Issue



Enhancing yield with flux and solder paste process control


08/01/2001







A key element for successful Chip scale and ball grid array packaging

BY STACY KALISZ

Advances in semiconductor process technology are leading to increasing solder bump densities and, therefore, tighter bump pitches. The ability to tightly control each assembly process to ensure high yields has become a key element in the success of chip scale packaging (CSP) and ball grid array (BGA) packaging.

Solder flux and solder paste deposition is a standard step during the chip attach and sphere attach portions of BGA and CSP assembly. A technique called ultra-violet fluorescence intensity mapping (UV FIM) has been introduced that allows flux measurement for system set up, process characterization and production process control. The UV FIM technique can also be used for solder paste inspection and is faster than most existing technologies for this application.


Table 1. Summary of fluxes tested to date.
Click here to enlarge image

The purpose of this article is to discuss the application space for flux measurement using UV FIM and to describe the latest experimental work done with this technique.

The Need to Control Flux

The use of solder flux has become standard in BGA and CSP package manufacturing during chip attach, sphere attach and in the solder paste for board level attach. Flux is required to provide tack to hold die and spheres in place during reflow and remove oxides on solder spheres and board pads. It is important to apply an appropriate amount of flux that will activate and burn off during reflow, limiting flux residue.


Figure 1. Process flows for various flux and solder processes.
Click here to enlarge image

Using flux in manufacturing of BGA and CSP packages such that it meets these requirements is, in principle, straight forward and has been done for solder paste deposition.1 Flux is a non-Newtonian material, however, meaning its viscosity changes as it is worked. Also, solder fluxes are often transparent in color and, therefore, cannot be measured using standard optics. Despite these issues, the industry is beginning to recognize the need to measure and monitor flux amount and presence during deposition and chip attach and sphere attach.2,3 If errors are detected before chip attach or sphere attach, die or package yield loss can be reduced by reworking the flux or adjusting deposition process parameters.


Figure 2. Image of solder pads with and without solder flux under visible lighting.
Click here to enlarge image

Flux thickness has been linked to higher numbers of missing spheres at reflow4 and to an increase in solder sphere voids.5,6,7 Voids in solder spheres decrease joint reliability and reduce the joint's high frequency signal propagation.5,6,8 Defect rates at sphere attach have been shown to increase with increasing flux thickness.9 Researchers concluded that for flip chip soldering, less than 25 µm of flux thickness was insufficient to compensate for sphere height variations (i.e., in ball dipping, flux did not properly wet to all balls), and above 70 µm caused fuzziness in the vision location in the placement tools.10 Philips researchers refined this criterion by looking at flux residues, underfill adhesion issues, soldering requirements and reliability, and determined that a 20 µm minimum and 50 µm nominal thickness was recommended, based on the solder sphere coplanarity specification and the solder sphere height.11

While the importance of flux thickness and presence on pads or bumps is well-established, the methods of measuring flux thickness, height and volume are not. Existing assembly and inspection equipment can detect missing or skewed parts, but no flux measurement techniques have been available with sufficient speed and precision for commercial use. Researchers have used measuring microscopes2 and average weight of flux dots on test coupons,10 but these tools are inadequate to measure to the tolerance required.2 The UV FIM technique is a patent pending technique that has the capability to inspect and measure flux location, area, height and volume.3,12

Application Space for Flux Measurement

Several different package processes use solder flux during attach. The use of flux inspection with UV FIM was investigated for sphere attach at the package and wafer level, flip chip on board (FCOB) and flux on die bump applications. Flux inspection fits into the process flow immediately after flux application (Figure 1).


Figure 3. Image of solder flux deposits under UV FIM lighting. Note that Ni/AU pads without flux are not visible.
Click here to enlarge image

Sphere Attach, Package Level: Initial flux monitoring work done with the UV FIM technique focused on BGA and CSP sphere attach applications.3 Flux monitoring was done between the flux print and sphere placement process steps. Although most of the work was done using screen printing for flux deposition, the measurement technique is not affected by how the flux is deposited (screen printed or pin dip transfer).

Using standard lighting, flux visibility was limited (Figure 2). Although the flux can be seen if the image is examined closely (flux is on left side of the image in Figure 2), it is very difficult to detect and impossible to measure. With the development of the UV FIM lighting scheme, the solder flux can be easily detected (Figure 3).


