Thermo-Compression Sinter-Bonding in Air Using Cu Formate/Cu Particles Mixed During Reduction of Cu2O
Article information
Abstract
A Cu-based paste containing Cu formate and Cu particles was prepared for the compressionassisted sinter-bonding of Cu-finished wide-bandgap power devices onto a Cu-finished substrate at a relatively low bonding temperature of 250 °C in air. A mixture of Cu formate and Cu particles was designed to mitigate the tremendous volume shrinkage during reduction of Cu formate, which approaches approximately 90%, and could be a significant obstacle in the formation of a high-density bond-line. The mixture was spontaneously formed during the 15-min reduction of the initial Cu2O particles by a simple wet process using formic acid. In the bonding, pure Cu generated in situ from the Cu formate at a temperature exceeding 200 °C exhibited significant sinterability, and the generated hydrogen reduced oxide layers on the Cu finishes. Furthermore, the mixed particles resulted in low volume shrinkage in the bond-line during bonding, compared to the use of Cu formate particles alone. Consequently, a robust die shear strength of 22.2 MPa was achieved by sinterbonding for even 10 min at low temperature and the compression of 10 MPa, even though Cu oxide shells were formed in the bond-line because of the long sintering in air. The simple wet process provided an efficient preparation of an effective filler system before the paste formulation for the sinter-bonding.
1. INTRODUCTION
As Si semiconductors are replaced by SiC and GaN in power modules to enhance their efficiency and reduce switching loss, the current solder alloy chip bonding materials need to be substituted with a metal that has both a high melting point and thermal conductivity, because the chip junction temperature can reach 300 °C [1-6]. Transient liquid phase (TLP) sinter-bonding, developed as a better alternative to soldering, has not been considered as an eventual alternative for automotive applications because of its long bonding time and brittle bond-line characteristics [7-9]. Efficient sinter-bonding techniques based on silver (Ag) have been researched [10-12], however, the Ag powder-based sinter-bonding paste is a large burden to industry because of the inherently high cost of Ag and long bonding time [13-15]. Accordingly, Cu-based pastes are considered to be more practical sinter-bonding materials, and relevant studies are being conducted [16,17].
Copper easily oxidizes in air, and as a result Cu has a native oxide layer on its surface. Moreover, this oxidation tendency accelerates when temperature increases [18] and the layer acts as an obstacle to sinter-bonding. To address this, the adoption of Ag-coated Cu particles and in situ reduction of the oxide layers on Cu particles at sinter-bonding temperatures using a reducing atmosphere and ingredient have been suggested [19,20].
In this study, a paste containing Cu formate particles was chosen as the material for sinter-bonding, using a more aggressive concept. A pioneering study by Yabuki et al. on the application of a copper complex paste containing Cu formate particles introduced the low-temperature fabrication of pure Cu films by thermal decomposition under a nitrogen atmosphere. [21,22]. At the same time, Lee et al. reported the formation of Cu films under an identical atmosphere using a paste formulation containing Cu complex [23-25]. The results of these studies indicated that pastes containing Cu complex could be employed as effective sinter-bonding materials when assisted by the application of external pressure. Therefore, the objective of the present study was to implement high-speed sinter-bonding at a relatively low bonding temperature of 250 °C in air using low-cost Cubased (Cu formate and pure Cu) particles and a Cu-finished chip and substrate. In addition, we synthesized filler particles mixed in situ with Cu formate and pure Cu during the reduction of Cu2O particles, because the reduction of Cu formate causes the volume to shrink significantly (approximately 90%) [26].
2. EXPERIMENT
2.1 Particle synthesis and paste preparation
Filler particles mixed with Cu and Cu formate were synthesized using a simple wet reaction [27]. First, 25–75 g of Cu2O powder (95%, Daejung Chemical) was added into 252 mL of formic acid (HCOOH, 85%, Samchun Pure Chemical). Second, the slurry was stirred at 250 rpm for 15-25 min. The resultant slurry was washed three times with ethanol. Finally, the sludge was decanted and dried in a lowvacuum oven for 5 h. Subsequently, a paste containing the obtained particles was prepared by mixing the dried particles with α-terpineol (98.5%, Samchun Pure Chemical) at the weight ratio of 7:3 using a spatula.
