Effect of electric current on grain orientation and mechanical properties of Cu-Sn intermetallic compounds joints

Accepted Manuscript Effect of electric current on grain orientation and mechanical properties of Cu-Sn intermetallic compounds joints Jiayun Feng, Chunjin Hang, Yanhong Tian, Chenxi Wang, Baolei Liu PII:

S0925-8388(18)31327-6

DOI:

10.1016/j.jallcom.2018.04.041

Reference:

JALCOM 45679

To appear in:

Journal of Alloys and Compounds

Received Date: 16 December 2017 Revised Date:

14 March 2018

Accepted Date: 3 April 2018

Please cite this article as: J. Feng, C. Hang, Y. Tian, C. Wang, B. Liu, Effect of electric current on grain orientation and mechanical properties of Cu-Sn intermetallic compounds joints, Journal of Alloys and Compounds (2018), doi: 10.1016/j.jallcom.2018.04.041. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

Effect of electric current on grain orientation and mechanical properties of Cu-Sn intermetallic compounds joints Jiayun Feng, Chunjin Hang, Yanhong Tian*, Chenxi Wang, Baolei Liu State Key Laboratory of Advanced Welding and Joining, Harbin Institute of

RI PT

Technology, Harbin 150001, China * Corresponding author: Tel: +86(0)451-86418359

SC

[email protected] (Tian Yanhong)

M AN U

Abstract

Electric current-assisted bonding experiments have been performed on Cu/Sn/Cu joints using an electric current density of 2.0×102 A/cm2 at 260 °C. The effect of electric current on the grain orientation of Cu–Sn intermetallic compounds (IMCs) and mechanical properties of Cu/Sn/Cu joints were evaluated. The scanning electron

TE D

microscopy (SEM) observation indicated that the electric current had obvious polarity effect on Cu6Sn5 growth. The Electron backscatter diffraction (EBSD) analysis revealed that Cu6Sn5 showed a preferred orientation of [0001] direction being parallel to the electric current. On the contrary, the morphology and grain orientation of Cu3Sn

EP

were barely influenced under the electric current density of 2.0×102 A/cm2. The calculation results indicated that the Cu6Sn5 (0001) plane showed the lowest

AC C

projection atomic density of 27.9 atoms/nm2. Therefore, the Cu6Sn5 grains with their [0001] directions oriented along electric current could grow faster because of smaller electron wind force and greater Cu atom flux. The results of nanoindentation experiment and shear test also proved that the full-Cu6Sn5 joints formed under electric current had lower Young’s modulus mismatch between Cu6Sn5 and Cu3Sn, and had a joint shear strength of 44.87 MPa. The present work offers a method to adjust the texture of Cu–Sn IMCs in the joints for achieving better electrical and mechanical properties. Keyword: Cu/Sn/Cu interconnect; Intermetallic compounds; Electron backscatter

ACCEPTED MANUSCRIPT diffraction (EBSD); Nanoindentation.

1. Introduction Electronic products have developed significantly towards miniaturization and in

recent

years.

Through

Silicon

Via

(TSV)

in

RI PT

multi-functionalization

three-dimensional (3D) integration is considered to be the most promising technique to satisfy these requirements in the coming period because of its high integration density, low signal latency and parallel processing capability [1-6]. The Sn-based

SC

solder alloy is commonly used for micro-bumps between under-bump metallization (UBM) layers on the chips, where Cu is usually used to contact with the solder for its

M AN U

good wettability [7, 8].

One critical requirement for TSV is that solder joints must endure repeated multilevel 3D stacking of additional layers without remelting of the joints at lower levels. Therefore, the solid-liquid interdiffusion (SLID) bonding process was developed to form entire high-melting-point Cu-Sn intermetallic compound joints at

TE D

lower temperatures, which can serve at higher temperatures [9]. However, it is known that the Cu6Sn5 and Cu3Sn compounds exhibit obvious anisotropic mechanical, electrical and electromigration resistance properties. When the microbump diameter

EP

decreases to the 1-10 µm range [10-12], the entire solder joint may contain only a few Cu-Sn IMC grains. Therefore, the orientation of Cu–Sn IMCs will greatly influence the reliability issues related to microbumps. It is critical to adjust the texture of Cu–Sn

AC C

IMCs in the joints to achieve better electrical and mechanical properties. As recently reported by Liu et al. [13], liquid-phase electroepitaxy (LPEE)-grown Cu3Sn and Cu6Sn5 grew in the direction of least electrical resistance. Along these particular directions, the traveling electrons were scattered least by the lattices, and the newly formed Cu-Sn unit cells thus oriented themselves in a particular orientation to facilitate the electron flow. In this way, the grain orientation adjustment by electric current became feasible. However, the reaction temperature for adjustment reached more than 400 °C, which is not suitable for electronics manufacturing.

