3.1.1 Organization of the chapter

In order to better explain what anisotropic adhesive (ACA, ACF, ACP, see list of abbreviations in the Appendix at the end of this chapter) interconnections are and how these perform, their nature is described first, directly followed by a summary account of the literature. The third section contains basic information about materials and processing. In line with our approach to focus on the industrial aspects of ACF-bonds, such as their reliability, the critical stress loads will be discussed and illustrated with examples in the fourth section. As a result of the performance under various types of stress, the particular sensitivity of adhesive interconnections to moisture and temperature justifies a dedicated section on evaluation methods. Much of what has been treated up to this point culminates in the case studies of the sixth section: heat seal connectors and an ultra-thin ball grid array package. Some concluding remarks are formulated in the final section.

3.3.1 Materials

Within the limitation outlined in the introduction, relevant materials are the flexible substrate, the dies with bumps, and the adhesives. Flexible foil for the electronic applications that are the subject of this chapter can be divided in three-layer and two-layer types. The three-layer foils have a polymeric base material with metallic traces attached by an adhesive layer. Such an adhesive interlayer may form a risk in the bonding process of very narrow structures because the bond pads may slide away (see Section 3.2). The two-layer foils, in contrast, have the metallic conductors attached adhesive-less to the base material. For the present treatment of adhesive interconnections applied to very fine pitches of 40 μm, three-layer foils are not very suitable. The traces are usually Cu with a Ni/Au-finish, but Al or Cu can be used as low-cost alternative metallizations.

For the investigations described in this work, the dies and the flexible foils were designed to incorporate daisy chains to monitor the connectivity of the interconnections. But the die-to-foil assembly also contained a number of single bump contacts of which resistance could be measured using a four-wire method (see Fig. 3.3 for an example of such a structure). The used flexible foils with a pitch of 100 μm had a polyimide layer of 50 μm thick, a solder resist layer of 25 μm thick, and in total a 15 μm thick Cu/Ni/Au metallization. For the 60 μm and 40 μm pitch type, the polyimide was 40 μm thick, solder resist was 12 μm, and the metallization consisted of 12 μm Cu, 2 μm Ni, and flash Au of 0.06 μm. In Figs 3.4 and 3.5, the layout of such foils is shown. Similarly, for heat seal connectors, a four-wire contact resistance measurement can be done using a dedicated test structure as given in Fig. 3.6. To overcome problems with contacting the heat seal connector, the two foil edges are bonded to the printed circuit board too. This way, a triple-bond structure is created; the contact resistance of the central interconnects can be monitored.

Adhesives can be divided according to the content of filler particles. Those with a high content of conducting particles, typically 70% or more, are isotropic conductive adhesives (ICA). The filler material is usually made of silver flakes. For much lower filling factors, in the range of 5–15% or even less, the conducting particles become effectively isolated from each other, thus blocking conduction in the lateral direction, and only conduction in the direction perpendicular to the metallic surfaces of the chip and the foil can occur. These are anisotropic conductive adhesives (ACA). The filler can consist of hard particles such as Ni, which serve to crush oxide layers on the surfaces that must be connected. Alternatively, one can apply flexible filler material that is made from Au-coated polymer spheres. Figure 3.7 shows an example of such a contact in a 100 μm-pitch assembly. The advantage of metal coated polymer spheres is that they can accommodate the relaxation of the adhesive to some degree. For completeness, non-conductive adhesives (NCA, NCF, NCP, see list of abbreviations in the Appendix at the end of the chapter) have a filler content of nil, but one may regard these essentially as anisotropic conductive adhesives as conduction can occur only in one direction. In Fig. 3.8, the distribution of 96 single-contact resistances is shown for 60 μm-pitch interconnections. For assemblies made with ACF (Au-coated polymer), the values are slightly but clearly lower than for the corresponding NCF-samples. For three adhesives, a selection of the physical properties is compiled in Table 3.2. At the glass transition temperature (T_g) these materials change from a glassy state below the transition temperature to a rubbery state above, and vice versa. In fact, the transition temperature is the center of a transition region that can extend over 10–20 degrees. Material properties such as coefficient of thermal expansion (CTE) and storage modulus change dramatically from one to the other state. In the next section on processing, it will be seen that the transition temperature depends on the curing of the adhesive and aging.

Dies with various types of contact bumps exist. In case of relatively hard Ni/Au-bumps (see Fig. 3.9), the bumps will keep their shape and the contact with the bond pad on the flexible foil is not as intimate as when using much softer galvanic Au-bumps or Au-stud bumps (see Figures 3.10 and 3.11). The Au-bump depicted in Fig. 3.10 even follows the shape of the bond pad.

In most cases described in this chapter, the dies have dimensions of 5 × 5 mm². In Table 3.3, a number of relevant parameters of these and other test dies are compiled.

3.3.2 Processing

As one can see from Table 3.1, the bonding pressure there is always low in comparison with our own settings, as detailed below. This means that the conductive particles are deformed but they are never pressed into the Au-bumps. The reason for using higher pressure is to achieve a more intimate contact between bump and track, and so open the way to use a non-conducting adhesive instead of an anisotropic adhesive.

A typical processing sequence is as follows. The flexible substrate is placed on the bonding table of a flip chip bonder and stretched over a glass plate by means of a vacuum to ensure its flatness. Then the desired amount of adhesive paste or film is applied to the substrate – in the latter case the adhesive is already partly cured which makes it suitable for cutting to size and easier handling. The component is aligned and pressed onto the substrate. Together with the temperature, bonding force and time are the important process settings.

