Bo,
Please, always "reply to all" to keep the list aware of the
discussion. This will improve your chances of hopefully being helped
by more knowledgeable people.
Take my next comments with skepticism as I am far fro being an expert
in this topic. The last phase transition research I did was years ago
and never used NPH ensembles.
Charlos,
Thanks for your reply! Just as you said, I tried the "fix temp/rescale"
command to rescale the temperature of the system of liquid Si in a series of
consecutive NPH runs, each NPH run consists of a former temperature
rescaling (10000 steps, 15ps) and a latter singlely NPH equibration (2000000
step, 3ns). In each next run, the "Tstart" was set the value averaged from
the equibration part of its former run, and "Tstop" was set as "Tstart-5",
but I encountered some difficulties:
I am a bit confused by your reference to Tstart and Tstop. Do you have
a thermostat coupled to
the system? I don't think that's what you are suppose to do (if that
is indeed what you did). The idea
is just to rescale the velocities at time=0 and let the system
equilibrate. Then compute the temperature (T) after equilibration. As
long as you have kinetic energy in your system you have temperature.
For example
an NVE ensemble has a well define T in the absence of a thermostat.
Next, repeat this process in a series of steps where the input of each
run is the equilibrated system of the previous one. Each run gives you
a (T,H) pair that you then plot to search for the transition. The
monotonic behavior of that curve should aid you identify the
transition. There is no T changing in time during each simulation (in
the production part). That's my take from my quick read to the
reference I sent.
Maybe someone else with more tangible expertise in this particular
problem will scoop in to help you further.
Carlos
PS: I like the "Charlos" instead of Carlos in your previous response.
Was that a mix of Charles and Carlos?