[lammps-users] Question on fix command

Dear all,

I want to find the equilibrium state of a bulk metal (such as Cu and Al)
system under constant temperature condition and pressure control. It
seems very difficult to find such a state because large pressue flucuations
(-100000 ~ 100000 bars) and temperature fluctuations (150~450K) exist.
Did anyone encount this problem and how to get the desired results?

The input lists are


To decrease T and P fluctuations, you could simulate a
bigger system, i.e.:

region box block 0 100 0 100 0 100

instead of:

region box block 0 10 0 10 0 10

Also, your timestep seems very small. I'd guess you
should be able to use "timestep 1.0" when in
production, although you may have to start out smaller
than that.


--- YONGGANG ZHENG <[email protected]>

NPT (Nose/Hoover) can have oscillatory behavior with solids (e.g.
a metal) - there is a "drag" option on the NPT fix that will damp
these oscillations ... see the doc page


I increased the system size to be a biger one (region box block 0 20 0
20 0 20) and changed the timestep to be 1.0, 0.1 or 0.01. It was found
that if the timestep is >= 0.1, errors will be encountered (e.g. for
dt=1, NANs appear, for dt=0.1, the program collapses).
In addition, if I set the initial temperature to be twice of the desired
value (300K in my simulation), temperature of the system approaches 300K very
quickly with small fluctuations). My question is: what is the optimal
initial temperature we should set for a simulation to make the system
reaches a desired value quickly?
I will try with "drag" option next. Thanks.


I just noticed you are running in metal units, where time is in units of
picosecs, so you really need a timestep of 0.001 to 0.003, depending
on the stiffness of your potential.

If you are starting with a perfect lattice, then potenial energy is a minimun,
so you do need to set initial T at about 2x what you want ... then it will
equilibrate to roughly 1/2 of the eng in KE and 1/2 in PE. But if you
are running NVT or NPT this will happen automatically.