Selective plasma enhanced atomic layer deposition for deep trench filling: A design-of-experiments approach

Bottom-up gap-filling processes, aiming to preferentially deposit materials from the bottom of high-aspect-ratio features frequently generate defects, such as voids or seams. Despite the implementation of highly conformal PE-ALD processes, non-uniform growth and/or early clogging at the top of 3D structures are commonly observed. To address this issue, super-cycle strategies are developed, which consist in periodically inserting a radical-enhanced inhibition step within PE-ALD cycles, to selectively hinder deposition on upper sidewalls, thereby promoting uniform, defect-free filling. Due to the absence of reliable non-destructive scientific diagnostic techniques to assess defects in deep trenches, we propose a statistical approach as an alternative. A design-of-experiments approach is developed based on three experimental parameters (plasma duration, plasma discharge power and inhibiting gas flow) and reflecting the inhibition depth. Although these parameters linearly affect the inhibition depth, obtained results show that plasma duration and flow of gas of interest have the most significant impact, while factor-to-factor first order interactions have a more limited effect. To confirm the validity of our design-of-experiments approach, measured inhibition depths in trenches were compared with the values given by the predictive model. Finally, the obtained model is generalized by being extended to trenches of various dimensions (both in terms of depth and density) by evaluating the equivalent total inhibited surface rather than inhibition depth which is only made possible by the inherent self-limiting surface-dependent and opening-independent characteristics of the inhibiting step.