fabrication technique of a compressible ... - KULeuven - ESAT

0 downloads 0 Views 1MB Size Report
2 ESAT-Micas , KULeuven, Kasteelpark Arenberg 10, 3001 Heverlee, Belgium. 3EECS .... ing sodium aurosulfite Na3Au(SO3)2 and additional compounds to ...
FABRICATION TECHNIQUE OF A COMPRESSIBLE BIOCOMPATIBLE INTERCONNECT USING A THIN FILM TRANSFER PROCESS A.A.A. Aarts1,2,3, O. Srivannavit3, K.D. Wise3, E. Yoon3, H.P. Neves1, R. Puers1,2 and C. Van Hoof1,2 1

Technology Unit, IMEC, Kapeldreef 75, 3001 Leuven, Belgium ESAT-Micas , KULeuven, Kasteelpark Arenberg 10, 3001 Heverlee, Belgium 3 EECS, University of Michigan, 1301 Beal Avenue Ann Arbor, MI 48109-2122, USA 2

Abstract — A compressible multifunctional interconnect for out-of-plane MEMS structures has been fabricated using a thin film transfer bonding technique and bio-tolerable materials. The bulk material of the compressible film consists of photo patternable poly-dimethylsiloxane (PDMS) and is fabricated on a carrier substrate. The film is bonded to a slimbase platform. The carrier substrate of the thin film is released using an aluminum anodic dissolution technique. Probe arrays can be assembled perpendicular into a slim base platform. Ones the probes are assembled a non-separatable electrical connection is made. This interconnect can also facilitate fluidic probes for direct drug delivery applications. Keywords : Stretchable interconnect, Transfer bonding, Anodic dissolution, Photo sensitive PDMS I - Introduction During the past few years there has been a lot of activities in the fabrication of probe arrays. Very successful probe arrays have been demonstrated [1]. However with increasing functionality as CMOS integration (multiplexers, filters, amplifiers, etc.) the base of the probe array becomes taller. Since the available space on the shafts of the probe array is very limited, most of the active components are positioned at the base of the probe arrays. Implantation of such probe arrays requires extra space beyond the implanted area which is not always available. As a result, the probe array can get stuck between the skull and cortical tissue resulting in tissue damage or probe failure. We have demonstrated a slim-base 3D integration technique that can be used in space-limited sites while maintaining the CMOS functionality [2]. The interconnect consists of a gold clip which leads over the edge of the cavity. However, the interconnect has no elastic behavior nor spring effect. This requires a very accurate thickness control of the assembled structures with a very tight thickness specification. Before assembly, the structures often need to be planarised by grinding or polishing techniques to ensure a proper fit. This paper describes an improved out-of-plane interconnect with stretchable, i.e. compressible, behavior and using significant less process steps. The interconnect consists of a stretchable layer containing several gold clips which are leading over the edge of a cavity, see Figure 1. The cavity acts as a socket for the assem-

bled probe arrays. Several different layouts have been designed and fabricated for the individual gold clips. During the assembly of the probe array, the stretchable interconnect will bend into the cavity and is squeezed against the contact pad of the probe array. The thus compressed layer leading over the edge of the cavity provides a spring effect that presses the overhanging gold contact against the contact pad of the assembled structure. The overhanging compressible contacts enable the assembly of structures containing higher total thickness variation (TTV). The compressible or stretchable film consists of a poly-dimethylsiloxane (PDMS) layer, which can also be used as bonding layer. PDMS is often used in microfluidics and biomedical applications. When assembling fluidic devices the overhanging PDMS layer can be used as sealing ring around the socket. PDMS patterns and structures are often created by casting/molding- or etching techniques. Patterning PDMS using a reactive ion etch (RIE) technique results in underetching of the PDMS layer. This work uses a very low viscosity photosensitive PDMS called WL5150 from DOW Corning. This is a commercially available product [3]. The process flow of the interconnect uses a thin film transfer bonding technique followed by an anodic release. The transferred thin film is fabricated on a temporary carrier substrate. Eventually the thin film is bonded to the slim-base platform and the carrier is released by aluminum anodic dissolution, separating the carrier substrate from the bonded thin film. Test structures have been assembled into the platform to gain information about the contact resistance.