Figure 4. A 3-D rendering of the flux volume/deposition represented by the square in Figure 3.
Click here to enlarge image

The capability to see the flux under the UV FIM lighting technique has been translated into 3-D measurement capability (Figure 4). Flux volume, area, location (offset from the pad) and height can now be measured. Measuring the flux attributes can allow for prediction of joint failure and defects. Previous work has shown that flux volume is known to contribute to placement errors and joint failures.3,15 A relationship between flux volume and ball shear strength was demonstrated,3,16 which is indicative of the relationship between flux volume and solder joint integrity.

Figure 5, which shows an image of 10 eutectic spheres after reflow, illustrates the correlation of the data obtained to actual sphere placement results. The UV FIM inspection tool, as used after flux print but before ball placement, flagged volume to be unusually low on pads 1, 2 and 3, therefore predicting an expected failure at placement or after reflow. Figure 5 shows that a missing ball occurred on pad 1 and non-wetting occurred on pads 2 and 3, consistent with the predictions and associated data from the UV FIM inspection technique.


Figure 5. post reflow BGA image. Pad 1 is a missing ball. Pads 2 and 3 demonstrate non-wets.
Click here to enlarge image

The utilization of UV FIM-produced flux data to predict joint failures, adjust process parameters and eliminate defects can be transferred to other flux deposition applications.

Sphere Attach, Wafer Level: In wafer level sphere attach, BGA spheres are attached directly to an uncut wafer. A stress compensation layer (SCL) is typically deposited on the wafer to eliminate the need for underfill and a substrate, and singulation is performed after the packaging process is complete.17

The shiny, mirrored surfaces of the wafer reflect light back into the camera, making many vision techniques impractical. However, the UV FIM lighting configuration allows for viewing wafers without reflection issues interfering. A comparison of Figures 6 and 7 shows the difference between standard lighting and UV FIM lighting to detect the presence or absence of flux. Further studies to calibrate for three-dimensional measurement are necessary, but these images show great promise for two-dimensional measurements.


Figure 6. Wafer pads with and without flux under normal lighting. The bottom left corner shows the flux wiped away from the pads.
Click here to enlarge image

Chip Attach, FCOB: In-line flux inspection in a flip chip on board application using UV FIM is performed after the screen printing of paste and flux, but before the components have been placed. The system detects both the paste and flux deposits at this process point. Gage studies of inline flux inspection demonstrate that this metrology tool is acceptable for production use, with a gage error of 2.4 percent for volume and 5.2 percent for height intensity measurements. Furthermore, this tool is capable of measuring flux deposition at around 400 deposits per second, which exceeds current line rate requirements.3,13

Chip Attach, Flux on Bump: To date, no known measurement technique is available for detecting flux on bumps. Most chip attach equipment is set up such that an image of the die is taken for alignment purposes, just before chip placement. Flux inspection could be done most efficiently if one image could be taken for both flux inspection and alignment purposes. To achieve this, the first step is to characterize the capability of measuring flux on bumps.


Figure 7. Wafer pads under UV FIM lighting head. Only the flux is visible. The pad locations can no longer be seen, as shown in the lower left corner.
Click here to enlarge image

Preliminary studies have demonstrated that the technique for measuring flux can be used in flux-on-bump applications during die attach.14 A study was done using bumped die where the bumps were manually dipped into a tacky chip flux. The presence or absence of flux, offset from bump and area of flux, can be resolved, but flux volume cannot be measured because of the round shape of the bump.

Further studies of flux on bumps with die that were dipped into a flux puddle using an automated chip placer were done. The resolution on the UV FIM lighting configuration used for these studies does not allow for flux on bump measurement as currently configured. Further work on increasing the amount of light emitted by the UV FIM configuration and increasing the magnification to optimize for measurement of bumps needs to be done.

Measuring Various Fluxes

Many different flux suppliers were contacted and samples of 40 commercially available fluxes were obtained for testing. These fluxes included chip attach and BGA fluxes in liquid and tacky formats. Of the 40 tested, 23 of the fluxes allow for two-dimensional characterization, including presence or absence of flux, offset, and area measurement with the standard UV FIM configuration. All of these fluxes were tacky fluxes.14

During the first phase of testing, it was evident that the fluxes that contained red, brown or amber dyes absorbed light in such a way that made them less suitable to the UV FIM 3-D or volume measurement techniques.14 The filtering scheme on the UV FIM lighting head was then upgraded to eliminate effects of visible room light, and all of the fluxes were re-tested. Again, all 23 of the tacky fluxes allowed for two-dimensional measurement. Additionally, the red, amber and brown fluxes re-test results were positive, and all 23 tacky fluxes are now capable of three-dimensional measurement.

