Characterization of Actinic Mask Blank Inspection for Improving Sensitivity Yoshihiro Tezuka, Toshihiko Tanaka, Tsuneo Terasawa, Toshihisa Tomie * M-ASET, Tsukuba, Japan * M-ASRC, AIST, Tsukuba, Japan E-mail: tezukay@mirai.aist.go.jp Y. Tezuka, M 25 EUVL Symposium, Nov.7 th 1
Outline Introduction Goal / positioning of M Experimental setup / Inspection image Opportunity factors PSF characterization Mask blank roughness Sensitivity prediction Risk factors Defect seed dependency Prototype design Summary Y. Tezuka, M 25 EUVL Symposium 2
Positioning of Research in M Goals Phase 1 (~Mar 4) Demonstrate POC of novel mask blank inspection Successfully Completed! Phase 2 (~Mar 6) Complete design of full-field prototype Opportunities and risks identified from POC tool Scope Development of advanced yet affordable mask blank inspection tool to: Support multilayer process development Study future generation capability -> Deliver top quality mask blank in EUVL pilot phase Y. Tezuka, M 25 EUVL Symposium 3
POC Tool & Inspection Image YAG Laser Zr Filter Ellipsoidal Mirror PC Zr filter Illuminator chamber Imaging optics chamber CCD camera EUV Spherical Mirror Tape Target CCD Camera (pixel=13µm) YAG Laser Mask blank doorway 625 Mask Blank Plane Mirror Schwarzschild Optics (2x, NA=.2) Light source chamber Clean Room Class 1 42x7nm FWHM x Height 8x5nm 7x3.5nm.5 mm Courtesy of Y. Tezuka, M 25 EUVL Symposium 4
Through Focus Images -4µm -3µm -2µm -1µm µm +1µm +2µm +3µm +4µm Fraction in Center Pixel.4.35.3.25.2.15.1.5 5x5 i, j A 3,3 BG ( A i BG ), j Local Local No. 25 No. 49-6 -4-2 2 4 6 Defocus (µm) Single pixel can collect only 25~35% even at best focus on optical axis +5µm +6µm +7µm -> Need quantitative analysis Y. Tezuka, M 25 EUVL Symposium 5
PSF Characterization Pixel Intensity P Fitting variables: x x PSF FWHM (µm) i,j D =, y D 25. 2. 15. 1. 5.. 1 2πσ σ x y P x x i,j : i+ 1 + 1 i D y j exp j, y, σ, σ y D : Defect coordinate within a pixel x 2 ( x x ) ( y y ) y D 2 2σ x + D 2 2σ y 2 dxdy 1 2 3 4 5 Defect No. Fitting 13µm PSF(FWHM) = 18.4 µm +/-.5 µm (95% Confidence Interval) Image height non-dominant -> Blur Budget Analysis Y. Tezuka, M 25 EUVL Symposium 6
Image Blur Budget Analysis Total PSF Optics PSF CCD PSF Field Curvature (Defocus) Aberration Encircled Energy Ratio 1.2 1.8.6.4.2 Estimate from 55 Fe photon counting image Good Agreement! Mirror data (geometric) Total PSF - CCD PSF (Experiment) 1 2 3 4 Estimate from Interferometry PSF Component : 66% CCD, 31% Optics Sensitivity will be improved by: 1. Higher magnification, even w/o optics quality improvement 2. CCD PSF improvement Y. Tezuka, M 25 EUVL Symposium 7
Defect Position Estimate Accuracy Y Coordinate (pixel) 479 478 477 476 475 474 473 472 471 47 469 468 1 pixel on mask = 65nm Programmed Defects 1 2 3 4 5 Defect No. Fitting Residual (pixel).5.4.3.2.1 -.1 -.2 -.3 -.4 -.5 1 2 3 4 5 Defect No. 3σ =.13 pixel = 87nm on mask Positional identification capability with sub-pixel resolution demonstrated Y. Tezuka, M 25 EUVL Symposium 8
Mask Blank Surface Roughness 1 B P i 2 16π fmax R ( ) 2 f 2 π λ fmin PSD( f ) df 1 rms =.145 nm in 1µm sq. PSD (nm4) 1 1 1 1.1 Experiment Assumed model.1.1.1.1 Spatial Frequency (1/nm) 1.8x BG intensity of assumed model Reduction of mask blank roughness critical, especially in spatial period range ~1nm Y. Tezuka, M 25 EUVL Symposium 9
Sensitivity Derivation Extrapolation from Experiments + Statistical scaling Signal Intensity (SBR) SNR 8 7 6 5 4 3 2 1 1 2 3 4 5 Volume (nm 3 ) S( volume) B( roughness, pixel) σ ( pixel) r ξ ( PSF) f ( PSF) pixel-to-pixel variability (%) = Normalized SNR 25 2 15 1 5 1.2 1.8.6.4.2.2.4.6.8 1 1.2 pixel size (µm).5 1 1.5 2 2.5 PSF FWHM (pixel) Y. Tezuka, M 25 EUVL Symposium 1
