Please include the lammps mail list, thanks.
Instead of using fix setforce and fix nve on group boundary, which is incorrect, I would just use fix rigid on group lower and fix move on group upper.
However, velocity ramp and fix nvt on group mobile does not seem to make sense. Also all these changes may still not work if the strain rate is not reduced.
Ray
The order in which fixes are defined can also affect your simulation. In addition to what Ray suggests, i would like to add/ask-
What fix is applied to control the temperature of boundary?( like nvt is done to control mobile region temperature). In other words, isnt your temperature increasing after a few steps?
Try to apply velocity in BOX units, not LATTICE units.
Sagar
Dear Ray
I have some question too:
1- I need loading the box in z direction and I should use P P m boundary condition. I don’t know that is it OK to use P P P boundary in relaxation mode and after that change the boundary to P P m for loading step?
2- is it OK to use the fix move for both of upper and lower group and move them in opposite direction. which one is better ( move both side or fix rigid one side and move other side)?
3- the documentation said that " IMPORTANT NOTE: You should not update the atoms in rigid bodies via other time-integration fixes (e.g. fix nve, fix nvt, fix npt)". I want to know if don’t need to control temperature in boundary groups?
Thanks
Hi Mohammad,
1- I need loading the box in z direction and I should use P P m boundary condition. I don't know that is it OK to use P P P boundary in relaxation mode and after that change the boundary to P P m for loading step?
These all depend on what commands you use for loading, e.g. change_box, velocity set, region move or fix move. If fix move, make sure the group that fix move is applied to (the rigid and moving group to perform the loading) does not move across the p boundary during equilibration.
2- is it OK to use the fix move for both of upper and lower group and move them in opposite direction. which one is better ( move both side or fix rigid one side and move other side)?
They are essentially the same if the strain rates are carefully chosen to be the same.
3- the documentation said that " IMPORTANT NOTE: You should not update the atoms in rigid bodies via other time-integration fixes (e.g. fix nve, fix nvt, fix npt)". I want to know if don't need to control temperature in boundary groups?
It depends on whether you want your moving group to be rigid or not.
Ray
Dear Ray
Thanks for reply.
In my simulation I want to interact tow crystal (Fe2o3/Fe) in z direction and after relaxation and bounding
the tow crystals at interface, I want to do the tension loading by “set boundary velocity” and " velocity ramp"
in tow sides in opposite direction.
In first step I just to simulate the tension loading for one crystal (Fe2O3).
I couldn’t relax the crystal in P P S boundary, but in P P P the relaxation result is better and in
“P P S” boundary with “setfors boundary 0 0 0” the relaxation result are good. but I think it is not a
regular method to use the “setforce boundary 0 0 0” during relaxation mode. please advise me?
in final, when I want to interact the tow crystals in z direction, the results is very bad and I couldn’t equlibrate the system
and the simulation diverged.
I want to know what is the regular method to equilibrate a system with interact tow crystal to each other. for instance
I think , I should equilibrate the crystals separately and then close up the equilibrated state of tow crystals to each other (in a single data file)
and then relax them again to interact in interface. please guide me.
Thanks
Hi Hadi,
In my simulation I want to interact tow crystal (Fe2o3/Fe) in z direction and after relaxation and bounding
the tow crystals at interface, I want to do the tension loading by "set boundary velocity" and " velocity ramp"
in tow sides in opposite direction.
In my opinion, velocity ramp should not be applied to the middle active region. The atoms that are not the boundary atoms for applying the tension/compression load to (or those that are active or mobile) should just respond freely and automatically to the load applied to the boundary groups.
In first step I just to simulate the tension loading for one crystal (Fe2O3).
I couldn't relax the crystal in P P S boundary, but in P P P the relaxation result is better and in
"P P S" boundary with "setfors boundary 0 0 0" the relaxation result are good. but I think it is not a
regular method to use the "setforce boundary 0 0 0" during relaxation mode. please advise me?
What errors were present with p p s boundary? How was "better" defined with the p p p boundary? Why is it not regular to fix some of the atoms (especially those kept for applying load to) immobile? If you don't care about the boundary atoms and care only about how the active and free atoms respond to the applied load, it does not matter if the boundary atoms are relaxed or not.
in final, when I want to interact the tow crystals in z direction, the results is very bad and I couldn't equlibrate the system
and the simulation diverged.