ACCEPTED MANUSCRIPT In this work, the electric current-assisted bonding process was performed under soldering temperature to reveal the microstructure and orientation evolution of the Cu-Sn IMCs in solder joints. A relatively low current density of 2.0×102 A/cm2 was chosen to slow Cu6Sn5 growth, which was helpful for observation. The morphology

RI PT

evolution of Cu–Sn IMCs was characterized by SEM. The grain orientations of Cu-Sn IMC textures were analyzed using EBSD technique. The nanoindentation and shear tests were also performed to evaluate the mechanical properties of full-Cu6Sn5 joints. 2. Experimental

SC

The samples used in this study were Cu/Sn foil/Cu sandwich structures. 1 mm-in-thickness, 99.99% purity Cu plates and 50 µm-in-thickness 99.95% purity Sn

M AN U

foils were chosen as base metal substrates and interlayers, respectively. Both of them were cut into 2.5×2.5 mm2 pieces. Cu plates were first polished with 0.5 µm polishing slurries and then cleaned using deionized water for 5 min and acetone for 5 min in an ultrasonic bath. Sn foils were first immersed in the ultrasonic bath with detergent solution for 3 min and then cleaned using the same procedures as those used in Cu

TE D

plate preparation.

The flux was used between Cu and Sn to remove the surface oxidations, in order to achieve better wettability. Then the sandwich samples were hold by two cylindrical

EP

Cu–Cr electrodes (5 mm-in-diameter). The electrodes were fixed on sliders that can move freely on the guide rail, as shown in Fig. 1. A pressure of 0.08 MPa was supplied by two springs to assure the close contact between Cu-Sn sample and the

AC C

electrodes. A dynamometer (HP-5N, HANDPI Instrument Co., Ltd, China) was used for force adjustment. The electrodes were connected to a DC power supply to form a sample–electrodes–power source circuit. Asbestos plates were used as the insulating layer between electrodes and sliders. An electric current density of 2.0×102 A/cm2 was applied in the above mentioned electrical circuit. For comparison, another two groups of samples were formed under an electric current density of 1.0×102 A/cm2 and without electric current. The experimental setup was then put into a vacuum tube furnace, heating at 260 °C for different processing times (15 min to 360 min). The soldering temperature was monitored in real time by a thermocouple.

ACCEPTED MANUSCRIPT The bonded samples were cold-embedded in resin and then ground and polished along the direction perpendicular to the bonding interfaces. Afterwards, a Quanta 200FEG SEM and the attached energy-dispersive X-ray spectroscopy (EDS) system were used to characterize the microstructure and determine the phase composition.

RI PT

The EBSD system (Quanta 200FEG, FEI Company, USA) was used to reveal the grain orientation of Cu–Sn IMCs. In the EBSD analysis, the ND (normal direction), RD (rolling direction), and the TD (transverse direction) denote the directions normal to the polished surface, parallel to the Cu/Sn/Cu interface, and parallel to the electric

SC

current direction, respectively.

The mechanical properties of Cu6Sn5 and Cu3Sn grains with certain orientations

M AN U

were evaluated using the nanoindentation technique (Nanoindenter G200™, Agilent Technologies, USA). The nanoindentation experiments were conducted with a three-sided pyramid Berkovich indenter. The expected indentation depth was 1000 nm with a strain rates of 0.05 s-1. Each sample was indented 10 times at different areas to obtain the average Young’s modulus and microhardness. In addition, shear tests

TE D

with the loading rate of 200 µm/s were performed on full-Cu6Sn5 joints using an XYZTEC Condor Sigma Bond Tester. The fractures after shear tests were then

AC C

EP

observed by SEM.

