What I understand from this is that nature of interactions is repulsive. So when we apply the soft potential energy doesnt go to infinity as there is no r in denominator, so the values can be calculated etc. Since the interactions are repulsive atoms do not move to another overlap potential.
But I still have to unoverlap them sufficiently which means running the simulation sufficiently long enough or ramping up the prefactor in the soft potential to large enough value (Hope sounds correct so far and I havent missed anything). I have tried moving A to (0.0, 10000000.0) and running the simulation for 2 * 1e6 steps still the same error persists. Where am I going amiss. Also is there any other way apart from hit and trial. Also can you give any inputs on setting rc. I looked through the lammps threads and this has been used mostly for polymers and atom_style bonds, In my case is my system too dense 75.5 is box length and dia 10 to allow eventual un-overlap.
Sorry if I seem too dense (pun not intended)
Regards
Anurag
How about putting the large particles on a large lattice
intitially, in a large box. Then shrinking the box,
either via fix npt or fix deform. If they start
with no overlaps, there will be no overlaps
as it shrinks.
Steve
From: Anurag Ranjak <[email protected]...>
Date: Fri, Jan 20, 2017 at 3:30 PM
Subject: Re: [lammps-users] creating fixed no of atoms
To: Axel Kohlmeyer <[email protected]>
What I understand from this is that nature of interactions is repulsive. So
when we apply the soft potential energy doesnt go to infinity as there is no
r in denominator, so the values can be calculated etc. Since the
interactions are repulsive atoms do not move to another overlap potential.
But I still have to unoverlap them sufficiently which means running the
simulation sufficiently long enough or ramping up the prefactor in the soft
potential to large enough value (Hope sounds correct so far and I havent
missed anything). I have tried moving A to (0.0, 10000000.0) and running the
simulation for 2 * 1e6 steps still the same error persists. Where am I going
amiss. Also is there any other way apart from hit and trial. Also can you
give any inputs on setting rc. I looked through the lammps threads and this
has been used mostly for polymers and atom_style bonds, In my case is my
system too dense 75.5 is box length and dia 10 to allow eventual un-overlap.
the soft potential has two parameters and both need to be chose suitably.
you don't just pick some random parameters, but have to find a useful set.
they don't have to be perfectly accurate, just good enough, since you
are going to switch to the proper pair style later.
a simple and straightforward method to find good settings, is to
simply plot both potential functions with selected parameters and
empirically adjust the ones for the soft potential until they are a
reasonable match.
axel.
The methods followed were -
Approach 1 (Mr. Kohlmeyer’s approach)
Plotted the potential for colloid using A_cc = 17.6 and rc = 10.6. Fit the soft potential function to the colloid potential . I get Large negative values for A and a large value for rc(soft). Since the box is only 75.5. I decided to keep rc as 37.5 Using both large negative values of A and large positive values of A. and varying the time of run from 10k to 1000k. I see that particles are clumping together. Maybe I did something wrong.
Approach 2 (Mr. Plimpton’s Approach)
Used an arbitrarily large box with particles placed in it randomly, initially no overlaps But as I shrink the box I get particles lost or the large/large overlap error.
Approach 3
Borrowing from Mr. Plimpton’s idea. I generate a set of co-ordinates assuming effective radii of particles as 5.1(original radius is 5.0) under the condition of no overlaps(so what I have is a 3-d lattice thing with spheres stacked at the intersection points), and write these co-ordinates to the lammps script using create_atom single command and these co-ordinates as arguments . This gives me 512 particles in a box size 81.6(volume fraction 0.493). The simulation run for the allotted duration.
But in my system the particles sort of just vibrate or hover in position(as seen through visualization tool). The mean square displacement is 0.7 after 100k timesteps. Is it because of the symmetrical nature of the system that the particles are being repelled equally in all directions? or are my parameters just bad?
Thanks in advance
Regards
Anurag