[gmx-users] Hydrogen bond autocorrelation function (HBAC)

[Senthil Kandasamy]

I have a few questions about the hydrogen bond autocorrelation function.

While using g_hbond with the -ac option, the relevent output is
hbac.xvg. "xmgrace -nxy hbac.xvg" shows two lines: one of them is the
actual autocorrelation function (the one that decays from 1 to zero
asymptotically and shows fluctuations. ) What exactly is being plotted
as the other line? (the smoother one which looks like a simple
exponential fit). Is it just the fit?

Also, the manual (ver 3.1.1, pg 160) says that a simple estimate for the
H-bond lifetime can be obtained from the integral of C(tau). It also
says that a more-sophisticated analysis can be used to get better
estimates for the H-bond lifetime using multi-exponential fits. I have
seen some papers in the literature which estimate these lifetimes using
multi-exponential fits, but am not convinced about the validity of some
of the methods used. Let's say, I fit the HBAC as a sum of three
exponentials

y=a0*exp(-t/a1) + a2*exp(-t/a3) +a4*exp(-t/a5) 

and I estimate a0 through a5, how would I get a good estimate of the HB
lifetimes from these values? I have seen some papers just average out
the three timeconstants (a1+a3+a5)/3. I am not totally convinced by it.
Is that a valid way to calculate a lifetime?

I would appreciate it if anybody would throw some light on this and
explain how "a better estimate of hydrogen bond lifetimes can be found
using multi-exponential fits".

Thanks.

Senthil