If the bonding force is too low, the resulting clamping force (see Section 3.2) is not high enough for a reliable interconnection. Too high a bonding force has the risk of inducing cracks in the Si-die or excessive deformation of the bond pads and the foil. In particular, when thin dies have to be mounted this is a risk, as one can see in Fig. 3.12 for a die that was thinned to 160 mm and then bonded with a 250 N/mm² bump area.

For industrial processes, which are the subject of this chapter, the bonding pressure is usually higher than in many research studies. This was already mentioned in Section 3.2. In the present investigations, bonding was usually done with pressures in the range of 100–400 MPa. To avoid any misunderstanding, the bonding pressure is calculated per contact area between bumps and bond pads, not with regard to the area of the die.

The temperature–time profile determines the curing of the adhesive. For two ACF-materials, the conversion was followed by means of a differential scanning calorimeter. Figures 3.13 and 3.14 show that different adhesives have different curing profiles, but at 180 °C or higher, the differences seem to vanish. The curing condition affects the glass transition temperature. Whereas most suppliers refer to the maximum glass transition temperature after curing at medium temperatures for several minutes, this is not practical in an industrial process where bonding has to proceed in seconds rather than minutes. In that case, the transition temperature can be 10–30 degrees lower. Realistically, in industrial processes the adhesives are seldom fully cured. As will be seen in a later section, during testing at elevated temperature and humidity, post-curing can occur.

3.3.3 Moisture absorption

Although this is a materials issue, the absorption and desorption of water is such an important property of the adhesives and the flexible foils that it will be treated separately. Not only the amount of moisture that can be taken up by the materials is a critical issue, rather this must be treated in combination with the desorption and permeability of the foil to moisture (de Vries et al., 2005a). An adhesively bonded chip on foil assembly can take up moisture within a relatively short period of time, as Fig. 3.15 shows – this process depends on the materials properties of the adhesive. Although the amount of moisture the two adhesives can absorb is somewhat larger for type S-1, the diffusion coefficient of type H-3 is so much higher that, in the end, this foil–adhesive combination takes up more moisture. In a reflow process step, the water vaporizes and is forced out of the assembly. Should the diffusion rate of the water through the adhesive be larger than through the foil, then water accumulates at the interface of the adhesive with the foil and the pressure can inflict damage to the interconnections. If this is the other way around, the water may pass safely from the adhesive, through the foil to the outside world. Making a reliable adhesive interconnection depends therefore, among other things, on the proper combination of materials. Desorption curves of water from saturated flexible foils from room temperature to reflow temperatures of 260 °C are compiled in Fig. 3.16. In Table 3.4 the moisture absorption and the diffusion coefficients for selected flexible foils and adhesives are listed. In all cases, the moisture diffusion coefficient of the adhesives is higher than for the flexible foils. The best combination therefore includes a foil with a high diffusion coefficient. Since adhesive-less foils are preferred for their better stability, only two flexible foil types from this list are applicable.

3.4.1 Bonding

Factors affecting the eventual clamping force inside adhesive interconnections are the bonding force, and the shrinkage of the adhesive during curing and during cooling down to room temperature. The higher the clamping force, the less sensitive the interconnection becomes to changes such as increases in temperature, and swelling of the adhesive because of moisture absorption. This can be seen clearly from Fig. 3.2. Figure 3.17 shows how the contact resistance develops during the bonding process. The contact resistance is monitored in-situ during the bonding process; in this example, bonding is done at 150 °C and at 180 °C. The lower part of Fig. 3.17 gives the temperature–force–time profile that was used during the bonding process. The top part of Fig. 3.17 shows how the electrical contact is formed, using the contact resistance as a criterion. If bonding is done at 180 °C, the electrical contact is formed during the bonding process. When the bonding pressure is released, the interconnect cools down and, as a result, the contact resistance decreases because of the T-dependence of the resistance and the shrinkage of the adhesive. If bonding is done at 150 °C, no good electrical contact is formed during the 10 second bond time. When the bonding pressure is released, the resistance increases. Only during cooling down to room temperature, does the shrinkage of the adhesive create an electrical contact. The eventual contact resistance, however, is higher than for bonding at 180 °C, and one can expect that if this assembly is used at a higher temperature, the electrical contact will be lost. This means that 10 seconds at 150 °C is too short to form a good interconnect for this type of adhesive.

3.4.2 Resistance to reflow soldering

Conductive adhesive interconnections are often used in multi chip modules. This means that afterwards, the modules are soldered to a printed circuit board. Hence, the first process step after bonding that needs attention often is (reflow) soldering of the semiconductor package. Accumulated moisture in the polymeric matrix of the flexible foil and the adhesive will evaporate vigorously and can damage the adhesive bond.

