Reliability of Eutectic Sn-Pb Solder Bumps and Flip Chip Assemblies

Reliability of Eutectic Sn-Pb Solder Bumps and Flip Chip Assemblies Xingjia Huang 1, Christine Kallmayer 2, Rolf Aschenbrenner 2, S.-W. Ricky Lee 1 1 Department of Mechanical Engineering Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong Phone: +852-2358-723 Fax: +852-2358-1543 e-mail: rickylee@ust.hk 2 Fraunhofer Institute for Reliability and Microintegration IZM 13355 Berlin, Germany Phone: +49-3-4643-228 Fax: +49-3-4643-161 e-mail: kallmayr@izm.fhg.de Abstract This paper presents an experimental study on the reliability of eutectic Sn-Pb (Sn63/Pb37) solder bumps and the reliability of eutectic Sn-Pb bumped flip chip (FC) assemblies mounted on an FR- 4 substrate. The growth kinetics of Sn-Ni intermetallic compound, Ni 3, on eutectic Sn-Pb solder bumped chips with Ni(P)/Au metallization was investigated. The growth of Ni 3 was found to be proportional to the square root of thermal aging time. The activation energy of Ni 3 growth was 31.23 kj/mol (.32eV). Accelerated reliability tests revealed that even after 1 cycles of the temperature cycling, the shear strength of eutectic Sn-Pb solder bumps did not change, while the shear strength showed a tendency to decrease after an extended period of high temperature storage at 125 C. The fracture mechanism for the shear test of solder bumps was a combination of fracture at the under-bump metallization/solder bump interface and in the bulk solder. For eutectic Sn-Pb solder bumped FC assemblies, accelerated reliability tests (temperature cycling (TC), high temperature storage and temperature humidity) indicated that the first 1 cycles or the first 1 hours was the most critical stage of the tests for the assemblies. In general, the temperature cycling was the most severe test. The premature failure of certain assemblies was due to weak interconnections between the bumps and the board due to cold soldering and small voids near the solder joints. For the specimens that exhibited drastic failure under the TC test, the delamination between the underfill and the passivation of the chip was the dominating mechanism. The International Journal of Microcircuits and Electronic Packaging, Volume 24, Number 2, Third Quarter, 21 (ISSN 163-1674) 246

Key words Eutectic Sn-Pb solder bump, eutectic Sn-Pb bumped FC assembly, accelerated reliability test, reliability, Sn-Ni intermetallic compound. 1. Introduction The flip chip (FC) technology has many advantages over other assembly techniques, such as better electrical performance, smaller footprints, higher I/O and lower profile [1]. In the past decade, much effort has devoted to solder-bumped flip chip on low cost organic substrate or printed circuit board (PCB). Bumping and bonding are two important steps of the FC assembly process. Electroless Ni(P) can act as not only an efficient adhesion layer and a diffusion barrier but also a solderable base on Al bonding pads of a die. Hence, electroless Ni(P) is a very promising candidate for under-bump metallization (UBM). A thin Au cover layer on Ni(P) protects the Ni from oxidation to preserve its solderability [2]. Good wettability, strong self-alignment effect and a relatively low melting point make eutectic Pb- Sn a predominant solder material for bonding assemblies onto the organic substrate [3, 4]. During the reflow process, intermetallic compounds (IMCs) form at the solder/ubm interface due to the interfacial reaction between the solder alloy and the UBM. The interfacial reaction will continue in the subsequent thermal aging, temperature cycling tests and/or even in service. The formation of IMCs is an indication of good metallurgical bonding. Nevertheless, if IMCs grow too thick, either during soldering or subsequent field aging, they may impose a deleterious effect on the long-term reliability of assemblies [5]. In the worst case, IMCs can cause the detachment of bumps completely from a chip [6]. It is imperative to study the growth kinetics of IMCs and their effect on the reliabilities of solder bumps and FC assemblies. In the present work, kinetics of the growth of Sn-Ni intermetallic compound, Ni 3, were studied. Accelerated reliability tests were performed to investigate the reliability of eutectic Sn-Pb solder bumps on Ni(P)/Au metallization and the reliability of eutectic Sn- Pb bumped flip chip assemblies in order to have a better understanding of the reliability and possible failure mechanisms. 2. Experimental Procedures Two kinds of chips, CHIP 1 and CHIP 2, were used. CHIP 1 and CHIP 2 with dimensions of 6.5 mm 6.5 mm had different via openings over Al pads where eutectic Sn-Pb solder bumps were made. Eutectic Sn-Pb solder bumps on chips were deposited by stencil printing and reflow. The UBM on the Al pads was electroless Ni(P). The Ni thickness was 5 µm, providing a nearly hermetic sealing to the pads. An immersion Au layer of.2 µm thick protected the Ni from oxidation. The FC substrate used was FR-4 PCB with a thickness of.5 mm. Each test board could carry up to six flip chips. A daisy chain pattern was designed so that the integrity of interconnects could be monitored by measuring the electrical resistance. The mounting of flip chips onto the substrate was carried out with a Fine Placer. A no-clean BS32R flux from Alpha Metals was manually dispensed using a syringe. The assemblies were then reflowed under N 2 atmosphere in a three-zone infrared reflow oven. The underfill process was performed after the reflow. Two commercial encapsulants, U 1 247

