Selective Electroless Nickel and Gold Plating of Individual Integrated Circuits for Thermocompression Gold Stud Bump Flip-Chip Attachment


Reading time ( words)

AbstractFlip chip bonding is the most desirable direct chip attachment approach for minimizing electronic assembly size as well as improving device performance. For most prototyping applications it is not cost-effective to purchase individual integrated circuits (ICs) that are solder-bumped as this typically requires the purchase of entire wafer. Also, many unpackaged IC's in die form are not available for purchase as an entire wafer for subsequent solder bumping. As an alternative to solder bumping, manufacturers of wire bond equipment have developed the gold stud bump process which allows single IC's to be automatically bumped using 1-mil gold wire. However, the rapid formation of brittle aluminum-gold (Al-Au) intermetallics at elevated temperatures (>200°C) precludes the use of thermocompression flip chip bonding due to the unreliability of the bond at the IC pad interface. To overcome the intermetallic problem at the IC's aluminum-metallized bonding pads, an electroless nickel and gold plating process was developed for making a gold-bondable diffusion barrier for use on individual, unpackaged silicon ICs. This process provided an electroless gold layer suitable for accepting the gold wire stud bumps as well as providing the necessary barrier to Al-Au intermetallic formation. A number of experiments were conducted using electroless nickel of various phosphorus contents to determine which would provide an optimal diffusion layer. Data will be presented comparing immersion and autocatalytic gold plating processes. Test wafers were stud-bumped and exposed to accelerated temperatures then shear tested.  Electroless nickel, immersion and autocatalytic gold plating process parameters were optimized to provide high reliability interconnections when using the high temperature thermocompression flip-chip bonding die-attach method. 

IntroductionThe gold stud bump flip chip process creates gold bumps on integrated circuit (IC) die bond pads. The die is then directly connected to a circuit board or substrate. Gold stud bumps are formed by a wire bonder utilizing a modified wire bonding technique. This process allows individual, off the shelf die to be gold stud bumped. This process is ideal for product development and prototyping.Aluminum is generally the metallization used for IC bond pads flip chip bonding of gold stud-bumped chips can be accomplished using an anisotropic adhesive or by thermocompression bonding. Thermocompression bonding is a much faster than waiting for anisotropic adhesive to cure. However, bonds made with stud bumps that are made directly on aluminum bond pads are unreliable due to the aluminum-gold intermetallics that form readily at thermocompression bonding temperatures (~300°C). It's well known that the formation of aluminum-gold (Al-Au) intermetallics at elevated temperatures (>200°C) degrades and compromises the bond at the gold bump and aluminum pad interface.A number of Under Bump Metallization (UBM) processes have been developed to overcome the intermetallic formation. UBM may be vacuum deposited by evaporation and sputtering techniques or electroless and immersion plating processes. However, vacuum deposition would is not suitable for metalizing individual die because of the number of masking steps they require. Electroless plating is an ideal alternative to vacuum deposition. This paper discusses an electroless nickel, immersion and autocatalytic gold plating process that was developed to plate aluminum bond pads on individual ICs.BackgroundWhen electroless nickel and gold (ENIG) plating aluminum, a zincation process is used to activate the exposed aluminum before the nickel is deposited. The aluminum oxide layer is removed and the surface is activated through zinc displacement plating using a zincate solution. A double-zincation process is used to assure maximum adhesion.  Electroless nickel (EN) is then deposited from a hypophosphate-based nickel bath. The EN is an autocatalytic process which chemically deposits a nickel-phosphorus alloy onto the zincated aluminum. The phosphorus content of EN baths can range from 3% up to 15%. Immersion gold is then deposited thru a galvanic reaction with the electroless nickel to deposit a thin layer of gold onto the nickel. For applications that require a thick gold layer such as stud bumping and wire bonding, autocatalytic electroless gold is then plated over the immersion gold layer.Figure 1 shows a crossection of an ENIG plated IC. The IPC ENIG Specification-4522 specifies 3 - 6 microns of nickel with 0.05 - 0.1 microns of immersion gold and 0.5-1.5 microns of autocatalytic gold.Lee Figure 1.jpgFigure 1: Cross-section of plated IC.Since all IC's are not fabricated the same way, ENIG plating process steps have to be tailored to each individual IC type. Variations in the aluminum alloy, bond pad size and thickness, passivation material and bond pad electrical potential have to be analyzed. Probe marks on bond pads and contamination also can make plating difficult. Experimental ProcedureA number of plating companies offer proprietary processes for ENIG plating on aluminum. The plating experiments in this paper were conducted using one of these proprietary processes. EN baths containing phosphorus contents of 2-3%, 7-8% and 11-12% were compared. EN deposits with both immersion and autocatalytic gold were evaluated. The pre-plate preparation steps, plating times and metal layer composition were varied. The adhesion strength of the ENIG layers was evaluated by measuring the force required to shear gold stud bumps from the plated aluminum pads. The reliability of the stud bump-to-plated pad interface was evaluated using a ball shear test after subjecting test vehicles to multiple temperature cycles and high temperature dwells.Silicon wafers with 2 microns of  evaporated aluminum were prepared as test coupons. The wafers were annealed at 200°C for 2 hours to increase the adhesion of the aluminum. The wafers were then diced into 1x1" test coupons. The silicon coupons were plated before the individual ICs to analyze the plating processes. The process steps for the ENIG under-bump metallization are shown in Table 1. The first step is a plasma etch in an oxygen/argon plasma. The plasma etching removes surface organics. The coupons are then immersed in a cleaner to remove any contaminates and help the wetting process. A micro etch is used to break down the oxide layer and to remove any alloying constituents from the surface of the aluminum. A nitric acid dip re-forms the oxide layer and prepares the aluminum to accept the zinc. A double-zincation process is used to obtain a uniform zinc layer and to assure maximum adhesion. Each wet process step is followed by a rinse in de-ionized water. The test coupons were plated with 4 microns of nickel and 0.05 - 0.1 microns of immersion gold. Coupons were also prepared with 1 micron of autocatalytic gold.Lee Table 1.jpgTable 1: Process steps.

