The Role of Metrology And Inspection In Semiconductor Processing

The Role of Metrology And Inspection In Semiconductor Processing

Loading
Loading Social Plug-ins...
Language: English
Save to myLibrary Download PDF
Go to Page # Page of 46

Description: Metrology and inspection systems can be broadly separated into three main classifications by application: critical dimension (CD) and overlay measurements, particle and pattern defect detection and thin film parameter measurement (such as resistivity, thickness, and stress). The typical processing steps and metrology and inspection equipment used to monitor and/or control them.

 
Author: Mark Keefer, Rebecca Pinto, Cheri Dennison, James Turlo  | Visits: 352 | Page Views: 782
Domain:  High Tech Category: Semiconductors 
Upload Date:
Link Back:
Short URL: https://www.wesrch.com/electronics/pdfEL1SE1000RKKF
Loading
Loading...



px *        px *

* Default width and height in pixels. Change it to your required dimensions.

 
Contents:
Chapter 6: Metrology and Inspection

241

6
The Role Of Metrology
And Inspection In
Semiconductor
Processing
Mark Keefer, Rebecca Pinto, Cheri Dennison,
and James Turlo

1.0

OVERVIEW

As integrated circuits (IC) are incorporated into more and more
products, the market demand for lower cost, higher performance devices
continues to grow. In order to design and manufacture a high performance
integrated circuit cost-effectively, the parameters of the manufacturing
process need to be carefully controlled: film thicknesses and material
properties must be accurate, uniform and controlled; linewidths and edge
profiles must fall within tight limits, and the devices need to be free of defects
that affect yield.
Thin film metrology and wafer inspection for defects are integral to
controlling the semiconductor manufacturing process. Film properties,
linewidths, and defect levels need to be measured, first to optimize the
manufacturing process, then later to ensure that it is operating under control.
241

242 Thin-Film Deposition Processes and Technologies
This chapter explores the subjects of metrology and inspection of
integrated circuits. After the introduction, implementation strategies for
metrology and inspection are examined from a historical perspective. Then,
as we anticipate increasingly complex devices having critical dimensions of
0.18 and 0.13 µm, manufactured on 300 mm wafers, we look at how
metrology and inspection will evolve to meet these measurement challenges,
while simultaneously meeting increasing pressure for automation, higher
throughput and higher reliability. In the final section we provide a technology
reference that discusses theory of operation, equipment design principles,
main applications, and strengths and limitations of the metrology and
inspection systems. The sections are organized as follows:
1.0 Overview
2.0 Introduction to Metrology and Inspection
3.0 Metrology and Inspection Trends: Past, Present and
Future
4.0 Theory of Operation, Equipment Design Principles,
Main Applications, and Strengths and Limitations of:
4.1 Film thickness measurement systems
4.2 Resistivity measurement systems
4.3 Stress measurement systems
4.4 Defect inspection systems
4.5 Automatic defect classification
4.6 Defect data analysis systems
2.0

INTRODUCTION TO METROLOGY AND INSPECTION

Metrology and inspection systems can be broadly separated into three
main classifications by application: critical dimension (CD) and overlay
measurements, particle and pattern defect detection, and thin film parameter
measurement (such as resistivity, thickness and stress). The typical processing steps, and metrology and inspection equipment used to monitor and/or
control them, are given in Table 1.
In the semiconductor industry, the continual demand for denser
integrated circuits with higher performance and higher speeds drives technological advances in all facets of manufacturing. A key to the success of
semiconductor processing is an understanding of the chemical, mechanical
and kinetic properties of the wide range of materials used to make a typical
circuit.

