# Definition of V in fix NVT command

Hi,

I had a hard time understanding the fix NVT command recently. In the canonical ensemble(NVT), if the Volume is a constant, can we still use f, s, m boundary condition? For instance if I use “s” boundary conditions to all directions of a system, the volume of the system will change during the simulation, but if I apply NVT command to this system, it will eventually reach thermal equilibrium. I am confused about the relations between V in NVT and the system’s volume. If they are the same, then the system’s volume should never change. If not, what does the V in NVT stands for and which “V” is controlled as a constant during the simulation process? Any answer will be helpful. Thanks.

Jim

Hi,
I had a hard time understanding the fix NVT command recently. In the
canonical ensemble(NVT), if the Volume is a constant, can we still use f, s,
m boundary condition? For instance if I use "s" boundary conditions to all

you can use non-periodic boundary conditions with fix nvt,
but your resulting ensemble is no longer a canonical one.

directions of a system, the volume of the system will change during the
simulation, but if I apply NVT command to this system, it will eventually
reach thermal equilibrium. I am confused about the relations between V in
NVT and the system's volume. If they are the same, then the system's volume
should never change. If not, what does the V in NVT stands for and which "V"
is controlled as a constant during the simulation process? Any answer will

your logic is upside-down "constant volume" in NVT means
that the fix will not _change_ the volume. in periodic boundaries
you don't need to do anything to keep the volume constant.
it just so happens. the problems start if you want the volume
to relax so that a certain pressure is maintained (on average).

think about the "N" part, i.e. the number of particles, which
also won't change unless to add or remove some.

to clarify these issues further, i suggest to administer
yourself a healthy does of statistical mechanics from
a suitable text book. the lammps documentation mostly
tells you what it does, not so much how and why.

cheers,
axel.

1 Like

Hi Axel,

Thank you for your quick responding. I have a further question on this problem. I built a 3x3x30 cubic nm pure Cu crystal system on xyz axes. Potential is Cu_u6.eam. I need to apply periodic boundary conditions to xy directions and free boundary to z direction. Here comes the problem, if I use “p” to xy direction and “s” to z direction. Then after 250 ps NVT control, the temperature of the system will oscillate sinusoidally, and if you check the VMD image you will find the system is shrinking and stretching like a spring in the z direction. But if I apply “p” to the z direction but leave enough space at both ends, then the system is stable after 250 ps NVT. I am pretty sure that the boundary setting is right, so I began to think maybe there is something wrong with the “s” boundary during NVT process. Theoretically I should get the same results in both conditions right? Thank you for your time.

Best,
Jim

Hi Axel,
Thank you for your quick responding. I have a further question on this
problem. I built a 3x3x30 cubic nm pure Cu crystal system on xyz axes.
Potential is Cu_u6.eam. I need to apply periodic boundary conditions to xy
directions and free boundary to z direction. Here comes the problem, if I
use "p" to xy direction and "s" to z direction. Then after 250 ps NVT
control, the temperature of the system will oscillate sinusoidally, and if
you check the VMD image you will find the system is shrinking and stretching
like a spring in the z direction. But if I apply "p" to the z direction but
leave enough space at both ends, then the system is stable after 250 ps NVT.
I am pretty sure that the boundary setting is right, so I began to think
maybe there is something wrong with the "s" boundary during NVT process.

no. you may have just hit a resonance to which the nose-hoover
thermostat is coupling to. it may go away with a different choice of
relaxation time constant for the temperature.

Theoretically I should get the same results in both conditions right? Thank

depends on the amount of space you left between the periodic boundaries.

in general, using fix nvt is no a good way to easily achieve equipartitioning
of the kinetic energy, which is a prerequisite for equilibration. i would first
start the calculation with fix nve + langevin and only later switch to fix nvt,
if at all needed. if you do have an intrinsic resonance in your system, you
may also experiment with the drag parameter to fix nvt.

axel.

Hi，Axel,

I have already tried to use different drag values in the NVT process and even I set drag parameter to 2.0 the problem still exist. My intention is to make the system stable first. I will try fix nve + langevin and then only nve to see if the results get better. Thank you!

Best,
Jim