The test coupons were plated with each of the conditions listed in Table 2. Each coupon was then gold stud bumped with 40 bumps using a F&K 6400 Wire Bonder with the following parameters: 1 mil gold wire, time = 30, bond force = 30, and ultrasonic power = 110. The coupons were subjected to a die shear test using a Dage Series 4000 Bondtester with the following parameters: speed 15.00mil/s, shear height 0.10mil. The bumping parameters were kept consent for all samples in this experiment.It was decided that 20 grams-force would constitute the minimum value acceptable for an individual stud bump shear strength. The silicon coupons were first shear-tested "as plated." The shear results were listed in three categories:

<20 g = poor, 20 - 40g = acceptable and >40g = good.

Table 2 indicates poor adhesion for the 2-3% phosphorus nickel regardless of the gold thickness. For this case there was no measurable adhesion at the nickel/immersion gold interface. The 7-8% and the 11-12% phosphorus nickel have poor adhesion with the immersion gold but had very good adhesion with the thicker autocatalytic gold.Lee Table 2.jpgTable 2: Stud bump adhesion.

Coupons from the two plating processes that passed the "as plated" test were then subjected to a thermal soak consisting of 1 hour at 300°C followed by 168 hours at 150°C. The 7-8% phosphorus nickel/autocatalytic gold showed no degradation in bond strength after the temperature soak. The 11-12% phosphorus nickel/autocatalytic gold showed a loss in of adhesion but the bonds could still be categorized as acceptable, as shown in Table 3. The two suitable plating baths were investigated and compared in the following plating tests.Lee Table 3.jpgTable 3: Adhesion test.Two types of IC's were selected for this experiment and are shown in Figures 2 and 3.  IC-A is a Voltage Regulator using bipolar technology.  IC-B is a Memory Chip using C-MOS technology.  Each IC was from a different manufacturer.  IC-A bond pads are 100x100 microns and the aluminum is 2 microns thick.  The passivation layer is 2 microns.  IC-B bond pads are 125x125 microns and the aluminum is 2 microns thick.  The passivation layer is 2 microns.Lee Figure 2.jpgFigure 2: IC-A.Lee Figure 3.jpg

