Highly Reliable Flip-Chip-on-Flex Package Using Multilayered Anisotropic Conductive Film

Journal of ELECTRONIC MATERIALS, Vol. 33, No. 1, 2004 Regular Issue Paper Highly Reliable Flip-Chip-on-Flex Package Using Multilayered Anisotropic Conductive Film MYUNG JIN YIM, 1,3 JIN-SANG HWANG, 1 JIN GU KIM, 1 JIN YONG AHN, 1 HYUNG JOON KIM, 2 WOONSEONG KWON, 2 and KYUNG-WOOK PAIK 2 1. Department of ACF Development, Telephus Inc., Taejon, 305-343 Korea. 2. KAIST, Department of Materials Science and Engineering. 3. E-mail: mjyim@ telephus.com Anisotropic conductive lm (ACF) has been used as interconnect material for at-panel display module packages, such as liquid crystal displays (LCDs) in the technologies of tape automated bonding (TAB), chip-on-glass (COG), chipon- lm (COF), and chip-on-board (COB). Among them, COF is a relatively new technology after TAB and COG bonding, and its requirement for ACF becomes more stringent because of the need of high adhesion and ne-pitch interconnection. To meet these demands, strong interfacial adhesion between the ACF, substrate, and chip is a major issue. We have developed a multilayered ACF that has functional layers on both sides of a conventional ACF layer to improve the wetting properties of the resin on two-layer ex for better interface adhesion and to control the ow of conductive particles during thermocompression bonding and the resulting reliability of the interconnection using ACF. To investigate the enhancement of electrical properties and reliability of multilayered ACF in COF assemblies, we evaluated the performance in contact resistance and adhesion strength of a multilayered ACF and single-layered ACF under various environmental tests, such as a thermal cycling test (255 C/1160 C, 1,000 cycles), a high-temperature humidity test (85 C/85% RH, 1,000 h), and a high-temperature storage test (150 C, 1,000 h). The contact resistance of the multilayered ACF joint was in an acceptable range of around a 10% increase of the initial value during the 85 C/85% RH test compared with the single-layered ACF because of the stronger moisture resistance of the multilayered ACF and ex substrate. The multilayered ACF has better adhesion properties compared with the conventional single-layered ACF during the 85 C/85% RH test because of the enhancement of the wetting to the surface of the polymide (PI) ex substrate with an adhesion-promoting nonconductive lm (NCF) layer of multilayered ACF. The new ACF of the multilayered structure was successfully demonstrated in a ne-pitch COF module with a two-layer ex substrate. Key words: Flip chip, chip-on- lm (COF), anisotropic conductive lm, adhesion, reliability INTRODUCTION Anisotropic conductive lm (ACF) has been widely used as an interconnect material in at-panel display module packages, such as liquid crystal displays (LCDs), plasma display panel, and organic electroluminescence display, and ip-chip module packages, such as chip-on-glass (COG), chip-on- lm (COF), and (Received March 7, 2003; accepted July 21, 2003) chip-on-board (COB) because of its advantages of low-temperature processing to cure the lm below 150 C, a ne-pitch interconnect possible below 50- m pitch, an environmental friendly interconnect without lead, and ux-free or solvent-free material and processing. The COF is a new technology compared with COG and COB in the production of at-panel modules. For example, LCD module production using COF technology is in a growth stage because of its advantages 76

Highly Reliable Flip-Chip-on-Flex Package Using Multilayered Anisotropic Conductive Film 77 of ne-pitch interconnection, low contact resistance, and pretest capability compared with COG in the high-density, multifunctional LCD module, such as in mobile phones and personal digital assistants. 1 Furthermore, this technology has great potential in many other product applications that demand a nepitch interconnect and thin packages with a high level of reliability on various types of substrates, such as chip-scale packages, multichip modules, and radio-frequency integrated circuit (IC) packages. 