inspection limits using a DUV-laser-based bright-field inspection tool
Raccah, Daniel Rogers, Ehud Tzuri, and Arik Shavit, Applied Materials
IC feature sizes continue to shrink well below the wavelength
of visible light, there is an increasing need for high-resolution, high-throughput
inspection tools to accelerate process development and optimize production.
At the 90-nm technology node and beyond, characterizing shrinking yield-critical
defects requires the detection of the new defect types that result from
the integration of new materials and processes. These requirements lead
to a natural trade-off between sensitivity and throughput. The increasing
cost of inspection suggests that new inspection approaches are needed.
deep-ultraviolet (DUV) laser-based bright-field inspection technology
developed by Applied Materials (Santa Clara, CA) was used to characterize
two advanced DRAM process modules. The inspection method enabled engineers
to address specific processing issues rapidly. This article focuses
on inspection strategies for three DRAM layers. In addition, the article
compares current optical inspection tools and correlates tool data results
with end-of-line probe results.
bright-field wafer inspection tools, which were introduced more than
20 years ago, have traditionally used lamps as illumination sources.
While these tools have evolved to become industry standards, they face
limitations as chipmakers begin to address the needs of 65-nm processes and below. The achievable
illumination levels of lamp-based bright-field inspection tools are
inherently limited. Time-delay-integration cameras can compensate in
part for the low signal levels of traditional bright-field tools, but
they too are limited in the low-light-flux regime. As a result, the
instruments' sensitivity for materials that are susceptible to light
damage or lack reflectivity can degrade substantially.
DUV-laser-based inspection tool discussed in this article offers simultaneous
bright-field and 3-D-channel imaging to perform in-line inspection of
advanced processes. With a photomultiplier tube (PMT) detector, the
tool operates at high resolution with a throughput of several wafers
per hour, making it suitable for production applications. To assess
the instrument's capabilities, tests were conducted on three layers
from advanced DRAM technology nodes. Containing challenging defect types,
the layers selected were shallow-trench isolation (STI), contact etch,
and metal line after photoresist developing. For each of the layers,
the challenges facing the inspection technique and the methodologies
that were developed to characterize the defects are discussed.
conduct the evaluation, many wafers with the targeted layers were inspected.
From previous probe-test data and failure analyses, specific difficult-to-detect
defect types were selected. Inspection recipes were created for optimum
sensitivity to the defects of interest. The recipes were subsequently
assessed for throughput and false rate to determine whether the inspection
recipe needed to be modified. For all three layers, the recipes met
the throughput requirement specifications, so sensitivity did not have
to be compromised.
tool's sensitivity and throughput performance can be attributed to its
design, which uses DUV-laser illumination, simultaneous bright-field
and 3-D-channel imaging, and a photomultiplier (PMT) module for light
collection. The combination of laser illumination and PMT detection
enables the system to maintain high detection sensitivity, even in low-light
situations. The bright-field channel is sensitive to shallow-pattern
defects, while the 3-D channel offers high sensitivity to defects with
some topography, such as particles and voids.
system's ability to collect signals from the bright-field and the 3-D
channels at the same time enables it to detect a broad range of defects
in a single scan. In addition, it can efficiently perform simultaneous
cell-to-cell and die-to-die inspection in one wafer pass, offering excellent
sensitivity across different regions of the die without requiring multiple
passes for different inspection conditions. Advanced filters and algorithms
were employed to suppress wafer noise, achieving a high signal-to-noise
ratio. Real-time nuisance filtering was also used to amplify defects
of interest, enabling efficient scanning electron microscope (SEM) defect
review and accelerating defect root-cause analysis.
1: STI Layer. The chemical-mechanical polishing (CMP) process
used in the formation of STI features can introduce particle contaminants
and surface scratches. In addition, void defects caused by inherent
oxide problems often emerge after CMP. STI void defects are known to
cause functional yield failures and reliability issues by inducing direct
shorts between consecutive active areas, as illustrated in Figure 1,
or two poly lines, as shown in Figure 2. If such void defects are not
detected in-line at an early stage of the transistor-formation sequence,
their presence may not be discovered until final probe test. During
that period of days or weeks, all the material in the production pipeline
can be placed at risk.
1: Schematic diagram of an oxide void shorting two active areas.
