Geometry and Bond Improvements for Wire Ball Bonding and Ball Bumping

Transcription

1 Geometry and Bond Improvements for Wire Ball Bonding and Ball Bumping Daniel D. Evans, Jr. Palomar Technologies, Inc Loker Avenue West Carlsbad, CA, Phone: (760) , Fax: (760) , Abstract Mobile electronic products continually require finer geometries for packaged integrated circuits. Wireless phones, PDA, and digital cameras continue to merge into a common device that will benefit from finer pitch and lower profile wire bonding or ball bumping for flip chip, stacked chip, and other advanced packaging technologies. Inherent variations in materials, tools, and process can cause variations in ball shape and size, stitch shape, bond quality, and thus yield. This paper will present Adaptive Bond Deformation, a method recently developed to control the geometry of both ball and stitch according to process parameter inputs supplied by the user. This technique adapts to normal variations in bond surface, bond tool coupling, part fixture, and other difficult-to-measure influences to produce bonded ball bumps and stitches with significant improvement in geometry consistency with similar or better results for ball shear and pull strength when compared to non-adaptive bonding. Key words: Ball Bump, Stud Bump, Wire Bond, Co-Planarity, GGI, Adaptability, Yield, and Traceability. Introduction Shrinking electronic package geometries are driving continuous improvements in interconnect technologies. Wire bonding is a mature technology that continues to meet challenges of finer geometric capability: loop shape, bond pitch, ball size, and stitch size control. This paper will focus on the ball bond and stitch bond as shown in Figure 1. Figure 1: Simplified representation of a wire interconnection showing the loop and change in ball and stitch geometries which occur during the ultrasonic bonding process. The packaging industry is looking for a common set of improvements as geometries continue to shrink: Uniformity of ball bump height and/or bump diameter to allow Reduced bond pitch for wires Reduced bond pitch for bumps Better bump coplanarity Uniformity of stitch impressions to Minimize adverse impacts to sensitive materials for stitch Process Reduced process development time Reduced sensitivity to material, environment, and setup variation Real-time capture/qualification of production data Data trending analysis of production data The motivation behind this work is to apply and validate Adaptive Bond Deformation (ABD) control technology (patent pending) for improving bond geometry. Adaptive Bond Deformation (ABD) was developed for a unique ball bump application where material, process, and tooling variations adversely affected yields due to occasional shorting of adjacent bumps.

2 The purpose of this investigation is to study the affect of ABD on both bump diameter and bump height for a standard ball bump application on a flat coupon. Although there has been promising work in applying Adaptive Bond Deformation to stitch-bonds, stacked-bumps, and stitch-on-bump, this paper will focus on ball bump. The definitions of terms used in this investigation are given in Figure 2 and Table 1. MBD BBD Figure 2. Ball bump shape and geometry terms Table 1. Ball geometry and shear definitions Measure Definition FABD Free air ball is the diameter of the ball created from the electronic flame off process while the ball is in free air and before it contacts the bond surface. MBD BBD TH SF SS TH Mashed ball diameter is the diameter of the ball after ultrasonic bonding is complete. (Largest diameter in plan view) Bonded ball diameter is the diameter of bond contact between ball and substrate. Top height is the height of ball measured from substrate surface to top of the ball. Shear force of the ball in grams Shear strength of the ball is calculated as the SF/BBD area in grams/mil 2. To show significant yield improvement, the goals of the investigation are to: Improve ball bond geometry consistency of MBD and TH by 20 % for a given process without adversely affecting shear strength more than 10%. Qualify an automated method of measuring MBD and TH to 0.2 µm, 3σ resolution, in order to verify improvements in MBD and TH consistency. Investigation This investigation involved three distinct sections: AUTOMATED BALL BUMP MEASUREMENT Develop and qualify an automated ball bump measurement method capable of resolving improvements in geometry control. ADAPTIVE BOND DEFORMATION Develop a bond control technique to modify the bond sequence based on bond geometry. BOND DATA MINER Develop a set of tools to capture in-situ bonding data for use in qualifying the improved geometry control technique, to use as a process development tool, and as a production monitoring tool. Automated Ball Bump Measurement One of the first things to be addressed was finding an efficient and capable method for measuring ball bump geometry. A Scanning Electron Microscope (SEM) was too slow and subject to operator interpretation while an optical microscope using conventional XY pattern recognition and laser height sensing was not capable of finely resolving ball bump geometry improvements (Measurement 3σ resolution requirement of MBD <= 0.2 µm and TH <= 0.2 µm) even using several different methods and tools available on the equipment. Although some gross cross checks on absolute accuracy were completed, NIST level verification was not necessary since relative bump geometries were the main interest of this study. Measurement System Selection Process Steps: Create bonded ball bump samples Draft measurement requirements and qualification protocol First pass survey of 19 suppliers Down select for refinement survey of 4 suppliers Analyze results Select measurement supplier based on capability to measure both geometry characteristics (TH and MBD) Confocal Multi-pinhole Microscopy was chosen as the measurement technology system based on supplier response and gage capability.

