I \ Moisture Effects on the Reliability of Anisotropic Conductive Films

Transcription

1 Moisture Effects on the Reliability of Anisotropic Conductive Films CY. Yin', H. Lu', C. Bailey, Y.C. Chan' #Computing and Mathematical Sciences School, The University of Greenwich, London, SEI0 9LS, UK *Department of Electronic Engineering, City University of Hong Kong + Tel: +44(0) , Fax: +44(0) Abstract Anisotropic conductive film (ACF) which consists of an adhesive epoxy matrix and randomly distributed conductive particles are widely used as the connection material for electronic devices with high U0 counts. However, for the semiconductor industry the reliability of the ACF is still a major conc1:m due to a lack of experimental reliability data. 'This paper reports an investigation into the moisture eficts on the reliability of ACF interconnections in the flip-chip-on-flex (FCOF) applications. A macro-micro 3-D finite element modeling technique was used in order to &e the multi-lengthscale modeling of the ACF flip chip possible. The purposes of this modeling work was to understand the role that moisture plays in the failure of ACF flip chips, and to look into the influence of physical properties and geometric characteristics, such as the coefficient of the moisture expansion (CME), Yoimg's modulus of the adhesive matrix and the bump height on the reliability of the ACF interconnections in a humid environment. Simulation results suggest that moisture-induced swelling of the adhesive matrix is the major cause of the ACF joint opening. Modeling results are consistent with the findings in the experimental work. 1. Introduction Anisotropic conductive Films (ACFs), more appropriately referred to as anisotropic conductive adhesive films (ACAFs) have been introduced as a promising flip chip interconnection material, due to its potential in achieving high density VO interconnection, low processing temperature and relatively mild impact on the environment. In particular, devices with flip chip on flexible substrate (FCOF) using ACFs are now widely used in smart cards, disk drivers and driver chips for LCDs [ 13. A typical ACF flip chip is shown in Figure 1. The conductive particles tmpped between the metallizations connect chip bumps and substrate pads Si chip Adhesive1 I \ Substrate Conductive Figure]. A typical ACF ini.erconnection system electrically. The mechanical integrity of the assembly is maintained by the adhesives. The adhesives also keep the trapped particles in a compressive state so that good contact between the particles and the bumpdpads. In spite of the increasingly important role of ACFs in the assembly of electronic products, there are still concerns about the reliability of any device with ACFs in it. After all, when compared with solder interconnect material, ACFs are new materials with many unknown properties. Among the many factors that affect the reliability of ACF devices, moisture is one of the most important ones. Previous studies, in experiment and modeling, have revealed that the reliability of ACF is strongly affected by moisture and it is even thought of as the dominant factor in ACF flip chip failures [2-51. However, previous modeling works [2-4], were mostly limited to the analyses of simplified 2-D models and 3-D effects were ignored. Even in the work in which 3-D models of ACF joints were used [6], the shear stress caused by the mismatch in the coefficient of thermal expansion (CTE) could be underestimated because the global effect of the model was ignored. The difficulty here is that due to the vast range of length-scales in an ACF flip chip, and the huge amount of conducting particles, an 'exact' model which includes all the particles and interconnections is simply not achievable with today's computer technology. In this paper, a 3-D macro-micro modeling method was used to overcome the difficulty caused by the multilength-scale nature of the problem. Two models, macro and micro models, with different mesh sizes were built. The macro model was used to predict the overall behavior of the whole assembly under moisture conditions. The displacement obtained fiom the global model was then used to set the boundary condition of the micro model so that the detailed stress analysis in the region of interest can be carried out. With this technique, the influence of physical and geometric parameters such as the coefficient of the moisture expansion (CME) and Young's modulus of the adhesive matrix, the bump height on the reliability of ACF interconnections was studied. 2. Computer modeling A typical FCOF assembly with a 1 I mmx3mmxlmm chip and a total of 368 Ni-Au peripheral pads was used throughout the analyses. The dimensions of a bump is 50pmx50px20p. The conductive particles in the ACF are assumed to be nickevgold coated poiymer particles &10$/$ s IEEE -I62-6th. Int. Con$ on ThennaL MechanicaE and Multiphysics Simulation and Erperiments in Micm-Electronics and Micro-System, EumSimE 2005

