On 23/06/2011 12:38 PM, xiaodong huang wrote:
Dear Justin and gmxers
Thank you so much for your helpful hints, but I am wondering how to
check if I have
obtained "proper distributions for the desired ensemble", as suggested
in your email?
My thoughts are:
1. As for NVT ensemble, I need to check if the velocity (or speed) of
molecules in my simulation
follows a Maxwell-boltzmann distribution, and the velocity is for
the center of mass of these
molecules, not for individual atoms, right?
Atoms are physical. Molecules are a convenient grouping of them to aid
human understanding.
2. Can I also check the distribution of T in my simulation (I output T
every 100 steps or so, and
make a statistics)? It is said in NVT ensemble, T follows a
gaussian distribution, according
to central limit theorem, and the square of its variance is
2T*T/(3N). Is that correct?
g_energy will report some descriptive statistics.
3. As for NPT ensemble, how to check if I have get a correction
distribution from pressure
coupling? Can I also check the P or V distribution? If so, what is
the correct distribution I
should get? I find some clues at "Understanding Modern Molecular
Dynamics
J. Phys. Chem. B 2000, 104, 159-178", but this paper just discuss
some simple systems
in T,P coupling section, and I still do not know the correct
distribution I should
get for my own simulations (e.g. protein in a box of water, NPT).
You need to use a small system that you can exhaustively sample, deduce
what the statistical distribution of your observables should be, and see
what you get. Or you can simply rely on the use of algorithms that have
been shown to produce correct distributions, and trust/verify that
GROMACS has implemented them correctly.
Mark
4.Are there any other quantities I need to check to make sure I have
sampled a correct
ensemble? I think to check the distribution of v, T, P V should be
enough for this purpose,
lthough to check more quantities is a plus, right?
Any suggestions or comments are very welcome. Thank you.
yours
xiaodong
Research School of Chemistry
ANU
On Tue, Jun 21, 2011 at 11:21 PM, Justin A. Lemkul <jalem...@vt.edu
<mailto:jalem...@vt.edu>> wrote:
xiaodong huang wrote:
Dear gromacs-ers
I am using stochastic dynamics integrator (integrator = sd),
so most of the time, I only need to adjust following
parameters for temperature and pressure coupling: tau_t pcoupl
tau_p (ref_t is room temperature and ref_p is 1 atm,
compressibility can be taken from experimental numbers).
In the manual, a tau_t between 1 and 2ps is recommended for
the sake of simulation stability, but I see someone use a
tau_t of 0.1ps with the same integrator (integration was
performed with Langevin dynamics,49 with a reference
temperature of 300 K and a weak frictional constant of 10
ps-1,) when simulating water solvent. I also see someone use a
tau_t of 0.2ps with the same integrator (To obtain a
isothermal–isobaric ensemble at 293 K, a leap-frog stochastic
dynamics integrator16 was used to integrate the equations of
motion. The inverse friction constant was set to 0.2 ps.) when
simulating some organic solvent. I check the references
mentioned in these gromacs papers so I am pretty sure they are
using the same integrator. So I am a bit confusing here, what
tau_t should I use, the number between 1 and 2ps as
recommended by the manual, or the numbers below 1ps, as
reported in these papers? Does this parameter depend on what
solvent (water, cyclohexane) I use? Can I just use any number
between 0.1ps and 2ps and check if my simulations look fine or
there is some ‘best’ number for a particular solvent?
This was discussed recently:
http://lists.gromacs.org/pipermail/gmx-users/2011-June/061992.html
When I use berendsen or Parrinello-Rahman, the same questions
apply: Is there some ‘best’ number for a particular solvent,
or I can just use any number between 0.5ps and 5ps and check
if my simulations run well?
When I read papers, I find many different pressure coupling
constant (tau_p) ranging from 0.5ps to 1ps (water), 1ps to 5ps
(organic solvent) with weak coupling scheme in gromacs. I am
wondering why they use bigger number for organic solvent? If I
use Parrinello-Rahman (it is said to be better than berendsen
in the manual), do I need to change tau_p, or I can just use
the same tau_p as berendsen?
Time constants are a bit empirical. The Berendsen algorithm is
more forgiving; it relaxes very quickly and thus low values of
tau_t/tau_p are stable. For methods that allow for wider
oscillations (N-H for temperature, P-R for pressure), small
tau_t/tau_p values are unstable due to the nature of these
algorithms. The most important information is whether or not you
obtain proper distributions for the desired ensemble. Any
algorithm can be made to behave artificially rigorously or
artificially relaxed.
-Justin
Thank you so much for your kind help, any suggestions or clues
are very welcome.
Yours
xiaodong huang
Research School of Chemistry
ANU
--
========================================
Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu <http://vt.edu/> | (540) 231-9080
<tel:%28540%29%20231-9080>
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
========================================
--
gmx-users mailing list gmx-users@gromacs.org
<mailto:gmx-users@gromacs.org>
http://lists.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
Please don't post (un)subscribe requests to the list. Use the www
interface or send it to gmx-users-requ...@gromacs.org
<mailto:gmx-users-requ...@gromacs.org>.
Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
--
gmx-users mailing list gmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-requ...@gromacs.org.
Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
