[gmx-users] PME

Michael Brunsteiner mbx0009 at yahoo.com
Thu Apr 7 15:14:06 CEST 2011


Elisabeth wrote:

> The point is in my case there is no option other than ignoring LR since LR
> is not covered by shift or switch functions but at least what PME reports
> for SR is more accurate.

not really ... PME electrostatic energy is a sum of two contributions
LR and SR ... and only the SUM is meaningful, therefore if you use
a switching function you might, in fact, get something thats MORE
accurate than using ONLY the SR part of PME.

> So the decomposed Coulmb. SR terms I am getting
> using energy groups from PME are "reliable ?

see above ... but then again, the PME (or Ewald) SR part
is in fact nothing but the normal 1/r function scaled by an
exponential term so that it goes smoothly to zero at the
cut-off radius, and in that sense you can regard it as the
total Coulomb interaction scaled with yet another tapering
function (like shift or switch)
If you want to make sure to get precise energies, I'd do
the procedure I laid out in my previous mail for a couple
of simple model coumpounds and than compare the energies
with what you get from PME-SR and/or shift/switch for
the same systems (or the same trajectories using
"mdrun -rerun")
And, mind you, here only the RELATIVE energies are
important, not the absolute numbers, as the latter might
differ a lot between different methods, but if your relative
numbers are OK then you should be fine.
by that i mean: if you use two methods X and Y to get
interaction energies (E) for a series of compounds A, B, C, ...
then you should find that (approximately)
E_AB_X/E_CD_X = E_AB_Y/E_CD_Y etc ...
if the two (or more) terms differ a lot than (at least) one
of your methods X or Y gives you the wrong answer.

> BTW: I am dealing with non polar particles i.e alkanes and carbon and
> hydrogen are the only species I have.

does this include the solvent? or ist that water? if so that
won't help, but if you really only have hydrocarbons you will probably get away
with cut-off/shift/switch ...

Anyway, what i would do is this: search for papers of people
who did LIE calculations (linear interaction energy, a method for
approximate binding free energies) and see what they did ... 
Jorgenson, Aqvist, ...

If you are in a rush, just use reaction field electrostatics,
its probably good enough for your purposes.

mic


      



More information about the gromacs.org_gmx-users mailing list