Modeling and experiments of non-telecentric thick mask effects for EUV lithography

Abstract

Various issues related to non-telecentric mask effects for EUV lithography will be discussed in this paper. First, a raytracing approach will provide a conceptual description as to the nature of non-telecentric thick mask effects, highlighting the behavior of various edge types and corners. Rigorous modeling of these effects are discussed along with a few consequences of improper modeling. A series of simulation and experimental studies are presented to probe both the one-and two-dimensional impact of thick mask effects. It will be shown that a simple constant edge bias appears sufficient for 1D features, but begins to break down when space-widths are less than about 45 nm. Investigation into the impact of corners and small 2D features indicates that a simple edge-based bias also breaks down for edge lengths less than about 60nm. A sample set of rules-based post-OPC HV corrections for 22nm node dimensions are proposed, although based on experimental results, it is concluded that more accurate resist modeling and scanner stability are required before being able to precisely predict and control HV effects. Finally, with some simplifying assumptions, simulation is used to predict the extent of potential HV effects of future EUV imaging systems.

Depth of focus in high-NA EUV lithography: a simulation study

“…Mask structure details are reported at the bottom-left of each table: these values are at wafer level, for mask dimensions they must be multiplied by 8 in case of horizontal structures, by 4 if vertical. It should be noted how the SMO naturally leads to having strong asymmetry along the vertical axis of both illumination and mask SRAF CD and SRAF Spacing to counterbalance mask shadowing effects which naturally distort positive and negative diffraction orders 5,6,7 .…”

“…This comes after almost 20 years of steady ~ 100 nm DoF, from the last nodes printed with ArFi lithography to current nodes where EUV has been implemented. Unfortunately, due to the reflective nature of EUV optical systems, several across-focus effects such as best focus shift 5,6 and Pattern Placement Errors (PPE) 7,8 further limit the Process Window (PW) when different structures must be combined.…”

Fast 3D thick mask model for full-chip EUVL simulations

“…One of such effects is asymmetric shadowing effect that presents itself as HV print bias between horizontal (H) and vertical (V) features. Although it can be modeled to a certain degree using the rule-based HV mask bias in the thin mask model 4 , the accuracy becomes inadequate as the feature size reduces 5 . Pattern shift is another such effect caused by mask-side non-telecentricity due to the oblique chief ray angle.…”

“…Pattern shift is another such effect caused by mask-side non-telecentricity due to the oblique chief ray angle. It is important to note that the large global pattern shift observed in early EUV simulations using 3D mask models is largely a modeling artifact and can be corrected by transforming the mask diffraction coefficient to the optimum object plane before passing it to the subsequent simulation steps [4][5] . This process mimics the EUV scanner setup stage that moves the mask to the optimum mask focal plane.…”

Thin absorber EUV photomask based on mixed Ni and TaN material

“…In order to achieve the typical requirement for the reflectivity to be below 2%, a Ta-based layer requires an approximate ,,Silicon substrate thickness 60-70 nm. The typical thickness is nearly five times the wavelength of the EUV light, which could lead to an increased Horizontal-Vertical (HV) print bias [9][10][11][12][13], increased number of defects, and would also require a larger input power [6]. EUVL requires oblique incidence, typically 6 degrees or greater, due to the use of all reflective optics.…”