System-level line-edge roughness limits in extreme ultraviolet lithography

Abstract

As critical dimensions shrink, line edge and width roughness (LER and LWR) become of increasing concern. Traditionally LER is viewed as a resist-limited effect; however, as critical dimensions shrink and LER requirements become proportionally more stringent, system-level effects begin to play an important role. Recent advanced EUV resist testing results have demonstrated lower bounds on achievable LER at the level of approximately 2 to 3 nm. Here we use modeling to demonstrate that a significant portion of this low bound may in fact be do to system-level effects and in particular the mask. Of concern are both LER on the mask as well as roughness of the multilayer reflector. Modeling also shows roughness (flare) in the projection optics not to be of concern.

Correlation method for the measure of mask-induced line-edge roughness in extreme ultraviolet lithography

“…As demonstrated by the modeled mask-induced LER data in Table 1 (see Ref. [4] for details on the modeling procedure), coherence and defocus parameters have a more significant impact on image LER coupling coming from phase roughness on the mask rather than mask absorber LER. Note that defocus has a more significant impact with annular illumination for both sources of mask-induced LER.…”

“…Recent modeling results [4] have suggested that mask effects are indeed important In advanced EUV development tools such as the SEMA TECH Berkeley microfield exposure tool (BMET) [5]. Here we present the first resist-based measurements of such effects.…”

“…Perhaps the most significant of these system level effects for EUV is the mask. [1][2][3][4] The mask can contribute to LER in two different ways. The first and most obvious source is from LER on the mask itself.…”

Validity of the thin mask approximation in extreme ultraviolet mask roughness simulations

“…Mask contributors, especially multilayer phase roughness have recently been shown to be significant contributors to LER [1][2][3][4].…”

“…Since this effect has been experimentally verified [5], there has been significant interest in the implications it has on EUV mask phase roughness specifications and thus substrate roughness specifications. This question has been addressed in the past [6] using two dimensional (2D) scalar modeling with the thin mask approximation [4] yielding the stringent specification of 50 pm replicated surface roughness [6] for 22-nm half-pitch device fabrication. Thin mask modeling was used due to the significant memory and speed burdens that would be incurred with full three dimensional finite-difference time-domain (FDTD) modeling.…”