Low-Temperature High-Throughput Assembly Technology for Transducer Array in Medical Imaging Applications

Transcription

1 2017 IEEE 67th Electronic Components and Technology Conference Low-Temperature High-Throughput Assembly Technology for Transducer Array in Medical Imaging Applications Hoang-Vu Nguyen, Nu Bich Duyen Do and Knut E. Aasmundtveit Department of Microsystems University College of Southeast Norway Borre, Norway Abstract Isotropic conductive adhesives (ICAs) based on metal-coated polymer spheres (MPS) have shown high potential for low-temperature, high-throughput assembly of a transducer array on a substrate in ultrasound imaging applications. The process of bonding and subsequently dicing a transducer stack on a flexible substrate was evaluated. The bonding material was MPS-based ICAs containing a commercial epoxy filled with Ø5 μm Au-coated mono-disperse polymer spheres in two different concentrations, the lower being somewhat above the percolation threshold whereas the higher being close to the maximum concentration for processability (such as dispensing, printing). ICA interconnects in test samples undergoing the entire assembly process exhibit high mechanical shear strength and sufficient electrical resistance even after the tough dicing. In addition, a high process yield obtained in this work is very promising for implementation in industry. The present work has demonstrated the applicability of MPS-based ICAs for assembling ultrasound transducer arrays on system substrates. (Lead Zirconate Titanate), the process temperature should be well below 150 C. The rapid development of new imaging diagnostic modalities demands high-resolution 3D ultrasound images, even in real time. That means an advanced ultrasound probe must have a 2D transducer array with an extremely high number of elements, typically several thousands. The assembly of such transducer array includes bonding a large transducer stack to a system substrate, and then a controlleddepth dicing of the bonded stack to form an array of individual elements, as illustrated in Fig. 1. Such a process requires bonding technologies that can provide a high yield as well as reliable interconnects with good electrical and mechanical performance to withstand the tough dicing. In addition, the chosen technologies should be compatible with high-throughput manufacturing and avoid occupying manufacturing equipment (such as bonders) for an extensive amount of time. Keywords-assembly; transducer array; isotropic conductive adhesives; metal-coated polymer spheres I. INTRODUCTION Medical imaging is an important tool to perform medical diagnosis. Ultrasound imaging technique has become one of the most utilized forms of diagnostic imaging available today. Ultrasound equipment is easy to use, mobile and relatively cheap compared to other techniques such as MRI (Magnetic Resonance Imaging) and x-ray CT (Computed Tomography). Most medical ultrasound transducers are normally built as a multilayered structure with different functional layers such as dense materials for reflection; active piezoelectric layer; layers for acoustic impedance matching to tissue [1]. The piezoelectric material is polarized to act as sensing and actuating components of the transducers. The material cannot sustain polarization at a temperature above its Curie temperature. Therefore, the temperature of the ultrasound transducers during manufacturing and operating must not exceed the Curie temperature of the piezoelectric material. In the manufacturing of the transducers, exceeding this depoling temperature requires re-polarization of the piezoelectric elements, adding more complexity in the process. For standard piezoelectric materials such as PZT Figure 1. The assembly process of an ultrasound transducer array on a system substrate; (a) bonding a large transducer stack to a system substrate and (b) subsequently dicing the bonded stack to form an array of individual ultrasound transducer elements. Non-conductive adhesives (NCA) have been a common solution for the assembly of an ultrasound transducer array on a substrate. The electrical conduction of NCA interconnects is established through direct metal-metal contact between a transducer element and a substrate pad. Thus, bonding of the transducers to the substrate requires simultaneous application of heat and pressure during the adhesive curing. The demand of low curing temperature to /17 $ IEEE DOI /ECTC

