Adhesion Improvement on Smooth Cu Wiring Surfaces of Printed Circuit Boards

[Technical Paper] Adhesion Improvement on Smooth Cu Wiring Surfaces of Printed Circuit Boards Motoaki Tani*, Shinya Sasaki*, and Keisuke Uenishi** *Next-Generation Manufacturing Technologies Research Center, Fujitsu Laboratories Ltd., 10-1, Morinosato-Wakamiya, Atsugi City, Kanagawa 243-0197, Japan **Graduate School of Engineering, Osaka University, 2-1, Yamadaoka, Suita City, Osaka 565-0871, Japan (Received August 5, 2011; accepted October 13, 2011) Abstract A novel Cu surface treatment method was developed to improve the adhesion between smooth Cu surfaces and epoxy dielectric layers using a silane coupling agent. In this process, the Cu surface was first modified with a functional group, which was then treated with the silane coupling agent. Here, we investigated the effect of the functional group and the silane coupling agent on the adhesion strength between Cu and epoxy dielectric layers. We also investigated the influence of the Cu/dielectric adhesion layer on smooth Cu wirings of the printed circuit board manufacturing process compatibility. From this study, it was found that two-step Cu surface modification with triazine trithiol followed by treatment of mercapto-group containing silane coupling agent would dramatically enhance the adhesion strength. It was found that a thin triazine trithiol layer was very effective to improve desmear-resistance. Keywords: Cu Wirings, Adhesion, Triazine Trithiol, Silane Coupling Agent, Desmear 1. Introduction Currently, there is a continuous and rapid increase in the performance of electronic devices. This has been possible because of the significant developments that have been realized in LSIs and LSI integration technologies, which enable higher data transmissions. In order to realize the high speed data transmission requirements for next generation electronics, further developments are necessary in the design and fabrication of smooth Cu wirings with sufficient adhesion to the dielectric layers in multilayer motherboards and high density packaging substrates. One of the major problems of these Cu wirings is the weak adhesion that exists between Cu and dielectric layers. Therefore, various studies aimed at improving the adhesion between the Cu and dielectric layer in packaging substrates have been reported, especially for the bottom Cu surface and the dielectric layer, using chemical treatments on packaging substrates, instead of the conventional physical treatment.[1 5] On the other hand, the physical adhesion between the upper surface of the Cu wiring and the dielectric layer has been improved using etching to increase the surface roughness of Cu to more than 2 μm.[6, 7] However, the surface roughness would induce significant resistive losses at high transmission frequencies as given in the following expression. d = 1/(ω μ σ) 1/2 where d is the skin depth, ω = πf (f is the frequency), μ is the permeability, and σ is the conductivity. Therefore, the dimensions of the ridges at the Cu/ dielectric interface are larger than the skin depths at high frequencies over 1 GHz. Consequently, the current will flow along the contour of the ridges and accordingly increase the effective resistance and inductance, resulting in considerable transmission losses. Therefore, there is an urgent need to develop technologies for fabrication of smooth Cu wirings with strong adhesion to the dielectric layers. This paper describes a novel surface treatment method for improving the adhesion between smooth Cu wiring surfaces and epoxy dielectric layers using mercapto-group containing silane coupling agent and triazine trithiol. We further explain that the introduction of a thin layer of triazine trithiol would significantly enhance the desmearresistance. 24

