Proceedings Article | 23 November 2024
H. Raut, M. Hasan, A. Das, M. Pak
KEYWORDS: Optical lithography, Extreme ultraviolet lithography, Semiconducting wafers, Printing, Photoresist materials, Lithography, Inspection, Electron beam lithography, Displays, Critical dimension metrology
Successful scale-down of emerging memory devices, implementing the cross-point memory architecture, require further compaction of the orthogonal or 4F2 cell configuration. A higher cell densification is typically achieved by pitch scaling of an orthogonal array of vertical pillars. Such aggressive pitch scaling is nowadays enabled by single exposure extreme ultraviolet (EUV) lithography, albeit with lingering challenges of simultaneous control over resolution, local critical dimension uniformity (LCDU) and edge roughness. In this respect, the current study focuses on single exposure EUV (NXE:3400) lithography at 34nm half-pitch to establish robust process conditions and minimized defectivity. To that end, we screen 4 different mask geometries. In particular, masks comprising square geometry (typically used to print circular structures) is compared with other geometries, namely hexagon, corner cut and circle. The choice of photoresist for the single patterning application is metal-oxide resist (MOR). The as-printed pillar arrays with 34nm pitch and 17nm target CD, are evaluated for both process throughput performance and post-lithography performance by comparing dose to size (DTS), LCDU, wafer critical dimension uniformity (WCDU), contact edge roughness (CER) and defectivity, which also contribute to potential device performance. To help select a process that simultaneously meets high throughput and precision requirements, a figure of merit, combining dose to size and local CD uniformity is determined and compared for all mask geometries. The results show that the corner cut mask geometry, outperforms all other mask geometries by utilizing a relatively lower DTS, to print orthogonal pillar arrays of mean CD 17.22 ± 0.01nm, WCDU of ~0.6nm, LCDU of 2.58 ± 0.01nm, and CER of 3.32 ± 0.007nm. Additionally, an investigation of defectivity using e-beam inspection shows a lower defectivity count of ~8 defective pillars in every 100,000 printed features.
