Microelectronics Reliability 45 (2005) 657–663 www.elsevier.com/locate/microrel
Intermediate wafer level bonding and interface behavior C.T. Pan
a,*
, P.J. Cheng b, M.F. Chen a, C.K. Yen
a
a
b
Department of Mechanical and Electro-Mechanical Engineering, Center for Nanoscience and Nanotechnology, National Sun Yat-Sen University, Kaoshiung 804, Taiwan Department of Electrical Engineering, Nan Jeon Institute of Technology No. 178, Chaocin Rd., Yanshuei Township, Tainan County 737, Taiwan Received 17 May 2004; received in revised form 21 October 2004 Available online 30 December 2004
Abstract The paper presents a new silicon wafer bonding technique. The high-resolution bonding pad is defined through photolithography process. Photosensitive materials with patternable characteristics are served as the adhesive intermediate bonding layer between the silicon wafers. Several types of photosensitive materials such as SU-8 (negative photoresist), AZ-4620 (positive photoresist), SP341 (polyimide), JSR (negative photoresist) and BCB (benzocylbutene) are tested and characterized for their bonding strength. An infrared (IR) imaging system is established to examine the bonding results. The results indicate that SU-8 is the best bonding material with a bonding strength up to 213 kg/cm2 (20.6 MPa) at bonding temperature less than 90 °C. The resolution of bonding pad of 10 lm can be achieved. The developed low temperature bonding technique is particularly suitable for the integration of microstructures and microelectronics involved in MEMS and VLSI packaging processes. Ó 2004 Published by Elsevier Ltd.
1. Introduction Micro-fabrication technique has been applied widely to the micro-component manufacturing and other key optics and communications process elements. As a result, the fabrications of MEMS, sensors, actuators, and VLSI can be extended to high density, multilayer 3-D structures [1,2]. Silicon wafer bonding technology has become a critical technique in micro-component manufacturing. Normally, bonding technology can be divided into two major branches, i.e., intermediate layer bonding and wafer direct bonding. Fusion bonding tech-
* Corresponding author. Tel.: +886 7 5252000x4239; fax: +886 7 5254299. E-mail address:
[email protected] (C.T. Pan).
0026-2714/$ - see front matter Ó 2004 Published by Elsevier Ltd. doi:10.1016/j.microrel.2004.10.019
nology with layer material is vulnerable to the processing temperature. Deposition, doping, diffusion, and evaporation will affect the bonded material due to the difference in the coefficient of thermal expansion [3,4]. Thus, thermal stress can be easily induced, which results in substrate bending or cracks after bonding wafer. Fusion bonding technology also affects the yield rate of component fabrication. Even though anodic bonding technology, using a non-interface method to bond wafers, can reduce the temperature effect on the microcomponents, the high electric field generated by the high voltage could seriously damage the electrical circuits on micro-component, especially in VLSI systems [5]. The bonding strength in direct wafer bonding is produced by the strength of the O–H bond on the surface of the hydrophilic wafer. Thus, both wafer surface cleanliness and roughness play a critical role on the bonding
658
C.T. Pan et al. / Microelectronics Reliability 45 (2005) 657–663
strength [6–9]. The major reason for low yield rate is the surface roughness. Silicon wafer etching, doping, sputter, and other processes make the wafer surfaces too rough [8–11]. Organic materials as the bonding materials were presented by Besten et al. [12] in 1992 and Arquint et al. [13] in 1995. They discussed about the phenomena of voids after bonding process. Their studies did not focus on how to define a high-resolution bonding pad, and how to test bonding strength, etc. SU-8 (photoresist) was used to pattern micro-channel and to bond a thin glass wafer for packaging micro-fluidic components [14–17]. Besides, photosensitive materials such as negative photoresist (ULTRA-I 300) from Dow Chemical and positive photoresist (S1818) from Shipley were investigated [18–21]. Park et al. used B-stage epoxy for wafer level packaging of RF-MEMS device with tensile strength over 20 MPa [22]. In this study, a new bonding method using photosensitive material as adhesive intermediate layer was developed. Through photolithography process, the resolution of bonding pad can be enhanced significantly. The method can reduce the residual stress during the packaging process. The effects of bonding strength in terms of the different bonding materials, bonding force, and bonding temperature are discussed in detail. This new bonding technique using photosensitive material has potential for solving some bonding problems of the applications to VLSI and MEMS devices.