In the electronic assembly industry, it is common practice to classify packages by their robustness in the reflow-soldering process. For this, a standardized method has been defined. Once any type of package has been given such moisture classification or moisture sensitivity level (MSL), process engineers will know how to handle the packages in order not to inflict any damage to the assembly. Usually, one follows the JEDEC-standard (JEDEC, 2008) as shown in Table 3.5. In brief, through this procedure, products are preconditioned by subjecting them to one of the standard or accelerated conditions, and then exposing them three times to the appropriate reflow profile. Prior to preconditioning, and after reflowing, the devices are evaluated for delamination and also one or more relevant parameter is measured; the occurrence of delamination or changes in the measured parameters determine whether or not the moisture sensitivity level is met. If it is met, the test is repeated at the next higher (more severe) level, until eventually the packages fail the test. Then the last successful test level is assigned as the moisture sensitivity level for that particular type of package. For adhesive interconnections, the value of the contact resistance is a suitable criterion since it is sensitive to variations in the clamping force, as explained above.

It must be noted that the adhesives that are discussed in this section are not fully cured, as was already mentioned in the introductory section. Therefore, in contrast to the observations of Yin et al. (2003), the adhesives are not completely resistant to absorption of water and thus will take up water with due consequences.

An example for such assessment of the moisture sensitivity level is given in Fig. 3.18 for a foil design with 300 μm and 200 μm pitch dies and NiAu-bumps of 20 μm height bonded with a 150 N/mm² bump area, see De Vries and Janssen (2003). For this experiment, the foil labeled D-1 from Table 3.4

Ambient was used. Sixty daisy-chained samples were taken and split in three series of twenty. These samples were characterized initially, divided in three groups, and then each set was treated differently: keeping under ambient conditions (20 °C/50%RH/24 hours); soaked (85 °C/85%RH/24 hours); and dried (85 °C/24 hours). Subsequently, the assemblies were fed through a reflow furnace set at a peak temperature of 230 °C. This does not exactly correspond to the JEDEC-levels, but one can clearly see the effect of the moisture content on the resistance. Temperature had a minor effect on the resistance value which had increased also after drying. Storing at ambient also caused one daisy chain to fail, which could have been an early failure. But most impressive is the result after humidity storage. All resistance values increased, and there were three failures.

The second example shows the additional influence of bonding pressure. The same type of foil and adhesive was used as described above. This time the samples did receive the standard preconditioning at MSL2 (see Table 3.5, accelerated equivalent condition). From Fig. 3.19 one sees that increasing the bonding pressure from 50 g/bump to 100 g/bump reduced the initial resistance of the daisy chains. (The dies with 300 μm pitch had 56 bumps, the dies with 200 μm pitch counted 84 bumps.) After preconditioning and reflow, in all cases the resistance increased. But the samples with the higher bonding pressure had still the lowest resistance value. This is in line with Section 3.4.1.

The third example serves to illustrate the aforementioned damage model of moisture being expelled and thus inflicting damage to the interconnection.

For the purpose of enhancing the flatness of the flexible foil during the bonding process, a thin metallic stiffener was applied to the backside of the foil; see the cross-section in Fig. 3.20. Such assemblies were subjected to the moisture sensitivity test as explained previously – preconditioning (MSL2, see Table 3.5) and three times through a reflow oven. The initial state is comparable to the cross section of Figures 3.7 and 3.10. The adhesive was of type H-3 and the foil was of type N-6 (see Table 3.4). The layout of Fig. 3.4 was used with a pitch of 100 μm. Without the stiffener, the moisture can evaporate through the adhesive and the foil, the resistance after reflow was still good – in Fig. 3.21 a cross-section after the MSL-test is shown. In contrast, with such a stiffener, the accumulated water could not escape fast enough: only the route parallel to the surface of the foil is available, which is two orders of magnitude longer than its thickness. The result was a lift-off of the die with the bump from the bond pad, as one can see in the cross-section shown in Fig. 3.22. From this it is clear that this method of applying a non-permeable stiffener is not feasible, but the results of this experiment support the failure mechanism proposed in Section 3.3.3.

3.4.3 Accelerated testing

Apart from testing the resistance against processing, assessment of the long-term behavior of adhesive interconnections under external stress conditions is relevant. Here, one is led by the nature of an adhesive interconnection to select the most suitable environmental tests. This means that thermal cycling and exposure to a humid environment, either constant or periodic, are the most severe test conditions. In particular, it is of importance to investigate the possibility to design accelerated tests. To this end, similar assemblies as were used for the moisture sensitivity experiments described in Section 3.4.2 were exposed to a damp heat test (85 °C/85%RH). The resistance of the daisy chains was periodically measured by taking the samples out of the climate chamber. In Fig. 3.23 such a series of measurements is shown (de Vries and Janssen, 2003). Each symbol in the graph represents the moment when the samples were taken out of the test for inspection. The differences in the initial values of the resistance have to do with the different length of the daisy chains. The 200–300 mm pitch assemblies consisted of four dies, two for each pitch, and had ten daisy chains on one side of a die and two that extended over three sides of a die. A failure was – arbitrarily – defined as a factor of 1.5 increase of the resistance compared to the initial value. The prolonged exposure to damp heat caused the bond to detach. A gap between the adhesive and the foil of several micrometers occurred and the NiAu-bump was lifted from the bond pad (Fig. 3.24). In Fig. 3.25 the failure distributions are given, which make clear that as the bonding pressure increased, the lifetime of the samples is longer.