2,3 In at-panel module fabrication, the geometry of the COF is very similar to that of tape automated bonding (TAB), and the bonding process is similar to that of COG. The difference is in the substrate, COF has a two-layer structure, normally Cu, and polyimide (PI), which is thinner, has a higher density and better exibility and is more durable in high temperature than TAB with a three-layer structure (Cu, adhesive, and PI), as depicted in Fig. 1. In COF technologies, there are several alternatives for interconnect materials and processes, such as Au-to-Sn joining, solder joining, ACF joining, and nonconductive paste. Among them, ACF joining technology has been applied as the main bonding method in LCD application. 4,5 The ACFs, which are composite materials composed of an adhesive polymer resin and conductive particles of metal-coated polymer particles, have been developed to form ne-pitch and reliable interconnects of IC bumps on ex substrates. The COF s substrate is a two-layer structure without an adhesive layer and, therefore, has a weak adhesion property for ACF materials. It is necessary to improve the adhesion property between the IC chip, ACF, and two-layer ex substrate for the reliability requirements of the COF module. In addition, nepitch interconnection is the basic requirement using the ACF method for IC packaging and other potential high-end applications. In this paper, we present a multilayer ACF that has functional layers on both sides of a conventional ACF layer to improve interface adhesion and control the bonding property during thermocompression bonding and the resulting reliability of the interconnection using the developed ACF for COF module assembly. Material Preparation EXPERIMENT The multilayered ACF was prepared by multiple coating of the release lm and slitting processes that cut the ACF to the speci c width. Figure 2a and b shows that the multilayered ACF has functional layers on both sides of the ACF layer that has conductive llers and nonconductive llers (NCFs). a Fig. 1. Schematic drawings of (a) TAB (3 layer structure) and COF (2 layer structure) lm and their applications in LCD module assembly using ACFs. b a b Fig. 2. Schematic drawings of (a) multiple-layered ACF, (b) cross-section view of ACF by SEM and (c) COF bonding process using developed ACFs. c

78 Yim, Hwang, Kim, Ahn, Kim, Kwon, and Paik Table I. Speci cation of Test Samples Test Driver IC Chip size X 5 14 mm, Y 5 1.7 mm Bump material Au (electroplated) Bump height 18 m Bump size 50 m 3 50 m Bump pitch 70 m Test Substrate Base lm PI, 22 m thick Conductor Cu, 12 m thick ACF Conductive Filler 4 m diameter Thickness 40 m Width 3 mm Functional layers have no particles inside the layer and good ow properties during thermocompression bonding. The functional layers have a low modulus of less than 100,000 dyn/cm 2 around 100 C before cure to be owed easily on the ne-pitched surface of the COF substrate. The ACF layer with conductive metal-coated polymer balls of 4- m diameter has silica llers of 0.8 m to reduce the mobility of conductive llers during bonding to have more conductive particles between the ne-pitch bumps and electrodes of the ex substrate. Figure 2c shows the COF bonding process using the multilayered (triple-layered structure) ACF in this study. The speci cation of the test samples used in this COF experiment using multilayered ACFs is summarized in Table I. ACF Characterization To characterize the material properties of multilayered ACF, we measured the curing properties and the Tg value by differential scanning calorimetry (DSC). The cured ACF samples were prepared by placing the adhesive mixture in a convection oven at 150 C for 30 min then thinning to a 0.6-mm-thick dimension for thermomechanical characterization, such as thermomechanical analysis (TMA) for the coef cient of thermal expansion (CTE) measurement, thermogravimetric analysis (TGA) for decomposition temperature, and dynamic mechanical analysis (DMA) for modulus measurement, and the moisture absorption rate test. COF Bonding Test There are three process steps for the COF bonding. First, the gold bumps on the chip and the input/output pads on the ex substrates were aligned. The multilayered ACF was then laminated on the substrate (prebonding). Finally, bonding pressure