(349b) Understanding the Relationship Between True and Measured Nanoscale Feature Size and Roughness Using a Detailed SEM Simulator

In the coming years, there will be more and more commercial applications that use nanotechnology. As in all manufacturing, there will be a need for process control and inspection of devices and materials that use this nanotechnology. While a variety of metrology and inspection tools will be used, electron microscopy, both SEM and TEM, will likely be some of the primary and most important inspection tools used. SEM is especially useful because it allows for inspections over a wide variety of length scales from the multi-micron to sub-10 nm. It is also minimally invasive and can be applied to a wide variety of materials and has minimal substrate requirements. As an example, one only need to look at integrated circuit fabrication, which even now can be considered nanomanufacturing. Current commercially available chips have minimal features on the order of 32 nm, and a major manufacturing metric is the smoothness of sides of these features which is on the order of 1-4 nm. SEM is the major source of metrology for these devices.

Despite these advantages, SEM has some limitations. It converts a three-dimensional structure into a two dimensional image. Likewise, since it is effectively integrating over a surface volume, the effect of some small features is lost. As SEM is used for more and more nanoscale devices, it is imperative that a better understanding is obtained between what the 2D SEM means about the 3D structure. To probe this question, we have developed a new highly rigorous SEM simulation tool. It considers both discrete elastic and inelastic scattering events, and as a result the generation of secondary electrons that are crucial to the SEM imaging process can be simulated with high fidelity. Inelastic scattering is determined using a combination of the complex dielectric function for the material in which the electron is propagating and Penn's algorithm, while elastic scattering is determined using the partial wave expansion of the Mott cross section for the material. Our model is capable of handling the nanoscale roughness found on such materials and devices. For the first application of this model, we have examined the SEM analysis of photoresist lines for metrology of high resolution lithography, since even now it is an important use of nanotechnology. We have determined that certain rough structures will appear completely smooth due to the integrating nature of the SEM analysis. Likewise, the measured line width and line edge roughness determined from the SEM do not necessarily match the true feature width and surface roughness. The limitations of the SEM for measuring such small features will be discussed and imaging strategies to improve the accuracy of the SEM compared to the true features will also be discussed.