Topic Options
#15177 - 07/31/07 01:52 PM counter ion RDF normalization NPT OCTA
ilja Offline
Forum Member

Registered: 05/19/04
Posts: 34
Loc: La Jolla, California
Hi,

I was wondering as to what the standard practice is for computing RDFs from NPT trajectories with truncated octahedron water box.

I have a 1.0 M Na+Cl- solution and would like to compute the Na-Cl RDF.

CHARMM has two options for that within COOR ANAL, using either CROSs or MULTi, and another option RDFSol.

My box parameter A is about 40 Angstroms and I would like to compute the RDF out to 16 Angstroms.

I tried the RDFSol, but that segfaults at 16 A, however works with 8 A:

rdfsol rdf seta select type sod end setb select type cla end -
rmax 8.0 nbin 80 iout 20 precise -
firstu @istart nunit @pcs skip 1


Hence I moved to the remaining two options, which give identical result.

I use these keywords to compute the RDF:

coor anal select type sod end - ! sodium
site multi select type cla end - ! chloride
firstu @istart nunit @pcs skip 1 - ! trajectory specification
isdist 20 - ! do the g(r) (here solute-solvent)
mgn 160 dr 0.1 - ! comp. g(r) at MGN points separated by DR
rdsp 16.0 - ! sphere to compute density
rsph 999.9 ! ! we use ALL ions for the calculation

and I don't specify anything for the PBC, although I do setup crystal parameters prior to reading the trajectory. I wonder if that is sufficient for this RDF tool to work with NPT and OCTA crystal facility.

I get somewhat strange looking RDF which does not seem to plateu even at 16 Angstroms and instead of setlling at 1 goes way below to below 0.5.

I wonder if there are experts out there, who could provide some advise as to how properly compute these kinds of RDFs using CHARMM.

Thanks,
-Ilja


Edited by ilja (07/31/07 04:21 PM)

Top
#15178 - 08/16/07 02:41 AM Re: counter ion RDF normalization NPT OCTA [Re: ilja]
ilja Offline
Forum Member

Registered: 05/19/04
Posts: 34
Loc: La Jolla, California
Hi again,
Trying to help myself... and may be others...

I verified that the options I used (see the original post) in connection with CRYStal to compute RDFs indeed give wrong results, particularly at the tail. I checked that for O-O RDFs of water.

So I decided to try my luck with the old? PBC specifications, basically pretending that I had a cubic box with relatively small dimensions that would fit inside the truncated octahedron. I don't think that this is a good solution, though.

Does anybody have a better suggestion?

Thank you for your time.

Top
#15179 - 08/16/07 03:07 AM Re: counter ion RDF normalization NPT OCTA [Re: ilja]
lennart Online   content

Forum Member

Registered: 09/25/03
Posts: 4741
Loc: ~ 59N, 15E
You have two options:
1/ Rerun the simulations with a cubic box. The octahedral box does not save you anything for a solvent system; it is only useful if you want to solvate a macromolecule.
or
2/ Code the more general transformations for COOR ANAL; it is not difficult (see eg the COOR HBOND code) and if a real need for this can be demonstrated I'd be willing to put it higher on my priority list.
_________________________
Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden

Top
#15180 - 08/18/07 02:35 AM Re: counter ion RDF normalization NPT OCTA [Re: lennart]
ilja Offline
Forum Member

Registered: 05/19/04
Posts: 34
Loc: La Jolla, California
Thank you, Lennart, for your reply.

The reason I used OCTA was to get longer separation PMF/RDF for the ion-pair. I computed the PMF/RDF using a distance-based restraint (via the revised HMCR (due to Michael Feig)) and 1D WHAM first, and then wanted to compare it to the PMF/RDF from the free MD. I felt it would be a good case to demonstrate the role of the Jacobian (2k_BTlnR) correction necessary for the distance-based PMFs that many people unfortunately forget or even never think about (see a very nice recent paper about this in ChemPhysChem 2007, 8, p162-169).

I thought that for a given number of water molecules and ion pairs OCTA will give me a longer range for the PMF/RDF calculations than CUBI when using distance-based umbrella sampling.

If you could find some time to implement the transformation for the OCTA to compute correct RDFs, I would greatly appreciate and acknowledge that.

Thank you,
-Ilja

Top
#15181 - 08/27/07 01:40 PM Re: counter ion RDF normalization NPT OCTA [Re: lennart]
ilja Offline
Forum Member

Registered: 05/19/04
Posts: 34
Loc: La Jolla, California
Hi,

I tried to follow one of Lennart's suggestions and re-run the simulation with the CUBIc box, again in NPT ensemble. It seems that RDF calculations cannot properly handle NPT trajectories even with the CUBIc box.

My question is, if I were to compute the average box size over the NPT trajectory and then feed it to the RDF through [x,y,z]box, would that give me a correct RDF?

-Ilja

Top
#15182 - 08/27/07 04:07 PM Re: counter ion RDF normalization NPT OCTA [Re: ilja]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8373
Loc: 39 03 48 N, 77 06 54 W
You don't need the average values, but you do need to specify values (initial values would be okay) for XBOX etc. for COOR ANAL, in order to for it to understand that there are PBC present.

Top
#15183 - 08/27/07 04:22 PM Re: counter ion RDF normalization NPT OCTA [Re: rmv]
ilja Offline
Forum Member

Registered: 05/19/04
Posts: 34
Loc: La Jolla, California
Thank you Rick,
Indeed, the output is not sensitive to the specified [x,y,z]box values. My final script looks like that:

#######################################################
coor anal select type sod end - ! sodium
site multi select type cla end - ! chloride
firstu @istart nunit @pcs skip 1 - ! trajectory specification
isdist 20 - ! do the g(r) (here solute-solute)
mgn 160 dr 0.1 - ! comp. g(r) at MGN points separated by DR
rdsp 16.0 - ! sphere to compute density explicitely
xbox @size ybox @size zbox @size - ! arbitrary size
rsph 999.9 !
#######################################################

-Ilja

Top

Moderator:  John Legato, lennart