Sensitivity Prediction FWHM of 2nm-high Detectable Defects at SNR>7 Detection Probability (%) 1 Current MAG x 1.3 + PSF x.7 + BG x.7 33nm 39nm 4nm 42nm 47nm 48nm 48nm 55nm defect pixel Worst case location = largest undetectable Best case location = smallest detectable Defect Width FWHM (nm) Blank Roughness Reduction is Essential! Y. Tezuka, M 25 EUVL Symposium 11
Risk Factors Signal Intensity (SBR) 8 7 6 5 4 3 2 1 y =.153 x 6nm w x 1nm h Binarized Image Natural Defect 1 2 3 4 5 Volume (nm 3 ) AFM Profile 6nm Question: Weakness for 1nm-high defects? 1nm Y. Tezuka, M 25 EUVL Symposium 12
Height Dependence 7nm 1nm Intensity (arb. units) 3 25 2 15 1 5 7nm 1nm 2 4 6 8 1 Pixel 3µm 1nm-high step intensity ~1/6 -> FDTD EM Simulation Y. Tezuka, M 25 EUVL Symposium 13
2D wide line defect model Gaussian FWHM Gaussian + Flat +Gaussian Slope width: SW 9% 1% Defect height: H Shape: Gaussian Si layer Mo layer Period: 6.98 nm Bilayer #: 4 Simulated by EM-Suite TM Y. Tezuka, M 25 EUVL Symposium 14
Signal Intensity Simulation Intensity (arb. units) 1 8 6 4 2 FWHM 6nm 9nm 12nm NA in =.1 NA out =.2 (SW) (43nm) (65nm) (86nm) 7nm 1nm AFM profile of a truncated pyramid defect 2 4 6 8 1 12 Defect Height (nm) 53nm EM Simulation corroborates Intensity difference between 7nm and 1nm EM Simulation suggests intensity also depends on slope width Y. Tezuka, M 25 EUVL Symposium 15
Smoothing Mode Dependence Mode 1 Linear Layer 4 Mode 1 Decelerated Layer 4 Mode 3 Accelerated Layer 4 Intensity (arb. units) Top Height = 2nm Fixed Bottom height & smooth speed dependence 35 3 25 2 15 1 5 mode 1 mode 2 mode 3 NA in =.1 NA out =.2 2 4 6 8 1 12 Bottom Height (nm) 1nm-seeded defect may show low signal intensity uncorrelated with printability Y. Tezuka, M 25 EUVL Symposium 16
EUV Scattering Simulation Angular distribution from 2D Gaussian Line defect Intensity (Arb. units).7.6.5.4.3.2.1. FWHM:6nm.3.2.1 NA.1.2.3 Height 2nm 4nm 6nm 8nm 1nm -3-2 -1 1 2 3 Angle (deg) Higher NA is NOT a universal solution - background increase will surpass signal increase for low defects Y. Tezuka, M 25 EUVL Symposium 17 Background Intensity (%).7.6.5.4.3.2.1 Background intensity dependence on NA Assumed model Actual PSD.15.2.25.3.35 Outer NA
Full-field Prototype Design CCD Camera (TDI Operation) EUV Light Source 26x optics Mask Stage Synchronized Control Load Lock Model MIRACL-1 (1) Specification Objects 625 EUVL Mask Blanks Sensitivity < 4nm w x 2nm h (2) Throughput 2hrs/blank Pos. Accuracy < 25nm Light Source Illumination Imaging Optics Sensor Stage Software Alignment Interface Configuration wavelength=13.5nm, DPP, Every pulse Triggerble Critical Illumination, Ellipsoidal mirror + Plane mirror, Illum. area >.5mm sq. 26x Schwarzshild Optics, Inner NA =.1, Outer NA =.2~.3 Back-illuminated CCD, TDI Operation, Synchronized with light source pulses Continuous move, Interferometer feedback Automatic defect cllasification, Size inference, Position identification Focus / Position alignment function by fiducal mark SMIF-capable, EUV mask handling standard compliant (1: Maskblank Inspection for Reflective multilayer by ACtinic Light) (2: in use of ultra smooth blank) Y. Tezuka, M 25 EUVL Symposium 18
Summary Sensitivity-limiting factors characterized, improvement path identified Magnification, PSF, blank roughness are critical to bring sensitivity to hp 32nm node Potential risks identified Seeds height dependence, profile / smoothing dependence need continuous study Full-field prototype design nearly complete Risk mitigation paths explored Y. Tezuka, M 25 EUVL Symposium 19
Acknowledgments We would like to acknowledge : T. Shoki, K. Yamashiro, Y. Usui and O. Nagarekawa of HOYA Corporation for their fabrication of the programmed defect mask blank. Y. Sugiyama of Nikon Corporation for providing data on mirror quality of the Schwarzschild optics This work was performed as part of a Ministry of Economy, Trade and Industry (METI) Project of Japan under contract with the New Energy and Industrial Technology Development Organization (NEDO). Y. Tezuka, M 25 EUVL Symposium 2