Metrology for Characterization of Wafer Thickness Uniformity During 3D-IC Processing SEMATECH Workshop on 3D Interconnect Metrology Chris Lee July 11, 2012
Outline Introduction Motivation For New Metrology Tools Experimental Results Summary & Conclusions Acknowledgements 2
Motivation: Wafer applications have demanding quality specifications IC Substrates Circuit feature size is proportional to depth of focus Decreasing depth of focus leads to much tighter wafer flatness requirements to avoid circuit failure To deliver flat wafers, you must be able to accurately characterize them Carrier Wafers Carrier Bond Wafer Thinned & Processed Wafer De-bond Wafer 3
Motivation: Wafer applications have demanding quality specifications Demand for very low TTV wafers and carriers requires precise full surface measurement of TTV and thickness FlatMaster MSP has demonstrated capabilities: - Nanometer scale reproducibility on TTV and thickness of transparent materials - Simultaneous Bow/Warp and substrate TTV/Thickness measurement on full surface 4
FlatMaster MSP Measurement Principle Tunable Laser Source 5
Transmission Intensity Transmission Intensity FlatMaster MSP Measurement Principle Tunable Laser Source Both diode cavity and external cavity have longitudinal laser modes c 2L External Cavity Diode Cavity Desirable to tune the length of the external cavity to mode match the diode cavity Mode-matching optimizes frequency stability, guarantees equal frequency steps, and optimizes fringe contrast Frequency Frequency 6
tio Mode # FlatMaster MSP Measurement Principle Frequency Scanning Interferometry Laser has a step-like tuning behavior - Eliminates frequency errors caused by positioning errors - Allows fast yet precise tuning 8070 LightGage TLD Tuning Characteristics 8068 8066 8064 8062 8060 8058 8056 8054-0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 P (mm) 0.5 7
FlatMaster MSP Measurement Principle Different surface pairs modulate at different frequencies as we step through the different wavelengths of the laser Use frequency domain information to isolate these surface pairs and construct the thickness map and the Bow and Warp data on the entire wafer surface Front surface to Fizeau Thickness fringes Back surface to Fizeau 8
Motivation: Historical silicon methods do not easily transfer to large, thin wafers Metrology strategies have evolved from methods used to characterize smaller, lower aspect ratio geometries Large, thin wafers have inherently low relative stiffness, leading to large gravity induced deflections Conventionally, 3-point mounts have been used to measure flatness/warp of wafer along with the gravity compensation Calibration techniques Multiple measurements (Side A/B) As gravity effect becomes larger, accurate compensation becomes more complex Corning s Flatmaster MSP 300 provides new methods to overcome the shortfalls of traditional methods 9
Wafers are mounted in several ways 3-point mounts are the only kinematic configuration, however they will also have very large deflections and can be sensitive to part positioning 4-point and ring supports have redundant support and can not be deterministically compensated for Once you depart from kinematic, more support points will result in less gravity induced shape. Wire support minimizes deflection due to gravity 3-Ball Support at 0.7r 3-Ball Support at 1R 4-Ball Support at 1R Ring Support Perimeter 10
Motivation: Current approaches have limited data coverage Most traditional methods generate flatness map from multiple profiles generated by scanning a single point probe Scan NW Scan Y Scan X Scan SW This map is generated by 4 scans lots of interpolation 11
Motivation: Limitations of current metrology techniques highlighted in SEMI standards Differences in diameter, thickness, fiducials, or crystal orientation from wafer used for gravitation compensation procedure, may yield incorrect results. Different methods of implementing gravitational compensation give different results. Different geometric configurations of wafer support (e.g. 3-point, 4-point, ring support, etc.) will yield different results. The quantity of data points and their spacing may affect the measurement results. Results obtained with different data point spacing using the same tests may be different. TTV and warp are determined using partial scan patterns; thus, the entire surface is not sampled and use of another scan pattern may not yield the same results. Certain test methods do not completely separate TTV from warp. Running probes off the test specimen during the scan sequence gives false readings. SEMI MF1390-0707 SEMI MF657-0707 E 12
Case Study: 3-Point Mount Wafer material: Si Density: 2.33 g/cm 3 Elastic modulus: 141 GPa Poisson s ratio: 0.22 Wafer thickness: 0.7mm +/- 0.01mm Glass Material: Corning SGW3 Density: 2.38 g/cm 3 Elastic modulus: 74 GPa Poisson s ratio: 0.23 Wafer thickness: 0.7mm +/- 0.01mm Material Diameter (mm) Thickness (mm) Support Radius (mm) Sag (um) Si 50 0.40 22 0.35 Si 300 0.69 147 212 Si 300 0.70 147 206 Si 300 0.71 147 200 Si 450 0.70 222 1060 SGW3 300 0.70 147 404 13
