Assembly of planar array components using anisotropic conducting adhesives: a benchmark study. Part I - experiment

Transcription

1 Loughborough University Institutional Repository Assembly of planar array components using anisotropic conducting adhesives: a benchmark study. Part I - experiment This item was submitted to Loughborough University's Institutional Repository by the/an author. Citation: OGUNJIMI, A. O....et al, Assembly of planar array components using anisotropic conducting adhesives: a benchmark study. Part I - experiment. IEEE transactions on components, packaging and manufacturing technology - Part C, 19 (4), pp Additional Information: This is a journal article. It was published in the journal IEEE transactions on components, packaging and manufacturing technology - Part C [ c IEEE], and is available from: Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE. Metadata Record: Publisher: c Institute of Electrical and Electronics Engineers (IEEE) Please cite the published version.

2 This item was submitted to Loughborough s Institutional Repository ( by the author and is made available under the following Creative Commons Licence conditions. For the full text of this licence, please go to:

3 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART C, VOL. 19, NO. 4, OCTOBER Assembly of Planar Array Components Using Anisotropic Conducting Adhesives A Benchmark Study: Part I Experiment Adebayo Oluyinka Ogunjimi, Samjid H. Mannan, David C. Whalley, and David J. Williams Abstract This paper presents new results from an experimental and theoretical program to evaluate relevant process parameters in the assembly of a 500 m pitch area array component using anisotropic conductive adhesive (ACA) materials. This experimental configuration has features of micro ball grid array (BGA), chip scale packaging (CSP), and also flip-chip and conventional ball grid array (BGA) package structures. A range of materials combinations have been evaluated, including (random filled) adhesive materials based on both thermoplastic and thermosetting resin systems, combined with both organic and thick-film on ceramic substrate materials. The ACA s used have all been applied as films, and hence are also known as anisotropic conducting films (ACF). The test assemblies have been constructed using a specially developed instrumented assembly system which allows the measurement of the process temperatures and pressures and the consequent bondline thickness reduction and conductivity development. The effects of the process parameters on the resulting properties, particularly conductivity and yield, are reported. A complementary paper [1] indicates the results of computational fluid dynamics (CFD) models of the early stages of the assembly process which allow the extrapolation of the present results to finer pitch geometries. Index Terms Anisotropic conducting adhesive, flip-chip, micro-electronics assembly. I. INTRODUCTION THE APPLICATION of anisotropic conductive adhesives (ACA s) to electronic interconnection has a number of potential advantages [2], but the impact of the assembly process on the properties of these materials is not fully understood, despite an extensive amount of work done in the area [3] [7]. The last reference [7] in particular examines many different approaches to the subject. The work presented in this paper is part of an ongoing program focused on understanding the behavior of these materials during the assembly/manufacturing process and how this contributes to their resulting properties and long term Manuscript received May 1996; revised November This work was supported by the Engineering and Physical Sciences Research Council (EP- SRC) Contract GR/J68410 and a consortium of industrial partners which include Cooksons Technology Centre and their associated company Alpha Metals Advanced Products, Eltek Semiconductors Ltd., Lucas Electronics, Rolls-Royce Plc, Design 2 Distribution Ltd., IBM (U.K.) Ltd., and Mitel Telecom. This paper was presented at the 2nd International Conference Adhesives in Electronics, Stockholm, Sweden, June 3 5, The authors are with the Department of Manufacturing Engineering, Loughborough University, Leics, LE11 3TU UK. Publisher Item Identifier S (96)01529-X. performance. This present work addresses the choice of manufacturing process parameters, subsequent work will address the interrelationships between processing conditions and reliability. The key issues in the use of anisotropic conductive adhesives are the achievement of acceptable interconnection conductance with satisfactory yield and the retention of this conductivity during the product life. Most anisotropic adhesives achieve conduction through the formation of a pressure contact between the conductive filler particles and the substrate/component metalizations. The overall electrical performance of an anisotropic adhesive joint is therefore dependent not only on the resistivity of the joining materials but also on the final conductor particle distribution and the contact pressures locked in to the material by processing. The properties of these materials therefore evolve with the process and there is a clear need to understand the effects of the chosen processing parameters on this process and the resulting quality of the interconnections formed. II. DESIGN OF THE EXPERIMENTAL TRIALS The trials reported here were designed to provide a benchmark comparison of a range of existing commercially available materials combined with substrate geometries readily achievable with current technology, and to establish the relative sensitivity of these materials combinations to the processing conditions. A. Materials Selection and Test Vehicle Design The trials were designed around a 10 mm square device with 500 m pitch area array connections. The bump metallization was thick film printed AgPd. This experimental configuration is equivalent to that of style CSP, but is also broadly representative of BGA and flip-chip devices. The test vehicles were designed to enable a specially instrumented assembly system to monitor 16 joints (of the total of 400 connections) throughout the assembly process. Two different substrate technologies were employed AgPd thick film on alumina and conventional PCB using 1 oz Cu clad FR5 laminate (with a sub-micron coating of Au on top of the Cu). The decision to use FR5 laminate was prompted by the reportedly poor performance of the more generally used FR4 laminates in ACA bonded assemblies [8]. FR5 is much stiffer at the high process temperatures being used and has a more uniform distribution /96$ IEEE