2: Schematic diagram of an oxide void shorting two poly lines.
performance of the STI module was evaluated on products being fabricated
at both the 95- and 80-nm technology nodes. While other optical inspection
systems were being used to detect void defects, the researchers found
that the laser-based bright-field tool could detect voids as small as
30 nm. Since STI voids are typically associated with surface topography,
their signal-to-noise level was improved dramatically by employing the
3-D channel. A proprietary spatial filtering method, optimized for the
repetitive small-pitch patterns that are typical in DRAM devices, was
used to eliminate pattern noise and further enhance the defect signal.
inspection technology was highly sensitive to void defects down to 30
nm for both the 95- and 80-nm products. Figure 3a presents a laser bright-field
wafer map of void defects at the STI inspection step, while Figure 3b
shows a SEM image of a void. The tool's high detection capacity led
to a successful reduction of overall defect levels. The tool's detection
accuracy was confirmed by SEM review to be 97%. In addition, tool repeatability
was measured at 98%.
3: Laser-based bright-field wafer map (a) showing void defects from
the STI inspection step. Typical defects of interest (b) include
particles, scratches, and voids.
2: Contact Etch after Clean Layer. The second critical layer
that was inspected during these tests was contact etch immediately after
a cleaning step was performed. At that level, typical defects of interest
include overetch contacts, blocked contacts, and deformed contacts.
This study focused on detecting overetch contacts on a contact etch
layer in 80-nm devices. Overetched contacts create a direct short between
the contact and the word line.
test parameters used to inspect wafers for overetched contacts were
similar to those established in the STI layer study. Using a combination
of the high-resolution 3-D channel, the bright-field channel, and a
unique spatial filtering mask to block the diffraction lobes, the inspection
tool detected defects with a distinct wafer-level signature, as illustrated
in Figure 4.
4: (a) Laser-based bright-field contact etch wafer map showing a
clear defect signature on the wafer edge, and (b) SEM image of an
effect of overetch on end-of-line yield was then determined by correlating
laser bright-field inspection maps to an end-of-line probe map. Laser
bright-field and probe maps from two wafers used in the study are presented
in Figures 5a and 5b, respectively. Occurring primarily at the wafer
edge, the die that failed the probe test appear black in Figure 5b.
Overetched contacts were also clustered at the wafer edge. Similar to
STI layer inspection, contact etch after clean inspection detected defects
of interest with a <3% false-alarm rate and >95% repeatability.
5: Edge-shading signature resulting from overetched contacts: (a)
laser-based bright-field map, and (b) end-of-line probe map. The
bright-field signature correlated strongly to the probe results.
3: Metal Line after Photoresist Developing. The third process
targeted in this evaluation was the metal line after photoresist developing
but before metal etch. The ability of the inspection tool to isolate
and detect photoresist microbridges in an after-develop inspection (ADI)
step enables engineers to make process adjustments that can prevent
the migration of defects to later process steps. More importantly, the
wafer can be reworked as long as it has not undergone etch.
metal line layer after photoresist developing is known for having a
variety of defect types, including full and partial bridging defects,
pattern defects, peeling (delamination), residues, bumps, particles,
and embedded particles. Some of these defects may have originated at
previous layers, making it difficult to find and focus on defects of
interest quickly. In this study, the researchers identified microbridges
as their highest priority. Their goal was to detect isolated micro-bridges
and review them using a SEM tool. For efficient SEM review, the inspection
results should ideally include defects of interest while ignoring nuisance
defects on ADI layers is challenging for optical inspection tools that
use visible light, because the percentage of light transmission increases
with wavelength. Transmitted visible light scatters from the underlying
pattern and generates noise from the previous layer, especially in the
case of rough and grainy metal films. Two fundamental features of the
laser bright-field technology help avoid this problem.
first feature is the tool's DUV wavelength. As shown in Figure 6, the
DUV antireflective coating (DARC) layer associated with the photoresist
strongly absorbs DUV light at 266 nm, effectively eliminating unwanted
signals that originate from underlying layers. The second feature is
simultaneous signal detection in the bright-field and 3-D channels.
With that arrangement, it is possible to achieve high sensitivity for
detecting a wide range of defect types in a single scan.
