Steve, LAMMPS users,
I am trying to do a thermal simulation across a silicon/aluminum interface. In the silicon (atom type 1) layer, I plan to use Stillinger-Weber potential, while in aluminum (atom type 2) layer, I plan to use the EAM potential. I am wondering if (and how) I should specify the pair style and pair coeff for the interactions between atoms types 1 and 2 at the interface.
Thanks,
Sreekant
Steve, LAMMPS users,
I am trying to do a thermal simulation across a silicon/aluminum interface.
In the silicon (atom type 1) layer, I plan to use Stillinger-Weber
potential, while in aluminum (atom type 2) layer, I plan to use the EAM
potential. I am wondering if (and how) I should specify the pair style and
pair coeff for the interactions between atoms types 1 and 2 at the
interface.
well, you'll need to find a potential that can describe the
interaction of silicon with aluminum and use that, or use
a potential, that can handle as the same time silicon and
aluminium.
you should also be aware of the fact, that in general the
interactions across an interface can not necessarily be
well described with the setup that LAMMPS allows (pair style hybrid).
this implements only a mechanical coupling of the two components,
which totally disregards (long-range) electrostatic and other
many-body interactions.
i don't know so much about the kind of system that you
describe but similar problems occur in QM/MM coupling,
where quantum chemistry and classical force field calculations
are combined (e.g. to study chemical reactions in enzymes).
using only a mechanical coupling here would be introducing
a significant error, and the typical QM/MM implementations today
use an electrostatic coupling scheme. there are also other schemes
(ONIOM) to estimate the effect of the embedding.
when coupling a metal and a semiconductor, there may also be
effects on on the electronic structure at the interface, which may
impact the reconstruction of both components near the interface.
again, i have little idea about this kind of system, but with common
elements like silicon and aluminium, i would expect that there should
be calculations using some (approximate?) quantum mechanics
method (DFT, or more semi-empirical methods).
HTH,
axel.
Hi Sreekant,
These two references may be of some help to you.
Noreyan, A., Qi, Y., and Stoilov, V., 2008, “Critical shear stresses at aluminum-silicon interfaces”, Acta Materialia, 56, pp. 3461-3469.
Gall K, Hostemeyer, MF, Van Schilfgaarde M, Baskes MI. Atomistic simulations on the tensile debonding of an aluminum-silicon interface. Journal of the Mechanics and Physics of Solids 2000; 48:2183-2212.
Best regards,
CD
Axel Kohlmeyer wrote:
you should also be aware of the fact, that in general the
interactions across an interface can not necessarily be
well described with the setup that LAMMPS allows (pair style hybrid).
this implements only a mechanical coupling of the two components,
which totally disregards (long-range) electrostatic and other
many-body interactions.
Not clear what Axel means by this. But if you specify the
cross interaction A-B as lj/cut/coul/long (for example), and
also define a kspace solver (e.g. PPPM) then you will
get long-range electrostatics across the interface. And
everywhere else in the model (A-A, B-B), which is presumably
what you want.
Steve
> you should also be aware of the fact, that in general the
> interactions across an interface can not necessarily be
> well described with the setup that LAMMPS allows (pair style
hybrid).
> this implements only a mechanical coupling of the two components,
> which totally disregards (long-range) electrostatic and other
> many-body interactions.Not clear what Axel means by this. But if you specify the
cross interaction A-B as lj/cut/coul/long (for example), and
also define a kspace solver (e.g. PPPM) then you will
get long-range electrostatics across the interface. And
yes. but you have to use an */coul/long pair style throughout
for _all_ interactions. if you have coul/long contributions only
in the A-B interaction, but with the kspace contribution to
consider all charges equally.
the point i wanted to make is that using many-body interactions
(which include long rage electrostatics) with pair_style hybrid
requires extra care in validating the selected model compared
to pairwise additive potentials.
everywhere else in the model (A-A, B-B), which is presumably
what you want.
the original question was about combining EAM and SW
with pair_style hybrid, btw.
cheers,
axel.
If what you want is:
EAM for A-A interactions
SW for B-B
assign charges to atoms and compute coul/long and PPPM for A-B
then I agree with Axel that this is a flawed model. LAMMPS
would let you do it, but that doesn't mean it makes sense.
You would be ignoring charge-charge interactions for A-A. In
addition EAM and SW are uncharged models, and they
account for charge effects implicitly in the interactions
they compute. So adding charges on top of EAM/SW
does not seem correct.
Steve
LJ