Monte Carlo sensitivity analysis of EUV mask reflectivity and its impact on OPC accuracy

Unlike optical masks which are transmissive optical elements, use of extreme ultraviolet (EUV) radiation requires a reflective mask structure - a multi-layer coating consisting of alternating layers of high-Z (wave impedance) and low-Z materials that provide enhanced reflectivity over a narrow wavelength band peaked at the Bragg wavelength. (1) Absorber side wall angle, corner rounding,(2) surface roughness,(3) and defects(4) affect mask performance, but even seemingly simple parameters like bulk reflectivity on mirror and absorber surfaces can have a profound influence on imaging. For instance, using inaccurate reflectivity values at small and large incident angles would diminish the bene fits of source mask co-optimization (SMO) and result in larger than expected pattern shifts. The goal of our work is to calculate the variation in mask reflectivity due to various sources of inaccuracies using Monte Carlo simulations. Such calculation is necessary as small changes in the thickness and optical properties of the high-Z and low-Z materials can cause substantial variations in reflectivity. This is further complicated by undesirable intermixing between the two materials used to create the reflector.(5) One of the key contributors to mask reflectivity fluctuation is identified to be the intermixing layer thickness. We also investigate the impacts on OPC when the wrong mask information is provided, and evaluate the deterioration of overlapping process window. For a hypothetical N7 via layer, the lack of accurate mask information costs 25% of the depth of focus at 5% exposure latitude. Our work would allow the determination of major contributors to mask reflectivity variation, drive experimental efforts of measuring such contributors, provide strategies to optimize mask reflectivity, and quantize the OPC errors due to imperfect mask modeling.