2 avoid depolarization of piezoelectric material in the transducers imposes a long curing time for NCA. Assembling of transducer array using NCA therefore occupies bonding equipment for a long time, and hence is a low-throughput process. Au-to-Au thermosonic bonding has recently been demonstrated as another promising approach for lowtemperature bonding of ultrasound transducers to substrates [2, 3]. Reliable, metallurgical interconnects was achieved at moderate temperatures, even down to room temperature, with bonding times in the order of seconds. However, assembling a transducer array with high number of elements demands bonding equipment with extremely high bonding force or high ultrasonic power. This is a huge limitation in terms of cost and process flexibility. Isotropic conductive adhesives (ICAs) are considered in our studies. The materials are composites of a polymeric adhesive filled with a high concentration of electrical conductive particles to ensure electrical conductions in 3D. The mechanical strength of the ICA joint comes from the matrix, while the particles provide electrical conductivity. The assembly of a transducer array on a substrate using ICAs has potential as a low-temperature, high-throughput process thanks to the particular manufacturing approach for ultrasound transducer arrays. The process starts with depositing a sufficient amount of ICA paste on a substrate by means of dispensing or screen/stencil printing. A large transducer stack is then placed on the substrate with ICA covering the entire bonding area. The curing of ICA can happen at a relatively low temperature, such as 100 C; and it does not require a force continuously applied. This allows curing many samples at the same time in a temperature chamber. After that, individual transducer array elements are singulated through a controlled-depth dicing process. With a dicing depth slightly cut into the substrate, no neighboring elements are short-circuited due to ICA. Traditional ICA compounds are loaded with vol% (up to 80 wt%) of solid silver (Ag) particles (typically in flake form) to ensure the required electrical conductivity [4]. Such high metal particle loading causes significant changes in the mechanical properties of the adhesives, including increased bulk modulus and reduced flexibility (a more brittle response). This is believed to be the main source for the poor mechanical performance, particularly poor impact strength, of the conventional ICA materials under demanding conditions [5]. Using conventional ICAs filled with Ag particles in the assembly of ultrasound transducer arrays has thus a high risk of failing during the tough dicing process. A novel type of ICAs based on metal-coated polymer spheres (MPS) have emerged as a promising alternative that satisfies all demands for the assembly process of ultrasound transducer arrays. The novel ICAs are similar to the conventional ICAs, but using MPS as electrical conductive fillers instead of solid metal particles [6, 7]. The coefficient of thermal expansion and the elastic modulus of the MPS are better matched to those of the adhesive matrix, compared to what is possible when loading the adhesive with solid metal particles. Thereby, the local stress induced in the bulk of the material can be reduced. Besides that, MPS in the adhesive matrix can also absorb mechanical energy. Consequently, the ICA resistance to dynamic mechanical loading (typically vibration and shock) can be improved. The MPS-based ICAs are thus believed to have better ductility, and hence higher reliability under mechanically tough conditions. Recent studies have demonstrated that ICAs based on MPS exhibit excellent impact strength [8], improved mechanical die shear performance [9], competitive electrical conductivity [10], as well as a good potential for stencil/screen printing and dispensing processes [11]. These indicate the high potential of using MPS-based ICA in the assembly of ultrasound transducer arrays. In the present work, we address the assembly of an ultrasound transducer array on a substrate using ICAs based on MPS. The process of bonding and subsequently dicing an ultrasound transducer stack on a substrate is evaluated using relevant dummy dies to represent the transducer stacks and flexible substrates mimicking the real substrates in ultrasound applications. The MPS-based ICAs contain a commercial epoxy matrix filled with a high concentration of Ø5 m Au-coated mono-disperse polymer spheres. Two particle concentration values were selected in this work. The lower value is somewhat above the percolation threshold whereas the higher is close to the maximum concentration for processability (such as dispensing, printing). The objective of comparing ICAs with such particle loadings has been