[#5, MA] Defect printability of thin absorber mask in EUV lithography with refined LER resist Takashi Kamo, Hajime Aoyama, Yukiyasu Arisawa, Mihoko Kijima, Toshihiko Tanaka and Osamu Suga e-mail: kamo.takashi@selete.co.jp EUVL Symposium (October 19 ~ 1, 9) 1
Outline [1] Introduction - Hurdle on Defect Printability Evaluation - Breakdown of Printed LER/LWR [] Experimental Method - How to improve LER/LWR of Printed Image - Experimental condition [3] Results [] Summary EUVL Symposium (October 19 ~ 1, 9)
Hurdle on Defect Printability Evaluation Table Mask Pattern Image and Resist Pattern Image (3nm HP) EUVL Symposium (October 19 ~ 1, 9) T. Kamo, et al., SPIE Advanced Lithography 9 It is difficult to measure critical defect size precisely because printed pattern's lineedge/width-roughness (LER/LWR) is larger than the CD tolerance of 3nm HP and beyond. 3
[1] Mask/Blanks Process Breakdown of Printed LER/LWR - LER/LWR of Absorber Pattern - Roughness of Multilayer Surface Systematic LER/LWR [] Contrast/NILS of Aerial Image - Exposure Tool / Condition (NA,, Flare, ) - Mask Structure (Binary, Att-PSM *, ) [3] Resist Process -Resist Material - Resist Stack (Under Layer, ) - Post Exposure Treatment Random LER/LWR is reduced by CD averaging method of multiple exposure shots. *) T. Kamo, et al., Effect of mask absorber thickness on printability in EUV lithography with high resolution resist, Proc. SPIE 7 () EUVL Symposium (October 19 ~ 1, 9)
How to Improve LER/LWR of Printed Image? Improvement of resist and mask process Resist: SSR3, Mask: Conv. (3 : ~.nm) Resist: SMR3, Mask: Conv. (3 : ~5.nm) Resist: SMR3, Mask: Improved (3 : ~.nm) 3nm L/S printed image Resist Improvement Mask Improvement CD averaging method of multiple exposure shots to extract systematic component from printed pattern with LWR by reducing random components Mask Pattern Printed Image CD Ave. Shot1 Shot Shot3 Shot Shot5 Shot Shot7 EUVL Symposium (October 19 ~ 1, 9) CD averaging 5
Experimental Condition Blanks/Masks Blanks structure : LR-TaBN(51nm) / CrN buffer(nm) / Si cap (11nm) / M.L. pairs (Mo/Si) / substrate Mask defect evaluation: Mask CD-SEM NGR (NGR) SFET Experiment Exposure condition: Resist : Resist CD evaluation: NA=.3 (central obscuration: 3%), sigma (inner/outer)=.3/.7, Incident angle=deg, Magnification=1/5 Selete Model Resist 3 (nm thickness) S93II (HITACHI High-Technologies) Lithography Simulation Simulator : Exposure Condition : Pattern : EM-Suite TM (Panoramic Technology Inc.) NA=.3(Central Obscuration : 3%), sigma (inner /outer)=.3/.7, Incident angle=deg, Magnification=1/5, =13.5nm, No lens aberration, No flare, Resist blur model (sigma=9nm) 3nm L/S with program isolated/edge defect (Parallel to EUV light projection) EUVL Symposium (October 19 ~ 1, 9)
Definition of Defect Size and Pattern Orientation Mask (Design) Mask (Measured) Wafer (Measured) Averaged CD a Min. a Mask SEM image is flipped to the same direction as wafer SEM image. Defect Size (Design) = a Square shape programmed mask defect for simulation and design CD at programmed defect EUV Light Defect Size = S (S: Area) EUVL Symposium (October 19 ~ 1, 9) 7
Gauge of LWR, Printed CD at Programmed Defect Raw data 1 plot: Ave. of 1pixes (1.1nm Length) LWR at No Programmed Defect 3 =5.nm 3 =.5nm LWR value depends on parameters of CD analysis. Printed CD at Programmed Defect -3 - - 3 Average of 1pixels (1.1nm length) is too large to measure accurate CD at programmed defect. 5-5 - 5-3 - - 3-5 - EUVL Symposium (October 19 ~ 1, 9)
Mask and Printed Image at No Programmed Defect Mask Pattern Printed Image CD Shot1 Shot Shot3 Shot Shot5 Shot Shot7 Ave. 5.5nm.nm 5.5nm.nm 5.nm.5nm.9nm 5.7nm 7.nm.nm 5.5nm.1nm 5.nm.nm.7nm.3nm EUVL Symposium (October 19 ~ 1, 9) 9 Line - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 5.3nm.3nm 5.1nm.1nm5.1nm.nm5.nm.nm5.nm.nm.9nm.1nm5.nm.nm.nm 1.5nm Space - - 1 13 1 15 1 17 1 19 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 1 13 1 15 1 17 1 19 - - - - LWR (raw data) LWR (1plot: 1.1nm Length) LWR of 1 shot: Systematic LWR, Random1 shot LWR (raw data) LWR (1plot: 1.1nm Length) LWR of CD averaging: Systematic, Random1 shot / N (N: 7shots) Systematic component of printed LWR of 1shot is smaller than random component and is estimated to be less than % of 3HP.
Mask and Printed Image at Programmed Defect Edge Opaque Edge Clear EUVL Symposium (October 19 ~ 1, 9)
Mask and Printed Image at Programmed Defect Edge Opaque Edge Clear EUVL Symposium (October 19 ~ 1, 9) 11
Mask and Printed Image at Programmed Defect Isolated Opaque Isolated Clear EUVL Symposium (October 19 ~ 1, 9)
Printed CD vs Programmed Mask Defect Size Isolated Edge Space Width [nm] Space Width [nm] 3 3 3 3 1 1 1 3 3 3 3 1 1 1 Opaque Simulation Measured 1 Defect Size (1x) [nm] Sim. (Defect: Simulation 1nm t) Sim. (Defect: 3nm t) Sim. (Defect: Measured nm t) Sim. (Defect: nm t) Measured 1 Defect Size (1x) [nm] Except isolated opaque defect, good agreement is achieved between measured and simulated results. Reduction of isolated opaque defect height is possibly the reason of difference between measured and simulated results. EUVL Symposium (October 19 ~ 1, 9) 13 Line Width [nm] Line Width [nm] 3 3 3 3 1 1 1 3 3 3 3 1 1 1 Clear Simulation Measured 1 Defect Size (1x) [nm] Simulation Measured 1 Defect Size (1x) [nm]
Summary With conventional experimental procedure, it is difficult to measure precise critical defect size because printed pattern s line-edge/width roughness (LER/LWR) is larger than the CD tolerance of 3nm HP and beyond. In order to reduce systematic LER/LWR, mask process is improved. In order to reduce random LER/LWR, low LER resist material and CD averaging method of multiple exposure shots is introduced. Systematic component of printed LWR is smaller than random component and is estimated to be less than % of 3nm HP. Except isolated opaque defect, good agreement is achieved between measured and simulated results. Reduction of isolated opaque defect height is possibly the reason of difference between measured and simulated results. EUVL Symposium (October 19 ~ 1, 9) 1
Acknowledgements The authors would like to thank; Kosuke Takai, Kazuki Hagihara, Koji Murano, Masamitsu Ito of TOSHIBA Corporation for mask fabrication. Eiji Yamanaka of TOSHIBA Corporation for support of programmed mask defect measurement. This work was supported by NEDO. EUVL Symposium (October 19 ~ 1, 9) 15