Enabling Thin Wafer Metal to Metal Bonding through Integration of High Temperature Polyimide Adhesives and Effective Copper Surface Cleans

Enabling Thin Wafer Metal to Metal Bonding through Integration of High Temperature Polyimide Adhesives and Effective Copper Surface Cleans Anthony Rardin and Simon Kirk 1 Dr. Mel Zussman 2 1 DuPont Wafer Level Packaging Solutions, 2520 Barrington Ct. Hayward CA 94545 2 HD MicroSystems, Barley Mill Plaza #30, Wilmington, DE 19880 Contact: Anthony.B.Rardin@usa.dupont.com phone: 510-784-7514 Abstract This paper will examine the use of high temperature Polyimide temporary adhesives and advanced copper cleaning solutions for thin wafer handling during metal to metal (- ) bonding. In the case of to solid state diffusion bonding, temperature requirements can be up to 400C for as long as 40 minutes. These extreme conditions pose a significant challenge with existing materials used for temporary bonding. Polyimides are ideally suited for high temperature applications and polyimide adhesives have been developed which can withstand 400 C processing without significant failure in adhesion. As well, to bonding processes will require the need for contamination free surfaces, which becomes increasingly important if lower temperature processes (<400 o C) are desired. Close-coupled copper cleaning immediately prior to the bonding step offers an excellent approach for providing process reproducibility while achieving increased electrical yields. Utilizing advanced wet cleaning solutions to obtain optimal surface preparation by removing undesirable copper oxide and Benzotriazole (BTA) layers, and potentially adding a copper oxidation barrier (COB) functionality will be discussed in detail. Introduction As the number of thinned wafers being processed increases,, the challenge of handling these wafers without breaking them becomes a real issue. This paper will present an approach that enables the handling of thin wafers by using high temperature polyimides and to bonding avoiding handling the thin wafer altogether. Essentially the thin wafer is always supported by either the carrier or the device wafer it will ultimately be bonded to. Figure 1 shows the use of a high temperature polyimide adhesive used to temporarily bond the wafer to a carrier wafer. The wafer is thinned while being supported by the carrier wafer. Prior to bonding the wafer to the other device wafer a close-coupled clean of any residual BTA or oxide is essential for void free bonding. The wafer, still bonded to the carrier, is then flipped for the to bond avoiding the need to handle the thin wafer without any support. Improved Temporary Bonding Scheme Metal 2 Metal bonding used to avoid thin wafers HD3007 Coat / Prebake HD3007 re Thermo compression Back grinding polish & PCMP Clean CoppeReady & EKC4000 Thermo compression to Sub. EKC Closecoupled cleaning Laser Ablation/Solvent Release Detach glass HD3007 Adhesive Removal Figure 1 Solvent Release & Adhesive Removal EKC865 Adhesive Remover Thermal Stability of Polyimide Adhesive rrently available temporary adhesives can not withstand the 400 o C needed for to bonding, however, polyimide (PI) adhesives can meet these criteria. The graph in Figure 2 shows the TGA weight loss of HD3007 PI adhesive vs. temperature. HD-3007, from Hitachi Chemical DuPont MicroSystems (HDMS), is a spin applied formulation which cures to a thermoplastic polyimide film. TGA Weight Loss HD3007, 10 C/min Figure 2

These data show that the PI adhesive does not have significant weight loss until beyond 500 o C. Further more no delamination was observed at temperatures of 500 o C and below. Delamination was initiated at 510 o C and severe delamination at 5 o C. At these temperatures micrograph results shown below in Figure 3 suggest the PI will delaminate from the glass carrier and remain on the device wafer. Micrographs of delamination after thermal exposure. View through glass plate. From wafer fab last CMP / PCMP process. Assume wafers will have remaining Ox as well as BTA/organic residues Integrated Cleaning Stand-alone Cleaning Integrated Cleaning - - Wafer Bonding Cluster Tool) Pre-Bonding Clean Pre-Bonding Clean Wafer Alignment Wafer Alignment - Wafer Bonding - Wafer Bonding - Wafer Bonding Module SUSS-MicroTec XBC300 Wafer Bonder Cluster Tool with Integrated Cleaner CL300 Cleaning Module SUSS-MicroTec XBC300 Wafer Bonder Figure 4 Figure 3 Surface Cleaning and Preparation Prior to Bonding Key Cleaning Needs The need for contamination-free Copper surfaces prior to - wafer bonding is critical and becomes increasingly important if lower temperature bonding processes (<400 o C) are desired. Two key requirements for a clean copper surface include 1) complete removal of undesirable copper oxide layers, and 2) complete removal of organic contaminants and inhibitors from previous steps (e.g. Benzotriazole (BTA)). Additionally, it may be advantageous to protect the copper surface with a thin copper oxidation barrier (COB) that can be easily applied and easily removed prior to bonding. This would protect the clean, more reactive surface and prevent further environmental contamination from attaching to the surface, especially in