[gmx-users] Long range forces, periodic boundaries and aggregation

[Steven Kirk]

Hello,

Firstly, many thanks to David and the others who answer users questions on this list. Your help has been invaluable to me in the past.

I have a large, approximately spherical macromolecule (silica nanoparticle, approx. 4 nm in diameter) which I am investigating in TIP4P water. I put the molecule centered in a cubic box of side 6 nm, then build another cell which is effectively two of these boxes side by side in the X-direction. The resulting simulation cell is then 12x6x6 nm, and I then add TIP4P water using genbox. I am using PME electrostatics, with a FFT grid spacing of 0.12 nm, a pme_order of 6, a time step of 0.001 ps and a simulation run time of 80 ps. 

My original thought was that the two nanoparticles (labelled 'left' and 'right' in the following, due to the x-axis alignment of the cuboidal cell) would drift slowly towards each other along the x-axis due to long-range forces. Instead, what seems to happen in the simulation is that the 'left particle' drifts *downward*, while the  'right particle' drifts *upward*. Neither particle comes very close to the upper or lower side of the simulation box, but the drift is visible in the output trajectory. g_dist tells me that, on average, the centers of mass of the particles get further apart. So far as I can tell, there is no net charge on each particle (how could I check this easily?).

The only reason for this I can come up with is simply a built-in asymmetry due to the non-perfect spherical shape of the particles, such that the nearest copy of each particle lies in the y-direction, corresponding to the direction of drift.

Could I force the relative motion of the 2 particles in the unit cell to be *only* along the x-axis by constraining the motion of their respective centers of mass, and would this introduce any unanticipated artifacts in the calculation ? Is there anything else I can do to make this 'slow aggregation' calculation faster ?

Many thanks in advance,
Steve Kirk

Dr. Steven R. Kirk          <Steven.Kirk at htu.se, S.R.Kirk at physics.org>
Dept. of Computer Science, TMD               http://taconet.webhop.org
Univ. of Trollhattan/Uddevalla   Ph: +46 520 475318 Fax:+46 520 475399
P.O. Box 957 Trollhattan 461 29 SWEDEN