Figure 3: IC-B.The back and sides of each IC was coated with liquid resist to prevent unwanted plating on any metalized areas except the bond pads.  The IC's were ENIG plated with the process shown in Table 1.  Electroless nickel layers of 4 microns were deposited on each specimen.  The electroless nickel baths containing, 7-8% and 11-12% phosphorous was compared.  Immersion gold layers of 0.05 - 0.1 microns were deposited followed by a 1 micron deposition of autocatalytic gold. The nickel and gold thicknesses were determined by using Seiko Instrument's SFT 7000 Fluorescence X-Ray Gauge. Figures 4 and 5 show a bond pad both before and after plating, respectively.Lee Figures 4 and 5.jpgFigures 4 & 5: Before ENIG (left); after ENIG (right).Process Issues A number of processing problems were encountered when plating the IC specimens. As mentioned earlier, surface preparation is a critical step in preparing the aluminum for the zincate step. Also, the etching step must be carefully controlled. The amount of time to etch and the concentration of the etchant are dictated by the aluminum alloy and thickness. Figure 6 shows bond pads that have been over etched. The dark pits are largely copper that is added to the aluminum to retard the solid-state diffusion of gold into aluminum.Lee Figure 6.jpgFigure 6: Over-etched pads.

Uneven plating can occur if a non uniform zincation layer has been applied to the bond pad as shown in Figure 7.Lee Figure 7.jpgFigure 7: Uneven plating.Probe marks on bond pads also can present processing problems.  Figures 8 shows probe marks before and after plating.Lee Figure 8.jpgFigure 8: Pads prior to plating (left) and after plating (right).Figure 9 shows an example of incomplete plating, i.e, some of the pads do not plate at all. Some devices have internal blocking diodes built into the chip that separate the ground and the power bond pads. If the plating bath has the right conductivity, a current can be generated through these blocking diodes producing a voltage potential sufficient to affect the plating process for these pads. Isolating the backside of the chip from the plating bath can alleviate this problem.Lee Figure 9.jpgFigure 9: Example of a bipolar IC exhibiting incomplete plating.

Gold Stud Bumping

Four ICs from each plating process were gold stud bumped using the same parameters that were used on the silicon test coupons. Figures 10, 11 and 12 show IC-A and IC-B samples after gold stud bumping.Lee Figure 10.jpgFigure 10: Sample IC-A after bumping.Lee Figure 11.jpg

Figure 11: IC-B after bumping.Lee Figure 12.jpg

Figure 12: Close-up view of gold stud-bumps on IC-B.

After stud-bumping, samples from each plating bath were subjected to a thermal test consisting of 1 hour at 300°C and 168 hours at 150°C. Die shear test were then preformed on "as plated" samples and the heat treated samples. Results and Discussion

Twenty stud bumps from each IC were shear-tested. Figures 13 and 14 show the mean value of the shear tests. Excellent stud-bump bonds were achieved with both plating processes. Some decrease in bond strength can be seen after the samples were exposed to the thermal test however, bond strengths remained within acceptable values. Lee Figure 13.jpgFigure 13: Shear test results for 7-8% PEN.Lee Figure 14.jpg

Figure 14: Shear test results for 11-12% PEN.Failure modes from the shear tests are shown in Figure 15. Type 1: shear failure was in the gold bump; Type 2: shear failure at the bump/gold interface; Type 3: shear failure at the nickel/gold interface, Type 4: shear failure at the nickel/aluminum interface.Lee Figure 15.jpgFigure 15: Failure types.

Figures 16 and 17 show the die shear test failure types. As seen in the chart in figure 16, the majority of failures in the 7-8% PEN are at the stud bump/gold surface interface. The chart in figure 17 shows that most of the bond failures in the 11-12% PEN occurred at the nickel/gold interface. No failures were seen at the nickel/aluminum interface. Type 4 failures occurred only in the 2-3% PEN samples.Lee Figure 16.jpgFigure 16: Shear failure types for 7-8% PEN.Lee Figure 17.jpgFigure 17: Shear failure types for 11-12% PEN.Deposit Morphology The surface roughness was measured across 4 bond pads on IC-A and IC-B. Measurements were taken before and after the ENIG plating using a VEECO Dektak 6m Stylus Profiler. As shown in Table 3, each sample was slightly smoother after plating.Lee Table 4.jpgTable 4: Surface roughness (RA).

