Growth of Surface Oxide Layers on Dendritic Cu Particles by Wet Treatment and Enhancement of Sinter-Bondability by Using Cu Paste Containing the Particles

Abstract

Pastes were prepared using dendritic Cu particles as fillers, and a compression die attachment process was implemented to establish a pure Cu joint using low-cost materials and high-speed sinter bonding. We aimed to grow an oxidation layer on the particle surface to improve sinter-bondability. Because the growth of the oxidation layer by general thermal oxidation methods makes it difficult to use as a filler owing to agglomeration between particles, we induced oxidation growth by wet surface treatment. Consequently, when the oxidation layer was appropriately grown by surface treatment using an acetic acid–ethanol solution, we obtained an improved joint strength, approximately 2.8 times higher than the existing excellent result based on a bonding time of 10 s. The joint formed in just 10 s at 300 °C in the air under 10 MPa compression showed a shear strength of 28.4 MPa. When the bonding time was increased to 60 s, the joint exhibited a higher strength (35.1 MPa) and a very dense microstructure without voids. These results were attributed to the acceleration of sintering by the in situ formation of more Cu nanoparticles, which effectively reduced the increased oxide layers in the particles using a reducing solvent.

1. Introduction

Wide-bandgap semiconductors, such as SiC, replace traditional Si power semiconductors owing to their advantages such as high power output due to high current capability, fast switching speed, and low energy conversion loss [1,2,3,4,5]. SiC power semiconductors may operate in high-temperature environments above 200 °C, so when using a SiC chip, the die-attach material under the chip should be replaced from the current solder to a metal with thermal stability [6,7,8,9,10,11,12]. In addition, the formed joint should exhibit excellent thermal conductivity, which is desirable for effectively dissipating the heat generated from the chip. Therefore, sinter-bonding processes using pastes or film materials containing Ag particles have been developed and used in the industry [13,14,15,16,17,18,19,20].

Recent sinter-bonding research has focused on low-cost Cu-particle-containing pastes [21,22,23,24,25,26,27,28]. Unlike Ag, Cu is easily oxidized, which hinders sintering; thus, research results introducing special oxidation suppression measures, such as sintering in a reducing atmosphere with formic acid or introducing an effective reducing agent into the paste, have been reported [29,30,31,32]. However, reducing atmospheres have poor industrial applicability because they cause environmental pollution, and the introduction of a reducing agent cannot completely prevent the oxidation of Cu during long-term sintering. The authors reported that excellent compression sinter-bonding characteristics could be achieved even in air by fabricating a paste based on modified dendrite-shaped Cu particles and an effective reducing agent [33,34]. In other words, dendritic particles undergo bending owing to compression, filling the gaps between particles, and the oxidation layers on the enhanced surface area in situ generate nanoparticles through a reaction with an effective reducing agent, leading to rapid sintering between particles [34]. Consequently, factors such as the degree of miniaturization of the dendrites and the compression pressure influence the required bonding time. However, a maximum shear strength of more than 28 MPa was rapidly obtained at 60 s of bonding, showing significantly superior high-speed sinter-bondability compared to previous research results [34]. In addition, the rapid increase in joint density due to rapid sintering minimizes the degree of oxidation in the formed joint, even during bonding in air [33,34].

In this study, considering the sintering behavior of dendritic Cu particles, we aim to improve sinter-bondability through surface treatment. Considering that Cu nanoparticles are generated from the surface oxidation layer of dendritic Cu particles, an increase in the thickness of the surface oxidation layer can improve the sinter-bondability by generating more Cu nanoparticles. However, the thermal oxidation of dendritic Cu particles in air causes particle agglomeration and growth of the oxidation layer; therefore, this process is unsuitable for dendritic Cu particles that need to be mixed into the paste. Therefore, this study aims to evaluate the sinter-bondability of a paste fabricated from dendritic Cu particles with grown oxidation layers by implementing an oxidation layer growth process on the surface of dendritic Cu particles via wet surface treatment.