Tani et al.: Adhesion Improvement on Smooth Cu Wiring Surfaces (2/7) 2. Experimental Procedure 2.1 Materials and sample preparation In this evaluation, 30 μm thick electroplated smooth Cu foils were used. Figure 1 shows the cross sectional SEM micrographs of the Cu surfaces. The average roughness (Rz) of the Cu foil was found to be 0.5 μm, which is very low when compared to the conventional etched (2 μm) Cu foils. The adhesive improvement procedure on the Cu surface involves two steps: (1) formation of functional groups on the Cu surface and (2) treatment of coupling agents on the functional groups. The dielectric film is composed of an epoxy resin and 18 wt% silica filler. After laminating the dielectric film and the Cu foil which has been treated with coupling agents, the laminates were cured at 170 C for 1 hour, and the peel strength was then measured. The coupling agents selected were trialkoxysilane compounds consisting of other functional groups which readily react with epoxy resins including amino-, mercapto- and epoxy-group. The Cu surface was treated with a 0.1 1 wt% solution of the coupling agent in water. The functional groups used in this study readily react with the alkoxy groups of the silane coupling agent including hydroxyl-groups, which can be generated by thermal treatment of Cu foils, and the functional groups in the chemical substances given in Fig. 2. Considering that these functional groups should readily react with the Cu surface (CuO) and silane coupling agents, multi functional groups containing compounds such as benzotriazole (carboxylic acid/triazole) and triazine trithiol (mercaptogroup) were selected as the adhesion promoters. 2.2 Measurement In the test specimens, a series of 1 cm wide cuts was made through the Cu layer. The peel strength was measured with a peel strength tester equipped with a digital force gauge. The number of measurement was more than three at each specimen. Considering the total thermal impact of the 4-layer double sided build-up printed circuit board (PCB) fabrication process, two sets of specimens, one cured at 170 C for 1 hour after laminating and the other cured additional four times at 170 C for 1 hour, were subjected to the peel test mainly. 2.3 Via-hole formation process Figure 3 shows the conventional via-hole formation procedure in build-up PCB manufacturing. After formation of the Cu wiring, the Cu surface is treated with the adhesion promoter, and the dielectric layer is then laminated on the (a) Smooth surface (Rz = 0.5 μm) (1) Surface treatment of Cu wiring (2) Lamination of dielectric film (b) Conventional (Rz = 2 μm) Fig. 1 SEM cross-sectional micrographs of Cu surfaces. (3) Laser Via Drilling (a) Triazine trithiol (b) Carboxy benzotriazole Fig. 2 Molecular structures of the adhesion promoters with functional groups. (4) Desmear treatment Fig. 3 Via-hole formation procedure of build-up printed circuit boards. 25

wiring layer. After curing the dielectric layer, via-holes were formed by CO 2 laser drilling, and then cleaned by desmear treatment. These desmear solutions consist of an alkaline swelling solution, an alkaline permanganate (permanganese acid) etching solution, and a neutralizing solution. 3. Results and Discussion 3.1 Selection of coupling agents Figure 4 shows Cu/dielectric peel strengths of the conventional etched Cu surface and the Cu surface treated with 1 wt% silane coupling agents after forming hydroxylgroup on Cu oxide. Irrespective of the treatment method, the initial peel strengths of all the specimens were found to be high (more than 10 N/cm). However, with the exception of the mercapto-silane treated specimen, after thermal treatment (additional four cure cycles), there was a drastic reduction in the peel strengths of the specimens. It is considered that the dielectric resin we used contains the ingredient which reacts with mercapto-groups strongly. Therefore, γ-mercaptopropyltrimethoxysilane was used as the silane coupling agent for the following experiments. 3.2 Effect of the functional group on Cu surface Figure 5 shows the peel strengths of specimens prepared by three Cu surface treatment methods: (1) formation of functional groups followed by reaction with the silane coupling agent, (2) treatment with only triazine trithiol, (3) treatment with only the silane coupling agent. For specimens with hydroxyl-groups or mercaptogroups followed by reaction with the silane coupling agent on the Cu surface, high peel strength of 10 N/cm and more than 8 N/cm were obtained with initial and heat treated (additional four cure cycles) samples, respectively. As shown in Fig. 6(a), the failure mode of the peel test was also found to be the destruction of the resin. It is considered that because hydroxyl-groups have formed on Cu oxide directly, all of the bonds between the Cu oxide and hydroxyl-group as well as bonds between the hydroxyl-group and the silane coupling agent are sufficiently strong, resulting in high peel strength. On the other hand, with respect to the mercapto-group, the triazine trithiol used in this study has three mercapto-groups, two of which make coordination bonds with the Cu surface and the remaining group reacts with the silane coupling agent. Therefore, triazine trithiol treated specimens also Fig. 5 Effect of adhesion promoter functional group and silane coupling agent on peel strength. (a) Treatment of triazine trithiol and coupling agent Fig. 4 Effect of functional groups of the silane coupling agents on peel strength. (b) Treatment of carboxy benzotriazole Fig. 6 Surface morphologies of Cu foils after peel test. 26