with low viscosity and U 2 with higher viscosities, were dispensed using a Cam/Alot 1414-L2 liquid dispensing system, respectively. An L-shape dispensing pattern was applied along two adjacent edges of the die. The L-shape pattern started and ended at a distance of 1 mm to the two opposite corners of the chip to avoid air entrapment. A sealing pass along the two other edges was added to obtain uniform fillet geometry around the flip chip. Temperature cycling and high temperature storage tests were conducted to investigate the reliability of the eutectic Sn-Pb solder bumps on Ni(P)-Au metallization for CHIP 2 by measuring the shear force of the bumps. Temperature cycling was performed in a Heraeus VMS 3/8/22/12 equipment from - 55 C to +125 C for 15 cycles. The duration time of each cycle was 3 min. The dwelling time at each temperature extreme was 1 min. High temperature storage was performed in a Heraeus UT62 oven at a temperature of 125 C for 15 hours. At certain intervals, dies were taken out to conduct shear test. Shear test was performed using a DAGE BT24 shear tester according to the specification of ASTM F1269-89. After the shear test, the fracture surfaces were observed using scanning electron microscope (SEM) with energy dispersive X-ray (EDX) and microprobe analyses to characterize the shear failure mode. To investigate the kinetics of the growth of Sn-Ni intermetallic compound, CHIP 2 was thermally aged at 85 C, 125 C and 15 C for 15 hours, respectively. At certain intervals, a die at each thermal temperature was taken out, mounted with epoxy and cross-sectioned. The thickness of the intermetallic compound was measured using SEM. Besides the temperature cycling and high temperature storage tests, temperature/ humidity (TH) testing was also conducted to evaluate the reliability of eutectic Sn-Pb bumped FC assemblies. Table 1 summarizes the accelerated reliability test conditions for the FC assemblies. In addition, 6 FC assemblies without underfill were also performed the temperature cycling to investigate the effect of underfill on the reliability and failure mechanism of FC assemblies. Altogether four (4) kinds of FC assemblies were conducted for the reliability tests, as listed in Table 2. During the aforementioned test program, FC assemblies were taken out at certain intervals for checking the integrity of interconnects using a SUSS PA2 Wafer Prober by measuring the electrical resistance. The contact resistance of two bumps together with the resistance of a Cu trace was also measured using Four Point Kelvin (FPK) method, as shown schematically in Figure 1. Furthermore, a scanning acoustic microscope (SAM), Sonoscan series D6, was used to detect small voids and delamination in the FC assemblies. 3. Results and Discussion 3.1 Reliability of Eutectic Sn-Pb Solder Bumps Figure 2 shows the effect of the reliability test (temperature cycling and high temperature storage) on the shear strength of eutectic Sn-Pb solder bumps of CHIP 2. There was no decrease in shear strength of eutectic Sn-Pb bump on CHIP 2 even after 15 cycles of temperature cycling test. In addition, the shear strength showed also no significant changes after long time of high temperature storage at 125 C (Figure 2(b)). 248

