[gmx-users] Force Fields and Electrostatics

[David van der Spoel]

On Wed, 2004-03-31 at 01:21, Dallas Warren wrote:
> Just wondering on a couple of things to do with force fields and the 
> electrostatics used in them.
> 
> Force fields are developed with particular electrostatics in mind, right? 
> i.e. cut-off of a certain length or PME etc.
> 
> What happens then if you then use something different?  Does the ff take 
> into "account" the irregularities (eg with cut-off you get concentration of 
> oppositely charged ions at the cut-off distance when looking at the RDF) 
> that occur and therefore if you use a different electrostatic treatment 
> things start to breakdown?  In other words, do you just stick to what is 
> defined by the ff to be used?
> 
> I have run three simulations of the same system with different treatment of 
> electrostatics.  This is been done using the G43a2 force field.  The 
> different conditions I have used are listed below.  The behaviour of all of 
> them is different to each other.
>          Cut-off 1.5 nm
>          Cut-off 2.0 nm
>          PME
> 
> I have noted that Marrink + Mark (BioChem 41(17) 5375 2002) used a similar 
> system with 1.5 nm, tested out 1.8 nm and saw not overall difference, using 
> Gromos96.  So they settled at 1.5 nm being OK to use.
> 
> I was thinking that may be that the ff is developed on simpler systems such 
> that the important interactions that produce the properties things are 
> adjusted for are under these cut off distances.  So when electrostatic 
> interactions that are over a similar distance to the cut-off, or a bit 
> larger, things can be adversely effected and behave in an unusual manner.
> 
> This is something that probably goes to the heart of the black art of force 
> fields ;-)
Indeed.

My approach is always that when you have a decent model (i.e. potential)
it should not deteriorate when you improve the accuracy by going to a
larger cut-off or using PME. However, things like SPC water were tuned
for a certain cut-off, i.e. they reproduce experimental energy and
density at that cut-off. If you change the cut-off you will get slightly
different properties. For protein simulations I usually use a twin range
for Van der Waals (0.9/1.4) and PME for electrostatic (with rcoulomb =
0.9). Some people argue that the distribution of particles with
considerable Van der Waals interactions is not homogeneous within a
distance of 1.4 nm, and that one henceforth should use a longer cut-off.
There must be some systematic tests out there, but I'm not aware of any
particular paper. For GROMOS96 it was determined that 0.8/1.4 was "good
enough". However most people use slightly larger short range than 0.8.
For OPLS anything goes, that is Jorgensen is quite strict in stating for
his water models that 0.9 nm is the recommended cut-off, however for
alcohols he uses 1.1 for some and 1.5 for others...

So, it's a mess...

-- 
David.
________________________________________________________________________
David van der Spoel, PhD, Assist. Prof., Molecular Biophysics group,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,  	75124 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://xray.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++