Behaviour of anisotropic conductive joints under mechanical loading

Behaviour of anisotropic conductive joints under mechanical loading

Microelectronics Reliability 43 (2003) 481–486 www.elsevier.com/locate/microrel Behaviour of anisotropic conductive joints under mechanical loading C...

289KB Sizes 2 Downloads 26 Views

Microelectronics Reliability 43 (2003) 481–486 www.elsevier.com/locate/microrel

Behaviour of anisotropic conductive joints under mechanical loading C.W. Tan, Y.C. Chan *, N.H. Yeung Department of Electronic Engineering, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong Received 25 July 2002; received in revised form 28 September 2002

Abstract Anisotropic conductive films (ACFs) received a great deal of attention in recent years for interconnection applications in electronic packaging. This paper reports the behaviour of ACF joints under various mechanical loading, i.e., die shear and cyclic fatigue in shear. The mechanical behaviour of ACF joints that have been exposed to environmental effects, i.e., high moisture and elevated temperature (autoclave test conditions) has been examined using die shear test. The maximum shear force prior to fracture was determined as 465.0 N, at which the contact resistance is found extremely high and can be considered as open circuits. Epoxy based ACF exhibits insignificant plastic deformation, especially for samples that have undergone autoclave test. Reduction trend was observed in the shear moduli over autoclave test time for ACF joints. Fracture surface of ACF that failed in shear test shows spalling and less plastic deformation after exposed to autoclave test. For cyclic fatigue test, the endurance limit is determined at about 143.5 N and the corresponding calculated endurance ratio is around 32.0%. Ó 2002 Elsevier Science Ltd. All rights reserved.

1. Introduction Flip chip assembly technologies have been of much interest because they are not only making electronic products thinner and smaller but also offer reduction in interconnecting distance and inductance. Increasing attention has been paid to the conductive adhesives, which are used as interconnecting materials for electronics applications. This is because conductive adhesive joint offers many advantages when comparing with traditional lead-containing soldering technology; lower processing temperature, environmental compatibility, flux-less bonding, high density interconnection, simpler processing and low fabrication cost [1–4]. The I/O density required is extremely demanding–– down to a pad pitch of 70 lm with separation 10 lm. Conventional tin–lead solder joining is not capable of handling such fine geometry. Flip chip interconnection

*

Corresponding author. Tel.: +852-2788-7130; fax: +8522788-7579. E-mail address: [email protected] (Y.C. Chan).

using anisotropic conductive adhesive films (ACFs) has been introduced as a promising procedure in flip chip technology. ACF is mainly consisting of metal or metal-coated fillers in a polymer matrix. Fine pitch requirement, in order to meet future chip size and I/O density, has ruled out most of the materials. Materials selection of ACFs focuses on materials physical properties, application pitch limitations and processing windows [5]. Epoxy based systems offer good post-assembly solder re-flow resistance, controlled cross-linked density, and are normally solvent free. Because of the anisotropic property, ACFs may be deposited over the entire contact region, thus greatly facilitating material application [1]. ACF has a low percentage of metal fillers in volume; the electric conduction through a number of conductive pads exhibits weak dependence on the external bonding force [1]. The concentration of particles is controlled in such a way that just enough particles are present to assure reliable electrical conductivity in the z direction while concentration is far below a critical value to achieve percolation conduction in the x–y plane [2]. It was indicated that the

0026-2714/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0026-2714(02)00318-9

482

C.W. Tan et al. / Microelectronics Reliability 43 (2003) 481–486

connection resistance rapidly decreased and then saturated following the increase of the conductive particle content [3]. Applications on flexible substrate such as smart cards, disk drives and driver chips for LCDs have attracted much interests and widespread uses [6]. Many flat panel displays are using aluminium metallurgy on the bonding pads. Therefore, chips with 1 lm aluminium metallization (bumpless) have been used in this study. This construction is also used to eliminate the influence of bump height and bump pitch when external stress is applied. In addition, bumpless chip costs less when comparing with conventional Au bump or Au/Ni bump. The objective of this paper is to add to the fundamental understanding of degradation in ACF joints, and to identify the dominant failure mechanisms for different service environment regimes, including tensile and cyclic mechanical loading under elevated temperature and humidity. Gupta et al. [7] reported that metal surfaces were relatively low on surface energy due to organic contamination and absorbed moisture, therefore, resulted in weak adhesion. The scope of this study includes an in-depth analysis of failure mechanisms and failure modes for several of parameters. Failure surfaces are examined for identification of degradation mechanisms in the ACF joints interlayer. Measurements and observations are related to damage processes; failure modes and the results are assessed with respect to the relevance of existing theories and failure criteria.

