This thesis investigates the possibility of substituting optical imaging of nano-electro-mechanical systems (NEMS) for the lengthy and costly process of scanning electron microscope inspection. Although image details are both modified and degraded by diffraction, nanoscale features can be seen in experimental images from conventional light microscopes, and excellent contrast is achieved in dark field (DF). The motivation is therefore to analyse scattering from sub-wavelength structures with visible electromagnetic (EM) waves and to establish a simple method of simulating their DF images, an otherwise extremely lengthy process.

A comprehensive review of different theories of rigorous two-dimensional (2D) diffraction theory is then presented. Geometrical theory of diffraction (GTD) is chosen as the best approach and validated by numerical simulation of the imaging of a set of periodic 2D canonical structures using the finite element method, which is compared with modal diffraction theory. It is demonstrated that only edge waves contribute to dark field images, and that GTD wedge diffraction coefficients provide good approximations to their amplitudes. A 2D model is then developed to account for direct scattering and re-scattering of in-plane radiation. The model is valid for both TE and TM polarizations.

Scattering from simulated optical imaging of 3D canonical structures is then investigated. The numerical results are carefully considered to establish the minimum number of waves needed to simulate diffuse and polychromatic illumination. A novel simulation tool is then developed to reduce the time taken to simulate dark field imaging of general NEMS. This model uses edge detection to identify the scattering edges, a convolution method to simulate in-plane scattering, and wedge diffraction coefficients from GTD to define edge wave amplitudes. The results are compared with 3D FEM simulations, and excellent agreement is obtained. Deficiencies in the model are identified and suggestions for future developments are made.