VMD-L Mailing List

From: cipitaua_at_gmail.com
Date: Wed Apr 23 2008 - 07:50:12 CDT

Ok I verified that using a shorter interval the error decreases. But I
think that the algorithm should center on the interval and not on the
upper exreme.
As it is, the peaks ALWAYS overestimate the correct average.

Best,
Rob.

On marted́ 22 aprile 2008, Axel Kohlmeyer wrote:
> On Tue, 22 Apr 2008, cipitaua_at_gmail.com wrote:
>
> RC>
> RC> Hello Axel,
> RC>
> RC> the attached file is a simple Si10 cluster in which you can
> manually RC> measure (through Mouse->Label->bonds) the nearest
> neighbour distance to RC> be less or equal to 2.60 angstroms.
>
> please re-run the g(r) analysis with a smaller interval size.
> your case is _very_ borderline and depends on how you define
> the r value that is plotted. the current output does not center
> on the interval but prints the r up to which the g(r) is computed.
> if you use a shorter interval, e.g. 0.01 you'll see that most of
> your data got "rounded up" into the 2.7 bin (which is the 2.6 to 2.7
> bin in practice).
>
> you cannot take the label values as reference, since you don't
> see whether you have 2.603 or 2.597. you are in so far correct
> that numbers seem to get rounded up a little bit too much. but
> the deltar=0.01 results look more reasonable.
>
> RC> Now evaluate the g(r) and click on the highest point of the first
> peak; RC> it is centered around 2.71 Angstroms instead of ~2.60.
>
> please note that those numbers also depend a little bit on
> where you have the cursor focus. just click multiple times
> when moving the mouse a tiny bit and you'd see a difference.
> the proper numbers you can only get when writing them to a file.
>
> cheers,
> axel.
>
> RC>
> RC>
> RC> On marted́ 22 aprile 2008, Axel Kohlmeyer wrote:
> RC> > On Tue, 22 Apr 2008, cipitaua_at_gmail.com wrote:
> RC> >
> RC> > RC> Hello,
> RC> > RC>
> RC> > RC> I noticed that the first peak of a g(r) calculation with
> the vmd RC> > RC> extension is slightly shifted respect to the
> expected value. RC> > RC> Check this out with a very well ordered
> structure (for example RC> > RC> graphene).
> RC> > RC> In my case (cristalline silicon nanoclusters) it gives 2.70
> RC> > instead of RC> 2.43.
> RC> > RC>
> RC> > RC> Anyone confirms?
> RC> >
> RC> > no. the peaks are exactly where they should be in my tests.
> RC> > e.g. in rigid spc/e water it get the intramolecular peak at
> RC> > exactly 1. angstrom.
> RC> >
> RC> > please provide a coordinate/trajectory file that reproduces
> this, RC> > explain how you obtained it, and with which code you get
> the RC> > different g(r).
> RC> >
> RC> > thanks,
> RC> > axel.
> RC> >
> RC> > RC>
> RC> > RC> Thanx,
> RC> > RC> Rob.
> RC> > RC>
> RC>
> RC>