In order to achieve the accurate interconnect design and manufacturing, we have adopted the following two approaches: accurate interconnect modeling using the RLGC model with skin effect considerations, and development of a low-k and low-loss resin featuring fine adhesion strength with Cu interconnects through the removal of the weak boundary layer. The experimental results revealed that the RLGC model with skin effect consideration showed good agreement with the measurement results of microstrip lines over a wide frequency range. The developed low-k (ε_r=2.75) and low-loss (tan δ=0.0099) resin does not require a rough anchoring surface, resulting in the reduction of metal and dielectric losses. The signal propagation loss is reduced to 0.37 dB/cm at 10 GHz with 50 Ω microstrip lines (line width=88μm, line thickness =15μm) on a 37μm thick resin with fine plated-Cu adhesion (9.2 N/cm) . These technologies will be indispensable for both the accurate modeling of interconnects and high signal integrity in the coming generation of electronics.

View full abstract

In this study, the formation of Kirkendall voids in the reactive diffusion between a Cu plate and lead-rich solders (90Pb-l0Sn and 95Pb-5Sn) in a reflow process was investigated. In addition, to study the conditions for void formation, a short time reflow test was carried out to observe the initial stage of its formation. As a result, conspicuous large Kirkendall voids were observed at the interface between the Cu₃Sn layer and Cu after both a conventional reflow treatment and a 613K×60 s reflow treatment. These voids were observed even after final mechanical polishing. When the sample was aged for 250hr or more at 423K followed by a 613K×10s reflow, large Kirkendall voids were observed in the sample in spite of observing only a small void after reflow. Large Kirkendall voids were observed in the case that only Cu₃Sn was formed during reflow.

View full abstract

In this study, we optimized the vacuum condition for direct bonding of CMP-Cu films using the surface activated bonding (SAB) method at room temperature. The quality of bonding with the SAB method depends strongly on the cleanness of the surface activated by the Ar atom beam irradiation. Therefore, the relationship between the vacuum condition and bonded area was evaluated with the amount of exposure (Pa⋅s) as the parameter. Our results show that the critical exposure condition to obtain a full bonded area on the surface is around 0.1 Pa⋅s. The numbers of molecules colliding with the active surface conceivably determined the cleanness of the surface, and XPS observation showed that oxidation of the surface could be prevented at this level of exposure.

View full abstract

Recently embedded passive devices have been actively investigated in an effort to save the surface area on printed circuit boards. About 60% of the area is occupied by the passive parts. As suggested in the previous paper, thin-film capacitors prepared using a semiconductor technology are one of the most promising candidates. In this paper, these film capacitors were embedded using two different methods. One was a conventional method using solder bump, and the other was a newly-developed method using conductive paste. The solder bump was formed using screen-printing following the wafer-level chip-size package (W-CSP) process. Since the solder bump is currently in practical use, high reliability should be expected. However, the chips formed by this method became thicker due to copper posts, solder bumps, and so on. The other method, using conductive paste, has the advantage of connection without any thickness problem. After embedding and solder heat tests, the capacitance of all the chips did not change, and the differences between before and after the tests are within the acceptable margin of error. However, in most cases, the loss (tan δ) increased slightly after the embedding and solder heat tests. In only one case did the solder heat test for an embedded BST capacitor chip connected by conductive paste show a lower tan δ value than that in the previous test. Based on all the results, the two methods used in this study are concluded to have a high potential for use in embedding methods. Further reliability tests are necessary for practical use.

View full abstract

Rapidly increasing broadband access has led to the requirement for much higher transmission throughput within a network equipment. The most promising solution is optical interconnection using an optical backplane. We have developed a high-density optical backplane using a small-diameter optical fiber to realize multi-tera bit/s throughput. To achieve this, we developed a new high-delta and small-diameter optical fiber (clad diameter=80μm, coating diameter=125μm), a high-den-sity optical fiberboard with small bending radius (r=5mm), a high-density MT ferrule, an easy assemble MT ferrule and compact optical right angle connector. Using these new devices, we fabricated a high-density optical backplane, which complies with the structure of an ATCA backplane. It has achieved a high throughput performance of 3 Tbit/s, which is five times that of an electric backplane.

View full abstract