to investigate the trade-off between the particle concentration and the performance of ICAs under the assembly of transducer arrays. Using an appropriate particle concentration for ICAs to achieve required manufacturing quality is an important advantage in terms of cost and process flexibility. In this work, a low-temperature, highthroughput bonding process of a dummy die to a substrate was employed. The characterization of manufactured samples was performed by means of electrical resistance, process yield as well as mechanical strength of ICA interconnects. II. EXPERIMENTAL A. Sample description Dummy dies used to represent ultrasound transducer stacks in this work are the actual de-matching layer (DML) component in the transducers. Using DML to represent the transducer is relevant since it is the backing layer for reflection to be bonded to a flex substrate in the real application. The DML used in this work is a material with an extreme hardness and a low electrical resistivity. The DML die has an area about 100 mm 2 with an Au coating on both sides. Flexible substrates are made of polyimide with bond pads populated identically on both sides. The bond pads are composed of Cu-Ni layers with an Au flash on top. The pads are distributed in the form of m-by-n matrix over an area suitable for bonding a DML die on it. The pad pitch is in the range of μm. 2122

3 B. Bonding materials and preparation of ICA pastes The adhesive matrix of the MPS-based ICAs was a commercially available two-component epoxy system. Electrically conductive particles were monodisperse polymer spheres coated with Ni and then Au as the outer shell (AuPS). The diameter of the particles is Ø5 μm. All materials were used as received. ICA pastes were prepared by manually mixing the appropriate amount of AuPS into the epoxy system with hardener. After mixing, the ICA pastes were vacuumed at about 6 kpa in 1.5 hours in order to minimize the presence of air bubbles in the ICA. ICAs filled with two different particle concentrations were evaluated in this work, as discussed in the Introduction. The ICA with higher concentration of particles is denoted ICA-1. The ICA filled with lower particle concentration is denoted ICA-2. C. Assembly process The assembly process of a DML die on a flexible substrate using ICA includes bonding the die to the substrate and subsequently dicing the bonded DML die to form an array of DML elements. The bonding process of a DML die to a flexible substrate was performed using a flip-chip bonder, FINEPLACER pico from Finetech GmbH & Co. KG. Prior to the ICA bonding, bond surfaces of the die and the substrate were cleaned with acetone and isopropanol, and dried with nitrogen gas. A line of ICA paste was then dispensed on the bond surface of the flex substrate. After that, the DML and the flex substrate were aligned while the flipped DML was kept on a bond tool by vacuum. The flipped DML was then pressed onto the flex substrate. Bonding parameters were selected based on an initial characterization study. For all samples bonded, the ICA bonding was performed with a bond pressure of about 3 MPa, and a bond time of 30 seconds. Temperature was not applied during the pick and place process. Test samples were then cured in a temperature chamber at 100 C in 3 hours. Such a low curing temperature was selected to avoid depolarization of the piezoelectric material inside the ultrasound transducers. In addition, the results from a previous study for the same epoxy matrix filled with silver-coated polymer spheres (Ag-PS) have shown a degree of curing of 96% with a curing schedule of 100 C for 2 hours [9]. Hence, the curing condition of 100 C in 3 hours after flip chip bonding is believed to provide sufficient degree of curing for the ICAs, and hence to ensure good electrical and mechanical performances of ICA interconnects. The controlled-depth dicing process was applied for bonded samples. This process is similar to the one applied for dicing ultrasound transducer stacks in real applications. During the dicing process, the continuous DML of each bonded sample was singulated into an array of individual elements. Each of these elements is connected to a corresponding flex pad by ICA forming an interconnect. D. Characterization The characterization of test samples was mainly based on the electrical resistance of ICA interconnects between a DML element and a flex pad. Another important factor was the yield of the assembly process, which is based on the electrical conductivity of interconnects in the samples. In addition, the shear strength of interconnects was tested. 