Low-temperature and low pressure copper-to-copper

direct bonding enabled by creep on highly (111)-

oriented Cu surfaces

Abstract—Cu-to-Cu microbumps have many advantages, such

as better electrical and thermal conductivities. The Cu-to-Cu

microbumps may be a potential solution to the heat dissipation

problem in 3D IC integration. In addition, for packaging of CMOS

image sensors, the bonding temperature is preferred to below 200

°C. Therefore, low temperature and low pressure Cu-to-Cu direct is

of great interests for packaging industry. However, it is a

challenging issue because Cu atoms diffuse slowly below 200 °C. In

this study, we achieved low temperature and low pressure Cu

bonding using highly (111)-orientated Cu films. Because the

diffusivity of Cu atoms on the (111) surface diffusivity is the fastest

among all the crystallographic planes of Cu. The bonding

temperature can be lowered to 150°C at a compressive stress of 114

psi held for 60 min at 10-3 torr. Since the melting points of popular

Pb-free solders such as SnAg and SnAgCu are over 230°C.

Therefore the present technique can be applied to packaging of

CMOS image sensors and 3D IC integration.

Keywords—Cu-Cu; direct bonding; nanotwin; 3DIC

I. INTRODUCTION

A change from two-dimensional to three dimensional integrated circuits (3DIC) is under way in the microelectronics industry as the limits of very-large-scale integration in silicon chip technology are approached [1]. The eutectic SnPb solders have been widely adopted as interconnect materials for microelectronic devices since 1970[2], because solders have low melting points and solder joint technology is very useful [3-5]. The melting points for the eutectic SnPb solder is 183oC and the reflow temperature is approximately 200oC. With increasing environment concerns, The Pb-free solders have replaced the Pb-containing solders for interconnects. The popular Pb-free solders are Sn-rich such as SnAg, SnAgCu and their melting points are approximately 250oC[6]. Furthermore, during the solid-state aging of the microbumps, the solder will transform into intermetallic compounds (IMC). These compounds will dominate the properties of the microbumps because the solder thickness decreases to about 10 mm or less in microbumps to connect the through-Silicon-vias(TSVs).

When the scaling of large-scale integrated circuits (ICs)

approaches physical limits, three dimensional (3D) ICs have become to be the most potential solution[7-9]. Silicon chips are stacked vertically to shorten the wiring between chips, and TSV and microbumps are adopted for interconnects between chips. The thickness of solder in microbumps decreases to about 5-10 μm[7-13]. However, the volume fraction of IMCs in a microbump increases obviously. The mechanical properties of Ni-Sn and Cu-Sn IMCS are brittle in nature [14,15]. The properties of the microbumps will become a reliability and yield issue. However, the solder may become necking in microbumps during reflow and solder-state aging[16]. Therefore, the solder-based microbumps for 3DIC will produce many new reliability issues.

One of method for the vertical interconnects is to remove solder and to use Cu-to-Cu direct bonding. The Cu-to-Cu microbumps have better electrical and thermal conductivity properties than solder microbumps. Many efforts have been made on the Cu-to-Cu durect bonding. It can be realized in an ultrahigh vacuum at room temperature, provided that the surfaces of the two Cu components are atomically flat, and the surfaces must be cleaned by the special process [17]. Thereby, this method is time wasting and expensive. The Cu direct bonding can also be accomplished by thermal compression bonding at 300oC in ordinary vacuum of 10-3 torr [18-20]. Nevertheless, some of applications will be restricted by using high temperature to bonding.

In our report, we attention a simple process to accomplish

the Cu-to-Cu direct bonding at 200oC for 30 min under 114 psi

(0.78MPa) of compression in ordinary vacuum conditions. Qne

of Cu films in direct bonding must possess a highly (111)-

oriented surface at least, because it has the fastest surface

diffusivity among all the crystallographic planes [21-23]. The

bonding interfaces (grain boundaries)Cross-sectional was

observed by transmission electron microscopy (TEM) and the

results indicate that oxide-free and almost void-free bonding

interfaces have been obtained through this method.

Chih-Han Tseng1,Chien-Min Liu1, Han-wen Lin1, Yi-Cheng Chu1, Chih Chen1*, Dian-Rong Lyu2, Kuan-Neng Chen2, K. N.

