Re: RDF calculation with NAMD

From: Ana Celia Vila Verde (
Date: Tue Jan 21 2014 - 11:23:57 CST


See replies to your specific questions below.

I hope it helps,


On 1/14/14 10:05 PM, Dian Jiao wrote:
> Hi,
> I am trying to calculate the radial distribution function for some
> metal ions in water (zn, cs, etc). Here is what I plan to do:
> 1. solvate ion in water box
> 2. run npt to equilibrate until volume of the water box stabilizes.
> 3. switch to nvt for production run.
> Since I am not an expert of NAMD, I have the following technical
> questions:
> 1. for step 2, should I run several short npt (update cellBasisVector1
> etc every time) or a long npt?
> 2. How to plot the dimension fluctuation of the simulation box vs
> simulation time?
Load the equilibration run into VMD, then select
"extensions"->"analysis"-> "namd plot" -> "volume" to see how the volume
changes in time. In reply to your question 1, a single equilibration
run should suffice but it should be long enough that the volume of the
box is stable for a few ns. *
*Keep in mind, though, that the fact that you've equilibrated the box
volume does not mean that your system as a whole is equilibrated.**
Equilibration should always be done for at least the same time as the
longest relaxation time in your system that can reasonably be captured
in a simulation.* This means that 1) you need to think about the
processes that matter to you in your system, 2) you should calculate
their relaxation times and, 3) if necessary, prolong your equilibration
run. For example, for your system you might care about the residence
times of waters in the first and second hydration shells of the ions.
Using your initial equilibration run, you should calculate survival
functions of waters in these layers, and then extract residence times.
It might happen that the residence time of waters in the first hydration
shell is so long that they never leave that shell during your
simulation; if this happens, then this residence time might be longer
than what you can reasonably simulate for, i.e., you cannot fully relax
your system. This is fine provided you are aware of it and make it
clear in your papers. However, let's say that the residence time of
waters in the second hydration shell is 2 ns; this means that you
should prolong your equilibration so that it is at least 2 ns. The
inaccessible relaxation times in your simulated system imply that your
production runs will give some, but not complete, insight into your
system, because they will not fully sample configuration space. This is
fine, again, provided you are aware of it and make it clear in your papers.
> 3. For prodution run, is it better to run nvt or npt?
It depends on what you want to calculate. Provided you have a large
system I would do NVT with a very weak thermostat, in case you are
interested in dynamic quantities.
> 4. what is the best way to calculate rdf? Is there any script available?
Again, load the trajectory in VMD, then select
"extensions"->"analysis"-> "radial distribution function".

> Thanks,
> Dian

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:20:25 CST