2. Materials and Methods

Cu dendrite particles with a D50 of 3.78 µm were used, which were manufactured in-house using a previously reported synthesis method [34]. After synthesis, the particles were washed with deionized water and dried in a vacuum chamber for 10 h. Meanwhile, 100 mL of acetic acid (CH₃COOH, 99%, Duksan Pure Chemicals Co., Ltd., Ansan, Republic of Korea)-ethanol solution at 87.4 mM and 175 mM was prepared, respectively, heated to 40 °C on a hot plate, and 1 g of Cu dendrite particles was loaded into the solution while stirring at 200 rpm using a magnetic bar for 1 min. Subsequently, sonification was performed for 4 or 9 min to result in a total of 5 or 10 min of surface treatment, respectively. Finally, the surface-treated Cu dendrite particles were obtained by repeated washing (using ethanol), centrifugation, and drying in a vacuum chamber at 5 × 10⁻² torr for 10 h.

The shape change of the dendritic Cu particles due to the surface treatment was observed under an acceleration voltage of 10 kV using a high-resolution field-emission scanning electron microscope (SEM, SU8010, Hitachi High-Technologies Corp., Tokyo, Japan). The phase analysis of the particles was conducted using X-ray diffraction (XRD, DE/D8 Advance, Bruker, Billerica, MA, USA), where measurements were made at a speed of 0.2 deg./s using a Cu Kα source in the 2θ range of 10−90°. A precise analysis of the surface change in the particles by the surface treatment was conducted through X-ray photoelectron spectroscopy (XPS, K-Alpha+, Thermo Fisher Scientific, Waltham, MA, USA) measurement using an Al Kα source.

The dried surface-treated particles were then mixed with a reducing solvent (EW-10, Epsilon Epowder, Seongnam, Republic of Korea) at a weight ratio of 71:29 to produce a paste [34]. The heating characteristics of the manufactured paste were identified through thermo gravimetry-differential thermal analysis (TG-DTA, DTG-60, Shimadzu, Kyoto, Japan) analysis, where analysis was conducted in a dynamic heating mode that continuously heated to 400 °C at a heating rate of 10 °C/min in air.

The paste was printed on a dummy Cu substrate using a stencil mask with a slit volume of 3 × 3 × 0.1 mm³, and a dummy Cu chip with dimensions of 3 × 3 × 1 mm³ was aligned and placed on a printed pattern. Subsequently, the printed pattern in the sandwich structure was dried for 30 s on a hot plate at 150 °C. Die-attach was carried out in the air by heating to 300 °C as soon as the chip upper center was compressed with a 14 mm diameter collet at 10 MPa and was carried out for 10−60 s, including approximately 10 s of heating time to 300 °C. Therefore, a bonding time of 10 s means a process that releases pressure and cools as soon as the joint reaches 300 °C, and a bonding time of 30 s means a process that is maintained for 20 s after the joint reaches 300 °C. A graphic scheme of the compressive die-attach process is shown in Figure 1. The rapid density increase behavior in the joint was observed using a spherical aberration-corrected scanning transmission electron microscope (Cs-STEM, NEO-ARM, JEOL Ltd., Tokyo, Japan) at an acceleration voltage of 200 kV, and fast Fourier transform (FFT) analysis was also conducted.

The robustness of the joint formed in the sandwich-structured sample was evaluated through a shear test using a multipurpose bond tester (HAWK 8200 S, KOVIS, Seoul, Republic of Korea). The shear strength of a joint was determined as a 5 samples-average value of the maximum stress value measured during the shear testing at a speed of 200 μm/s with a shear-tip height of 200 μm above the substrate surface. Changes in the microstructure of the joint according to the bonding time and fracture structure after the shear test were observed using SEM. An intuitive comparison of the degree of oxidation in the bonding area and confirmation of the fracture location on the fracture surface were made using optical microscopy.

3. Results and Discussion

Figure 2 shows the morphological changes in the dendritic Cu particles according to the surface-treatment conditions compared to pristine dendritic Cu particles. Unlike typical dendrites, pristine dendritic Cu particles (Figure 2a) show short branches of several hundred nanometers, with Cu nanoparticles exhibiting a directional configuration and forming tangled branches on the central stem [33,34]. In contrast, the surface roughened sequentially for the surface-treated Cu particles and showed a trend of further oxidation in the order of Figure 2b−e. In particular, considering the specimen shown in Figure 2e, which underwent surface treatment for a relatively long time (10 min) with a high-concentration surface-treatment solution of 175 mM, the formation of rough-feeling nodules on the particle surface was observed, and it was inferred that the largest surface oxidation layer growth occurred.