2. Experimental methods The flex substrates that are used in this study are of 50 lm thickness and the height of gold/electroless nickel coated copper (Au/Ni/Cu) electrodes is about 14 lm. ACF that has the capability for the fine pitch on flex is used in this study. The thickness of ACFs is about 35 lm. Conductive particles with diameter of 3.5 lm (Resin þ Ni/Au plating þ Insulating coating) are distributed in the adhesive matrix, with concentration of 3.5 million/ mm3 . The test chips have a size of 11  3 mm2 , a total of 368 aluminium metallizations. Each bump has a height of 1 lm and a bump area of 70  50 lm2 . The pre-bonding process is carried out using the Karl Suss FCM manual flip chip bonder at temperature of 90 °C and pressure of 0.3 MPa. Then, followed by final bonding by using the Toray FC 2000 semi automatic flip chip bonder at 200 °C and 160 N for 10 s. The die shear test was carried out by using INSTRON Mini 44 Tester with a cross-head speed of 5 mm/min on ten samples. The shear blade was placed approximately 3 m from the surface of flexible substrate. At various displacements, the contact resistance of the samples was measured by using four-point probe method. The strain was found by dividing the observed displacement prior to fracture over the joint width in the direction of shear force. Samples were also subjected to the autoclave test that was designed according to JEDEC Standard no. 22 Method A102-B, with test con-

Fig. 1. Experimental set-up and specimen geometry for fatigue testing in shear.

C.W. Tan et al. / Microelectronics Reliability 43 (2003) 481–486

ditions set at 121 °C, 100% relative humidity and 1 atm. The test readout points were selected as 0, 48, 96, 192, 288, 336 and 384 h. At these points, the shear strength of samples was determined by using shear test; the microstructural analysis of the fracture surfaces were obtained by using Philips XL 40 FEG scanning electron microscope. Ten samples were selected for each readout point. The shear cyclic fatigue testing was performed using an INSTRON MINI-44 universal tensile tester with the fixture shown in Fig. 1. The displacement controlled shear fatigue tests were performed with peak displacement of 0.5 mm, at a speed of 5 mm/min and cycling frequency of 0.42 Hz. The shear peak loading change with time was recorded at each shear cycle. The cycle after which the peak loading drops to 50% of that of the initial cycle is defined as the fatigue life. This test was performed at room temperature and about 60% relative humidity. At various numbers of cycles, the contact resistance of the samples was measured by using fourpoint probe method.

3. Results and discussion 3.1. Shear test Die shear tests were used to obtain the range of specimen durability. Fig. 2 shows shear force and contact resistance curves as a function of displacement for ACF joints. As expected from elastic displacement, the force increased linearly with displacement. In this elastic region, i.e. in between a displacement of 0.00–1.25 mm, the contact resistance was approximately constant but did typically show a small increase. ACF joints show an insignificant viscoelastic deformation after a displacement of 1.25 mm. The maximum shear force prior to

Fig. 2. Load–displacement curve of ACF joint and its average contact resistance at various displacements.

483

Fig. 3. Stress–strain curves of ACF joints at various autoclave test time, i.e., 0, 48, 96, 192, 288, 384 h.

fracture is about 465.0 N at a displacement of 1.725 mm. In this moderately plastic region, the contact resistance exhibits a shoot up prior to separation of ACF joints at a displacement of 1.75 mm. For most of the time, epoxies display the same behaviour as a brittle metal or ceramic in tensile test. Fig. 3 shows stress–strain curves of ACF joints that have undergone autoclave test for various test times. The maximum shear stress prior to fracture was obtained in samples prior to autoclave test at approximately 14.35 MN m2 . Maximum shear stress for ACF joints reduced with autoclave test time. The viscoelastic deformation region also became insignificant for samples that have undergone autoclave test where its breaking points were in the elastic region. Stress–strain curve of ACF does not demonstrate a perfect linear correlation in the elastic region because it is a visco-elastic material. Therefore, the initial part of Fig. 3 (shear strain < 0:01), which demonstrates a linear stress–strain curve, was chosen and enlarged, as shown in Fig. 4. Shear moduli were reduced with autoclave test time, from 48.53 MN m2 before test to 20.13 MN m2 after 384 h of autoclave test. While shear modulus decreases, plasticization due to absorption of moisture will result in decreases of its ductility at elevated temperature. Liu [8] has reported that oxidation growth on metal surface causes increases in electrical resistance but does not cause catastrophic failure. Crack formation must also be involved as a failure mechanism in order to explain the resistance change during the humidity testing (85 °C/85%RH). Plasticization due to absorption of moisture causes expansion of adhesive joint and the crack normally starts at the interface between the adhesive and the adherent [8]. Fracture surfaces of ACF joints underwent various autoclave test times have been exhibited in Fig. 5. For sample prior to autoclave test, the scanning electron

484

C.W. Tan et al. / Microelectronics Reliability 43 (2003) 481–486

Fig. 4. Shear moduli for ACF joints obtained from experimental results at different autoclave test time.

