2008 IMPACT Workshop. Faculty Presentation: Lithography. By Andy Neureuther, Costas Spanos, Kameshwar Poolla, EECS and ME, UC Berkeley

Size: px

Start display at page:

Download "2008 IMPACT Workshop. Faculty Presentation: Lithography. By Andy Neureuther, Costas Spanos, Kameshwar Poolla, EECS and ME, UC Berkeley"

Garry Banks
6 years ago
Views:

1 2008 IMPACT Workshop Faculty Presentation: Lithography By Andy Neureuther, Costas Spanos, Kameshwar Poolla, EECS and ME, UC Berkeley IMPACT Lithography 1

2 Current Milestones Litho 1: Develop and experimentally verify process effect monitoring and full chip assessment methodologies Parameter specific test patterns and circuits; Sim & Exp. Verify Litho 2: Electromagnetic modeling methods and electromagnetic phenomena Mask edge effects; Plasmon monitors; Inspection signals Litho 3: Fast-CAD for full process window characterization of double patterning and guidance during decomposition Pattern Matching of post-decomposition and real-time decomposition Litho 4: Assessment tools for double patterning decomposition State dependent learning to maximize process window NT 4: ODP-based parameter extraction and silicon verification Optimized dual period 1-D gratings for optical aberration extraction IMPACT Lithography 2

Gate Variation Sources and Monitors Goal: Use process/device/circuit simulation to understand process variation electrical contributions and develop parameter specific electronic monitors 1443 with

3 Gate Variation Sources and Monitors Goal: Use process/device/circuit simulation to understand process variation electrical contributions and develop parameter specific electronic monitors 1443 with Lateral interactions between standard cells using pattern matching (BACUS 07) Interpreting RO results from BWRC 90 nm and 45 nm New evidence of significant effects from TI-SPIE Lynn Wang Hyper-Sensitive Parameter-Identifying Ring Oscillators for Lithography Process Monitoring (SPIE 08) 5X sensitivity enhancement; Pulse RO circuits Dry lab case for Si experiment Methodology for SEM to electrical performance sensitivity IMPACT Lithography 3

4 Cell-to-Cell Interaction vs. Distance Oscillates Pattern Matching Metal 1 Focus P 2 = 0.5 λ NA Periodic oscillatory behavior Good spacing and bad spacing, saves area Lynn Wang PMF Peak-to-Peak Range: 0.07, ~ 7% variation in linewidth IMPACT Lithography 4

5 Hyper-Sensitive Parameter-Identifying RO Focus Sensitive Large P-MOS to enable fast pull-up Patterns J. Holwill. SPIE 06 J. Holwill. BAUCS 07 W. Poppe. Ph.D. Thesis 07 Circuits Vfb Vout Lynn Wang Gate under Test Focus Monitors Dense 3.0% -80nm to 40nm Focus Isolated 5.5% Bulge (nonrectangular) 12.8% Phase Shift 27.8% BWRC RO IMPACT Lithography 5 5X more sensitive

Interconnect Variation Sources and Monitors Goal: Use process/device/circuit simulation to understand process variation electrical contributions and develop parameter specific electronic monitors

6 Interconnect Variation Sources and Monitors Goal: Use process/device/circuit simulation to understand process variation electrical contributions and develop parameter specific electronic monitors 1443 with Eric Chin Prediction of interconnect delay variations using pattern matching (SPIE 07) Slack time order in a design changes through focus (BWRC) Modeling timing across the lithographic process window (SPIE 08) Various interconnect scenarios and sensitivities Bosson plot for R, C and delay (F-E algebraic model) feasible Emperical F-E models; Multilayer Metal; Double patterning IMPACT Lithography 6

7 Second Interconnect Topology Eric Chin Worst case change in delay across FE window: 0.78ps (0.82%) Self compensating R, C minimizes impact on delay. IMPACT Lithography 7

Phase Shift Layout as a Process Monitor Eric Chin Bossung curves shows an asymmetric change in linewidth with focus, while linespace remains relatively flat.

8 Phase Shift Layout as a Process Monitor Eric Chin Bossung curves shows an asymmetric change in linewidth with focus, while linespace remains relatively flat. Worst case change across the FE window is 41.8ps (35%). This can be amplified by increasing the load capacitance to S3=64, resulting in the worst case change of 59.5ps (40%). An 11 stage ring oscillator with this phase shift interconnect topology between each stage would experience a frequency shift of 244 MHz (32%) from focus variations. IMPACT Lithography 8

Goals: Characterize ATT-PSM edge effects, explore plasmon monitors, and examine speed-ups in simulation and interpretation of signals in mask and wafer inspection.