6: Photoresist DARC light absorption versus wavelength. A much higher
percentage (80%) of DUV (266-nm) light is absorbed than UV and visible
metal line layer after photoresist developing was evaluated at the 95-nm
technology node. The tool's inspection range was optimized to cover
the full wafer, including both array and periphery regions. The array
region, in which microbridges were abundant, was scanned using the system's
cell-to-cell mode with the 3-D channel. The wafer's periphery area,
which suffered from peeling defects, was scanned using the tool's die-to-die
mode with the bright-field channel. The schematic diagram in Figure
7 highlights the inspection strategy. The system's ability to scan the
array and periphery areas of the wafer simultaneously without compromising
sensitivity in either area represents a throughput advantage over other
bright-field tools. More importantly, the tool was highly sensitive
to photoresist microbridge defects, as shown in the wafer maps in Figure
7: Schematic diagram (a) depicting how simultaneous bright-field
and 3-D-channel detection is used to detect microbridges (b) and
peeling defects (c) in a single scan.
8: Metal line ADI wafer maps from two types of lamp-based bright-field
tools (a and b) and a laser-based bright-field tool (c). The laser-based
wafer image (d) shows a typical photoresist array bridge.
SEM review confirmed that the laser-based system is more sensitive to
microbridges than other optical inspection tools and returns fewer reports
of nuisance defects. The system's ability to filter out nuisance defects
and selectively report defects of interest provides a more reliable
process indicator and enables faster feedback than other tools.
the 95- and 80-nm technology nodes evaluated in this study, a laser-based
bright-field technology was able to detect and resolve yield-limiting
defects. Contact etch inspection results also correlated nicely to final
wafer yield data. The system achieved high detection sensitivity without
lowering throughput or missing defects of interest. The system's detection
capability can be attributed to several key technologies: First, a 266-nm
laser source and an efficient PMT detector render the tool sensitive
to small and challenging defects; second, simultaneous bright-field
and 3-D-channel imaging enables users to detect all critical defect
types in a single scan; and third, effective filtering strategies reduce
the rate of nuisance defects.
forward, given its sensitivity and throughput, laser-based bright-field
technology promises to be extendable to the 65-nm technology node and
beyond, since it detects increasingly small defects as well as critical
defects in complex structures.
authors would like to thank Justin Arrington and Pete Frank of Micron
Technology; and Noam Shachar, Ido Hammer, Silviu Reinhorn, and Ofir
Montal of Applied Materials' PDC group.
YH. "Effect of UV Inspection on Defect Management for
Advanced Devices." Paper presented at the Semicon Europa
Yield Management Solution Seminar. April 24–26, 2001. Munich.
L, M Cheng, and T Han. "Detecting and Classifying DRAM Contact Defects
in Real Time." MICRO 21, no. 7 (2003): 71–75.
S, et al. "UVision for Advanced DRAM Process Monitoring."
Paper presented at the Applied Materials Engineering Technology Conference.
May 24–26, 2004. Santa Clara, CA.
Raccah is a product marketing manager
in the wafer inspection division at Applied Materials Process Diagnostic
and Control business group. In the past five years, she has served in
various positions in the PDC division, including as a technologist and
an application development engineer. Before joining Applied Materials,
Raccah worked at Intel Electronics, Israel, as an etch process engineer.
She holds a BS in chemical engineering and an MS in technology management
from the Hebrew University in Jerusalem. (Raccah can be reached at firstname.lastname@example.org.)
Rogers is a wafer inspection process engineer at Applied
Materials. He received a BS in physics from Boise State University in
Idaho. (Rogers can be reached at email@example.com.)
Tzuri is the general product manager of the inspection division
at Applied Materials. He began his career as a wafer inspection application
engineer and then held positions in product and marketing management.
For the last four years, Tzuri has been responsible for the management
of all marketing activities in the inspection division. Previously,
he held key positions at Orbot-Instruments. He received a BS in electronics
and computer engineering from the Ben-Gurion University in Beer-Sheva,
Israel. (Tzuri can be reached at firstname.lastname@example.org.)
Shavit is an application development engineer in the wafer
inspection division at Applied Materials Process Diagnostic and Control
group. He has served in this position for the last two years. He received
a BS in electrical and electronics engineering and a BS in physics and
astronomy, both from Tel Aviv University in Israel. (Shavit can be reached