Tu3 1 Department of Materials Science and Engineering,

National Chiao Tung University, Hsinchu, Taiwan 30010, Republic of China 2 Department of Electronics Engineering,

National Chiao Tung University, Hsinchu, Taiwan 30010, Republic of China 3 Department of Materials Science and Engineering,

University of California at Los Angeles, Los Angeles, California 90095, USA

ICEP-IAAC 2015 Proceedings

523

FA2-1

Fig. 1 (a)Dark-field TEM image using (11-1) diffraction. The selected area diffraction (SAD) pattern is shown in the bottom-right corner. (b) Bright-field TEM image of two highly (111)-oriented films bonded at 200oC for 1 h.

Fig. 2 (a) Cross-sectional high resolution TEM image of the bonding interface after annealing 200oC for 1 h.(b) TEM EDS line-scan across the entire sample in figure 1. (b)

II. EXPERIMENTAL

In our study, we used two types of Cu films for our

experiment. One Cu film was highly (111)-oriented, and the

other Cu film was randomly oriented. Highly (111)-oriented nt-

Cu film was fabricated by a 100 nm thick Ti adhesion layer and

a 200 nm thick Cu film were sequentially sputtered on a 20 cm

Si wafer. For Cu-Cu direct bonding, the samples were cut into

3 × 3 mm2 pieces. Randomly oriented Cu films were prepared

by sputtering using Ion Tech Microvac 450CB. The samples

were cleaned by ultrasonically in acetone for 5 min, dried with

a N2 purge, cleaned with a mixed solution of HCl and

deionized water at the ratio of 1:1 for 30 s, rinsed with DI water,

and purged with N2 gas again. Next, the two Cu films were

placed face to face and used a screw clamp immediately. The

bonding condition of pressure was measured to be 114 psi

(0.78 MPa) using a mechanical gauge at room temperature. The

samples were loaded into a quartz furnace tube, the pressure

was approximately 1×10-3 torr to avoid the sample surface

being oxidized at high temperature. The furnace was heated to

200oC over the necessary time period for a ramping rate of

1.3°C/sec, and all cooling events were carried out by simply

allowing the furnace to cool naturally. The bonding condition

of temperature was 200oC that we adopted three times for 10

min, 30 min and 1 h respectively. The microstructures of the

bonding interfaces were observed by scanning transmission

electron microscope (STEM). The TEM examinations were

performed at 200 kV with a point-to-point resolution of 0.23

nm and a lattice resolution of 0.14 nm. Electron backscatter

diffraction (EBSD) and X-ray diffraction were used to analyze

the grain orientation and orientations of the Cu films. The

EBSD measurements were performed using a JEOL 7001F

field emission scanning electron microscope with an

EDAX/TSL system operated at 25 kV. OIMTM software was

used to analyze the orientation maps and crystallographic

textures based on Kikuchi patterns. Energy-dispersive x-ray

spectrometer (EDS) analysis was performed on a Link ISIS

300 EDS attached to the JEOL-2100 TEM and was used to

determine the compositional distribution of the local areas of

the samples in the Cu-Cu bonded interfaces.

III. RESULTS AND DISCCUTION

The image can indicate that the bonding interface almost

void-free between the two highly (111)-oriented Cu films can

be achieved at 200oC. Figure 1(a) shows the corresponding

dark-field TEM image using the (11-1) diffraction. The selected area diffraction pattern is shown in the bottom-right

corner, and has a [011] pole. The results indicate that both of

the Cu films are columnar Cu grains with highly -

oriented. Figure 1(b) shows a bright-field cross-sectional TEM image for two highly (111)-oriented Cu films bonded at 200oC

for 1 h.

Figure 2(a) illustrates the high resolution TEM for the bonding interface. No amorphous layer was found and the lattice points were continuous across the bonding interface. Figure 2(b) describes the EDS line scan across the bonded sample, and no obvious oxygen signals was detected at the interface. These results reveal the formation of excellent bonding interfaces or grain boundaries.

The strong bonding interfaces can also be achieved when the bonding time was shortened to 30 min. Figure 3(a) presents the cross-sectional bright-field TEM image for the bonded

Identify applicable sponsor/s here. If no sponsors, delete this text box (sponsors).