Figure 1: Schematic view of the compressible interconnects showing the gold contact supported by the surrounded PDMS. Probe arrays can be assembled perpendicular into the platform.

II - Fabrication The process flow of the compressible interconnect is shown in Figure 2. As explained in the introduction the

thin film containing the gold metallization is fabricated on a carrier wafer while the platform, containing the cavities, is fabricated on another wafer. A die-to-wafer transfer bonding technique is used to bond both structures together. This section is divided in two subsections. Section A describes the fabrication of the thin film, illustrated by Figure 2a-d. Section B describes the thin film transfer bonding technique, illustrated by Figure 2e-g. A. Thin Film Fabrication The carrier substrate is a standard 4 inch silicon wafer covered with a 200nm thick thermal oxide layer. The backside oxide prevents any gold deposition at the backside of the wafer during the gold electroplating process.

Figure 2: Process flow of the stretchable interconnects.

On top of the silicon oxide layer a Ti/Al/Cr/Au metal layer is deposited using evaporation. The aluminum layer functions as a sacrificial layer during the release process and has a thickness of 500nm. The titanium layer is used as electrical conductive layer during the anodic release process. The gold layer functions as seed layer during the gold electroplating process. The chrome layer in between acts as diffusion barrier preventing the gold to diffuse into the aluminum layer. As shown in Figure 2a, a resist layer is spin coated on top of the gold seed layer and defines the position of the

gold electroplated structures. During the gold electroplating only the exposed gold seed layer will be covered with electroplated gold, see Figure 2b. The gold electroplating rate and morphology is mainly controlled by the current density so it is important to know the total exposed surface area during electroplating. Too high current densities can lead to rough plated structures while too low current densities can result in inhomogeneous structures. The growth rate of the electroplated bath is often dependant of the electro plating setup. The gold plating process is performed in a cyanide free electroplating bath, see Figure 2c. The bath consists of a mixture of BDT-510 plating solution and BDT Brightener, which is commercially available (Enthone-OMI). The electrolyte consists of a gold sulfite solution holding sodium aurosulfite Na3Au(SO3)2 and additional compounds to maintain the conductivity and pH of the bath at a constant level [4]. The salt, sodium aurosulfite, is already dissolved in the plating solution and the aurosulfite ion dissociates in a gold ion (1) which forms solid gold (2) at the surface of the sample which is electrically connected to the cathode of the current source. [Au(SO3)2]3- → Au+ + 2SO32(1) Au+ + e- Au(s) (2) Different plated structures with thickness of 1 µm up to 5µm have been fabricated. Current densities of 1.5 to 2 mA/cm2 have been used. Next step is to remove the plating mask with acetone and isopropanol followed by a DI rinse. The gold metallization layer will be encapsulated with PDMS. PDMS has a bad adhesion to gold. Prior spin coating of the PDMS a chrome adhesion layer is deposited by evaporation deposition followed by an oxygen plasma. The O2 plasma improves the adhesion of the PDMS and the chrome layer. A 15 µm thick layer of WL5150 is spin coated on top of the chromium layer followed by a soft bake step using a hotplate. During the soft bake all residual solvents are removed however, the PDMS film remains tacky. Spacers were used during alignment and exposure to prevent stiction between the PDMS and mask. UV exposure of WL5150 causes activation of the photosensitive compound. During the postexposure bake a selective cross-linking process takes place. The non-exposed areas are removed during the development step using a standard negative resist developer like SU-8 Developer. A post exposure bake is required to initiate the cross linking of the PDMS. Without cross-linking the whole film will be removed during the development step. The recommended post exposure bake is at 150°C for 120 seconds. Longer bake times will increase the development time significantly and can leave residues at the bottom corner of the nonexposed areas [5]. A hard bake is used to complete the film curing process. During film curing there is no formation of cure by-products, see Figure 2d. The hard bake is done in a convection oven at 180°C for one hour. After the hard bake a RIE can be used to remove some silicone residues of the developed regions. This plasma etch requires an oxygen/fluorine gas mixture [3]. High plasma densities can damage and crack the

PDMS film. Different interconnect layouts have been fabricated, see Figure 3.

