Process Modeling and Thermal/Mechanical Behavior of ACA/ACF Type Flip-Chip Packages

Process Modeling and Thermal/Mechanical Behavior of ACA/ACF Type Flip-Chip Packages K. N. Chiang Associate Professor e-mail: knchiang@pme.nthu.edu.tw C. W. Chang Graduate Student C. T. Lin Graduate Student Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan, R.O.C. Development of flip-chip-on-glass (FCOG) assembly technology using anisotropic conductive adhesive/film (ACA/ACF) is currently underway to achieve fine pitch interconnections between driver IC and flat panel display. Conductive adhesives are characterized by fine-pitch capability and more environment compatibility. Anisotropic conductive adhesive/film (ACA/ACF) is composed of an adhesive resin and conductive particles, such as metallic or metal-coated polymer particles. In contrast to a solder type flip chip interconnection, the electric current passing through conductive particles becomes the dominant conduction paths. The interconnection between the particles and the conductive surfaces is constructed by the elastic/plastic deformation of conductive particles with contact pressure, which is maintained by tensile stress in the adhesive. Although loss of electric contact can occur when the adhesive expands or swells in the Z- axis direction, delamination and cracking can occur in the adhesive layer while the tensile stress is excessive. In addition to performing processing simulations as well as reliability modeling, this research investigates the contact force that is developed and relaxed within the interconnection during the process sequence by using nonlinear finite element simulations. Environmental effects, such as high temperature and thermal loading, are also discussed. Moreover, a parametric study is performed for process design. To improve performance and reliability, variables such as ACF materials, proper processing conditions are discussed as well. Furthermore, this study presents a novel method called equivalent spring method, capable of significantly reducing the analysis CPU time and make process modeling and contact analysis of the 3D ACA/ACF process possible. DOI: 10.1115/1.1389847 Introduction Liquid crystal displays LCD have increasingly found industrial applications owing to their low power consumption, lower weight, better compatibility with LSI chips, and more compact shape than the conventionally used cathode ray tube CRT. Having found extensive industrial use, the flip chip technology with a solder type of interconnects was developed to provide mechanical and electrical interconnection between the LSI chips and the substrates within LCDs. Despite its merits, such as including a lower profile, better electrical characteristics, and higher packaging density, this technology has its limitations, such as an expensive underfill and rework process. Thus, an advanced LCD packaging technology, with increasing throughput and fine-pitch interconnection, must be developed to effectively respond to the need for better quality of LCD panels, particularly in terms of resolution, color, and size. The Chip-on-Glass COG, as shown in Fig. 1 assembly technology e.g., Hatada et al. 1 ; Hatada et al. 2 ; Kristiansen and Liu 3 provides a simple, yet viable solution for a fine-pitch connection, largely owing to its simple, inexpensive, and repairable manufacturing process. The COG assembly generally consists of three major components: an IC chip with bumps, a circuit substrate, and an ACA/ACF that are typically made of a lightsetting insulating resin. The ACA/ACF serves mainly as mechanical bonding based on its adhesion force and electrical connection based on its shrinkage force during a curing process. When the adhesive contracts during hardening, a pressure contact is developed on the particle between the bumps. ACA/ACF is composed of an adhesive resin and conductive particles Fig. 2, such as metallic or metal-coated polymer particles. In contrast to a solder type flip chip interconnection, the electric current passing through conductive particle becomes the dominant conduction path. In addition, the interconnection between the particles and the conductive surfaces is constructed by the elastic/plastic deformation of conductive particles with a contact pressure, which is maintained by tensile stress in the adhesive. Herein, electrical conduction in the COG assembly using an anisotropic conductive adhesive occurs through the connection of the bump and the electrodes. In doing so, the contact resistance can be applied in evaluating the design quality as well as the overall reliability of a particular assembly. Similar to findings in literature e.g., Liu 4 ; Kristiansen et al. 5 ; Nicewarner 6 ; Timsit 7, Chiang et al. 8, the contact resistance between the Contributed by the Electronic and Photonic Packaging Division for publication in the JOURNAL OF ELECTRONIC PACKAGING. Manuscript received by the EPPD May 10, 2001. Associate Editor: B. Courtois. Fig. 1 FCOG with flexible circuit Journal of Electronic Packaging Copyright 2001 by ASME DECEMBER 2001, Vol. 123 Õ 331