ICEP-IAAC 2015 Proceedings

524

FA2-1

Fig. 3 (a) Bright-field TEM image of two highly (111)-oriented films bonded at 200oC for 30 min. (b) Bright-field TEM image of two highly (111)-oriented films bonded at 200oC for 10 min.

Fig. 4 The schematic of bonding process.

Fig. 5 (a) Plan-view EBSD orientation image map of the highly (111)-oriented film. (b) X-ray diffraction pattern for the highly (111)-oriented film.

(111)-oriented Cu films at 200oC for 30 min. Almost no voids can be found at the smooth bonding interface. But, many voids may be found at the boning interface when the bonding time was further decreased to 10 min. Figure 3(b) shows the bonded (111)-oriented Cu films at 200oC for 10 min, and many voids existed in the interface after the 10 min Tensile tests were performed on 2×2 cm specimens bonded at 200oC for 1 h, and the bonded interface possesses high strength values. Three of the specimens that exhibited strong bonding quality survived tensile testing in excess of 175 kg. Therefore, the bonding interfaces have excellent bonding strength. [25, 26].

The bonding temperature can be lowered to 200oC or 150oC which is lower than that in using solders to microbump. The fast diffusion of Cu atoms on the Cu (111) surfaces is the critical reason for the breakthrough. The bonding process was caused and achieved by Nabarro-Herring (or Coble) creep [24], because the two Si chips and Cu films were under compressive stress. Figure 4.is the schematic of bonding process. The Cu at the contact regions was under a high compressive stress at the initial stage, while the Cu at the un-contact regions should be less or no compressive stress. The calculated theoretical diffusivity is 9.42*10-6 cm2/sec, 1.19*10-9 cm2/sec and 5.98*10-11 cm2/sec for(111), (100) and (110) plane,

respectively. The much faster surface diffusion on (111)-oriented grains will facility the Cu-to-Cu direct bonding. Figure 5(a) presents the plan-view EBSD orientation image map, and the inverse pole figure is illustrated in the bottom-right corner. The results indicate that 97% of the surface grains are (111)-oriented and the grain size is 195 nm on the average. Figure 5(b) shows the XRD, and the intensity for (111) plane appears very high, yet the (200) peak is invisible in the pattern. Therefore, the Cu film has a very strong (111) preferred orientation.

In addition, the experiment can also be achieved between a randomly-oriented Cu film and a (111)-oriented Cu film. Figure 6(a) presents the plan-view TEM image for the randomly-oriented Cu film and the selected area diffraction pattern was shown in the bottom-left corner. The image indicates the randomly-oriented Cu film is polycrystalline. Figure 6(b) illustrated the X-ray diffraction, in which the Cu (200) is visible. The results indicate the grains were randomly

ICEP-IAAC 2015 Proceedings

525

FA2-1

Fig. 6(a) Plan-view TEM image map of the randomly oriented Cu film. The SAD diffraction was shown in the bottom-left corner. (b) X-ray diffraction pattern for the randomly-oriented Cu film. (c) Bright-field TEM image of a highly (111)-oriented film bonded to a randomly-oriented Cu film on the top at 200oC for 30 min (d) Corresponding dark-field TEM image for the sample in (a) using (11-1) diffraction of the highly (111)-oriented film.

oriented. Figure 6(c) presents the cross-sectional bright-field TEM image for a randomly-oriented Cu film (top) bonded to a highly (111)-oriented Cu film (bottom) at 200oC for 30 min. The bonding interface is excellent, but with more voids than the results in Figure 1.It can be attributed to the fact that only one sample of the Cu films is (111)-oriented and the other Cu film is randomly-oriented. Figure 6(d) shows the corresponding dark-field TEM image using (11-1) diffraction. The grains on the top film become invisible because they were not (111)-oriented. The initial grain size on the randomly-oriented Cu film is about 20 nm, so there are many grain boundaries. Grain growth may occur inside the film first and thus many large grains were observed in the film after the bonding by annealing temperature was 200oC. In addition, the bonding interface moved toward to the (111)-oriented film, which may be attributed to ripening effect.

