rdf normalization confusion

aj...@nyu.edu

unread,

Nov 30, 2023, 5:36:51 PM11/30/23

to MDnalysis discussion

When calculating rdf based on this documentation: https://docs.mdanalysis.org/1.1.0/documentation_pages/analysis/rdf.html#module-MDAnalysis.analysis.rdf

I never get the rdf normalized to 1 at large distances. Also tried setting density=True/False, and that does not help.

Also, rdf.count does not sum up to the number of pairs.

I am unsure about what could be wrong here.

It seems that if I do

density*4*np.pi*np.square(rdf.bins)*rdf.rdf*np.diff(rdf.edges)

I get roughly the correct count in each bin, but according to the documentation, it seems that I already should have a value like that in the rdf structure.

Would be grateful for any advice.

Thank you!

Oliver Beckstein

unread,

Nov 30, 2023, 6:38:42 PM11/30/23

to mdnalysis-discussion

Hello,

On Nov 30, 2023, at 15:36, aj...@nyu.edu <aj...@nyu.edu> wrote:

When calculating rdf based on this documentation: https://docs.mdanalysis.org/1.1.0/documentation_pages/analysis/rdf.html#module-MDAnalysis.analysis.rdf

Is the current version of this document https://docs.mdanalysis.org/stable/documentation_pages/analysis/rdf.html#module-MDAnalysis.analysis.rdf

Which version of MDAnalysis are you using?

import MDAnalysis as mda

print(mda.__version__)

I never get the rdf normalized to 1 at large distances. Also tried setting density=True/False, and that does not help.

What kind of system are you looking at?

Can you share your code and some actual results that show the problem?

Did you try the analysis on a system where you know the answer, e.g., a box of water?

Best,

Oliver

Also, rdf.count does not sum up to the number of pairs.

I am unsure about what could be wrong here.

It seems that if I do

density*4*np.pi*np.square(rdf.bins)*rdf.rdf*np.diff(rdf.edges)

I get roughly the correct count in each bin, but according to the documentation, it seems that I already should have a value like that in the rdf structure.

Would be grateful for any advice.
Thank you!

--

Oliver Beckstein (he/his/him)

email: orbe...@mdanalysis.org
twitter: @orbeckst

GitHub: @orbeckst

MDAnalysis – a NumFOCUS fiscally sponsored project

https://www.mdanalysis.org/

aj...@nyu.edu

unread,

Nov 30, 2023, 9:30:57 PM11/30/23

to MDnalysis discussion

Super, thank you. The problem was the old documentation. When I use norm='none', it seems I get the proper count in rdf.rdf.