Table 1. Conditions of the accelerated reliability tests for FC assemblies Test items Conditions Facilities Temperature cycling -55 C / +125 C for 1 cycles Heraeus VMS 3/8/22/12 High temperature +125 C for 1 hours Heraeus VLK 8/15 storage Temperature humidity 85 C/85% RH for 1 hours Heraeus UT62 Table 2. FC assemblies for the accelerated reliability tests FC assembly Specification A-1 CHIP 1, FR-4 substrate, without underfill CHIP 1, FR-4 substrate, underfilled with U1 CHIP 1, FR-4 substrate, underfilled with U2 CHIP 2, FR-4 substrate, underfilled with U2 I 2out U out I in U in Bump U in I 1out Figure 1. Schematic illustration of Four Point Kelvin measurement SEM secondary electron (SE) and back scattered electron (BSE) images of a typical fracture surface are shown in Figures 3(a) and (b). Figure 3(c) is the high magnification of the initial area in the fracture surface. EDX analysis found that the darker grey area in the initial area of the fracture surface was the UBM, while the lighter grey area was Sn-Ni intermetallic compound, Ni 3. SEM analyses revealed that the fracture mechanism of shear test was a combination of fracture at the solder/ubm interface and in the bulk solder bump. X-ray mapping of elements Sn, Pb, Ni and P on the corresponding fracture surface is shown in Figure 4. 249

Shear Strength (gf) 4 35 3 25 2 15 1 5 2 4 6 8 1 12 14 16 Number of Cycles (a) Temperature cycling Shear Strength (gf) 4 35 3 25 2 15 1 5 2 4 6 8 1 12 14 16 Time (h) (b) High temperature storage Figure 2. Effect of accelerated reliability tests on shear strength of eutectic Sn-Pb solder bump on CHIP 2 3.2 Kinetics of growth of Sn-Ni IMC The microstructure of the eutectic Sn-Pb in the as-received state for CHIP 2 was fine and Pb-rich phase (white areas in Figure 5(a)) distributed relatively homogeneously in the whole cross section. Figure 5(b) shows the formed Ni 3 IMC phase in the as-received state of CHIP 2. With increasing the thermal aging time and temperature, the fine eutectic microstructure coarsened, and Ni 3 continued to grow, increasing in thickness, as shown in Figure 6. Figure 7 shows the 25

(a) (a) SEM SEM micrograph micrograph in in SE SE mode mode (b) SEM micrograph in BSE mode Ni 3 UBM (c) (c) High magnification of of the initial area on the fracture surface area on the fracture surface Figure 3. SEM micrographs of the fracture surface of the solder bump on CHIP 2 after 15 temperature cycles (a) Sn (b) Pb (c) Ni (d) P Figure 4. X-ray mapping of elements Sn, Pb, Ni and P on the corresponding fracture surface shown in Figure 3(a) 251

