this behavior?
Is there a better fix than just using X and Y and assuming symmetry?
Thanks!
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Mar 08 2016 - 13:30:03 PST