To further accelerate the degradation process it is common practice to increase the stress level, provided that the failure mechanism remains the same. For the damp heat test, this can be achieved only by increasing the temperature. So called ‘highly accelerated stress testing’ (HAST) is a suitable tool for this. Such equipment allows temperatures of up to 140 °C while the humidity can be controlled to 96%RH. With respect to the glass transition temperature of the adhesives in the range of 125 °C and higher, the temperature in these highly accelerated test was limited to 110 °C. The resistance of single contacts in a 100 |mm-pitch assembly, subjected to 110 °C/85%RH after MSL2 preconditioning and reflow, was recorded periodically and the result is shown in Fig. 3.26 (de Vries et al., 2005a). One can see the same generic behavior as for the larger pitch assembly tested at 85 °C/85%RH shown in Fig. 3.23. Also in this test, the failure mode is lifting of the contact bump from the bond pad, as one can see in Fig. 3.27. Thus both tests, the standard or accelerated damp heat test of 85 °C/85%RH and the highly accelerated version at 110 °C/85%RH, can be applied to investigating the reliability of conductive adhesive interconnections.

Another illustration to this effect is Fig. 3.28, where the failure distributions of the two damp heat tests on 100 μm assemblies is given for two adhesives. At 85 °C/85%RH, the two adhesives behaved quite similarly, while at 110 °C/85%RH there was some difference. But in both adhesives, the acceleration of the test is clear. The statistical result is: for H-3 a Weibull slope of β = 1 in 110 °C/85%RH and β = 3.8 in 85 °C/85%RH; for S-1 this is β = 0.8 and β = 2.1 respectively.

For reasons of completeness, also temperature cycling results are presented. The same type of assemblies as those of Fig. 3.26 was subjected to a thermal shock test (− 55 °C/+ 125 °C/20′−20′-cycle), after MSL2 and reflow. Two things are worth noting. First, the increase of resistance was far less pronounced than in the humidity test, even though the maximal temperature was higher. Second, one notices the irregular behavior of some contacts which also was found in the humidity tests. There is strong reason to ascribe this to taking the samples out of the test chamber for off-line electrical measurements. This will be addressed in Section 3.5 on evaluation methods.

3.4.4 On-line versus off-line monitoring

Figure 3.29 illustrates the power of on-line monitoring of the adhesive interconnects: in this example, showing the resistance drift over time, we can clearly distinguish three different phases. First, we get a steep increase in contact resistance as a result of temperature increase and of moisture take-up by the adhesive, when going from room temperature conditions to 85 °C/85%RH conditions. In the next phase, the contact resistance shows a decreasing trend, most likely due to a post-cure effect. Finally, the contact resistance starts to increase due to aging at high temperature and humidity conditions. Off-line monitoring would have shown only that the contact resistance did slightly change as a result of the accelerated test.

In Fig. 3.30 we show typical differences as observed between off-line and on-line resistance monitoring. The on-line resistance data show a decreasing trend, suggesting a post-cure. The off-line measurements clearly show an increasing contact resistance. Possible explanations for this difference are damage from handling during off-line monitoring or from the sudden change in temperature and moisture every time when taking the samples out of and putting them back into the test chamber. A similar observation was made during another series of experiments: the samples were continuously monitored on-line, but the test chamber needed to be opened shortly periodically to refill the water reservoir. Figure 3.31 shows the decrease in the measured contact resistance every time the test chamber was opened, which resulted in a decrease in temperature of the samples under test. In some cases, however, a sudden increase in the contact resistance could be observed after opening and closing the chamber, as is shown in Fig. 3.31. Sometimes, this then was the start of a visible further higher degradation rate.

3.6.1 Heat seal connector

A heat seal connector is an example of the use of an interconnection with a thermoplastic anisotropic conductive adhesive. The connector consists of a flex foil with a C-pattern screen printed on it. On top of this, the adhesive layer with conductive particles is applied. The interconnect is made by pressing the flex to the substrate using a hot bar (thermode). It is a low-cost option to connect, for example, an LCD with the printed circuit board containing the driver electronics. The aging kinetics of this type of adhesive interconnect were determined under static temperature loading. Heat seal connectors to a printed circuit board were aged at 60 °C, 70 °C, and 80 °C. To be able to monitor the contact resistance on-line, the triple track test structure as shown in Fig. 3.6 was used. Figure 3.34 shows the actual samples. The resistance of the inner row of contacts is monitored. The outer contacts are used to be able to electrically contact the flex during the accelerated test. The test structure allows one to monitor continuously the contact resistance drift of all the interconnects of the heat seal connector during the high temperature storage. Figure 3.35 shows the result for some interconnects at 80 °C. It shows that the resistance of the interconnect increases in the range of a few ohms to 50 ohms over 500 hrs testing at 80 °C. The interconnects with the highest drift in resistance are the outer ones (Track 1 and 2). The rounding at the thermode edge explains the large difference with the inner contacts in this example. Applying an arbitrary or functional failure criterion, e.g. of 10% resistance increase, allows one to determine times to failure for all the individual interconnections in the test and to make cumulative failure distributions for the three temperature levels. The cumulative failure distributions for the tests at 60 °C, 70 °C, and 80 °C, assuming a lognormal distribution, are shown in Fig. 3.36. If an Arrhenius T-dependence of the contact degradation is assumed, the activation energy for aging at static temperature loading can be estimated. For the type of HSC connectors used in this study, the estimated activation energy for dry aging, Ea ≈ 1 eV. From the test results and from this activation energy, the acceleration transform and the expected life time (TTF) at any field temperature level can be calculated for this type of interconnect:

[3.4]

3.6.2 Ultra-thin fine pitch ball grid array package

In this section, the application of conductive adhesives as first level interconnects in actual flip-chip packages will be treated. This concerns an ultra-thin ball grid array package of about 300 μm thickness; Fig. 3.37 shows a schematic of such a device (in Fig. 3.5 the footprint is shown). Because of this so-called ‘die down’ layout, the thickness of the die must be in the range of 160 μm. This poses a limitation to the bonding pressure in order not to damage the thin die. A further requirement concerns the subsequent processing: the second level solder balls are made by stencil printing of solder paste followed by a reflow step. Finally, these packages are meant for soldering on a rigid or flexible substrate. Thus, the flip chip on flex construction has to withstand at least two reflow process steps.