1) Measurement of interconnect resistance and process yield Fig. 2 shows a sketch of a part of a test sample after the assembly process. The design of the flex substrates enables four-wire resistance measurement for every pair of two adjacent interconnects having bond pads connected via a track on the substrate. The four-wire resistance measurement for a pair of interconnects was carried out by means of two tips from a probe card contacting the back of each DML element and a Keithley 3706A Digital Multi-meter. The resistance of an individual interconnect was determined by halving the measured value between two adjacent interconnects with resistance of the track on substrate being deducted. An ICA interconnect was considered acceptable for ultrasound applications when its resistance was less than 5. The process yield was determined by dividing the number of interconnects with sufficient conductance by the total interconnects of all samples. As the highest resolution of probing is two interconnects, both interconnects were claimed to be open if an open pair was found. Thus, the value given for process yield is a conservative value. Figure 2. Setup for the electrical resistance measurement of interconnects after the assembly process. Each DML element is contacted on the back by two tips from a probe card. This enables four-wire resistance measurement for every pair of two adjacent interconnects having flex pads connected via a track on substrate. This is a sketch of a part of a test sample and is not to scale. 2) Die shear testing Shear strength of ICA interconnects between a DML element and a flex pad was evaluated using a shear tester F&K Delvotec Due to the limits of force resolution of the shear tester and the dimensions of shear tools available, the shear tests were performed for groups of 5 interconnects in a row with an assumption that the shear force is evenly distributed to all interconnects in the same group. Prior to the destructive die shear tests, the backside of the flex substrate was glued to a glass substrate with a commercial epoxy. Following the application of the epoxy, the sample was 2123

4 placed in an oven at 150 C in 30 minutes to cure the epoxy. The substrate was then kept fixed on the substrate holder of the shear tester. Several rows of DML elements were removed to make room for the shear tool to access the ICA interconnects of interest. In this experiment, only interconnects at the peripheral areas of the sample were tested. 3) Visual inspection All test samples were visually inspected after dicing in order to evaluate the dicing process in terms of dicing yield and failure modes (if any) that might happen during the dicing. The dicing yield was determined by dividing the number of DML elements remained intact on the corresponding flex pads in the bonding area by the total interconnects (DML-flex pad) of all samples. In addition, cross-sectioning was performed for selected samples to inspect the bond-line of ICA interconnects as well as the separation between neighboring interconnects after dicing. III. RESULTS Eleven ICA-1 and four ICA-2 samples underwent the entire assembly process; bonding and subsequently dicing the DML. ICA-1 and ICA-2 are ICAs filled with higher and lower concentration of AuPS, respectively. Fig. 3 shows an example of test samples after the assembly process. A. Dicing yield, interconnect resistance and process yield The yield of the dicing process was calculated using data from all 11 ICA-1 samples and all 4 ICA-2 samples. The dicing yield of ICA-1 and ICA-2 samples are comparable; 99.7 % and 99 %, respectively. The resistance measurement results of ICA-1 test samples are shown in Fig. 4. The resistance of ICA-1 interconnects is mainly in the range of with an average value of The yield of the entire assembly process is 99.2 %. Fig. 5 shows the resistance distribution within a typical ICA-1 sample. In general, interconnects with lower resistance were found at the peripheral areas of test samples, particularly the top and bottom edge (see Fig. 5). Interconnects close to the central area of manufactured samples exhibited slightly higher electrical resistance. Interconnects with high resistance, including failed interconnects, were found randomly in the bonding area of test samples. Fig. 6 shows the results of ICA-2 test samples. The resistance of ICA-2 interconnects is mainly in the range of The average resistance and the process yield are about 0.35 and 98 %, respectively. Fig. 7 shows the resistance distribution within a typical ICA-2 sample. In general, the interconnect resistance is relatively uniform in most of the bonding area. Interconnects with higher resistance, including failed interconnects, were mainly found at the bonding regions close to the edges of the samples. Figure 3. A typical sample after the assembly process. The sample shown is a ICA-1 sample. Figure 4. Histogram for measured resistance of individual DML elementflex pad interconnects of all 11 ICA-1 samples. Figure 5. Distribution of interconnect resistance of a typical ICA-1 sample. The color scale bar varies from 0 to 5. Figure 6. Histogram for measured resistance of individual DML elementflex pad interconnects of all 4 ICA-2 samples. 2124

