Synthesis of Hierarchical Structured Cu–Sn Alloy Mesoparticles and Its Application of Cu–Cu Joint Materials

Toshihiro Kuzuya; Toma Takedachi; Tetsuya Ando; Yasuharu Matsunaga; Ryouya Kobayashi; Yoshihiro Shimotori; Naofumi Nakazato; Hiroshi Nishikawa; Takuya Naoe

doi:10.2320/matertrans.MT-MC2022008

Abstract

In the synthesis of Cu–Sn alloy nanoparticles, we found that the addition of Tin-Ethylhexanoate provided Cu–Sn alloy mesoparticles, which are aggregates of nanoparticles with a diameter of less than 100 nm. HRTEM observation revealed that Cu–Sn alloy constituent nanoparticles contain plane defects and domains with various crystal orientations. Also, mesoparticles have a Cu₂O layer with approximately 4 nm thickness on their surface, in which the Cu₂O (111) plane is parallel to the Cu (111) plane of the lower layer or is inclined slightly. When Cu–Sn alloy mesoparticles with a composition of Sn: 0.42 at% were used as joint materials of the Cu–Cu joint, the maximum shear stress of the joint interface was measured to be more than 11 MPa. In the case of Cu–Sn alloy mesoparticles with 1.5 at% Sn, the maximum shear stress decreased significantly, which is considered to be attributed to the formation of the Cu–Sn intermetallic compound phase. Therefore, mesoparticles with 0.42 at% Sn may be a strong candidate for a low-cost Cu–Cu joint material, which could be used as joint materials for electronic devices under high temperatures than conventional ones.

1. Introduction

The melting temperatures of metal nanoparticles (NPs) are lower than those of bulk materials. Therefore, metal NPs are strong candidates for bonding materials for electronic components and metal inks for printed wiring boards.¹^–⁴⁾ In addition, since they have heat resistance and heat fatigue resistance comparable to those of Pb solder, they can be an indispensable material for bonding materials for power semiconductors that operate under high-temperature conditions. Au and Ag NP pastes have already achieved target values of maximum shear stress (20 MPa) and thermal conductivity (50 W/mK) required for power device mounting.⁵⁾ And they have high chemical stability and have been put into practical use as products. In recent years, industrial and scientific interest has been shifting to Cu NPs with lower cost and higher migration resistance.⁶⁾

The melting point of NPs depends on their size. When the size of Cu NP is smaller than 20 nm, its melting point decreases drastically with its size and is estimated to be approximately 700 K at 4 nm of its size.⁷⁾ Therefore, it is necessary to reduce the size of Cu NPs to 20 nm or less to realize low-temperature sinterability. Compared with Au and Ag, Cu has relatively poor chemical stability. Therefore, oxygen and H₂O easily react with Cu NPs to form oxide and hydroxide gel, making long-term storage of Cu NPs difficult. In addition, the NPs fabricated by the solution-phase synthesis method have a passivation layer consisting of organic compounds on their surface, which contributes to the prevention of oxidation and the improvement of dispersibility in the solvent. However, compared to the particle size, when the thickness of this passivation layer is not negligible, shrinkage of the passivation layer leads to the generation of cracks in sintered compact.

Yonezawa et al. synthesized Cu NPs with an average size of 100 nm and investigated their sintering behavior.⁸⁾ They confirmed that the introduction of 2-step sintering enables sufficient low-temperature sinterability. In this method, the NPs are first heated in an oxidizing atmosphere to precipitate fine oxide NPs on the surface. Then a reducing gas is introduced to convert the oxide NPs into metal NPs and connect the NPs by necking. In addition, Yuan et al. added metal-organic compounds into Cu NPs paste.⁹⁾ Thermolysis of these compounds leads to the generation of fine copper NPs during sintering, enabling low-temperature sintering. On the other hand, Kamikoriyama et al. formed an oxidation-resistant film that decomposes at low temperature on the surface of Cu NPs to achieve low-temperature sinterability and chemical stability.¹⁰⁾ In the field of power devices, which require high reliability, much research has been conducted on copper bonding using Ag, Cu, and both alloy particles.¹¹⁾

Herein, Cu–Sn mesoparticles (MPs) with a hierarchical structure were prepared by using the decomposition of metal-organic compounds. We employed Tri-n-octylamine (TOCA) as a solvent, because it serves as a mild reducing agent and weak ligand. TEM and STEM-EDX analysis revealed detailed information on the nanostructure of MPs. The dependence of the melting point of NPs on their curvature reminds us that hierarchical structured MPs consisting of fine NPs may have low-temperature sinterability while maintaining high chemical stability.¹²⁾ In this study, we investigated the Cu–Cu joints properties of this MPs at relatively low sintering temperature.