Table 1 Properties of the photosensitive material
2. Experimental procedure
from Microchem Co.), BCB (from Dow Chemical), and SP-341 (a positive photoresist from Toray Co.) were selected for the bonding experiments. The detailed material properties and the chemical formula of the photosensitive materials are shown in Tables 1 and 2, respectively. A 4-in. h1 0 0i single-side polished silicon wafer with 525 ± 25 lm in thickness was used. The bonding process is shown schematically in Fig. 1. To avoid particle adherence to the silicon wafer surface, the silicon wafer was cleaned through the standard RCA process and dried with the N2 gas. After that, the photosensitive materials were spin-coated onto the surface of the silicon wafer. The film thickness of the photosensitive materials used is listed in Table 3. To evaporate the solvent con-
There are several important processes involved in the bonding technique. First, the adhesive material layer was spin-coated onto the silicon wafer. Then photolithography process was applied to pattern the bonding pad. After the two silicon wafers were aligned pad to pad, the wafers were placed into a bonding chamber. The intermediate layer needs to have low glass transition temperature (Tg), and high viscosity. Besides, after bonding process was finished, excellent adhesive strength with a low residual stress is also required. Therefore, AZ-4620 (a positive photoresist from Shipley), JSR-137N (a negative photoresist from Japan Synthesis Rubber Co.), SU-8 (a negative photoresist
Characteristics
Value
Benezocy-clobutene (BCB) Modulus of elasticity (E) Poisson ration Polymer shrinkage Coefficient of thermal
2.9 GPa 0.34 – 52 ppm/°C
SU-8 Modulus of elasticity (E) Poisson ration Polymer shrinkage Coefficient of thermal
4.02 GPa 0.22 75% 52 ppm/°C
AZ-4620 Modulus of elasticity (E) Poisson ration Polymer shrinkage Coefficient of thermal
2.0 GPa 0.25 – 42 ppm/°C
JSR Modulus of elasticity (E) Poisson ration Polymer shrinkage Coefficient of thermal
3.0 GPa 0.22 – 45 ppm/°C
SP-341 Modulus of elasticity (E) Poisson ration Polymer shrinkage Coefficient of thermal
2.9 GPa 0.3 80% 40 ppm/°C
Table 2 Chemical formula of photosensitive materials Polymer
Chemical formula
SU-8 BCB JSR-137N AZ-4620 SP-341
Epoxy Novolac resin, c-Butyrolactone and Triaryl sulfoniumsalt–HSbF6 etc. Vinyl–triacetoxysilane 1-methoxy-2-propanol which is solvent and concentration Methyl-3-methoxypropionate (MMP) Naphthoquinone Diazide derivative, Nobolakresin Derivative and Propylene Glycol Monomethyl Ether Acetate Polyimide and N-methy-2-pyrrolidone (NMP)
C.T. Pan et al. / Microelectronics Reliability 45 (2005) 657–663
Fig. 1. Process flow chart of bonding with flip chip and wafer level alignment: (a) define the bonding pad, (b) wafer level alignment and (c) bonding process.
Table 3 Range of film thickness of the photosensitive used Polymer
Film thickness (lm)
SU-8 BCB JSR-137N AZ-4620 SP-341
20–30 15–25 20–30 10–20 15–20
tained in the photosensitive materials, the material was baked for a period of time determined by the thickness of the materials. The wafer with photosensitive material was then exposed and developed to define the bonding pad, as shown in Fig. 1(a). Two wafers were then aligned pad to pad (see Fig. 1(b)) and placed into a bonding machine under a certain bonding force and temperature. For the bonding process, various levels of bonding temperatures and different bonding materials were tested. The pressure of bonding environment under 10E4 mbar in the bonder was created for all experiments. The bonding tool is a commercially available
659
EV501 (Electronic Visions) bonder with a PC controller and chuck. The bonder allows wafers to contact and to be annealed between the hot planes. The vacuum environment is provided in the bonding machine. After creating a vacuum environment in the bonding chamber, two wafers can be bonded together. Several methods of non-destructive and destructive test exist for mechanical characterization of the bonding results. The most common techniques are bond imaging, cross-sectional analysis, and bond-strength measurement. The bond imaging method is non-destructive and can be used as in-process monitors, while the cross-sectional analysis and bond strength measurements are destructive and require control wafers for characterization. There exist three dominant methods of bond imaging to inspect bonded pairs including infrared transmission, ultrasonic, and X-ray topography. In the study, infrared transmission was used to inspect the bonding structure. The IR method has the advantage of being simple, fast, and inexpensive. It can be used directly in the clean room to image the bonded wafers. The other two imaging methods offer higher resolution at the expense of speed, cost, and incompatibility with clean room processing [23,24]. Infrared rays and infrared photography was used to check for voids. Any defects or cavities can be examined using microscope through CCD. This method costs less and easily produces the best bonding parameters for various bonding materials. In the tensile strength test, bonded silicon pair is cut into chips with 1 cm2 in area using a dicing saw. Later, the chip is attached to a clamping apparatus in the Material Test Station (MTS) for bonding strength test. The clamping apparatus is pulled apart using a selected pull force and pull speed with the separation time recorded. Software is available to control this machine and the record tensile and strain test data.