5 Figure 7. Distribution of interconnect resistance of a typical ICA-2 sample. The color scale bar varies from 0 to 5. Figure 8. Typical fracture surface of DML element-flex pad interconnects after a shear testing trial. Similar failure modes were observed for both ICA-1 and ICA-2 interconnects. The sample shown is a ICA-1 sample. B. Shear strength Destructive die shear tests were conducted for 3 ICA-1 and 2 ICA-2 samples. Four shear testing trials were performed for each sample. For each trial, a group of 5 DML elements were tested. The shear strength of individual interconnects between a DML element and a flex pad for ICA-1 and ICA-2 is in the same range, as shown in Table I. The visual inspection of fracture surfaces after the shear tests showed similar failure modes for ICA-1 and ICA-2 interconnects; predominantly adhesive failure at the DML/ICA and flex pad/ica interfaces. Fig 8 shows a typical fracture surface of interconnects after shear testing. C. Cross-section Cross-sectioning and visual inspection of one ICA-1 and one ICA-2 sample (after dicing) were performed along the short edges close to the central area of the sample, as illustrated in Fig. 9. Typical cross-sectional images of ICA-1 and ICA-2 interconnects are shown in Fig 10 and Fig. 11, respectively. The bond-line thickness was found in the range of μm for ICA-1 and μm for ICA-2. The cross-sectional images also show that the dicing process was well depth-controlled, with DML and ICA being cut through. This ensures a complete separation between neighboring ICA interconnects consisting of a DML element and a flex pad. TABLE I. RESULTS FROM THE SHEAR TESTS FOR ICA-1 AND ICA-2 SAMPLES Sample Shear strength* Standard deviation ICA-1 29 MPa 2 MPa ICA-2 27 MPa 4 MPa * Average of all shear testing trials Figure 9. Cross-section of a typical sample. The red lines show the locations where the sample was cut. The arrows show the directions where the cross-sectioned pieces were looked into. Figure 10. Optical micrograph of an ICA-1 interconnect in a crosssectioned sample Figure 11. Optical micrograph of an ICA-2 interconnect in a crosssectioned sample 2125

6 IV. DISCUSSION Table II compares the experimental results from the ICA with higher particle loading (ICA-1) and the ICA with lower particle loading (ICA-2). Compared to ICA-2, ICA-1 exhibits better process yield, somewhat higher electrical resistance, and comparable mechanical shear strength. TABLE II. SUMMARY OF EXPERIMENTAL RESULTS FROM THE ICA WITH PARTICLE CONCENTRATION CLOSE TO THE MAXIMUM LIMIT FOR PROCESSABILITY (ICA-1) AND THE ICA WITH PARTICLE CONCENTRATION SOMEWHAT HIGHER THAN THE PERCOLATION THRESHOLD (ICA-2) Samples ICA-1 ICA-2 Dicing yield* 99.7 % 99 % Process yield* 99.2 % 98 % Resistance* Shear strength** 29 (± 2) MPa 27 (± 4) MPa Bond-line thickness μm μm * Average value from all test samples ** Average value from all shear testing trials A. Mechanical performance of ICA interconnects The high dicing yields of ICA-1 and ICA-2 samples indicate good mechanical strength of ICA interconnects between a DML element and a flex pad. The observation is also confirmed from the shear testing results, showing that the shear strength of individual ICA-1 and ICA-2 interconnects is in the same range as that of the conventional ICAs filled with Ag flakes and the traditional solders (60Sn- 40Pb and 63Sn-37Pb) [9, 12]. Both ICA-1 and ICA-2 exhibited comparable shear strength and similar failure modes after the shear testing trials. The comparable shear strength between ICA-1 and ICA-2 with different particle concentration agrees with the results from a previous study in which the die shear strength of an epoxy filled with uncoated monodisperse polymer spheres is stable with particle loading up to 45 vol% and slightly decreases for higher particle loading up to 55 vol% [13]. The fracture mechanism of ICA interconnects after shear tests (see Fig. 8) reveals the weakest locations in an interconnect: the adhesion of ICA to the Au