Proceedings Article | 14 September 2001
KEYWORDS: Photomasks, Diffraction, Phase shifts, Lithography, Nanoimprint lithography, Etching, Image transmission, Electromagnetism, Semiconducting wafers, Critical dimension metrology
Our previous work showed that for 100nm lines, the Sidewall Chrome Alternating Aperture (SCAA) mask structure could overcome the problem of transmission and phase imbalance among multiple pitch structures. In that work, we explained the SCAA mask concept, showed a brief electromagnetic field (EMF) simulated comparison to two subtractive etch techniques and proposed a fabrication paradigm that could make SCAA a reality. What we did not show, however, was the detail of our EMF simulation work for any of these masks. Our current work provides this missing item and explores across pitch performance at 248nm wavelength for several masks meant to optimize alternating phase-shift (altPSM) mask phase and transmission: SCAA, asymmetric lateral biased, additive, undercut, dual trench (with and without undercut), mask-phase-only, and uncompensated. First, we discuss why vector electromagnetic field (EMF) simulation is necessary. Then we describe a typical optimization approach. There we describe how two simulators, ProMAX (FINLE Technologies, Inc.) and TEMPESTpr (Panoramic Technologies), were set up to reduce grid snapping and other simulation pitfalls, as well as EMF output analysis and topography optimization techniques using one mask type as an example. The optimization approach was to find the best topography for the 100:200nm line:space mask of each type according to the phase and transmission errors extracted from the EMF simulated diffraction orders. Because phase and transmission errors in an alternating PSM are both coupled to the existence of a non-zero central diffraction order, we screened mask topographies according to the zero diffraction order power, relative to power in the first orders. Monitoring the central diffraction order did prove be a useful technique for optimizing topographies because it is a single attribute that correlates to both phase and transmission errors, which are coupled and thus difficult to optimize concurrently. The same topography adjustments from the 300nm pitch optimization were then applied through pitch with fixed 100nm line. Next we summarize the EMF results for each mask compensation technique. Mask types were ranked according to best sum of central diffraction order power through pitch, effectively ranking phase and transmission performance across pitch by mask type. The highest ranking masks were SCAA (with 15nm ARC on chrome and no topography adjustments from ideal) and the asymmetric biased mask (with no ARC but with 40nm increase in each side of shifter space width at mask scale). The lowest performing masks were dual-trench (mainly because of phase errors) and the unadjusted mask (mainly due to transmission errors). Finally we move from EMF to lithographic simulation of the best two masks according to EMF simulation. For SCAA and asymmetric bias we examine the NILS and MEEF (with line size 90nm, 100nm, and 110nm) for 300nm pitch. Responses for the process window analysis include resist linewidth, resist retention, sidewall angle and feature placement. The analysis showed that SCAA and optimized asymmetric bias had identical NILS through focus, but that image CD was less sensitive to focus on a SCAA mask than on an asymmetric biased mask. The MEEF results were 0.9 for both masks, while SCAA had better depth of focus than the asymmetric biased mask for single line sizes. While the asymmetric biased mask is simpler to build with existing mask production processes, it requires EMF simulation to determine optimum topography (as do all the other compensation techniques in this study). SCAA requires a non-standard chrome deposition, but performed well according to lithographic simulations without any EMF simulation and topography adjustment. Both SCAA and asymmetric biased masks, it should be noted, did not require any undercut. Future work aimed at the most promising altPSM mask types is needed to further quantify sensitivity to expected fabrication variations and to gain experience with physical wafer prints.
