[gmx-users] g(r) does not go to 1 at long r -- bug in g_rdf?

[Pablo Englebienne]

Hi,

I tried to calculate the radial distribution functions for a simple system: a 5nm a side cubic box with 10 Ne atoms and 10 Ar atoms, simulated for 100ns in NVT @ 300K. I was expecting to get an RDF with a peak, stabilizing to 1.0 at long distances.

This was the case for the Ne-Ar RDF, but not for the Ne-Ne or Ar-Ar RDFs, which stabilize to about 0.9. I believe this is due to a problem in the normalization of the histograms with respect to the number of pairs available: there are N*N pairs for the Ne-Ar, while N*(N-1) for the Ne-Ne and Ar-Ar case.

Did somebody else find an issue like this? I think that the issue may become not evident for a relatively large system, as N*N ~ N*(N-1) for large N.

I put the relevant files if somebody wishes to reproduce it here: https://gist.github.com/4085292

I'll appreciate input on this and I can also file a bug if deemed necessary.

Take care,
Pablo

-- 

Dr. Pablo Englebienne
Postdoctoral Researcher

*TU Delft / 3mE / Process & Energy*
/Engineering Thermodynamics (ETh) group/

Building 46
Leeghwaterstraat 44, room 030
2628 CA Delft
The Netherlands

*T* +31 (0)15 27 86662 <tel:+31152786662>
*E* p.englebienne at tudelft.nl <mailto:p.englebienne at tudelft.nl>