3 Optical surface measurement technologies have made much progress in the last few years and some of them are now as accurate as high grade stylus profilometers, with the added advantage of avoiding damage to the surface being measured. One of the non-contact systems is a new confocal microscope that is especially designed for application in the field of engineering surfaces [C1]. The principle is based on the confocal-multi-pinhole (CMP) technique with dynamic real time synchronization, which leads to a high signal to noise ratio for each measured data point [C2]. The principle of confocal white light microscopy is based upon increasing the effect of focal discrimination by means of an optical microscope and space filter technique (pinholes). Light intensity which does not originate in the focal plane of the objective will be suppressed. In order to use this principle for not only one point, but for an area, a multi pinhole disc is used as spatial filter. Pinholes are arranged on this rotating, micromechanically-manufactured disc in a spiral form as shown in Figure 3. results in hundreds of single intensity pictures, of which each represents the height information in a certain plane (like contour lines in a map). A fast mathematical algorithm finds the maximum intensity for each surface point. Interpolation effects lead to a 10 times higher accuracy in the height measurement than the number of z steps that the objective makes. The vertical scanning or shift of the objective is achieved with a closed loop controlled piezo transducer with 350µm travel length or, for larger height measurements, with a stepper motor and a mechanical z-axis. The confocal microscope ball bump scan is shown in Figure 4. (a) 3-D image Vertical scan causes the MBD to shadow the ball outer diameter wrap around toward the BBD/Substrate interface Figure 3. Schematic diagram showing the operating principal of confocal multi-pinhole microscopy The disc is oriented between a beam splitter and the microscope objective lens (objective). Each individual pinhole represents a point light source that is focused on the surface by the objective lens. The light that is reflected or scattered back from the surface will be exactly imaged into the same pinholes and transmitted via the beam splitter into a CCD camera detector. The entire image in the microscope objective lens field of view is scanned in real-time onto the CCD camera during one revolution of the disk. The maximum light intensity will be detected when a sample surface point is exactly in the focal point of the objective. Precise shifting of the microscope objective in the z-direction (b) Contour Image shows Z versus XY Figure 4. Confocal Microscope ball bump vertically scanned images (a) 3D view, (b) Contour Map show two different views of the scan data. Algorithms and methods were developed to analyze the scanned ball bump data to automatically measure MBD and TH. Figure 5 shows the repeatability of measuring a single ball, 100 times. Other tests were

4 run on the same ball at another angle and on another ball bump however results are not shown in this paper um 3S (a) Top-Height Qualification Table 2. Bond Deformation Definitions Measure Touch Height Pre Bond Height Post Bond Height Touch Deformation Ultrasonic Deformation Total Deformation Description Recorded Z position at touch sense Recorded Z position at ultrasonic start Recorded Z position at ultrasonic end Deformation between touch and start of ultrasonics = Pre Bond Height Touch Height Deformation between start and end of ultrasonics = Post Bond Height Pre Bond Height Touch deformation + Ultrasonic Deformation = Post Bond Height Touch Height 0.2 um 3S (b) Mash Ball Diameter Qualification Figure 5. Confocal system results measuring 1 bump 100 times (a) TH at 0.12 µm, (b) MBD at 0.2 µm Adaptive Bond Deformation Technology Adaptive Bond Deformation records the Z height at touch sense, ultrasonic start, and ultrasonic end. From these three heights, three separate deformations can be calculated: touch, ultrasonic, and total. Schematic representations of the height and deformation measures are described in Table 2 and depicted in Figures 6 and 7. The same measures apply to ball and stitch separately, although the detail dimensions are shown only for the ball. Adaptive Bond Deformation automatically modifies the bond process to achieve a specific bond deformation according to specified process limits for deformation and time. Figure 6. Ball Deformation Diagram Figure 7. Stitch Deformation Diagram Adaptive Bond Deformation Theory of Operation Theoretically, FABD consistency, Touch Consistency, and Deformation Consistency should produce TH and MBD Consistency. Since the bonder cannot directly measure the actual ball bump shape, Adaptive Bond Deformation assumes there is a consistent set of free air balls input to the bond process and the bonder consistently senses touch. Both of these assumptions are valid and were checked for in this experiment. Although there is a portion of the ball deformation which is nonobservable, given consistent FABD and Touch inputs

