Strain engineering is a key technology to continue Moore's law with silicon or any other foreseen semiconductor in very large scale integration. The characterization of strain in nanostructures is important to determine the potential of these technologies, and it is typically performed using micro-Raman when investigating strained silicon. Here, we report on the Raman shift-stress behavior from the (001) silicon surface of highly strained ultra-thin (15 nm-thick) suspended nanowires with stresses in the range of 0–6.3 GPa along the [110] direction. We employ a strain technology that offers a precise control of stress values at large sampling while reducing variability. The stress level of the nanostructures has been accurately evaluated by the finite element method simulations and further correlated to the Raman spectra. For stresses below 4.5 GPa, the aforementioned behavior was linear and the extracted stress shift coefficient was in agreement with those reported in the literature. For stresses greater than 4.5 GPa, we show that the Raman shift-stress behavior resembles a quadratic function.

The mechanical properties characterization of silicon nanowires is generally performed by tensile nanomechanical loading tests with in situ strain quantification. While the strain is characterized by electron beam (e-beam) microscopy techniques, the understanding of the sample-electron interaction is essential to guarantee artifact-free measurements. In this work, we investigated suspended strained silicon nanowires under electron beam exposure in a scanning electron microscope (SEM). The fabricated nanowires had their initial stress profile characterized by Raman spectroscopy and finite element method simulations. Then, the sample was exposed to an e-beam where we observed a gradual electrical charging of the sample, verified by the image drift, and down deflection of the suspended nanowire caused by electrostatic forces. These additional stresses induced the mechanical fracture of the nanowires in the corner region due to accumulated stress. These results ascribe electrostatic mechanical loading concerns that may generate undesirable additional stresses in nanomechanical tests performed in SEM, demonstrating the importance of proper sample preparation to avoid electrostatic charging effects. Here, we propose a simple and effective method for imposing the structures under an impinging electron beam at an equipotential, which mitigates the charging effects acting on the nanowire.

No abstract

ResumoO tema deste projeto é a localização dos limites de uma estrada off-road através de visão. O trabalho consiste em duas técnicas diferentes para detectar a representação dos limites da pista. A primeira técnica usa processamento de imagem para detectar faixas brancas e representa-las em duas equações de segundo grau. Além disso, essa técnica foi implementada no projeto VERDE (Veículo Elétrico Robótico com Diferencial Eletrônico), uma plataforma robótica. A segunda técnica é na área de Aprendizado de Máquina, com a comparação de 3 modelos de Detecção de Objetos, detectando uma referência que representa os limites da pista. A escolha desses modelos de detecção foi feita entre um modelo preciso (Faster R-CNN), um modelo rápido (FastYOLOv2) e um modelo intermediário (SSD300).Palavras chave: Aprendizado de máquina; Aprendizagem supervisionada (Aprendizado do computador); Veículos autônomos; Redes neurais (Computação); Visão por computador.