Tani et al.: Adhesion Improvement on Smooth Cu Wiring Surfaces (4/7) exhibited high peel strength. However, in the case of specimens in which the Cu surfaces have been treated with the mercapto-group by treating with triazine trithiol (but without the silane coupling agent on the mercapto-group), the initial peel strength was found to be about 8 N/cm, which is 2 N/cm less than that of the samples obtained with both treatments. Further, when the Cu surface is treated with only the silane coupling agent without mercapto-group of triazine trithiol, the initial peel strength was 6 N/cm and the peel strength after the thermal treatment was very low (less than 1 N/cm). The introduction of hydroxyl-groups or mercaptogroups on Cu surface resulted in high peel strengths when these functional groups are treated with the silane coupling agent. On the other hand, the initial peel strengths of the samples treated with nitrogen containing carboxylic acid was found to be 1 N/cm, and peeling off occurred after the thermal treatment. In this case, as can be seen from Fig. 6(b), the failure mode of the peel test was found to be the Cu/dielectric interface. Here, we presume that the carboxylic acid groups and nitrogen in carboxy benzotriazole form chelates with the Cu surface, and there are no other functional groups to react with the silane coupling agent. As a result, it exhibited very low peel strength. For this reason, we consider that the adhesion promoter molecule should have more than one functional group, of which one should be a free functional group that does not form chelates with the Cu surface, thus making it possible for reaction with the silane coupling agent. The formation of a hydroxyl-group on the Cu oxide surface followed by the coupling agent treatment resulted in high peel strength, even after thermal treatment. However the de-lamination between the dielectric layer and Cu surface occurred due to the dissolution of Cu oxide during desmear treatment of the PCB manufacturing process. Figure 7 shows the de-lamination Cu oxide surface around the via-holes. Therefore, in our next study, we investigate the mechanism for the formation of mercapto-group by triazine trithiol and the adhesion improvement on the Cu surface by detailed analysis at the interface between the Cu and dielectric layer. 3.3 Adhesion layer structure and adhesion mechanism on Cu surface Figure 8 shows the depth profile of auger electron spectroscopy of the Cu surface treated with both triazine trithiol and the silane coupling agent. S and Si from the silane coupling agent were detected at the immediate vicinity of the surface, whereas N and S from triazine trithiol were detected a little further from the surface. A maximum of 12% Si was measured from the silane coupling agent. From this result, it can be deduced that the immediate surface is composed of a high density silane coupling agent layer, and beneath this layer there is a mercapto-group layer formed by triazine trithiol on the Cu surface. Figure 9 shows the depth profile of auger electron spectroscopy of the Cu surface treated with only triazine trithiol. Here, N and S from triazine trithiol were detected at Fig. 8 Auger Electron Spectroscopy Depth profile of Cu surface with triazine trithiol and the silane coupling agent. Fig. 7 Effect of de-smear treatment of via-hole on oxide layer on Cu surface. Fig. 9 Auger Electron Spectroscopy Depth profile of Cu surface with triazine trithiol. 27

the immediate surface, and a mercapto-group layer formed by triazine trithiol was also detected in this area. A maximum of 17% S was detected at the surface from mercaptogroups. The amount of S was found to be the same as that of the treated triazine trithiol and the silane coupling agent. However the peel strength of the sample with the mercapto-group from triazine trithiol was 2 N/cm less than that of the sample with the mercapto-group from both triazine trithio and the silane coupling agent. Triazine trithiol has a ring structure, and it therefore reacts with the dielectric layer only at its immediate surface at the range of the length from the mercapto-group. On the other hand, the silane coupling agent has a methylene chain, and it can make bonds at a length that is the sum of the lengths of the mercapto-group and methylene chain. Therefore, it is considered that the treatment with both triazine trithiol and the silane coupling agent would make it possible for the mercapto-group to react with the resin not only at the immediate resin surface, but also inside the resin at a depth of the length of about a methylene chain. Consequently, we believe that treating with both triazine trithiol and the silane coupling agent can result in high peel strength although the same amount of mercapto-groups exist on the Cu surface. Figure 10 shows the depth profile of auger electron spectroscopy of the Cu surface treated with only the silane coupling agent. We confirmed that there exists 5% Si from the silane coupling agent. We believe that the lower peel strength of these samples were due to the low functional group density on the Cu surface. Therefore, it can be deduced that the formation of functional groups (especially mercapto-groups), which can react with the silane coupling agent on the Cu surface, is necessary to achieve high peel strength. 3.4 Effect of the triazine trithiol treatment time Figure 11 gives the peel strengths of the initial and heat treated specimens prepared by three surface treatment methods prior to the silane treatment. They are as follows: (1) without triazine trithiol treatment, (2) 30 sec triazine trithiol treatment, (3) 60 sec triazine trithiol treatment with 0.01 wt% solution of triazine trithiol in water. Figure 12 shows the photographs around the via-holes of specimens with 30 sec and 60 sec triazine trithiol treatment after the desmear process. Although the peel strengths of both specimens after thermal treatment were found to be more than 8 N/cm, for the 60 sec treated specimens, a color change was observed in the region surrounding the viaholes. This phenomenon may be attributed to the de-lami- Fig. 11 Effect of triazine trithiol treatment time on peel strength. (a) 30 sec treatment Fig. 10 Auger Electron Spectroscopy Depth profile of Cu surface treated with the silane coupling agent. (b) 60 sec treatment Fig. 12 Effect of desmear treatments at via-holes on Cu surface treated with triazine trithiol. 28

