[lammps-users] pppm with increasing box size: energy and forces not converging

I am trying to simulate fluid in square nanochannels, so I have 1D periodic boundary conditions. Although I may end up using a cut and shifted potential for the Coulombic interactions (CHARMM or force shifted Wolf potential) instead of a long range method, I want to compare them under the same conditions. Similar to the slab method in LAMMPS (but without a correction or modification of how the Ewald sum is calculated), I have simply used 3D periodic boundary conditions and increased the box size in 2 directions to introduce a vacuum space between the periodic images. I would expect that the long range Coulomb energy and the forces on an atom to converge once the vacuum space is large enough. Using ewald in LAMMPS or PME in NAMD, the change in the forces is down to the 4th decimal place by the time the box size with vacuum is around 3 times the channel width (5.4 nm). However, when using pppm in LAMMPS, the energy and forces never seem to converge. The forces are nearly the same as when using ewald when the box sizes are equal to the channel width, but never stop changing even if the box is 4000 angstroms wide. Is this a problem with pppm, or can I just not use the PPPM method in this way? The results of single point energy and force calculations are below.

Thanks,
Brian

test of vacuum space size for pppm

3 X 3 nm inner X 5nm channel

forces are on atom id 8453, a water oxygen near the wall

boxx (A) boxy (A) Elong fx fy fz

53.972 53.974 -54980.37 10.80908008 4.36380881 -0.10042917
100.000 100.000 -50790.098 10.80876253 4.34555091 -0.08344371
200.000 200.000 -46462.011 10.83453645 4.32803379 -0.04675206
300.000 300.000 -45179.524 10.84430522 4.31743035 -0.02828637
400.000 400.000 -42451.356 10.87369322 4.27927482 0.02616893
500.000 500.000 -40438.77 10.91288781 4.26150540 0.09223980
600.000 600.000 -40551.614 10.91298214 4.26290367 0.08944146
700.000 700.000 -39762.131 10.92869239 4.25732040 0.11986751
800.000 800.000 -39554.806 10.90888970 4.26783174 0.10622168
900.000 900.000 -38639.516 10.93889175 4.24644411 0.15926205
1000.00 1000.00 -38719.035 10.93905155 4.24771912 0.15669713
1100.00 1100.00 -39157.377 10.93754062 4.25280615 0.14244711
1200.00 1200.00 -38808.815 10.93924054 4.24914719 0.15384185
1300.00 1300.00 -38162.86 10.93920202 4.24107474 0.17385646
1400.00 1400.00 -38247.481 10.93935159 4.24234340 0.17110888
1500.00 1500.00 -35208.296 10.94000182 4.18787593 0.30172775
1600.00 1600.00 -35212.121 10.94055211 4.18591346 0.30522862
1700.00 1700.00 -34884.112 10.94107297 4.18207976 0.31775742
1800.00 1800.00 -34699.459 10.94248600 4.17939916 0.32720936
1900.00 1900.00 -34436.838 10.94695957 4.17714020 0.34192838
2000.00 2000.00 -34367.044 10.94676190 4.17513535 0.34598775
2100.00 2100.00 -34354.678 10.94453023 4.17267057 0.34676264
2200.00 2200.00 -34030.492 10.95421089 4.17214412 0.36760796
2300.00 2300.00 -33936.078 10.95538261 4.17045509 0.37370040
2400.00 2400.00 -33977.084 10.94993252 4.16743043 0.36978090
2500.00 2500.00 -33689.793 10.96169118 4.16812996 0.39111506
3000.00 3000.00 -29995.937 10.99142127 4.07365522 0.68140344
3500.00 3500.00 -29520.536 11.03057522 4.08010542 0.74431350
4000.00 4000.00 -27280.629 11.02949388 4.00030969 0.95573536

test of vacuum space size for ewald

3 X 3 nm inner X 5nm channel

forces are on atom id 8453, a water oxygen near the wall

boxx (A) boxy (A) Elong fx fy fz

53.972 53.974 -61216.287 10.79736285 4.37177624 -0.10102428
70.000 70.000 -61216.267 10.78521758 4.36624089 -0.10300801
90.000 90.000 -61216.254 10.78108524 4.36404957 -0.10503890
100.00 100.00 -61216.25 10.78016570 4.36345442 -0.10564891
120.00 120.00 -61216.246 10.77898069 4.36280361 -0.10643741
140.00 140.00 -61216.244 10.77831080 4.36241728 -0.10691606
160.00 160.00 -61216.242 10.77795483 4.36222659 -0.10718116
180.00 180.00 -61216.241 10.77767677 4.36205417 -0.10736444
200.00 200.00 -61216.24 10.77750654 4.36193845 -0.10749968
400.00 400.00 -61216.237 10.77693067 4.36162080 -0.10789422

I'll let Paul comment on this.

Steve

Brian,

Make sure that you’re using a larger PPPM grid when you increase your box size. LAMMPS should set that automatically for you if you allow it to choose your grid size. If you’re already doing that, you might also try adjusting the tolerance to be more accurate (i.e. 1e-5 or 1e-6), or setting the grid size manually yourself. See the mesh keyword here: http://lammps.sandia.gov/doc/kspace_modify.html

Paul