Molecular Dynamics Simulation Of Mechanical Behavior Of Nanofilms

Nanofilms have been extensively used as structural as well as electrical components of NEMS devices, especially ultrasensitive sensors for ultrafine resolution applications. With decreased size, a nanofilm possesses different mechanical properties compared with its macroscopic counterparts. Using molecular dynamics method combined with classic atomic potentials, the mechanical properties of nanofilms, including surface stresses and size-dependent elastic properties, were studied in this thesis.The fundamentals, especially the "philosophical" aspects of molecular dynamics method were elucidated in chapter 2. Molecular dynamics method was introduced in the light of statistical physics, in that the orbital average (orbital sampling in practice) in MD simulations can serve as an alternative to the ensemble average in statistical physics according to ergodic hypothesis. By integrating out the incalculable degrees of freedom of external system to one or several fictional degrees of freedom, one can treat the physical system in question plus the external system as a whole isolated system, called extended system, of which the motion equations were easily obtained. Beside the motion equations of molecular dynamics method, atomic interactions are also of fundamental importance. The principles of constructing and selecting classic atomic potentials were discussed. The derivation of embedded-atom method potentials was presented as a case in point.In chapter 3, a systematic comparison between two different atomic stress definitions, i.e., the decomposed virial formula and Tsai formula, shows that they are mathematically equivalent in calculating the overall average stress of an atomic system. But in the case of calculating local stress distribution, the former gives ambiguous results, e.g. it gives nonzero normal stress at free surfaces ant it typically "underestimates" the inhomogeneity of microstructures and deformations in material. With a highly degenerate atomic chain model, it was shown mathematically that the results obtained by the decomposed virial formula are accurate only if the deformation is homogenous within the neighborhood of an interaction-cutoff radius, centered at the atomic site considered. Thus it is worth noting that Tsai formula is more adequate for calculating both the overall average stress and local stress distribution.In chapter 4, molecular dynamics simulations were carried out to calculate the distributions of energy and bulk stresses in an EAM Cu nanofilm, and to examine the effect of Cu(001) surface relaxation on the distribution patterns of energy and bulk stresses in the surface region. The results show that a release of energy and bulk stresses in the surface region occurs upon surface relaxation. In addition to an oscillatory release pattern of the energy in the top layers, we also found a rather 'anomalous' release pattern of the in-plane bulk stresses in the surface region, in which the decrease of the local in-plane bulk stresses,σ_xx. andσ_yy, in the vicinity ofCu(001) surface, dose not occur in the topmost layer, but in the two layers beneath. The reduction of the in-plane bulk stresses in the second and third top layers converts the third top layer from a tensile layer to a compressive one, though the resultant surface stress is still a tensile one.Adsorbate-induced surface stress is the change in surface stress due to adsorption of atoms on substrate, which not only has a decisive role in epitaxy nanofilm growth, but also has its technical implications e.g. fabrication of nanofilm-based sensor. It has been observed in many experiments that even for fairly low coverage, a significant adsorbate-induced surface stress develops. In chapter 5, molecular dynamics simulations were carried out on Cu/Cu(001) and Al/Ni(001) to invest the dependence of adsorbate-induced surface stress on adatom concentration. A moment relaxation algorithm is used to minimize the total energy. Non-monotonic coverage dependence of adsorbate-induced surface stress was observed in the present atomistic simulations. In low coverage regime where interactions between neighboring adatoms are negligible and the surface stress mainly originates from the local interactions of each adatom with the underlying surface atoms in the neighborhood, a negative adsorbate-induced surface stress develops and changes linearly associated with adatom concentration. This trend could extend to full monolayer coverage if the interactions between neighboring adatoms were discarded in the atomistic simulations. In the medium coverage regime where the neighboring adatoms come into interaction with each other and dominates the contribution of adsorbate-induced surface stress, the compressive adsorbate-induced surface stress inverses into a tensile one and increase sharply to a maximum at the adatom concentration of half monolayer. In the large coverage regime, the tensile stress between neighboring adatoms drops as their separations reduce. This adsorbate-induced tensile stress is also counteracted by the continuing negative contribution of local adatom-substrate interactions, and the nanofilm eventually arrive back at the state with equally stressed double surfaces. The results burn out the opinion that neighboring adatom-adatom interactions are core-core repulsions. We further discuss the results in the light of a semiempirical framework based on embedded-atom method to explore the origin of surface stress. The present results can add to our knowledge base about systematics of surface stress. Strain gradient effect of micro- and nanoscale structures in plasticity is well known. Elastic strain gradient effect has also been reported, but so far only theoretical work about this issue could be found. No confirmatory data obtained from experiments and atomistic simulation has been published. In chapter 6, molecular dynamics simulation of Cu and Si nanofilms bending was carried out and compared to a strain gradient elasticity solution to ascertain whether there is also any strain gradient effect in elastic range, and its significance if presence. It is shown that there exists a noticeable inconsistency between atomistic simulation and theoretical prediction. In contrary to hardening effect predicted by the strain gradient elasticity, the atomistic simulation results reveal a significant softening effect in diamond silicon nanofilm bending when the thickness is less than 10nm. Although a hardening effect is exhibited in the atomistic simulation of FCC copper nanofilm bending, it is negligible. We regard this study as an attempt to draw more attention and discussion about this issue in the community of this area.