[gmx-users] About EWD3DC

[David L. Bostick]

The previous tests seem to suggest that the rule of thumb should be to
increase Lz by 3*Lx or 3*Ly, whichever is larger.  Keep in mind that the
3dc term with sufficient Lz can be viewed as a kind of asymtotic solution
for approximating a 2D ewald sum in x and y.  As such, each image must be
far away along z, compared to the x and y dimensions of the cell. If you
adopt a shorter Lz, you should do a test against actual 2D ewald to be sure
you are okay.

You are right however to be conservative about how long you make z, because
when you add length to Lz, you are increasing the number of k-vectors in
the sum, and this increases computational overhead.  Nonetheless you should
realize that even with the slight computational overhead you will be
getting something very close to 2D-ewald, and since you can use PME, this
will be the fastest 2D-ewald available at this time.  In other words, you
will still be getting your money's worth. Actual 2D ewald is not yet as
fast as 3D PME with one extended cell dimension.

David




-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
David Bostick					Office: 262 Venable Hall
Dept. of Physics and Astronomy			Phone:  (919)962-0165
Program in Molecular and Cellular Biophysics
UNC-Chapel Hill
CB #3255 Phillips Hall				dbostick at physics.unc.edu
Chapel Hill, NC 27599	           		http://www.unc.edu/~dbostick
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

On Wed, 21 Sep 2005, Janne Hirvi wrote:

> Hello!
>
> I have read couple of articles and mailing list comments of David Bostick about
> the slab geometry correction in Ewald sum or actually in my case in PME. But I
> am still wondering if needed empty space in z-dimension perpendicular to the
> slab plane is really comparable to x- and y-dimensions (z = 3*(x or y)) and not
> directly to real z-dimension of the slab. I mean that if I need large
> (x,y-dimensions = 16nm) and thin (z-dimension = 2nm) film for simulations of
> liquid droplets on it, do I have to use huge box (48nm) or is quite ok to use
> box, which is somewhat elongated (droplet + surface + free space).
>
> Thanks for any help and comments!
>
> Janne
>
> -----------------------------------------------------------------------------
> Janne Hirvi, MSc(Physical Chemistry), Researcher
> University of Joensuu, Department of Chemistry, P.O.Box 111 80101 Joensuu, FI
> Tel: +358 13 2513326 & +358 50 3474223
> E-mail: Janne.Hirvi at joensuu.fi & hirvi at cc.joensuu.fi
> -----------------------------------------------------------------------------
>
>
>
>
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
>