Y.C. Chan *, D.Y. Luk

Transcription

1 Microelectronics Reliability 42 (2002) Effects of bonding parameters on the reliability performance of anisotropic conductive adhesive interconnects for flip-chip-on-flex packages assembly I. Different bonding temperature Y.C. Chan *, D.Y. Luk Department of Electronic Engineering, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon Tong, Hong Kong Received 11 February 2002; received in revised form 27 March 2002 Abstract The effects of different bonding temperatures during flip-chip-on-flex (FCOF) assembly in relation to the performance of anisotropic conductive adhesive (ACF) interconnect were investigated. Two types of flip chips were used in this study. It was found that Ni bumps formed better interconnections than bumpless FCOF packages. Aluminium oxide was observed and was thought to be the main cause of the increased in contact resistance after the moisture-soak tests. The conductive particles were not fully compressed by the bumps and pads and gaps were observed between the conductive particles and Cu pads in bumpless packages. Conductive particles in the Ni bump FCOF packages were tightly trapped between the bumps and pads and hence gave better connections. The performance of the ACF interconnects were affected by the degree of curing of the ACF, which was determined by the bonding temperature. Ó 2002 Published by Elsevier Science Ltd. 1. Introduction Before the invention of anisotropic conductive adhesives (ACFs), solder alloys were used as interconnection materials in flip chip packages. These packages were bulky, hard to work with and the lengthy assembly processes were complicated and required high temperatures. In addition, conventional lead tin soldering used in flip chip interconnections is incompatible with extremely fine pitch interconnection and is undesirable due to the toxic effects of lead. ACFs possess many distinct advantages that solder alloys can not offer, namely being flexible, capable of fine pitch interconnections, environmental friendly, and cheaper to manufacture as the assembly processes are simpler, shorter and with lower temperatures. Hence, many chip on flex electronics * Corresponding author. Tel.: ; fax: address: eeycchan@cityu.edu.hk (Y.C. Chan). packages involve interconnect applications using ACFs, for example, mobile phones, personal digital assistants [1] and smart cards [2]. Despite the advantages mentioned above, there are two major drawbacks. The contact resistance of ACF joints becomes increasingly unstable through time, particularly under high temperature (85 C) and high humidity (85% RH) conditions (so called 85/85 conditions) [1]. This is such an important issue because these conditions are well known as the qualification standards throughout the electronic industry. Mechanisms that thought to affect the stability of contact resistance include water absorption, electrochemical corrosion and metal oxidation [3]. These degradation mechanisms interfere with the contact resistance of ACF joints and hence limiting the performance of electronic packages. In order to solve this problem, one must have a clear understanding of how exactly does the failure mechanisms occur. In addition, the ACF is unable to self-align [4] and hence high precision bonding processes are required /02/$ - see front matter Ó 2002 Published by Elsevier Science Ltd. PII: S (02)