4 258 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART C, VOL. 19, NO. 4, OCTOBER 1996 TABLE I ADHESIVE CHARACTERISTICS TABLE II INTERPRETATION OF LEVELS Fig. 1. Instrumented test rig setup. of resin and reinforcement. It was therefore expected to suffer less from the problems associated with FR4. Three different adhesive films supplied by two vendors were investigated. The adhesives were selected to reflect the state of the art in thermoset and thermoplastic based adhesives and to cover a range of conducting particle sizes and compositions. Table I summarizes the characteristics of the adhesives used in the trials. B. The Instrumented Test Rig The test assemblies were constructed using a specially developed instrumented assembly system which allows measurement of the process temperatures and pressures and the consequent bondline thickness reduction and conductivity development. This test rig is shown schematically in Fig. 1. The device was aligned, placed and tacked down on the substrate separately on a manual flip-chip bonder with split beam optics to obtain alignment, and then brought over to the instrumented test rig for application of pressure and curing. The planarity was checked using pressure sensitive paper (Fuji prescale film). C. Factor Selection and Experimental Design Since the level of interaction between the relevant processing parameters was not known full factorial experiments were designed. The key parameters thought to affect the quality of the interconnection were pressure, rate of application of pressure, temperature and substrate type which gave four factors which were each tested at two levels. Table II shows the factors and levels considered in the experimental design. This resulted in a total of 16 experiments which were repeated three times for each of the adhesives. The difference in material properties (especially thermal properties) of the test vehicle substrates had to be taken into consideration in designing and performing experiments. The parameters used in the experiment were designed around the recommended parameters as shown in Table I. The quantifiable dependent variables in the experiments were the joint contact resistances, of which 16 spatially distributed joints were monitored on each substrate, and the yield as defined in Section III. III. RESULTS There are three important parameters that have been derived from the raw experimental results: the average conductance of the monitored connections; the proportion of pads that actually conducted i.e. the yield (any contact resistances below 1 have been taken to be conducting in the analyzes presented); and the coefficient of variance of conductance (which is the ratio of the standard deviation to the mean of the conductances). The coefficient of variance gives an indication of the uniformity of the conductances achieved. Analysis of variance (ANOVA) was also conducted in order to understand the

5 OGUNJIMI et al.: ASSEMBLY OF PLANAR ARRAY COMPONENTS 259 TABLE III RESULTS (a) (b) (c) Fig. 2. (d) (e) (f) Effect of different factors on process yield. statistical level of significance of the effect of the different parameters (factors) and the first order interactions between the different process parameters (Table III). The results for each type of substrate were analyzed individually as the initial analysis showed this to be the most significant factor in all of the experiments and to interact strongly with the other factors. A. Process Yields Fig. 2(a) (f) shows the effect on the average (for three sets of trials) process yield for the different factors. Here the yield is defined as the percentage of good contacts. It may be clearly seen that the yield obtained for the six possible materials combinations have significant differences in their sensitivity to the process parameters. While these figures show the average over three different set of trials, they are interpreted in conjunction with the results of the more accurate estimate of the significance of the factors as investigated using ANOVA. 1) Ceramic Substrate: Temperature and rate of application of pressure are shown to have equal effect on the process yield for adhesive A1 [Fig. 2(a)], the thermoplastic adhesive with a polymer cored conducting particle. Pressure has a smaller effect on yield. Fig. 2(a) also shows that high temperature