The XRD measurement results for the pristine Cu dendrites and Cu dendrites with different surface-treatment conditions are summarized in Figure 3. Cu oxides, such as Cu₂O, were detected even in the pristine Cu dendrites, but their amounts were relatively small. In contrast, the Cu₂O phase in the surface-treated Cu dendrites increased as the concentration of the surface-treatment solution and the surface-treatment time increased. This confirms that the growth of surface oxides is possible by wet surface treatment, and that their amount can be controlled depending on the processing conditions.

XPS analysis was conducted on the surface-treated powders to better understand the changes in the surface oxidation layer due to the surface treatment. Figure 4 shows the Cu 2P XPS spectra of pristine dendritic Cu particles and dendritic Cu particles surface-treated under various conditions. In the case of pristine dendritic Cu particles (Figure 4a), CuO and Cu₂O phases were detected mainly because of the unavoidable oxidation occurring on the surface of the Cu particles after synthesis and drying. In contrast, for the surface-treated Cu particles, Cu₂O peaks were observed to grow sequentially, as shown in Figure 4b–e, indicating that oxidation on the particle surface gradually progressed as the concentration of the surface-treatment solution and the surface-treatment time increases. In particular, for the sample shown in Figure 4e, which was exposed to a harsh oxidation environment, the largest Cu₂O peaks were detected, where no CuO peaks were detected. Such harsh surface-treatment conditions resulted in the transition of all the CuO phases to the equilibrium phase of Cu₂O via a chemical reaction.

The TG-DTA results for the five paste types are shown in Figure 5. Based on the TG results, all pastes showed a significant weight reduction behavior owing to the evaporation of the reducing solvent at 183−193 °C. As the surface became rough, the capillary effect caused the reducing solvent to be retained for a longer period, resulting in delayed evaporation. Also, an apparent weight increase during oxidation was observed at 283−290 °C, and this weight increase can be minimized via a rapid density increase during actual compression sinter-bonding. Conversely, considering the DTA results, the start of the first exothermic peak at 224−236 °C and the start of the second exothermic peak at 283−290 °C were observed, which results form a combination of the exothermic behavior by sintering between particles and the oxidation of particles. The area values (ignoring units) from Figure 5a–e were measured to be 100, 138.3, 159.8, 193.9, and 134.8, respectively (note that Figure 5d had the largest value). Therefore, assuming that the exothermic quantities of oxidation were similar, it was indirectly inferred that the degree of sintering was the largest in the specimen shown in Figure 5d.

The results of sinter bonding using five types of Cu pastes are shown in Figure 6. Figure 6 indicates the change in shear strength values of joints sinter-bonded by heating to 300 °C in the air under 10 MPa compressions. Generally, the shear strength of all the Cu pastes increased as the bonding time increased up to 60 s. However, they clearly showed different strength results depending on whether surface treatment was performed or the conditions. Therefore, pastes containing particles surface-treated with 87.4 mM acetic acid–ethanol solution showed better strength values at the same time of bonding compared to pastes containing pristine Cu particles. Furthermore, the paste containing particles surface-treated with 175 mM acetic acid–ethanol solution for 5 min showed the best values (even after a very short bonding time of 10 s, a sufficient value of as much as 28.4 MPa was shown, and a strength value of up to 35.1 MPa was shown at 60 s bonding). However, the paste containing particles surface-treated with 175 mM acetic acid–ethanol solution for 10 min showed lower values than those of the standard paste containing pristine particles at 10–30 s bonding, which was interpreted as being due to hindered sintering because the surface oxidation layer formed by excessive oxidation could not be entirely reduced. In addition, in particles surface-treated with 175 mM acetic acid–ethanol solution for 10 min, no CuO phase was observed. However, considering that the reducing solvent used in this study was a polyol component, the absence of CuO phases also could not induce effective reduction of the Cu₂O phases [35].

To evaluate the density of the formed joints and the change in density according to the pretreatment conditions, the joints of the specimens sintered for 60 s were polished and their cross-sections were observed. Figure 7 shows the microstructures of the cross-sections of the joints formed by heating to 300 °C in the air under 10 MPa for 60 s. The degree of density, which was of the most interest, was proportional to the strength, as shown in Figure 6, with the densest structure observed in Figure 7c. In contrast, Figure 7d shows a relatively porous structure. This is the result of a decrease in sinterability; that is, sintering between particles was suppressed owing to the incomplete reduction of the excessively grown oxide layer on the surface of the Cu particles.

