Page 46 - Chip Scale Review_January-February_2024-digital

Summary
The investigation of an alternative
approach to evaluate bond strength for
actual die-level direct bonded samples was
executed. The estimation of bond strength
at the die level is crucial in assessing the
reliability of advanced chiplets processed
using D2W hybrid bonding. The use of the
nanoindentation test made it possible to
measure the in-plane uniformity of W2W
bonding. Furthermore, it is now possible
to measure the bonding strength of D2W
samples, which could not be measured
in the DCB test. Although the accuracy
of the measurement is still a challenge,
the measurement can accelerate the
development of D2W hybrid bonding with a
Figure 9: a) An SPM image of indentation; and b) An analysis image from the interface (thermal SiO 2 ). precise bond strength value.
Acknowledgments
This article was presented at the 25th
Electronics Packaging Technology
Conference (EPTC 2023) and was edited for
publication in Chip Scale Review. This work
was supported by the “Intensive Support for
Young Promising Researchers” project from
the New Energy and Industrial Technology
Development Organization (NEDO) and
Strategic Research Seed Cultivation Project
from KISTEC.

References
1. F. Inoue, et al., “Inf luence of
composition of SiCN as interfacial
layer on plasma activated direct
b o n d i ng ,” E C S J. of S o l i d
State Science and Tech., vol. 8,
pp. 346-350, 2019.
2. J. H. Lau, “Recent advances and
trends in advanced packaging,” IEEE
Trans. on Components, Packaging and
Manufacturing Tech., vol. 12, no. 2, pp.
Figure 10: The bonding energy of D2W samples for different bonding dielectric interfaces and locations.
228-252, 2022.
existence of voids and the impact of bonding the die as they do not completely overlap. 3. A. Elsherbini, et al., “Enabling
voids on bond strength is another area of In the case of LT-SiO 2 , the average bonding next generation 3D heterogeneous
2
interest for our investigations. energy was 0.58J/m . This indicated that integration architectures on Intel
process,” 2022 International Electron
For the case of D2W bond strength the process for D2W direct bonding that we Devices Meeting (IEDM), San
measurement, three different samples were performed is not well optimized compared Francisco, CA, USA, pp. 27.3.1-27.3.4.
prepared. The bonding is homogeneous (i.e., to the one used for W2W. 4. T. Uhrmann, “Accelerating 3D and
the same dielectric layer is used for both top Continuing the above discussion for our heterogeneous integration with high-
die and the bottom wafer). Figure 9 shows case in particular, the plasma activation volume D2W hybrid bonding,” Chip
the SPM image of D2W bonded samples and bonding had some time interval. As we Scale Review Mar/Apr 2023.
with Th-SiO 2 as the bonding dielectric layer. observed for the case of W2W bonding, the 5. F. Inoue, et al., “Inorganic temporary
The delamination area of D2W is clearly plasma condition makes a large difference direct bonding for collective die to
larger than that of W2W, even though the with respect to the bond strength. The wafer hybrid bonding,” 2023 IEEE
interface of W2W and D2W is the same activated site, e.g., the OH group, or even a 73rd Electronic Components and Tech.
thermal SiO 2 film. dangling bond, might have been deactivated Conf. (ECTC), Orlando, FL, USA, pp.
556-563.
In the case of Th-SiO 2 , the average due to the queue time for the D2W samples. 6. I. Nathan, N. Nejadsadeghi, N.
bonding energy of the Center (in this case, The further optimization of the D2W direct Kohama, H. Tanoue, K. Motoda,
2
the center of the die) was 0.53J/m , the bonding process is being evaluated based on “Distortion simulation for direct wafer-
average bonding energy of the Edge was the obtained bond strength values. to-wafer bonding process,” 2023 IEEE
0.58J/m (Figure 10). Comparing the Center 73rd ECTC, Orlando, FL, USA, pp.
2
and Edge quartiles, there is a variation within 694-698.
44
44 Chip Scale Review January • February • 2024 [ChipScaleReview.com]