Discussion on hydrogen bond correlation function

[Rajorshi Chattopadhyay]

Dear all,

I am a new user of MD Analysis and want to use this tool to calculate hydrogen bond correlation functions for simple aqueous systems (my trajectory files are in .xyz format).

I was going through the page on Hydrogen Bond Autocorrelation in the MD Analysis manual and have a question on the methodology. The method 'MDAnalysis.analysis.hydrogenbonds.hbond_autocorrel.HydrogenBondAutoCorrel' takes 3 arguments in addition to others - nrun , nsample and sample_time. My questions are:

1. If I specify nrun = x and sample_time = t, then the calculation will create x different time slices of length t from the trajectory. In each run, one of these slices will be used for Cx calculation. Is this so ?

2. If I specify nsample = n and Cx is being calculated for a lag time of t, then n frames each separated by time t will be used to calculate the numerator of Cx. Am I right in my understanding ?

I have never worked with correlation functions before and my questions may be really amateurish. I thank you in advance for your help and suggestions. 

[Hugo Macdermott-Opeskin]

Hi, 

No worries, correlation functions can be a challenge. 

If my understanding is correct each run calculates a Cx of time length t using nsample sampling points within the interval t. 

Each run then takes a slightly different time offset to give different time origins.

Then each  Cx is combined and the whole thing normalised.

 You can then fit to a double exponential (or whatever functional form) and extract characteristic times. 

Hopefully this helps (and that I got this right). Others feel free to weigh in if I havn't.

Cheers

Hugo