processes where there may be an extended queue time between cleaning and bonding. Integration Approach The most effective approach for cleaning would be to implement a clean immediately prior to the actual - wafer bonding step, as can be seen in Figure 4. Leading edge bonding tools will have clustered modules which include a pre-bonding clean station. Close-coupling the clean with the bonding step through this clustered arrangement allows for better control of the copper surface prior to bonding. In the case of our studies, a XBC300 wafer bonder was used which has a dedicated cleaning module, CL300. However, not all bonding equipment will have integrated cleaning modules, and it is important to ensure that the cleaning chemistries and process will also be effective in stand-alone cleaning tools as well. Various pre-bonding cleaning solutions were compared for their ability to 1) remove copper oxides, 2) remove BTA films, 3) remove a combination of BTA in the presence of thin copper oxides, and 4) add copper oxidation barrier (COB) functionality to prevent oxide (re)growth. Benzotriazole (BTA) Removal Benzotriazole (BTA) is commonly used to protect surfaces in wafer fab CMP processes. It can be found in the polishing slurries to prevent copper corrosion and static etching during the CMP process. It is also added to interstation rinse steps in the CMP tool to ensure copper protection as the wafers are processed. In wafer fab processing, the BTA needs to be removed prior to subsequent liner deposition steps as part of the copper metallization process. One of the goals of the PCMP cleaning step is to remove any residual BTA left on the wafer surface. However, not all commercially available PCMP cleaners can completely remove the BTA film. In this case, the remaining BTA must be removed during a pre-clean in the liner deposition step. However, at the last CMP step, there is no subsequent liner clean/deposition needed, and so the wafers would leave the fab with residual BTA on the surfaces. On the other hand, if the PCMP clean step completely removes the BTA film, the copper is very reactive towards the environment and can grow copper oxide rapidly. Therefore the result would be wafers leaving the wafer fab with significant oxide growth. To solve this, DuPont/EKC Technology offers cleaning solutions that can completely remove the BTA and oxide, and then immediately protect the copper using a COB. This approach can be seen in Figure 5. Removal of BTA through Cleaning -BTA Incoming Optimized Cleaning Complete BTA removal Incomplete Cleaning Incomplete BTA removal Complete BTA removal Copper Oxidation Barrier (COB) Possible Downstream Issues Ox (re)growth Process Benefits No loss / etching No Ox (re)growth Localized etching Organic contamination Bonding issues Electrical / yield loss loss Bonding issues Electrical / yield loss COB easily removed by thermal desorption Enhanced bonding Improved electrical / yield performance Figure 5

A common way to measure BTA removal from a surface is through contact angle measurements using DI water. Pristine copper typically has a low contact angle in the range of 15-30 o and is hydrophilic. When BTA is added to a cleaned copper surface, the surface becomes hydrophobic and the contact angle increases significantly to > o. Figure 6 shows this effect on contact angle measurements. The contact angle measurements correlate quite well with surface analysis measurements using Fourier-Transform Infrared Reflection Absorption Spectroscopy (FT-IR-RAS). In this case, cleaned wafer pieces were contaminated with a BTA solution (0.1-1.0%) and then cleaned using different solutions. The surface was then measured using FT-IR- RAS and contact angle. The analysis showed good correlation between FT-IR-RAS peak intensities at 750 cm -1 (aromatic C-H of the BTA) and contact angle measurements. Test for BTA Removal Performance Hydrophilic Surface low contact angle Contact Angle DIW contact angle on [deg] DI Water 70.0.0 50.0 40.0 30.0 20.0 10.0 0.0 DIW contact angle [w/o BTA] Hydrophobic Surface high contact angle Figure 6 Contact Angle -BTA DI Water Organic residue [ ex -BTA] The data can be seen in Figure 7. This allows us to use contact angle for screening of cleaning chemistries, which is a quicker and less costly method than sending samples to 3 d party labs for FT-IR-RAS analysis. Copper Oxidation Barriers (COB) Copper oxidation barriers (COB) allow protection of the cleaned copper surface from environmental contamination including copper oxide growth. The previous Figure 5 shows that in a typical cleaning process when BTA is completely removed from the surface the bare copper can quickly grow undesirable oxides. The optimized cleaning process flow illustrates how DuPont/EKC Technology cleaners solve the technical need of protecting the cleaned copper surface with a COB. The COB forms a thin organic layer that prevents the formation of copper oxide. Typically, the layer is deposited on an oxide free copper surface from the cleaning solution. Surface analysis using XPS/ESCA, IR, and ToF- SIMS data confirm the presence of a thin organic layer. XPS/ESCA data allows assignment of the metal oxidation states and thickness of