Fig. 4 Delamination of adhesive layer of ACA package Yim and Paik 10 Fig. 2 ACF package and its electrical conduction mechanism bump and the electrode on the substrate strongly depends on the contact stress and the contact area. Therefore, the contact force can be used to assess the reliability of the assembly. To an extent, a higher reliability of the packaging relies on a better contact stability as well as larger bonding stresses. The COG Assembly Technology Using an ACAÕACF Figure 3 schematically depicts the center cross section of a typical COG assembly. In this figure, the chip is flipped, aligned, pressed, and connected to the substrate using an anisotropic conductive adhesive. In contrast to conventional solder joining types of flip chip technologies, the bumps on the LSI chips/substrate directly initiates surface contact with the metal particles based on the compressive force. This force is owing to the external pressure during the bonding process and the shrinkage force of the anisotropic conductive adhesive during the curing process. Notably, a more compliant joining mechanism can be obtained since the bump and the metal particles are not fully bonded. In this situation, the induced mechanical and thermal stresses are absorbed by the adhesive instead of the bumps/particles, thus leading to a higher reliability. Notably, the pressure bonding between the bumps and the metal pads serves mainly as electrical and thermal conductive passages from the LSI chip to the glass substrate. The adhesive applied in this method is the UV light-setting resin. Among the advantages that it has over heat curing adhesives include a faster bonding process and a lower bonding temperature that results in lower stresses on the die. Importantly, the compressive force in related applications is usually larger than the minimum requirement to ensure a higher contact reliability at a high-temperature environment. However, a higher compressive force may induce higher peeling stresses in the adhesive s interface with the substrate and with the chip. Moreover, an adhesive with high tensile and adhesion strengths is preferred not only to avert possible delamination and cracking Fig. 4, but also to secure high bonding reliability. Contact Behavior This study also examines the contact characteristics of the bumps and the conductive particle during the COG assembly process sequences by performing nonlinear finite element contact analysis. When two solid objects are forced into contact with a wide range of loads, only a finite number of small areas i.e., A-spots, Fig. 5 within these two objects are actually in contact. Instead of the entire nominal contact surface, these A- spots construct the junction of the electrical current between these two objects. Owing to the conductive particle of ACA/ACF is extremely small i.e., a diameter of around 5 m, each conductive metal particle can be assumed to act as an A-spot. For ACA/ACF type packaging, Fig. 6 illustrates the relation between contact force and electrical resistance, indicating that a larger contact force produces lower contact resistance. Although the contact force must be sufficiently large to obtain a low and stable contact resistance, excessive contact force may cause stress failure or cracking of conductive metal particles and I/O pad. This research presents an Fig. 3 FCOG manufacturing processes Fig. 5 Surface contact mechanical and electrical behavior 332 Õ Vol. 123, DECEMBER 2001 Transactions of the ASME

equivalent spring model for conductive particle to effectively estimate the contact force, contact resistance and long-term reliability of ACA/ACF packaging. The contact electrical resistance (R cr ) between a single particle and I/O bump surface can be expressed as R cr 1 2 4r r 3 3PE*r 1 4 (1) (2) E* 1 1 2 1 2 2 (3) E 1 E 2 where are the electrical resistance coefficients of I/O bump and particle, respectively, r 1 is the radius of contact surface of metal particle, P is the external loading, and E 1,E 1, are the Young s modulus, poisson s ratio of particle and I/O bump, respectively. Furthermore, if there are N particles between two I/O bumps, the average contact pressure for each particle can be assumed as P/N. Therefore, Eq. 2 can be rewritten as Fig. 7 Equivalent spring model r 3 3 PE*r 1 4 N Moreover, substituting Eq. 4 into Eq. 1 leads to Eq. 5 (4) R cr 1 4 4N 1 2 3 (5) 3r 1 PE* The foregoing equation indicates that the electrical resistance can be estimated if the material properties of two contact bodies, contact pressure, and radius of the deformed surface are known. Finite Element Modeling of ACA package An equivalent spring method is developed herein to accurately predict the contact force and contact surface area of the ACA/ACF metal particles Fig. 6. This methodology can significantly reduce the analysis CPU time and make large-scale ACA/ACF nonlinear contact analysis possible. Figure 7 presents the 2D plane strain metal particle model used herein. Based on the contact force versus displacement curve of 2D model, a 1D equivalent spring model can be developed to match the mechanical behavior of the 2D particle model. Once the 2D model validates the single equivalent spring, the equivalent spring Fig. 8 can replace the particles of the ACA/ACF package for finite element contact analysis. However, a 2D particle model will remain to obtain the contact area data. Fig. 6 Contact pressure versus contact resistance Fig. 8 System level equivalent spring model Process Modeling and Validation of Equivalent Spring Model A special modeling technique, commonly referred to as the element-dead and element-birth method, is applied for the ACA/ ACF processing finite element analysis. ACA/ACF processes normally comprise of four major steps Fig. 3 : 1 Apply the external pressure with a high temperature, e.g., 125 C, on the top-side of the die. In this stage, the adhesive is in liquidity status. For finite element analysis, the adhesive is in the element-dead status, particles and pads are in contact condition; 2 Irradiate the UV light on the adhesive through the substrate side to accomplish a light setting for the adhesive. The adhesive is then cured at 125 C and, finally, transferred from element-dead to element-birth status; 3 Remove the external pressure loading at 125 C; and 4 Cool down the whole device to room temperature of 25 C. Table 1 lists the material properties of ACF/ACA used herein. To reduce the FEM analysis CPU time, a benchmark 2D model of conductive particle must be constructed to get the 1D equivalent spring force-displacement curve. Figure 7 illustrates a nickel metal particle with 5 m diameter, in which the lower and upper pads are gold metal with 15 m thickness. After an external loading is applied on the above model, the results of contact force versus displacement and contact radius versus contact force are summarized in Fig. 9 and 10. Figure 11 depicts a typical deformation of conductive particle. Incorporating these results into Eq. 1 one could get the curve of electrical resistance versus contact force,and the result is shown in Fig. 12. Notably, in Fig. 9, the Journal of Electronic Packaging DECEMBER 2001, Vol. 123 Õ 333