3. Results and discussion Photosensitive material was applied as the intermediate layer in this bonding technique. It has several features including low bonding temperature and excellent capability of surface planarization. In this experiment, the single polished silicon wafers were bonded with different photosensitive materials according to the mentioned above bonding procedure. The characteristics of the bonding material, bonding force, and bonding temperature were investigated. Various photosensitive materials and parameters for the bonding process were realized in the experiments. The wafer bonding strength of the intermediate layer was verified in this study. The bonding pad can be patterned on the silicon wafer through photolithography. The silicon wafers can be adhered to each other with
660
C.T. Pan et al. / Microelectronics Reliability 45 (2005) 657–663
excellent bonding strength just at a lower bonding temperature. Besides, the photosensitive material as the adhesive layer is very flexible, which would not cause the residual stress after wafer bonding process. The bonding temperatures of five kinds of photosensitive materials were tested. The curves for the bonding strengths are shown in Figs. 2–5. Each data represent the average of three measurements. Fig. 2 shows the tensile test curves as a function of bonding temperature under a constant bonding force (50 N). From the result, it can be seen that when the bonding temperatures for SU-8, JSR, AZ-4620, BCB, and SP-341 are between 80 and 120 °C, the bond strength reaches their maximum, respectively. When the bonding temperatures are higher than 140 °C, AZ-4620 would be scorched. Besides, when the bonding temperature is higher than 200 °C, SU-8, JSR, BCB, and SP-341 exhibits the same bonding strength as those at 100 °C. It also reveals that SU-8 served as adhesive layer at bonding temperature 90 °C has maximum bonding strength about 213 kg/cm2 (20.6 MPa). SU-8 has many attractive properties as an intermediate adhesive layer described as follows. SU-8 has an epoxy feature with very high bonding strength. It is a negative photoresist and is cross-linked after exposed to UV light. It exhibits excellent chemical resistance after UV exposure. In addition, it just requires very low bonding temperature (about 90 °C) to form excellent bonding strength. On the other hand, for AZ-4620, when the bonding temperature is about 90 °C, the bonding strength reaches to 86 kg/cm2. When a thinner bonding pad is
Fig. 2. Experimental results of bonding strength and bonding temperature for various intermediate adhesive layers.
Fig. 3. Experimental results of bonding strength and bonding force for various intermediate adhesive layers.
Fig. 4. Schematic illustrations of void and gap in the interface of intermediate adhesive layers.
C.T. Pan et al. / Microelectronics Reliability 45 (2005) 657–663
Fig. 5. Experimental results of bonding strength and bonding strain for various intermediate adhesive layers.
required, the AZ-4620 photoresist is a good choice for the bonding process. In addition, when the bonding temperature is about 90 °C, the bonding strength of SP-341, JSR, and BCB would reach to 100, 88 and 92 kg/cm2, respectively. Fig. 3 shows the curves of tensile test under different bonding force. Each data represent the average of three
661
measurements. The bonding temperatures at 90 °C for SU-8, JSR, AZ-4620, BCB, and SP-341 were tested. The result shows that when the bonding force was reduced, the bonding strength became weaker between the two wafers due of voids and gaps created by the total thickness variation (TTV) of wafer surface and by the air trapped in photosensitive materials as schematically illustrated in Fig. 4. It is worth noticing that with a lower bonding force, the bonding strength between the substrate and adhesive layer can not be formed tightly because the bonding force is not strong enough to eliminate the void and gap in the interface. Besides, when larger bonding force is applied, the adjacent microstructure and IC may be destroyed. The selection of bonding force is an important issue in the bonding process. The bonding force is very important factors in wafer bonding technology. Furthermore, how to choose the photosensitive material as intermediate layer is also an important issue. In this paper, the stress versus strain relationship of photosensitive material as intermediate layer was examined, too. The laser displacement sensor is set up in the normal
Fig. 6. IR transmission images of two bonded wafer pairs: (a) successful wafer bonding and (b) failed wafer bonding with void.
