Feature-level Compensation & Control. Sensors and Control September 15, 2005 A UC Discovery Project

Transcription

1 Feature-level Compensation & Control Sensors and Control September 15, 2005 A UC Discovery Project

2 2 Current Milestones Integrated sensor platform development 2 (M26 YII.16) Gather CMP and etching rate data and correlate with process variables. Complete preliminary experimental study for CD non-uniformity reducing across the litho-etch sequence (M27 YII.17) Assess predictive capability of mode, and build optimizing software to compute optimal changes in control parameters. Provide proof of concept test of CD non-uniformity reduction scheme based on direct CD metrology. Zero-footprint Optical Metrology Wafer (Milestone Added, YII.18) Evaluate and calibrate dielectric thickness monitoring (resolution, sensitivity, and stability). Metal etch endpoint and pre-endpoint (<50nm) detection and monitoring. Testing the prototype metrology wafer in vacuum environment. Using Spatial CD Correlation in IC Design (M30 Major Rev., YII.19) Initial experiments on test structures and measurement for extracting spatial correlation characteristics. Aerial Image Metrology (M31 YII.20) Integrate prototype transducer for use and deployment on a silicon wafer.

3 3 Zero-Footprint Optical Metrology Wafer Student(s): Vorrada Loryuenyong and ZhongSheng Luo* PI. Professor Nathan Cheung * Currently at KLA-Tencor Prototype and develop methodology for in-situ process monitoring with zero-footprint metrology wafer.

4 Main Objectives Evaluate and calibrate dielectric thickness monitoring (resolution, sensitivity, and stability). [Resolution and sensitivity analysis completed] Testing the prototype metrology wafer in vacuum environment. [Completed] Metal etch endpoint and pre-endpoint (<50nm) detection and monitoring. [In progress]

5 5 3x3 pixel Zero-Footprint Metrology Wafer RPD LED Detecting Window PPD Top Wafer Backside Contact Via Sketch of the cross-section of the prototype. Bottom Wafer

6 6 Pixel-to-Point Transfer 1. Pick-up of the pixel 2. Release of the growth substrate:laser Liftoff 3. Registration of the LED pixel to the target substrate Pick-up Rod Adhesive 1 Pd-In Sapphire Laser beam LED 4. Selective removal of the pick-up rod Target Substrate

7 Film to be grown/etched R R θ i Reference Photo- Detector (RPD) PPD reading (a.u.) R P Air (Slope= ) Water (Slope= ) P.R. (Slope= ) 7 Methodology Dielectric Window LED Primary Photo- Detector (PPD) RPD reading (a.u.) R F F = P ( δv R R P 0 P 0 P / δv R ( δv ) ( δv P / δv A new function F is defined to eliminate non-measurable constants: Errors due to misalignment of optical components and Variation in the detector circuits, photo-detectors, and light intensity F function depends on incident angle, wavelength, refractive index and thickness. R P ) / δv 0 R ) 0

8 Calibration of the Prototype with a Plasma Etch Process of Silicon Oxide Experimental Fitted λ (nm) Measured 463 (peak) Extracted n.a θ i ( ) n.a. 56 F(θ) -0.4 d w (nm) 650 ± (n w, k w ) (n f, k f ) (2.054,0)* (1.464,0)* n.a. n.a Oxide Thickness (nm) *D.L. Windt, IMD Software. The good fit between experimental data and calculation demonstrated that the methodology worked as expected. As expected, effective incident angle, detection window thickness and even effective incident wavelength can be determined by a calibration process.

9 F(θ) F as a function of incident angle at different refractive index n=1.0, k=0 n=1.3, k=0 n=1.6, k=0 n=1.9, k=0 n=2.2, k=0 n=2.5, k=0 Simulation Condition: Vacuum ambient, infinity thickness for thin film, nitride window thickness 649nm, LED peak wavelength 463nm. *D.L. Windt, IMD Software Incident Angle (Degree) F is a function of refractive index and incident angle. The effective incident angle can be precisely determined by using media with tunable refractive index.