layer. IR data confirms the presence of the COB on the copper surface and give structural information on the surface complex. ToF-SIMS data confirm presence of COB and give an indication of the stability of the film. Cyclic voltammetry and linear sweep voltammetry are used to characterize the oxidation of the copper surfaces treated with COB and exposed to ambient atmosphere. The graph in Figure 8 shows linear sweep voltammetry measurements on a copper surface comparing cleaning chemistries with and without COB. As can be seen in the graph, the non-cob containing sample has multiple oxide peaks when measured after 18hrs exposure to an ambient environment. In comparison, the COB containing sample shows no evidence of oxide peaks when measured after 68hrs exposure to the same ambient environment. Comparison of COB Treated and Untreated Surfaces with Exposure to Ambient Atmosphere COB treated surface after 68 hours exposure No Copper Oxide formed Untreated sample showing Copper oxides from 18 hours exposure BTA Removal and Contact Angle Correlation Surface sensitive FT-IR (RAS) of BTA treated wafer Aromatic C-H of BTA Figure 7 Conditions BTA only Chemical A (Partial BTA Removal) Chemical B (Complete BTA Removal) Good correlation between peak intensities (750 cm -1 ) and contact angles was observed Contact Angle.3 42.6 30.3 Figure 8 While having COB functionality is important in a cleaning formulation, it is also very important that the COB layer can be easily removed from the surface without any special processing that would decrease process throughput. Figure 9 shows the use of X-ray Photoelectron Spectroscopy (XPS) / Electron Spectroscopy for Chemical Analysis (ESCA) analysis to compare a pristine surface that has been sputtered clean using Argon, and a COB treated surface that has subsequently had the COB layer removed by thermal desorption.

Process Flow Removal of Copper Oxidation Barrier (COB) from Surface surface cleaned in EKC copper clean ing chemistry containing COB Sample allowed to sit in ambient environment for 24hrs. Sample put in XPS and heated for 2-3 mins at 200 o C to remove COB surface analyzed for COB removal Sample is compared to a clean surface that was Argon sputtered Normalized Intensity 1 0.9 0.8 0.7. 0.6 0.5 0.4 0.3 Normalized 2p3/2 spectra Brief post treatment yields (0) surface similar to pristine Ar sputtered film with COB barrier removed after 24hr Ar Sputter cleaned film Cleaning Performance DuPont/EKC Technology cleaning solutions showing excellent cleaning performance for BTA and oxide removal resulting in very low surface contact angle results can be seen in Figure 11. The Group A cleaning formulations seen in the far left side of the graph provide contact angle measurements in the 10-15 o range. These cleaning formulations were all processed at room temperature with a chemical exposure time of seconds. BTA and Oxide Cleaning Performance 0.2 0.1 0 Figure 9 950 945 940 935 Binding Energy (ev) 930 925 Group A PCMP cleaners exhibit excellent BTA and Ox removal resulting in low contact angle The data show that the two curves are virtually identical, indicating that a heating process of 2-3 minutes at 200 o C successfully removed the COB layer from the surface. These process conditions would be easily achieved in a wafer bonding chamber. As was the case for measuring BTA removal on a surface, it can also be very time consuming and expensive to perform large numbers of screening tests to evaluate COB functionality on a copper surface. We have found that a good proxy test to simulate COB functionality is through accelerated DI water rinsing. In this testing, first the surfaces are pre-cleaned to remove any residual BTA/organics and oxides; second the cleaned copper surfaces are immediately immersed into the different cleaning solutions; third the surfaces are then rinsed in DI water from anywhere between 5 30 minutes; finally the pieces are dried and then inspected in a Scanning Electron Microscope (SEM) for copper oxide formation. The SEM images can then be quantified through contrast comparisons (copper oxide formations are a very bright white in the SEM vs. copper which is darker) for efficiency. Figure 10 outlines this process, which allows for faster screening of large numbers of potential cleaning formulations. COB Functionality Screening Using Accelerated Conditions Wafer : blanket [ Polished with barrier CMP] Flow i. Pre- Treatment : Remove Ox and -BTA to prepare pristine. ii. Cleaning Chemistry : Dipped in cleaning solution then rinsed DIW. iii. SEM inspection : Top down view iv. Measurement : Calculate Ox area (through contrast) Group A Group C Group D Figure 11 Evaluation of potential cleaning formulations for COB functionality was performed using the accelerated method previously referred to in Figure 10. Once again the cleaning conditions were kept constant at room temperature and seconds chemical exposure time. The results can be seen in Figure 12, and clearly show that the COB-containing cleaning formulation prevented oxide growth during the accelerated DI testing while the cleaning formulation without the COB had significant copper oxide growth on the copper surface. Copper Oxidation Barrier (COB) Performance Wafer : blanket Process: 1) Preclean to remove Ox 1min DIW 30sec N2 Dry 2) Dip in cleaner 1min DIW 30sec 3) Use extended DIW rinsing 5min to promote Ox growth SEM check -BTA COB COB Pre 1. Pre- Treatment 2. PCMP Clean + DIW dip 3. SEM inspection Method allows rapid evaluation of COB functionality on surfaces % of White Area = ~ % Clean Chemistry with COB Ox regrowth on unprotected surface Clean Chemistry without COB Figure 10 4. Measurement Figure 12

