I can't provide a definitive answer-there's not enough information here for that. But here are a few questions to think about:
* How are you measuring diffusion coefficients accurately when the size of your box can adjust, displacing the positions of the water molecules along with it?
* Are you sure that you are handling periodic boundary conditions correctly?
* How many molecules are you using to determine the diffusion coefficent?
From: Moultos, Othon
Sent: Sunday, August 25, 2013 11:48 AM
To: '[email protected]'
Subject: RE: [lammps-users] ewald and pppm give different results for
diffusion of CO2 in water
Thanks for the answer. Well the diffusion is known to have a dependence on
system size. I have checked the case of 5 co2s diffusing in 500, 1000 and
2000 waters. I am calculating the diffusion from the info acquired by the
npt file which contains the velocities, positions and continuation vectors
of the com of molecules. Since I construct the unfolded positions from these
positions it doesn’t matter in which ensemble I perform the simulation. I
calculate both msds and vacfs and I found it strange to have differences
while using ewald and pppm. Am I missing something?
do you have any proof that there is a *systematic* difference between
the two methods?
just getting different results from two trajectories is not
sufficient, it could happen simply from different starting conditions.
since your kspace convergence parameter of 1.0e-4 is rather low,
differences between trajectories (but less so in properly converged
averages) are to be expected.to build quickly.
have you checked for statistical convergence? with so few molecules
obtaining converged results can be difficult.