10 10 Dependence of F function on incident wavelength (λ) and window thickness F(θ) λ=463nm d w =659nm λ=468nm d w =649nm λ=463nm d w =639nm λ=463nm d w =649nm Simulation Condition: Vacuum ambient, and incident angle of 60, nitride window thickness, LED peak wavelength 463nm. *D.L. Windt, IMD Software λ=458nm d w =649nm Oxide Thicknesss (nm) Both wavelength and detecting window thickness have similar effects on the F function 10 nm change in detecting window thickness = 5 nm change in incident wavelength

11 Polishing fixture Metal layer e.g. Cu 11 Metal CMP Endpoint Detection Setup Polishing pad Data Acquisition System Simulation Condition: Vacuum ambient, nitride window thickness 649nm, LED peak wavelength 463nm, the refractive index of Cu*: n=1.16,k=2.43. Slurry F(θ) Detection Window Metrology wafer 80Deg 70Deg 40Deg 60Deg 50Deg *D.L. Windt, IMD Software Cu Thickness (nm)

12 12 Work in Progress and Proposed Work Evaluate and calibrate the stability of dielectric thickness monitoring. Demonstrate Metal etch endpoint and preendpoint (<50nm) detection and monitoring. Model and demonstrate monitoring of thin-film thickness roughness. Prototyping of wireless data acquisition/transmission and evaluate performance with measurements taken in processing systems.

13 Workshop Student(s): Jing Xue Faculty: Costas Spanos Title: Integrated Aerial Image Sensor (IAIS)

14 light Mask Image system 14 Motivation Defocus Lens aberration Partial Coherence Magnification CD Uniformity Aerial Image Sensor Aerial image Latent image Resist image Wafer Sensor on equipment

15 Main Objective Complete design of transducer capable of nm-scale aerial image resolution Integrate prototype transducer for use and deployment on a silicon wafer Complete the micro-assembly of the commercial CCD with the Si carrier wafer; Integrate the aperture mask and the CCD arrays Complete the IAIS working prototype with front-illu. CCD, and test IAIS in GCAWS/ASML stepper in Berkeley Micro-lab Complete the aerial image and detector image reconstruction; Complete the over-topography simulation of the aberration part

16 16 Integrated Aerial Image Sensor (IAIS) Concept High spatial frequency aerial image ( m + n) P + x x Aperture mask transmission Low spatial frequency detector signal

17 17 IASI Design Aperture Mask nm 30nm 50nm 90nm 70nm nm 90nm 70nm 60nm m n width(nm ) Max & min intensity vs. aperture thickness and width width(nm ) C DIC vs. aperture thickness and width contrast width(nm) Aperture mask thickness in the range of 70nm & aperture mask width in the range of 30nm contrast m ax current m in current 0.90 detector noise width(nm ) photocurrent(pa)

18 18 Summary of Aperture Mask Design for 130nm Periodic AI (65nm nodes) w a W t W i Wd W g l α -Si t w a = 30 nm n = 154 t = 70 nm l = 20 µ m m N = 307 = 21 = x 5 nm W W 3.1µ m = P = 130nm = µ m d mp i d = ( m + 1) P + x W W t w g a = µ m = µ m

19 19 IAIS Modeling Aerial Image and Detector Image Reconstruction: 88 coherence groups Aerial Image Detector Image Annular Illumination: σ = 0.89/0.59 NA=0.85, PSM, CD = 65nm

20 20 Defocus Testing: IAIS Modeling TM TM TE TE (a) (c) (b) (d) Dipole Illumination: σ = 0.3, NA= 0.78, Attenuated PSM, CD =90nm (a) Illumination discretizing ( 2 points illustration); (b) Aerial image intensity with defocus (c) Detector Image intensity with defocus; (d) Integrated intensity of detector Image vs. defocus

21 IAIS Modeling Defocus Testing (Focus vs. Contrast): Aerial Im age Contrast vs. Defocus Detector Im age Contrast vs. Defocus Contrast Image Contrast Image Contrast Defocus (µ m) Defocus Contrast Change % % % % % C o n t r a s t C h a n g e v s. D e f o c u s D e t e c t o r i m a g e c o n t r a s t C h a n g e = A e r i a l i m a g e c o n t r a s t Dipole Illumination: s = 0.3, NA= 0.78, Attenuated PSM, CD =90nm Aperture mask: w a = 40nm, t = 90nm 5 0 % D e f o c u s ( µ m ) Aperture mask improve the contrast value as defocus, making the aberration detecting easy and meaningful

22 Wafer Reconstituting: 22 IAIS Assembly Si Substrate SiO 2 CCD Polymer ARC Dyed PMMA

23 Flip-Chip Bonding: 23 IAIS Assembly Solder bump Wire bonding pad on the CCD chip Si Substrate Amorphous Si SiO 2 CCD chip Flip-chip Bonding

