[gmx-users] nonormalize option in g_rotacf

[Andrew DeYoung]

Hi, 

I am having some seeming difficulty with the -normalize option in g_rotacf,
which calculates rotational autocorrelation functions of molecules.  

My system consists of 254 SPC/E water molecules.  I equilibrated (3 ns, md
integrator) and ran production dynamics (2 ns, md integrator) in the NVT
ensemble, using Nose-Hoover temperature coupling.  I chose the cubic box
size such that the density of the system is set to ~1 g/cm^3.  After I
completed my production run, I used following command:

g_rotacf -f nvt-md.trr -s nvt-md.tpr -n allatoms.ndx -P 1 -b 0 -e 10 -o
finalresults/normalized_10ps.xvg

where allatoms.ndx specifies all 762 atoms in triplets of atoms, with each
triplet corresponding to one water molecule.  The default setting for
-normalize is yes, so the above command should give me the _normalized_
rotational autocorrelation function, averaged over all molecules.  I do
indeed get a normalized ACF, which has C(t=0)=1 by definition.  (Recently
(http://lists.gromacs.org/pipermail/gmx-users/2011-June/062005.html), Justin
kindly pointed out to me that, per the normalize_acf() function in
src/tools/autocorr.c, the normalization is such that the ACF is 1 at time
t=0.)

But, on the other hand, when I use the option -nonormalize, using the
following command, I get a seemingly strange result:

g_rotacf -f nvt-md.trr -s nvt-md.tpr -n allatoms.ndx -nonormalize -P 1 -b 0
-e 10 -o finalresults/unnormalized_10ps.xvg

When I execute this command, I obtain an ACF that is 1 at time t=0.  This
cannot really be the correct unnormalized result, unless by some remote
coincidence the unnormalized ACF happens to be 1 at time t=0.  

I have posted my results at
http://www.andrew.cmu.edu/user/adeyoung/june12/june12.pdf .  If you have
time, could you please help me think where I may have gone wrong?  At first,
I was convinced that I mixed up the files or did something else silly.  But,
I've redone it, carefully naming, opening, and plotting the results, and I
get the same: both the normalized and the unnormalized ACFs are 1 at time
t=0, and the coordinates in the .xvg files are identical.  I am sure that I
am doing something wrong, but I have not found it yet.  Can you please help
me think what to try next?  I am using Gromacs version 4.5.4.

When I want to turn off normalization of the ACF, should I use
"-nonormalize" (as I have done above, and as seems to be suggested by the
manual), or should I use "-normalize no"?

Thank you very much for your time!

Andrew DeYoung
Carnegie Mellon University