New Approach To Measure High Aspect Ratio Wafers Frequency stepping interferometer avoids limitations of standard phase measuring interferometers - Avoids 2π ambiguity from phase measuring interferometry Wire support approach minimizes gravity effects Simultaneously measure flatness and TTV of glass wafer (<1minute/wafer) Full aperture interferometric approach gives sub-millimeter lateral resolution (~3,000,000 data points /300 mm wafer) Wire Support 14
FlatMaster MSP specifications 300 mm system Field of View (Circular) Z-Resolution Lateral Resolution Measurement Time Measurement Method Data Points Repeatability > 300 mm 10 nm (0.40 μinch) 0.15 mm (0.006 inch) 30 seconds typical Frequency Scanning Interferometry 2kx2k up to 4,000,000/measurement 25 nm 15
Peak-to-Valley: Deflection (um) from FEA Deflection due to gravity (prior to compensation) Material (thickness) 3-Pt. perimeter (0.7 mm) 3-Pt. 0.7 radius (0.7 mm) Wire Support (0.7 mm) Si 217 61 0.6 SGW3 422 121 0.6 Techniques that use calculation/calibration to remove gravity effects assume many consistencies which may affect data (material, thickness etc.) Wire support minimizes gravity effects, which eliminates this complication, while eliminating dead zones where the surface is not visible due to part support structures 16
Flatmaster MSP output Data maps showing (a) TTV (~1.4 um) and (b) flatness (~17 um) of a glass wafer with sub-millimeter lateral resolution. Each data point shown is representative of an actual data point collected There is no compensation from gravity effects flatness is measured as it sits on the wire mount. (a) TTV 1.4 um (b) Flatness 17 um 17
Flatness (um) Consistency of 3-point & wire support Evaluate repeatability from mount techniques on MSP 80 70 60 50 40 30 20 10 0-10 Warp - 3-Pt Bow - 3-Pt Warp - Wire Bow - Wire 1 2 3 4 5 6 7 8 9 10 Iteration 200 mm diameter glass wafers Measure same part 10x with 2 mount techniques (3-point at 0.7R, wire support) 3-point support creates larger warp and standard deviations compared to the wire support 10 um variation with 3-point for the same part Compensations strategies do not account for this error 18
Thickness (um) TTV (um) Flatmaster MSP Thickness, TTV: Highly Repeatable 701 1.25 700.8 1.23 700.6 1.21 700.4 700.2 700 Glass wafer measured 10X Thickness - 3-Pt Thickness - Wire TTV - 3 Pt TTV - Wire 1 2 3 4 5 6 7 8 9 10 Iteration Wire support method gives thickness repeatability better than 0.03 µm & TTV repeatability < 0.003 µm. Large deflection from 3-point mount affects repeatability of thickness and TTV measurements, but it is still quite good at 0.1µm and 0.01µm respectively 1.19 1.17 1.15 19
Glass TTV & wafer stack TTV Data taken using 3M WSS process Corning wafers (SGW3) of specified TTV (low/high) used with 3M WSS to study effect of wafer TTV on bonded stack TTV Data is highly correlated - low glass TTV gives low wafer stack TTV as expected Glass wafers from other established Japanese wafer supplier reporting TTV < 1µm (5 points/wafer) Actual thickness variation >1µm which will clearly impact bonded stack TTV, and therefore performance New demanding applications require accuracy and data density provided by MSP 20
Summary: Current Challenges Large wafer diameter presents challenges to existing techniques Very low resolution to build 3D maps from 4 scans Many assumptions are required Typically use a calibration step, or side A/B measurement and analysis that require excellent repeatability and geometric consistency Thickness, material, exact mount support (and R&R) significantly impact gravity compensation techniques Large sag due to gravity for conventional mounting techniques lends itself to large errors Limited number of data points can lead to low estimation of total warp/ttv Leads to poor understanding of wafer quality and impact on subsequent processes 21
Summary: Flatmaster MSP Utilizes mount that has 2-3 orders of magnitude lower gravity deflection Leverages technology for ultra-precise photo mask characterization Avoids gravity sag vs. accounting for it - eliminates the need for significant assumptions Data shows significantly better measurement repeatability using wire support over 3-point mount Correlation of wafer TTV and stack TTV demonstrates MSP value Actual data points every <1 mm gives high fidelity data with little/no prior assumptions on the part shape Data density invaluable to characterize Si wafers and carrier wafers You must be able to see the error to achieve required performance Substantial opportunity to use data for process enhancements (IC manufacture (silicon wafer flatness), glass wafer flatness/ttv, carrier wafer recycling) 22
Acknowledgements The authors acknowledge colleagues: Bor Kai Wang for measurement and setup, Matt Cempa and Keith Hanford for numerical modeling, and substantial support provided by the Corning Advanced Technology Center (CATC) in Taipei, Taiwan. 23
Thank You! 24