[lammps-users] ewald/pppm restart inaccuracy during volume changes

Okay. I did not pick up on the change in the G vector before.
If you don't like this behavior, you can always set the PPPM parameters
manually using the kspace_modify command.

PPPM initialization ...
  G vector = 0.307172
  grid = 32 32 32
  stencil order = 5
  RMS precision = 3.40947e-05
  brick FFT buffer size/proc = 59319 32768 18252
Memory usage per processor = 53.0592 Mbytes
write_restart d3z.rst.is
PPPM initialization ...
  G vector = 0.308201
  grid = 32 32 32
  stencil order = 5
  RMS precision = 3.29425e-05
  brick FFT buffer size/proc = 59319 32768 18252

This corresponds to case B. Anytime one of the PPPM parameters is modified,
you should expect to get somewhat different energies.


I tried setting the pppm parameters (grid, order and gvector), but still
have the same behavior. I noticed the "RMS precision" isn't the same, so
something must be changing with pppm? See www.hhatch.com/lmp/setpppm

Dear Steve,

Thank you for looking into this. What I would like to do is have no
discontinuity in the energy/forces upon reading a restart. I tried setting
all of the kspace parameters in the example www.hhatch.com/lmp/setpppm but I
still get discontinuity even with the same grid, order, gvector etc. I did
notice a difference in the value for RMS precision, and 'something' appears
to be different between the kspace at the end of the first run and beginning
of the second run.

I apologize that my example is such a large system. I am working on setting
up a water simulation with <1000 atoms to reproduce these behaviors.

Thank you,

ok - please send me a small example that illustrates
the discontinuity (when you have one) and I will take
a look.


Dear Steve and Aidan,

Attached (hhatchPPPM.tar.gz) is a smaller demonstration of restart
discontinuity and enthalpy increases using 216 SPC/E water molecules.

Folder setpppm shows a discontinuity in the kspace energy, and also the
pressure, after reading a restart file. I attempted to set all of the
kspace parameters. It is important for me to remove this discontinuity, if
possible, so that I may use my minimum configurations for future

Folder enthalpyup demonstrates enthalpy increase during fix box/relax
minimization after 26 steps.

Thank you for looking into this for me,

hhatchPPPM.tar.gz (54.7 KB)