Fig. 1 Schematic of the experimental setup. 3. Results and discussion

ACCEPTED MANUSCRIPT 3.1. Morphology evolution of Cu6Sn5 and Cu3Sn under electric current stressing Cross-sectional SEM micrographs of the Cu/Sn/Cu joints subjected to electric current stressing for 15 min, 30 min, 60 min and 360 min at 260 °C are presented in Fig. 2. The EDS analysis revealed that the two adjacent layers on Cu/Sn interfaces are

RI PT

η-phase (Cu6Sn5) and ε-phase (Cu3Sn). The dotted arrows in the figures indicate the direction of electron flow. As shown in Fig. 2 (a) and (c), after 15 min, Cu–Sn IMCs at the cathode became thicker than those at the anode. The possible explanation is that the Cu atoms dissolving in the liquid solder traveled from the cathode to the anode

SC

under electromigration, causing the lack of Cu atoms at the cathode and the Cu atoms accumulation at the anode. The IMCs at the cathode kept dissolving while the IMCs

as [14]: ሬሬሬሬሬറ el ሬሬሬሬሬሬሬሬറ ሬറ=F ሬറ+Zwd eE ሬറ=Z* eE ሬറ +‫ ܨ‬௪ௗ =Zel eE F

M AN U

at the anode kept precipitating. The driving force of electromigration can be expressed

(1)

ሬሬሬሬሬറ el in which ሬFറ is the total driving force, F is the direct electrostatic force caused by

TE D

ሬሬሬሬሬሬറ wd electric field, F is the force pushes the atoms towards the anode due to the momentum exchange between electrons and atoms. Zel is the electrostatic charge number, and Zwd is the charge number due to the electron wind force. e is the

EP

ሬറ represents the electric field. The unsymmetrical growth of electron charge and E Cu-Sn IMCs is consistent with the polarity effect reported in Refs. [15-18], which

AC C

means ሬሬሬሬሬሬറ Fwd is larger than ሬሬሬሬሬറ Fel , and the sign of Z* is negative. Channels between two Cu6Sn5 grains were found at the cathode side, and Cu atoms

can easily travel through these channels into the liquid. At the anode side, the channels were blocked by IMCs, implying that Cu atoms came mostly from the cathode. Hence, the Cu6Sn5 grains at the anode side continued to coarsen according to the Ostwald ripening mechanism. Larger grains expanded in volume while smaller grains gradually disappeared. After 30 min, a few Cu6Sn5 grains growing from the anode side had touched and combined with the Cu6Sn5 from the cathode side, as shown in Fig. 2 (b) and (d). After 60 min, nearly all the liquid Sn was consumed, and

ACCEPTED MANUSCRIPT the gaps between adjacent Cu6Sn5 grains gradually disappeared. A full-Cu6Sn5 joint formed in 60 min (Fig. 2 (e) and (g)). The serrated cathode Cu substrate indicated that severe Cu depletion happened during soldering. By contrast, the anode surface kept flat, indicating a lower Cu consumption. After this stage, the coarsening process of

RI PT

Cu6Sn5 in solid state and the transition from Cu6Sn5 to Cu3Sn became the dominant reactions. Finally, after 360 min, the Cu6Sn5 phase was totally depleted (Fig. 2 (f) and (h)). The Cu3Sn layers at both sides were found to have similar thicknesses. This phenomenon illustrated that electric current had little effect on Cu3Sn growth. The

SC

possible explanation is that Cu3Sn cannot form directly in liquid Sn under low Cu concentration at 260 °C, so the formation of Cu3Sn happened mostly at the solid

M AN U

interfaces between Cu and Cu6Sn5, in which the diffusion activation energy of Cu atoms is much smaller than that in the liquid [19]. So the Cu atoms could not migrate freely under electric current. Besides, the serrated surface of cathode Cu was found to become flat again because long-time reaction between Cu6Sn5 and Cu. A honeycomb structure formed between the two Cu3Sn layers because their uneven surfaces, which

TE D

may reduce the interfacial strength. This structure could eventually disappear after

AC C

EP

long-time reaction. The SEM results are in agreement with previous reports [20].

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 2 Full-view and magnified SEM micrographs of the cross-sectional Cu/Sn/Cu joints under the condition of 260°C, 2.0×102 A/cm2 for (a) (c) 15 min, (b) (d) 30 min, (e) (g) 60 min and (f) (h) 360 min. 3.2. Grain orientation evolution of Cu6Sn5 and Cu3Sn under electric current stressing Fig. 3 provides the corresponding EBSD analysis results for the cross-sectional surfaces shown in Fig. 2. The corresponding planar inverse polar figures in the transverse direction (TD) are shown in Fig. 3 (a)-(d). The (0001) polar figures of the

ACCEPTED MANUSCRIPT Cu6Sn5 phase and (100) polar figures of the Cu3Sn phase are shown in Fig. 3 (e)-(h). Different grain orientations were indicated by different colors, as reflected in the legend. To reveal different phases more clearly, the contours of the different grain boundaries were depicted.