9 Goals: Characterize ATT-PSM edge effects, explore plasmon monitors, and examine speed-ups in simulation and interpretation of signals in mask and wafer inspection. Electromagnetic Modeling Marshal Miller Characterization and monitoring of photomask edge effects (BACUS 07) ATT-PSM has 20 nm of quadrature light on each edge This tilts the Bossong plot and it is noticeable at 45 nm Impact of Photomask Quadrature Edge Effects through Focus (SPIE 08) MoSi line-ends are about 3X worse than line edges MoSi has only a slight incident angle effect Effects are larger for Ta oxide masks than MoSi (waveguiding?) Dry lab experiment for monitoring ATT-PSM and plasmons IMPACT Lithography 9

10 MoSi: Real and Imaginary Edge Contributions TE TM TM TE MoSi mask shows variation through both incident angle and pitch of about 5nm Clear polarization dependence Real: TE effect is double TM Marshal Miller Imaginary: Opposite, TM is worse IMPACT Lithography 10

Marshal Miller 3-D Simulations: MoSi 3-D FDTD simulations of MoSi line

far more sensitive through focus (up to 5x) Tilt in LES curve apparent

Defocus - 80 nm - 40 nm 0 nm 40 nm 80 nm σ = 0 was used to emulate

11 Marshal Miller 3-D Simulations: MoSi 3-D FDTD simulations of MoSi line Wafer image calculated with Panoramic (contours shown below) Line ends far more sensitive through focus (up to 5x) Tilt in LES curve apparent in fields imported from FDTD Line end: solid Line width: dashed Defocus - 80 nm - 40 nm 0 nm 40 nm 80 nm σ = 0 was used to emulate plane wave illumination with NA = 1.0 and 2x reduction. IMPACT Lithography 11

EM Simulation Framework Goal: Provide cross-cutting EM simulation capabilities to guide technology innovation in maskless and emerging lithography issues.

plasmonic, lithography and laser anneal New computational capabilities include: Half-cell material boundary placement control Pulsed, surface plasmon, and induced polarization sources

12 EM Simulation Framework Goal: Provide cross-cutting EM simulation capabilities to guide technology innovation in maskless and emerging lithography issues. DARPA/SRC Project and Dan is on FLCC/IMPACT Spring 2008 to assist Litho 2 (EM simulation) Developed TEMPEST v7 to provide computational and physical insight to optoelectronic, plasmonic, lithography and laser anneal New computational capabilities include: Half-cell material boundary placement control Pulsed, surface plasmon, and induced polarization sources Post-processing device under test system Applications include: Dan Ceperley 460 Sub-Wavelength Grating Plasmonic couplers, nanoparticles, and lithographic mask elements Rapid Thermal Annealing PhD 5/08 IMPACT Lithography 12

13 Surface Plasmon Generation Efficiencies Surface plasmon generation by small surface features is surprisingly strong. Small surface features can have an effective width 2x larger than their physical widths. Surface plasmon generation is strongly angle dependent. A 25 shift in illumination angle can double the plasmon conversion length. Engineering the pedestal beneath the feature is critical for optimizing surface plasmon grating couplers. Incident Wave 100nm Feature Incidence Conversion Length Angle Leftward Rightward 0 85nm 85nm nm 9nm 56 7nm 98nm IMPACT Lithography 13

Laser Spike Annealing (Ultratech) Gabor (IBM): circuit performance variation dominated by RTA*. Pattern density dependence confirmed by TEMPEST v7. Gate geometry dependence discovered.