2 with a diameter of 3. 5 each. ~ The substrate is 25pm thick polyimide and the metal pads on the substrate are made of nickel-gold plated copper. The thicknesses are 12, 4 and 0. 5 for ~ the copper pad, nickel and gold coatings respectively. The gold layer was neglected in the model. The layout of the chip is shown in Fig.2. PHYSICA, a multi-physics software package, has been used in this modeling work [7]. Figure 2. The layout of the Si chip o( Methodology The temperature, humidity and pressure of the Autoclave test condition for this modelling work are 121 C, 1OO%RH and lam respectively. During the Autoclave test, it is assumed that the moisture uptake in the materials follows the Fick s Law of Diffusion [2,31 (see Equation 1). ltjk = il&,sjk -E 2,uE jk - Where 1, v, E and,u are the Lame constant, Poisson s ratio, Young s modulus and shear modulus respectively, /I is the CME, a is the CTE, dt is the temperature change, and C is the moisture concentration. The material properties used in this modeling work were based on the work described in [3]. A macro-micro modeling method was used in order to make the full scale modeling analysis of the ACF flip chip possible. At the package level (macro model), a coarse mesh was used to predict the displacement and moisture concentration through the assembly; at the ACF joint level (micro model), a tiner mesh was used that captures the detail of an ACF particle. The displacement extracted from the macro model will be used as the boundary condition of the micro model and the detailed stress analysis will be performed using the micro model. 2.2 Global modeling The macro model of the ACF flip chip is shown in Fig. 3. Only one quarter of the package was simulated due to the symmetry of the ACF package. The model consists of a total of brick elements. 8G dzc d2c a2c -= D(---t-+-) at axz +2 az2 Where C is the moisture concentration, decribbg the ability of material to absorb moisture, D is the coefficient of moisture diffusion, decribing how fast the mass diffuses. Unlike temperature distribution, which is continuous, the distribution of the moisture concentration C is discontinuous at material interfaces because the saturated concentration values are not the same for different materials. This problem can be easily solved by using the wetness faction approach [8]. When the temperature changes in an assembly made of more than one material, stresses will build up because of CTE mismatch. Similarly, when moisture absorption takes in a ACF assembly, stresses will also build up due to the mismatches in the coefficient of moisture expansion (CME). These are calied hygroscopic stresses. Assuming that the mechanical, thermal, and moisture induced strains are independent of each other, equation 2 shows that the mechanical strain is the total strain less the thermal strain and the hygro strain [Z]. The stresses can be calculated from Equation 3. Figure 3. The mesh of the macro model ACF flip chip bonding is a complicated process which involves heat transfer, fluid flow and solid deformation [9]. In order to simplify the analysis, the residual stresses created in the bonding process are assumed to be negligible. This means that the model is stress free at the reference temperature. As a further assumption, all materials in the model were assumed to be homogenous, isotropic, and the temperature dependence of material properties was neglected. All the metal materials were assumed to be elastic-plastic, and other materials were treated as elastic materials. The moisture diffusion analysis was coupled with the stress analysis s6 that the displacement field and the moisture concentration were solved simultaneously. The Von-Mises stress distribution and the warpage of the th. Int. Con$ on Thermal, Mechanical und Multiphysics Simulation and Experiments in Micro-Ekcironics and Micro-Systems, EumSimE 2005

3 whole package due to the moisture absorption are shown in Fig. 4. The package bends because the flex substrate absorbs water and expands while the chip absorbs virtually no water at all and therefore does not expand. the flex substrate and adhesive meet. In autoclave experiments, this is usually the location where micro cracks are found..'i !lS c i ri ~, x I Figure 4. the deformation and stress distribution of the whole package 2.3 Local modeling The mesh of the micro model is shown in Fig. 5. This model represents the ACF joint located at the comer of the ACF flip chip where the stress level is usually the highest. A representative conductive particle was included and finer elements were employed around the particle in order to capture the detail of the p;uticle. A total of