Fig. 7. Patternable intermediate layer for selective wafer bonding in low temperature.
662
C.T. Pan et al. / Microelectronics Reliability 45 (2005) 657–663
direction of pull on MTS for the strength test. The measurement range and resolution of laser displacement sensor is ±0.2 mm and 0.04 lm, respectively. The laser
displacement sensor measures the micro-displacement before adhesive layer fractured. The strain of intermediate layer can be obtained from micro-displacement. The stress versus strain relationship of intermediate adhesive layer can be plotted on MTS. Fig. 5 shows the stress versus strain relationship of intermediate adhesive layer. It shows that when the bonding strength reaches their maximum about 80–215 kg/cm2, respectively. The maximum strain of SP-341 is 155E-3, SU-8 is 63E-3, AZ-4620 is 75E-3, BCB is 70E-3, and JSR is 65E-3, respectively. In the stress versus strain diagram, SU-8 exhibits the steepest slope. It means that SU-8 will become more brittle behavior in the interface after bonding, but the intermediate layer using SP-341 will show elastic characteristics after bonding. The choice of intermediate layer can be based on the requirement of bonding process. A simplified set-up of an infrared bond imaging system was established. It consists of an IR source and an IR sensitive camera. A camera has sufficient sensitivity in the near-IR range that it can be used when outfitted with a filter for visible light. The bonded wafer pair is located between the IR source and camera. Any defect in the bond shows up. Examples of the images obtained by the method for two bonded 4-inch silicon wafer pairs are shown in Fig. 6. Fig. 6(a) shows a successful bonding result. On the other hand, as shown in Fig. 6(b), if NewtonÕs Ring appears, it means void or gap existing in the bonding area. This imaging method generally cannot image voids with a dimension less than one quarter of the wavelength of the IR source. Fig. 7 shows bonding pads. The linewidth of pad can be achieved as small as 10 lm which is very suitable for high density VLSI and MEMS device packaging. Fig. 8 shows SEM photograph of SU-8, AZ-4620, SP-341, BCB, and JSR as the intermediate layer. The sandwich structure is silicon, intermediate layer, and silicon in the cross-section profile. The MEMS and IC devices can be placed on the fabricated cavities through the bonding technique. In Fig. 9, a brittle fracture can be
Fig. 8. Cross-section SEM photographs of adhesive materials: (a) SU-8, (b) AZ-4620, (c) SP341, (d) JSR and (e) BCB.
Fig. 9. SEM micrograph of morphology of SU-8 between the bonding interfaces.
C.T. Pan et al. / Microelectronics Reliability 45 (2005) 657–663
observed from the surface morphology of the bonding sample. On the left side of Fig. 9 the brittle fracture of SU-8 can be observed. This result indicates that the SU-8 material attaches to the silicon very well with excellent bonding strength. 4. Conclusion Photosensitive materials as adhesive lays for bonding technology have been successfully demonstrated in this study. It has advantages over conventional wafer bonding technology. This technique takes advantage of photosensitive material characteristics to define the bonding pad served as the bonding adhesive layer between silicon wafers. This bonding method can be applied to bond 3-D microstructures, microsensors, MEMS, and VLSI devices. This bonding technique can reduce the residual stress in packaging micro-components, and can prevent the damaging from high electric fields and high temperature process effect. It can also solve the problem of bonding stress caused by the roughness of the wafer surface. Several types of photosensitive material were tested for bonding strength. The results indicate that SU-8 is the best bonding material with bonding strength up to 213 kg/cm2 (20.6 MPa). This method offers selective area bonding, low temperature, and non-electric field bonding techniques, which is particularly suitable for the integration of microstructure with microelectronics involved in MEMS packaging.