24 Workshop Student(s): Paul Friedberg, Willy Cheung Faculty: Costas J. Spanos Title: Modeling Gate Length Spatial Variation for Process/Design Co-Optimization

25 25 Motivation Manufacturing-induced variation in device parameters leads to variability in circuit performance Two approaches to address this concern: Tailor IC design to minimize sensitivity to parameter variation Use process control to reduce manufacturing variation Both approaches can be investigated through Monte Carlo analysis of canonical circuits Various design styles can tested for susceptibility to variation Hypothetical control scenarios can be mapped directly into circuit performance space to determine robustness For accurate, useful predictions, Monte Carlo framework must model reality very well Specific focus of this work: spatial variation effects (correlation)

26 Main Objective Milestone M30: Spatial CD Correlation in IC Design Extract within-die spatial variation components from dense gate length measurements (historical study) Investigate effects of spatial variation on circuit performance variability using Monte Carlo framework based on historical study results Design new test structures to explore mid-range ( micron) spatial variability Submit new test structures for manufacture; gather measurements from fabricated test structures

27 27 Departure Point: Spatial Correlation Calculation Exhaustive ELM poly-cd measurements (280/field): Standardize each CD measurement, using wafer-wide distribution: z i = ( x x) / σ i For each spatial separation considered, calculate correlation r among all within-field pairs of points using: ( ) jk = z j * z k / n r ELM data provided by Jason Cain, UC Berkeley

28 Spatial Correlation Results Within-field correlation vs. horizontal/vertical distance, evaluated for entire wafer: 28 Shape of correlation curve is confounded by non-stationary (systematic) components of variation

29 29 Decomposition of Nonstationary Variation Components CD variation can be thought of as nested systematic variations about a true mean: CD ij = µ + f i + w j + σ Spatial components True mean Across-field Across-wafer Random Wafer Field

30 30 WIF Systematic Variation Component Within-field variation: - = Slit Scan Average Field Scaled Mask Errors Non-mask related acrossfield systematic variation Polynomial model of across-field systematic variation Removing this component of variation will simulate WIF process control

31 31 AW Systematic Variation Component Across-wafer variation: - - Average Wafer Scaled Mask Errors Across-Field Systematic Variation = Across-Wafer Systematic Variation Polynomial Model Removing this component of variation will simulate AW process control

32 32 Die-to-Die Dose Control One more round of control: die-to-die (D2D) dose control - =

33 33 Simulated Full Process Control Removing WIF, AW, and D2D variation components: Large(mm)-scale spatial correlation is largely accounted for by systematic variation; smaller(µm)-scale correlation may still have structure, to be investigated in future work

34 34 Test Structure for Mid-Range CD Variation 2x10 Probe frame: 100um x 100um pads, 150um pitch Dense ELM base case test structure:

35 Variant ELM Submodules Dummy lines used to extend measureable range, explore effects of pattern density and regularity 35

36 Workshop Student(s): Qiaolin(Charlie) Zhang Faculty: Kameshwar Poolla, Costas Spanos Title: CD Uniformity Control Across Litho-etch Sequence

37 37 Motivation Across-wafer CD uniformity (CDU) is critical for Advanced logic devices, MPU and memory Yield improvement Etch tool sets have limited control authority to address spatial non-uniformity Dual-zone He chuck is often the only knob Litho tool sets have much more control authority to address spatial non-uniformity Multi-zone PEB bake plate Variable dose settings at exposure

38 Main Objectives Build process models for PEB step: (done) CD offset model & temperature offset model Assess potential DI & FI CDU improvement (done) Based on CD offset model Based on temperature offset model Expand CDU control concept to simultaneous CDU control for multiple CD targets (new) Experimentally extract baseline CD signature of dense, iso and semi-iso CD targets Formulate simultaneous CDU control as a minimax optimization problem Experimentally verify DI & FI CDU improvement using our approach (ongoing)

39 39 The Problem Wafer Litho Etch Processing Tool Poor Across-Wafer CD Uniformity How can we improve the across-wafer CDU?

