Processing parameters and reliability performance
BY CHRISTINE KALLMAYER, KARL-FRIEDRICH BECKER AND TOM ADAMS
Flip chip packages hold great promise for high performance and low manufacturing cost, but several criteria must be met to achieve these goals. Some of the criteria – a high throughput rate and high reliability, for example – can be difficult to achieve in the same package design.
Reflowable (or no-flow) underfill techniques may improve the overall manufacturability of a flip chip package because the need for the fluid underfill material to flow by capillary action under the chip is eliminated. Instead, the fluid underfill is deposited onto the substrate, and the chip is then put in place. During placement the solder bumps push through the fluid underfill to make a first mechanical contact with the substrate (Figure 1). In the subsequent reflow step, the solder joints are formed.
 Figure 1. Typical reflowable underfill process flow. |
The material properties of the underfill material, along with the material's flow behavior during compression by the die, are critical parameters in designing reflowable flip chip processes. There have been advances in this area recently, and this article reports on work carried out at the Fraunhöfer Institute to characterize newly available materials and to establish processing guidelines.
Test and Measurement Procedures
The test and measurement procedures described here permit the standardized inspection of incoming reflowable underfill materials and the analysis of prototype assembly processes. The following methods were used to determine material properties:
- Rotational viscosimeter to determine viscosity
- Optical microscopy to determine wetting angle
- Differential scanning calorimetry (DSC) to determine each material's cure profile
- Thermomechanical analysis (TMA) to determine coefficient of thermal expansion (CTE) and glass transition temperature (Tg)
- Dynamic mechanical analysis (DMA) to determine Tg and Young's Modulus
To evaluate assembly processes, the following approaches were used:
- Specially designed test board to simulate various substrate topographies
- X-ray inspection to evaluate placement accuracy and contact shape
- Scanning acoustic microscopy (C-SAM) to characterize solder joints and to detect underfill defects
- Physical cross-sectioning to visually evaluate solder joints and underfill quality
Most of the reflowable underfill materials currently available do not contain filler particles. Of the four materials (designated A, B, C and D) tested during this study, only material C contained filler particles (10 percent by weight). The absence of filler particles renders the analysis more straightforward, although under some conditions underfill materials without filler particles can themselves be a limiting factor in reliability.1
The die used in the study were 10 mm x 10 mm silicon chips having 136 peripheral I/Os at 300 µm pitch. The chips were bumped with 5 µm electroless Ni/Au and with 100 µm stencil-printed eutectic Pb/Sn. Two different substrate types were used. To evaluate process development and reliability, a 1 mm thick FR4 board with 18 µm Cu/Ni/Au conductor lines and daisy chain structures was used. To evaluate flow properties, a 2 mm-thick FR4 board having 35 micron-thick Cu/Ni/Au conductor lines was used. This test board contains 10 different patterns representing different substrate topographies, and it was useful in determining flow behavior and the presence of defects, such as voids, over the various topographies.
Because the material properties of a reflowable underfill material have considerable influence on its suitability, several procedures were used to measure material parameters of each material.
 Table 1. Material properties of the four no-flow underfill materials studied. |
The thermomechanical properties of the four materials A, B, C and D are shown in Table 1. There are no extreme differences among the four materials, but the properties of all four differ from the properties of conventional flowed underfill materials. Specifically, the materials tested here have CTEs that are higher than those of conventional flowed underfill materials by a factor of about three, a lower Young's Modulus (typically 8 GPa for flowable materials), and a Tg lower than the 145°C typical of flowable materials. Together, these properties suggest that the reliability of these four reflowable materials would be lower than the reliability of conventional flowable materials. Additionally, materials A and C have such low Tg values that their use would probably be limited to low-performance consumer applications.
The cure behavior of the four materials was evaluated with differential scanning calorimetry (DSC). The four materials had similar profiles, and all four could be expected to cure in less than 3 minutes only if the temperature were at or above the Pb/Sn eutectic temperature.
 Figure 2. Board pre-treatment resulted in faster wetting and a smaller final wetting angle. |
The time needed for 95 percent of each material to cure was also measured. Materials A and C cure with a standard reflow profile, it was found, while materials B and D require shorter profiles in order to avoid premature gelation.
Before the evaluation of the wetting behavior and wetting angle for the four materials, two separate 1 mm thick board types were prepared. One type was stored under N2 atmosphere but not otherwise treated. The second type was dried for three hours at 125°C in a vacuum, and then treated with O2 plasma for five minutes. The second method resulted in faster wetting and, therefore, permitted faster placement of the chip. It also reduces the possibility that the chip may lift after placement and thereby break the contact between the solder bumps and the substrate. The wetting angle resulting from the plasma method was further reduced by reflow to less than 10°, resulting in excellent wetting of the sides of the die and a wider fillet than is typical with conventional underfill processes (Figure 2).
The Assembly Process
Reflowable underfill materials may be deposited onto the substrate by screen printing, dipping or dispensing. For this study dispensing was used. In initial samples, two patterns – a dot and a star – were used in dispensing, and the cured underfill was then examined by C-SAM to determine whether either pattern gave an advantage. C-SAM imaging showed that there were no voids in the underfills produced by either pattern, so a dot in the center of the chip site was used for the balance of the study.
