Growth and Printability of Multilayer Phase Defects on EUV Mask Blanks

Transcription

1 Growth and Printability of Multilayer Phase Defects on EUV Mask Blanks Ted Liang, Erdem Ultanir, Guojing Zhang, Seh-Jin Park Components Research Intel Corporation Erik Anderson, Eric Gullikson, Patrick Naulleau, Farhad Salmassi LBNL Paul Mirkarimi, Eberhard Spiller, Sherry Baker LLNL 1

2 Outline Types of defects on EUV mask Growth modes ML phase defects MP-PDM PDM test sample Printability on resist patterns Defect specification Resist effects Summary and final analysis 2

3 Four Major Categories of EUV Mask Defects I: Substrate defects: propagate to ML surface II: ML defects: substrate and deposition process III: Absorber pattern defects: mask fabrication process IV: Soft defects: contaminations from handling and use (Wed. talk DI-02) (Poster MA-P03) Absorber (TaN, 80nm) ML cap (Ru, 2.5nm) ML (Mo-Si, 280nm) (~3nm Mo/4nm Si) Substrate (LTEM, ¼ ) Conductive layer 3

4 Nature and Growth of ML Defects IBD is widely used for ML coating for mask blanks Defect nature: bumps (particles) or pits Defect growth/evolution depends on deposition conditions Off-normal Near-normal Magnetron 60nm Au 60nm Au Optical manifestations in resist printing Phase Amplitude Or both components 50nm Au P. Mirkarimi 4

5 Acceptable ML Defect Spec ML defects are extremely complex Defect understanding necessary for defect reduction Defect sources Impact to resist printing of mask patterns Detection and disposition Defect specifications shall achieve universal acceptance Specs define the amount of development required for blank quality and inspection tool capability ($$) suppliers vs. users Specs must be data-based resist printing and validated modeling We use well-characterized model system PDM Programmed defect mask 5

6 MP-PDM PDM Design MP-PDM: PDM: ML L defects placed near absorber Patterns with full range of sizes/shapes and proximity Cell layout 10 bands 10 absorber line CDs 25nm to 90nm 3 absorber pattern L:S ratios 1:1 1:3 1:5 1x1 1x2 1x3 1x5 4 ML defect shapes: W x L ML defects were produced from substrate patterns Specially tuned ML deposition process Wide range of defects on the same test plate 6

7 MP-PDM PDM sub-cell: ML bump size and shape Label for substrate pattern shape Absorber patterns: CD and L:S ratio (50nm 1:1) 17 ML defect sizes 20nm to 100nm Varying proximity Defect pointers 7

8 MP-PDM PDM Sub-cell: ML Bump Proximity ML bumps in full range of proximity to patterns Line Space Line SEM image of 50nm 1:1 lines d proximity offset a b c d e f 10nm 50nm, 1X d = 0 Each defect size/location repeated 3 times for better statistics in printed resist CD measurements 8

9 MP-PDM PDM Fabrication: Key Steps Three-layer patterning with alignment marks for registration Materials preparation and pattern etch at Intel ML coating at LLNL EB resist patterning at LBNL (Now capable at Intel) 1. Super-smooth substrate 2. Backside CrN 3. Thin Ru/Si 5. 48nm HSQ (2 st layer) 6. ML coat w/ 8. Resist (3 rd layer) 4. TaN (1 st layer) 7. TaN Final MP-PDM 9. Absorber Use HSQ CD to label ML defect size 9

10 ML Phase Defect Characterization ML bump size measurements by fine AFM scans Surface profile 75nm 70nm HSQ 65nm Line scan FWHM H ML defect characterized as surface bump Height x FWHM HSQ CD 10

11 ML Phase Defect Characterization (cont d) Wide range of sizes Surface height (nm) Height: 0 to 8nm FWHM: 30nm to 70nm Measured sizes of ML phase bumps Defect label as HSQ size (nm) Height FWHM Rectangular ML phase bumps This talk focuses on square ML bumps (point defects) FWHM (nm) Surface height (nm) x2 HSQ 1x3 HSW Defect label as HSQ size (nm) 11