2.2 Sinter-bonding
Dummy Cu substrates and chips were polished with a 2000-mesh sandpaper and then etched in a 10% H2SO4 solution for 1 min. The prepared paste was screen-printed on a dummy Cu substrate over an area of 3×3 mm, followed by drying at 150 °C for 3 min to decrease the amount of solvent by evaporation. Subsequently, a 3×3×1 mm dummy Cu chip was placed on the printed pattern to form a sandwich structure. For sinter-bonding, the sandwich-structured sample was heated at 250 °C at a continuously applied external pressure of 5 or 10 MPa using a thermo-compression bonder, and held for 1-10 min at that temperature.
2.3 Characterization
The morphology of the initial Cu2O and synthesized Cu complex particles, the cross-sectional microstructures of the bond-lines, and the fracture surfaces were characterized using high-resolution scanning electron microscopy (HR-SEM, SU8010, Hitachi High-Technologies Co.). Thermal analysis of the particles was carried out while heating to 400 °C in air, at a heating rate of 10 °C/min using thermogravimetry-differential thermal analysis (TG-DTA, DTG-60, Shimadzu) to estimate the thermal behavior of the Cu complex particles during heating. Phase transformation in the particles after wet synthesis or heating was determined by X-ray diffraction (XRD, DE/D8 Advance, Bruker). The solidity of the formed bond-line was measured during shearing at 200 μm/s at a shear height of 200 μm using a bond tester (DAGE 4000, Nordson Corp.), and the bonding strength was defined as the measured maximum stress.
3. RESULTS AND DISCUSSION
3.1 Synthesis of Cu complex particles
Fig 1 shows the morphologies of the initial Cu2O particles and back-scattered electron (BSE) images of particles mixed with Cu formate, and Cu particles obtained using different amounts of Cu2O, following wet synthesis for 15 min in formic acid. The synthesized particles indicated two completely different morphologies. The white and tiny pseudo-spherical shapes and grey flake-type particles in the BSE images represent pure Cu and Cu formate phases, respectively [27], which were in situ mixed during the synthesis. However, the degree of aggregation of the synthesized particles differed significantly based on the initial amount of Cu2O. The best dispersion among the samples was accomplished when 50 g Cu2O was used.
Fig 2 shows the BSE images of the Cu complex particles obtained at various synthesis times when 50 g Cu2O particles was added. In principle, the aggregation between particles intensified and the number of pure Cu particles decreased with increasing synthesis time. The XRD patterns of the particles measured as a function of synthesis time are shown in Fig 3.
In every sample of 15-25 min synthesis times, pure Cu, Cu formate, and slight Cu oxide phases were indexed. However, the peaks of Cu formate were slightly higher with an increase in synthesis time, indicating that the amount of Cu formate increased with time. The Cu oxide phase seemed to be the result of the oxidation of the synthesized pure Cu in air.
The formation of Cu formate from Cu2O can be accomplished through following two reactions [27]:
Thus, with insufficient synthesis time, a pure Cu phase will be obtained. The 15-min synthesis using 50 g Cu2O particles was chosen as an optimal synthesis condition to prepare a filler for a paste with both less aggregation and a higher total amount of pure Cu particles in the filler material.
3.2 Thermal properties of the synthesized particles
The TG-DTA results of the synthesized particles mixed with Cu formate and Cu are presented in Fig 4.
The Cu formate in the particles began to decompose at approximately 200 °C via the sequential reactions of Equations (3) and (4) [28,29], resulting in abrupt weight loss and heat generation:
Accordingly, exothermic peaks of the two steps were observed in the DTA curve during the weight-loss period; the appearance of the peaks was slightly delayed compared to that observed in a previous report due to the fast heating rate [28]. The generated heat (26±2 kJ/mol [30]) can substantially contribute to sinter-bonding behavior. The weight loss approached approximately 30% at 251 °C, which implies that the amount of Cu formate in the mixed particles was 62 % in weight.