The microstructure observed in Figure 7c is very dense compared to the joint formed by the paste using spherical micro-sized Cu particles, primarily because the bending behavior of the dendrites effectively fills the gaps between particles [33,34]. Despite the same densification behavior, high-magnification TEM images were obtained to determine the cause of the significantly higher density observed compared to that achieved when using Cu dendrites without surface treatment. Figure 8a–c shows the results of observing the joint of a 60 s bonding sample using a paste containing particles surface-treated for 5 min with a 175 mM acetic acid–ethanol solution, which is shown in Figure 7c, with TEM. Numerous nanoparticles were generated and clumped in the gaps between the bent dendrite particles, and these particle phases were identified as Cu by mapping using energy dispersive X-ray spectrometry (EDS) (Figure 8d) and FFT analysis (Figure 8e). Therefore, both the increase in strength (Figure 6) and density (Figure 7) were attributed to the generation of a large number of Cu nanoparticles by reducing the sufficiently grown oxidation layers.

Figure 9a–d shows images of the fracture surfaces observed by optical microscopy and SEM, respectively, after the shear test of the samples bonded for 60 s using pastes with the four types of surface-treated Cu dendrites. In Figure 9a–c, all specimens showed a cohesive failure mode (Figure 9e); however, in Figure 9d, a fracture occurred mainly at the interface between the chip and the joint, showing an interface failure mode (Figure 9f). This change in the failure mode matches well with the result of the low-strength value shown in Figure 6. In addition, because sinter bonding was performed for 60 s in air at high temperatures, severe oxidation occurred around the joint, which was observed as black areas by optical microscopy. In Figure 9a–c, a Cu-colored fracture is observed in most areas, including the center of the joint. However, in Figure 9d, a somewhat black color is observed at the center of the fracture, which indirectly indicates that oxidation of the Cu particles occurred in the joint. In addition, in the SEM microstructure observation results presented below the optical image in Figure 9a, a fracture microstructure in which the particle shape remained because sintering was insufficient was observed, whereas in Figure 9b, a slightly porous elongated fracture microstructure without a particle shape was observed. Figure 9c shows an elongated fracture microstructure without pores. The XRD measurement results (Figure 10) of the center of the fracture in Figure 9c shows that there was almost no formation of Cu oxide in the formed joint. However, as shown in Figure 9d, a relatively rough fracture microstructure with slight elongation and remaining fracture fragments was observed, and a Cu oxide phase was found on the fracture. The XRD measurement results at the center of the fracture (Figure 10) indicated the presence of a residual Cu₂O phase. Therefore, the specimen in Figure 9d did not quickly densify like the bulk material because sintering did not proceed effectively, and the excessive growth of the oxidation layers on the surface of the Cu particles did not induce complete oxide removal. Hence, the specimen shown in Figure 9d was severely oxidized, and the most insufficiently sintered and oxidized joint was observed among all samples.

The edges of the fracture surfaces on the dummy Cu chips shown in Figure 9 were oxidized by heating in air, which was clearly observed after 60 s of bonding. Therefore, bonding times exceeding 60 s in air are not desirable because the expansion of the oxidized regions at the edges owing to the increased bonding time can lead to a decrease in the shear strength. To minimize the edge oxidation, bonding should be completed in a very short time or under a nitrogen atmosphere.

4. Conclusions

To improve the sinter-bondability of the paste using dendritic Cu particles as the filler, we induced the growth of an oxidation layer on the dendritic Cu particles via wet surface treatment. As the concentration of the surface-treatment solution and the surface-treatment time increased, growth of the Cu₂O oxidation layer was observed; however, in the case of particles surface-treated with 175 mM acetic acid–ethanol solution for 10 min, excessive oxidation layer growth and the disappearance of the CuO phase were confirmed through XPS analysis. In the case of the Cu particles, surface-treated under optimal conditions (175 mM acetic acid–ethanol solution for 10 min), there was a remarkable improvement in the sinter-bondability of the paste and the largest sintering exothermic amount in the TGA results. As a result, the joint formed in just 10 s at 300 °C in air showed an excellent shear strength value of 28.4 MPa, and when the bonding time increased to 60 s, the strength value increased to 35.1 MPa. This improved sinter-bondability was attributed to the formation of more Cu nanoparticles by reducing the oxidation layers grown by the reducing solvent used. Therefore, the joint manufactured under optimal surface-treatment conditions exhibited a very dense microstructure without voids. After the shear test, the fracture surface exhibited an excellent structure with an elongated metallic surface without excessive oxidation, and a desirable cohesive failure mode was observed.