Chapter 6: Metrology and Inspection

243

Table 1. Typical Metrology and Inspection Parameters Monitored, by
Process Step
PROCESS STEP
USED

MEASURED ATTRIBUTE

METROLOGY

SYSTEM

Si manufacturing

resistivity

4-point probe, eddy current

Inspection of incoming
wafers

flatness
defects:
particles
micro-scratches
crystalline defects
haze

flatness tester
defect inspection system

Si epitaxy

resistivity
thickness

4-point probe, eddy current
FTIR

Conductor deposition
(PVD, MOCVD)

resistivity
particulate contamination

4-point probe, eddy current
defect inspection system

Dielectric deposition
(CVD)

thickness, RI
stress
particulate contamination
dielectric constant

reflectometer, ellipsometer
stress gauge
defect inspection system
C-V tester

Dopant processes
(ion implant, diffusion)

uniformity
depth profile

4-point probe, thermal wave, FTIR
SIMS, SRP

Planarization

removal rate and uniformity
local/global planarity
slurry particles,
micro-scratches

reflectometer, ellipsometer
surface profiler (high resolution)
defect inspection system

Etch

removal rate and uniformity
etch selectivity
etch profile
particulate contamination
pattern defects

reflectometer
reflectometer
SEM, AFM
defect inspection system
defect inspection system

Lithography

critical dimension
overlay
pattern defects
particulate contamination

SEM
optical overlay tool
defect inspection system
defect inspection system

Yield monitoring

correlation of metrology and
inspection results to yield

fab-wide data management
system



244 Thin-Film Deposition Processes and Technologies
To maintain control of the product, thin film quality is regularly
monitored by tools that measure electrical, physical, optical and mechanical
properties of films. Film parameters typically monitored include thickness
and refractive index, resistivity and stress.
Another key area for process control is the reduction of defects that
affect production yield. Defect reduction is typically achieved through an
iterative process that involves detection of defects, classification of defects,
identification of the source of the yield-limiting defects, correction of the
process to eliminate or reduce the defect mechanism, then monitoring the
process for yield excursions. The process is iterative in that, while the process
is monitored for defect excursions, further defect analysis is conducted in
parallel to drive continuous improvement of the yield.
Defect reduction is employed for five main applications:
• Inspection of bare wafers for contamination and surface
quality, either during the wafer manufacturing process or as
incoming material at the IC manufacturer
• Inspection of bare wafers used to monitor the cleanliness of
process (or metrology) equipment
• Inspection of product wafers to decrease defectivity during
the early phases of the product life cycle, and throughout the
life cycle for continuous improvement
• Inspection of product wafers to monitor processes that introduce
contamination, scratches or pattern defects
• Prediction of yield
Specific tool sets have been developed to address each of these needs.
Bare wafer or unpatterned wafer inspection systems are optimized for
scanning wafers at high throughput without the constraint of coping with the
interference of pattern signals. Inspection of patterned wafers for yield
improvement or process monitoring requires a system that not only copes
with pattern, but provides high capture of key defect types at reasonable
throughput. For yield prediction, the ability to integrate defect information—
number, type, spatial signature—with other parameters such as electrical test
results, becomes the key. Theory of operation as well as equipment design
principles for each of these categories of defect inspection systems are
described in Sec. 4.0 of this chapter.

Chapter 6: Metrology and Inspection
3.0

245

METROLOGY AND INSPECTION TRENDS: PAST,
PRESENT, AND FUTURE

Metrology systems have undergone tremendous changes since the
home-built bench top characterization tools of the 1960s. Inspection systems
have gone through similar changes, moving from manual inspection by
operators to automated inspection tools. As the semiconductor industry has
grown and matured, metrology and inspection systems have kept pace.
Measurement performance—predominantly sensitivity and repeatability—
has steadily improved. The level of automation has dramatically increased,
beginning with automated wafer handling, then pattern recognition, remote
operation through development of the Semiconductor Equipment Communication Standard (SECS) protocol, development of automated algorithms that
“learn” best measurement setups for monitoring a given product and process,
and now automatic defect classification. These developments have supported the growing practice of making measurements on product wafers
(rather than designated monitor wafers), a practice driven by the increase in
wafer diameter to 200 mm, with a consequent rise in monitor wafer cost.
Now, increasing attention is being paid to reliability, up-time and ease-of-use
attributes, with the goal of increasing overall effectiveness of the equipment.
In the future, price-performance pressures on IC manufacturers will
continue to be passed on to equipment manufacturers. The shift from off-line
sampling to on-line control will accelerate, with increasing use of in-line and
in situ measurements. Reliability and ease-of-use emphasis will drive the
implementation of integrated, automated systems for measurement optimization, data interpretation, and adaptive feedback to the process equipment.
3.1