2 1186 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) ACFs consist of mixtures of conducting fillers in an insulating matrix. This arrangement allows the material to conduct in the z-direction while remaining insulators in the x y plane [3]. The aim of these adhesives is to trap at least one conductive particle between the conductive bumps on the flip chip and the corresponding pads on the substrate. This has to be achieved without the occurrence of bridging between the pads. The particles are randomly distributed in the matrix in most anisotropic materials, which can cause problems especially in ultrafine pitch applications. This is because the concentration of particles within the material varies at different locations, and hence may result in open or short circuit. Epoxy resin based ACFs are thermosetting [2]. They are temperature sensitive therefore their structure is highly dependent upon the bonding temperature chosen. The mobility of the conductive particles is different at different stages during ACF curing. During the bonding process of flip-chip-on-flex (FCOF) assembly, the ACF is being cured and becomes soft and rubbery. This transformation allows the ACF to flow, which in turn allows the conductive particles within to move and distribute themselves evenly throughout the ACF joints. When the curing process is completed, the ACF becomes hardened and the mobility of the conductive particles is lost. A reliable interconnect should have sufficient amount of conductive particles between the bump and pad in close contact and that they do not flow away during bonding [5]. The root cause of the instability of contact resistance maybe due to the incorrect selection of bonding temperature during the assembly of FCOF packages. This series of studies concentrate on the effects of different bonding parameters during the assembly of FCOF packages in relation to the reliability of the ACF joints. The aim of this study was to investigate the effects of different bonding temperatures on the contact resistance of ACF joints, with special focus on the chip/ conductive particle metallization interface. The results of this study would allow development of ACF joints using fine pitch flip chips on flexible substrates with better reliability and longer fatigue life. 2. Experimental procedure The FCOF packages are made up of three different materials, namely silicon (Si) chip, ACF and flexible substrate Silicon chips The dimensions of the silicon (Si) chips are 10:87 mm 3:14 mm, with rectangular bumps (70 lm 50 lm). The bumps are arranged in sets of five as a group; with two adjacent bumps for measuring insula- Fig. 1. Schematic diagram showing a corner of a Si chip with daisy-chained bumps. Fig. 2. Schematic diagram showing the structure of the ACF and their conductive particles.

3 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) Table 1 Specifications of the ACF Description Specification Film thickness (lm) 30 Conductive particle Au/Ni coated polymer Insulation coated No Particle size (lm) 3 Pre-bonding temperature 110 ( C) Pre-bonding time (s) 5 Pre-bonding force (MPa) 10 per unit area of bump Bonding temperature ( C) 180 Bonding time (s) 10 Bonding force (MPa) per 100 unit area of bump T g ( C) 145 Table 2 Bonding temperatures used Bonding temperature ( C) Standard 180 Tests 160, 200, 220, 240 tion resistance and three for contact resistance. There are a total of 12 sets of these daisy-chained bumps within the chip. The layout of the chip is shown in Fig. 1. Electroless nickel (Ni) bumping process involves aluminium (Al) cleaning, Al activation, electroless Ni deposition and immersion gold (Au) coating [3]. The bump height of the Ni bump and Al pad are 4 and 1 lm, respectively. The last two steps in the bumping process of bumpless chips were omitted Anisotropic conductive adhesives The type of ACF used in this study was a double layer ACF that consists of an epoxy layer and another one filled with conductive and insulation particles. The conductive particles are made up of polymers plated with a thin layer of nickel followed by a thin layer of gold. Fig. 2 shows the structure of the ACF and its specifications are summarized in Table Flexible substrate The flex substrates used in this study were about 40 lm thick and the electrode is gold/electroless nickel coated copper (Au/Ni/Cu). Twelve micrometer thick of copper (Cu) traces was electrodeposited onto a 25 lm thick polyimide (PI), followed by 4 5 lm thick of electroless nickel (Ni) and finally sputtered with a 0.4 lm thick gold (Au) layer. Since the flex substrate is of ultrafine pitch (the smallest gap between the traces was 10 lm), Ni was plated onto the Cu traces to prevent Cu migration. Au sputtering was necessary to prevent the Ni layer from oxidation. During the pre-bonding process, the ACF was laminated onto the flexible substrates, by using the Karl Suss manual flip chip bonder. The final bonding of flip chip onto the ACF/flex was carried out using the Toray semiautomatic flip chip bonder. The alignment accuracy is 2 lm. Different bonding temperatures were used in this study, as shown in Table 2 and the schematics of the bonding process is shown in Fig. 3. Fig. 3. Schematic diagram showing the formation of flip chip interconnections with (a) bumped chip and (b) bumpless chip using ACFs.