Figure 3: Different PDMS structures containing gold metallization with pitch of 70µm (left and center) and 35µm.

B. Thin Film Transfer Bonding The edges of the patterned PDMS layer are taller compared to the rest of the film resulting in a local thicker edge around the photo definable structures. The local thickening of the layer (dimple) is mainly caused due to internal stresses in the film [6]. The dimple can be reduced by optimizing the post exposure bake and decreasing the proximity gap during UV exposure [5]. The dimple can cause problems when using too low bonding forces. The dimples can be compressed easily when increasing the bonding temperature. As shown in Figure 2e, the platform is bonded to the carrier substrate holding the thin metallization film. The front side of the platform is covered with a silicon oxide layer which has a good adhesion with the PDMS layer. Prior bonding an oxygen plasma is used to activate the PDMS surface and to make the silicon oxide surface hydrophilic. The PDMS layer consists of repeated groups of -OSi(CH3)2- which forms a silanol group (Si-OH) in place of one methyl group (-CH3) when treated by an oxygen plasma [7]. Silanol groups are polar which results in a hydrophilic surface. Bringing the oxygen treated PDMS in contact with the silicon oxide surface results in a strong Si-O-Si covalent bond. Too long exposure time of the PDMS in the oxygen plasma as well as too high plasma densities degrades the bonding properties [7]. The alignment and bonding was done using a flip-chip bonder.

Figure 4: Photograph of bonded structures to the thin film and Schematic view of the alignment and bonding step (upper left).

The release process, see Figure 2f, is done using anodic dissolution. Anodic dissolution of a sacrificial metal can be used to release (complete or partial) thin film structures or other microstructures [8]. It is an electrochemical etch technique which results in a higher etch rate compared to conventional chemical wet etch techniques. Aluminum is used as sacrificial metal and can be selectively (each metal has a different polarization

potential) removed using a neutral sodium chloride solution and applying a small positive potential to the metal layer. The titanium conductive layer under the aluminum is electrically connected to the anode of a voltage source. The cathode is electrically connected to a stainless steel plate. As electrolyte a 2 molar sodium chloride (NaCl) solution has been used. The aluminum dissolution is done in a beaker setup. The aluminum layer starts to dissolve at an applied potential of 0.5V. During aluminum dissolution several electrochemical reactions occur [8][9][10]. In addition to the aluminum oxide formation and aluminum reduction: Al → Al3+ + 3e(1) + 2Al + 3H2O → Al2O3 + 6H + 6e (2) Al + 3Cl- → AlCl3 + 3e(3) Aluminum corrodes in water which involves oxidation (1) and reduction. During oxidation the metal loses electrons and dissociates into a free electron and ion resulting in an anodic current. This current will be transported from the metal to the electrolyte solution and is mainly driven by diffusion, potential gradient and convection. This reaction is normally balanced by the reduction reaction where ions capture the free electrons resulting in a cathodic current. Oxidation and reduction happen at distinct sites of the metal surface and without connecting the metal to an external power source these reactions are in equilibrium. Reaction (2) describes the oxide film formation where a spontaneous oxide layer covers the aluminum. The actual dissolution of aluminum under applied potential starts with pitting where ions (often chloride ions) penetrate the oxide film and start to oxidize the aluminum. Aluminum dissolution in a neutral sodium chloride electrolyte involves several species like Al3+, Cl-, Na+, H2O, OH- and H+. The rate of aluminum dissolution involves the depletion layer of the electrolyte species and the solubility of the anodic reaction products (the salt). The dissolved metal near the anode reacts with the anodic reaction products that have to dissolve in the electrolyte. The mass transport is limited by the salt transport mechanism (3). AlCl3 is the salt that dissolves into the electrolyte. The mass transfer mechanism of the anodic reaction products influences the dissolution rate. The dissolution rate decreases in time because of the increasing underetch paths. The etch rate of the aluminum is about 16 µm per minute. The complete release process of the film takes about 2.5 hours. To speed up the release time and to decrease the underetch path we designed supply channels for the bulk electrolyte in the PDMS layer. When the carrier substrate is removed the transferred film is flipped with the gold seed layer and chromium adhesion layer facing up, see Figure 2g. The gold seed is removed using a wet etch, based on a potassium io-

dide and iodine chemistry (Gold Etchant TFA, Transene Company). The chromium adhesion layer is etched in diluted chromium etchant (Cr-14, Cyantek Corporation). The interconnect is shown in Figure 5.

