Hi,
I have got few questions on the fix deform which I am not very clear.
"What is the difference between rescaling velocities and rescaling x. "
Consider the following simulation in which I have a box of dimensions 10X10X10
For simplicity we consider only one dimension X, which is bound by XLO and XHI (let it be 0, and 10)
Let the atomic co-ordinates be {x0,v0}.
Now if I compress the box using deform by say 5% I understand that
XLO becomes XLO*.95
YLO becomes YLO*.95
Now for me it appears obvious that the co-ordinates of atoms also should be multiplied by the same factor thus
each of x= .95*x0.
If the position coordinates are not rescaled this does'nt happen. Hence some of the atoms in the above case tend to go outside the simulation box because the simulation box is now shrinked. Due to periodicitity these atoms are remapped backed into the simulation box. What I observe is a non uniform density with atoms clustering at the ends and possible overlap of some of the atoms.
I get similar result when I use "remap v" also because the positions are not rescaled. I am very confused at this.
If any one has done this simulations, please help me.
The second one I want to bring to the notice of Developers is
There are two options given for outputting the configuration while running a simulation. One is "write_restart' and the second option is 'dump'.
I tried to do a set of simulations one after the other with input file like this.
I have got few questions on the fix deform which I am not very clear.
"What is the difference between rescaling velocities and rescaling x. "
Now for me it appears obvious that the co-ordinates of atoms also should be multiplied by the same factor thus
each of x= .95*x0.
If the position coordinates are not rescaled this does'nt happen. Hence some of the atoms in the above case tend to go outside the simulation box because the simulation box is now shrinked. Due to periodicitity these atoms are remapped backed into the simulation box. What I observe is a non uniform density with atoms clustering at the ends and possible overlap of some of the atoms.
I get similar result when I use "remap v" also because the positions are not rescaled. I am very confused at this.
Use remap x if you want the atom coordinates scaled to the new box -
that's what it does.
The second one I want to bring to the notice of Developers is
There are two options given for outputting the configuration while running a simulation. One is "write_restart' and the second option is 'dump'.
I tried to do a set of simulations one after the other with input file like this.
***********************************************************************************
variable r index 1 2 3 4 ...
boundary p p p
Some LAMMPS commands
write_restart restart.xyz
clear
boundary s s p
read_restart restart.xyz
Some other
next r
jump in.file
*************************************************************************************
What I found is the restart file remembers lot of information which includes the boundary conditions (f, p or s).
What if I don't want to repeat the simulation with the same boundary condition.
Generally in the first simulation I will do a bulk system melting and quenching and in the second simulation I will cut out some atoms and remove periodicity so as to run a finite sample simulation. I have no way to wait for the result of first simulation, take the restart file and the convert manually from restart2data and run the simulation from dat file.
Had there been a command similar to restart which dumps only the coordinates and velocities in the format of dat.lammps which can be easily read by lammps, it would have been much easier.
The restart file does contain boundary and other info needed to
restart the same simulation. So you do need
a data file to restart a simulation with different parameters.
Restart2data is the tool to do that. Writing out a new
data file from a simulation is an interesting idea, but not easy b/c
of its format, especially if there are bonds, etc.
Steve