4 1188 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) microstructure and microjoints of the FCOF packages, especially the chip/conductive particle metallization interface. To simulate the curing reaction of ACF during FCOF assembly, the ACF was laminated onto the flex substrate followed by curing for 10 s at the temperatures selected as shown in Table 2. The degree of curing of ACF was measured by using the Perkin Elmer spectrum one Fourier transform infra-red (FT-IR) Spectrometer as demonstrated by Chiu et al. [6]. 3. Results and discussion Fig. 4. Contact resistance measurement of ACF joints using the four-point probe method (I ¼ 1 ma). The contact resistance of the ACF joints of the FCOF packages was measured by using the four-point probe method as shown in Fig. 4. In the four-point probe test, 1 ma was applied to the circuit constantly and the voltage was measured for each set of bumps using the Hewlett Packard 3478 A Multimeter. The contact resistance was calculated by using R ¼ V =I. One set of samples was stored under dry conditions (20 C/30% RH) and another set of samples was moisture-soaked under 60 C/95% RH conditions for 336 h. The samples were then mounted in epoxy resin and cross-sectioned. The Philips XL40 FEG scanning electron microscope (SEM) equipped with energy dispersive X-ray (EDX) was used to inspect and analyse the From Fig. 5, one can see that the contact resistance for both Ni bump and bumpless (flip chips with Al pads rather than Ni bumps) FCOF packages show the same trend. The contact resistance of the packages bonded at C decreased steadily but then increased when the bonding temperature used was above 200 C. FCOF packages assembled at 200 C gave the best contact when compared to those assembled at a different temperature. For the Ni bump and bumpless FCOF packages assembled at 200 C, the initial contact resistance was 2.77 and 4.18 mx, respectively. After 336 h of dry storage, the contact resistance dropped only slightly to 2.56 and 4.06 mx, respectively. After 336 h of moisture absorption, the contact resistance increased slightly to 2.88 and 4.93 mx, respectively. Fig. 5 shows that bumpless FCOF packages gave higher contact resistance values than Ni bump packages, especially with bonding temperatures above 200 C, the contact resistance doubled. This may be due to the heat being absorbed by the FCOF packages and acting as a catalyst for oxidation of the Al pads. Humid environments favour oxidation and the increase in contact Fig. 5. Average contact resistance of Ni bump and bumpless (b/less) FCOF packages assembled at various temperatures with different storage conditions.

5 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) Fig. 6. Schematics showing the path of an electron flowing through the (a) non-oxidised and (b) oxidised Al pad. Fig. 7. SEM micrograph showing the ACF interconnect chose for EDX analysis at the (a) middle of the Ni bump and (b) Ni bump/ ACF interface. Fig. 8. SEM micrograph showing the ACF interconnect chose for EDX analysis at the (c) middle of the Al pad and (d) Al pad/acf interface.

6 1190 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) Fig. 9. Curing percentage of ACF at various temperatures (curing time ¼ 10 s). resistance with FCOF packages after 60/95 treatment was evident, as shown in Fig. 5. Materials like aluminium oxidises very easily and the aluminium oxide layer formed may act as a barrier between the Al pad-conductive particle and conductive particle-cu pad, making it difficult for the electrons to flow through this Al padconductive particle-cu pad path. This phenomenon is shown in Fig. 6. The SEM-EDX results showed that the oxygen content in bumpless packages was much higher than those with Ni bumps. Referring to Fig. 7, there was no oxygen detectable at point (a), the middle of the Ni bump, but Fig. 10. Optical micrographs showing the morphology of ACF (a) before curing and after curing for 10 s at (b) 160 C (c) 180 C (d) 200 C (e) 220 C and (f) 240 C (magnification ¼200).