fractograph showing parabolic-shaped dimples characteristic of moderately ductile fracture resulting from shear loading, as shown in Fig. 5(a). Fig. 5(a)–(f) show that ACF joints lost its ductility and yield a relatively flat fracture surface as the test time increases. Brittle fracture takes place without any appreciable deformation, and by rapid crack propagation. The direction of crack motion is approximately perpendicular to the direction of the applied stress. In Fig. 5(e) and (f), the fracture surfaces of ACF joints that have experienced autoclave test for 336 and 384 h respectively, underwent spalling in which the surfaces were smoother and less plastic deformation was observed. For fracture to occur without significant plastic deformation, adjoining atoms must be completely separated. The force or stress, which is required, is a function of the sublimation energy of material [9]. Two note-

Fig. 5. Fracture surface of ACF joint underwent autoclave test for (a) 0, (b) 48, (c) 96, (d) 192, (e) 336 and (f) 384 h after failing the shear test.

C.W. Tan et al. / Microelectronics Reliability 43 (2003) 481–486

worthy characteristics of these brittle fractures are the following: (1) involves minimal energy absorption and plastic deformation, and (2) the fracture stress is highly variable. 3.2. Fatigue test Fatigue refers to fracture arising from cyclic stresses that alternate between tension and compression. The number of stress cycles prior to fracture is a function of the applied stress [10]. The samples were clamped in the test fixture exercising care not to damage the Cu traces on flex. There was typically a small measured force after the clamping process. The force was typically negative producing a compressive stress in the Cu traces, and may have resulted from the elongation of the copper traces when they were clamped. Since the force resulting from clamping was small when compared with the cyclic force, it was neglected and the initial clamping position was defined to be the position of zero displacement [11]. Since the maximum load for fatigue test should not exceed the fracture limit of ACF joints, cyclic displacement was set at 0.5 mm at which the joints are still in elastic region. Fig. 6 shows the fatigue data for ACF joints. The figure plots the values of force in shear at the maximum displacement and average contact resistance of ACF joints. Only minimum force is shown because the maximum and minimum force is only different in direction but has same magnitude. At the beginning of test, the force in shear shows a drastic degradation from about 189.1 to 151.7 N. Another degradation is seen prior to an endurance limit at about 143.5 N after 1000 cycles. This unfavourable observation could be due to the property of ACF joints that they are prone to brittle and

Fig. 6. The relationship of cyclic shear force and average contact resistance at various fatigue cycles in shear.

485

have experienced plastic deformation before endurance limit. The endurance ratio can be determined by dividing the endurance limit over tensile strength. In this case, the endurance ratio calculated by using experimental data is only about 32.0%, which means the stress below endurance limit has about 32.0% possibility that fatigue failure will never occur. Many materials have an endurance limit below which it is not subjected to fatigue failure. Lower-stress fatigue failure commonly starts as surface cracks at 10–20% of the fatigue life. Therefore, micro-crack in ACF joint is suspected that has led to further degradation of force in shear after endurance limit. The cracks progress slowly from the initiation point to the final with rapid failure. Prior to crack initiation, extrusions and intrusions arises from irreversible slip during successive stress reversals. Strain hardening after each slip movement prevents back-slip on the same plane. A decrease in usable strength under cyclic loading is directly attribute to the fact that the material is not a homogeneous and isotropic solid. Fatigue cracks can be initiated under relatively low stresses at surface imperfections. In normal service, the stress concentrations are initially not severe enough to produce catastrophic failure, however, they are sufficient to cause slow propagation of a crack into material. Eventually the crack may become sufficiently deep so that the stress concentrations exceed the fracture strength and catastrophic failure occurs. As the mechanical force in shear degraded radically at the initial of the cyclic loading, the contact resistance of ACF joints have been improved from about 110.5 to 91.6 mX. This observation is similar to Constable et al. [11] report. There are a few possible mechanisms can result in this observation. Firstly, before the cyclic loading has degraded the rigid ACF joints, it further compresses the joints and gives larger contact area of particles sandwiched between the two metallizations. Secondly, slipped ACF joints by cyclic loading induced frictions that might have scrubbed away the aluminium oxide layer at the surface of aluminium metallization. At normal environment, aluminium will form an oxide film,  for most of the cases. This approximately 25–40 kA oxide layer has higher resistance than that of bare aluminium. No doubt scrub-off of this oxide layer can improve the contact resistance of ACF joints. Thirdly, particles used in this ACF consist of a thin insulation layer at the outer surface. This layer should be eliminated during the thermo-compression bonding process. If there is any left behind, sliding in surfaces of ACF joints might help in eliminating this insulation layer due to frictions. However, after about 1000 cycles, an increasing trend can be seen in Fig. 6. It is believed that micro-cracks induced in ACF joints that has been discussed earlier,

486

C.W. Tan et al. / Microelectronics Reliability 43 (2003) 481–486

have reduced the contact area of particles and metallizations. This mechanism started to dominate the failure mechanism of ACF joints since then. Until the contact resistance reached about 357.6 mX after 9000 cycles, fatigue failures were observed at the interface between adhesive and Cu traces.