For assessment of the moisture sensitivity level, the same procedure as described in Section 3.4.2 applies. This part of the investigation was done by mounting dies with a pitch of 100 μm to foil substrates. Two different anisotropic adhesives were used (S-1 and H-3, see Table 3.4). The die of 160 μm thickness – as was mentioned in Section 3.3.2 (Fig. 3.12) – is more vulnerable than the standard type of 600 μm and therefore the bonding pressure was kept at a moderate level of 200 MPa. Further, preconditioning at one level lower (MSL3, see Table 3.5, accelerated equivalent condition) should be done as well. After the moisture–reflow test, the endurance performance was assessed by subjecting the assemblies to damp heat and thermal shock tests. The result of the moisture sensitivity experiment is presented in Fig. 3.38. Apart from a few outliers, between the two adhesives no significant differences occur. Of more importance is the fact that the two preconditioning levels do not lead to any significant difference in the distribution of the contact resistance. The yield of the contact, defined as the number of good contacts after the test relative to the initial number of contacts, was 100% at MSL2 and MSL3. Subjecting the same assemblies – after receiving the MSL3 reflow treatment – to stress tests, showed no difference between the two adhesives in the damp heat test (85 °C/85%RH) and thermal shock test (− 55 °C/+ 125 °C/20′−20′ cycle) as is seen in the Weibull failure distributions (Figs 3.39 and 3.40). In the highly accelerated damp heat test (110 °C/85%RH, Fig. 3.41), the Weibull distributions do differ in that the adhesive S-1 performs slightly better, but both distributions have become broader than in the other damp heat test. (Note that the data of Figs 3.39 and 3.41 are the same as those in Fig. 3.28.) This indicates, in principle, the possibility of another failure mechanism, but earlier experience as described in Section 3.4.3 has shown that the failure modes for the two types of humidity stress test are the same.

The next step brings us to assemblies with still smaller pitches of 60 μm and 40 μm (see de Vries et al., 2005b). In this case also a non-conductive adhesive was applied. Cross-sections of typical examples of these assemblies are shown in Figs 3.42 and 3.43. One can see the very intimate contact between the bumps and the bond-pads. MSL3 was chosen for preconditioning. The result is given in Fig. 3.44. Of the 60 μm-pitch assemblies, 96 contacts were tested and 128 in the 40 μm-pitch version. A difference between anisotropic and non-conducting adhesives was observed: the samples made with the latter performed significantly better in the reflow test. Tentatively, this can be attributed to post curing of the adhesive, as has been reported elsewhere for assemblies on FR4 (Seppälä et al., 2001) and on flexible foils (Yin et al., 2003). The same difference between the two types of adhesive occured in the stress tests carried out after the reflow test. In the damp heat test at 110 °C/85%RH, the 40 μm-pitch assemblies made with non-conductive adhesive were even better than the 100 μm-pitch version based on conductive adhesive, as shown in Fig. 3.45. Similar observation can be made for the test at the milder condition of 85 °C/85%RH, where the results in Fig. 3.46 show that the ACF-samples with a pitch of 100 μm failed earlier than the 60 μm samples, but in the non-conductive adhesive assemblies of 60 mm- pitch only one failure was detected after 2400 hours in test. That very first result is represented in Fig. 3.46 by the probability line through this one data point assuming the same Weibull-slope (β = 2.4) as for the 100 μm version (dashed line in Fig. 3.46).

The longevity of fine-pitch anisotropic adhesive interconnections has thus been well established. In 80 μm pitch assemblies, Palm et al. (2001a) reported no failures for two types of adhesive after 2000 hours accelerated damp heat testing (85 °C/85%RH), while in one other type, three failures were detected. However, in the thermal shock test (− 40 °C/+ 125 °C), more failures occurred in two of these adhesives while one passed the test that was terminated after 1880 cycles. Assemblies of similar pitch were made by Kim et al. (2004) and the resistance increased almost linearly with time in 1000 hours damp heat testing (80 °C/85%RH). In the accompanying thermal shock test (− 15 °C/+ 100 °C), no increase of the resistance was found after 1000 cycles. The mechanical robustness of adhesively-bonded samples with a pitch of 80 mm was studied (Lu and Chen, 2008), together with the performance in the damp heat test (85 °C/85%RH). Under these circumstances the contact resistance increased sharply in the first 100 hours, and then remained constant depending on the bonding conditions. If the bonding process was done at low temperature (160 °C or lower) the resistance continuously increased. Bonding at temperatures up to 190 °C lead to resistance values that were constant until after 900 hours, when the resistance became slightly higher. All the samples reported in this paragraph were not subjected to a reflow process, as were those of the earlier mentioned study of Yin et al. (2003). In this latter case, the ~ 80 μm pitch assemblies survived at least 500 hours in the accelerated damp heat test (85 °C/85%RH) when the test was stopped. The resistance only moderately increased.

