Page 40 - ChipScale_May-June_2020-digital

Demand for 3D devices drives inspection

and metrology innovation

CSR asked Raj Jammy, President of the Process Control Solutions Business Unit at Carl Zeiss SMT
group and President of Carl Zeiss SMT Inc., the North America subsidiary of Carl Zeiss SMT GmbH,
to provide an update on how demand for 3D devices is driving inspection and metrology innovation.

CSR: 3D devices drive a good deal scanning electron microscopy (SEM) RJ: Generally, advanced packages are
of semiconductor technologies today imaging to FIB prep and TEM imaging, used for the advanced Si nodes that use
and in turn, such devices need to have today we see the cycle repeat with package fragile dielectric materials in the back end
3D integrated information when it fine pitches driving a transition from of line (BEOL) Si interconnects. To achieve
comes to 3D inspection and metrology. traditional cross-sectional methods to new fine pitches, Cu-pillar solder bumps are
Why is imaging alone not sufficient for cross-section approaches with the speed being adopted for advanced package
these devices? and accuracy for package interconnects at interconnects. Cu is a hard material, and
RJ: In the pursuit of root cause of 50 micron and finer pitches. the combination of low-modulus brittle
failures and process defects, structural materials and hard materials can lead to
information tells only part of the story. CSR: What other considerations come cracks and delaminations, whether from
Something may look str ucturally into play with respect to cross-sectional handling steps of sample preparation or
sound, but be electrically or chemically methods for advanced packages? from reliability failures caused by chip-
defective. For example: often one
wants to do energy dispersive X-ray
spectroscopy (EDS) analysis of a focused
ion beam (FIB) cross section before
final transmission electron microscopy
(TEM) prep and electron energy loss
spectroscopy (EELS) analysis. It can
be important to understand if any
oxidation present at a defect site occurred
before exposure to atmosphere, or is a
side effect from exposure during the
transfer from FIB to TEM. 3D device
architectures, indeed, pose significant
challenges for all three types of analyses,
because of shrinking features that are
buried beneath the surface and often not
accessible for testing.
Figure 1: Crossbeam Laser FIB-SEM throughput for each step of the workflow for a customer’s application.
CSR: With respect to shrinking
features and imaging, how is cross-section
accuracy impacted?
RJ: Traditional package cross-section
approaches have an accuracy in the range
of about 15 microns, and it can take two
days or longer to prepare a high-quality
advanced-package cross section at exactly
the right location with minimal artifact. As
package interconnects continue to shrink,
the ability to cross section to exactly the
right location becomes less successful.
We are dealing with technologies today
in packaging that require nanometer
resolution for micron-sized objects located
within millimeters of volume. Just as
shrinking semiconductor dimensions
in the 1990s drove the transition from Figure 2: Crossbeam Laser FIB-SEM provides fast, high-quality cross sections of Cu-pillar microbumps buried
760µm deep with total time to results of <1 hour.

38
38 Chip Scale Review May • June • 2020 [ChipScaleReview.com]