Further work was carried out jointly with SUSS MicroTec on their XBC300 wafer bonder and CL300 cleaning module. Prior to any bonding tests, DuPont/EKC Technology cleaning chemistries were sent to SUSS and screened for efficiency in removing thin and thick copper oxide films, and BTA films on both bare and thin copper oxide. SUSS also compared various other chemistries including their existing process of record (PoR). Contact angle testing was used for this study. All cleaning solutions were run at room temperature for 30 seconds in a static beaker environment and contact angle measurements were taken immediately after the clean step. The electroplated wafers used for this testing simulated the type of oxide that is typical of wafers that would have received depositions a few days to weeks prior to the bonding. The results of the oxide removal tests can be seen in Figure 13. the contact angle, indicating the BTA and/or oxide films were still remaining on the surface. Pre-Bonding Cleaning Chemistry Performance: BTA Removal BTA was coated on the surfaces by dipping the samples in a 0.01M BTA solution and rinsing BTA creates a nearly perfect phobic surface with a contact angle of 80-90 degrees. This prevents other species in the environment from attaching to the surface. BTA is less effective on an oxide surface vs. bare, but is good for simulating BTA and Oxthat can be present on incoming surfaces Contact Angle (deg) 80 70 50 40 30 Removal of BTA on Oxide from Surfaces with Different Clean Treatments Chemical Conditions = 30sec @ Room Temperature (Puddle) 78 73 61 Pre-Bonding Cleaning Chemistry Performance: Copper Oxide Removal 20 10 9 This type of oxide is typical of wafers that had depositionsa few days to weeks prior to bonding Note that oxide layers do not self limit. The oxidation continues over time and thickens in response to time, temperature and humidity. Contact Angle (Deg) 80 70 50 40 30 20 10 0 Removal of Oxide from Surfaces with Different Clean Treatments Chemical Conditions = 30sec @ Room Temperature (Puddle) 19 37 Native Ox Citric Acid EKC4000 Chem 1 Chem 2 Chem 3 Figure 13 The data shows that the untreated oxide has a contact angle of approximately o. Several chemistries appear to be successful at removing the oxide as evidenced by the contact angle reduction to the 15-30 o range. Citric acid provided the lowest contact angle, with EKC4000 TM and several other chemistries achieving a lower contact angle. BTA removal was also tested on both bare and thin oxide films. While BTA is less effective on oxide surfaces compared to bare, it is still useful for simulating BTA and Oxide that can both be present on incoming surfaces due ineffective PCMP cleaning. The BTA was coated on the surfaces by dipping the samples in a 0.01M BTA solution and rinsing, and the cleaning process was identical to the one used for the oxide removal tests. Figure 14 shows the contact angle as a function of different cleaning chemistries. The contact angle of untreated BTA on oxide is around o. The data indicates that only EKC4000 was able to clean the BTA and oxide contamination from the surface, resulting in a very low contact angle of 9 o. The other cleaning chemicals tested either showed no improvement or increased 34 12 70 0 BTA on Ox Citric Acid EKC4000 Chem 1 Chem 2 Figure 14 Once the screening tests for BTA and oxide removal were completed, the most promising candidates were chosen for the wafer bonding tests. Several candidates from DuPont/EKC Technology including EKC4000 TM, EKC5500 TM, and EKC520 TM were investigated. Bonding was performed on blanket wafers made up of 200Å Ti /1KÅ PVD /3KÅ Electroplated. Scanning Acoustic Microscopy (SAM) was used to evaluate the effectiveness of bonding. A scanning acoustic microscope (SAM) uses focused sound to investigate, measure, or image an object and is commonly used in non-destructive failure analysis and useful in detecting voids, cracks, and delaminations within microelectronic packages. Multiple splits were run varying different clean and bonding process parameters. The main cleaning splits had two different process conditions. The first split was an aggressive condition in which the copper wafers were left exposed in a high temperature (80 o F) and humidity open environment for 48hrs to enhance oxide growth; the bonding time was shortened; and no forming gas was used during bonding. The SAM results for the first split can be seen in Figure 15. The SAM images showed quite a large area where voids were located (shown as white areas on the blue background) on the wafer with no wet cleaning treatment. Improvement was seen when EKC4000 was used as a pre-cleaning chemistry and the EKC5500 sample exhibited significant reduction in voids.

