To gain insight into the experimental findings and to develop predictive capabilities, we also pursue multiscale modeling. Direct comparison with experiments is pursued to bridge the gap between experimental findings and computational predictions. We follow a multiscale (Quantum Mechanics / Molecular dynamics ) QM/MM/MD approach to understand fundamental behavior of nanoscale materials. QM modeling is done to capture the electronic structure and associated polarization (piezoelectricity) and/or bond breaking (fracture). These simulations also serve to validate semiempirical force fields, which are then employed for large-scale MM/MD simulations (size effect studies). This validation across different length scales is necessary, given that the size of the models which can be studied via QM is limited by the state-of-the-art computational power.
Computations are performed in an in house 8-node quad core cluster for performing large scale simulations. We also have access to High Performance Computing (HPC) systems available at Argonne National LabNL (Bluegene) and the Teragrid computing network.
We have applied this approach to investigate elasticity size effects in zinc oxide (ZnO) nanowires and to understand failure mechanisms under uniaxial tensile loading. By employing a Buckingham type pairwise potential, a size dependent elastic modulus was identified consistent with the experimental findings (Figure 2a). The simulations revealed that a reduced interatomic spacing of surface atoms, as a result of surface reconstruction, leads to a stiffer shell and a softer core (Figure 2b-c) when compared to bulk behavior. The increasing surface-to-volume ratio, as wire diameter decreases, combined with this stiffening effect leads to the observed size dependence.
Even though the elastic properties are well predicted by the semiempirical Buckingham potential, questions remain concerning its prediction of failure mechanisms. Experiments reveal brittle fracture. By contrast, a phase-transformation is predicted by the MD simulations. This casts doubt on the applicability of pairwise potentials in predicting failure. To address this discrepancy we are also involved in QM calculations using density functional theory.
Our current efforts, therefore, involve multiscale QM/MM/MD approaches to ensure that the employed semiempirical potentials are appropriate for predicting the properties of interest. We are also extending the QM calculations to calculate the electronic properties including piezoelectricity of nanowires as a function of wire diameter. This approach is being applied to nanowires of different materials, in close coordination with the experimental work being carried out by the group.
Recently, we have also performed computations on GaN nanowires revealing their mechanical properties. These results were coupled with experiments to give an unambiguous picture of the mechanical properties of these nanowires.