7 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) Fig. 11. SEM micrographs showing the morphology of ACF (a) before curing and after curing for 10 s at (b) 160 C (c) 180 C (d) 200 C (e) 220 C and (f) 240 C wt.% oxygen was present at point (b), the Ni bump/ ACF interface. The oxygen is suspected to come from the ACF rather than the thin layer of Au on the Ni bump being oxidised, as gold is inert to oxidation. Comparing these values to the results obtained from the locations shown in Fig. 8, point (c), the middle of the Al pad, contained 22.4 wt.% of oxygen and 49.3 wt.% at point (d), the Al pad/acf interface. In addition, from the SEM micrograph shown in Fig. 8, it is clear that the Al pad consists of two layers; one being the Al pad itself and the other is suspected to be the aluminium oxide layer. The degree of curing of ACF plays an important role in determining the reliability of the ACF joints. ACFs are thermosetting polymers, which deform and degrade easily at high temperatures. The degree of curing of ACF is very much dependent upon the bonding temperature. Previous study shows that 80% curing of ACF would be achieved for ACF prepared at 180 C for 10 s [1,6]. From the curing percentage of ACF results shown in Fig. 9, we found that the ACF was about 74% and 83% cured at 180 and 200 C for 10 s, respectively. When the bonding temperature was below 160 C, the ACF was only 26% cured. The cross-linkage within the

8 1192 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) Fig. 12. SEM micrograph showing the conductive particles trapped within the ACF joint of Ni bump FCOF assembled at 200 C. Fig. 13. SEM micrograph showing the conductive particles unable to distribute within the ACF joint of Ni bump FCOF assembled at 240 C. polymer may be incomplete. In contrast, when the bonding temperature was above 240 C, the ACF was about 95% cured, which would not be desired to use as a bonding temperature. This is because the epoxy may set too quickly without sufficient flowing at this temperature and hence the conductive particles would not have enough time to distribute themselves in between the bumps and pads in close contact. The glass transition temperature, T g, of the ACF used is 145 C. When this temperature is reached, the ACF softens and begins to flow. It becomes hardened when the curing reaction is completed [5]. This is because as the curing of ACF proceeds, the linear polymer chain in the epoxy resin grows and branches to form cross-links [5]. The polymer chain is no stronger than its weakest link, and the temperature of initial degradation is usually the temperature at which the least thermally stable bonds fail. The bulk of the polymer may be stable, but the failure of the weakest bonds often produces results such as discolouration [7]. When the polymer has reached its failure point, it will decompose and its physical integrity would be lost. These effects were observed in the ACF being cured at different temperatures for 10 s, as shown in Fig. 10 as the temperature in-

9 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) Fig. 14. SEM micrograph showing the conductive particles being compressed by the Ni bump and Cu pad leaving no gaps. Fig. 15. SEM micrograph showing the conductive particles not being fully compressed by the Al pad and Cu pad leaving small gaps. creased, the epoxy started to break and eventually the epoxy layer was degraded into lumps. The appearance of the ACF being cured at 160 and 180 C, Fig. 10(b) and (c) respectively, was similar to the uncured ACF, Fig. 10(a) no significant break down of the ACF was observed. When the curing temperature of ACF was at 200 C, the linear polymer chains within started to grow and its physical appearance began to change as shown in Fig. 10(d). At 220 C, the growing polymer chain branched out to form cross-links, Fig. 10(e), until the chemical bonds within the cross-links extended to their maximum and broke and lumps were observed as shown in Fig. 10(f). This change in morphology was examined by using the SEM and the micrographs are shown in Fig. 11. During the bonding process of FCOF assembly, the ACF is being cured and becomes soft and rubbery. This transformation allows the ACF to flow, which in turn allows the conductive particles within to move and distribute themselves evenly throughout the ACF joints. When the curing process is completed, the ACF becomes hardened and the mobility of the conductive particles is lost. When the curing temperature was at 160 C, the conductive particles carried within the epoxy layer were able to move but only in a slow rate. At a higher temperature, 180 C, the particles gained more energy and