Table 1 Material properties Fig. 11 Typical contact deformation of conductive particle contact force versus displacement pattern of loading/unloading closely resembles the experimental data Fig. 13, Fu et al. 9. Furthermore, the curve of Fig. 9 is used as the mechanical behavior of 1D equivalent spring, in which the material properties of equivalent spring model are adjusted so that the forcedisplacement curve is the same as in Fig. 9. Fig. 9 Contact force versus displacement Fig. 12 Contact resistance versus contact force Fig. 10 Contact radius versus contact force Fig. 13 Experimental result of conductive particle 334 Õ Vol. 123, DECEMBER 2001 Transactions of the ASME

Fig. 14 2D ACAÕACF multiple particles system Fig. 15 Average contact forces at each step Fig. 16 3D conductive particle model to 1D equivalent spring model Once the 1D equivalent spring material behavior is obtained, this model can be used to perform contact analysis of multiple conductive particles ACA/ACF system. Herein, two benchmark models were constructed to confirm that the equivalent spring concept works for the multiple particles system. Model 1 Fig. 14 is a true 2D model for ACA/ACF system with 4 conductive particles. In addition, the configuration of model 2 is the same as in model 1, except all of the 2D particles were replaced by the 1D equivalent spring model. The process modeling steps of ACA/ACF type flip-chip packages are as follows: Step 1: Apply a compressive loading of 2 kg/cm 2 at 125 C, and the adhesive is in liquidity status. Step 2: Release the external loading and cure the adhesive at 125 C adhesive is in solid condition Step 3: Reduce the packaging temperature to 25 C Figure 15 summarizes the results of average contact force at each step per particle for 2D and equivalent spring model. According to this figure, the average contact forces of these two models are extremely close, indicating that the equivalent spring method could be a highly effective model for ACA/ACF multiple particles analysis. A 3D particle model Fig. 16 is also developed for equivalent spring method. According to our results, although the contact force versus the displacement Fig. 17 has a similar pattern as the 2D particle model Fig. 9, the 3D model more closely corresponds to the experimental results Fig. 13 than the 2D model when the displacement is lower than 0.06 m. Once the mechanical behavior of 1D equivalent spring model is understood, the model can be easily applied to 3D ACA/ACF Fig. 18 stress/ strain analysis for the process and reliability study. To optimize the process parameters, parametric analysis is necessary to obtain the optimum process design of the ACA/ACF package. This highly effective equivalent spring model can be used in design of 3D ACA/ACF process parameters. Fig. 17 Contact force versus Displacement of 3D model Journal of Electronic Packaging DECEMBER 2001, Vol. 123 Õ 335