surfaces of DML elements and flex pads. If the mechanical performance (e.g. shear strength) of ICA interconnects needs to be improved, the adhesion between ICA and Au surfaces of bonding components should be addressed, for example by using adhesion promoters for Au surfaces. B. Interconnect resistance and process yield The electrical resistance of ICA interconnects presented in this work includes the interconnect resistance as well as the bulk resistance of DML elements. Since the electrical resistivity of the DML material is low (on the order of 10-7 m), the contribution of the bulk resistance of the DML elements in the measured resistance is negligible. As shown in Fig. 4, the resistance of ICA-1 interconnects is mainly in the range of with an average value of For ICA-2, the interconnect resistance is mainly in the range of and the average resistance is about 0.35 (see Fig. 6). The interconnect resistance in this range is acceptable for electrical connections between a transducer element and a pad on a system substrate in ultrasound applications. In addition, the electrical resistance of AuPS-based ICA interconnects in this work agrees with results from a previous study in which a Si-based chip was bonded to a printed circuit board using ICAs based on MPS [8]. The AuPS-based ICAs exhibit somewhat higher electrical resistance compared to other adhesive-based interconnection technologies, such as non-conductive adhesives (NCA) and anisotropic conductive adhesives (ACA), that have been used for bonding similar structures (resistance normally less than 0.2 ) [14, 15]. Such electrical resistance of ICAs based on AuPS is attributed to two reasons; 1) the thicker bond-line of ICA interconnects (10-20 μm vs. less than 3 μm for ACA/NCA interconnects); and 2) the hardness of the Ni coating layer between the polymer core and the outer Au coating of the AuPS. While the bondline thickness of ICA interconnects can be reduced by increasing bonding pressure, the effect of the Ni coating is discussed. Ni is relatively hard compared to other metals used as coating layers for polymer spheres such as Au, Ag, Cu [16]. In ICA interconnects, the particle-particle contacts and the particle-dml/flex pad contacts are maintained by the shrinkage of the adhesive matrix after curing. Therefore, hard Ni coating causes less deformation of the particles, and hence less contact areas between neighboring particles as well as between the particles and the bonding surfaces of DML and flex pads. In order to improve the electrical resistance of MPS-based ICAs, particles with softer metal coating such as Ag should be used. ICA-1 samples exhibited somewhat higher interconnect resistance compared to ICA-2 samples even though the concentration of AuPS in ICA-1 is higher than that of ICA-2 (see Table II). It should be noted that the electrical resistivity, and hence resistance of interconnects with the same dimensions, of an MPS-based ICA with particle concentration close to the higher limit for processability is considerably lower than that of an ICA with MPS loading close to the percolation threshold [10]. Therefore, the bondline thickness of ICA-1 and ICA-2 samples is probably the main reason for the higher electrical resistance of ICA-1 interconnects compared to ICA-2 interconnects. With a 1.4 times thinner bond-line, ICA-2 interconnects showed about 1.6 times lower electrical resistance. The thinner bond-line of ICA-2 samples, compared to ICA-1 samples, is attributed to the viscosity of the ICA pastes during bonding. It has been known that filling an adhesive matrix with increasing concentration of polymer spheres results in increasing viscosity of the adhesive paste [11]. With the same bonding conditions, particularly the same bonding pressure, ICA-2 with lower particle concentration, hence lower viscosity, is thus squeezed out more than ICA-1 with higher particle concentration. This results in a thinner layer, and hence thinner bond-line, of ICA-2 paste remained between a DML die and a flex substrate after bonding. As shown in Fig. 4 and Fig. 6, the interconnect resistance of ICA-1 samples spreads more than that of ICA-2 samples. Such higher spreading of ICA-1 interconnect resistance is 2126