14 Laser Spike Annealing (Ultratech) Gabor (IBM): circuit performance variation dominated by RTA*. Pattern density dependence confirmed by TEMPEST v7. Gate geometry dependence discovered. 50% of power absorbed by nitride capping layer. Simulation challenges include small feature sizes and highangle waves. *I. Ahsan, et al, RTA-Driven Intra-Die Variations in Stage Delay, and Parametric Sensitivities for 65nm Technology, IEEE 2006 Symposium on VLSI Tech, IMPACT Lithography 14 Illumination: Perpendicular to grating Illumination: Parallel to grating Grating orientation Gates Directly on Wafer Reflectio n Heating Power % 100% % 100% % 100% % 100% % 97% 45 11% 89% 60 31% 69% 75 63% 37% 85 78% 22% % 20% 90 80% 20%

Gate Variation Sources and Monitors Goals: Develop fast approximate methods with physically based models to assess and guide decomposition for double patterning and spacer lithography and develop

15 Gate Variation Sources and Monitors Goals: Develop fast approximate methods with physically based models to assess and guide decomposition for double patterning and spacer lithography and develop parameter specific electronic monitors 1443 with Juliet Rubinstein Images in Photoresist for Self-Interferometric Electrical Image Monitors (BACUS 07) {based on double exposure, open/short} Six parameter DOE resist printing => < 0.3 Rayleigh Unit Post-Decomposition Assessment of Double Patterning Layout (SPIE 08) PV-band vs. PM shows proximity and focus are distinct Defocus is not a small aberration => OPD 2 => PM with Z 0, Z 3, Z 9 Pre-OPC PM and Post-OPC PM similar but some differences Found instances where an alternative split improves Z 0,Z 3,Z 9 F-E model; Interpret (IMEC-SPIE) IMPACT Lithography 15

16 Matching Agrees with PV Bands There is a region of necking due to the high match with Z 0. There is a region of high focus variation where the high match for Z 3. Illumination: Clearfield with attenuated features, σ = 0.3, NA = 0.85, 0.04 λrms defocus. Focus Z 0 Juliet Rubinstein IMPACT Lithography 16

17 Matching on Post Decomposition Layout Juliet Rubinstein Match Location Illumination: Darkfield with openings Annular , NA=1.2. Layout supplied by IMEC. IMPACT Lithography 17

Fast-EM Methods for EUV Masks Goal: Extend fast ray-tracing methods for EUV buried defect feature interactions and establish rules-of-thumb and defect compensation

Chris Clifford Fast Three-Dimensional Simulation of Buried EUV Mask Defect Interaction with Absorber Features (BACUS 07) Established fast simulation capability in 3D

18 Fast-EM Methods for EUV Masks Goal: Extend fast ray-tracing methods for EUV buried defect feature interactions and establish rules-of-thumb and defect compensation strategies. This work is supported by Intel and will provide a leg-up for IMPACT in Y3 and Y4. Chris Clifford Fast Three-Dimensional Simulation of Buried EUV Mask Defect Interaction with Absorber Features (BACUS 07) Established fast simulation capability in 3D Smoothing based model for images of isolated buried EUV multilayer defects (SPIE 08) Smoothing process imposes asymptotic surface shape Emphasizes center of lens => PSF broader and less ringing How does this affect defect compensation strategies? IMPACT Lithography 18

19 Integrated Simulation Methodology ourier Transform Reference Plane Transmission through absorber features Reflection from multilayer Thin Mask Model Ray Tracing Absorber Pattern Specifications Defect and Multilayer Specifications Incident Absorber Layout Near Wave FT{ } Simulator Field Multilayer Simulator Plane Wave Near Field Final Result Absorber Layout Simulator Near FT{ Field } IMPACT Lithography 19 Chris Clifford

Defect Printability: Algebraic Model Verification Algebraic model ΔL = I edge m defect h ImageSlope Plug in parameters I m h edge defect = 0.3 SurfaceDefect = 0.191 = 3.

20 Defect Printability: Algebraic Model Verification Algebraic model ΔL = I edge m defect h ImageSlope Plug in parameters I m h edge defect = 0.3 SurfaceDefect = = 3.19 SurfaceDefect ImageSlope = Compare to FDTD simulation ΔL = 7. 08nm A ΔL = 3. 8nm R Δ L = 3. 5 R+ C nm Chris Clifford Distance (nm) TEMPEST (top view) Distance (nm) Results Do Not Match until the absorber coverage is included IMPACT Lithography 20

Probe-Pattern Grating (PPG) Focus Monitor through Scatterometry Jing Xue, Costas Spanos and Andy Neureuther Sensitive, accurate focus metrology needed for high NA lithography Scatterometry: a

21 Probe-Pattern Grating (PPG) Focus Monitor through Scatterometry Jing Xue, Costas Spanos and Andy Neureuther Sensitive, accurate focus metrology needed for high NA lithography Scatterometry: a metrology platform with the significant benefits of being non- destructive, accurate, fast and repeatable Illustration of the high sensitivity of the Probe-Pattern Grating Focus Monitor IMPACT Lithography 21