(a) Cross section of a solder (a) Cross-section of a solder bump of CHIP 2 in as-received st bump of CHIP in as-received state (b) (b) Ni Ni 3 phase 3 intermetallic compound phase formed formed in the as-received in the as-received state of CHIP state 2 of CHIP 2 Figure 5. SEM micrograph of a solder bump of CHIP 2 in as-received state (a) Coarsened microstructure of eutectic Pb-Sn solder after 15 hours at 15 C (b) Ni (b) 3 Sn Ni 43 Sn intermetallic 4 compound formed compound after formed 15 hours after at 15 15 o C hours at 15 C Figure 6. SEM micrograph of a solder bump of CHIP 2 after 15 hours of thermal aging at 15 C relationships between the thermal aging time and the thickness of Ni 3 IMC at different thermal temperatures. The growth of Ni 3 was proportional to the square root of thermal aging time, t 1/2. As a general rule, the higher the aging temperature, the greater the growth rate of Ni 3 IMC. The relationship between the aging time t and the thickness d can be expressed [7] as d = D t 1/2 252

Thickness ( m) 3.5 3. 2.5 2. 1.5 1..5. 15 C 125 C 85 C 1 2 3 4 1/2 1/2 Time (h ) Figure 7. Relationships between the thickness of Ni 3 intermetallic compound and the thermal aging time where d = the intermetallic compound thickness t = thermal aging time D = growth constant In the present study, the growth constant D for thermal aging temperature of 85 C, 125 C and 15 C was derived from Figure 7 as 9.3 1-3, 25.1 1-3 and 47.3 1-3 µm/h 1/2, respectively. Usually, over a wide range, the growth constant fits the Arrhenius relationship [7] D = D exp(-q/rt) where R = Gas constant T = absolute temperature Q = activation energy for Ni 3 layer growth Generally Q and factor D are considered to be independent of temperature. By plotting the natural logarithm of D against 1/T, as shown in Figure 8, the activation energy Q for the Ni 3 growth was calculated to be 31.23 1 3 J/mol (.32 ev), and the factor D was 33 µm/h 1/2. 253

-2.5 1/2 lnd ( m/h ) -3. -3.5-4. -4.5-5. 2.3 2.4 2.5 2.6 2.7 2.8 2.9-3 -1 1/T x 1 (K ) Figure 8. Plot of lnd against 1/T 3.3 Reliability of FC assemblies bonded with eutectic Sn-Pb bumps For A-1 FC assemblies without underfill, only the temperature cycling test was performed. It was found that all 6 assemblies failed after 5 cycles. Two types of failure mechanism were observed. One is a crack propagating through the solder joint as shown in Figure 9(a). The close view in Figure 9(b) indicated that the crack started from the interface between the UBM and the solder. Further EDX (Energy Dispersive X-ray) analysis revealed that the dark grey band along the interface was Ni-Sn intermetallic compound, Ni 3 (Figure 9(b)). Therefore, to be more precise, the crack in Figure 9(a) was originally initiated at the interface edge of the Sn-Ni IMC and Sn-Pb solder bump, and then propagated into the solder bump (Figure 9(b)). Another failure mode observed was the separation of the UBM from Al bond pad, as shown in Figure 1. Although such damage was due to the defect of the UBM deposition instead of the failure of solder, it would also lead to the malfunction of electronic device. Therefore, this failure mode is not acceptable either. Nevertheless, this failure can be overcome by optimizing the adhesion of the Al during the wafer fabrication. After underfill, the reliability of FC assemblies increased significantly. Figure 11 shows the cumulative failure rate of underfilled FC assemblies under various types of accelerated reliability tests. In general, higher percentage of failure rate was observed for FC assemblies under the TC test. It was found that about 2% 254

Ni Ni 3 Sn 3 4 (a) Crack path in in the the solder solder (b) Crack initiation initation Figure 9. SEM micrographs of the cross section of the failed FC assembly without underfill Separation Figure 1. Separation of UBM from Al bond pad 255

