Hi!
On 02/06/2014 01:05 PM, Arash Azari wrote:
> 1- I would like
to simulate a rod-like (stiff) polymer and I have
> tried using dihedral angle potential.
What other potentials do you use? Without pair
bonds, dihedrals are impossible, since nothing is
keeping the particle s together. Also, you need an
angular potential which makes sure that the
dihedrals do not degenerate.
> “Note that usage
of dihedrals increases the interaction range of
> bonded interactions to 2 times the maximal
bond length!”
> I cannot fully understand the reason and the
meaning of the above
> statement.
It just means that the largest distance at which two
particles interact is twice the bond length, which
is simply
the largest possible distance from one of the inner
particles of a dihedral to the tips.
> To prevent bond
broken message, we can increase the
> maximum bond length a little and on the other
hand the dihedral
> potential multiplies it by 2.
No interactions are modified. Just the maximal range
of all short ranged bonded interactions doubles, as
the message states. Also, you cannot just modify the
parameters at will, they are determined by the force
field you want to implement.
> What is the
correct way of using
> dihedral potential for rod-like system? I
mean time step, warm up,
> settings, and simulation.
There is no "correct" way, just different models.
For example, if you use harmonic bonds with a
constant of about 30 to form the pair connections
and angular bonds with a constant around 200 and
angle pi, that would already stabilize a rod fairly
well at the regular timestep of 0.01. I don't
understand what you need dihedrals for, but of
course you can add them as well. That is just not a
standard model, and I cannot tell you how you would
"correctly" set it up. Just what I can tell you that
it will not work without pair bonds and angles in
any case, and then the dihedrals are pretty
superfluous.
> I had a look at
the manual and I found
> these features there, GAUSSRANDOM and
GAUSSRANDOMCUT , only in the
> manual and nowhere else. Did they already
implemented in the codes?
First of all, GAUSSRANDOM/CUT will not solve your
problemm since that is just a wrong setup, no matter
what RNG you use. Both are implemented, but might be
only in the git repository, so just check that out.
> I think one way
to simulate rod-like polymer is using the virtual
> sites, but honestly, in my opinion, the user
manual for virtual sites
> is poor. It will be great if anyone could add
some examples (which
> work!) to this section. I have tried to use
virtual sites a number of
> times, but I could not (sorry, I cannot
remember the details).
No, using virtual sites is not a good idea, since
you potentially create long objects, which
experience high torques. In fact, you don't want
fully stiff rods in MD, they should be a tiny but
flexible. Just like you avoid hard spheres and
rather use Weeks-Chandler-Anderson. As for some
examples, you can take a look at the testsuite tests
for virtual sites. Of course it would be great to
have some more examples, you are welcome to
contribute!
> 2- When we do not use the periodic boundary
condition, why we still
> need the box length? Does the code need the
box length for domain
> decomposition for parallel run?
Not just in parallel, but even on a single core, a
domain decomposition linked cell scheme is used. All
particles outside the given box will be in a
boundary box, making interactions N^2. So to be
fast, better make sure your particles are in the
box.
> 3- I would like to impose spherical
confinement constraint on my
> system and I start with a very large sphere,
then decreasing the
> radius of the sphere gradually. I thought
that the sphere should be
> surrounded by the simulation box, say the
diameter of the sphere
> should be smaller than the box length. In
this way, the simulation is
> too slow because of the large box. If the box
length in this case is
> only used to decompose the domains for
parallel run, may I start with
> a smaller box length?
Again, that will be very expensive, since then the
particle interactions will be calculated in a N^2
fashion. You can shrink the box with the sphere, and
maybe try to limit the number of cells by "setmd
max_num_cells 10000" or so. On the other hand, I
also don't see a point in setting up a box that is
more than twice as large as your target box. We
could even create polymer melts like that.
> 4- Another
question related to the pressure
> calculation in the above system. Does the
pressure calculation
> procedure work in constrained and confined
systems? Especially non
> rectangular system where we cannot use the
change_volume command.
The pressure is just the virial pressure, using the
box volume for normalization. If you have
constraints, you need to correct for the accessible
volume manually by multiplying the pressure with
V_box/V_acc, because Espresso cannot compute the
accessible volume. However, if you have soft
constraints, it is sometimes not clear what is the
accessible volume, that is up to you to decide.
> 5- When we get
the force on the constraint, constraint force
command,
> is it the force per area or just net force
exerted on the constraint,
> e.g. boundary walls.
That is the total force, since force per area would
only make sense for a planar wall. For all others,
you actually need surface perpendicular force per
area, which currently cannot be computed in
Espresso.
Regards,
Axel
--
JP Dr. Axel Arnold
ICP, Universität Stuttgart
Allmandring 3
70569 Stuttgart, Germany
Email:
address@hidden
Tel: +49 711 685 67609