Coulombic energy on a single flexible SPC water

Hi Steve,

I checked that. Starting from a box of 3x3x3 nm and going up until 20x20x20 nm, the Coulombic energy is not sensitive to the box size.
I checked as well the ewald vs ewald/n summation methods, no changes. The value is almost insensitive to the cut-off for the real/reciprocal space.
The energy changes with the requested precision of the ewald summation.

It really looks like there is a non-zero coulombic interaction somewhere, but I cannot see where it comes from. And why then the total energy is zero also puzzles me.

Thank you for your answer!

Dr. Hegoi Manzano

Molecular Spectroscopy Laboratory
Department of Physical Chemistry, UPV/EHU
Barrio Sarriena s/n, 48940 Leioa, Bizkaia, Spain

e-mail: hegoi.manzano@ehu.es // hegoi.manzano@…24…
http://sites.google.com/site/hegoimanzano

Dear Steve,

I kept working with the checking. I found that the cut-off in the coul/long interaction has a lot of influence on the coulombic energy value. The longer the cut-off the closer to zero. This is the value for the change from real to reciprocal space calculation of the coulombic energy, so I believed it shouldn’t affect so much to the absolute value. Anyhow, increasing this value will increase quite a lot my computational time (from 10 to 100 cut-off the energy decreases from 40Kcal/mol to 3Kcal/mol), so I don’t know how to solve this. Should I move to other k_space styles?

Best

Pd. This stuff is consistent trough several lammps versions (January12 and June12)

Maybe Paul has time to look at this. If I understand
what you are trying to do, you have a single water in a
big box with Coulomb forces within the water turned
off (via special bonds), and long-range Coulombics,
but are seeing large variations in the energy as a function of
cutoff. I doubt there is any issue with LAMMPS for
this, with Ewald or PPPM, and Paul has tested many problem
like this with the long-range
solvers, so possibly he can help.

Steve

Dear Steve,

I kept working with the checking. I found that the cut-off in the coul/long
interaction has a lot of influence on the coulombic energy value. The longer
the cut-off the closer to zero. This is the value for the change from real
to reciprocal space calculation of the coulombic energy, so I believed it
shouldn't affect so much to the absolute value. Anyhow, increasing this
value will increase quite a lot my computational time (from 10 to 100
cut-off the energy decreases from 40Kcal/mol to 3Kcal/mol), so I don't know
how to solve this. Should I move to other k_space styles?

why don't you just post the exact input files that you are using?
that would make discussing them much easier.

axel.

I can take a look at this if you can send me your LAMMPS input script and data file, along with a thorough description of the specific problematic behavior you are currently observing. I don't think that there is a problem with LAMMPS on this, but I may be able to help you interpret your results.

Paul

Thanks for sending me your input (offline). I've run your input, and I think that the answer is a very simple one: you were not looking at the long-range energy ("elong"), which almost perfectly balances out the short-range coulombic energy ("ecoul"). Using regular ewald instead of ewald/n, and changing your thermo_style command to include "elong" like this:

thermo_style custom etotal pe ebond eangle ecoul elong evdwl

I'm seeing the following output:

TotEng PotEng E_bond E_angle E_coul E_long E_vdwl
   0.1760905 0.1760905 0.085831889 0.092906118 33.432795 -33.435442 0
0.035099237 0.035099237 0.00012943707 0.037519857 33.439439 -33.441989 0
0.017449247 0.017449247 0.0046069118 0.015352509 33.442225 -33.444735 0
0.0046059874 0.0046059874 0.00012555741 0.0069825844 33.442863 -33.445365 0
0.00069605974 0.00069605974 0.00065594556 0.002524794 33.44409 -33.446575 0
-0.0014387765 -0.0014387765 0.00028377642 0.00075509007 33.444614 -33.447091 0
-0.0023458011 -0.0023458011 2.5562737e-06 0.00012288346 33.445076 -33.447547 0
-0.0024016372 -0.0024016372 1.4756209e-05 5.233311e-05 33.44525 -33.447718 0
-0.0024625886 -0.0024625886 3.1172603e-06 1.1111524e-06 33.445393 -33.44786 0
-0.0024659248 -0.0024659248 1.0534007e-07 4.0168398e-07 33.445418 -33.447885 0
-0.0024660364 -0.0024660364 6.3074542e-08 2.4602984e-07 33.445425 -33.447891 0
-0.0024661241 -0.0024661241 2.9880312e-09 2.3764998e-07 33.445423 -33.44789 0
-0.0024661552 -0.0024661552 7.1200278e-09 1.722474e-07 33.445425 -33.447891 0
-0.0024661877 -0.0024661877 2.2633884e-09 9.4845169e-08 33.445428 -33.447895 0
-0.0024661877 -0.0024661877 2.2633884e-09 9.4845169e-08 33.445428 -33.447895 0
Loop time of 0.0150001 on 1 procs for 14 steps with 3 atoms

You can tell from the output that the "elong" contribution is where your "missing energy" was hiding.

Paul