[gmx-users] Re: Neigborsearching, Electrostatics and vdw options

[ahmet yıldırım]

Dear users,

*Berk Hess says:
[gmx-users] a) rlist vs rvdw/rcoulomb size confusion, and b) reduced units
Mon Jul 16 14:17:12 CEST 2007*

There are three options in Gromacs.

The option you want is rcoulomb < rlist and rvdw < rlist.
This works and gives the most accurate and also the most costly integration.

A faster and very commonly used setting is: rlist=rcoulomb=rvdw.
With PME the Coulomb interactions are very small at the cut-off,
as are the LJ interactions.
So with a small sacrifice in integration accuracy one can gain a lot
of performance, also because analytical LJ is cheaper than tabulated.

The last option is rcoulomb > rlist and/or rvdw > rlist.
Then the energies and forces beyond rlist are only updated every nslist
steps.
This gives less integration accuracy but can give a lot of "interaction"
accuracy
at a small computational cost.

*Justin A. Lemkul says:
[gmx-users] Twin-range cut-off
Tue Sep 13 01:25:28 CEST 2011*


A twin-range cutoff just means that your short-range cutoffs aren't
all the same
value, such that they form two interaction zones.  Within the shortest, the
neighbor list is updated every step.  Between the shortest and longest cutoffs,
the neighbor list is updated every nstlist steps.  For instance:

rlist = 0.9
rcoulomb = 0.9
rvdw = 1.4

are common settings for Gromos96 force fields (in conjunction with PME).  Thus
there are two interaction zones - the first is if two atoms (or charge groups,
depending on the algorithm) are within 0.9 nm, and the second is if the two
interacting species are beyond 0.9 nm but within 1.4 nm of each other.

rcoulomb: distance for Coulomb cut-off (nm)
rvdw:distance for LJ or Buckingham cut-off (nm)
nstlist: neighbor list update frequency
rlist: cut-off distance of the short-range neighbor
Twin range cutoff consists of rcoulomb and rvdw, isnt it?

Let's imagine a sphere(two concentric spheres):
radius of the inside small sphere=rvdw
radius of the big sphere=rcoulomb
distance between two of our nested spheres:rlist
is this approach correct?
I could not understand the fourierspacing and rlist.


Thanks in advance



21 Mart 2012 20:53 tarihinde ahmet yıldırım <ahmedo047 at gmail.com> yazdı:

> Dear users,
>
> I have two configuration as the following related to Neigborsearching,
> Electrostatics and vdw options. I checked the literature:
> Generally the rlist, rcoulomb and rvdw have used as the following.
> rlist=1
> rcoulomb=0.8
> rvdw=1.4
>
> *Is there much difference between the following two options in the
> calculation/the results? Is there one significant difference between the
> two options. If yes, then what is it? What is relationship between rlist,
> nstlist and rvdw/rcoulomb?*
>
> Furthermore,
> fourierspacing         = 0.16
> or
> fourierspacing         = 0.12
> *difference between these two options?*
>
> *1.choice*
> .....
> ; NEIGHBORSEARCHING PARAMETERS
> nstlist                 = 5
> ns-type               = Grid
> pbc                    = xyz
> *rlist                    = 1.0  *
>
> ; OPTIONS FOR ELECTROSTATICS AND VDW
> coulombtype              = PME
> pme_order                = 4
> fourierspacing          = 0.16
> *rcoulomb                 = 1.0*
> vdw-type                 = Cut-off
> *rvdw                       = 1.0*
> ...
>
> *2.choice*
> ..
> ; NEIGHBORSEARCHING PARAMETERS
> nstlist                  = 5
> ns-type                = Grid
> pbc                      = xyz
> *rlist                     = 0.9   *
>
> ; OPTIONS FOR ELECTROSTATICS AND VDW
> coulombtype           = PME
> pme_order              = 4
> fourierspacing         = 0.16
> *rcoulomb                 = 0.9   *
> vdw-type                 = Cut-off
> *rvdw                       = 1.4    *
> ...
>
> Thanks in advance
> --
> Ahmet Yıldırım
>
>


-- 
Ahmet Yıldırım
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20120322/930b83f1/attachment.html>