3.6.3 Anisotropic conducting film (ACF) versus nonconducting film (NCF)

In this final section of the case studies, the differences between these two types of anisotropic conductive adhesives is addressed. Remember that, in principle, NCA is also an anisotropic conductive adhesive.

In the first place, basic reasoning tells us that the conductivity is less for NCA than for ACA, but the electrical isolation is better. In particular, for very small pitches this is an important issue, because the gap between the bumps is perhaps just one half of the pitch. Typically, the diameter of the conductive filler particles is in the range of 5 μm, while the gap is 20 μm. Thus, a distinct risk for short circuiting exists, in particular if one realizes that the number of interconnections is extremely high. A brief statistical argumentation may be illustrative. Assume a two-dimensional situation. By means of the Poisson distribution, one can describe the probability that a bridge of conductive filler particles is formed between two adjacent bumps:

[3.5]

The parameter k is the critical number of particles needed to form such bridge, and equals g/s, where g is the width of the gap between the two bumps, and s the diameter of the particles (see Fig. 3.47). Further, there are a number of possible bridges equal to L/s, where L is the length of the bump. The parameter μ is the average number of particles per surface area, which can be estimated as follows. Taking the particle density as 1.5 × 10⁶ /mm³ (De Vries et al., 2005b), one calculates the number of particles in the space between the bumps (see Table 3.3). Here it is assumed that the particle size distribution is monodisperse. In the next step we state that the particles are evenly distributed over the height (h) of the bump, so that there are h/s layers with particles. This gives a value for the average μ. For various widths of the gap and diameters of the particles, the probability was calculated for a flip chip with 400 bumps that a conducting bridge occurs. In Fig. 3.48 the result makes clear that the combination of particle size and gap width is critical in the region below 40–50 μm. For the calculations the gap was taken as one half of the pitch. The authors emphasize that this statistical approach is meant only to illustrate the issue of conductive adhesives in a very small pitch, and that a handful of assumptions were made.

In the second place, we have presented various experimental results. The remarkable difference in performance between non-conductive adhesives and those with (Au-coated polymer) filler particles after the reflow test procedure needs further explanation. The filling factor is comparatively low, and both adhesives have the same base material. still, a different failure mechanism can be conceived. Upon heating up, the adhesive passes through the glass transition point (which is in the range of 125–145 °C, see Table 3.2) and thus possesses a low elasticity and accordingly a higher coefficient of thermal expansion. In contrast, the metallized polymer particles with a much higher glass transition temperature (~ 300 °C) will retain their elasticity and relatively low thermal expansion coefficient. The connection between the bump and the track can thus become partly lost during heating up, and the stiffer filler particles, in the absence of an external pressure, will hamper restoration of the connection upon cooling down. The diagram of Fig. 3.2 gives circumstantial support for this theory. It shows the resistance of a chain of conducting adhesive interconnections after 2400 hours at 110 °C/85%RH, when the assembly was again put under gradually increasing pressure. Thus, part of the contact area between the bumps and the tracks might be permanently lost. in non-conductive adhesives, this mechanism is obviously not present.

Regarding the differences in the reliability performance between interconnections made from conductive and non-conductive adhesives, Liu and Lai (2002) were one of the first to report. They tested ACF- and NCF-based samples on FR4-substrate. The pitch was 100 μm. in general, the failure distributions of the anisotropic conductive adhesive joints were broader than those of the non-conductive adhesive. As was mentioned in section 3.2, Kim et al. (2008a) investigated the resistance against reflow of ACF- and NCF-bonds in 200 μm-pitch samples. The ohmic behavior of the interconnections changed in the non-conductive adhesives after one or two reflow cycles, while in the anisotropic conductive version the resistance was ohmic up to two or three reflow cycles. On the other hand, the value of the resistance of the NCF-samples remained constant whereas in the ACF-samples it gradually increased. No endurance stress tests were done. Flip chip assemblies with a pitch of 60 μm were made (Chiang et al., 2006). These were put to test in a pressure cooker (121 °C/100%RH). The resistance was measured at regular intervals and was found to increase by a factor of four after 192 hours. The same type of samples showed a factor of 2.5 higher resistance after 500 hours in high-temperature storage (150 °C).

Farley et al. (2009) observed that with non-conductive adhesive interconnects, the bonding conditions could result in a diffusion bond between the Au-bump and the Au-finish on the substrate. As a result, the bonding mechanism, and hence also the degradation mechanism, will become completely different. This observation could be an alternative explanation for the different behavior between ACAs and NCAs.

Caers, J., de Vries, J., Zhao, X.J., Wong, E.H. Some characteristics of anisotropic conductive and non-conductive adhesive flip chip on flex interconnection. J. Semiconductor Techn. and Science. 2003; 3:122–131.

Caers, J., Zhao, X.J., Lekens, G., Dreesen, R., Croes, K., Wong, E.H. Moisture induced failures in flip chip on flex interconnections using anisotropic conductive adhesives. Proc. Int. IEEE Conf. on the Business of Electronic Product Reliability and Liability. 2003; 171–176.

Caers, J., Zhao, X.J., Wong, E.H., Ong, C.K., Wu, Z.X., Ranjan, R. ‘Prediction of moisture induced failures in flip chip on flex interconnections with non-conductive adhesives’, 53rd Electr. Comp. and Techn. Conf. 2003; 1176–1180.