22 Concept of Interferometric Focus Monitor Inverse mapping Defocus Pupil Defocus Mask 1D defocus OPD 2D defocus mask cross-section IMPACT Lithography 22 Monitor Mask Design Rule

23 PPG Focus Monitor Design Pitch: P Pattern line: W pa 90o Probe line: W pb W pa (a) PPG mask design Imag 0-1 K pb +1 A pa K pa (b) Diffracted Beams at pupil plane K pa A pb K pb A pb K pb A pa K pa Real Fourier Space IMPACT Lithography 23 (c) In Fourier space, the phase and magnitude of the diffracted orders

24 PPG Focus Monitor Design Where S is the sensitivity coefficient, which determines the curvature of the cosine function. K 0 and K 1 are the magnitude of 0 th and 1 st order Illustration of aerial image with defocus IMPACT Lithography 24 Where Linear approximation can be obtained at small defocus

25 Resist Image of PPG Illustration of the probe-pattern line behavior through defocus; The top plots show the resist profiles IMPACT Lithography 25

26 Resist Profile Characterization through Scatterometry Measurement Indices characterization: T_TCD, T_MCD, T_BCD T_HT, P_TCD, P_MTCD, P_MBCD, P_BCD, P_HT, TP_HT Multiple parameter characterizations of Probe-Pattern Grating IMPACT Lithography 26

27 Scatterometry Measurement Scatterometry measurement wafer map and measured spectra IMPACT Lithography 27

28 Scatterometry Measurement Results Trench depth to Defocus slope: ~ 94.4nm/R.U. in the approximately linear range IMPACT Lithography 28

29 Conclusions A significantly high sensitivity of aerial image to defocus is obtained by PPG focus monitor. A linear model can be developed to translate the probe line trench depth into the focus error through scatterometry based ODP techniques. The average slope of the probe trench depth to focus is around 94.4nm / RU, which indicates that the sensitivity of the measurement is around 1.1nm defocus / nm trench depth, and the PPG focus monitor can detect the defocus distance to well under 0.05 Rayleigh Unit. Thanks to Jeffrey Schefske, Kwame Eason, and Phillip Jones in Spansion Inc. for supporting lithography and scatterometry measurements of this project Thanks to TEL/Timbre for supporting the scatterometry simulation tool at UC Berkeley. IMPACT Lithography 29

30 Optical System Characterization through Scatterometry Yu Ben and C. J. Spanos, EECS / UCB Test pattern design Measurements with ellipsometer ψ(λ) Δ(λ) CD SWA Data analysis to extract process parameters IMPACT Lithography 30 Profile extraction

31 Phase-shifted Dual-bar Grating Aberration causes CD difference between the two bars within one period of dual-bar grating In binary dual-bar grating, CD difference is insensitive to even aberrations Solution: 90 degree phase shift is introduced to one of the two bars Linearity is obtained between CD difference and even aberrations (shown in red for spherical aberration Z9) CD Difference (μm) Binary PSM Z9 (Wavelengths) IMPACT Lithography 31

32 System Generation & Verification IMPACT Lithography 32

System Performance / Goals Comparison of the original and extracted Zernike coefficients. The original coefficients are generated randomly to imitate a lithographic system with Strehl ratio of 97%.

33 System Performance / Goals Comparison of the original and extracted Zernike coefficients. The original coefficients are generated randomly to imitate a lithographic system with Strehl ratio of 97%. The extracted coefficients are obtained by using linear model without measurement noise. Experimental Verification Improve system by taking into account dose and focus information Improve the resist model Modify the linear model for better robustness against noise Difference of original and extracted Zernike coefficients in wavelengths. The extraction is done with virtual measurement noise. 50 different aberration settings are randomly generated to test the system. Normally distributed random noise with 3σ = 0.9 nm is introduced. The virtual measurement result is repeated 20 times to average out the noisy result. IMPACT Lithography 33

34 The Calculus of Clips Poolla, Neureuther, Spanos + TBD student Basic Assertion: Working with clips is more efficient and natural than distances/rules Potential opportunities for clip calculus Faster printability analysis Hot-spot detection and repair Mask fragmentation for multiple-patterning Clips Could be non-rectangular Standard cells, macros, etc Core Central part of a clip Context Outer part of a clip Library Collection of clips Core & Context depend on target application Ex: DRC, OPC, Printability analysis OPC re-use Faster DRC Faster RCX context clip mask The problem: algorithms to efficiently deal with clips core IMPACT Lithography 34

