Scale dependence in mechanical behavior of silicon nanowires

Based on fabrication, characterization, and modeling, this thesis explores the multifaceted fundamentals of silicon nanowire mechanical behavior. Despite their significant potential as building blocks in nano electromechanical systems, studies on silicon nanowires have so far failed to converge to draw a comprehensive picture of the scale dependence of their fundamental mechanical properties. Aiming to clarify the existing lack of consensus in this field, nanowires obtained through top-down fabrication are tested in-situ in three-point bending configuration, where the monolithic technology provides i) well-defined boundary conditions without generating any interface, and hence any additional compliance, between nanowire samples and their supports and ii) ample space for accommodating bending deflections during testing. Metrology work is carried out for the quantification of both nanowire dimensions and intrinsic stresses. Bending behavior is then interpreted through a newly developed multiscale model that primarily incorporates the surface effect through surface-induced intrinsic stresses by considering the native oxide surface state and crystalline anisotropy of surfaces. This is achieved by a combined approach based on molecular dynamics simulations and modified core-shell model. Furthermore, the utilization of machine learning facilitated by comprehensive molecular dynamics simulations enhances the scope of the methodology established in this study, employing artificial intelligence as a practical instrument for understanding the nanomechanics of silicon nanowires. Finally, a thorough analysis of uncertainties attributed to the interplay between processing-structure-property in silicon nanowires is conducted. The findings 1) bridge the 10-nm barrier that exists between computational and experimental studies due to various challenges in both realms; 2) link the evolution of the modulus of elasticity to nanowire critical dimension through the parameterization of surface inhomogeneity. A stiffening trend is quantified with increasing critical dimension. Experimental trendlines are validated through independent analytical and machine learning methods enabled through atomistic simulations; 3) reveal theoretical strength limit achieved by nanowires below 20 nm. This robust finding is in stark contrast to a substantial body of literature on nanowires of similar dimensions reporting strength values ranging from a few MPa to 20 GPa; 4) yield a confidence level surpassing 95% for fracture strength and 85% for modulus of elasticity estimations of silicon nanowires via error analysis of systematic and random error sources; 5) shed light on the crucial influence of the native oxide on the mechanical properties of silicon nanowires. Following the miniaturization trend in electronics as predicted by Moore's law, silicon nanowires due to their minuscule size resulting in exceptionally high resonance frequencies and significant surface-to-volume ratios emerge as primary candidates for transducers in the next generation of miniaturized sensors based on nano electromechanical systems. Employing a methodology that integrates comprehensive theoretical and multi-scale modeling along with state-of-the-art fabrication and characterization techniques, a series of enduring issues related to the interpretation of the scale effect in silicon nanowires have been conclusively resolved. The same approach can be applied to other low-dimensional materials including SiC, ZnO, core-shell and alloy nanowires as well as 2D materials including MoS2 and WS2.