Cao, L., Li, S., Lai, Z., Liu, J. Formulation and characterization of anisotropic conductive adhesive paste for microelectronics packaging applications. J. Electr. Mater. 2005; 34:1420–1427.

Chan, Y.C., Tan, S.C., Lui, N.S.M., Tan, C.W. ‘Electrical characterization of NCP-and NCF-bonded fine-pitch flip-chip-on-flexible packages’, IEEE Trans. Adv. Pack. 2006; 29:735–740.

Chang, S.M., Jou, J.H., Hsieh, A., Chen, T.H., Chang, C.Y., Wang, Y.H., Huang, C.M. Characteristic study of anisotropic-conductive film for chip-on-film packaging. Microelectronics Reliability. 2001; 41:2001–2009.

Chiang, W.K., Chan, Y.C., Ralph, B., Holland, A. Processability and reliability of non-conductive adhesives (NCAs) in fine-pitch chip-on-flex applications. J. Electr. Mater. 2006; 35:443–452.

Divigalpitiya, R., Hogerton, P. Contact resistance of anisotropic conductive adhesives. Proc. IMAPS. 2003; 471–476.

Farley, D., Kahnert, T., Sinha, K., Solares, S., Dasgupta, A., Caers, J., Zhao, X.J. ‘Cold Welding: A New Factor Governing the Robustness of Adhesively Bonded Flip-Chip Interconnects’, 59th Electr. Comp. and Techn. Conf. 2009; 67–73.

Frisk, L., Cumini, A. Effect of substrate material and thickness on reliability of ACA bonded FC joints. Soldering & Surface Mount Techn. 2009; 21(3):16–23.

Frisk, L., Ristolainen, E. Flip chip attachment on flexible LCP substrate using an ACF. Microelectronics Relilability. 2005; 45:583–588.

Ikeda, T., Kim, W.K., Miyazaki, N. ‘Evaluation of the delamination in a flip chip using anisotropic conductive adhesive films under moisture/reflow sensitivity test’, IEEE Trans. Comp. Pack. Techn. 2006; 29:551–559.

Jagt, J. Reliability of electrically conductive adhesive joints in surface mount applications. In: Liu J., ed. Conductive Adhesives for Electronics Packaging. Electrochemical Publications, 1999.

Jagt, J., Buijsman, A., Lijten, G. ‘Flip chip technologies with adhesives – possibilities and challenges’, Adhesives in Electronics, 4th International Conference on Adhesive Joining & Coating Technology in Electronics Manufacturing, 18–21 June 2000. Espoo, Finland: Helsinki University of Technology; 2000.

Janssen, E., Kums, G. The development of an industrial technology for flip chip on flex circuitry. Proc. Semicon. 2000; 179–184.

JEDEC. Moisture/Reflow Sensitivity Classification for Nonhermetic Solid State surface Mount Devices, 2008. [IPC/JEDEC J-STD-020D.1].

Jokinen, E., Ristolainen, E. Flip chip joining of thin chips on flexible PEN substrate, 2001. [Proc. 2001 Int. Symp. Microelectr, 600–604.].

Jokinen, E., Seppälä, A., Pienimaa, A., Ristolainen, E. Reliability of ACF flip chip joint on flexible substrate, 2000. [37th IMAPS Nordic Conf., 192–196.].

Kim, J.Y., Kwon, S., Ihm, D.W. Reliability and thermodynamic studies of an anisotropic conductive adhesive film (ACAF) prepared from epoxy/rubber resins. J. Mater. Proc. Techn. 2004; 152:357–362.

Kim, J.W., Jung, S.B. Behavior of anisotropic conductive film joints bonded with various forces under temperature fluctuation. J. Electr. Mater. 2007; 36:1199–1205.

Kim, J.W., Jung, S.B. Effect of bonding force on the reliability of the flip chip packages employing anisotropic conductive film with reflow process. Mater. Sc. Eng. 2007; A452–453:267–272.

Kim, J.W., Kim, D.G., Lee, Y.C., Jung, S.B. ‘Analysis of failure mechanism in anisotropic conductive and non-conductive film interconnections’, IEEE Trans. Comp. Pack. Techn. 2008; 31:65–73.

Kim, J.W., Koo, J.M., Lee, C.Y., Noh, B.I., Yoon, J.W., Kim, D.G., Park, S.K., Jung, S.B. Thermal degradation of anisotropic conductive film joints under temperature fluctuation. Int. J. Adh. Adh. 2008; 28:314–320.

Lai, Z., Liu, J. ‘Anisotropically conductive adhesive flip-chip bonding on rigid and flexible printed circuit systems’, IEEE Trans. Comp. Pack. Man. Techn. 1996; B19:644–660.

Lam, D., Yang, F., Tong, P. ‘Chemical kinetic model of interfacial degradation of adhesive joints’, IEEE Trans. Comp. Pack. Techn. 1999; 22:215–220.

Lawrence Wu, C.M., Chau, M.L. Degradation of flip chip on glass interconnection with ACF under high humidity and thermal aging. Soldering & Surface Mount Techn. 2002; 14:51–58.

Li, L., Morris, J., Liu, J., Lai, Z., Ljungkrona, L., Lai, C. Reliability and failure mechanism of isotropically conductive adhesive joints, 1995. [Proc. 45th Electr. Comp. & Techn. Conf, 114–120].