A crossection of a 7-8% phosphorus nickel/immersion gold/autocatalytic gold bond pad was observed with a scanning electron microscope (SEM). The nickel morphology is observed to have a fine grained structure and the gold layer is crystalline (Figure18).Lee Figure 18.jpgFigure 18: Cross-section of a plated bond pad.Thermal Diffusion Test and ResultsSilicon wafer specimen's were plated with 4 microns of 7-8% phosphorous nickel, 0.05 microns of immersion gold and 1 micron of autocatalytic gold were subjected to a thermal test consisting of 1-3 hours at 300°C and 168 hours at 150°C. Thermal diffusion of the nickel into the gold was analyzed by energy dispersive spectroscopy (EDS). The EDS technique is not quantitative and thus, it is not possible to measure the quantity of the diffused species (i.e., amount of gold diffused into the nickel). However, the diffusion length can be measured. The data does show that there was no significant diffusion of gold into or through the nickel. Figures 19-21 show EDS analysis of specimens: As Plated, 3 hours at 300°C and 168 hours at 150°C. Lee Figure 19.jpgFigure 19: As plated.Lee Figure 20.jpgFigure 20: Three hours at 300°C. Lee Figure 21.jpgFigure 21: 168 hours at 150°C.

An IC-A sample was flip chip bonded to a gold-plated molybdenum substrate. Figure 22 depicts a cross-section of the two stud-bump bonds indicating a high-quality gold-to-gold attachment.Lee Figure 22.jpgFigure 22: Cross-section of flip chip IC.Summary of Results The results indicated that a 7-8% phosphorus nickel deposit resulted in better reliability after temperature soaks testing. A paper published by Hashimoto et al. (2001)  [1]reported that the preferred phosphorus content in nickel films was between 6 and 8.5. In this paper, it was noted that at low phosphorus levels (< 7% by weight), the electroless nickel deposit is microcrystalline (tensile film stress). As the amount of alloyed phosphorus increases, the microstructure changes to a mixture of amorphous and microcrystalline phases, and finally to a totally amorphous phase (compressive film stress). Masayuki and Gudeczauskas et al. (1997) [2] also noted that a nickel deposit containing 7-8% phosphorus is desirable when immersion gold plating.Poor stud bump adhesion resulted from ENIG deposits containing 2-3% phosphorus. After temperature soaking, stud bumps made on ENIG deposits with 7-8% and 11-12% phosphorus exhibited excellent adhesion when used with the thicker autocatalytic gold. Although the ENIG plating on the samples with the 11-12% phosphorus nickel had high bond strengths, the majority of these bonds failed at the gold-nickel interface when shear-tested. In the case of the 7-8% phosphorus samples, shear failures occurred at the bump/pad interface. As noted by Johal, Roberts, Lamprecht and Wunderlich et al. (2005) [3], the internal film stress within the nickel deposit will shift from tension at medium phosphorus content (7-9%) to the compression at (10-13%) phosphorus content. A high compressive stress in the nickel would likely produce a failure at the gold-nickel interface where as a low tensile film stress would not.    After temperature soaking at 300°C for 1 hour and 150°C for 168 hours, EDS analysis indicated that there was little if any appreciable diffusion of the gold atoms into or through the nickel film. This study indicated that ENIG plating recipes and IC pre-plate preparation are required depending upon the IC manufacturing process (i.e., bipolar versus CMOS and analog versus digital). Conclusion Multiple tests were conducted to compare the effect of phosphorus content, pre-plating preparation and gold thickness on gold stud bump reliability at thermocompression bonding temperatures. The results indicated that a reliable ENIG deposit was obtained using a 7-8% phosphorus nickel with a thick gold layer. Pre-plating die preparation was highly dependent upon device technology.

References

1. Hashimoto, S., Kiso, M., Nakatani, S., Uyemmura, C., Gudeczauskas, D. (2001), Proceedings of IPC Printed Circuits Expo 2001, IPC, Northbrook, Illinois, pp.S14-3-1.

2. Masayuki, K., Gudeczauskas, D. (1997), Proceedings of IPC Printed Circuits Expo 2001, IPC, San Jose, California, pp.S16-1-1.

3. Johal, K., Roberts, H., Lamprecht, S., Wunderlich, C., SMTA Pan-Pacific Microelectronics Symposium Sheraton Kauai Resort, Kauai, Hawaii, January 25 - 27, 2005.-Published with permission.-

Share



Copyright © 2015 I-Connect007. All rights reserved.