10 1194 Y.C. Chan, D.Y. Luk / Microelectronics Reliability 42 (2002) hence became more mobile and were able to move slightly faster. At 200 C, the ACF softened and was flowing at the correct rate. At this temperature, the conductive particles within the epoxy layer were able to distribute themselves throughout the ACF joints, and hence creating the best contact between the chip bumps and substrate Cu pads, as shown in Fig. 12. A reliable interconnect should have enough conductive particles within the ACF joint and that they do not flow away during bonding [5]. However, when the bonding temperature was above 200 C, the ACF was cured and set too fast preventing the conductive particles from spreading evenly between the bumps and pads, therefore not creating good interconnections as shown in Fig. 13. When compared to the Ni bump FCOF packages, the bumpless ones gave a higher contact resistance owing to the way the conductive particles within are trapped. As shown in Fig. 14, the conductive particles are being compressed slightly and trapped within the bumps and pads. There was hardly any space in between the conductive particles and bump or pad. This combination gave a better contact and hence easier for the electrons to flow through. However, in the bumpless FCOF packages, the conductive particles are not being fully compressed by the bumps and pads, and hence leaving a gap in between, as shown in Fig. 15. This combination may impede the flow of electrons through the interconnects therefore bumpless FCOF gave a higher contact resistance. 4. Conclusions Most materials increase in resistance with temperature, since the higher the temperature the more vigorously the atoms vibrate, so the more they hinder the passage of drifting electrons. It was found that there was only a gradual increase in contact resistance after the FCOF packages were being moisture soaked. Since the order of magnitude in contact resistance did not change, one would consider the packages were reliable. In this study, the optimum temperature for bonding FCOF with ACF was concluded to be at 200 C. The bonding temperature determines the degree of curing of the ACF. At temperatures below 200 C, the degree of curing was <80% and the flow of ACF provided the conductive particles sufficient mobility to distribute themselves evenly between the bumps and pads. In contrast, if the bonding temperature was >200 C, 95% curing of the ACF is achieved and the ACF would set too quickly before the conductive particles have a chance to locate themselves throughout the interconnects. Hence, the ACF joints of the FCOF packages assembled at 200 C performed better with lower contact resistance values when compared to those assembled at different temperatures, especially at temperatures above 200 C. It is therefore concluded that the performance of the ACF interconnects is greatly influenced by the bonding temperature during the assembly of FCOF packages. In addition, it was found that the conductive particles were trapped tightly between the Ni bumps and Cu pads in the Ni bump FCOF packages. However, the conductive particles within the bumpless FCOF packages were not fully compressed between the Al and Cu pads and hence leaving small gaps. This finding was thought to be another factor that caused the ACF interconnects in bumpless FCOF packages to be less effective and may be influenced by the bonding pressure. Therefore, to fully understand the performance and reliability of the ACF interconnects, other bonding parameters should also be considered. Acknowledgement The authors would like to acknowledge the Strategic Research Grants (project no ) of the City University of Hong Kong. References [1] Chan YC, Hung KC, Tang CW, Wu CML. Degradation mechanisms of anisotropic conductive adhesive joints for flip chip on flex applications. In: Proceedings of the 4th International Conference on Adhesive Joining and Coating Technology in Electronics Manufacturing, p [2] Aintila A, Sarkka J, Kivilahti JK. Development of highdensity interconnection techniques for contactless smart cards. In: Proceedings of the 4th International Conference on Adhesive Joining and Coating Technology in Electronics Manufacturing, p [3] Aschenbrenner R, Ostmann A, Motulla G, Zakel E, Reichl H. Flip chip attachment using anisotropic conductive adhesives and electroless nickel bumps. IEEE Trans CPMT Part C 1997;20(2):96. [4] Fan SH, Chan YC. Effect of misalignment on the electrical characteristics of anisotropic conductive adhesive joints. Microelectron Reliab 2002;42(7): [5] Connell G, Zenner RLD, Gerber JA. Conductive adhesive flip-chip bonding for bumped and unbumped die. In: Proceedings of the 47th Electronic Components and Technology Conference, p [6] Chiu YW, Chan YC, Lui SM. Electric field effects on shortcircuiting between adjacent joints in fine pitch anisotropically conductive adhesive interconnects. In: 52nd Electronic Components and Technology Conference. San Diego, California, USA, May [7] Buch X, Shanahan MER. Thermal and thermo-oxidative ageing of an epoxy adhesive. Polym Degrad Stab 2000; 68(3):