elements were used in this micro model. NUAu Bump { NilCu Pad z i" Figure 5. Pattern of the Von-Mises stress distribution The displacement, temperature and moisture concentration results &om the macro model were used as the boundary conditions of the local model. In this work, the boundary mesh of the micro model is finer than in the corresponding parts of the macro model so that the nodes of the two models do not coincidr: at the boundaries. An interpolation technique had to be used to transfer the displacement values from the macro model to the micro model's boundary nodes. a. Moisture induced stresses The Von-Mises stress distribution in this ACF joint is shown in Fig. 6. Stress concentriition was found at the interfaces between the adhesive and the bumplpad. The stress level is the highest around {he spot where the pad, Figure 6. Pattem of the Von-Mises stress distribution The interfacial stresses along the top surface of the substrate pad are presented in Fig. 7. It was found that both the normal and the shear stresses were higher in the area around the conductive particle. The shear stress is not significant compared to the normal stresseven though 1.4Et02 -Normal stress 1.2E+02 -Shear stress 1.OE+02 B.OE+Ol 23 6.OE+01 - $ 4.OE+01 2.OE+Q1 O.OE+OO -2.OE+OD OE+01-6.OE+01 Position along the pad (m Figure 7. The interfacial stresses with moisture OdY the modeled ACF joint is located at the corner of the flip chip. The loading condition around the conductive particle is mostly tensile. This means that due to the moisture absorption of the adhesive, the ACF swelling effect pushes the die upwards resulting in the high stresses at the interface between the conductive particle and metallization. The electric connection between conducting particles and the surrounding metallization are formed through the contact pressure caused by the elastic/plastic deformation of the particles. This contact pressure is maintained by the residual stress in the adhesive. The loss of the electrical contact may occur when the adhesive expands excessively in the vertical direction [l]. The joint opening occurs when the contact area is completely lost. The normal stress distribution around the conductive particle is shown in Fig 8. Higher stress was found at the interfaces between the conductive particle and adhesive matrix. -I66 6th. lnt. Con$ an Thennal, Mechanical and Mulriphysics Simulation and fiperiments in Micro-Elecironics and Micro-System, EuroSimE 2W5

4 and the pad with different adhesive CME values (CME1=200 mm3/g, CME2=400mm3/g and CME3-800 mm3/g) were recorded at a fixed moisture concentration and the results are shown in Fig. 10. It was found that the CME of adhesive matrix had a great effect on the stresses at the interface. The interfacial stress increases with the CME of the adhesive matrix. Therefore higher adhesive matrix CME value results in an increased possibility of joint opening. Figure S. Pattem of the tensile stress distribution b. Temperature induced stresses Apart from the moisture-induced stresses during the autoclave test, the temperature-induced stresses due to the CTE mismatch need to be simulated and analyzed as well. The final bonding temperature of 190T was used as the stress-fiee reference temperature. Fig.9 shows the normal stress distribution in the ACF joint. The result suggests that the interfaces between the conductive particle and metallization are under compression so that the interface cracks wilt not open and the compressive forces can help maintain the good contact between the conductive particles and metallization j , E-647 i-726 i-~es 1."... ". I Figure 9. Pattem of the thermal stress distribution 4. Optimization analysis The modeling results presented above have proved that the reliability of ACF is strongly affected by moisture rather than temperature during the Autoclave test. In order to improve the reliability of ACF flip chips in such circumstances, a detailed parametric analysis was carried out. The effect of the CME, Young's modulus of the adhesive matrix and the bump height on the interfacial stress was analyzed. 4.1 Effect of the CME of the adhesive matrix According to the Equation 2. The value of hygroscopic stress is proportional to the product of CME and moisture concentration. The CME of the adhesive, p, was chosen as an example to show how moisture property affect the reliability of the ACF flip chips. The values of the normal stress q,. at the interface between the particle - 3.OEW2 2.5E+02 2.OE & 1.5Ei OEa2 I E 5.0E+01 ti O.OE+QO -5.OE+01-1.OE+02 O.O +0O 50E-04 1.OE43 1.5E43 2.0E E-03 Along the interface (m) Figure 10. The interfacial stresses with different CME values of the adhesive matrix. Also according to Equation 2, hygroscopic stress is proportional to the moisture concentration C. In the autoclave test, moisture saturation occurs quickly so the maximum moisture concentrations in the assembly are dictated by the materials' saturated moisture concentration, Csa,, Therefore, to reduce the damaging hygroscopic stress both B and C,, should be minimized. 