Page 32 - Chip Scale Review_January February_2023-digital

measurements were excluded from the thickness. Figures 4a and b show us
calculations. Based on the remaining cross sections of a bump + mold at the
sampling, for the best-case samples center and edge of a wafer (sample 1).
with granular resin mold, we see a TTV The bump was a bit off-center for both
Table 4: Mold thickness mean, standard deviation, of around 19µm (+/-9.5µm). In addition, cross sections, hence, the difference
and TTV average for different wafer processing skews. while in terms of coverage, some of the in width dimensions between the two
skews show incomplete fill, the TTV bump cross sections. However, the
incomplete fill at the edges with liquid for the rest of the area is controlled mold area clearly showed that for a
mold material, their mold thickness rather tightly (~21µm or +/-10.5µm), on wafer with complete coverage, we can
range wasn’t calculated here. The average, for all the different processing obtain a higher mold thickness at the
mold thickness range over the entire condition skews. wafer center compared to the edge—
wafer gives us the mold total thickness Wafer X-ray scanning electron this is consistent with our estimates
va r iat ion (T T V ) for t he va r ious m icroscopy ( X- SEM ) and SEM from bump height measurements.
processing conditions. The average energy-dispersive X-ray (EDX). SEM measurements were carried
of the data collected over 9 samples Wafer cross-sectioning was carried out using a Hitachi SU3900 SEM
at different processing conditions is out at the center and edge of a wafer microscope. Image and EDX data
shown in Table 4. For incomplete to confirm the adherence of the mold were collected at the center and the
molded samples, out of distribution and provide an estimate of the mold edge of the wafer for the samples.
Wafers with full coverage of the mold
clearly showed solder balls completely
encapsulated by the mold material.
The solder ball itself shows a lack of
mold signal, which suggests minimal
contamination onto the solder ball
during the process (Figure 4c) owing
to the use of the optimized release film
described previously (in the section
entitled, “Materials and experimental
setup”).

Results and discussion
The sections below address the
effects of the liquid vs. the granular
mold, the liquid mold itself, the
effects of different mold cut sizes and
temperature, as well as the effect of
process skews.
Effect of liquid vs. granular mold.
Under standard processing conditions,
most of the g ranula r resi n mold
samples managed to show complete
and uniform fill of the mold compound
from center to edge. The reason for
this is possibly because granular
compound can be dispensed evenly
over the whole mold area (wafer area)
during the pre-compression molding
step as seen in Figure 5d. This, in turn,
reduces the compound flow distance
during the molding process and allows
the compound to fill up to the edge
of the wafer. In cases where filler
size is smaller (e.g., liquid resin), we
see incomplete fill or resin bleed out
(Figure 5b, 5c)
I n compa r ison to the g ranula r
mold, when using a liquid mold the
liquid resin needs to be dispensed in
Figure 4: Mold height at the a) center and b) edge of the wafer; c) SEM microscopy and EDX profile a specific pattern pre-mold. In the
showing solder ball encapsulated by mold material all around. The center and edge both showed comparable final dispense pattern chosen, liquid
encapsulation and minimal contamination.

30
30 Chip Scale Review January • February • 2023 [ChipScaleReview.com]