Packaging (Project no. 8720003) and grant from Innovation Technology Commission, The Government of the Hong Kong Special Administrative Region (ITS/182/ 00)––Conductive Adhesive Technology Programme for Fine Pitch Electronic Interconnect (Project no. 9440006).

4. Conclusions

References

ACF joints achieved maximum shear adhesion force of 465.0 N prior to breaking. In between a displacement of 0.00 to 1.25 mm, the contact resistance of ACF joints shows a small increase. Fracture of these joints led to open circuits. The mechanical properties of ACF joints are dependent on the working environment. Its shear moduli reduce with autoclave test time, from 48.53 MN m2 before test to 20.13 MN m2 after 384 h of autoclave test time. Samples, that have undergone autoclave test display brittle behaviour with less plastic deformation, were observed at fracture surface and stress–strain curves. These fracture surfaces are smoother when comparing with that of the samples before test. Spalling was observed in samples that have undergone autoclave test for 336 and 384 h, which indicates a brittle property of these joints. At the beginning of fatigue test, the force shows a drastic degradation from about 189.1 to 151.7 N. As the mechanical force in shear degraded radically, the contact resistance of ACF joints have been reduced from about 110.5 to 91.6 mX. This improvement could be due to either scrub away of aluminium oxide film at metallization surface and insulation layer on particles or increase of contact area due to excess compression force. Adhesion is also an important factor for having stable and low contact resistance values [12]. Another degradation in force is seen prior to an endurance limit at about 143.5 N after 1000 cycles. Since then contact resistance has been dominant by microcracks at adhesion interface, which exhibited an increasing trend. The endurance ratio is approximately 32.0%, which means the stress below endurance limit has about 32.0% possibility that no fatigue failure will occur.

[1] Fu Y, Willander M, Liu J. Statistic of electric conductance through anisotropically conductive adhesive. IEEE Trans Component Packaging Technol 2000;24(2):250–5. [2] Chang DD, Fulton JA, Lyons AM, Nis JR. Design considerations for the implementation of anisotropically conductive adhesive interconnections. In: Proceedings of Nepcon W., 1992. p. 1381. [3] Yim M, Paik K. Design and understanding of anisotropic conductive films (ACFs) for LCD packaging. IEEE Trans Component Packaging Technol––Part A 1998;21:226–34. [4] Nagai A, Takemure K, Isaka K, Watanabe O, Kojima K, Matsuda K, Watanabe I. Anisotropic conductive adhesive films for flip chip interconnection onto organic substrates. In: Proceedings of IEMT/IMC, 1998. p. 353–7. [5] Fan SH, Chan YC, Tang CW, Lai JKL. Effect of cooling rate on the isothermal fatigue behavior of CBGA solder joints in shear. IEEE Trans Adv Packaging 2001;24(1):10–6. [6] Chan YC, Hung KC, Tang CW, Wu CML. Degradation mechanism of anisotropic conductive adhesive joints for flip chip on flex applications. In: Proceedings of IEEE 4th International Conference on Adhesive Joining and Coating Technology in Electronics Manufacturing, 2000. p. 141–6. [7] Gupta S, Hydro RM, Pearson RA. Fracture behaviour of isotropically conductive adhesives. IEEE Trans Components Packaging Technol 1999;22(2):209–14. [8] Liu J. On the failure mechanism of anisotropically conductive adhesive joints on copper metallisation. Int J Adhes Adhes 1996;16(4):285–7. [9] Grellmann W, Seidler S. Deformation and fracture behavior of polymers. Berlin: Springer; 2001. [10] Gomatam R, Sancaktar E, Boismier D, Schue D, Malik I. Behavior of electronically conductive filled adhesive joints under cyclic loading part 1: Experimental approach. In: Proceedings of International Symposium on Advanced Packaging Materials, 2001. p. 6–12. [11] Constable JH, Kache T, Teichmann H, Muhle S, Gaynes MA. Continuous electrical resistance monitoring, pull strength, and fatigue life of isotropically conductive adhesive joints. IEEE Trans Compon Packag Technol 1999;22(2):191–9. [12] Liong S, Wong CP. An alternative to epoxy resin for application in isotropically conductive adhesive. In: Proceedings of International Symposium on Advanced Packaging Materials, 2001. p. 13–8.

Acknowledgements The authors would like to acknowledge the financial supports from Research Grants Council of Hong Kong––Cooperative Research Centre on Conductive Adhesive Technology for High Density Electronic