of 30 100 MPa and temperature of 210 C for 5 15 sec were applied to bond the chip on the ex substrate ( nal bonding). The chip is electrically connected to the substrate via entrapped conductive llers of the ACF. In ne-pitch interconnection, the number of conductive particles on the bump is very important to achieve low contact resistance. As the effective number of conductive particles trapped between the small area of bumps and substrate electrodes is higher, the contact resistance is lower because of the many conductive paths. Therefore, the effect of bonding parameters, such as bonding pressure, on the number of conductive particles on the bump and contact resistance were evaluated. The initial contact resistance was measured using a four-point probe, and adhesion properties of the IC chip on the ex substrate were measured using the 90 peel-strength test. Reliability Test To investigate the reliability of the COF module using multilayered ACF, the stability of the contact resistance of a single interconnect and adhesion strength of the driver IC on the ex substrate under environmental stress were evaluated. Contact resistance and adhesion strength were measured at each time interval until completion of the reliability tests. For the reliability test conditions, 85 C/85% RH high humidity and 150 C temperature condition for 1,000 h, 255 125 C thermal cycling for 1,000 cycles condition with a precondition test of 30 C/70% RH for 192 h and two times of re ow at 275 C peak temperature were used. RESULTS AND DISCUSSION Material Characterization DSC Results The DSC curves in Fig. 3a show that the curing reaction was started when the temperature reached 118 C, the typical temperature of conventional ACF. The general bonding temperature of ACF for COF is in the range of 180 220 C for 5 20 sec. Figure 3b shows the isothermal curing curves and time for full cure of the multilayered ACF at 180 C, 200 C, and 220 C. From those curves, the developed multilayered ACF can be cured 20 sec at 180 C, 15 sec at 200 C, and 10 sec at 220 C. The time for full cure of the ACF can be measured by this method. Thermomechanical Analysis Results The CTEs of multilayered ACF cured at 180 C for 20 sec were measured using TMA. The in ection point of the thermal expansion curve is de ned as TMA Tg (Tg TMA ). The Tg TMA was 121 C. The CTE of the multilayered ACF below the Tg TMA, de ned as 1, and the CTE above the Tg TMA, de ned as 2, are important parameters in determining the reliability of the COF assembly using ACF. A low CTE of adhesive can reduce the thermal strain induced by thermomechanical stress caused by thermal mismatch between chip and organic substrate, resulting in the high electrical stability during reliability test. 6 Figure 4 shows that the 1 is 100 ppm/ C higher than normal ACF for a COF of 60 70 ppm/ C for its multilayered ACF structure with NCF layers on both sides of ACF layer. DMA Results Dynamic mechanical analysis was performed to evaluate the elasticity behavior of the multilayered

Highly Reliable Flip-Chip-on-Flex Package Using Multilayered Anisotropic Conductive Film 79 a Fig. 3. DSC curves of multi-layerd ACF samples. (a) Dynamic scan at 10 C/min ramp rate and (b) Isothermal scan at different temperatures. b Fig. 4. TMA analysis of the multi-layered ACF cured. Fig. 5. DMA analysis of the multi-layered ACF cured. ACF cured as a function of temperature. The DMA test specimens were prepared by placing a multilayered ACF to be cured in the oven at 150 C for 5 min. After that, the sample was removed from the oven, cooled to room temperature, and cut into a square with dimensions of about 2 3 11 3 35 mm. The measurement was performed in single cantilever mode under 1-Hz sinusoidal strain loading. The elasticity of the multilayered ACF at 25 C was 1.45 GPa and 0.02 GPa at 150 C. The elastic modulus has an effect on adhesion strength, 7 and a modulus above 1 GPa at 25 C is preferred for high adhesion and low contact resistance. Therefore, the relatively low modulus of the multilayered ACF for ne-pitch COF application is better than the conventional ACF, which is single layered with a high density of conductive llers and a high modulus of 3 GPa based on the literature. Moisture Absorption Moisture absorption has a large in uence on the re ow stability and reliability of ip-chip-on- ex using ACF. When water is absorbed in the ACF and the ex, interfacial adhesion between chip-adhesive- ex becomes weak, and delamination occurs because of the moisture accumulation and moisture release from the adhesive in the center of the bonded area. Therefore, the amount of absorbed water and the diffusion rate along the interface are important material characteristics. 