# Can I use two NVT for different group in one simulation?

Dear Prof. Axel,

Recently, I am working on some water confinement MD simulations using LAMMPS. About my model, there are mainly two parts, i.e. the first one is water, and the other is an unfixed nanotube. I have two questions:

1. If I impose NVT ensemble for the whole system (like the command below), although the temperature of the system is maintained at 300K very well, the temperature of water is slightly larger than the targeted 300K with a decreasing trend. And the energy (total, potential) is not a constant averagely, which have a slight decreasing trend (The slope is only about 5, the intercept is in the order of 10^7). I wonder if this system can be considered as equilibrium state. I discussed that the slightly decreasing trend of energy is caused by the calculation of electrostatic force (PPPM method) and it’s difficult to eliminate, is it right?

Fix 1 movepart nvt temp 300.0 300.0 100.0

1. I tried to impose NVT separately (like the command s below), the temperature can be maintained very well. However, I am confused that can I use these two NVT meanwhile?

fix 1 NT nvt temp 300.0 300.0 100.0

fix 2 water nvt temp 300.0 300.0 100.0

Hope to hear from you soon. Thank you very much.

Best Regards,

Timmy

Dear Prof. Axel,

Recently, I am working on some water confinement MD simulations using
LAMMPS. About my model, there are mainly two parts, i.e. the first one is
water, and the other is an unfixed nanotube. I have two questions:

1. If I impose NVT ensemble for the whole system (like the command below),
although the temperature of the system is maintained at 300K very well, the
temperature of water is slightly larger than the targeted 300K with a
decreasing trend. And the energy (total, potential) is not a constant
averagely, which have a slight decreasing trend (The slope is only about 5,
the intercept is in the order of 10^7). I wonder if this system can be
considered as equilibrium state. I discussed that the slightly decreasing
trend of energy is caused by the calculation of electrostatic force (PPPM
method) and it’s difficult to eliminate, is it right?

​what you are observing are indications on an ongoing equilibration. there
may be a drift in energy due to rearranging of the geometry originating
from lennard-jones interactions. also from pppm when the convergence
parameter is not very tight (many examples use 1.0e-4, which is lax. 1.0e-6
should be better. going to even tighter convergence usually does not help
as much, since other errors will dominate.

a simple way to speed up equipartitioning (i.e. all parts of the system
have the same temperature) is to use a dissipative themostat at the
beginning of the equilibration (e.g. langevin, temp/csld, gld, gle, ...)
and then switch to fix nvt. the nose-hoover thermostat (and also
temp/berendsen or temp/rescale) work globally and do not promote
equipartitioning and thus may require much longer equilibration times.​

Fix 1 movepart nvt temp 300.0 300.0 100.0

2. I tried to impose NVT separately (like the command s below), the
temperature can be maintained very well. However, I am confused that can I
use these two NVT meanwhile?

fix 1 NT nvt temp 300.0 300.0 100.0

fix 2 water nvt temp 300.0 300.0 100.0

​you can do that. but, same as mentioned above, it is usually best to do
this for a while until the temperature is well maintained at equal levels
through the entire system, but for production calculation, i would see, if
you can switch to a single fix nvt. if the system cannot maintain
equipartitioning then, there may be some other issue causing this and that
would have to be understood.

axel.​

Dear Prof. Axel,

I will try to equilibrate the system using langevin first.

About the question 2, I have a continuous confusion: For the system, if I use two NVT ensembles, what differences will cause compared with only a single NVT? Will this operation possibly lead to missing of the couple between these groups?

Highly appreciate the guidance.

Best

Timmy

Dear Prof. Axel,

I will try to equilibrate the system using langevin first.

About the question 2, I have a continuous confusion: For the system, if I
use two NVT ensembles, what differences will cause compared with only a
single NVT? Will this operation possibly lead to missing of the couple
between these groups?

​the two subsystems are still in contact and thus still couple. only that
each of them is also algorithmically and independently coupled (as the
whole group) to an external reservoir. using two thermostats like this is a
viable procedure to promote equipartitioning of kinetic energy during
equilibration. it can even by meaningful for production calculations in
some cases. however, for a system in equilibrium a single global
nose-hoover thermostat chain should be sufficient and will model the
usually desired NVT ensemble (assuming a dense periodic system).

axel.