2. Materials and Methods

2.1 Synthesis of Cu–Sn alloy MPs

All reagents were used as received. When the Cu–Sn alloy MPs were synthesized, 4 mmol of copper oleate (Cu-OA; Wako, contain 8.8 mass% of Cu) was mixed with a ligand mixture solution (4.22 mmol oleic acid (OA; Wako, 60%); 2.1 mmol oleylamine (OLA; Wako); and 20 mL tri-n-octylamine (TOCA; Wako, 97%)) in a vial bottle. This mixture solution was then degassed using a vacuum pump until the pressure in a bottle reached 13 hPa. After degassing, Tin-ethylhexanonate (Sn-EH; TCI, >85%) was injected into the mixture solution, and it was annealed for 1.8 ks at 393 K in an Ar atmosphere (Low-temperature annealing). Then, this mixture solution was annealed at 473 K for 1.8 ks (High-temperature annealing). After quenching, ethanol (Wako, 99.5%) was added to the reaction mixture at 298 K to obtain a dark red precipitate. The precipitate was separated by centrifugation to remove excess reagents and redispersed in hexane (Wako, 96%). The precipitation/redispersion process was repeated several times to purify the NPs.

2.2 Characterization

In order to prepare the TEM/STEM sample, a drop of the NP-hexane dispersion was placed on a carbon-coated Cu or Mo microgrid. TEM, annular dark field (ADF) images, and energy–dispersive X-ray spectroscopy (EDS) measurements were acquired using a field emission TEM/STEM (JEOL, JEM2100F with a point-to-point resolution of 0.23 nm) at 200 kV.

2.3 Image analysis and creation of the lattice mapping image

HRTEM images were converted into fast Fourier transform (FFT) images using Image J software. In order to obtain a lattice pattern distribution image, only the spots corresponding to the specific lattice pattern in the FFT image were inverse transformed. Each lattice pattern distribution was colored with a specific color and superimposed to create a lattice mapping image.¹³⁾

2.4 Evaluations of mechanical strength and thermal conductivity

A slurry of Cu–Sn hierarchical MPs was used as Cu–Cu joint materials. Cu (C1100) plate with dimensions of (a) 10 × 20 × t1.0 or (b) 10 × 30 × t0.5 mm was etched by 1 mol% HNO₃ aqueous solution to eliminate organic contamination. Then, Cu-plate was rinsed with distilled water and ethanol. After drying, 0.1 ml of a slurry of Cu–Sn hierarchical MPs (0.2 g/ml) was deposited on the Cu plate. And Cu plate with Cu–Sn MPs was dried in air. The Cu plate with the same dimension was placed onto the deposited MPs layer to fabricate the sandwich structure samples. Schematic illustrations of Cu–Cu joint test pieces for the evaluation of mechanical strength and thermal conductivity are shown in Figs. 1(a) and (b), respectively. In the ambient atmosphere, they were pressed by hot plates pre-heated at 573∼613 K under the pressure of 40∼120 MPa for 600 ks. Then, it was quenched to ambient temperature.

Fig. 1

Schematic illustration of a Cu–Cu joint test piece for (a) mechanical strength and (b) thermal conductivity measurement.

In reference to the procedure of JIS K 6850, the evaluation of the mechanical strength of the Cu–Cu joint was conducted by using a tensile tester under a tensile rate of 0.008 mm/s in an ambient atmosphere.¹⁵⁾ Experimental conditions are summarized in Table 1.

Table 1 Experimental conditions and results of shear stress measurements.

The relative thermal conductivities of Cu–Cu joint test pieces containing MPs sintered layer were evaluated by the temperature gradient method.¹⁴⁾ We normalized the thermal conductivity of samples (Cu plates(a) were used as reference sample) by using measurement values of Cu plate with a dimension of 10 × 20 × t2 mm. A sample was inserted between two aluminum blocks. Each block was equipped with four thermocouples (T-type) at regular intervals. An effective thermal conductivity (k_eff) of the sample is represented as the following equation;

\begin{equation*} k_{\textit{eff}} = q_{S}\cdot t_{S}/\Delta T_{S} \end{equation*}

where q_S, t_S, and ΔT_S, are the heat flux at the center of the sample, the sample thickness, and the temperature difference between very upper and very lower sides of the sample, respectively. In order to evaluate q_S, the temperature distribution of the heat transfer aluminum block was measured. Temperatures of the very upper and very lower sides of the sample were obtained by extrapolating the temperature distributions of the heat transfer aluminum blocks.