Scanning Acoustic Microscopy Results after - Wafer Bonding (1 of 2) Aggressive Split Conditions Enhanced oxide growth, shortened bonding time Wafers: 200A Ti /1KA PVD /3KA Electroplated Blanket Wafers Wafers left exposed in high humidity, high temperature (80 o F) open environment for 48hrs to enhance oxide growth Bonding Conditions: Standard temperature and bonding force; Shortened bonding time Cleaning Conditions: Room Temperature, 30 seconds puddle process (no spray / mechanical agitation) Path Forward The cleaning data seen so far is very promising. Ongoing work is underway to further evaluate the efficiency of the cleaning steps on bonding integrity. This includes: Review SEM /and Transmission Electron Microscopy (TEM) data of the - bond interface on bonded samples using different cleaning chemistries (wafers from split # 1 were sent out for analysis) No Chemical Pretreatment No Forming Gas EKC4000 TM No Forming Gas Figure 15 EKC5500 TM No Forming Gas The second split simulated more standard conditions in which the copper oxide growth was not enhanced; the bonding time was not shortened; and forming gas was used during the bonding. Figure 16 shows the SAM results from the second split. The results show that when using the standard Note - Wafers from this split are currently being processed by SEM / TEM for analysis of the - bond interface Scanning Acoustic Microscopy Results after - Wafer Bonding (2 of 2) Standard Split Conditions no accelerated oxide growth, standard bonding time Wafers: 200A Ti /1KA PVD /3KA electroplated Blanket wafers Wafers left in sealed wafer carrier box Bonding Conditions: Standard temperature, bonding force, and bonding time Cleaning Conditions: Forming gas used; Chemical run at room temperature, 30 seconds puddle process (no spray / mechanical agitation) No Chemical Pretreatment Forming Gas Used EKC4000 TM Forming Gas Used EKC5500 TM Forming Gas Used Figure 16 conditions the amount of voids seen during the bonding process is significantly reduced, as shown in the image with no wet cleaning treatment. The EKC4000 treatment results in much lower voiding, especially close to the wafer edge. Following the same trend as the aggressive split, the EKC5500 exhibits the best performance resulting in a practically void-free surface. It is important to note in both splits that the CL300 cleaning tool did not use the spray arms for cleaning due to some equipment upgrades being installed. In the bonding clean tests the chemical was manually poured onto the wafer and allowed to puddle for 30 seconds, and then the wafers were rinsed and dried. Once the cleaning module becomes fully operational, there will be a significant cleaning enhancement due to the mechanical agitation and flow of the chemistry. Obtain electrical results from bonded wafers tested with different cleaning chemistries and/or conditions (this work is currently being performed by a 3 d party) As well, further work on the process application side needs to take place to better characterize and generate a process of record for the cleaning process: Continue process window testing on cleaning when CL300 cleaning module is fully functional Evaluate performance at different cleaning temperatures and times Also of interest is extending the cleaning evaluation to include different films that may be present or of interest in bonding: Further evaluate cleaning chemistries for added functionality of removal of Tin (Sn) oxide (initial work confirms Tin oxide removal) Looking at the overall process of integrating the high temperature polyimide adhesives and effective surface cleans the data indicates that the HD-3007 demonstrates a thermal stability strong enough to withstand 400 o C, and the EKC4000 and EKC5500 cleaning materials result in very low void - bonds. Further collaborative work will continue with to identify and optimize a complete - bonding solution. Acknowledgments The authors would like to acknowledge Dr. Shari Farrens and Sumant Sood of for their time and efforts performing the cleaning and copper bonding tests at their facility in Waterbury Center, VT. Cass Shang, Paul Bernatis, Akira Kuroda, and Atsushi Otake of DuPont/EKC Technology and Jeff Thompson of DuPont CR&D for their contributions in advanced cleaning in the areas of copper oxide removal, BTA removal, and copper oxidation barriers.