RI PT

Hexagonal-structure η-Cu6Sn5 (a=b=0.4218 nm and c=0.5106 nm [21]) and orthorhombic-structure ε-Cu3Sn (a=0.4320 nm, b=0.5490 nm and c=0.4740 nm [21]) are normally stable during soldering and can be retained at room temperature during the soldering and the subsequent cooling process [19]. Thus, the Cu6Sn5 and Cu3Sn

SC

grain orientations were marked below using the hexagonal prism and long-strip shaped unit cells intuitively.

M AN U

From the warm color of the Cu6Sn5 grains in Fig. 3 (a)-(c), it was deduced that the c-axis of Cu6Sn5 tended to align with the electric current direction (TD direction). Therefore, the angle θ between the c-axis of Cu6Sn5 and the TD direction was calculated using a mathematical relationship between the Euler angle of the crystal orientation (φ1, φ, φ2) and the Cartesian coordinate system transformation matrix g.

TE D

The corresponding angles were listed in Table 1. Some detailed features reflecting the orientation evolution tendencies with time were also found in these figures. For the joint subjected to electric current stressing for 15 min (Fig. 3 (a)), the newly nucleated

EP

Cu6Sn5 grains at the anode side started to touch each other with the relatively random distribution of θ values ranging from 8° to 54°. This angle range was similar to that of Cu6Sn5 grains obtained without electric current stressing. The possible explanation is

AC C

that at the initial stage, the Cu6Sn5 grain orientation was mainly dominated by other factors, such as the crystal orientation and surface structure of Cu pad, instead of the electric current. After another 15 min, some Cu6Sn5 grains grew rapidly and became larger grains, such as grains 3, 6 and 7 in (Fig. 3 (b)). There still were some smaller Cu6Sn5 grains such as grains 1, 2, 4 and 5. According to the Ostwald ripening mechanism [22, 23], the grains with higher volumes would grow larger and “swallow” the smaller neighboring grains. It was worth mentioning that all larger Cu6Sn5 grains had relatively smaller θ values, as indicated in Fig. 3 (b). After 60 min (Fig. 3 (c)), the remaining Cu6Sn5 grains started to combine with each other. Grain 2 with the smallest

ACCEPTED MANUSCRIPT θ value of 10° was the largest grain. After 360 min, only Cu3Sn clusters were left in the whole joint. The planar inverse polar figures indicated that the [100] directions of Cu3Sn were nearly parallel to the TD orientation, which was consistent with the results of annealing without electric current stressing [20].

RI PT

To demonstrate the orientation evolution tendency more clearly, (0001) polar figures of the Cu6Sn5 grains formed at 15 min, 30 min, and 60 min and (100) polar figures of the Cu3Sn grains formed at 360 min are presented in Fig. 3(e-h). In Fig. 3 (e), the red regions are mainly located at the middle of the center and at the ends of

SC

the TD axis, meaning that the c-axis of Cu6Sn5 was not parallel to the TD direction after electric current stressing for 15 min. However, after 30 min (Fig. 3 (f)), the red

M AN U

regions scattered near the ends of the TD axis and some circles were observed around these regions. After 60 min, in the polar figure of the full-IMCs joint (Fig. 3 (g)), the red regions were more concentrated near the TD ends, which indicated that the proportion of Cu6Sn5 grains with smaller θ values increased. This evolution process was consistent with the aforementioned analysis, in which the preferred orientation of

TE D

Cu6Sn5 [0001] direction being parallel to electric current direction became stronger with time. In addition, the preferred orientation of the Cu3Sn [100] direction being parallel to the TD orientation is also shown in Fig. 3 (h).

Grain 1

Grain 2

Grain 3

Grain 4

Grain 5

Grain 6

Grain 7

Grain 8

54°

12°

8°

12°

12°

52°

52°

44°

25°

24°

AC C

15 min

EP

Table 1. Angle θ between the c-axis of Cu6Sn5 and the TD direction

30 min

17°

17°

18°

40°

41°

60 min

28°

10°

26°

26°

15°

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 3 EBSD images of cross-sectional planar inverse polar figures under the conditions of 260°C, 2.0×102 A/cm2 for (a) 15 min, (b) 30 min, (c) 60 min, and (d)

TE D

360 min; polar figure of Cu6Sn5: (e) 15 min, (f) 30 min, and (g) 60 min; polar figure of Cu3Sn: (h) 360 min.

