[gmx-users] Determining onset of melting

[Bjørn Steen Sæthre]

Hi people,
I an relatively new to MD-simulations, having acquainted myself with the
GROMACS system for almost a year now.
I am using Gromacs 3.3.1(i.e the last stable distribution)
I am studying a solid-liquid interface at the moment, in conjunction
with some melting simulations, and have some questions

I am using the Nose-Hoover thermostat, with coupling time tau_t=0.1ps,
and the Parrinello-Rahman barostat with coupling time tau_p=0.5ps,
simulating 600000steps at dt=2fs, close to the experimentally measured
melting point of the solid(which of course may not correspond well to
the melting point in the model system.)
My pressure statistics is not very precise, with a typical average of
-200+-600Bar(The aimed for pressure being 1Bar, 
Concerning this I have one questions:
1. How do I make the statistics better without increasing the number of
simulation steps? I have already tried several different coupling
strengths(times) with no significant improvement in the statistics. 
I have also tried berendsen pressure coupling, to try avoiding
oscillations, with limited success. Any ideas? 
(Are large pressure fluctuations necessarily unavoidable close to
phase-transitions?) 

I further have some questions regarding analysis of my output
2. What is the best quantitative criteria, easily obtainable in GROMACS
for ascertaining onset of melting in a solid crystal where hydrogen
bonding is important?

I come to think of the following 3 criteria myself:
-The rms fluctuations of the atoms from their lattice positions
exceeding a certain fraction of the lattice spacing. (Lindemann's
criterion)[Phys Rev A 184(1), p. 233-]
3. How can I use the analysis tools(e.g g_rmsf) to get a quick readout
of Lindemann's criterion with minimal further processing 

-The "smearing-out" of peaks in a suitably normalized rdf. 
4. Does there exist an objective measure of this valid for solid-liquid
transitions in general(something like a width-of-peak measure) , or do I
have to construct one for each particular solid-liquid system.

-The ratio of hydrogen bonds in the crystal to hydrogen bonds in the
liquid, not involved in the crystal bonding configuration.

5. Have I forgotten another relevant measure easily accessible in the
Gromacs tools? (Can I use the extent of the qasi-liquid-layer via
g_density perhaps?)

Any relevant suggestions received with thanks
-- 
---------------------
Bjørn Steen Saethre 
PhD-student
Theoretical and Energy Physics Unit
Institute of Physics and Technology
Allegt, 41
N-5020 Bergen
Norway 

Tel(office) +47 55582869