IV. CONCLUSION

In summary, we accomplish low-temperature bonding using

highly (111)-oriented Cu films. With the fastest Cu diffusion

on (111) planes, the bonding conditions can be lowered even

to 200oC for 30 min. Almost void-free and oxide free interface

can be accomplished by bonding two highly (111)-oriented Cu

films. In addition, excellent interface with few voids can also

be achieved between a highly (111)-oriented Cu film and a

randomly oriented Cu film. This method provide a potential

solution with a enormous impact to the interconnect materials

in 3D IC. Financial support from the National Science Council,

Taiwan, under contract NSC 99-2221-E-009-040-MY3, is acknowledged. The authors thank the Center for Micro/Nano Science and Technology (CMNST) at National Cheng Kung University for assistance with the analytical equipment.

References

[1] J. C. Lin et al., paper presented at IEEE International

Electron Devices Meeting, San Francisco, CA, 6 December 2010.

[2] L. F. Miller, IBM J. Res. Develop. 13(3) (1969) 239. [3] K. N. Tu, Solder joint technology, Springer, New York, NY

(2007).

[4] N. Chawla, Inter. Mater. Rev. 54 (2009) 368. [5] C. K. Chung, J. G. Duh, C. R. Kao, Scripta Mater. 63 (2010)

258.

[6] K. Zeng, K. N. Tu, Mater. Sci. Eng. R 38 (2002) 55. [7] R. S. Patti, Proc. IEEE 94 (2006) 1214.

[8] M. T. Fukushima, T. Tanaka, Proc. IEEE 97 (2009) 49.

[9] J.-Q. Lu, Proc. IEEE 97 (2009) 18. [10] H. Y. Hsiao, C. M. Liu, H.W. Lin, T. C. Liu, C. L. Lu, , Y. S.

Huang, C. Chen, K. N. Tu, Science 336 (2012) 1007.

[11] K. N. Tu, Microelectron, Reliab. 51 (2011) 517. [12] T. C. Liu, C. M. Liu, Y. S. Huang, C. Chen, K. N. Tu, Scripta

Mater. 68 (2013) 241.

[13] Y. S. Huang, H. Y. Hsiao, C. Chen, K. N. Tu, Scripta Mater. 66 (2012) 741.

[14] K.E. Yazzie, H.X. Xie, J. J. Williams, N. Chawla, Scripta

Mater. 66 (2012) 586. [15] K. E. Yazzie, H. E. Fei, H. Jiang, N. Chawla, Acta Mater. 60

(2012) 4336.

[16] Y. C. Liang, H. W. Lin, H. P,. Chen, C. Chen, K. N. Tu, Y. S. Lai, Scripta Mater. 69 (2013) 25.

[17] T. H. Kim, M. M. R.Howlader, T. Itoh, T. Suga, J. Vac. Sci.

Technol. A 21 (2003) 449. [18] C. S. Tan, L. Peng, J. Fan, H. Li, S. Gao, IEEE Trans. Device

Mater. Relia. 12 (2012) 194.

[19] J. Lee, D. M. Fernandez, M. Paing, Y. C. Yeo, S. Gao, IEEE Trans. Compon., Packag. Manuf. Technol. 2 (2012) 964.

[20] K. N. Chen, S. H. Lee, Paul S. Andry, Cornelia K. Tsang, Anna

W. Topol, Y. M. Lin, J. Q. Lu, Albert M. Young, M. Ieong, W. Haensch, 2006 International Electron Devices Meeting (IEDM), San

Francisco CA, December 11-13, 2006.

[21] P. M. Agrawal, B. M. Rice, D. L. Thompson, Surf. Sci. 515 (2003) 21.

[22] D. B. Butrymowicz, , J. R. Manning, M. E. Read, J. Phys.

Chem. Ref. Data 2 (1973) 643. [23] C. N. Liao, Y. C. Lu, D. Xu, J. Electrochem. Soci. 160 (2013)

D207.

[24] F. A. Mohamed, K. L.Murty, J. W. Jr. Morris, Mater. Trans. 4

(1973) 935.

[25] K. N. Chen, S. M. Chang, L. C. Shen, and R. Reif, J. Electron. Mater. 35(2006) 1082.

[26] K. N. Chen, T. M. Shaw, C. Cabral, Jr., and G. Zuo, 2010

International Electron Devices Meeting (IEDM), San Francisco CA, December 6-8, 2010.

ICEP-IAAC 2015 Proceedings

526

FA2-1