4 Microstructure and Mechanical Reliability Issues of TSV
89
Fig. 4.18 SEM micrograph showing delamination of the capping layer at the top of a row of TSVs
following thermal excursions
have little if any effect, but the pumping is greater on average for larger via diameters
[50, 56]. However, although the average protrusion of an array of TSVs is larger for
larger TSVs, the maximum copper pumping appears to be independent of the TSV
diameter, which suggests that TSV diameter has little effect on BEOL reliability [56].
Copper overburden typically has no effect, but a higher pre-CMP annealing temperature (∼430 °C) reduces pumping. Finally, an additional annealing step, following
pre-CMP anneal, reduces pumping significantly [50].
Recent work has demonstrated that the surface relief that is often seen on the endsurface of the TSVs after thermal cycling (e.g., Figure 4.12b), is associated with grain
boundary sliding along incoherent 3 boundaries of the copper [60]. As noted earlier,
EBSD typically reveals no strong preferred orientation (texture) in the Cu-TSV, along
with the presence of numerous high angle grain boundaries. In particular, the presence
of 3 boundaries (both coherent and incoherent) are particularly prolific in the
electroplated Cu TSVs. However, the incoherent 3 boundaries with misorientations
of ~59
o (e.g., the boundaries between B and C, C and D, and D and E in Fig. 4.19)
are the ones susceptible to sliding because of their high interfacial energy [60]. The
coherent 3 boundaries, which have much lower energy, do not slide. This implies
that if the electroplating process for TSVs can be controlled to yield only coherent
3 boundaries near the top surface, copper-pumping due to boundary sliding can be
largely eliminated.
4.3.1.4 Heating Rate Dependence of Copper Pumping
The extent and mechanisms of copper-pumping during thermal cycling depend on
the heating/cooling rate, as well as the range of temperatures over which the package
is thermally cycled. The influence of heating rate during in situ heating has been
reported [60] on dies with 200 µm pitch TSV arrays, where the samples were rapidly
heated at a rate of 0.1 °C/sec to 300 °C, and then heated further to 425 °C at various
heating rates.
No change in the TSV surface occurs up to 300 °C, but during holding at 425 °C
for 90 min, the existing surface features start protruding. As shown in Figs. 4.20 and
4.21, little change occurs during holding at 425 °C following heating at 0.02 °C/sec;
89
Fig. 4.18 SEM micrograph showing delamination of the capping layer at the top of a row of TSVs
following thermal excursions
have little if any effect, but the pumping is greater on average for larger via diameters
[50, 56]. However, although the average protrusion of an array of TSVs is larger for
larger TSVs, the maximum copper pumping appears to be independent of the TSV
diameter, which suggests that TSV diameter has little effect on BEOL reliability [56].
Copper overburden typically has no effect, but a higher pre-CMP annealing temperature (∼430 °C) reduces pumping. Finally, an additional annealing step, following
pre-CMP anneal, reduces pumping significantly [50].
Recent work has demonstrated that the surface relief that is often seen on the endsurface of the TSVs after thermal cycling (e.g., Figure 4.12b), is associated with grain
boundary sliding along incoherent 3 boundaries of the copper [60]. As noted earlier,
EBSD typically reveals no strong preferred orientation (texture) in the Cu-TSV, along
with the presence of numerous high angle grain boundaries. In particular, the presence
of 3 boundaries (both coherent and incoherent) are particularly prolific in the
electroplated Cu TSVs. However, the incoherent 3 boundaries with misorientations
of ~59
o (e.g., the boundaries between B and C, C and D, and D and E in Fig. 4.19)
are the ones susceptible to sliding because of their high interfacial energy [60]. The
coherent 3 boundaries, which have much lower energy, do not slide. This implies
that if the electroplating process for TSVs can be controlled to yield only coherent
3 boundaries near the top surface, copper-pumping due to boundary sliding can be
largely eliminated.
4.3.1.4 Heating Rate Dependence of Copper Pumping
The extent and mechanisms of copper-pumping during thermal cycling depend on
the heating/cooling rate, as well as the range of temperatures over which the package
is thermally cycled. The influence of heating rate during in situ heating has been
reported [60] on dies with 200 µm pitch TSV arrays, where the samples were rapidly
heated at a rate of 0.1 °C/sec to 300 °C, and then heated further to 425 °C at various
heating rates.
No change in the TSV surface occurs up to 300 °C, but during holding at 425 °C
for 90 min, the existing surface features start protruding. As shown in Figs. 4.20 and
4.21, little change occurs during holding at 425 °C following heating at 0.02 °C/sec;