Ling, S., Binh, L., Lew, A., Nhan, E. A study on advanced flip-chip interconnect technologies for space application. Proc,HD Int Conf. High-dens. Interconnect and Syst. Pack. 2001; 131–136.

Liu, J. ‘On the failure mechanism of anisotropically conductive adhesive joints on copper metallization’, Int. J. Adhesion and Adhesives. 1996; 16:285–287.

Liu, J. Recent advances in conductive adhesives for direct chip attach applications. Microsystem Technologies. 1998; 5:72–80.

Liu, J., Lai, Z. Reliability of anisotropically conductive adhesive joints on a flip-chip/FR4 substrate. Advances in Electronic Packaging. 1999; 26:1691–1697.

Liu, J., Lai, Z. Reliability of anisotropically conductive adhesive joints on a flip-chip/FR4 substrate. J. Electr. Packaging. 2002; 124:240–245.

Lu, S.T., Chen, W.H. ‘Reliability of Ultra-thin Chip-on-Flex (UTCOF) with Anisotropic Conductive Adhesive (ACA) Joints’, 58th Electr. Comp. and Techn. Conf. 2008; 1287–1293.

Määttänen, J., Palm, P., De Maquillé, Y., Bauduin, N. Development of fine pitch (54 mm) flip chip on flex interconnection process. IEEE Polytronic: Proc; 2002. [155159].

Majeed, B., Paul, I., Razeeb, K.M., Barton, J., Ó’Mathúna, S.C. ‘Effect of gold stud bump topology on reliability of flip chip on flex interconnections’, IEEE Trans. Adv. Pack. 2007; 30:605–615.

Miessner, R., Aschenbrenner, R., Reichl, H. ‘Reliability study of flip chip on FR4 interconnections with acá’, 45th Electr. Comp. and Techn. Conf. 1999; 595–601.

Nysaether, J., Lai, Z., Liu, J. ‘Thermal cycling lifetime of flip chip on board circuits with solder bumps and isotropically conductive adhesive joints’, IEEE Trans. Adv. Pack. 2000; 23:743–749.

Olliff, D., Gaynes, M., Kodnani, R., Zubelewicz, A. Characterizing the failure envelope of a conductive adhesive. Int. Symp. Adv. Pack. Mater. 1997; 124–126.

Palm, P., Määttänen, J., Tuominen, A., Ristolainen, E. Reliability of 80 μm pitch flip chip attachment on flex. Microelectronics Reliability. 2001; 41:633–638.

Palm, P., Määttänen, J., De Maquillé, Y., Picault, A., Vanfleteren, J., Vandecasteele, B. Reliability of different flex materials in high density flip chip on flex applications. First Int. IEEE Conf. Polymers and Adhesives in Microelectronics and Photonics. 2001; 224–229.

Palm, P., Määttänen, J., De Maquillé, Y., Picault, A., Vanfleteren, J., Vandecasteele, B. Comparison of different flex materials in high density flip chip on flex applications. Microelectronics Reliability. 2003; 43:445–451.

Rusanen, O., Lenkkeri, J. Thermal stress induced failures in adhesive flip chip joints. Int. J. Microcircuits and Electr. Pack. 1999; 22:363–369.

Seppälä, A., Pienimaa, S., Ristolainen, E. Flip chip joining on FR4 substrate using ACFs. Int. J. Microcircuits and Electr. Pack. 2001; 24:148–159.

de Vries, J. ‘Failure mechanism of anisotropic conductive adhesive interconnections in flip chip ICs on flexible substrates’, IEEE Trans. Comp. Pack. Techn. 2004; 27:161–166.

de Vries, J., Janssen, E. Humidity and reflow resistance of flip chip on foil assemblies with conductive adhesive joints. IEEE Trans. Comp. Pack. Techn. 2003; 26(3):563–568.

de Vries, J., van Delft, J., Slob, C. 100 mm pitch flip chip on foil assemblies with adhesive interconnections. Microelectronics Reliability. 2005; 45:527–534.

de Vries, J., Delft, J. van, Slob, C. SMT-compatibility of adhesive flip chip on foil interconnections with 40 mm pitch. IEEE Trans. Comp. Pack. Techn. 2005; 28(3):499–505.

Wu, S.X., Hu, K.X., Yeh, C.P., et al. Contact reliability modeling and material behavior of conductive adhesives under thermomechanical loads. In: Conductive Adhesives for Electronics Packaging. Electrochemical Publications; 1999.

Wu, Y.P., Alam, M.O., Chan, Y.C., Wu, B.Y. Dynamic strength of anisotropic conductive joints in flip chip on glass and flip chip on flex packages. Microelectronics Reliability. 2004; 44:295–302.

van Wuytswinkel, G., Dreezen, G., Luijckx, G. The effects of temperature and humidity aging on the contact resistance of novel electrically conductive adhesives. Proc. IEEE Polytronic. 2002; 225–230.

Yim, M.J., Paik, K.W. ‘Effect of non-conducting filler additions on ACA properties and the reliability of ACA flip chip on organic substrates’, IEEE Trans. Comp. Pack. Techn. 2001; 24:24–32.

Yin, C.Y., Alam, M.O., Chan, Y.C., Bailey, C., Lu, H. The effect of reflow process on the contact resistance and reliability of anisotropic conductive film interconnection for flip chip on flex applications. Microelectronics Reliability. 2003; 43:625–633.