Proceedings Article | 20 March 2019
KEYWORDS: Etching, Bridges, Photoresist processing, Extreme ultraviolet lithography, Metrology, Computer simulations, Focus stacking software, Stochastic processes, Time metrology, Optical lithography
Despite the innumerable advances in EUV lithography for materials, optics, and process in recent years, the N7 and N5 targets for resolution, line roughness, and sensitivity (RLS, collectively) have not simultaneously been achieved from both a yield and cost perspective due to their interdependent nature. For example, reducing dose to an economically practical level results in resolution and roughness levels that generate unacceptable yield and performance. These limitations have traditionally been viewed in a light where only conventional post-lithographic processing by etch is used for the subsequent pattern transfer. Recent post-lithography defect mitigation techniques, however, combine the use of high etch selectivity underlayers, atomic layer etch (ALE) descum, and fast-switching deposition and etch resist linespace pattern repair have overcome the limitations of the RLS interdependency by correcting for resolution and stochastics related defects and improving LER associated with lower exposure dose [1]. Here, an aspect ratio dependent deposition can be used to protect the resist lines from further notching and damage while allowing for residual scum between the lines to be etched [2]. A lithography, etch, and metrology feedback loop can be envisioned in which a minimum dose requirement is found where both the LER and lithography related defects are still correctable using post-exposure processing; however, due to the extremely long metrology times for e-beam inspection combined with the large quantity of adjustable parameters, traditional experimental DOEs quickly become unmanageable. The intractability of this situation necessitates a simulation-based parameter space optimization to reduce the required number of feedback cycles. In this study, Coventor SEMulator3D® is used to find optimized solutions for minimizing resist line bridges and breaks as well as line smoothing. Here, distributions of line roughness and resist divot and scum dimensions can be subjected to simulated process recipes with tunable parameters like etch selectivities, aspect ratio dependence of deposition and etch, deposition and etch rates, number of ALE cycles, etc. that one would typically explore in this defect mitigation strategy. The resulting defects can then be analyzed to determine how each defect in the distribution reacts to a given treatment. Additionally, further insight can be gleaned regarding the types and dimensions of defects that can be corrected and that would otherwise not be measurable using any physical metrology. For example, at low dose, a portion of the defect size distribution is comparable to the drawn features and can no longer be corrected by the post-exposure treatment. Due to the randomness of the defects and small size, three-dimensional characterization is difficult, but using simulation, it is possible to show at what dimension the defect mitigation strategy begins to fail.