This can be due to a lot of things:
1. bad structure, especially overlapping atoms at the interface,
2. bad potential model for the particular interfacial system, and
3. bad setup, e.g. misuse of commands.
I want to know what is the regular method to equilibrate a system with interact tow crystal to each other. for instance
I think , I should equilibrate the crystals separately and then close up the equilibrated state of tow crystals to each other (in a single data file)
and then relax them again to interact in interface. please guide me.
Yes, this is usually considered a good practice.
Cheers,
Ray
Hi Hadi,
> In my simulation I want to interact tow crystal (Fe2o3/Fe) in z
direction and after relaxation and bounding
> the tow crystals at interface, I want to do the tension loading by "set
boundary velocity" and " velocity ramp"
> in tow sides in opposite direction.In my opinion, velocity ramp should not be applied to the middle active
region. The atoms that are not the boundary atoms for applying the
tension/compression load to (or those that are active or mobile) should
just respond freely and automatically to the load applied to the boundary
groups.Dear Ray
Thanks for your reply.
Absolutely you are correct to don't use velocity ramp. but because of use
high strain rate in tension loading
(to reduce the simulation time) I used velocity ramp to avoid shock wave
during simulation.
my reference to use this method is the following paragraph of "Molecular
dynamics simulation based cohesive surface representation
of mixed mode fracture, X.W. Zhou , J.A. Zimmerman, E.D. Reedy Jr., N.R.
Moody"
"During molecular dynamics simulations of mode I crack
propagation, the system is uniformly stretched (by moving
each atom a distance corresponding to a uniform normal
strain increment) in the y direction each time step. Numerical
approaches are then used to update the atom positions
based upon interatomic potential and Newton’s equations
of motion under the condition that the y coordinates of the
top and bottom horizontal layers of boundary atoms
remain fixed. Unlike displacing boundary atoms alone,
a uniform stretching of the system
avoids the creation of shock wave during simulations"
In Lammps Examples and tutorial this method used too. please let me know if
I
am in wrong way
> In first step I just to simulate the tension loading for one crystal
(Fe2O3).
> I couldn't relax the crystal in P P S boundary, but in P P P the
relaxation result is better and in
> "P P S" boundary with "setfors boundary 0 0 0" the relaxation result are
good. but I think it is not a
> regular method to use the "setforce boundary 0 0 0" during relaxation
mode. please advise me?What errors were present with p p s boundary? How was "better" defined
with the p p p boundary? Why is it not regular to fix some of the atoms
(especially those kept for applying load to) immobile? If you don't care
about the boundary atoms and care only about how the active and free atoms
respond to the applied load, it does not matter if the boundary atoms are
relaxed or not.when I use "P P S" to relax the Crystal after a few time step the
structure warped immediately and convert to a
amorphous structure and reduce the volume in z direction. I used the
following command. when I used P P P
the relaxation continued till 60 pico second and then diverged. I prefer to
relax P P S with setforce 0 but
because of finally I should interact one side to another crystal then its
correct to setforce just one side and
other side must relaxed freely. ( I am not sure and need help)
units metal
dimension 3
atom_style atomic
boundary p p s
read_data atom.data
timestep 0.0002
pair_style meam
pair_coeff * * library.meam Fe3 O2 FeO.meam Fe3 O2
replicate 4 4 4
compute enperatom all pe/atom
compute pe all reduce ave c_enperatom
compute msdd all msd
thermo 10
thermo_style custom step vol pe c_pe temp c_msdd[1] c_msdd[2] c_msdd[3]
c_msdd[4] press cpu
min_style cg
minimize 1.0e-10 1.0e-10 1000 10000
velocity all create 300.0 234 dist gaussian
fix 1 all nvt temp 300 300 0.05
run 500000
I am not sure if this is relevant to this discussion but might be useful. One way to get rid of the high strain rate is to use quasi static loading. You can apply a force or a strain and let the structure minimize. This should give you the equilibrium structures at various stress/strain values.