7 probably due to the thicker bond-line, compared to ICA-2 interconnects. The bond-line of ICA-1 samples is about 50 % thicker than that of ICA-2 samples (see Table II). With the same x-y dimensions of bonded DML dies before dicing, the volume of ICA paste between a DML die and a flexible substrate of ICA-1 samples is about 50 % larger than that of ICA-2 samples. Such a larger volume between the die and the substrate in ICA-1 samples provides more possibilities for AuPS particles to arrange in the paste, compared to ICA- 2 samples. In the same manner, particles in individual ICA-1 interconnects also have more possibilities to arrange in the paste thanks to the larger volume, compared to ICA-2 interconnects. It should be noted that the electrical resistance of ICA interconnects is considerably influenced by the number of conductive paths between the die and the substrate [10]. Thus ICA-1 samples with more possibilities for particles arrangement in individual interconnects probably have more spreading in the number of conductive paths between a DML element and a flex pad. This explains the more spreading in the interconnect resistance of ICA-1 samples, compared to ICA-2 samples. A thinner bond-line seems to be better in terms of the resistance value and the uniformity of interconnect resistance within a transducer array. Failures in ICA interconnects could happen during the bonding and/or the dicing process of DML dies in test samples. Since the DML dies and the ICAs after curing are electrically conducted in 3D, all interconnects in the samples were short-circuited before dicing. It was thus not possible to determine the bonding yield prior to the dicing process. Therefore, the process yield presented in this work is a combination of the bonding yield and the dicing yield. The process yield of ICA-1 samples is better than that of ICA-2 samples (see Table II). The reasons are probably related to the particle concentration of the ICAs, and that process yield is determined based on electrical resistance of ICA interconnects. ICA-2 with particle concentration somewhat above the percolation threshold probably have random sites with insufficient amount of particles. This leads to a higher measured resistance for interconnects corresponding to those sites, and hence a higher number of interconnects being considered as open circuit. This explains why ICA-2 has lower yield than ICA-1 even though both ICAs exhibited comparable mechanical shear strength and similar failure modes during shear testing. The trade-off between the particle concentration and the performance of ICAs based on AuPS under the assembly of ultrasound transducer arrays on flexible substrates are seen from the results in this work. Using ICA with the concentration of AuPS close to the higher limit for processability provides a better process yield. The disadvantages of using high particle concentration are cost (using more particles) and limited process flexibility in terms of approaches, parameters. Using ICA with lower AuPS concentration has advantages in terms of cost (using less particles) and process flexibility. However, the process yield may be reduced. V. CONCLUSIONS A low-temperature, high-throughput assembly process of an ultrasound transducer array on a flexible substrate using isotropic conductive adhesives (ICAs) based on Au-coated polymer spheres (AuPS) was studied. The process of bonding and subsequently dicing a transducer stack on a system substrate was evaluated using relevant dummy dies (DML) to represent the transducers and flexible substrates mimicking the real substrates in ultrasound applications. The entire assembly process has a maximum temperature of about 100 C. The time that a bonding equipment being occupied during bonding is less than a minute. The ICAs contain a commercial epoxy matrix filled with Ø5 m AuPS. The trade-off between the particle concentration and the performance of ICAs under the assembly of transducer arrays was investigated using ICAs with two different concentration values, the lower being somewhat above the percolation threshold whereas the higher being close to the maximum concentration for processability (such as dispensing, printing). The results showed good integrity of AuPS-based ICA interconnects, composed of a DML element and a flex pad, in test samples undergoing the entire assembly process. For the ICA with particle concentration close to the maximum limit for processability, high process yield (99.2 %), acceptable electrical resistance (average value ~ 0.59 ) and high mechanical shear strength (~ 29 MPa) were obtained. For the ICA with particle concentration somewhat above the percolation threshold, a process yield of 98 %, average resistance of about 0.35 and high mechanical shear strength of about 27 MPa were obtained. The trade-off between the particle concentration and the performance of ICAs based on AuPS under the assembly of ultrasound transducer arrays on flexible substrates are observed. Using ICAs with