8 The amount of moisture absorption was measured by weighing the multilayered ACF cured before and after 85 C/85% RH for 500 h. An average 2.1 wt.% of moisture absorption rate was found, and this value is typical for epoxy-based ACF. Degree of Cure Measurement of Multilayered ACF Each specimen of multilayered ACF was cured in a hot plate for a speci ed time at different temperatures, the amount of heat generated was measured with a DSC unit, and the reaction rate was determined with the following formula: Reaction rate 5 (Q o 2 Q T )/Q o 3 100 where Q o is the initial amount of heat generated, and Q T is the amount of heat generated after hardening. The degree of cure of ACF is one of the critical parameters in uencing assembly reliability because

80 Yim, Hwang, Kim, Ahn, Kim, Kwon, and Paik Fig. 6. Reaction rates (Degree of cure) of multi-layered ACFs for COF package at different temperature and time. as the degree of cure of the resin in ACF increased, the peel strength of the ACF interconnect increased. Figure 6 shows that curing at 230 C for 5 sec is necessary to have 90% cure of the multilayered ACF during the COF bonding process. Over 90% cure is essential for reliable adhesion of the ACF joint during nal bonding. COF Bonding Characteristics Figures 7 and 8 show the bonding pressure effect on the contact resistance and deformation of the conductive ball of the COF module using a multilayered ACF bonded at 210 C for 5 sec, 10 sec, and 15 sec. The results indicate that the contact resistance decreases with high bonding pressure and increasing bonding time. The effect of bonding pressure on contact resistance can be explained by the conductive ball deformation, as shown in Fig. 8, scanning electron microscopy pictures of the conductive ball trapped between the Au bump of the IC and the pad of the ex in the COF module with different bonding pressures at 210 C for 10 sec. As the gap between the Au bump and pad decrease with increasing bonding pressure, the conductive ball deforms more, and the contact area between conductive ball, bump, and pad surface also becomes larger. This is the Fig. 7. Effect of bonding pressure on the contact resistance of the COF assembly using multi-layered ACF at different bonding time at 210 C. relationship between bonding pressure and contact resistance. The contact resistance is dependent on the number of conductive particles on the bump. The number of conductive particles was measured as a function of bonding pressure and is shown in Fig. 9. The number of conductive particles was not in uenced by bonding time and pressure. The contact resistance is mainly dependent on the particle deformation and reaction rate of the epoxy in the COF bonding process. As characterized in Fig. 6, it is necessary to have the ACF over 90% reacted during the bonding process for high adhesion and low contact resistance of the ACF interconnect. The adhesion properties were evaluated by a 90 peel test, as shown in Fig. 10. The multilayered ACF has better adhesion properties under a high humidity and temperature environment compared with a conventional single-layered ACF. This is due to the enhancement of the wetting to the surface of the PI ex substrate with the adhesion-promoting NCF layer of the multilayered ACF. The moisture absorption was also prevented at the interface between the multilayered ACF, IC, and ex substrate because of the better wetting property. Adhesion strength of the COF assembly using the multilayered ACF was a b c Fig. 8. SEM pictures of deformed conductive particles in the COF assemblies using multi-layered ACF bonded at 210 C for 10 seconds with (a) 20, (b) 50, and (c) 80 g/bumps.