3. Results and Discussion

3.1 The formation of Cu–Sn MPs

When Cu-OA was annealed in a mixture of oleylamine, oleic acid, and TOCA, the color of the solution turned from deep blue to reddish-brown at higher than 473 K. This color change indicates the formation of Cu NPs. Yamamoto et al. reported that highly monodispersed Ag NPs could be synthesized by reducing the Ag carboxylate (RCOOAg) with a tertiary amine (R′₃N; R and R′ represents an alkyl group).¹⁶^,¹⁷⁾ The tertiary amine reacts with Ag oleate to form bis (amine) silver (I) carboxylate (R′₃NAgRCOO), which is thermally unstable. In this complex, the lone electron pair of the N atom moves to the Ag ion. The tertiary amine changes to a polyamine or an amine oxide (R₃NO). To our best knowledge, there is no report dealing with the redox potential of TOCA. The redox potential of 3-aminophenol is measured by cyclic voltammetry to be approximately 0.7 to 1.0 V (vs. Standard Hydrogen Electrode (SHE)).¹⁷⁾ Whereas the redox potential of Cu is 0.52 V for Cu⁺ + e⁻ = Cu and 0.34 V for Cu²⁺ + 2e⁻ = Cu, so Cu precipitation is considered to be due to thermal decomposition of Cu-OA.

Figures 2(a)–(b) shows SEM and TEM micrographs of Cu NPs. These figures indicate the formation of nearly monodispersed Cu NPs with an average size of 404 nm and the standard size deviation of 19% of the average size (see Fig. 2(a)). As shown in the HRTEM image of Cu NP (Fig. 2(b)), two different fringes in the core and shell region were observed with interfringe distances of 0.20 (dCu (111)) and 0.23 nm (dCu₂O (111)), corresponding to the {111} planes of Cu (Fm-3m) and Cu₂O (Pn-3m), respectively.

Fig. 2

(a) SEM image and (b) HRTEM image of Cu NPs obtained by thermal decomposition of Cu-OA.

Figures 3(a)–(b) show SEM and TEM micrographs of Cu–Sn alloy precipitates synthesized with Sn-EH of 0.09 ml (Sa. #1) (a) and 0.28 ml (Sa. #2) (b). The addition of Sn-EH led to a significant change in the morphology and size of a precipitate. Cu–Sn alloy precipitates synthesized with Sn-EH take a mesoscale hierarchical structure consisting of primary NPs. The sizes of primary NPs were estimated from FWHM values of XRD patterns to be 21∼34 nm for Sa. #1 and 15∼26 nm for Sa. #2. The electron diffraction patterns of Sa. #2 (Fig. 3(c)) exhibits characteristic ring patterns corresponding to Cu (111), Cu (200), Cu (220), Cu (311), and Cu₂O (111), but we could not observe the ring patterns attributed to Sn compounds such as SnO₂ and Cu–Sn intermetallic compounds. TEM micrographs shown in Figs. 3(a) and (b) indicate the formation of nearly monodispersed MPs with the average particle size of 440 nm for Sa. #1 and 680 nm for Sa. #2. The average size of MPs increased with the amount of Sn-EH.

Fig. 3

SEM and TEM micrographs of (a) SA. #1, (b) SA. #2 MPs, and (c) electron diffraction pattern of SA. #2 MPs.

We have investigated the element distribution in Cu–Sn alloy MPs by using STEM-EDX. The elemental mapping image indicated that Sn was distributed uniformly in the MPs (see Fig. 4). Because of the relatively high affinity of Sn for oxygen, Sn is considered to be concentrated in the oxide phase. However, the Sn-rich oxide shell could not be confirmed in Fig. 4. Because the thickness of the oxide layer is comparable to probe size, STEM-EDX resolution is insufficient to reveal the existence of a surface oxide layer. Furthermore, the relatively low metal density of the oxide phase reduces the intensity of the characteristic X-ray signal. The average concentrations of Sn and O in the MPs were Sn: 0.42 ± 0.066 at% (0.79 ± 0.12 mass%), O: 1.7 ± 0.44 at% (0.43 ± 0.11 mass%) at Sa. #1, and Sn: 1.5 ± 0.32 at% (2.8 ± 0.58 mass%), O: 2.1 ± 0.86 at% (0.54 ± 0.22 mass%) at Sa. #2. Though the composition of each particle was also measured, there is no correlation between Sn and O concentrations in the MPs. The Sn concentration in the MPs was smaller than the dosage ratio. The maximum solid solubility limit of Sn in the Cu phase is about ∼0.7 at% (1.3 mass%) at synthesis temperature (473 K). Therefore, quenching of Cu–Sn alloy NPs to the ambient temperature may lead to the precipitation of the ε-phase. Regardless of our careful TEM observation, the ε-phase could not be confirmed.