Cumulative Failure (%) 1 8 6 4 2 2 4 6 8 1 Cumulative Failure (%) 1 8 6 4 2 2 4 6 8 1 Number of Cycles Time (hours) (a) Temperature cycling (b) High temperature storage Cumulative Failure (%) 1 8 6 4 2 2 4 6 8 1 Time (hours) (c) Temperature humidity test Figure 11. Cumulative failure of FC assemblies under accelerated reliability tests of specimens failed in the first 1 cycles (for TC) or the first hours (for high temperature storage and TB tests). This is considered as the mortality of FC assemblies. However, this phenomenon mainly appeared for the specimens with larger via opening (CHIP 1) on Al pad and high viscosity underfill (U 2 ). For and assemblies, the premature failure only occurred under the TC test. Furthermore, for the same category of specimens ( under TC test), drastic failure was triggered after 25 cycles and eventually all FC assemblies failed after 5 cycles. Figure 12 shows the typical electrical resistance change of daisy chains of the assemblies with longer lifetime. For the assemblies passing the reliability tests, the resistances changed very little, as shown clearly in Figure 12. The relationships between the time of accelerated reliability tests and the contact resistance (including Cu trace resistance) are illustrated in Figure 13, showing also little change during the testing. As aforementioned, all FC assemblies underfilled with U1 failed after 5 cycles though they functioned normally up to 25 cycles. Such phenomenon could be understood from the SAM pictures shown in Figure 14. Before the temperature cycling test, 256

Resistance ( ) 14 12 1 8 6 4 2 2 4 6 8 1 Number of Cycles (a) Temperature cycling Resistance ( ) 14 12 1 8 6 4 2 2 4 6 8 1 Time (hours) (b) High temperature storage Resistance ( ) 14 12 1 8 6 4 2 2 4 6 8 1 Time (hours) (c) Temperature humidity test Figure 12. Typical electrical resistance change of daisy chains of FC assemblies with longer lifetime 257

Resistance ( ) Resistance ( ) 6 5 4 3 2 1 6 5 4 3 2 1 2 4 6 8 1 Number of Cycles (a) Temperature cycling 2 4 6 8 1 Time (hours) (b) High temperature storage Resistance ( ) 7 6 5 4 3 2 1 2 4 6 8 1 Time (hours) (c) Temperature humidity test Figure 13. Typical electrical resistance change of contact resistance of FC assemblies with longer lifetime 258

(a) Before temperature cycling (b) After 5 cycles (a) Before temperature cycling (b) After 5 cycles Figure 14. Typical C-SAM images of FC specimens there was no delamination in the assembly. However, after 5 cycles of temperature cycling, the FC assembly was almost totally delaminated. The lighter area in Figure 14(b) implied delamination. Further cross section analysis of the failed assembly confirmed the result of the SAM observation, as shown in Figure 15. An SEM micrograph shows that the delamination existed along the interface between the underfill and the chip. The delamination in turn resulted in the solder failure finally. Therefore, improving the adhesion between the underfill material and the chip could increase the reliability of FC assemblies. Moreover, the material combinations can be optimised by material screening experiments. In order to further understand the failure mechanism for FC assemblies with early mortality rate under the high temperature storage and temperature humidity tests, more cross section analyses were performed. From the SEM micrographs shown in Figures 16 and 17, two dominating failure modes were identified. One cause of failure was cold solder joints of solder bumps to the board (Figure 16), while the other was due to the small voids in the underfill near the solder joints (Figure 17). The former obviously resulted in imperfect interconnections and the latter led to local stress concentration in the solder. Consequently, premature failure occurred during the early stage of reliability tests. 4. Concluding Remarks The kinetics of the growth of Sn-Ni intermetallic compound, Ni 3, on eutectic Sn-Pb solder bumped chips with Ni(P)/Au metallization has been studied. Accelerated reliability tests have been conducted to investigate the reliability of the eutectic Sn- Pb solder bumps and the reliability of eutectic Sn-Pb bumped flip chip assemblies with FR- 4 substrate. The results are summarized as follows. (1) Sn-Ni intermetallic compound, Ni 3, formed at the interface of the solder bump and UBM. The growth of Ni 3 was proportional to the square root of time. In general, the higher the aging temperature, the 259

