Graphene Tensile Test - Critical strain is too high

Hello LAMMPS users, I have been performing several tensile test simulations on graphene and all of the critical strains that they produce are higher than what is published in the literature. I have attached one of the input files that I have been using below. I tried reducing the strain rate and using the NVE ensemble with the Berendsen thermostat (shown commented out), but I still obtain higher critical strains than what is published. Any suggestions would be greatly appreciated.

Andy Fox

Hi,

Are those references from MD simulations or experimental data ? .
Usually in experimental data the strain rate of deformation is of the
orden of microseconds. Remember MD simulation are done at high strain
rates (>10^9 1/s) . I'm not sure this may be the case, but my
experience in MD simulation, tells me that materials behave different
at hight strain rate than lower strain rate. Usually all the material
science models predict plasti-elastic transition are based on a
critical stress (defects appears when a maxium shear move the atoms
from it position) . However that may not be the case at the time
scales of MD simulations. those simulation predict defect nucleation
occours at higher strain rates for the maxium shear (CRSS).

Or maybe i'm complety wrong , and its a problem with your input script ....

cheers
Oscar G

Hi Andy,

Glad to hear that because some of the results in that “Science” paper was wrong. We have recently published a paper trying to correct the misleading results. The paper is attached. If any questions, please let me know.

with best regards,

AC

APL_100_21_211912_1.pdf (1.86 MB)

Hi Andy,

Glad to hear that because some of the results in that “Science” paper was wrong. We have recently published a paper trying to correct the misleading results. The paper is attached. If any questions, please let me know. (looks like the attachment is bigger than 500 KB, I’m attaching the URL as follows.)

Citation: Appl. Phys. Lett. 100, 211912 (2012); doi: 10.1063/1.4722786

View online: http://dx.doi.org/10.1063/1.4722786

I looked at your paper “Atomistic study on the strength of symmetric tilt grain boundaries in graphene” and your critical strain (~31%) for the 13.2 degree zigzag sheet is closer to the critical strain (~28-33% depending on other parameters I modify) that I am now getting for the same sheet using the AIREBO potential (also using NVE integration and a Berendsen thermostat at 300K) as opposed to the critical strain (~13%) indicated in the “Anomalous Strength Characteristics of Tilt Grain Boundaries in Graphene” paper I mentioned before.

In your paper you mention that you use the REBO potential. Why are you using this instead of AIREBO? It is my understanding that to use the REBO potential you use the following line:

pair_style airebo 2.0 0 0

and for the AIREBO potential you use the following line:

pair_style airebo 2.0 1 1

Do you know if this is correct?

Also, how did you set the cutoff radius between 1.92 and 2.0? I did this by changing both parameters in the CH.airebo file (rcmax_CC and rcmaxp_CC).

Best,

Andy

Andy,

It is nice to hear that your results are consistent with mine. As far as I know, the motivation for developing AIREBO potential is to include C-H interaction in REBO. It does not matter to include LJ potential for C-C. I have tried simulations with both REBO and AIREBO and the results turned out to be pretty close. Plus, because of long range LJ potential, AIREBO is much time-consuming. These are the reasons I did not use AREBO.

For REBO, I just used pair_style airebo 1.0 0 0

for AIREBO, I used pair_style airebo 3.0 1 1 (3.0 is from literature)

What I meant changing cutoff is to change the value of rcmax_CC in CH.airebo file.

Best,

AC

Hi Albert and Andy,

REBO potential (1st gen, Brenner 1990) was a pure short range potential. However it was found the repulsive wall was too soft so rebo is often splined with a 12-6 LJ potential. This yields a double well potential; one well at equilibrium, the other at much shorter distances.

AIREBO potential (Stuart, 2000) was developed to adaptively turning on or off the LJ interaction based on several criteria and to avoid the double well. At equilibrium and tensions, this LJ is often off so that REBO and AIREBO should perform similarly at these conditions. However at compressions, this LJ repulsive wall is quite important and REBO and AIREBO will give very different results. For example, REBO predicts graphite-to-diamond transition at compressions far too easily compared to AIREBO.

It matters little to include LJ for graphite/graphene tensions, but much greater for compressions.

Best,
Ray

RAY,

Thanks for pointing this out. Another thing for the LJ interaction of C-C is to model graphene/multi-wall carbon nanotube interlayer interaction.

Cheers

AC