# problem with ramp increasing of temperature

Dear Lammps users and Developers
I have a problem and I don’t understand why this happend.
I tried to equilibrate the fe2o3 crystal, by:
1- minimize the force and energy.
2-use “fix nve” for a few steps
3- use “fix nvt” to ramp the temperature to 300 K.
4-use “fix nvt” in 300 K. (following commands).
problem: I found that after minimization the potential energy dont change by increasing in temprature. I checked the “cfg” file of crystal by atoeye and found that all the atoms moved whith each other instead of each atom oscillate in their places.
after that I changed the ramp increasing of temprature and used “velocity all create 300.0 234 dist gaussian” instead, and then the atoms oscillated at their places. also when I dont persist to minimize force and use “minimize 1.0e-10 1.0e-10 100000 1000000” command the problem have been solved too.
1- why this happend when I increase the temp with ramping method?
2- is it necessary to minimise the force (& press) in minimisation and is there any rule for it?

Thanks

boundary p p p
timestep 0.0002
fix relaxbox all box/relax aniso 0
min_style cg
minimize 0 1.0e-10 100000 1000000
unfix relaxbox

fix 1 all nve
run 10000
unfix 1

variable tt equal temp
variable ttt equal \${tt}

fix 1 all nvt temp \${ttt} 300 0.1
run 50000
unfix 1

fix 1 all nvt temp 300 300 0.1
run 500000

If your initial velocity is 0.0 (which it appears to be
in the input script), then ramping a thermostat is

Try giving the atoms a finite velocity/temp with
the temperature command, then use fix nvt.

Steve

Dear Steve

as I said in previuse Emaile , I persist to minimize the force and energy together in first step by

use :
fix relaxbox all box/relax aniso 0
min_style cg
minimize 0 1.0e-10 100000 1000000

as the force minimization take more time when the simulation box is big,

I want to know is it necesery to minimize the fmax

in minimization for relaxation ?

Dear Steve
as I said in previuse Emaile , I persist to minimize the force and energy
together in first step by
use :

fix relaxbox all box/relax aniso 0
min_style cg
minimize 0 1.0e-10 100000 1000000

as the force minimization take more time when the simulation box is big,
I want to know is it necesery to minimize the fmax
in minimization for relaxation ?

there is no single answer for such a question. it always depends on what
the overall purpose is.

your original post said that you want to equilibrate a crystal at finite
temperature. why do you think you need to do a minimization for that in the
first place?

a "full" minimization is conceptually the same as an "equilibration to 0K"
so it would actually contradict your purpose of equilibration to a finite
temperature.

axel.

Dear Axel
Thanks for your reply. as you know I want to equilibrate the system at finite temprature (300K). for this purpose I do ( from study the most mailing list discription about equilibration) the following steps:
1- minimization + fix box/relax
2- velocity create 300
3- fix nve
4- fix nvt 300k
but my simulation till now don’t pass the nve step good. when I use minimization the system behave smoothly and continued even for 500000 steps (dt=0.0002 PicoSec), but without the minimization the temprature raises after a few steps ( nve) and lost atoms. also in most input scripts and mailing lists I found using the minimization at first step.I think my simulation time step is enough small for using MEAM potential (I found in some articles the dt=0.001 PicoSec) but I couldnt equilibrate the system. please advise me.
Thanks

Why don’t you show your full input script so that those interested in helping you can see your logic in full? The lines you pasted in the original email have no pair potential defined anywhere.

Carlos

Dear Axel
Thanks for your reply. as you know I want to equilibrate the system at
finite temprature (300K). for this purpose I do ( from study the most
mailing list discription about equilibration) the following steps:
1- minimization + fix box/relax
2- velocity create 300
3- fix nve
4- fix nvt 300k
but my simulation till now don't pass the nve step good. when I use
minimization the system behave smoothly and continued even for 500000 steps
(dt=0.0002 PicoSec), but without the minimization the temprature raises
after a few steps ( nve) and lost atoms. also in most input scripts and
mailing lists I found using the minimization at first step.I think my
simulation time step is enough small for using MEAM potential (I found in
some articles the dt=0.001 PicoSec) but I couldnt equilibrate the system.

"because somebody else did this" is not a good justification for making
choices. e.g. the choice of timestep depends on the steepness of the
individual potential and the mass of the particles (think about harmonic
oscillators and the relation between force constant and particle mass as
the simplest possible model).

to make the right choices, you'd rather have to think about what is causing
a problem and what do you want to achieve.
if your original geometry is already an (ideal) crystal, there is little
point in running a minimization, it will only consume time and take you
farther away from the 300K equilibrium. however, you have to make *certain*
that your crystal structure is correct and correctly continued across
periodic boundaries in all directions. this can be easily verified using
visualization and careful inspection of the geometry data. if you lose
atoms, that means atoms move too fast (locally), so you should investigate
*where* this is happening and *why*. if you work around this somehow with a
minimization, you don't solve the problem; you only postpone it. a more
direct approach would be to use fix nve/limit that would reduce the motion
of atoms that move too fast. however, in the case of a reasonable initial
crystal, this *should* not happen and that is why you need to investigate
carefully and not just blindly follow what somebody else does.

the next point to consider is that for equilibration, you need as a
prerequisite a reasonable equipartitioning of kinetic energy. this means
making atoms that move too fast slower and atoms that move too little
faster. fix nve cannot do this and fix nvt only acts on the total system.
better is to use fix nve+langevin for example. but even initializing the
velocity to a rather low temperature and then using fix nvt with a
temperature ramp will likely work.
since you already noticed the "flying icecube syndrome", you have to make
sure that the total momentum of your system is zero at the beginning and
you don't introduce any interactions that will change that (which is one of
the problems with using fix nve/limit).

from here on you need to think about what is a good way to *determine* when
your system has reached equilibrium (basic rule: it always takes longer
than what one expects). but that is something to worry about *after* you