IV - Acknowledgements This work was performed in the Lurie Nanofabrication Facility (LNF) in collaboration with the Wireless Integrated MicroSystems (WIMS) group and the Integrated Microsystems Laboratory of the University of Michigan, Ann Arbor. A.A.A. Aarts would like to thank the staff of the LNF for the technical support and the Research Foundation - Flanders (FWO) together with the agency for Innovation by Science and Technology (IWT) for the financial support. References

[1] Q. Bai, K.D. Wise, and D.J. Anderson, “A high-

Figure 5: Microscopic picture of the transferred film showing the overhanging electrical contacts.

[2]

III – Results and Conclusion A. Out-of-Plane Assembly Several test structures have been fabricated to get an idea about the electrical contact resistance. The fabrication of the test structure will not be described in this paper. The test structures are assembled perpendicular into the platform using a flip-chip bonder [2]. The assembly goes smooth and no lubrication, for example methanol, is required to facilitate the assembly [11]. Figure 5 shows the assembled test structure. Preliminary four-point measurements show a contact resistance around 3.2Ω.

[3] [4]

[5]

[6] [7] Figure 5: Picture of the out-of-plane assembled structure (left). Macroscopic picture of the compressible interconnects(right).

[8] B. Conclusions A biocompatible compressible interconnect for out of plane structures has been demonstrated. The number of process steps have been pushed to a minimum by using a transfer bonding technique followed by an anodic release process. The interconnect should be fully scalable because all the metallization steps have been done on planar surfaces. Contact resistances of 3.2 Ω have been measured. Future experiments are planned to test the high density interconnect shown in Figure 3.

[9] [10] [11]

yield microassembly structure for threedimensional microelectrode arrays,” IEEE Trans. Biomed. Eng., vol. 47, no. 3, pp. 281- 289, March 2000. A.A.A. Aarts, H.P. Neves, R.P. Puers and C. Van Hoof, “Interconnect for out-of-plane assembled biomedical probe arrays”, J. Micromech. Microeng. vol. 18, 064004 (7pp) 2008. Product information for WL-5351 and WL-5150, Dow Corning http://www.dowcorning.com/applications/search Kenneth W. Yee, “Gold-Electroplating Technology for X-Ray-Mask”, master thesis department of electrical engineering and computer Science, MIT, June 1996. O. Graudejus, C. Tsay, Z. Yu, B. Morrison, S. P. Lacour, S. Wagner “Advances in Encapsulating Elastically Stretchable Microelectrode Arrays”, Materials Research Society Symposium Proceedings, Vol. 1009E, U04.2 (2007). Salil P. Desai, Brian M. Taff and Joel Voldman, “A Photopatternable Silicone for Biological Applications”, Langmuir 2008, 24, 575-581. S. Bhattacharya, A. Datta, J. M. Berg and S. Gangopadhyay, “Studies on surface wettability of poly(dimethyl) siloxane (PDMS) and glass under oxygen-plasma treatment and correlation with bond strength”, J. Microelectromechanical syst., vol 14, NO. 3, June 2005. S. Metz, A. Bertsch, and P. Renaud, “Partial release and detachment of microfabricated metal and polymer structures by anodic metal dissolution”, J. Microelectromechanical syst., vol. 14, NO. 2, April 2005. Vargel, Christian, “ The corrosion of Aluminium” Book: part B. Corrosion of Aluminium © 2004 Elsevier. M. Datta, “Anodic dissolution of metals at high rates”, IBM J. research and Development, vol. 37, pp.207-226, 1993. Y. Tanye Tang et. Al., “Microfluidic Chamber with Active Suction Ports for Localized Chemical Stimulation of Brain Slices,” to be presented in MEMS 2010, January 24-28, Hong Kong, China.