6 260 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART C, VOL. 19, NO. 4, OCTOBER 1996 Fig. 3. Device process yield. (a) (b) (c) and rate of application of pressure combined with low pressure are the best combination of parameters for high yield. However, ANOVA reveals that for this material, temperature is the only factor that is statistically significant above 90% confidence level. As depicted in Fig. 2(b) and (c), adhesives A2 (thermosetting matrix with a polymer cored particle) and A3 (thermosetting matrix with a Ni particle) behaved similarly, as is perhaps expected since they are understood to have similar resin matrix materials properties (similar curing reactions based on an epoxy chemistry). None of the factors showed a statistical significance above 90% confidence level, but temperature seems to have the smallest effect on yield. The rate of application of pressure has the greatest effect on yield for material A3 but this result is only significant at 87% confidence level. However, for both A2 and A3 thermosetting materials low level factors (closer to the originally recommended factors) gave the highest yields. 2) Organic Substrate: Fig. 2(d) (f) (A1 A3) shows for the organic substrates the effect of the process factors on yield. The observed behavior with the organic substrate is significantly different to that with the ceramic substrates. In this case, temperature has the largest effect for A1 at a significance of above the 90% confidence level. The interaction between pressure and temperature has the next largest effect, also at above the 90% confidence level. The combination of high pressure and low temperature and rate of change of pressure gave best yield. For A2, pressure has the greatest effect and is significant at the 95% confidence level. The interaction of pressure and temperature have the next greatest effect, but this is not a significant result at 90% confidence levels. For A3, the thermosetting material with solid conducting particles, temperature has the greatest effect and is significant at the 95% confidence level. The effect of interaction of temperature and pressure are the next significant but not at 90% confidence level. Rate of application of pressure has the least effect for this material, Therefore, a low temperature and pressure together with a high rate of application of pressure lead to a high process yield for adhesive A2 and all low process factors, i.e. closest to those recommended by the manufacturer, give a high yield for A3. Since the manufacturer s data sheets specify only temperature and pressure, the recommendations for settings based on these results agree with the data sheet. A more practical definition of yield is that if any of the connections on a board have contact resistances greater than 1 then the whole board is classed as a failure. Since there are only three replications for each combination of factors, the data is rather sparse and the yield can only be 0%, 33%, 66%, or 100%. Some useful conclusions can however be obtained from an analysis of these component yields. The plots in Fig. 3(a) (c) (A1 A3) compare the yields obtained with the FR5 substrate with those obtained with ceramic substrates from which it can be seen that for all three adhesive materials far better yields are obtainable with the organic substrates and that whilst reasonable component yields are obtainable with both A1 and A2, the material containing very small particles, A3, gives generally poor results. B. Conductance Fig. 4(a) (f) shows the effect of the different factors on the average conductance at the end of the manufacturing process. The distribution of resistances is discussed in Section IV. 1) Ceramic Substrate: On a ceramic substrate, the level of conductance can be seen to be higher with the thermoplastic adhesive with polymer cored particles, A1. ANOVA reveals that the first order interaction of pressure and rate of change of application of pressure was found to have the greatest effect followed by temperature which has a marginally greater effect than the two other main factors. All other factors and their interactions however had a statistical level of significance lower than 90%. Adhesive A1 [Fig. 4(a)] performs better at the high level of the tested factors, perhaps reflecting the importance of improved matrix flow and increased particle deformation under these conditions. For A3 [Fig. 4(c)], the thermosetting material with small solid particles, the rate of application of pressure was found to have the most significant effect at more than a 95%

