[gmx-users] using PME option

[Erik Lindahl]

Alan Wilter Sousa da Silva wrote:

>Dear Drs. Lindahl and van der Spoel,
>
>	First, thanks a lot.
>
>	PME values is now clear to me.  However since LJ interactions take
>the most of computational time I would like to discuss more about the VDW
>options.
>
All nonbonded interactions (coulomb+LJ) take a lot of computational time.

>
>	So after the GMX manual.  If I opt for Cut-off (with rlist = 1.0)
>I have to set rvdw >= rlist (e.g. rvdw = 1.4) and rvdw-switch become
>useless, right?
>
>	If I opt for Shift, according to the GMX manual (p. 122) rlist
>should be 0.1 to 0.3 nm larger than rvdw, so I have, for example, to set
>rvdw = 0.9 and rvdw_switch = 0.6 (or even 0), am I right?
>
>	Which option is faster: Cut-off or Shift?  I think the first
>	Which option is more precise? I think Shift
>
>Many thanks in advance,
>
I would use long-range LJ (i.e. cut-off & rvdw=1.4) if you need high 
accuracy.

Erik