35 Ex: DRC Current Practice DRC brick is 2,000 pages and exploding, Conventional rule-based DRC at 22nm will be unmanageable Alternatives: Work with clips not distances Leverage the speed of pattern match Produce library of good, bad, or graded patterns Use library to detect and correct new design layouts Open problems Clip-based DRC, Hybrid Rule-Clip DRC, Redundancy removal in rules, Correction! Core and Context Core is the region that is DRC clean given the fixed Context Context may not be DRC clean as that depends on Context(Context) Use Case DRC Clean core Library of known DRC clean (in core) clips In a mask M, use PatternMatch against library L Can eliminate the core of every matched clip Context Will have to do DRC on remaining areas IMPACT Lithography 35

36 Clip Metrics What is a good metric on the space of clips? Difficult problem Must also extend to Alternating PSM, Attenuating PSM Metric must also be computable in the language of rectangles ci, c = # of Standard pixel based metrics fail: j Treats each pixel independently Does not respect proximity When are two clips similar? If the images in Silicon of the core of both clips are similar Litho model Clips Projection onto core agreeing pixels Suggests that we need application dependent weightings Exposure & Dose sensitivity analysis will have different weights IMPACT Lithography 36

37 Some Computational Problems mask M, clip c, library L, N = number of clips 1. ExactMatch: Find all instances of c in M [done] 2. RoughMatch: Find all sub-patterns p in M with d( p i c) pε 3. VolFind: Find Area(core union) 4. WhereNext: Find largest rectangle not covered by clips 5. ExactTile: Tile M with core of clips drawn from library i.e. choose tiling to maximize Area(core union) 6. RoughTile: Tile M with core of clips drawn from library such that context of mask sub-pattern p and library clip c are close: d( p i c) pε New Result For 3 & 4 we have N log(n) time algorithms with N log(n) pre-processing time Based on planar point location methods IMPACT Lithography 37

38 Some computational problems There are other very interesting problems Net-list covering by clips Library generation These are all problems in CS, with a twist Must compute based on rectangles Must respect hierarchy Must work Must work on huge problems Algorithm Classes Randomized Adaptive 2008 Goals Devise and analyze algorithms for basic clip operations Test clip-based DRC on modest layout for speed-up IMPACT Lithography 38

39 Collaborative Verification Discussion Sandboxes Berkeley Microlab BWRC DATA Industry Data SVTC via ASML Industry tools Industry CMOS SVTC via ASML ST Micro via BWRC IBM? Foundry via Marvell? Investigations Collab. Platform DfM Tool SEM impact to device Cell-to-Cell Interactions Double Patterning ODP Aberrations (May 08) Hyper-sensitive layouts RO for focus (May or Sep 08) IMPACT Lithography 39

40 Future Milestones Litho 1: Process effect monitoring and full chip assessment Lynn Wang (Gate): Interpret (BWRC, TI-SPIE); Litho SEM to electrical; RO concepts and CMOS layouts Eric Chin (Interconnect): DP issues; F-E model; layouts Litho 2: Electromagnetics Marshal Miller: Dry Lab monitors for ATT-PSM and plasmons Litho 3: Fast-CAD for double patterning Juliet Rubinstein: Z 0,Z 3,Z 9 F-E model; Interpret (IMEC-SPIE) Litho 4: Assessment tools for double patterning decomposition TBD: Explore application dependent clip weightings NT 4: ODP-based parameter extraction and silicon verification Yi Ben: Sim, Layouts and Exp. Prg. Aberrations NA 0.85, 1.35 IMPACT Lithography 40

2008 IMPACT Workshop. Faculty Presentation: Lithography. By Andy Neureuther, Costas Spanos, Kameshwar Poolla, EECS and ME, UC Berkeley

2008 IMPACT Workshop. Faculty Presentation: Lithography. By Andy Neureuther, Costas Spanos, Kameshwar Poolla, EECS and ME, UC Berkeley 2008 IMPACT Workshop Faculty Presentation: Lithography By Andy Neureuther, Costas Spanos, Kameshwar Poolla, EECS and ME, UC Berkeley IMPACT Lithography 1 Current Milestones Litho 1: Develop and experimentally