7 OGUNJIMI et al.: ASSEMBLY OF PLANAR ARRAY COMPONENTS 261 (a) (b) (c) Fig. 4. (d) (e) (f) Effect of different factors on conductance. confidence level. Higher order interactions of the factors were found to be less significant for either of the thermosetting materials A2 or A3 and all main factors had statistical levels of significance considerably lower than 90%. Both materials however followed the same trend with low level combination of parameters yielding higher conductance. 2) Organic Substrate: [Fig. 4(d) (f)] The interaction between pressure and temperature has the greatest effect on conductance of the thermoplastic material A1, with polymer cored conducting particles. This is followed closely by pressure [Fig. 4(d)]. Pressure and rate of change of pressure are significant for A2 followed by pressure and temperature interaction. Pressure however, is the most significant factor. Low temperature and pressure combined with high rate of application of pressure generally give high conductance for both A2 and A3. 3) Uniformity of resistance across the board: The uniformity of conduction across the board was investigated by computing the coefficient of variance of the conductances (i.e. the ratio of standard deviation to the mean). It can be seen from the data in Table III that the coefficient of variance is generally lower for the organic substrates. ANOVA however reveals that the rate of application of pressure and its first order interaction with temperature are the most significant at greater than 90% confidence level for the thermosetting material A2 on alumina substrate. Pressure was also found to be the most significant parameter (factor) at greater than 95% confidence level for the same material on FR5. None of the other combination of factors investigated shows significance at statistical confidence level greater than 90%. IV. DISCUSSION OF EXPERIMENTAL RESULTS The good performance of the organic substrate in these experiments may be attributable to the topography of the pads and the mating bump on the component. The component was made out of alumina printed with thick film Ag Pd Fig. 5. (a) Cross sections of the substrate pads. bumps. The topography of the two substrates is as is indicated in Fig. 5. It is considered that the flat top of the pads on the organic substrate makes it easy to retain a sufficient number of conductive particles for electrical conduction. It was also observed that the Au coated chemically etched Cu PCB pads on organic substrate are more consistent in shape than those on the thick film AgPd printed on ceramic substrates. This is perhaps responsible for the more uniform distribution of resistance observed with the organic substrates. It also shows that the organic substrate has a wider tolerance for misalignment than ceramic substrates. It should be observed that although CFD modeling indicates that the inclined walls of the ceramic bump will encourage flow of particles back onto the pad [1] during the process the difference in the effective areas of the ceramic and organic bumps is thought to have a more significant effect on the likelihood of achieving satisfactory conductance. The coefficient of thermal expansion (CTE) of the substrate with the device is worse for the organic substrate and hence is not the reason for its better results. The experiments show that the preferred processing parameters for the thermosetting matrix adhesives A2 and A3 are those originally specified by the manufacturers. This suggests that the process has probably been optimized for these products, and that these optimized conditions must be achieved to give good process yields. The thermoplastic adhesive A1 is more tolerant to variations in process conditions and has a wide process window which allows for flexibility in the process set up. (b)

8 262 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART C, VOL. 19, NO. 4, OCTOBER 1996 TABLE IV OPTIMUM PROCESS PARAMETERS FOR THE DIFFERENT MATERIALS COMBINATIONS Fig. 6. Scaled bump and joint resistances for A1 (peak at 20 m is probably a measurement error). Fig. 7. Scaled bump and joint resistances for A2. Further factors that may influence the performance of the adhesives is the conducting particle size and composition. Fig. 3 shows that the larger, polymer cored particles are more forgiving for the joint configurations investigated here since the smallest particle size adhesive, A3, performed worst overall. This conclusion is further supported by the fact that the largest particle size adhesive, A1 performed better than A2 on the ceramic substrate, where the gaps between substrate and device were most variable. The optimum process conditions for the different materials combinations which have been determined from the experimental results are given in Table IV. Finally Figs. 6 8 show the distribution of resistances found for each adhesive in these experiments. In each graph, the distribution of internal device resistances (bump resistance), measured using a standard wafer probe tester, are also plotted Fig. 8. Scaled bump and joint resistances for A3. (the total measured resistance equals the actual joint resistance this internal resistance). Only 38 internal resistance measurements were taken, but on the graphs the frequencies have been scaled up so that the total number of measurements for each of the three lines are the same. The final point on each line at 1 represents the number of opens. The results from all the trials with different parameter settings have been lumped together. If the actual joint resistances were zero, then all the lines on each graph would coincide but the graphs all show long tails in the joint resistance distributions extending well above 1 indicating that many actual joint resistances exceed this value. The reason for these high resistance joints is either that the particles have been crushed and the outer conducting skin ruptured, or that a thin film of insulating material separates the conducting particles from the electrode surfaces. This phenomena is currently under investigation. V. CONCLUSION It can be concluded that uniform conductivity and high yield are more readily obtained with the organic PCB than the thick-film ceramic substrates. However, improvements to the thick film pad geometries may improve the thick film substrate performance. It has also been shown that the optimum process conditions and adhesive material choice can be very different for ceramic and organic substrates. Significant differences in assembly performance between the adhesive materials also emerged whilst finer particle sizes have been shown to have statistical yield advantages in fine pitch applications, the larger

