Hello,
I Have a system made of multiple chains where each chain is represented by a bead and a spring model. The chains are arranged randomly and stacked on top of each other to ensure each chain don’t intersect the other to form a network of fibers interacting Via a LJ potential. Currently I am trying to conduct a stress Vs strain test. I deform the network by applying a very small displacement to the the atoms in my network based on initial x coordinate i.e:- 0.001*x. Then I hold my right most atoms and my most left atoms between fixed grips then I minimize the potential energy for the atoms in the middle then i run MD for the same atoms at very small temperature using NVT for around 1000 steps. I choose my Tstart to be the temperate from thermo output during minimization around 0.91K and my Tfinal to be 1K. My timestep is 0.001ps and I use Tdamp to be 0.1 as suggested by the manual however the temperature fluctuate vigorously to temperatures up to 150K and it never reaches 1k it reaches around 22k. I tried using drag command with NVT but there is no significant change. My question how is how to keep temp around 1k. Is there a way to solve this issue
Hello,
I Have a system made of multiple chains where each chain is represented by a
bead and a spring model. The chains are arranged randomly and stacked on top
of each other to ensure each chain don't intersect the other to form a
network of fibers interacting Via a LJ potential. Currently I am trying to
conduct a stress Vs strain test. I deform the network by applying a very
small displacement to the the atoms in my network based on initial x
coordinate i.e:- 0.001*x. Then I hold my right most atoms and my most left
atoms between fixed grips then I minimize the potential energy for the
atoms in the middle then i run MD for the same atoms at very small
temperature using NVT for around 1000 steps. I choose my Tstart to be the
temperate from thermo output during minimization around 0.91K and my Tfinal
to be 1K. My timestep is 0.001ps and I use Tdamp to be 0.1 as suggested by
the manual however the temperature fluctuate vigorously to temperatures up
to 150K and it never reaches 1k it reaches around 22k. I tried using drag
command with NVT but there is no significant change. My question how is how
to keep temp around 1k. Is there a way to solve this issue
this is all too vague, or (needlessly) convoluted and confusing. it is
near impossible to give any good advice on such a mess of a
description. specifically since very important information like the
total number of atoms in your system is missing. finally, you mention
extremely short simulation times on a system that - from its
description - appears to be very difficult to properly equilibrate.
to get help you need to a) break down your problem into something much
simpler and ask questions only for each part of setting up a
simulation that is adequate to yield dependable results. and need to
b) provide complete example/test inputs demonstrating your
observations. then people can actually provide you with specific and
meaningful advice.
it also looks to me, that you need to get a better understanding of
the MD methodology itself. you are asking about dealing with symptoms
like the observed temperature, when those are often an indication of
much more fundamental problems. suppressing the symptom does not solve
your problem.
axel.
Hi Axel,
Sorry for the poor Description. My system is made of 160000 atoms.
My input file is the following
units metal
boundary p p p
dimension 3
atom_style molecular
read_data /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/results22.txt
bond_style harmonic
bond_coeff 1 135 3.3875
angle_style harmonic
angle_coeff 1 3299 180
pair_style table linear 1836025
pair_coeff * * /lore/sengaa/LAMMPSNEW/lammps-16Feb16/potentials/Lj_cal.table Lj_cal 40.65
#pair_write 1 3 1836025 rsq 3e-2 40.65 /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/potential1.txt Lj_cal
velocity all create 1 9303
dump 1 all custom 1000 /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/network3.lammpstrj mass type id x y z
variable 1 equal “epair”
variable 2 equal “eangle”
variable 3 equal “ebond”
variable 4 equal “pxx”
variable 5 equal “x[124914]”
fix 11 all print 100 “{1}" file /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/epair121.txt screen no
fix 12 all print 100 "{2}” file /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/eangle121.txt screen no
fix 13 all print 100 “{3}" file /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/ebond121.txt screen no
fix 14 all print 100 "{4}” file /lore/sengaa/LAMMPSNEW/lammps-16Feb16/sengab/network/pxx121.txt screen no
thermo 1
thermo_style custom step time temp pe eangle epair ebond pxx v_5
minimiztation
min_style cg
minimize 1e-15 1e-15 5000 10000
Running MD
fix 1 all nvt temp 1 1 0.008
timestep 0.005
run 35000
variable x atom 0.001*x
looping over to conduct stress VS strain test
label loopa
variable a loop 264
displacing atoms in the x as a function of the current configuration
displace_atoms four move v_x 0 0 units box
holding the left and right edges between grips
fix 2 five setforce 0 0 0
minimization of the atoms between two grips
min_style cg
minimize 1e-15 1e-15 5000 10000
variable t equal “temp”
run 0
equilibration. The problem is here the temperature fluctuates vigorously in 1 loop
fix 3 three nvt temp ${t} 1 0.1 drag 0.5
timestep 0.005
run 1000
unfix 2
unfix 3
next a
jump SELF loopa
Sorry for asking the wrong question so my question would what causes the temperature to fluctate vigorously?
Regards
Hi Axel,
Sorry for the poor Description. My system is made of 160000 atoms.
My input file is the following
[...]
displace_atoms four move v_x 0 0 units box
# holding the left and right edges between grips
fix 2 five setforce 0 0 0
# minimization of the atoms between two grips
min_style cg
minimize 1e-15 1e-15 5000 10000
variable t equal "temp"
run 0
# equilibration. The problem is here the temperature fluctuates vigorously
in 1 loop
fix 3 three nvt temp ${t} 1 0.1 drag 0.5
timestep 0.005
run 1000
unfix 2
unfix 3
next a
jump SELF loopa
Sorry for asking the wrong question so my question would what causes the
temperature to fluctate vigorously?
impossible to say, even with a more complete input. there are still
lots of things messing, but several unusual things stand out:
- you have a *huge* cutoff and it is not clear what "size" your "atoms" are
- you are using a tabulated potential, so it is impossible to make any
assumptions about that. it could be completely bogus. you say, your
system interacts via LJ, why not use lj/cut?
- your force constants for bonds and angles seem high given your choice of units
- using a harmonic angle potential with an 180 degree equilibrium is
an odd choice. your restoring force is degenerate and thus cannot be
computed at the equilibrium angle.
- your initial "equlibration" at 1K over only 35000 time steps doesn't
deserve that name. nothing can happen in such a brief time and at such
a low temperature and then the 1000 time steps "equilibration" are
even more pointless
- i don't understand the purpose of feeding an instantaneous
temperature without any kind of averaging as target temperature to a
following run
- you were reporting significantly high instantaneous temperatures,
which suggest an initial configuration with very high potential energy
and an insufficient equilibration, relaxation or generally an
unphysical starting geometry.
- you mention a random initial structure, but not any for of
"unoverlapping" process, which would be obviously needed for such an
entangled starting structure.
this all looks to me, as if you moved ahead far too far in building
your input without even having figured out the basic process to go
from your initial random (and possibly highly unphysical)
configuration to something more meaningful. people have been
mentioning the GI-GO (= garbage in, garbage out) principle of MD
simulations on this list regularly, and there are still far too many
unknowns in what you are presenting to make any dependable assessment
of the validity of your overall model. it does look rather
questionable.
my suggestion is to step back and validate your model as such with a
much smaller and simpler setup and determine whether your table are
working correctly and can reproduce published data. same for your
bond/angle potentials. this can always be done in stages. as mentioned
previously if you don't understand why something doesn't work as
expected, make your test system simpler, until it is so simple that it
does work or you understand what is going wrong, and then work your
way backwards, step-by-step and check the validity of each step.
axel.
Thank you very much for your help