12 EUV Exposure Conditions MET: N.A = 0.3, 5X reduction, ~4 º Annular pupil fill Annular, Monopole, Dipole, 45 o Annular (0.35 < σ < 0.55) Resist: 112nm, ~20mJ/cm 2 Resolution limits: >40nm Process window (11x17 FEMs) DOF: ~ ±150nm Exposure latitude: ~ ± 5% Y-monopole (σ dia = 0.53) 45 o dipole (σ dia = 0.5) 1. P. Naulleau et al, Proc. SPIE, Vol (2006). 2. T. Liang, et al, Proc. SPIE, Vol (2006) 12

13 Effect of Illuminations Example: 3 sets of defects at between 50nm 1:1 lines Annular Dipole Space ΔCD/CD (%) ML bump Ht (nm) Dipole Annular 40 Y-mono Defect label Defects are more printable under dipole illumination Possibly due to higher resol. in both horiz. and vert. directions 13

14 Effect of Defect Proximity Resist CD change (%( % ΔCD/CD) Δ for 50nm 1:1 lines Space ΔCD/CD (%) ML bump Ht (nm) d=12nm d=22nm d=32nm d=42nm Y-monopole Defect label Space ΔCD/CD (%) ML bump Ht (nm) d=12nm d=22nm d=32nm Annular Defect label Space ΔCD/CD (%) ML bump Ht (nm) d=12nm d=22nm d=32nm d=42nm d=52nm Defect label Dipole Defects are most printable at center between 2 lines Defects are non-printable when 1/2 under absorber line Pattern covering is effective to render defect non-printable 14

15 Effect of Defocus Phase defect printability is expected to vary with defocus Space ΔCD/CD 40% 35% 30% 25% 20% 15% 10% 5% 0% 45nm lines vs. focus Focus = 0nm Defect Labels Effect reduced for larger ML defect Larger defects have amplitude components + Focus = =100nm 15

16 Comparison with Simulations There is mismatch with aerial image only simulations Under-predicts printability Disagrees in proximity effect Possible causes Resist effects (resolution, EL) ML defect size/shape accuracy Resist effects need further Investigation Resist model AIM measurements Space %ΔCD/CD 25% 20% 15% 10% 5% 0% SSA simulated results 50nm 1:1 L:S; Y-monopole d = 25nm d = 15nm d = 5nm d = -5nm Bump Ht (nm) 16

17 ML Defect Specification Discussion For example: 50nm 1:1 line printing 2.5nm x 50nm at center of line considered critical Non-printable at line edge Factors to consider in ML defect specifications Exposure conditions Allowable %ΔCD/CD:% layer dependent ~20% Proximity effect: Pattern placement with respect to defects Resist effects and limitations (resol., LWR, EL) Post-resist processing Non-linear resist effects makes scaling defect spec to smaller device patterns difficult, if not impossible Resist model unreliable to extend to patterns beyond the calibrated pattern geometries/size *Ref: Zhang and Liang, BACUS 2007 Aerial image simulation under-predicts printability when EL is small* Further defect printability studies require adequate resist performance 17

18 Defect Specification Discussion (cont d) Substrate defect specification is more complex strongly depends on deposition conditions Near-normal: <30nm (producing 2.5nm ML bump) Off-normal: << 30nm due to decoration phenomenon 30nm 25nm ML bump Off-normal process shall be avoided Moderate smoothing: >30nm Smoothing or other rendering scheme highly desirable, probably a must Impractical to expect a tool for substrate inspection with 100% capture Invisible substrate defects (sub-threshold) are high risk 18

19 Summary ML defects are complex use model MP-PDM PDM for comprehensive investigations of true ML phase defects ML phase defect printability is sensitive to exposure conditions and resist process Aerial image simulation seems to under-predict defect printability Covering defects, even partially, with absorber patterns is very effective ( rewarding( rewarding ) ) to render defects non- printable Useful ML blanks may not be necessarily defect-free Max. # of allowable defects depends on device layer structures 19

20 Acknowledgements MET exposures at Berkeley: Paul Denham, Brain Hoef, Gideon Jones, Jerrin Chui and Ken Goldberg MET exposures at Intel: Jeanette Roberts Gil Vandentop, Alan Stivers, Pei-Yang Yan, Emily Shu of Intel 20