Fig. 19 Manufacturing compressive loading effect Fig. 18 3D ACAÕACF model Process Parameters Versus Contact Force Fig. 20 Environmental temperature versus contact force at two different processing temperatures Fig. 21 Environmental temperature versus contact force for two different adhesives Development of flip chip package assembly technology using anisotropic conductive film ACF is currently underway to achieve fine pitch interconnections between driver IC and flat panel display. In contrast to solder type flip chip interconnection, the electric current passing through conductive particle becomes a dominant conduction path. Insufficient electric contact can occur when the adhesive expands or swells in the Z-axis direction. Herein, nonlinear finite element simulations are used to investigate the contact force developed and relaxed within the interconnection during process sequence as well as the environmental effects such as high temperature thermal loading. Moreover, processing modeling is performed to yield a better electrical and reliability performance, the effect of process variables such as ACA/ACF materials, processing pressure and temperature are investigated. A 3D FCOG package with 15 conductive particles under each I/O bump using equivalent spring technique Figs. 16 and 17 is modeled as shown in Fig. 18. The processing simulation starts from steps 1 3 and, after the manufacturing processes, the package gradually increases the environmental temperature from 25 C 155 C. Doing so allows us to examine how the process parameters affect the contact force behavior of ACA/ACF particles. Figure 12 reveals that the required minimum contact force for the package is around 0.5mN. Figure 19 shows the contact force of three different ACA/ACF packages when they are manufactured under different process pressures. Simulation results indicate that a higher process pressure could produce a better contact force when the package is subjected to thermal loading. The contact force falls below 0.5 mn if the environmental temperature is higher than 107 C for 2 kg/cm 2 process pressure, 117 C for 3 kg/cm 2 process pressure and 130 C for 4 kg/cm 2 process pressure. Furthermore, Figs. 20 and 21 reveal that a higher manufacturing temperature and stiffer ACA/ACF adhesive layer imply a better contact force when the package is subjected a thermal loading. Notably, the designer should carefully design the process pressure and select the appropriate ACA/ACF material since the excessive pressure loading might damage the conductive particles permanently. Furthermore, a stiff adhesive layer might generate an excessive peeling stress Fig. 4, subsequently incurring delamination along the layer interface. 336 Õ Vol. 123, DECEMBER 2001 Transactions of the ASME

Conclusions This study presents a novel element-dead, element-birth and nonlinear contact finite element analysis process modeling technique. The methodology proposed herein can be used to analyze the mechanical and electrical resistance behaviors of conductive particle during manufacturing. Furthermore, a 1D equivalent spring model is developed for 2D conductive particle, in which the method can significantly reduce the analysis CPU time with results similar to the 2D finite element model for ACA/ACF multiple particles system. Furthermore, a 3D equivalent spring model is also developed. The equivalent spring method can significantly reduce the 3D model CPU time as well as make the 3D parametric and reliability analysis of ACA/ACF COG package possible. Acknowledgment The author would like to thank the National Science Council, Taiwan, R.O.C., for financially supporting this research under grant NSC89-2212-E-007-105. References 1 Hatada, K., and Fujimoto, H., Kawakita, T., and Ochi, T., 1988, A New LSI Bonding Technology Micro Bump Bonding Assembly Technology, IEEE- CHMT Proceeding, pp. 45-49. 2 Hatada, K., Fujimoto, H., Ochi, T., and Ishida, Y., 1990, LED Array Modules by New Technology: Microbump Bonding Method, IEEE Trans. Compon., Packag. Manuf. Technol., Part A, 13, No. 3, Sept., pp. 521-527. 3 Kristiansen, H., and Liu, J., 1997, Overview of Conductive Adhesive Interconnection Technologies for LCD s, PEP 97, IEEE, pp. 223-232. 4 Liu, J., 1996, On the failure mechanism of anisotropically conductive adhesion joints on copper metallization, Int. J. Adhesion and Adhesives, 16, No. 4, pp. 285-287. 5 Kristiansen, H., Gulliksen, M., Haugerud, H., and Friberg, R., 1998, Characterization of Electrical Contacts Made by Non-Conductive Adhesive, Proceeding of Adhesive Joining and Coating Technology in Electronics Manufacturing 3rd International Conference, Sept. 28 30, 1998, pp. 345-350. 6 Nicewarner, E., 1999, Interconnect resistance characteristics of several flipchip bumping and assembly techniques, Microelectron. Reliab., 39, pp. 113-121. 7 Timsit, R. S., 1999, Electrical Contact Resistance: Properties of Stationary Interfaces, IEEE Trans. Compon., Packag. Technol., Part A, 22, No. 1, Mar., pp. 85-98. 8 Chiang, K. N., Chang, C. W., and Lin, C. T., 2000, Analysis of ACA/ACF Package Using Equivalent Spring Method, Proceeding of EPTC 2000 IEEE, Singapore, Dec. pp. 110-116. 9 Fu, Y., Wang, Y., Wang, X., Liu, J., Lai, Z., Chen, G., and Willander, M., 2000, Experimental and Theoretical Characterization of Electrical Contact in Anisotropically Conductive Adhesive, IEEE Trans. Compon., Packag Manuf. Technol., Part B, 23, No. 1, Feb., pp. 15-21. 10 Yim, M. J., and Paik, K. W., 1999, The Contact Resistance and Reliability of Anisotropically Conductive Film ACF, IEEE Trans. Adv. Packag., 22, No. 2, May, pp. 166-193. Journal of Electronic Packaging DECEMBER 2001, Vol. 123 Õ 337