Carbon naotube-silicon substrate simulation potentials

LAMMPS LAMMPS Mailing List Mirror
1

Hi,

I am trying to model interaction of cnt and pure Si (diamond). I have two silicon block and a sw-cnt in the middle along z direction, with periodic boundary in all directions.

I have two atom types, 1. Si, 2. C.

As I looked into previous forum posts, I came across several suggestions.

They are mainly:

Approach 1:

Use tersoff for all atom types.

pair_style tersoff

read_data cntsi_v1.lmp

pair_coeff * * SiC.tersoff Si C

I tried this and equilibrated the structure with NPT, NVT, 0.4 nanoseconds each. Firstly the nanotube becomes slightly misaligned with z axis, and stays there for the rest of the run. The temperature is nicely equilibrated at 300. But when I applied NVE for 0.4 nanoseconds, the temperature gradually dropped to 260 K at the end of run. Also Kinetic energy slowly drops.

Question: could this come from a bad input file? Or may be inappropriate relaxation time in NVT and NPT? I am using a Tdamp = 0.1 with metal units and Pdamp =1 and a time step of 0.001 (i.e. 1 fs)

Approach 2:

Use tersoff for Si-C and Si-Si interaction and airebo for cnt. This was discussed too, and the problem here was that since both airebo and tersoff are multibody potential , treating them like pair-wise additive potentials may not be a good idea. However, There are literatures using exactly this approach. (e.g. Zhang X. A Low-Frequency Wave Motion Mechanism Enables Efficient Energy Transport in Carbon Nanotubes at High Heat Fluxes. Nano letters. 2012-06;120618121952007 )

Question:

How could we do this in lammps (if at all)?

Approach 3:

Use tersoff or sw for silicon, airebo for cnt, lj/cut for Si-C interaction. The lj/cut pair coeffs can be obtained from Universal force field.

So, overall which approach is most realistic to model such a system?

Thanks,

Souvik.

2

Hi,

I am trying to model interaction of cnt and pure Si (diamond). I have two
silicon block and a sw-cnt in the middle along z direction, with periodic
boundary in all directions.

I have two atom types, 1. Si, 2. C.

As I looked into previous forum posts, I came across several suggestions.

They are mainly:

Approach 1:

Use tersoff for all atom types.

pair_style tersoff

read_data cntsi_v1.lmp

pair_coeff * * SiC.tersoff Si C

I tried this and equilibrated the structure with NPT, NVT, 0.4 nanoseconds
each. Firstly the nanotube becomes slightly misaligned with z axis, and
stays there for the rest of the run. The temperature is nicely equilibrated
at 300. But when I applied NVE for 0.4 nanoseconds, the temperature
gradually dropped to 260 K at the end of run. Also Kinetic energy slowly
drops.

Question: could this come from a bad input file? Or may be inappropriate
relaxation time in NVT and NPT? I am using a Tdamp = 0.1 with metal units
and Pdamp =1 and a time step of 0.001 (i.e. 1 fs)

hard to say at this abstract level. there are many different little things
that can "leech" kinetic energy and it is similarly difficult to say when
a system is equilibrated.

Approach 2:

Use tersoff for Si-C and Si-Si interaction and airebo for cnt. This was
discussed too, and the problem here was that since both airebo and tersoff
are multibody potential , treating them like pair-wise additive potentials
may not be a good idea. However, There are literatures using exactly this
approach. (e.g. Zhang X. A Low-Frequency Wave Motion Mechanism Enables
Efficient Energy Transport in Carbon Nanotubes at High Heat Fluxes. Nano
letters. 2012-06;120618121952007 )

not everything that gets published can be transferred
from one study to another.

Question:

How could we do this in lammps (if at all)?

just don't do it. it is a bad idea.

Approach 3:

Use tersoff or sw for silicon, airebo for cnt, lj/cut for Si-C interaction.
The lj/cut pair coeffs can be obtained from Universal force field.

So, overall which approach is most realistic to model such a system?

DFTB?

axel.

3

Use tersoff for Si-C and Si-Si interaction and airebo for cnt. This was
discussed too, and the problem here was that since both airebo and tersoff
are multibody potential , treating them like pair-wise additive potentials
may not be a good idea. However, There are literatures using exactly this
approach. (e.g. Zhang X. A Low-Frequency Wave Motion Mechanism Enables
Efficient Energy Transport in Carbon Nanotubes at High Heat Fluxes. Nano
letters. 2012-06;120618121952007 )

not everything that gets published can be transferred
from one study to another.

Question:

How could we do this in lammps (if at all)?

just don't do it. it is a bad idea.

This is possible with LAMMPS I believe. Aidan can comment further.
The way it could be done is to use a hybrid-overlay potential
using AIREBO for C only, and Tersoff for Si and C, but using
a Tersoff file that turns off the C-C interactions. I.e. I presume
you only want the Tersoff 3-body interaction if one or maybe
2 of the atoms is Si. I think this can be done by setting
the appropriate coeffs in SiC.tersoff to zero.

Steve

4

I don't know of any CNT studies using Tersoff. And it's probably not
because you are the first person to think of trying it.

The gradual cooling might be caused by disordering and stretching of the
CNT, with a net increase in potential energy, balanced by an increase in
entropy. Think of the last time you spilt gasoline on your hand.

You can certainly use hybrid overlay to run REBO (no need to use AIREBO,
unless you are doing multiwall CNT) for the CNT and Tersoff for the Si
block and the Si-C interactions. You must set the "C C C" A and B
parameters to zero, effectively turning off the Tersoff potential for C-C
and C-C-C interactions. However, it is important to recognize that
because both of these potentials are manybody, there is still a problem at
the interface (hence Axel's concern). Essentially, you are creating a new
potential for the interface, one that no one has ever used before, and one
that is difficult even to write down. So, you should test your new
potential on some simple cases e.g. relax a small Si-CNT-Si system to its
ground state and compare with DFT results.

You can also use hybrid overlay to create your Approach 3, an LJ
interface. This has the advantage that your new interface potential is
somewhat better defined. You can also examine the effect of increasing or
decreasing the LJ energy to increase and decrease the strength of the
interface.

Aidan