3.3. Confirmation of the lowest-resistance path in the Cu6Sn5 phase To explain the relationship between the preferred orientation of Cu6Sn5 and the

EP

electric current, the concept of “projection atomic density” is introduced, which refers to the projected planar atomic density of all atoms involved in a single compound unit

AC C

cell on certain plane. All visible atoms projected from a certain direction will be counted. Thus, the projection plane atomic density ρ projection can be expressed as

ρ projection =

Nvisible S projection

(2)

where N visible is the number of atoms visible on a certain projection plane in the unit cell and S projection is the area of the projection plane. Because of the electrical anisotropy of the Cu6Sn5 compound, the electrical conductivity of Cu6Sn5 varies among different crystal orientations. The conductivity

ACCEPTED MANUSCRIPT (σ) is the product of the carrier mobility (µ), electron charge (e) and the carrier concentration (n), as expressed by

σ = neµ

(3)

As the carrier concentration of a certain metal is nearly constant, the carrier mobility

RI PT

is the dominant factor influencing the electrical conductivity. Actually, the carrier mobility is inversely proportional the electrical resistance, and the resistance is attributed mostly to the scattering of the electrons by the crystal lattices. When electrons travelled through a plane with the lowest projection atomic density, they

SC

will encounter the lowest scattering by crystal lattices, and the direction perpendicular to such plane corresponds to the lowest resistance path.

M AN U

According to the aforementioned EBSD analysis, the (0001) plane of Cu6Sn5 was expected to be the primary plane for electrons to travel through, which should have the lowest projection atomic density. Calculations were performed to determine the lowest projection atomic density in the unit cell of hexagonal η-Cu6Sn5. The unit cell was established using molecular dynamic simulations with the structure containing 8

TE D

primitive cells, as shown in Fig. 4. The unit cell contains 61 Cu atoms (yellow) and 16 Sn atoms (purple) and the lattice parameters are a=b=2×4.218 Å, c=2×5.490 Å. The radii difference between Sn and Cu atomic was neglected to simplify the calculations.

EP

A Cartesian coordinate system was established with Z axis parallel to c axis, and Y axis parallel to b axis. Fig. 5 showed the atomic distributions of vertical projections

AC C

for an η-Cu6Sn5 unit cell rotated with different angles around the Y-axis. Fig. 5 (a) shows the vertical projection of the atoms on (0001) plane which contains 13 Cu atoms and 4 Sn atoms. Using equation (2), the atomic density was calculated to be 27.9 atoms/nm2. When the unit cell was rotated around Y axis away from (0001) plane by 45°, the atoms hidden behind gradually appeared, as shown in Fig. 5 (b). The projection atomic density increased significantly. Rotate by another 45° to (1010) plane, the ρ projection reached 43.4 atoms/nm2, as shown in Fig. 5 (c). When the rotation was maintained at 135°, the projection atomic density increased to the same value as that obtained at 45° (Fig. 5 (d)). Clearly, (0001) plane showed the lowest

ACCEPTED MANUSCRIPT projected atomic density during the rotation around Y axis. The projection atomic density change for η-Cu6Sn5 unit cell rotating around Z axis was then considered. The atomic distribution of the vertical projections for (1010) plane, which had a total of 55 visible atoms and an atomic density of 43.4 atoms/nm2, was chosen as the initial

RI PT

projection plane. When the unit cell was rotated around Z axis, the number of visible atoms firstly increased and then decreased until rotating into (2110) plane (as shown in Fig. 6). The atomic density of (2110) plane was 42.4 atoms/nm2.

SC

The aforementioned analysis did not cover all of the orientations because the other orientations were unlikely to show a projection plane atomic density lower than that

M AN U

of (0001) plane. Because the hexagonal system has only one symmetry axis, and more atoms will be eclipsed projected along the c-axis. Hence, (0001) plane showed the lowest projected atomic density and the electrons flowing through this plane would have the least scattering. Therefore, it was concluded that Cu6Sn5 grains with lower θ

AC C

EP

TE D

values exhibited lower electric resistances.

Fig. 4 Unit cell of hexagonal η-Cu6Sn5.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 5 Atomic distributions of vertically projected images for a single η-Cu6Sn5

AC C

EP

TE D

unit cell after rotation at different angles around Y axis.

