10 Fundamentals of Bonding Technology and Process Materials …
317
1100 μm down to 270 μm and becomes negligible when the solder diameter reaches
to 130 μm [59]. In fact, our SAC 305 solder with ~130 μm diameter in size led to
a conclusion that the difference in cooling rates between the two process does not
yield a significant difference in solder grain size, therefore, a similar grain boundary
density between the two samples. No difference in grain size between the two average
grain sizes is explained by the degree of undercooling, in which smaller solder balls
(i.e. ~130 μm in diameter) undercools more than the bigger solder balls [55, 59,
77, 78]. Consequently, grain boundary diffusion kinetics of Cu and Ni atoms would
have been similar. Therefore, grain size is not a determining factor for the premature
EM failure observed for TCB-joints. Rather, additional mechanism might have been
involved that could explain for this phenomenon, such as the effect of IMC and β-Sn
crystallographic orientation on susceptibility to EM failure.
Effects of IMC on EM Degradation. A representative solder joint microstructure
before the EM-aging test are shown in Fig. 10.61. Note that from the top to bottom,
silicon die, UBM, IMC, SAC 305, IMC, ENIG surface finish, Cu bond pad, and
PCB substrate are structurally layered to form an interconnection. Note that before
bonding processes, the IMCs on the chip-side with the solder joint had been already
generated via the bumping process. On the chip-side, IMC thickness is similar for
both samples; however, the low-aspect ratio Cu 6 Sn 5 IMC scallops are present in
reflow joints. This difference originates from the two bonding process difference,
although the IMC morphologies are more obvious at the bond pad interface. When
comparing the morphologies of IMCs on the substrate-side, the reflow joint shows
are uniformly structured and entirely covers ENIG surface (Fig. 10.61a) while the
TCB joint exhibits a significantly different interface, where only certain regions are
discerned with rod-like Cu 6 Sn 5 IMCs (Fig. 10.61b). The difference in thickness of
the IMC and its morphology affects the kinetics of the Cu and Ni atom diffusion
through the SAC 305 solder matrix. As suggested from Bashir’s study [54], there is
a faster diffusion of Cu and Ni from the bond pad and ENIG surface finishes if the
IMCs do not fully cover those areas or if the IMC scallops have high aspect ratio
structures. To better visualize these phenomena, readers are aided with a schematic
representation (Fig. 10.62). In terms of EM effect on these samples, it is likely that
TCB joints would suffer from premature failure due to a much faster Cu and Ni
diffusion. Eventually, the depletion of those atoms will lead to void formation and
propagation, hence delaminating the interface that will short the interconnect.
From our results, IMC morphology was shown to have the considerable difference
between the solder joints reflowed by two different processes (Fig. 10.61). The difference in IMC morphology could have been attributed to the cooling rates between
the two bonding processes. Because the reflow-processed joints endured a much
slower cooling rate, molten solder has more time to wet and form/grow the IMCs
before reaching its solidification temperature [59], in which its layer was eventually
deposited at the substrate side of the solder joint interface. How the layers of IMCs
are structured on the solder-Cu bond pad joint interfaces is vastly essential in understanding both the diffusion of Cu and Ni atoms and Joule Heating that eventually
leads to premature failure upon EM-aging test.
317
1100 μm down to 270 μm and becomes negligible when the solder diameter reaches
to 130 μm [59]. In fact, our SAC 305 solder with ~130 μm diameter in size led to
a conclusion that the difference in cooling rates between the two process does not
yield a significant difference in solder grain size, therefore, a similar grain boundary
density between the two samples. No difference in grain size between the two average
grain sizes is explained by the degree of undercooling, in which smaller solder balls
(i.e. ~130 μm in diameter) undercools more than the bigger solder balls [55, 59,
77, 78]. Consequently, grain boundary diffusion kinetics of Cu and Ni atoms would
have been similar. Therefore, grain size is not a determining factor for the premature
EM failure observed for TCB-joints. Rather, additional mechanism might have been
involved that could explain for this phenomenon, such as the effect of IMC and β-Sn
crystallographic orientation on susceptibility to EM failure.
Effects of IMC on EM Degradation. A representative solder joint microstructure
before the EM-aging test are shown in Fig. 10.61. Note that from the top to bottom,
silicon die, UBM, IMC, SAC 305, IMC, ENIG surface finish, Cu bond pad, and
PCB substrate are structurally layered to form an interconnection. Note that before
bonding processes, the IMCs on the chip-side with the solder joint had been already
generated via the bumping process. On the chip-side, IMC thickness is similar for
both samples; however, the low-aspect ratio Cu 6 Sn 5 IMC scallops are present in
reflow joints. This difference originates from the two bonding process difference,
although the IMC morphologies are more obvious at the bond pad interface. When
comparing the morphologies of IMCs on the substrate-side, the reflow joint shows
are uniformly structured and entirely covers ENIG surface (Fig. 10.61a) while the
TCB joint exhibits a significantly different interface, where only certain regions are
discerned with rod-like Cu 6 Sn 5 IMCs (Fig. 10.61b). The difference in thickness of
the IMC and its morphology affects the kinetics of the Cu and Ni atom diffusion
through the SAC 305 solder matrix. As suggested from Bashir’s study [54], there is
a faster diffusion of Cu and Ni from the bond pad and ENIG surface finishes if the
IMCs do not fully cover those areas or if the IMC scallops have high aspect ratio
structures. To better visualize these phenomena, readers are aided with a schematic
representation (Fig. 10.62). In terms of EM effect on these samples, it is likely that
TCB joints would suffer from premature failure due to a much faster Cu and Ni
diffusion. Eventually, the depletion of those atoms will lead to void formation and
propagation, hence delaminating the interface that will short the interconnect.
From our results, IMC morphology was shown to have the considerable difference
between the solder joints reflowed by two different processes (Fig. 10.61). The difference in IMC morphology could have been attributed to the cooling rates between
the two bonding processes. Because the reflow-processed joints endured a much
slower cooling rate, molten solder has more time to wet and form/grow the IMCs
before reaching its solidification temperature [59], in which its layer was eventually
deposited at the substrate side of the solder joint interface. How the layers of IMCs
are structured on the solder-Cu bond pad joint interfaces is vastly essential in understanding both the diffusion of Cu and Ni atoms and Joule Heating that eventually
leads to premature failure upon EM-aging test.