Fig. 4

STEM-EDX images of SA. #2 MPs. (a) ADF image, (b) Cu, (c) Sn, (d) O mapping images, and (e) mapping image in which (b), (c), and (d) are superimposed on ADF image (a).

Figure 5(a) shows an HRTEM micrograph of Sa. #2. Lattice patterns with interfringe distances of 0.2 nm corresponding to the Cu (111) and 0.23 nm corresponding to Cu₂O (111) planes could be observed in this figure. FFT image of the HRTEM micrograph (Fig. 5(b)) exhibits several spots corresponding to Cu (111) and Cu₂O (111). The angle formed by spots A and B corresponding to Cu {111} is 60°, which is smaller than the ideal angle (70°). Next, each lattice domains corresponding to spots A, B, C, and D shown in Fig. 5(b) were obtained by an inverse FFT of only a specific spot. Figure 5(c) is a lattice mapping image in which each lattice domain is colored and superimposed. The purple and light blue domains in Fig. 5(c) correspond to Cu (111) spots A and B in Fig. 5(b), indicating that there are several domains having different crystal orientations. Therefore, primary NP contains stacking defects between crystal domains. The red, green, and blue domains in Fig. 5(c) correspond to the spots C, D, and E (Cu₂O (111) plane). MPs have a Cu₂O layer with a thickness of approximately 4 nm on their surface, in which the Cu₂O (111) plane is parallel to the Cu (111) plane of the lower layer or is inclined by approximately ∼8°.¹⁸⁾

Fig. 5

HRTEM image and its FFT analysis of Sa. #2. (a) HRTEM image, (b) its FFT image, and (c) Lattice mapping image obtained by inverse FFT of (b).

The addition of Sn-EH led to the drastic morphological change of precipitates. As shown in Fig. 3, MPs had a hierarchical structure, which consisted of fine Cu–Sn primary NPs. Sun et al. synthesized hierarchical Cu metal NPs by reducing octahedral Cu₂O NPs with hydrazine.¹⁹^,²⁰⁾ The volume shrinkage associated with the reduction reaction from Cu₂O to Cu led to a porous structure. It is also possible to fabricate a hierarchical structure by using the so-called “bottom-up” process. The primary NPs were modified by a polymer or a surfactant in this process. Surfactant or polymer adsorbed on the NP’s surface mediates the self-assembly of primary NPs to form a hierarchical structure. For example, suspensions of iron oxide primary NPs modified with oleyl phosphate were evaporated to form a 3D superlattice via self-assembly of iron oxide primary NPs.²¹⁾ TEM images of Cu–Sn alloy MPs indicate that this hierarchical structure is due to the self-assembly of primary NPs. Sn (II) serves as a reduction agent for Cu²⁺ ions (R1) because the redox potential of 0.15 V (R2) is lower than that of Cu²⁺ (0.34 V).²²^,²³⁾

\begin{equation} \text{Cu$^{2+}$} + \text{Sn$^{2+}$} = \text{Cu(0)} + \text{Sn$^{4+}$} \end{equation}

(R1)

\begin{equation} \text{Sn$^{2+}$} = \text{2e$^{-}$} + \text{Sn$^{4+}$} \end{equation}

(R2)

Indeed, the addition of Sn-EH changed the color of the mixture solution from dark blue to dark brown during low-temperature annealing, indicating the formation of metal NPs. These NPs agglutinate into Cu–Sn MPs with a hierarchical structure to reduce the surface free energy. FT-IR spectrum of Sa. #2 indicated adsorption of oleic acid on their surface. The oleic acid prevents the aggregation of particles and serves as an excellent dispersant for non-polar solvents. In this synthesis system, TOCA with a weak polarity was used as s solvent. Therefore, the NPs with hydrophobic surface aggregate during high-temperature annealing. Then, fine primary NPs join because of their lower melting point. And the decomposition of remained Cu-precursor provides Cu monomer and aggregates grow uniformly, resulting in MPs with a hierarchical structure. On the other hand, when Sn-EH is not added, nucleation occurs during high-temperature annealing, and the nuclei grow rapidly before aggregation. Though the average size of MP increased with the amount of Sn-EH, contrary the size of primary NP which composed the MP decreased. Therefore, in order to reduce the surface free energy of the primary NPs system, the smaller the primary NPs, the more the aggregation proceeds. This fact indicated that an increase in the amount of Sn-EH enhanced the nucleation of Cu–Sn primary NPs.