ACCEPTED MANUSCRIPT Fig. 6 Atomic distributions of vertically projected images for a single η-Cu6Sn5 unit cell after rotating at different angles around Z axis.

At the initial reaction stage, newly nucleated Cu6Sn5 grains with different θ values ranging from 0° to 90° formed on the anode Cu pad surface, where their orientations

RI PT

were primarily determined by the texture of the underlying Cu. These small Cu6Sn5 grains then started to coarsen with the soldering time. Simultaneously, newly formed Cu3Sn grains gradually grew into an intermediate layer between Cu and Cu6Sn5. At

SC

this point, the electron flow rather than the Cu pad texture would influenced the growth of Cu6Sn5 grains. According to equation (3), the electrical resistance of

M AN U

Cu6Sn5 grains could be influenced by their orientations with electric current. And the EBSD analysis results above have demonstrated that Cu6Sn5 grains with lower resistance were those with lower θ values. These grains would experience smaller electron wind force and more Cu atom flux under electric current stressing, and thus grew faster than those with higher resistances. Therefore, the grains with high θ

TE D

values gradually shrank and then disappeared during the ripening process. Finally, the Cu6Sn5 grains with lower θ values gained an advantage in volume and transformed into larger grains in the later solid-state reactions. Therefore, it was concluded that the electric current played an important role of orienting Cu6Sn5 lattices along the

EP

low-resistance path.

3.4. Microhardness and shear strength of full-Cu6Sn5 joints

AC C

Fig. 7 (a) and (b) shows the nanoindentation on Cu6Sn5 grains and Fig. 7 (c) and (d)

shows the enlarged images of the indentations. Eighteen indentation experiments were performed on the Cu6Sn5 cross-sectional surfaces and the average values of Young’s modulus and microhardness were 120.337 ± 3.789 and 5.608 ± 0.313 GPa, respectively. Twelve indentations were performed on the Cu3Sn cross-sectional surface and the average values of Young’s modulus and microhardness were 139.907 ± 3.693 and 5.321 ± 0.675 GPa, respectively. Fig. 7 (e) showed the plots of Young’s modulus and microhardness of Cu6Sn5 as functions of θ value. The measured Young’s modulus increased with θ value, whereas the microhardness showed less dependence

ACCEPTED MANUSCRIPT on θ value. Typically, the crystal plane with lower atomic density has smaller interplanar crystal spacing (which means the orientation perpendicular to it is densely packed by atoms) and therefore has larger Young’s modulus, which well agreed with the nanoindentation results. Therefore, the (0001) plane had the largest Young’s

RI PT

modulus, which can reduce the Young’s modulus mismatch between Cu3Sn and Cu6Sn5, whereas the Young’s modulus in the shear directions are smaller, which can reduce the critical resolved shear stress under shear force. In short, the texture formed in full-Cu6Sn5 joints under electric current stressing could achieve better mechanical

SC

properties.

The shear tests were performed to further evaluate the shear strength of the

M AN U

full-Cu6Sn5 joints formed at 120min under different levels of electric current densities (0, 1.0×102 A/cm2, and 2.0×102 A/cm2). Fig. 8 (a) shows that the shear strength increased with electric density. The joints bonded under 2.0×102 A/cm2 exhibited the highest shear strength of 44.87 MPa. Fig. 8 (b-d) presents the top-view fracture images of the joints. In Fig. 8 (b) and (c), fractures tended to propagate along the

TE D

phase interfaces between Cu6Sn5 and Cu3Sn. By contrast, failures mainly occurred inside the Cu6Sn5 phase in Fig. 8 (d). In the case without or with small electric current, Cu6Sn5 at both sides could not combine tightly and there were many defects and weak

EP

neckings around the Cu6Sn5-Cu3Sn interfaces, which may induce extra stress concentration. Besides, it can be seen from Fig. 2 that there were some Kirkendall voids at the Cu/Cu3Sn and Cu6Sn5/Cu3Sn interfaces, which can influence the

AC C

electromigration lifetime of solder joints [24, 25]. These voids may also induce stress concentration under shear force and lead to the fracture. However, the larger electric current could enhance the Cu6Sn5 growth and may reduce the defects to achieve higher joint shear strength. As a result, a combination of good electrical properties in the [0001] direction and good mechanical properties in the shear direction were achieved in the obtained full-Cu6Sn5 joint.