9 OGUNJIMI et al.: ASSEMBLY OF PLANAR ARRAY COMPONENTS 263 particle size materials with soft cores have proved superior in these trials. It can also be seen that the thermoplastic material has a greater process latitude in addition to its rework potential. Reliability testing will now be used to determine whether these advantages carry through to long term product performance. REFERENCES [1] S. H. Mannan, D. C. Whalley, A. O. Ogunjimi, and D. J. Williams, Assembly of planar array components using anisotropic conducting adhesives A benchmark study: Part II Theory, this issue, pp. XXX XXX. [2] A. O. Ogunjimi, O. A. Boyle, D. C. Whalley, and D. J. Williams A review of the impact of conductive adhesive technology on interconnection, J. Electron. Manufact., vol. 2, pp , Sept [3] I. Watanabe, K. Takemura, N. Shiozawa, O. Watanabe, K. Kojima, and T. Ohta, Flip-chip interconnection to various substrates using anisotropic conducting adhesive films, in Proc. Latest Achievements Conducting Adhesive Joining Electron. Packag., Sept. 1995, Eindhoven, The Netherlands, pp [4] L. Li and J. E. Morris, Structure and selection models for anisotropic conductive adhesive films, in Proc. 1st Int. Conf. Adhesive Joining Technol. Electron. Manufact., Berlin, Germany, Nov [5] D. J. Williams and D. C. Whalley, The effects of conducting particle distribution on the behavior of anisotropic conducting adhesives: nonuniform conductivity and shorting between connections, J. Electron. Manufact., vol. 3, pp , [6] K. Gustafsson, J. Liu, and Z. Lai, Surface characteristics, reliability, and failure mechanisms of tin, copper, and gold metallizations, in Proc. Adhesives Electron. 96, Stockholm, Sweden, 1996, pp [7] J. H. Lau, Ed., Flip Chip Technologies. New York: McGraw-Hill, 1996, chs. 6 10, pp [8] H. M. van Noort, M. J. H. Kloos, and H. E. A. Schafer Anisotropic conductive adhesives for chip on glass and other flip chip applications, in Proc. 1st Int. Conf. Adhesive Joining Technol. Electron. Manufact., Berlin, Germany, Nov Adebayo Oluyinka Ogunjimi received the Ph.D. degree from Loughborough University, U.K. After graduating he has worked as a Postdoctoral Research Associate at both Loughborough and at CALCE EPRC, and is currently working for PERA International Ltd. He has been employed in electronic s manufacturing research and development since 1990 and has taken part in projects in surface mount technology including the use of conductive adhesives, both isotropic and anisotropic, as replacements for solder in electronics interconnection. He has undertaken research in the physics of failure approach to electronics reliability. His experience includes manufacturing process modeling and design evaluation using classical theory and finite element analysis. Samjid H. Mannan received the B.A. degree in physics from Oxford University, U.K., in 1988 and the Ph.D. degree in theoretical elementary particle physics from Southampton University, U.K., in After working as a Research Assistant at Salford University and Loughborough University in the area of electronics interconnection (solder paste and ACA s), he is currently working in the same area as a Research Fellow at Loughborough University. He is the author of over 20 scientific papers. David C. Whalley received the B.Sc. and MPhil. degrees from Loughborough University, U.K. Since then he has been involved in research into electronic interconnection reliability, the use of engineering analysis techniques both in electronic product design, in process simulation, and on new interconnection technologies such as conducting adhesives. He has worked as an Engineer both at Loughborough University and at Lucas Industries Advanced Engineering Centre, Solihull, Birmingham, before being appointed in 1990 as lecturer in Processes for Electronic Manufacture within Loughborough University s Department of Manufacturing Engineering. He is the author of over 60 papers and reports in the areas of electronics manufacturing processes, interconnection technology, and electronic thermal design. David J. Williams received the B.Sc. degree at UMIST, and the Ph.D. degree at Cambridge University, Manchester, U.K. He has been Professor of Manufacturing Processes, Loughborough University, since January He was Head of the Department of Manufacturing Engineering from 1991 to His present research focuses on the resolution of problems within electronics manufacturing. This research includes work at the process engineering science level, understanding of the interaction between process understanding and design for manufacture (encompassing take back), and business globalization. This follows a number of years of control oriented work in CIM. Before taking up his present post he was a Lecturer in Manufacturing Engineering and Design in the Engineering Department, Cambridge University, U.K., and worked for industry for Metal Box and GKN. He has published three books and more than 200 papers and contributions to books in the area of manufacturing. He is Editor-in-Chief of the IJCIM and JEM, and is an Active Consultant to the manufacturing industry and various governments.