Tani et al.: Adhesion Improvement on Smooth Cu Wiring Surfaces (6/7) nation of dielectric layer from the Cu surface, and this may be due to the dissolution of Cu oxide on the Cu surface by the desmear solution. We guessed that Cu oxide layer was formed on the Cu surface treated with the triazine trithiol solution for 60 sec. However, according to the detail observation, the appearance of this de-lamination is different from that of Fig. 7. We considered that the triazine trithiol reacted on Cu surface prevented from dissolving Cu oxide on the Cu surface a little. Figure 13 shows the depth profiles of auger electron spectroscopy of the Cu surfaces treated with triazine trithiol solution for 30 sec and 60 sec. When the treatment time was 30 sec, S and N from triazine trithiol were detected on the Cu surface, but the oxygen concentration was found to be very low. We consider the thickness of a Cu oxide layer is less than 5 nm. On the contrary, when (a) 30 sec treatment the treatment time was 60 sec, high concentrations of S and N from triazine trithiol were detected along with a high concentration of O from Cu oxide. From this data, the thickness of the Cu oxide layer was calculated to be 30 nm. Therefore, it can be concluded that the Cu surface treated with triazine trithiol solution has a triazine trithiol layer and a Cu oxide layer. Consequently, we found that a triazine trithiol layer on the Cu surface didn t dissolve in the desmear solutions. The treatment on the Cu surface using triazine trithiol paves the way for the realization of both higher peel strength and higher desmear-resistance. This would occur if the silane coupling agent is treated after forming the triazine trithiol layer on the Cu surface, which would prevent Cu oxide formation. 4. Conclusion We have developed a novel surface treatment method to improve the adhesion between smooth Cu surfaces and epoxy dielectric layers using a silane coupling agent together with a triazine trithiol compound. The introduction of the mercapto-groups of triazine trithiol followed by the treatment of the silane coupling agent on Cu surfaces resulted in high peel strengths of 10 N/cm and grater than 8 N/cm with initial and heat treated (additional four cure cycles) samples, respectively. From the peel test, it was confirmed that the failure mode resulted in the destruction of the resin. The peel strength of the Cu foils obtained by the treatment with only triazine trithiol exhibited weaker peel strength of less than 8 N/cm, compared with the samples with both treatments. We found that a Cu oxide layer is formed at the Cu surface during the triazine trithiol treatment. When the treatment was done for a longer time, a thicker layer of Cu oxide formed. Due to this thick oxide layer, dielectric/cu interface de-lamination occurred during the desmear process. Therefore, we believe that in order to prevent formation of Cu oxide layer, a thin layer, perhaps a molecular monolayer, of triazine trithiol would be essential. (b) 60 sec treatment Fig. 13 Auger Electron Spectroscopy Depth profiles of Cu surfaces with triazine trithiol. References [1] M. Tani, K. Nakagawa, M. Mizukoshi, and M. Kato, Fine-Pitch Multilayer Wiring Technology for Packages using Via Posts and Grinding Planarization, THE IEICE TRANSACTIONS ON ELECTRONICS, Vol. J90-C, No. 11, pp. 734 740, 2007. [2] S. Z. Shi, J. C. Wei, C. Rhodine, W. G. Kuhr, and S. Nakamura, Molecular Modification of PCB Sub- 29

strates: Demonstration of HAST Survivability of Fine- Line Patterned Structures, International Conference on Electronics Packaging, pp. 818 823, 2009. [3] M. Tani, K. Nakagawa, and M. Mizukoshi, Multilayer Wiring Technology with Grinding Planarization of Dielectric Layer and Via Posts, Transactions of The Japan Institute of Electronics Packaging, Vol. 3, No. 1, pp. 1 6, 2010. [4] K. Baba, Y. Nishimura, M. Watanabe, and H. Honma, Formation of Fine Circuit Patterns on Cyclo Olefin Polymer Film, Transactions of The Japan Institute of Electronics Packaging, Vol. 3, No. 1, pp. 73 77, 2010. [5] M. Horiuchi, T. Yamasaki, and Y. Shimizu, Metallization Technologies on a Smooth Resin Surface for the Next Generation of Flip Chip Packaging, Transactions of The Japan Institute of Electronics Packaging, Vol. 3, No. 1, pp. 110 115, 2010. [6] T. Nakagawa and H. Tojima, Copper Surface Treatment for the Production of Fine Circuits, Proceeding of 14th Microelectronics Symposium (MES2004), pp. 269 272, 2004. [7] T. Yamashita, F. Inoue, and A. Nakaso, New copper surface treatment method, which produces the highest level of adhesion between smooth copper surface and low dielectric resin, Proceeding of 15th Microelectronics Symposium (MES2005), pp. 377 380, 2005. 30