Delamination (a) Location of Delamination (b) Close view of a failed solder joint Figure 15. Cross section of a failed FC specimen Figure 16. Cold solder joint 26

Void s Figure 17. Voids near solder joints higher the growth rate of the intermetallic compound. The activation energy for Ni 3 growth was 31.23 kj/mol (.32 ev). (2) Temperature cycling did not affect the shear strength of the eutectic Pb-Sn solder bump. After long time of high temperature storage at 125 C, the shear strength of the eutectic Pb-Sn solder bump showed a tendency to decrease. For the shear test, the fracture mechanism was a combination of failure at the UBM and solder bump interface and failure within the solder bumps themselves. (3) For the eutectic Sn-Pb bump soldered FC assemblies without underfill, all specimens failed by 5 cycles. There were two failure modes; a crack initiated at the interface edge of the intermetallic compound Ni 3 and the solder bump, and separation of UBM from Al pad. (4) Underfill materials improved the reliability of FC assemblies significantly. Accelerated reliability tests revealed that the first 1 cycles (for temperature cycling) or the first 1 hours (for high temperature storage and temperature humidity tests) was the most critical stage of the testing. In general, temperature cycling was the most severe test. (5) It was identified that the failure of underfilled FC assemblies was mainly due to three causes: cold soldering of solder bumps, voids in the underfill near solder joints, and the delamination between the underfill and the chip passivation. 261

(6) In order to enhance the reliability of FC assemblies, the reflow of FC solder bumps needs to be monitored, the voiding during underfill dispensing and curing should be avoided, and the adhesion strength between the underfill and the chip passivation should be improved. Acknowledgments The Research Grant Council of Hong Kong sponsored this study through the grant HKUST6231/1E to the Hong Kong University of Science and Technology (HKUST). The authors wish to acknowledge this support. IN addition, the authors would like to express their appreciation to Mr. Carston. Nieland, Mr. Ralf Miessner, Mr. Erik Jung, Mrs. Bettina Otto, Miss Sabine Anhoeck, Mr. Joerg Gwiasda of Fraunhofer Institute IZM for helpful discussions and assistance during the experiment work. References 1. J. H. Lau, A brief introduction to flip chip technologies for multichip module application, Flip Chip Technologies, Edited by J. H. Lau, McGraw-Hill, Chapter 1, pp.1-82, 1996. 2. E. Zakel and T. Teutsch, A roadmap to low cost bumping for DCA, COF, CSP and BGA, Proceedings of IEEE/CPMT Berlin International Electronics Manufacturing Technology Symposium, Berlin, pp. 55-62, 1998 3. C. Tabubo, N. Hirano, K. Doi, H. Tazawa, E. Hosomi and Y. Hiruta, Eutectic solder flip chip technology for chip scale package, Proceedings of IEEE/CPMT International Electronics Manufacturing Technology Symposium, pp. 488-493, 1996 4. C.-J. Chen and K.-L. Lin, The reactions between Electroless Ni-Cu-P deposit and 63Sn-37Pb flip chip solder bumps during Reflow, Journal of Electronic Materials, Vol. 29, No. 8, pp. 114, 2. 5. H. D. Blair, Y.-Y. Pan, J. M. Nicholson, R. P. Cooper, S.-W. Oh and A. R. Farah, Manufacturing concerns of the electronic industry regarding intermetallic compound formation during the soldering stage, Proceedings of IEEE/CPMT International Electronics Manufacturing Technology Symposium, pp.282-292, 1996. 6. K. Kulojarvi, V. Vuorinen and J. K. Kivilahti, Effect of dissolution and intermetallic formation on the reliability of FC joints, Proceedings of 34th ISHM-Nordic Annual Conference, Oslo, Norway, pp. 51-58, 1997. 7. R. J. Klein Wassink, Soldering in Electronics, second edition, Van Nostrand Reinhold, New York, Chapter 4, pp.135-23, 1989. 262