The XRD result indexed after heating the Cu complex particles at 300 °C is shown in Fig 5. The peaks of Cu formate disappeared and those of pure Cu were strongly indexed as a product of the pyrolysis. The Cu generated in situ during the pyrolysis would be inherently active, thus we expected that it would exhibit strong sinterability under physical contact. In the results, the peaks of copper oxides were also indexed owing to the heating in air. However, the degree of oxidation can be significantly reduced in a real bond-line if the bond-line can be rapidly compacted by an external pressure, unlike the heating conducted without pressure for this measurement. This result clearly demonstrates that the Cu formate used in the Cu complex particles was transformed to Cu by pyrolysis through heating in air.
3.3 Sinter-bonding properties of the paste containing synthesized particles
Fig 6 shows the average shear strength values of dummy Cu chips sinter-bonded at 250 °C in air as a function of bonding time. Overall, the strength increased, with an increase in bonding time under 10 MPa pressure. The strength finally approached the excellent value of 22.2 MPa after bonding for 10 min, surpassing the strength of Pb-5Sn bond-lines [31]. When the bonding pressure was 5 MPa, however, the strength decreased to less than half for an identical bonding time. These results indicate that rapid sinter-bonding within 10 min at 250 °C in air using Cu-based particles is feasible under a pressure of 10 MPa, which is a significantly faster method compared to those performed at higher temperatures in inert atmospheres using Cu-based particles [32].
The BSE images of bond-lines sinter-bonded at 250 °C under 10 MPa are shown in Fig 7. In cases where the density of the formed bond-line is high, there may be instances where micro voids are filled due to the influence of debris during the polishing process. However, in this study, it was distinctly observed that there were hardly any instances of voids being filled, as the density of the bond-line was not high or only macro voids primarily exist.
For the samples bonded up to 3 min, compact microstructures were not formed in the bond-lines although three-dimensional connections between the remnant particles were formed. However, compact bond-line structures were observed in the 5-min bonded samples, and bonding areas at the chip/bond-line particles and bond-line particles/substrate interfaces were significantly enhanced in the samples. Detailed observation of the bond-lines revealed the formation of oxide shells on the surfaces of the remnant Cu particles. The formation of such oxide shells is anticipated to have a somewhat detrimental effect on the thermal conductivity property within the bond-line.
3.4 Fractography
Fig 8 shows images of the fracture surfaces on the dummy chips sinter-bonded at 250 °C with respect to bonding time. Bond-line/substrate interfacial failures were observed in the 1- and 3-min bonded samples, indicating that the interface was weakest owing to insufficient sintering. However, the failure mode changed to coherent failure within a bond-line from the bonding of 5 min, implying that sinter-bonding at the bond-line/substrate interface was reinforced. Furthermore, shear bands indicating the ductile fracture were clearly observed on the fracture surfaces, and the number of shear bands increased in the 10-min samples. Although oxide layers existed on both substrate and chip, the sinter-bonding at the interface seemed to be responsible for the in-situ reduction of the oxide layers during the bonding. The hydrogen gas emitted by Equations (3) and (4) during the pyrolysis facilitated robust interfacial sintering on the Cu finish by reducing the surface oxide at that temperature with the assistance of external pressure. Immediately after the reduction, active Cu already reduced in the bond-line began to sinter with the Cu finishes of both substrate and chip, and solidity at the interfaces was attained with an increase in the interfacial sintering area by the increase in bonding time.
4. CONCLUSIONS
Robust die bonding using a Cu-finished substrate and chip was achieved at high speed by sinter-bonding in air at a low temperature of 250 °C, using a paste containing a mixture of Cu formate and pure Cu. The filler mixture of Cu formate and Cu particles was spontaneously formed by the simple wet reaction of Cu2O particles for a short time of 15 min. During bonding, the Cu generated in situ by pyrolysis at temperature exceeding 200 °C exhibited significant sinterability, and the simultaneously emitted hydrogen reduced oxide layers on the Cu finishes. As a result, dies that were sinter-bonded even for 10 min under 10 MPa compression exhibited a sufficient shear strength of 22.2 MPa even though Cu oxide shells formed in the bond-line, since the process was conducted in air. Compared to the use of Cu formate particles alone, the filler mixture resulted in low volume shrinkage in the bond-line during bonding. The simple wet reaction provides an efficient preparation method for an effective filler system for sinter-bonding.
Acknowledgements
This study was supported by the Research Program (U2023-0102) funded by SeoulTech (Seoul National University of Science and Technology).