Flare compensation in EUV lithography Place your image on top of this gray box. If no graphic is applicable, delete gray box and notch-out behind gray box, from the Title Master Jonathan Cobb, Ruiqi Tian, Robert Boone, Kevin Lucas, Scott Hector, Vladimir Ivin*, Mikhail Silakov*, and George Babushkin* *SOFT-TEC MOTOROLA and the Stylized M Logo are registered in the US Patent & Trademark Office. All other product or service names are the property of their respective owners. Motorola, Inc. 2002.
Outline Description of EUV flare Calculation details Compensation strategies Conclusions
Flare in EUV lithography Cause: surface roughness on optics Scales as 1/λ 2, so more problematic at EUV wavelengths Effects: Scatters light out of bright regions and into dark regions reduces contrast Couples local light intensity to features 1000 s of mm away pattern dependent Simple calculations 1% (absolute) change in flare causes 0.86 nm CD change
Estimating impact of flare variation on CD control Statistical simulation parameters (CCI design) Average mask transmission: 80 90% Mean focus error: -0.05 0.05 µm Cross-slit focus variation (1σ): 0.006 0.008 µm Mean dose error: ±10% Cross-slit dose variation (1σ): 2.0 2.6% Mean flare: 10 15% Flare variation (1σ): 1 3% (absolute) 500 calculations per set of conditions Constant simulation parameters 45nm lines on a 110nm pitch Partial coherence = 0.7 NA = 0.25 Wavefront error = 0.045λ Absorber stack 120nm thick Normal incidence
Simulation results Term Flare_Var(1,3) Def_Err(-0.05,0.05) Dose_Mean(0.27,0.33) Flare_Mean(10,15) Mask_T(80,90) Def_Err(-0.05,0.05)*Flare_Var(1,3) Def_Var(0.02,0.03) Mask_T(80,90)*Flare_Mean(10,15) Def_Var(0.02,0.03)*Dose_Mean(0.27,0.33) Mask_T(80,90)*Dose_Mean(0.27,0.33) Orthog Estimate -0.0424507 0.03301723 0.01629922-0.0122251-0.0108554-0.0083664-0.0074315-0.0037086-0.0029393-0.0029023 Flare variation is largest factor influencing CD control Effect of flare variation is 3.5X the effect of mean flare Model predicts 1σ flare variation must be less than 1.7% (absolute) for ±10% CD control at -0.05 µm focus error
10 13 Flare calculations, p. 1 PSD on optics PSF in image 10 11 10 9 r = λzf 10-9 10-11 PSD (nm 4 ) 10 7 10 5 PSF (nm -2 ) 10-13 10-15 1000 10-17 10 10-19 0.1 10-8 10-7 10-6 10-5 0.00010.001 0.01 0.1 Spatial Frequency (nm -1 ) 10-21 0.1 1 10 100 1000 10 4 10 5 10 6 Radial Distance (nm) Stearns et al. (J. Appl. Phys. 84, 1998): I x, y) I ( x, y) + I ( x, y) PSF( x, ( 0 0 y )
Flare calculations, p. 2 Aberrations F = <T> PSF(d) 10-9 10-11 Stearns PSF (nm -2 ) 10-13 10-15 10-17 10-19 r = λ NR 2( NA) 10-21 0.1 1 10 100 1000 10 4 10 5 10 6 Radial Distance (nm)
Memory Logic Flare calculations Data scaled so that poly lines are 0.09 µm (appropriate for 0.1 NA ETS) Include measured aberrations from ETS field center Average flare over 0.09x0.09 µm 2 regions inside of a 2x2 mm 2 section of mask data
Calculation results 18 16 Percent Flare 14 12 10 8 Memory 6 Logic 4 0 50 100 150 200 250 300 Distance from Corner (µm)
Proposed compensation strategies (Krautschik et al.) Selective sizing Resize lines according to known d(cd)/df response Apply global resizing in middle of mask where flare variation assumed to be small Apply local resizing in corners where variation is largest Iterate to convergent solution Dummification (i.e., tiling) Reduce flare variation by reducing pattern density variation Add dummy features to areas of low pattern density Dummy features must not interfere with circuit function
No place to apply global resizing 20 15 Percent Flare 10 5 0 0 500 1000 1500 2000 2500 3000 Distance from Corner (µm)
Effects of sizing on process window I x, y) = I ( x, y) (1 F) + F ( 0 NA = 0.25, Θ i = 6, σ = 0.7, λ = 13.5 nm, <T> = 75% T 1.2 1 Scalar, no flare, no aberrations 0.035 0.03 5% flare and 17.5%, best dose Relative Intensity 0.8 0.6 0.4 0.2 17.5% flare 5% flare 0-0.06-0.04-0.02 0 0.02 0.04 0.06 Horizontal Position ( µm) CD (µm) 0.025 0.02 0.015 0.01 0.005 17.5% flare with bias 17.5% flare, common dose 0-0.2-0.15-0.1-0.05 0 0.05 0.1 0.15 0.2 Defocus (µm)
Overlapping process window for dense and isolated lines possible with biasing but smaller window than without flare variation -0.2 0 0.2 Defocus (microns) 25% average pattern density 0.037 0.035 0.033 0.031 0.029 0.027 0.025 Aerial image linewidth (microns) 30 nm isolated line, 10% flare 31 nm dense line, 10% flare 30 nm isolated line, 15% flare 31 nm dense line, 15% flare 35 nm in resist Dense: 90 nm pitch Isolated: 270 nm pitch
MEEF effects minimal, and mean-totarget CD control not critical for biasing MEEF 1.06 1.04 1.02 1 0.98 0 10 20 Mean flare (% absolute) Best focus -50 nm defocus 50 nm defocus MEEF varies with flare and focus Effect of focus larger but only 1-2% Aerial image linewidth (microns) 0.04 0.035 0.03 0.025 25 35 Mask linewidth (nm) Best focus 50 nm of defocus -50 nm of defocus Linear behavior implies MTT CD control not critical 35 nm in resist; 90 nm pitch
Various tiling algorithms tried Origins in CMP processes Rule-based Insert dummy features in all appropriate empty space Increases pattern density uniformity over short length scales Model-based CMP Insert tiles according to empirical model that relates pattern density and polish uniformity Considers weighted pattern densities over mm length scales Model-based EUV Place subresolution tiles Minimize pattern density variation with optimization calculation that attempts to consider all relevant length scales in PSF Extend tile placement into borders with model-based CMP algorithm
Tiles from EUV algorithm Circuit features Tiles
Effects of tiles on flare 20 Percent Flare 15 10 5 Uncompensated EUV tiling 0 Rule-based tiling 0 50 100 150 200 250 300 Dis t ance from Corne r ( µm)
Conclusions Barring significant improvements in EUV optical fabrication technology, mask compensation will be required to reduce flare variation Selective sizing is feasible but is computationally expensive and reduces the focus latitude Tiling also reduces variation but certain features are not tiling-friendly Both selective sizing and tiling will likely be required for full compensation