# NVE/NVT system equilibration

Hi,

I was calculating Pt atom deposition onto a TiO2 substrate after the substrate reaches the equilibrium. The substrate was divided by three slabs along the z-axis. The bottom slab was fixed and the middle one was in an NVT ensemble. And the top one was in an NVE ensemble. For the NVT ensemble its temperature was set 300 K. However, what I found while the equilibration of this system is that the temperature kept increasing higher than 900 K not 300 K and it have kept increasing unless I had stopped the calculation,
Can anyone figure out what was going on? Is there someone who suffered from the same problem?

Dooseok

I assume you are adding energy to the system (possibly
thru the deposition) and not removing it rapidly enough
with your thermostat. Check the time constant.
Can you run NVE and conserve energy and run good
dynamics? If not,
make that right first, before you turn on NVT.

Steve

Hi Steve,

thank you for your kind reply. Your advice was very helpful! I would like to ask you some basic question about systems of two layers and each layer is defined as different ensembles, e.g. upper is NVE and the lower NVT. Of course, the NVE ensemble does not allow energy to get in and out, as long as I understood in my statistical mechanics class. But what I found is that if the energy of the upper layer (NVE ensemble) increases due to deposited atoms then the lower NVT ensemble’s temperature increases as well! Does this mean that the energy of the NVE ensemble transfers to the NVT ensemble? If so, the NVT ensemble plays a role in heat sink?

Dooseok

Hi Steve,

ask you some basic question about systems of two layers and each layer is
defined as different ensembles, e.g. upper is NVE and the lower NVT. Of

no. they are not. this is a common misconception.

course, the NVE ensemble does not allow energy to get in and out, as long as

that is incorrect. the nve integrator doesn't add or remove energy,
doesn't add or remove particles and doesn't change the volume.
however, if *something else* does one of these things, the energy
changes, or the number of particles or the volume.

I understood in my statistical mechanics class. But what I found is that if
the energy of the upper layer (NVE ensemble) increases due to deposited
atoms then the lower NVT ensemble's temperature increases as well! Does this
mean that the energy of the NVE ensemble transfers to the NVT ensemble? If

please don't use the term ensemble in this context, but
group. ensemble in the statistical mechanical sense is
the *entire* system and due to its setup, it is not
in any well defined statistical mechanical ensemble.

the purpose of using a layer with nve style time integration
on top of a layer of nvt style time integration on top of
a layer of non-moving atoms is to emulate the behavior
of a bulk system with a truncated setup. the nvt layer
has to mimic the energy with the (non-existing) bulk,
while the nve layer allows to correctly simulate the
motions of the surface layer and the deposited atoms.
that layer will exchange kinetic energy with the nvt
layer, but not through external intervention, but simply
through natural ballistic transport.

so, the NVT ensemble plays a role in heat sink?

kinetic energy will flow both ways. however, your bigger problem
is to determine, if your system *without* depositing atoms can
conserve energy. this is would be much more troublesome.

axel.

Hi Axel,

You are right. I found that the energy of this system without deposition is NOT conserved. I am very confused now. Could you please tell me the reason for this problem? I dont understand why nve thermostat leads to the increase of the energy…

Thanks
Dooseok

1. 오후 10:59에 “Axel Kohlmeyer” <[email protected]>님이 작성:

Hi Axel,

You are right. I found that the energy of this system without deposition is NOT conserved. I am very confused now. Could you please tell me the reason for this problem? I dont understand why nve thermostat leads to the increase of the energy…

Could be for a number of reasons like too large a time step, bad potential parameters, bad geometry, or a typo somewhere. Hard to say without knowing any details.

Axel