5 then using ABD to produce a more consistent deformation during bonding should result in more consistent MBD and TH. Bond Data Miner - Automated In-Situ Bond Data Capture, Cataloging, and Retrieval Bond Data Miner (BDM) tools were used in development of Adaptive Bond Deformation and are useful in process setup and production process monitoring. The tools allow data gathering and analysis of the parameters defined in Table 1, bond time, and other measures for each ball bump as shown in Figures 8 and 9. ABD - OFF ABD - ON Figure 8. Bond deformation data are displayed for and. has tighter deformation grouping within specified upper and lower deformation control limits. ABD - OFF Experiment Setup Palomar Technologies Model 8000 Wire/Ball Bumper Au Kovar Lid Substrate 1x1 inch (bump matrix with ID, orientation, and row column ball numbering) Au wire, 25.4 µm diameter, <1% elongation Capillary FABD 73.5 µm +/ µm, 3σ Process Steps Bond samples with ABD = OFF ( ) Bond samples with ABD = ON (, ) Measure MBD, TH for ball bumps using the confocal measurement system Review samples and data to confirm measurements Shear ball bumps Analyze results BDM data saved (Touch, Pre Bond, Post Bond, Touch Deform, US Deform, Total Deform, Bond Time, and other measures for each ball bump) ABD parameters selected were based on BDM bond deformation statistics captured during ABD = OFF run. Parameter selection is a relatively simple process after reviewing the BDM data. Experimental Results and Interpretation Summary output of the experiment showed that Adaptive Bond Deformation performed to goals. Measurement comparison of 85 measured samples for vs. OFF ABD - ON Figure 9. Bond deformation distributions are displayed for and. The bimodal distribution was specified by the user through the selection of ABD parameters. Experiment Once the capabilities to measure the bumps, adaptively bond the bump (based on geometry), and capture and save data were all in place, the experiment was arranged to validate the geometry control improvements as a result of using Adaptive Bond Deformation. TH (µm) 60.8 (+1.7%) +/ σ (-30%) See Figure 10 MBD (µm) 83.5 (-1.0%) +/ σ (-21%) See Figure 11 SS (g/mil^2) 9.7 (-6.3%) +/ σ (-8.9% ) See Figure 12 Bond Deformation (µm) 14.3 (-0.4%) +/ σ (-65%) See Figure 13 Bond Time (ms) 4.4 (-70%) See Figure 14 Adaptive Bond Deformation has the following effects: TH Showed a 1.7% increase in average height and a 30% reduction in variance. This was well above the 20% variance improvement goal. MBD Showed a 1% reduction in average diameter and a 21% reduction in variance. This is above the 20% variance improvement goal.

6 SS Showed a 6.3% reduction in average shear strength and an 8.9% reduction in variance. This is within the goal of the investigation and produced ball shear strengths well within accepted quality standards. Bond Deformation Showed a 0.4% reduction in average deformation and a 65% reduction in variance. Bond Time Showed a 70% reduction in bond time when using Adaptive Bond Deformation. (Resulting in a higher wire per second rate) TH [um] Figure 10. Top height (TH) measurements showing a 30.3% improvement in three-sigma variation with MBD [um] Figure 11. Mashed Ball Diameter (MBD) measurements showing a 20.8% improvement in three-sigma variation with Total Deformation [um] Figure 12. Total Deformation measurements showing a 65.3% improvement in three-sigma variation with Shear Strength [gram/mil^2] Conclusion Adaptive Bond Deformation technology was validated as improving bump geometry with some reduction in shear strength. Goals of the project met were: ABD shows a 30% improvement in TH and a 21% improvement in MBD variance; There was an 8.9% improvement in ball shear strength variance and a 6.3% reduction in average shear strength; Automated Bump Measurement for MBD and TH was validated and used for 0.2 µm capability; Automated Data Management and Mining for in-situ deformation was developed and used for process development and verification. Additional benefits included: time savings per bond of 10.6 ms, or 70.7 % reduction of the bond time. Suggestions for further study would include: Increase the Adaptive Bond Deformation control parameter limits to increase deformation and validate that it also raises the shear strength; Better setup of the initial ball shape and process; Develop an automated ball bump measure for mashed ball height (MBH) and the height of the capillary lip on the ball from the substrate. MBH can cross check other measurements and process; Perform the same work on an aluminum substrate to verify the same benefits of using ABD; Perform modifications to background ultrasonics values to improve resolution of deformation control; Alternate investigations could be completed for Adaptive Bond Deformation improvements applied to Stitch Bond, Stacked Bump, Stitch on Bump, etc Figure 13. Shear Strength (SS) measurements showing an 8.9% reduction in three-sigma variation with and 6.3% reduction in average shear strength Bond Time [ms] Figure 14. Bond time measurements showing a 71% reduction in bond time with

7 Acknowledgements The author would like to thank the following people for their support on this paper: Palomar Technologies, Inc. Don Beck, Mark Greenwell, Albert Perez, Mike Artimez, James O Bryan. NanoFocus, Inc - Dr. Christian M. Wichern, Managing Director, References [C1] Jordan, H.-J., et al.: Quality assurance of HARMS and MOEMS surface structures using confocal white light microscopy, SPIE Proceedings 4440 (2001), 51. [C2] Weber, M.A., et al: Konfokale Mikroskopie zur Oberflächenuntersuchung mikrostrukturierter Materialien Tagungsband, unkenerosion 2002, WZL RWTH Aachen, Kapitel 20. Biography Dan Evans is Senior Scientist for Palomar Technologies (formerly Hughes Aircraft). He has developed equipment and processes for semiconductor and optoelectronic packaging since 1984 with several related patents. Dan holds a BSME/Purdue and MSME/Stanford both with emphasis in robotics and controls.