Trends in Metrology

Thin film measurements have progressed from simple single layer
thickness measurements to multiple layer thickness and refractive index
measurements. The trend towards multiple layer measurements has been
partially driven by the increasing use of cluster deposition systems, where no
opportunity exists for single layer measurements. Additionally, economics
plays a role, in that as wafer sizes increase, cost savings can be realized by
reducing the use of monitor wafers. In many cases, measurement of a layer
on a product wafer requires measurement of the underlying layer(s) as well.

246 Thin-Film Deposition Processes and Technologies
Measurement of film quality and stoichiometry has become as important as film thickness control. This is especially true with the advent of plasma
enhanced CVD films such as silicon-rich anti-reflective oxynitride layers,
since the film properties are strong functions of the process parameters such
as plasma energy and reactant flow rates.[1]
At the same time, higher accuracy and tighter system-to-system
matching are required to facilitate process transfer and reduce the time to
start up new wafer fabs.
In the future, significant changes in the approach to process monitoring
will occur. The trend will be to monitor processing parameters inside the
processing chamber, not film characteristics; that is, process variables, not
product variables. The shift from off-line sampling to on-line control will
continue, with increasing use of in-line and in situ measurements. Prevention
of process excursions by process control sensors should significantly reduce
product loss.[2] Sensors can be equipment state (mechanical, electrical),
process state (chemical/physical, temperature) or wafer state (product
parameter, uniformity).
3.2

Trends in Defect Inspection

In the 1980s the first inspection systems were entirely manual and thus
operator-intensive. Typically operators used a bright light source, and
conducted visual inspections of incoming silicon, then manual microscope
inspection of production wafers at various points in the process. Data was
only as accurate and repeatable as the operators themselves. Results from
these inspections were commonly written on paper and stored in binders and
cabinets. Operators manually correlated defects to yield using two sheets of
transparency paper for map-to-map comparisons. Over the intervening years
advancements in IC technology, as well as economic pressures, have driven
the need for enhancements in sensitivity, throughput, repeatability and
automation for the defect reduction process.
In this section, we examine a few key trends in defect inspection:
• How the sensitivity of defect inspection systems has tracked
critical dimensions of ICs—and how these systems are predicted
to meet sensitivity needs for inspection of 0.18 and 0.13 µm
devices, and beyond
• Challenges specific to inspecting 300 mm wafers

Chapter 6: Metrology and Inspection

247

• Shifts in inspection strategies: from monitoring processes using
bare wafers to using product wafers; towards differentiating
yield-learning from in-line inspection; and looking forward to
in situ inspection and adaptive process control
• Increasing automation of inspection equipment using automated
wafer handling, automated data transfer, mini-environments,
and remote operation
• The movement towards automated systems for measurement
optimization, data interpretation, and adaptive feedback to the
process equipment
• The growing emphasis on cost of ownership, overall equipment
effectiveness, and ease of use
Sensitivity Challenges as Critical Dimensions Decrease. A good
rule of thumb in defect inspection is that the critical dimension (smallest
linewidth) of the device determines the minimum size of defect likely to
affect yield. At early steps in the process, defects as small as one-third the
critical dimension can cause an electrical failure; at back-end levels, detecting
defects about as large as the CD is sufficient to protect yield. Until recently
DRAM devices had the smallest critical dimensions of any device on the
market; at present, logic devices also have leading-edge critical dimensions.
Historically manufacturers of inspection equipment watched DRAM manufacturers closely, striving to stay ahead of the design rule of the next memory
device, so that the inspection system would be able to detect yield-limiting
defects. Today we have various industry consortia, and in particular, the
National Technology Roadmap for Semiconductors,[3] to guide metrology
and inspection equipment manufacturers in the design of next-generation
inspection systems.
Figure 1 shows a history of how the detection limit of inspection
systems has kept ahead of the critical dimension of IC devices. Unpatterned
wafer inspection systems currently on the market can now detect defects as
small as 80 nm, whereas patterned-wafer systems can detect defects less
than 100 nm.* The advances that have allowed inspection systems to
continue to improve performance include:

