imaps presentation
Download as PPT, PDF1 like831 views
This document discusses using X-ray fluorescence (XRF) as a non-invasive method to measure the percentage of silver (%Ag) in tin-silver (SnAg) solders used in flip chip applications. It describes how XRF was calibrated and optimized through comparisons with invasive techniques like inductively coupled plasma (ICP) and differential scanning calorimetry (DSC). Factors that impacted XRF measurement accuracy included solder bump geometry, collection time, and background noise. Flattening solder bumps improved XRF reliability and consistency. With proper calibration and accounting for limitations, XRF was found to be an excellent non-invasive method for measuring solder composition variability across wafers, chips and individual solder
imaps presentation
- 1. © 2009 IBM Corporation Assessment of XRF Technique as a Method to Measure Percent Ag in SnAg Solders for Flip Chip Applications Jennifer D Schulera Chia-Hsin Shihb , Charles L Arvina , KyungMoon Kimc , Eric Perfectoa a: Semiconductor Research and Development Center, IBM, Hopewell Junction, New York 12533 b: STATS ChipPAC Taiwan Ltd, Hsin-Chu Hsien Taiwan, R.O.C 307 c: STATS ChipPAC Korea Ltd, Kyoungki-do, 467-814 South Korea
- 2. © 2009 IBM Corporation 2 Introduction X-Ray Fluorescence (XRF) is a non-invasive method of measuring Ag% composition in Pb-free solders. 1. Sample preparation - Experimental bumping variables 2. Tool configuration - XRF configuration, calibration, optimized measuring methodology and the importance of having known standards with the same dimensions of the bumps being measured 3. XRF recipe setting - Measuring accuracy and correlation with ICP and DSC 4. Data interpretation - Ag distribution study in the die and wafer level
- 3. © 2009 IBM Corporation 3 Why XRF? One area of interest for Pb-free solder manufacturing is the ability to control and measure the %Ag composition and its variation from wafer to wafer, chip to chip, and C4 to C4. Invasive Non-Invasive Atomic Absorption (AA) X-Ray Fluorescence (XRF) Differential Scanning Calorimetry (DSC) Inductively Coupled Plasma (ICP) Electron Probe Micro- Analyzer (EPMA) Methods to Measure Solder CompositionWafer to Wafer Chip to Chip C4 to C4 Ag% Composition Control Pb-free SnAg solder has become the industry standard for fabricating flip chip interconnects utilizing C4 (controlled collapse chip connection) technology.
- 4. © 2009 IBM Corporation 4 BumpVariablesWaferVariables Underlying Metallurgy Bottom Layer Metallurgy (BLM) Size UBM Stack C4 Height Ag% Target Mechanically Good (MG) Electrically Good (EG) 90 um 110 um Thick Cu No Cu 70 um 90 um 0.6% 1.7% Sample Preparation – Test Vehicle List VariablesImpactingXRFMeasurement
- 5. © 2009 IBM Corporation 5 Tool Configuration # BLM Size C4 Height Type Known Bump Composition UBM Stack SnAg Solder Process A 90 70 MG 0.5%Ag Ni/SnAg C4NP B 90 70 MG 1.8%Ag Ni/SnAg C4NP C 110 90 MG 0.5%Ag Ni/SnAg C4NP D 110 90 MG 1.8%Ag Ni/SnAg C4NP C4NP is a solder transfer process which controls the Ag% composition to ± 0.1%Ag. XRF Calibration Standards Tool Description XRF Anode material is made of tungsten or molybdenum ICP-OES Invasive method to get C4 composition on global zone DSC Invasive method to get C4 composition on global zone Verification Methods Through data comparison with ICP and DSC, a non-invasive XRF method was created and calibrated.