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 7 Micrographs of nanoindentation morphologies on the cross-sectional plane of

M AN U

Cu/Sn/Cu joints under the conditions of 260°C and 2.0×102 A/cm2 for 60 min: (a) full view of the joint; (b) planar inverse polar figure; (c)-(d) enlarged view of the indents;

AC C

EP

TE D

(e) curves of Young’s modulus and microhardness vs. θ.

Fig. 8 Shear strength of joints as a function of electric current density, and the top-view SEM images of the corresponding fractures at 120 min under the electric

ACCEPTED MANUSCRIPT density of: (b) 0; (c) 1.0×102 A/cm2; (d) 2.0×102 A/cm2. 4. Conclusions In this work, the effect of the electric current stressing on the Cu6Sn5 and Cu3Sn grain orientation and mechanical properties in the Cu/Sn/Cu joints was investigated,

RI PT

and the full-Cu6Sn5 joints with a combination of good electrical and mechanical properties were achieved. The results are summarized as follows:

(1) An obvious polarity effect on Cu6Sn5 growth rate was under a current density of 2.0×102 A/cm2 at 260 °C, whereas the growth of Cu3Sn was barely influenced by the

SC

electric current. The Cu atoms dissolved in the cathode migrated to the anode under electromigration, leading to the formation of a full-Cu6Sn5 joint in 60 min and a

M AN U

full-Cu3Sn joint in 360 min.

(2) The orientation evolution of the Cu6Sn5 grains with time was examined using EBSD analysis. A strengthening texture of the Cu6Sn5 [0001] direction being parallel to electric current was observed during solid-liquid reaction under electric current. Cu6Sn5 grains with smaller angles between the c axis and electric current grew faster,

TE D

and were more likely to grow into larger grains. The orientation of Cu3Sn was not influenced by the electric current.

(3) The lowest projection atomic density of the Cu6Sn5 unit cell was calculated as

EP

27.9 atoms/nm2, corresponding to the (0001) projection plane. Here, the electric current played the role of selecting the Cu6Sn5 grains with a low-resistance path. The Cu6Sn5 grains whose (0001) planes were perpendicular to the electric current grew

AC C

faster because of smaller electron wind force and greater Cu atom flux. (4) Nanoindentation experiments and shear tests were performed to full-Cu6Sn5

joints. The average Young’s modulus and microhardness of Cu6Sn5 were 120.337 ± 3.789 and 5.608 ± 0.313 GPa, and those of Cu3Sn were 139.907 ± 3.693 and 5.321 ± 0.675 GPa. The Cu6Sn5 texture formed under larger electric current could achieve better mechanical properties for full-Cu6Sn5 joints. The joints formed under 2.0×102 A/cm2 could achieve the highest shear strength of 44.87 MPa. Acknowledgement The authors are grateful for financial support from the National Natural Science

ACCEPTED MANUSCRIPT Foundation of China (Grant No. 51522503) and support from Program for New Century Excellent Talents in University (NCET-13-0175).

Referrences

international interconnect technology conference, 2007, pp. 213-15. [2]

Ritwik

Chatterjee, Murielle

Fayolle, Patrick

Leduc, Scott

RI PT

[1] Pozder S, Chatterjee R, Jain A, Zhihong H, Jones RE, Acosta E: Proceedings of the IEEE 2007

Pozder, Bob

Jones, Eddie

Acosta, Barbara Charlet, Thierry Enot, Michel Heitzmann, Marc Zussy, Antonio Roman, Olivier Maitrejean, Didier

Louis, Nelly

Kernevez, Nicolas

Sillon, Gerard

SC

Louveau, Sylvain

Passemard, Victor Pol, Varughese Mathew, Sam Garcia, Terry Sparks, Zhihong Huang, IEEE 2007

M AN U

international interconnect technology conference, 2007, pp. 81-83.

[3] Morrow PR, Park CM, Ramanathan S, Kobrinsky MJ, Harmes M, IEEE Electron Device Letters, 2006, vol. 27, pp. 335-7.

[4] Ko CT, Chen KN, Microelectron. Reliab. 50 (2010) 481–8.

[5] K. Hara, Y. Kurashima, N. Hashimoto, K. Matsui, Y. Matsuo, I. Miyazawa, T. Kobayashi, Y.