Acknowledgments The author would like to thank Dr. Tung-Chuan Wu at MIRL of ITRI in Taiwan for his guidance, and National Science Council (NSC) for their financial supports to the project (granted number: NSC-NSC93-2622-E110-003-CC3, NSC93-2212-E-110-028 and NSC932212-E-110-029).
References [1] Schmidt MA. Wafer-to-wafer bonding for microstructure formation. Proc IEEE 1998;86(8):1575–80. [2] Berthold A, Sarro PM, French PJ, Vellekoop MN. Silicon fusion for fabrication of sensor, actuators and microstructures. Sens Actuat A 1991;21:919–26. [3] Berthold A, Jakoby B, Vellekoop MJ. Wafer-to-wafer fusion bonding of oxidized silicon to silicon at low temperatures. Sens Actuat A 1998;68:410–3. [4] Xiao ZX, Wu G-Y, Li Z-H, Zhang G-B, Hao Y-L, Wang Y-Y. Silicon–glass wafer bonding with silicon hydrophilic fusion bonding. Sens Actuat A 1999;72:46–8.
663
[5] Quenzer HJ et al. Silicon–silicon anodic-bonding with intermediate glass layers using spin-on glasses. In: IEEE Proc Micro Electro Mechanical Syst (MEMSÕ96), 1996. p. 272–6. [6] Go¨sele U, Schumacher A, Kra¨uter G. Low temperature silicon direct bonding for application in micromechanics: bonding energies combinations of oxides. Sens Actuat A 1998;70:271–5. [7] Shimbo M et al. Silicon-to-silicon direct bonding method. J Appl Phys 1986;60:2987–9. [8] Jiao J, Lu D, Xiong B, Wang W. Low-temperature silicon direct bonding and interface behaviors. Sens Actuat A 1995;50:117–20. [9] Tong QY, Cha G, Grfiteanu R, Go¨sele U. Low temperature wafer direct bonding. J Microelectromech Syst 1994;3:29–35. [10] Wolffenbuttel RF. Low-temperature intermediate Au–Si wafer bonding; eutectic or silicide bond. Sens Actuat A 1997;62:680–6. [11] Quenzer HJ et al. Low-temperature silicon wafer bonding. Sens Actuat A 1992;32:340–4. [12] Besten C et al. Polymer bonding of micro-machined silicon structures. In: IEEE Conference of MEMS92, 1992. p. 104–7. [13] Arquint P et al. Flexible polysilicon interconnection between two substrates for microsystem assembly. In: 8th Int Conf Solid-State Sensors and Actuators (TransducersÕ95), 1995. p. 263–4. [14] Jackman RJ, Floyd TM, Ghodssi R, Schmidt MA, Jensen KF. Microfluidic systems with on-line UV detection fabricated in photo definable exposy. J Micromech Microeng 2001;11:263–9. [15] Wang T-H, Wong PK, Ho C-M. Electrical molecular focusing for laser induced fluorescence based single DNA detection. In: IEEE Conference of MEMS2002, 2002. p. 15–8. [16] Deval J, Tabeling P, Ho C-M. A dielectrophoretic chaotic mixer. In: IEEE Conference of MEMS2002, 2002. p. 36–9. [17] Suzuki H, Ho C-M. A magnetic force driven chaotic micro-mixer. In: IEEE Conference of MEMS2002, 2002. p. 40–3. [18] Shivkumar B, Kim CJ. Microrivets for MEMS packaging: concept, fabrication, and strength testing. J Micromech Microeng 1997;6(3):217–25. [19] Oberhammer J et al. In: Proceeding of MME2001. [20] Nicklaus F, Enoksson P, Ka¨lvesten E, Stemme G. Lowtemperature full wafer adhesive bonding. J Micromech Microeng 2001;11:100–7. [21] Jourdain A, Demoor P, Pamidighantam S, Tilmans AC. In: IEEE Conference of MEMS2002, 2002. p. 667–80. [22] Park Y-K, Park H-W, Lee D-J, Park J-H, Song I-S, Kim C-W, et al. In: IEEE Conference of MEMS2002, 2002. p. 681–4. [23] Maszara W, Jiang B-L, Yamada A, Rozgonyi GA, Baumgart H, de Kock AJR. Role of surface morphology in wafer bonding. J Appl Phys 1991;69(1):257–65. [24] Schmidt MA. Wafer-to-wafer bonding for microstructure formation. Proc IEEE 1998;86(8):1575–85.