The chips were put in position by a manual pick-and-place machine. Each chip was held in place for approximately three seconds to permit wetting of the sides of the die and to permit the fluid underfill material to relax sufficiently to prevent the die from lifting. In a production environment, it would probably be feasible to use a much shorter hold time. After the die was positioned, X-ray imaging was used to examine the position of the die.
Solder Joint Formation
Differential scanning calorimetry, mentioned earlier, had shown that materials A and C required standard reflow profiles, while materials B and D needed faster profiles. These profiles were observed during reflow. Even with the faster profile and with careful verification of chip placement, reproducible solder joints could not be achieved with material D. This material was therefore excluded from further investigation, including reliability studies. For materials B, C and D it was found that the presence of the underfill material had no impact in die self-alignment and that initial displacement as great as 60 percent of the pad size still yielded good interconnects. For material A, the self-alignment is inhibited by the fast curing process, so a residual displacement was found for nearly all samples (Figure 3).
Test Board Determination of Flow Behavior
The test board designed at the Fraunhöfer Institute has 10 different substrate patterns having, for example, lines of various widths, diagonal lines, and uncovered vias. Each of the 10 patterns appears twice on the test board, which is designed to predict the behavior of an underfill material in the presence of various substrate features. Because the features present a change in direction with respect to the flow of underfill material, the presence of these features often causes the formation of voids (Figure 4).
Previous experience with the 10 patterns on the test board had shown that the following three patterns are most likely to produce voids:
- Large grooves of increasing width, but without metallization structures, in the solder mask layer. This pattern simulates large uncovered areas.
- Metal pads of 200 µm diameter with solder mask openings of 300 µm. This pattern simulates small uncovered vias, and it is the most critical pattern for reflowable underfills.
- Metal pads of 300 µm diameter with solder mask openings of 500 µm. This pattern simulates large uncovered vias.
Each material was dispensed onto the pre-dried test pattern and reflowed. After reflow, C-SAM acoustic micro imaging was used to check for the presence of voids. The very few voids observed were all adjacent to the peripheral bumps when materials A and C were used (Figure 5). (The pattern with the unstructured solder stop mask was used.) This is, however, a location that may cause eventual device failure because the solder may creep into the void, a phenomenon noted in conventionally underfilled flip chips after approximately 2000 thermal cycles.
 Figure 5. C-SAM imaging after reflow showed voids only in materials A and C. |
Overall, however, it appears that reflowable underfill materials minimize the occurrence of voids in comparison with conventional underfill materials. This suggests that variations in substrate topography may not be as critical when reflowable underfill materials are used. Overall, however, it appears the reflowable underfill materials minimize the occurrence of voids in comparison with conventional underfill materials. Dispensing the reflowable underfill onto the most critical patterns created no voids. This suggests that variations in substrate topography may not be as critical when reflowable materials are used.2
Reliability Testing
Samples using materials A, B and C (there were no viable samples using material D) were exposed to thermal cycling (-55°C to 125°C) and humidity testing (85% R.H. at 85°C). At regular intervals during testing, samples were imaged acoustically by C-SAM to inspect for internal defects. After these tests, electrical testing for daisy chain integrity showed clearly that the current group of reflowable materials is superior in reliability to materials tested previously. There were no significant increases in resistance after 1,000 hours of humidity exposure and 700 thermal cycles (Figure 6). Material A did not show signs of degradation even after 1,000 cycles. Acoustic imaging, however, showed that material A had a tendency to form large delaminations rather quickly during humidity testing. Acoustic imaging of samples using materials B and C showed no similar defects.
Conclusions
The four materials tested during this study fall into two groups. Materials A and C are relatively soft materials having low Tg, and can be processed with standard reflow profiles. Materials B and D are more like conventional underfill materials and require faster reflow materials. For material D, the processing constraints are so limiting that no reproducible solder joints could be obtained.
 Figure 6. The change in daisy chain resistance after 700 thermal cycles (top) and 1,000 hours of humidity exposure (bottom). |
In general, the three materials subjected to reliability testing performed well, although the delaminations noted after humidity testing of material A might lead to eventual device failure. Reflowable underfill materials appear to be suitable for high volume production, particularly for consumer products where the reliability requirements are not extreme. AP
References
- K.-F. Becker et al., “Qualification of Underfill Materials for Flip Chip Assemblies Using Acoustic Microscopy,” IAMIS 1998 Symposium.
- K.F- Becker et al., “Basic Design Guidelines for a Robust Underfilling Process,” IAMIS 2000 Symposium.
Christine Kallmayer is a development engineer at Technical University of Berlin, FSP Technologien der Mikroperipherik. Karl-Friedrich Becker is a development engineer at Fraunhöfer IZM, Berlin. For information, contact Tom Adams, consultant for Sonoscan Inc., 20 Devon Avenue, Lawrenceville, NJ 08648; 609-883-5040; Fax: 609-883-5087; E-mail: [email protected].
null