Highly Reliable Flip-Chip-on-Flex Package Using Multilayered Anisotropic Conductive Film 81 Fig. 9. Number of conductive particles on the bump for different bonding at 210 C for 10 seconds. Fig. 11. Reliability of COF assemblies using multi-layered ACFs in thermal cycle test. Fig. 10. Adhesion properties of multi-layered ACF in the COF assembly before and during high-temperature humidity test (85 C/85%RH, 500 hrs). maintained above 600 gf/cm during the 85 C/85% RH test, compared with the COF assembly using single-layered ACFs, which was lowered below 600 gf/cm. Reliability Test Results The COF assembly using the multilayered ACF was subjected to three major environmental stresses to characterize the reliability of the assembly by monitoring of the contact resistance. Before performing the reliability test, a preconditioning test of 30 C/70% RH for 192 h and two times re ow at 275 C peak temperature was performed. Figure 11 shows the reliability result from a temperature cycle test of 255 125 C for 1,000 cycles to determine assembly integrity, such as electrical connection and adhesion between the ACF and ex when exposed to thermal stress from repeated expansion and shrinkage. The COF assemblies using the multilayered ACF and conventional ACF with a Fig. 12. Reliability of COF module using multi-layered ACFs in 85 C/ 85%RH test. single-layer structure showed stable contact resistance behavior that indicate suf cient durability under a thermal cycle environment. The contact resistance variations of COF assemblies using the multilayered ACF during the 85 C/85% RH condition and 150 C aging condition for 1,000 h are shown in the Figs. 12 and 13. The contact resistance increases of the multilayered ACF joint were in an acceptable range of around 10% during the 85 C/85% RH test compared with the single-layered ACF because of the stronger moisture resistance of the multilayered ACF and ex substrate. The contact resistances of both ACFs in the 150 C aging test were stable, con rming durability of COF assemblies. CONCLUSIONS In this paper, we presented a highly reliable COF technology using multilayered ACF with functional layers for the improvement of adhesion and reliabil-

82 Yim, Hwang, Kim, Ahn, Kim, Kwon, and Paik the single-layered ACF because of improved adhesion characteristics on the two-layer ex substrates. However, improvement of adhesion for reliability in severe environmental condition is needed. ACKNOWLEDGEMENTS Financial support from the Innovative Business Development Program under Small and Medium Business Administration and the Ministry of Commerce, Industry and Energy of Korea is gratefully acknowledged. Fig. 13. Reliability of COF assembly using multi-layered ACF in 150 C temperature storage test. ity. We manufactured a triple-layered ACF using multiple coating technology and measured its thermomechanical properties of curing property, CTE, modulus, etc. We also investigated the effect of bonding pressure, time, and temperature on contact resistance of the multilayered ACF and correlated its behavior with the number of conductive particles on the bump and reaction rate of the resin. The reliability tests of the COF assemblies of the multilayered ACF compared with the single-layered ACF showed that adhesion and electrical properties of multilayered ACF joints were superior to those of REFERENCES 1. P. Clot, J.F. Zeberli, J.M. Chenuz, F. Ferrando, and D. Styblo, Proc. of IEEE, US 24th Intl. Electronics Manufacturing Technology Symp. 1999, pp. 36 41. 2. C. Kallymayer, H. Oppermann, S. Anhock, R. Azadech, R. Aschenbrenner, and H. Reichl, Proc. Electronic Packaging and Technology Conf. (EPTC) 1998, pp. 303 310. 3. R. Fillion, B. Burdick, P. Piacente, L. Douglas, and D. Shaddock, Proc. Int. Conf. on Multichip Modules & High Density Packaging (New York: IEEE, 1998), pp. 242 246. 4. S.M. Chang, J.H. Jou, A. Hsieh, T.H. Chen, C.Y. Chang, Y.H. Wang, and C.M. Huang, Microelectron. Reliab. 41, 2001 (2001). 5. Y.C. Chan, K.C. Hung, C.W. Tang, and C.M.L. Wu, Proc. Adhesive Joining Coating Technology in Electronics Manufacturing (Piscataway, NJ: IEEE, 2000), pp. 141 146. 6. M.J. Yim and K.W. Paik, IEEE Trans. Comp. Packag. Technol. 24, 24 (2001). 7. T. Fujinawa, K. Kobayashi, M. Arifuku, and N. Fukushima, Hitachi Chem. Rep. 7, 21 (2002). 8. C.V. Veen, E. Janssen, B. Pahl, and J. Guenther, Proc. Eur. Microelectronics Packaging & Interconnection Symp. (Cracow, Poland: 2002), pp. 207 212.