Experimental and simulation studies of printability of buried EUV mask defects and study of EUV reflectivity loss mechanisms due to standard EUV mask cleaning processes

There's a big push for development and commercialization of extreme ultraviolet (EUV) lithography for high-volume semiconductor manufacturing of 14 nm half-pitch patterning and beyond. One of the primary concerns for making this a reality has been the ability to achieve defect-free masks. My study is focused on two aspects related to the performance degradation of the EUV masks namely EUV mask cleaning induced reflectivity loss mechanisms, and the buried multilayer phase defects in EUV masks. Standard EUV mask cleaning processes involve various steps and chemistries that lead to degradation of the multilayer structure, in turn leading to EUV reflectivity loss, hence impacting the throughput. In this work, we developed an understanding of the root causes of how different cleaning chemistries cause EUV reflectivity loss, and quantified these losses. Upon subjecting the mask blanks to multiple cleaning cycles, multilayer degradation observed was in terms of multilayer etching and pitting, surface roughness increase and surface oxidation. We studied each of these phenomena through experiments and simulations, and correlated the extent of EUV reflectivity loss to each of the degradation phenomena. The second aspect of this work involved investigating the printability behavior of buried native phase defects in EUV mask blanks. Since completely defect-free masks will be hard to achieve, it is essential to have a good understanding of the printability of EUV mask defects. In this work, the printability of native mask defects was studied using a novel level-set multilayer growth model that took into account the tool deposition conditions where the multilayer deposition took place. For this study, the native mask blank defects were characterized using atomic force microscopy (AFM) and cross-section transmission electron microscopy (TEM), and the defect printability of these defects was evaluated using simulations implementing the finite-difference time-domain (FDTD) and the waveguide algorithms. The simulation results were compared with through-focus aerial images obtained at the SEMATECH-Berkeley Actinic Inspection Tool (AIT), an EUV mask-imaging microscope at Lawrence Berkeley National Laboratory (LBNL), and a good match was obtained between them. Further, an approximate but robust method was developed for investigating the defect printability of arbitrarily-shaped native defects given the AFM defect profiles on top of the multilayer stack. Given the full-width at half-maximum (FWHM) and height of the defect, as obtained from top layer AFM, we were able to infer the bottom defect profile in terms of FWHM and height, through study of multiple native EUV mask defects. Multilayer growth over native, Gaussian and regular-shaped substrate defect profiles (having similar FWHM and heights) was simulated using the level-set multilayer growth model and their printability performances were compared, in terms of aerial image intensities, and reasonable comparison was obtained between them