4.2 Effect of the Young's modulus of the adhesive matrix Previous research work [3j has revealed that ACF materials with high Young's modulus can significantly reduce the joint opening of ACA interconnections in humid environment. To prove this and to estimate the extent at which the Young's modulus has on the reliability of ACF joints, the values of stress ow along the particleipad interface are shown in Fig. 11 for three different Young's modulus values. It was found that the r 2.OEW2 1.5EW2 I.OE+02 a 5.OE+01 m 3 O.OE+OO S.OE+Ol +-E= 725Mpa 1 -E = 1450 Mpa +E =29WMpa O.OE+OO 5.OE-W 1.OE E E E-03 Abng the interface (mm) Figure 11. The interfacial stresses with different Young's modulus value of the adhesive matrix - I65-6th. Inf. Coni on %ml, Mechanical and Multiphysics Simulation and Experiments in Micro-Electronics and Micro-System, EuroSimE 2005

5 interfacial stress decreases significantly as the Young's modulus of the adhesive increases, therefore the reliability of ACF flip chips can tie improved if the ACF with higher Young's modulus is used. One possible explanation of this phenomenon is that the ACF with higher modulus deforms less which can lead to less mismatch in moisture expansion between the adhesive and metal pads [3]. 4.3 Effect of the bump height The bump height was chosen to be as an example to show how geometric design parameters may affect the reliability of ACF flip chips in a humid environment, Models with three different bump heights were used for this analysis. The pad height was kept at a fixed value. The values of the stress ayy at the interface between the particle and the pad with different bump height are shown in Fig.12. It can be seen that the interfacial stress shows a modest increase as the bump height increases. This suggests that the iduence of this parameter is not as significant as the other two parameters discussed earlier. However, if thinner bumps are not difficult to achieve, then it is still a useful means of enhancing the reliability of the ACF flip chips in a humid eiivironment. 5. Experimental results In order to validate the computer modeling results, experiments were also used to study the moisture effect on the reliability of ACA flip Chip. Twenty flip chips on flex using ACFs were assembled and the samples were tested under the 121"C, loo%rh, latm conditions for up to 168 hours. Fig.13 shows the SEM photos of an ACF interconnection as manufactured and the ACF interconnection that has undergone 48 hours' of autoclave test. As shown in Fig. 13a, conductive particles have good contacts with the conductive metallization surfaces before the autoclave test. But as shown in Fig. 13b, the conduction gap between the conductive particles and the metallization was clearly visible after the 48 hours autoclave test. The formation of this conduction gap signals a loss of the contact area and this leads to an elevated contact resistance. Moreover, cracks can be seen along the adhesiveiflex interface and the interface between the flex and substrate pad. The propagation of these cracks may eventually cause the total failure of the ACF joints. 12E+02 I.OE+M 1 p" B.OE+Ot 6.0E+01 : g 4.0E E+01 O.M+W -2.DE+01 + bunp height is 6um -cr bunp height i!;lzum O.OE+OO 5.OE-W 1.OE Ed3 2.OE Plong the Interhe (nm) Figure 12. The interfacial stmses with different bump height Since the moisture properties of the polyimide substrate are not available for this modeling analysis, two groups of ad hoc material properties were used for trend analysis. These properties are listed in Table I. Since the chip absorb virtually no water, the whole package always shows the upwards deformation due to the swelling effect of the substrate. The extent of the warpage depends on how much water substrate absorbs and how high the CME value of the substrate is. When the second group of properties was used, the package showed greater deformation, and higher stress in the ACF joint at the corner of the package. The stress distribution was not affected very much. Figure 13. SEM photos showing the ACF interconnection with conductive particle (A) before Autoclave test (b) after 48 hours autoclave test In general, there are two possible causes for the increase in the contact resistance of ACF joints that have undergone autoclave tests: debonding at ACF joints and corrosion of the metallization. Both can reduce the -i66-6tk Inr. ConJ on ThermaI, Mechanical andmultiphysics Simulation and Experiments in Micro-Electronics and Micro-System, EuraSimE 2005

