Page 41 - Chip Scale Review_January February_2022-digital

Figure 5: Simulation of underfill flow propagation.
and 0.3µm Au. Two 5µm polyimide (PI) and 60x60mm have been evaluated. the one that is commonly used for the
layers with two 3µm Cu metal layers One 60x60mm TV sample is used to protection of packages on a printed
were built on a 12” glass wafer for TF- check the feasibility of the surface circuit board (PCB)—it has a low
RDL. Each 10x10mm unit has four mounting process for a sizable 2.2D viscosity for smooth underfilling. After
3x3mm daisy chain blocks that were substrate. The 40x40mm samples are underfilling, the assembly was cured in
populated on a 12” glass wafer. There used to get the warpage information an oven at 150ºC for 15mins.
is lead-free solder on top of each TF- of the 2.2D integrated substrate. To further implement the underfill
RDL pad. The Cu pillars were built The 20x20mm TVs were prepared process for the 2.2D substrate, a good
with a height of 80μm on top of the for the reliability qualification test. simulation tool is needed to optimize
ceramic substrate Ceramic substrates were placed on the selection of copper pillar height,
One of the designs of the 2.2D a 12” glass wafer by pick and place underfill gap and underfill material.
TV pattern used for this study can equ ipment, wh ich ha d a not too A computer aided engineering (CAE)
accommodate substrates underneath demanding placement accuracy of tool from Moldex3D is used for the
the TF-RDL with an incremental size +/-25µm. Later, the assembly went simulation of the underfill process.
of 10mm. Figure 4 demonstrates the t h roug h a conve nt ion al f u r n a ce For accurate simulation, we measure
substrate integration with two different with a peak temperature of 260ºC the underfill material A’s viscosity,
sizes of the ceramic substrate that for solder reflow. curing kinetic and surface tension vs.
can be placed at will on the surface After the 2.2D substrate assembly the process temperature for the data
of the 12” TF-R DL glass wafers. process, underfill is dispensed to input of the model. The simulation
For the 2.2D TV, the joining TF- protect and enhance the bonding of underf ill f low propagation by
RDL to the ceramic substrate with strength of the TF-RDL to the ceramic underf illing material A is shown
dimensions of 20x20mm, 40x40mm, substrate. The underfill material A is in Figure 5. The simulation result
at the end-of-filling state (351s) is
in good agreement with the actual
measurement. In addition, the liquid
underfill moving front will finally
meet at Corner B. The shape predicted
by the model also agrees well with the
experimental result, as indicated in
Figure 6. The fillet at Corner A has
a larger volume than Corner B. After
the underfill dispensing process, laser
debonding is used to remove the 12”
glass carrier wafer. All the ceramic
substrates mounted on the TF-RDL
are singulated into 2.2D integrated
substrates using laser cutting.
Figure 6: Comparison of the simulation of an underfill fillet formation vs. the experimental result.

39
Chip Scale Review January • February • 2022 [ChipScaleReview.com] 39