3.2 Mechanical and thermal properties of Cu–Sn MPs

In this study, the mechanical and thermal properties of Cu–Sn MPs joint materials were evaluated. Figure 6 shows shear stress (σ/MPa) - strain (ε (−)) curves of Cu–Cu joint test pieces. Shear stress (σ) is expressed as σ = f/S, where f and S are load force (N) and joint area (m²), respectively. In the case of Sa. #1, Cu–Cu joint obtained under (573 K; 50 MPa) was pealed at σ = 9 MPa. On the other hand, Cu–Cu joint obtained under (573 K; 80 MPa), (613 K; 50 MPa), and (613 K; 80 MPa) using Sa. #1, did not peel off and the Cu plate fractured. At f = 1 kN, the shear stress on the Cu–Cu joint and tensile stress on the Cu plate reach 10 MPa and 200 N/mm². The tensile strength of the Cu (C1100) plate ranges from 200 to 300 N/mm². Therefore, maximum shear stresses of the Cu–Cu joint obtained under (573 K; 80 MPa), (613 K; 50 MPa), and (613 K; 80 MPa) are considered to be larger than 11 MPa. When Sa. #2 was used as Cu–Cu joint material, maximum shear stresses of Cu–Cu joints sintered at 613 K under the pressure of 40∼120 MPa were estimated to be smaller than 5 MPa. The maximum shear stresses of Sa. #1 were larger than those of Sa. #2. When the Sa. #2 joint material is sintered, Cu–Sn intermetallic compounds which were estimated ε- and/or other phases were formed in the joint layer, leading to a decrease in the maximum shear stress. Also, compared with previous results, relatively high pressure was required to bring the metal cores into contact with each other due to the surfactant and/or oxide film on the surface. This may be resolved by annealing in an anaerobic atmosphere.

Fig. 6

Shear Stress (σ/MPa) - Strain (ε (−)) curves of Cu–Cu joint test pieces (SA. #1 MPs).

Finally, the thermal conductivities of the joint interface was evaluated. Cu plates were joined at a sintering temperature of 573∼613 K under 40∼120 MPa. Thermal conductivities of the samples jointed by Sa. #1 and Sa. #2 were estimated to be from 40 to 70% of the Cu reference’s value (see Table 2). The decrease in thermal conductivity is considered to be due to voids and the oxide layer on the MP’s surface.

Table 2 Measurement results of relative thermal conductivities of Cu–Cu joint test pieces.

4. Conclusion

Cu–Sn alloy MPs were synthesized by annealing Cu-OA and Sn-EH in a mixture solution of TOCA and surfactants. This MP has a hierarchical structure and consists of fine Cu–Sn alloy NPs with an average size of 100 nm. We tried to join Cu plates using these MPs as a joint material and evaluate their mechanical and thermal properties. The results are summarized below.

(1) Sn-EH reacts with Cu-OA to form Cu–Sn NPs with an average size of about 100 nm. Cu–Sn NPs agglutinate and grow uniformly, resulting in the formation of hierarchical MPs.
(2) MPs has an oxide layer on its surface. The orientation of Cu₂O crystal depended upon that of base-Cu crystal.
(3) The Sn concentration in MPs was smaller than the dosage ratio. The Sn concentrations in SA. #1 and SA. #2 MPs were estimated to be about 0.42 ± 0.066 at% and 1.5 ± 0.32 at%, respectively.
(4) The maximum shear stress of the joint interface was measured to be about 11 MPa at Sa. #1. In the case of Sa. #2, the maximum shear stress decreased because of the formation of the Cu–Sn intermetallic compound phase.
(5) The thermal conductivities of the joint interface ranges from 40 to 70% of the thermal conductivity of Cu.

Acknowledgements

We would like to acknowledge Dr. Kawamura for the assistance provided with the TEM observations.

Funding: This Work was partially performed under the Joint Usage/Research Center on Joining and Welding, Osaka University.

REFERENCES

Corresponding author

Register with J-STAGE for free!