6 contact area between the conductive particles and metallization 151. For the samples used in ths work, the metal pads have been plated with a gold layer so that the effect of the electrochemical corrosion is negligible. This makes the debonding caused by ACF swelling when water is absorbed as the dominant degradation mechanism. According to the modeling results, higher stress was identified along the interface between the fledadhesive and the interface between the adhesivekhip. However, according to the experimental results, most cracks were found along the substrate pad side. The possible reason is the weak adhesion strength at the interface between the epoxy adhesives and the polyimide substrate due to the similar surface tension energy values of these two materials. Epoxy has a relatively low surface tension energy compared to Au, Si02 and many other materials, therefore a good adhesion is usually expected when it applied to the surface of these materials[lo]. Therefore, the interface between the adhesive and flex is more vulnerable to water molecule attacks due to the respectively weak adhesion strength. Once the degradation at the adhesiveiflex interface has started, water can move along the interface, combing with the stress due to the swelling effect of the adhesive, this may came a further degradation of the interfaces. 6. Conclusions Finite element analysis on the reliability of ACF flip chips under autoclave test conditions was studied. The macro-micro modeling technique was used to overcome the multi-length-scale problem so that the full scale analysis of an ACF flip chip could be done. The modeling results showed that the reliability of ACF is strongly affected by moisture. Due to the CME mismatch, higher stress was found at the interface between the adhesive and bumpjpad. The predominantly tensile stress found at the interface between the conductive particle and metallization reduces the contact area, while the thermal stress has the opposite effect. Through the parametric analysis, it was found that the stresses at the particleipad interface increase with the CME of the adhesive matrix and decrease with the Young s modulus of the adhesive matrix. The interfacial stress showed slightly increase with bump height. It is believed that higher stresses would cause more loss of contact area and therefore a bigger increase in the contact resistance. The modeling results are consistent with the findings in experiments. In the experiments, most cracks were found along the interface between the ACF and flex substrate, few was found at the chip side. The weak adhesion strength between the ACF and flex substrate due to the similar surface tension energy was thought to be the cause. Acknowledgments The authors would like to acknowledge de financial support from the University of Greenwich and City University of Hong Kong. C. Y. Yin would also like to thank Miss Sai Choo Tan of City University of Hong Kong for her help in SEM experiments and helpful discussions, References 1. Y. C. Chan, K. C. Hung, C. W. Tang and C. M. L. Wu, Degradation Mechanism of Anisotropic Conductive Adhesive Joints for Flip Chip on Flex Applications Adhesive Joining and Coating Technology in Electronics Manufacturing, Proceedings of the 4~ International Conference on, 2000 Page(s): Lei L. Mercodo, Jeny White, Vijay Sarihan and Tien- Yu Tom Lee, Failure Mechanism Study of Anisotropic Conductive Film (ACF) Packages, IEEE Trans on Components and Packaging Technologies, Vol. 26, NO. 3,2003, pp Zhou Wei, Low Siu Waf, Neo Yong Loo, Eng Meow Koon, Mark Huang Studies on Moisture-induced Failures in ACF Interconnection, Electronics 4. Packaging Technology Conference, 2002, pp J.F.J.M. Caers, X.J.Zhao, G.Lekens, R.Dreesen, K. Croes and E. H. Wong Moisture Induced Failures in Flip Chip on Flex Interconnections using Anisotropic Conductive Adhesives, International IEEE Conference on the Business of Electronic Product Reliability and Liability, C.W. Tan, Y.C. Chan and N. H. Yeung Effect of Autoclave Test on Anisotropic Conductive Joints Microelectronics Reliability 43(2003): C.Y.Yin, H. Lu, C. BaiIey and Y.C. Chan, Experimental and Modeling Analysis on the Reliability of Anisotropic Conductive Films, Electronic Components and Technology Conference, PHYSICA Multi-Phy sics Ltd, uk/-physica 8. E.H. Wong, Y. C. Teo, T.B. Lim, Moisture Difhsion and Vapour Pressure Modelling of IC Packaging, Electronic Components and Technology Conference, D.C. Whalley, L. Chen, J. Liu Modelling of the Anisotropic Adhesive Assem a1 Symposium on Electronic Packaging Technology, Beijing, August 2001, pp , ISBN bly Process 10. Johan Liu et al, Conductive Adhesivs for Electronics Packaging, Electron chemical PubIications Ltd(London, U.K , pp rh. Int. Conf on Theml, Mechanical and Multiphysics Simulation and Experiments in Micro-Ekrfronics and Micm-Systems, EumSimE 2005