Conclusion

UV FIM, a novel inspection technology, has been developed to measure flux height, volume, area and offset for various CSP and BGA packaging needs, with current status summarized in Table 2. Flux on board measurements for both FCOB and sphere attach are mature and available for commercial use today.


Table 2. Summary of application space for flux measurement technology
Click here to enlarge image

Yield enhancement through defect prediction has been sufficiently proven for sphere attach applications using flux inspection and can be translated into other areas of packaging where flux is used.

Comprehensive testing on tacky fluxes for both chip attach and sphere attach have indicated that commercially available fluxes of this nature are suitable for measurement and compatible with the UV FIM configuration.

Future Work

The results of the wafer level flux inspection studies demonstrate two-dimensional characterization of flux. Further studies to incorporate a calibration for three-dimensional measurement are planned. The UV FIM technique should allow for three-dimensional measurement without modifications, but this work has not been completed.

Initial feasibility studies for flux on bump have shown promise. Further work needs to be done to solidify the use of UV FIM for chip attach flux on bump characterization.

Excitation and emission data from the liquid fluxes and other non-tacky fluxes show that the fluxes may be measurable with slight modifications to the current UV FIM configuration.

AP

References


  1. M. Owen and J. Hawthorne, "Process Control for Solder Paste Deposition," by SMTA pp 488-493, 1999.
  2. C. Clark, "Impact of Flux Height Deposit on Pre-Reflow Sphere Registration," an internal Motorola CSP Process Characterization Study, August 7, 1998.
  3. S. Kalisz et al., "Inline Flux Volume Measurement for CSP Process Control," Pan Pacific Microelectronics Symposium 2000 Proceedings, 2000.
  4. "Solder-Ball Manufacturing and Attachment for BGA," Panel Discussion Presentation in BGA Symposium, Nepcon West 1997.
  5. C.S. Chiu et al., "Voiding in BGA at Solder Bumping Stage," pp. 462-471, ISHM, 1997.
  6. "Voiding Mechanism in BGA Assembly", ISHM, 1995.
  7. W. Casey, "Reduction of BGA Eutectic Ball Solder Joint Voiding," SMI, pp 541-548, 1998.
  8. R.D. Banks et al, "The Affects of Solder Joint Voiding on Plastic Grid Array Reliability," SMI, pp 121-126, 1996.
  9. N. Lee and W. Casey, "Soldering Technology for Area Array Packages," SMTA, pp 282-297, 1999.
  10. Novak et al, "Process Limit Testing on Fluxes used for Flip Chip Soldering," SMI, pp 275-280, 1998.
  11. C. Beelen and M. Verguld, "Verification of Flip-Chip Assembly on FR4 Boards," Soldering and Surface Mount Technology, pp 23-28, 1998.
  12. International patent publication number WO0059671(A1).
  13. "Measurement Systems Analysis Reference Manual" Chrysler Corporation, Ford Motor Company, General Motors Corporation Internal Quality Control Procedures, pp. 29, 51, 60, 69, 88, 1995.
  14. S. Kalisz, "Inline Flux Measurement for BGA and CSP Process Control" HDI, 2000.
  15. M. Weller et al., "Optimized CSP Assembly," SMT, May 1998.
  16. R. Erich et al., "Shear Testing and Failure Mode Analysis for Evaluation of BGA Ball Attachment," by IEEE/CPMT International Electronics Manufacturing Technology Symposium, 1999.
  17. S. Berry and S. Winkler, "New Methods of Underfill Application for WLP Covers the Entire Wafer," Chip Scale Review, p. 7, September/October 2000.

Stacy Kalisz, senior applications specialist, can be contacted at Agilent Technologies/MV Technology Ltd, 279 West Desert Avenue, Gilbert, AZ 85233; 602-458-7085; Fax: 480-813-1905 E-mail: [email protected].

Article content and research was first published by the SMTA in the Pan Pacific Microelectronics Symposium 2001 Proceedings.