40 40 Our Approach Compensate for systematic across-wafer CD variation sources across the litho-etch sequence using all available control authority : Exposure step: die to die dose PEB step: temperature of multi-zone bake plate Etch: backside pressure of dual-zone He chuck Exposure PEB / Develop Etch dose temperature He pressure Optimizer Wafer-level CD Metrology Scatterometry/CDSEM

41 41 Multi-zone PEB Bake Plate PEB step is critical due to chemically amplified resist Spatially programmable bake plate is introduced into PEB to enable PEB temperature uniformity Schematic setup of multi-zone bake plate (approximate) 2 4

42 42 Develop Inspection (DI) CDU Control DI CD is a function of zone offsets T T1 g1 = =... T m g m CD DI T = = T T ( O, O... O ) 1... ( ) O1, O2... O7 S T resist baseline CD baseline CD DI CD =... CD 1 n f = f Seen as a constrained nonlinear programming problem 1 n ( O, O... O ) ( ) O1, O2... O7 CDDI CD t Minimize arget 2 Low Up Subject to: O O O i i = 1,2...7

43 43 Simulation Results of DI CDU Control Dense Line Semi-isolated Line Isolated Line Experimentally extracted baseline DI CDU Simulated optimal DI CDU after applying PEB tuning Dense Line Semi-isolated Line Isolated Line CDU Improvement 72% 61% 69%

44 44 Final Inspection (FI) CDU Control Across-wafer FI CD is function of zone offsets Plasma etch signature: CD p _ s = CD FI CD DI Assumed bowl shape plasma etch signature CD FI = CD DI + CD p _ s = g g 1 n ( O 1 ( O 1, O... 2, O 2... O 7... O 7 ) ) Minimize: CDFI CD t Low Up Subject to: O Oi O i = 1, arget 2

45 45 FI CDU Control Simulation - Bowl Plasma Signature Simulated baseline FI CD Dense Semi-isolated Isolated Simulated corrected DI CD after PEB tuning Simulated optimal FI CD after PEB tuning Note that DI CDU may actually worsen! Dense Semi-isolated FI CDU Improvement 68% 57% Isolated 65%

46 46 Simultaneous CDU Control for Multiple CD Targets It is good to have simultaneous CDU control for multiple CD targets Formulated as a minimax optimization problem Minimax finds optimal offsets O opt = arg min (max( Wi F O i i ( O))) F i = CD i CD _ T i 2 W i is the weighting factor for CD target i

47 47 Simultaneous CDU Control for Multiple CD Targets Simulation of simultaneous CDU control for dense, semi-iso and iso lines Wd =0.36; Ws =0.33 ; Wi =0.31 Wd = 0.90; Ws =0.05 ; Wi =0.05 Wd = 0.05; Ws =0.90 ; Wi =0.05 Wd = 0.05; Ws =0.05 ; Wi =0.90 Simulated baseline FI CD Dense Line 62.9% 66.8% 48.2% 60.1% Semi-iso Line 44.7% 15.9% 54.7% 32.4% Iso Line 58.4% 61.8% 54.1% 62.6% Dense Semi-isolated Isolated Simulated optimal FI CD after PEB tuning

48 48 Future Milestones (Year 3) Zero-footprint Optical Metrology Wafer (SENS Y3.1) Modeling and demonstration of metrology wafer for detection and thin-film roughness monitoring. Initiate prototyping of wireless data acquisition/transmission and evaluate performance with measurements made in experimental systems. Complete experimental study for CD non-uniformity reducing across the litho-etch sequence (SENS Y3.2) Experimentally verify DI & FI CDU improvement using model based optimal control of PEB with various CD objective functions. Using Spatial CD Correlation in IC Design (SENS Y3.3) Perform spatial variation analysis and incorporate results into Monte Carlo framework. Evaluate impact of updated variation/correlation models on circuit performance variability using Monte Carlo Framework. Aerial Image Metrology (SENS Y3.4) Complete the micro-assembly of the commercial CCD with the Si carrier wafer. Integrate the aperture mask and the CCD arrays.

49 49 Future Milestones (Year 4) Integrated sensor platform development 4 (M54) Incorporate optical spectroscopy capability with optical filters integration. Final phase of CD uniformity control project (M55) Complete study of feed-forward and feedback based schemes for process/equipment control to enhance feature level pattern transfer. Study various control architectures in terms of sensor integration, implementation cost, and expected benefit. Real time feature-level test structures (M58) Develop feature-level test structures that can be monitored for real-time insight in their evolution. Examples include real-time etch-rate monitors that are subject to micro-loading.