the concentration of AuPS close to the maximum limit for processability provides a better process yield. The disadvantages of using high particle concentration are cost (using more particles) and limited process flexibility. Using ICAs with AuPS concentration somewhat above the percolation threshold has advantages in terms of cost (using less particles) and process flexibility. However, the process yield may be reduced. This work has demonstrated the feasibility of using MPSbased ICAs as a low-temperature, high-throughput solution for assembling ultrasound transducer arrays on system substrates. ACKNOWLEDGMENT The present work was funded by the Research Council of Norway through the NANO2021 program (project number /O70; B-EAM Advanced Assembly Technologies for Electro Acoustic Module used in Ultrasound Cardiovascular Applications). The authors gratefully acknowledge Trym Eggen (GE Vingmed Ultrasound AS, Norway) and Tung Manh (University College of Southeast Norway) for their support in this work. 2127

8 REFERENCES [1] R. S. C. Cobbold, Foundations of Biomedical Ultrasound. Oxford, United Kingdom: Oxford University Press, [2] K. E. Aasmundtveit, T. T. Luu, T. Eggen, C. E. Baumgartner, N. Hoivik, K. Wang, et al., "Thermosonic Bonding for Ultrasound Transducers: Low-temperature Metallurgical Bonding," in The 4 th Electronics System Integration Technology Conferences, Amsterdam, The Netherlands, 2012, pp [3] J. H.-G. Ng, M. P. Y. Desmulliez, K. E. Aasmundtveit, H.-V. Nguyen, R. Ssekitoleko, Z. Qiu, et al., "Low Temperature Bonding of Piezoelectric Single Crystal Materials for Miniaturized High Resolution Ultrasound Transducers," in The 4 th Electronics System Integration Technology Conferences, Amsterdam, The Netherlands, 2012, pp [4] J. E. Morris and J. Liu, "Electrically Conductive Adhesives: A Research Status Review," in Micro- and Opto-Electronic Materials and Structures: Physics, Mechanics, Design, Reliability, Packaging. vol. 2, E. Suhir, Y. C. Lee, and C. P. Wong, Eds., ed U.S.A.: Springer US, 2007, pp. B527-B570. [5] J. E. Morris, "Isotropic Conductive Adhesives: Future Trends, Possibilities and Risks," Microelectronics Reliability, vol. 47, pp , [6] H. Kristiansen, K. Redford, and T. Helland, "Isotropic Conductive Adhesive," United States Patent Application No. US , [7] H. Kristiansen, K. Redford, and T. Helland, "Isotropic Conductive Adhesive," United Kingdom Patent GB PCT- EP , [8] J. Gakkestad, P. Dalsjo, H. Kristiansen, R. Johannessen, and M. M. V. Taklo, "Use of Conductive Adhesive for MEMS Interconnection in Ammunition Fuze Applications," Journal of Micro/Nanolithography, MEMS and MOEMS, vol. 9, p , [9] H.-V. Nguyen, E. Andreassen, H. Kristiansen, and K. E. Aasmundtveit, "Die Shear Testing of a Novel Isotropic Conductive Adhesive Epoxy Filled with Metal-Coated Polymer Spheres," IEEE Transactions on Components, Packaging and Manufacturing Technology, vol. 3, pp , [10] S. Jain, D. C. Whalley, M. Cottrill, H. Kristiansen, K. Redford, C. B. Nilsen, et al., "Electrical Properties of an Isotropic Conductive Adhesive Filled with Silver Coated Polymer Spheres," in The 18 th European Microelectronics and Packaging Conference, Brighton, United Kingdom, 2011, pp [11] H.-V. Nguyen, E. Andreassen, H. Kristiansen, R. Johannessen, N. Hoivik, and K. E. Aasmundtveit, "Rheological Characterization of A Novel Isotropic Conductive Adhesive Epoxy Filled With Metal-coated Polymer Spheres," Materials and Design, vol. 46, pp , [12] S. K. Kang, "Development of Lead (Pb)-free Interconnection Materials for Microelectronics," Metals and Materials International, vol. 5, no. 6, pp , [13] H.-V. Nguyen, H. Kristiansen, J. Gakkestad, R. Johannessen, N. Hoivik, and K. E. Aasmundtveit, "Spherical Polymer Particles in Isotropic Conductive Adhesives - A Study on Rheology and Mechanical Aspects," in The 3 rd Electronics System Integration Technology Conferences, Berlin, Germany, 2010, pp [14] H. Dong, Y. Li, M. J. Yim, K. S. Moon, and C. P. Wong, "Investigation of electrical contact resistance for nonconductive film functionalized with -conjugated self-assembled molecules," Applied Physics Letters, vol. 90, pp , [15] H.-V. Nguyen, T. Eggen, and K. E. Aasmundtveit, "Assembly of Transducer Array Using Anisotropic Conductive Film for Medical Imaging Applications," presented at the The 20 th European Microelectronics and Packaging Conference (EMPC 2015), Friedrichshafen, Germany, [16] J. William D. Callister, "Appendix B: Properties of Selected Engineering Materials," in Fundamentals of Materials Science and Engineering, 5 th ed U.S.A.: John Wiley & Sons, Inc., 2001, pp