TE D

Yokoyama, M. Fukazawa, IEEE Transactions on Advanced Packaging, 28 (2005) 367-76. [6] H. Huebner, S. Penka, B.Barchmann, M. Eigner, W. Gruber, M. Nobis, S.Janka, G. Kristen, M. Schneegans, Microelectron. Eng. 83 (2006) 2155-62. Hsi-Kuei Cheng,

Yu-Jie Lin,

Chih-Ming Chen,

Kuo-Chio Liu,

Ying-Lang Wang,

EP

[7]

Tzeng-Feng Liu, Metall. Trans. A 47 (2016) 3971-80.

AC C

[8] HSI-KUEI CHENG, YU-JIE LIN, HOU-CHIEN, CHANG, KUO-CHIO LIU, YING-LANG WANG, TZENG-FENG LIU, and CHIH-MING CHEN, Metall. Trans. A 46A (2015) 1834-37. [9] THI-THUY LUU, NILS HOIVIK, KAIYING WANG, KNUTE. AASMUNDTVEIT, and ASTRID-SOFIE B. VARDØY, Metall. Trans. A 46A (2015) 5266-74. [10] Chiu WL, Liu CM, Haung YS, Chen C, Appl. Phys. Lett. 104 (2014) 171902. [11] Hsiao HY, Liu CM, Lin HW, Liu TC, Lu CL, Huang YS, Chen C, Tu K, Sci. 336 (2012) 1007. [12] Huang M, Yang F, Sci. Rep. 4 (2014) 7117. [13] C.Y. Liu, Y. J. Hu, Y. S. Liu, H. W. Tseng, T. S. Huang, C. T. Lu, Y. C. Chuang, S. L. Cheng, Acta Mater. 61 (2013) 5713-19. [14] M.L. Huang, Z.J. Zhang, N. Zhao, F. Yang. J. Alloys Compd. 619 (2015) 667-675.

ACCEPTED MANUSCRIPT [15] Haitao Ma, Anil Kunwar, Junhao Sun, Bingfeng Guo, Haoran Ma: Scr. Mater. 107 (2015) 88. [16] Anil Kunwar, Haoran Ma, Haitao Ma, Bingfeng Guo, Zhixian Meng, Ning Zhao, Mingliang Huang, J. Mater. Sci. Mater Electron 27 (2016) 7699-706. [17] M.L. Huang, Z.J. Zhang, H.T. Ma, L.D. Chen, J. Mater. Sci. Technol. 30 (2014) 1235-42.

RI PT

[18] Laurila T, Vuorinen V, Kivilahti JK, Mater. Sci. Eng., R, 49 (2005) 1-60. [19] Li JF, Agyakwa PA, Johnson CM, Acta Mater. 59 (2011) 1198-211.

[20] Rui Zhang, Yanhong Tian, Chunjin Hang, Baolei Liu, Chunqing Wang, Mater. Lett. 110 (2013) 137-140.

SC

[21] P. Villars, K. Cenzual, Pearsons's Crystal Data - Crystal Structure Data Base for Inorganic Compounds (On CD-ROM), Release, ASM International, Materials Park, Ohio, USA, 2012/2013.

M AN U

[22]Kim HK, Tu KN. Physical Review B 53 (1996) 16027-16034. [23]Tu KN. Solder joint technology, Springer, New York, 2007.

[24] Ding M, Wang G, Chao B, Ho PS, Su P, Uehling T, J Appl Phys. 99 (2006) 094906. [25] Chae SH, Zhang X, Lu KH, Chao HL, Ho PS, Ding M, et al. J, Mater Sci: Mater Electr. 18

AC C

EP

TE D

(2007) 247.

ACCEPTED MANUSCRIPT Table 1. Angle θ between the c-axis of Cu6Sn5 and the TD direction 2

3

4

5

6

7

8

15 min

54°

12°

8°

12°

12°

52°

52°

44°

30 min

17°

17°

18°

40°

41°

25°

24°

60 min

28°

10°

26°

26°

15°

AC C

EP

TE D

M AN U

SC

RI PT

1

ACCEPTED MANUSCRIPT 1. Full-Cu6Sn5 interconnects was fabricated within 60min. 2. The microstructure and orientation evolution of Cu6Sn5 under electric current was analyzed.

RI PT

3. The orientation adjustment mechanism of Cu6Sn5 was studied.

AC C

EP

TE D

M AN U

SC

4. The adjusted full-Cu6Sn5 interconnects exhibited a high reliable mechanical property.

1