How to accelerate diffusion of C in Cu (COMB3)

Dear All!

I am currently trying to set up a simulation where (COMB3) Carbon diffuses through a (very thin) layer of copper in order to see if it allows the growth of graphene on the opposite face.

I set up the simulation by making a box, putting a few layers of copper in the center with pbc in xy and fixed in z (so the copper sheet extends into xy).

Then I fill the lower half of the box with carbon and make the bottom face into a fixed LJ96 wall in order to prevent carbon loss.

When I heat up the system the carbon slowly diffuses into the copper sheet but the process is painfully slow. I tried the following without success

  1. Start with higher concentration of carbon: I frequently get the warning of loss of atoms or atoms stuck in the wall during an update step

  2. Increase the number of carbon during the run using fix deposit but the command is not available since the system was set up without the MISC package and the admin doesn¡¯t want to include that package for now (I am just running this on a test system set up by my university)

  3. Increase the temperature more: the copper melts quite drastically (probably because it is such a thin sheet and COMB3 has been reported to have lower mp of Cu)

Can anyone suggest me a ¡°subtle¡± method to speed up the process at moderate temperatures whilst preventing atom loss at the walls?

I attached a snip of the setup below in case I forgot to mention something. The run is set up as npt

dimension 3

boundary p p f

timestep 0.0005

copper slab

lattice fcc 3.615

region box block -10 10 -20 20 -20 20

region carbon block -10 10 -20 20 -5 0

region sheet block -10 10 -20 20 1 3

create_box 2 box

create_atoms 1 region sheet

create_atoms 2 random 200 12345 carbon

mass 1 63.54

mass 2 12.011

pair_style comb3 polar_off

pair_coeff * * ffield.comb3 Cu C

fix wall1 all wall/lj93 zlo -20.0 1.0 1.0 2.5

fix addatoms all deposit 1000 2 100 29494 region carbon

velocity all create 100 53244

fix 1 all npt temp 100.0 100.0 0.1 x 0.0 0.0 1.0 y 0.0 0.0 1.0

run 1000

Dear All!

I am currently trying to set up a simulation where (COMB3) Carbon diffuses
through a (very thin) layer of copper in order to see if it allows the
growth of graphene on the opposite face.

I set up the simulation by making a box, putting a few layers of copper in
the center with pbc in xy and fixed in z (so the copper sheet extends into
xy).

Then I fill the lower half of the box with carbon and make the bottom face
into a fixed LJ96 wall in order to prevent carbon loss.

When I heat up the system the carbon slowly diffuses into the copper sheet
but the process is painfully slow. I tried the following without success

1) Start with higher concentration of carbon: I frequently get the
warning of loss of atoms or atoms stuck in the wall during an update step

2) Increase the number of carbon during the run using fix deposit but
the command is not available since the system was set up without the MISC
package and the admin doesn’t want to include that package for now (I am
just running this on a test system set up by my university)

3) Increase the temperature more: the copper melts quite drastically
(probably because it is such a thin sheet and COMB3 has been reported to
have lower mp of Cu)

Can anyone suggest me a “subtle” method to speed up the process at moderate
temperatures whilst preventing atom loss at the walls?

the only meaningful suggestion is: don't use molecular dynamics. the
process you want to model *is* slow and the steps you are looking at
have a low probability. there are some ways to accelerate the
occurance of individual "rare" events, but what you seem to be
expecting to see requires - in my personal opinion - the use of monte
carlo simulation methods.

axel.

I totally agree with Axel. If you use comb3 (or electronic structure method) to characterize the energy barriers for some of the steps involved in the diffusion of C through Cu, you can actually estimate the rate of this diffusion using Arrhenius or transition state theory. What you will discover is that this process may take in the order of microseconds, which is beyond the scale of classical MD.

You can use some of the accelerated MD methods to do this, but I agree that MC (such as kinetic Monte Carlo) is a better choice.

Ray

Dear Ray, dear Axel!

Thank you for your (as always) fast and helpful replies!

I will see if there is any chance to get a MC code running on the server I am allowed to use, I am just afraid the admin will not be very happy for these individual requests :wink:

Thank you for your help!

Yours,

Chris

2015-12-16 2:39 GMT+01:00 Christoph Wolf(신소재공학과) <[email protected]...>:

Dear All!

I am currently trying to set up a simulation where (COMB3) Carbon diffuses
through a (very thin) layer of copper in order to see if it allows the
growth of graphene on the opposite face.

I set up the simulation by making a box, putting a few layers of copper in
the center with pbc in xy and fixed in z (so the copper sheet extends into
xy).

Then I fill the lower half of the box with carbon and make the bottom face
into a fixed LJ96 wall in order to prevent carbon loss.

When I heat up the system the carbon slowly diffuses into the copper sheet
but the process is painfully slow. I tried the following without success

Carbon keeps diffusing into copper? Well, that's for sure not what you
want, with copper being one of the typical low C-solubility catalytic
metals... And it nicely mirrors my experience, in which I found that COMB3
Cu is basically a sponge that keeps absorbing C ad infinitum (or, at least,
way too much).

Kristof