Re: Setting PME Grid Size

From: Jim Phillips (
Date: Wed Apr 13 2011 - 08:55:23 CDT

I think you have it backwards. The PME grid sizes are fixed throughout
the simulation and if your cell shrinks you have a finer grid, which is
fine. The PMEGridSpacing parameter is optionally used to set the grid
sizes based on the initial cell dimensions. The calculation is just
dividing the initial cell dimension by the spacing and then rounding up to
an integer with small integer factors so that the FFT is efficient. If
you restart the simulation after the cell has shrunk the calculation will
be repeated with the new cell dimension, yielding a smaller grid. There
might be an error reported if the cell grows too much in some dimension.

The tolerance is the error at the cutoff distance.


On Wed, 30 Mar 2011, da xue wrote:

> Dear NAMD developers and users,
> Recently, I run a NPT ensemble for the glycerol solution by NAMD. The
> initial simulation box is definitely not at equilibrium and the distances
> between molecules are man-made large. So, after several steps, the
> simulation box will shrink obviously to achieve a reasonable distribution of
> molecules.
> In the Configuration file, I set the GridSizeX, Y and Z according to the
> used's guide without setting the PMEGridSpacing. For example,
> cellBasicVector1 31 0 0
> cellBasicVector2 0 44 0
> cellBasicVector3 0 0 51
> cellOrigin 0 0 0
> PMEGridSizeX 32
> PMEGridSizeY 45
> PMEGridSizeZ 54
> *My questions are below.*
> 1. What value will the PMEGridSpacing hold if I set the above PMEGridSizeX,
> Y and Z?
> 2. Considering the obvious shrink of the initial simulaiton box, will the
> PMEGridSizeX, Y and Z (set according to the initial size of the box) be
> effective throughout the simulation period? How do I set PMEGridSizeX, Y and
> Z when the initial simulation box is larger than the equilibirium box
> obviously?
> 3. How are PMEGridSizeX, Y and Z and PMEGridSpacing converted to each other?
> 4. I imagine. It is the PMEGridSpacing that takes effect throughout the
> simulation period. The initially-set PMEGridSizeX, Y and Z can only be used
> to calculate the PMEGridSpacing and they will be negelected when
> PMEGridSpacing is calculated?
> 5. What does *PME direct space tolerance* really mean? I found several
> different ways to express it. For example, "with a direct sum tolerance of
> 10E-6 at the 9 A direct space cutoff" or "with a direct space sum tolerance"
> .. Which one is more proper?
> With kind regards.
> Lindong Weng
> Dalian University of Technology
> Dalian, Liaoning Prov., P.R.China

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:56:58 CST