- 6. © 2009 IBM Corporation 6 Mass Transport Effect The need for the correct geometry comes from the learning that the %Ag is mass transport controlled. The concentration of Ag in a larger test structure on the edge of a wafer may have a different and most likely higher concentration of Ag than the Ag in an actual C4. C Site A has a largest height controlled by diffusion plating compared to B or C Site A has the lowest Ag% Diffusion Controlled Plating Convection Controlled Plating 10um Site C has the smallest height controlled by diffusion plating. Compared to A or B Site C has the highest Ag%
- 7. © 2009 IBM Corporation 7 Optimized collection time creates a stronger XRF feedback generating more stable readings. From C. Shih XRF Performance by Collection Time
- 8. © 2009 IBM Corporation 8 Process time 180sec on C4NP Standards 20 XRF Readings on Randomly C4s # BLM Size C4 Hei ght Ag% Target Mean Std Dev Range A 90 70 0.5% 0.66% 0.44% 1.84% B 90 70 1.8% 1.77% 0.16% 0.53% C 110 90 0.5% 0.48% 0.06% 0.19% D 110 90 1.8% 1.79% 0.05% 0.21% Remeasure on flattened C4NP Standards Process time 180sec on flattened C4NP standards 20 XRF Readings on Randomly C4s # BLM Size C4 Hei ght Ag% Target Mean Std Dev Range A 90 70 0.5% 0.50% 0.05% 0.20% B 90 70 1.8% 1.89% 0.06% 0.27% XRF inspection on C4NP Standards Tool issues such as stage movement accuracy and laser alignment were investigated and then eliminated as possible causes. The X-ray spot size is 40µm -> a C4 geometry issue was hypothesized. Reliability dramatically improved after the flattening, proving that C4 geometry impacts the intensity of the signal. Flattening is recommended if the bump height is less than 70um What caused the instability in A and B? Bump Geometry
- 9. © 2009 IBM Corporation 9 FlattenedUn-Flattened SEM photos comparing flattened and un-flattened C4’s The spot size of the X-ray opening is 40 µm. The detector is collecting photoelectrons from any region that is excited by the X-rays. Flattening reduces the chance of exciting larger regions which would cause artificially higher readings and thus a larger standard deviation. From S. McLaughlin
- 10. © 2009 IBM Corporation 10 The spectrum data shows a stronger feedback of X-ray fluorescence radiation on larger C4s. Ag% Un-reflowed Bump From C. Shih 90um, 1.8% 70um, 1.8%
- 11. © 2009 IBM Corporation 11 From C. Shih Wafer 1 -thick Cu Wafer 2 -thick Cu Wafer 3 -thin Cu Comparison Post Reflow
- 12. © 2009 IBM Corporation 12 From C. Shih Wafer 1 -thick Cu Wafer 2 -thick Cu Comparison Post Flattening Wafer 3 -thin Cu
- 13. © 2009 IBM Corporation 13 XRF readings on MG wafers are comparable with ICP data (within 0.2%Ag), but the XRF readings are not reliable on EG wafers due to high background noise To alleviate this, the XRF collection time was shortened to mitigate primary X-ray penetration. Collection Time on EG Wafers From C. Shih EG/MG Spectrum Comparison
- 14. © 2009 IBM Corporation 14 From C. Shih EG spectrum pre-flattening EG spectrum post-flattening
- 15. © 2009 IBM Corporation 15 Conclusion XRF is a popular, non-invasive inspection method for bump composition. Several challenges were encountered when establishing this technique. Balancing X-ray power voltage and current settings to obtain suitable X-ray penetrating ability. Mitigating background noise, especially from EG chips Finding an appropriate collection time Characterizing a tool ability limitation for very low Ag% detection (less than 0.2% Ag). Through XRF, ICP, DSC comparison, XRF recipes can be developed as an excellent monitor for non- invasive bump composition evaluations. As in all methods, XRF has a unique set of challenges and limitations which can be overcome through proper calibration and verification. As the packaging trend moves toward finer pitched products, studies of small diameter C4s will become pervasive. If XRF is preferred to measure the bump composition of smaller C4s, geometry will become a key issue. C4 flattening helps mitigate geometry issues on smaller C4s, but may be insufficient for micro- bump features. Future work will involve how to define C4 geometry and the possible creation of composition test sites which will need to have extensive correlation established.
- 16. © 2009 IBM Corporation 16 References Acknowledgements J. Sylvestre, et al.,”The Impact of Process Parameters on the Fracture of Brittle Structures During Chip Joining on Organic Laminates,” 2008 ECTC. E. Perfecto, et. al, “C4NP Technology: Present and Future,” 2008 IMAPS Device Packaging Conference. B.. Beckhoff B. Kanngießer N. Langhoff R.Wedell H.Wolff (Eds.) in Handbook of Practical X-Ray Fluorescence Analysis 2006. Special thanks go to Chia-Hsin Shih and KyungMoon Kim from STATSChipPAC as well as Charles Arvin and Eric Perfecto from IBM for their guidance and support. Thanks also go to John Pennacchia, Stephen McLaughlin and the entire C4 team.