* Quoted detection limits are referenced to NIST-traceable polystyrene latex spheres
of known size deposited on clean, well-polished bare silicon wafers.

248 Thin-Film Deposition Processes and Technologies
• New optical designs
• More powerful and/or shorter-wavelength light sources
• More accurate wafer stages
• Higher resolution cameras
• Better signal processing algorithms
These evolutionary changes are ongoing for optical-based inspection
systems, and the existing technology is proving to be extendible for inspection
of 0.18 and 0.13 µm devices.

Figure 1. Critical dimension of integrated circuits (upper line) determines detection limit
specifications for defect inspection systems (wide bands). Historically, smaller defects have
been detectable using unpatterned systems, where the additional challenge of coping with
pattern signal is not present.

An inspection technology based on scanning electron microscopy
(SEM) is being used in addition to optical-based inspection systems for
development of 0.25 and sub-0.25 µm processes. The biggest challenge for
such high sensitivity inspection is providing cost-effective throughput. Despite this challenge, SEM-based automated inspection systems are used in
advanced IC lines throughout the world.
SEM-based automated inspection systems provide two unique defectdetection capabilities. The first results from a SEM’s high resolution and
large depth of focus: these systems can find small defects hidden in dense
geometries where they can not be seen by optical microscopy. The second
unique capability is the result of new, nontraditional SEM designs that

Chapter 6: Metrology and Inspection

249

enable a properly optimized system to see contrast differences in electrically defective IC features. A common example of this type of defect is an
electrical fault caused by voids in the metal that fills a contact or via. Here the
structure of the metal plugs appears correct when viewed from the top
surface, but a cross-section reveals that the metal doesn’t fill the bottom of
the contact hole. In this and similar cases an electron-beam inspection tool
that is designed to maximize the charge-induced voltage contrast effect may
detect the problem as a slight difference in contrast of the feature in the SEM
image. This voltage contrast imaging ability has opened up new applications
for automated defect inspection tools. For further information on this
capability, see Ref. 4.
Inspection Challenges for 300 mm Wafers. As pilot lines are
coming up for production of devices on 300 mm wafers, inspection
systems that accommodate these wafers are entering the market as well.
Unpatterned inspection systems for 300 mm wafers have been on the
market since early 1997, as these are required well in advance of production to allow process equipment vendors to develop their 300 mm equipment. Patterned wafer inspection systems are expected to be introduced in
1998. The main challenge for inspecting 300 mm wafers is cost of
ownership; particularly the increased footprint of the system, and the
challenge of maintaining high throughput. Compared to a 200 mm wafer,
a 300 mm wafer has an area 2.25 times as large, and for most equipment
designs, throughput scales roughly linearly with inspected area. Maintaining the same wafer-per-hour throughput specifications for 300 mm as
currently available for 200 mm wafers could be achieved in one of several
ways:
• Evolutionary improvements to subsystems including faster
data rates, faster scanners, less reliance on scanning the stage
since its mass is relatively large.
• Revolutionary new scanning designs, such as the spinning
wafer strategy currently employed on some high throughput
unpatterned wafer inspection systems.
• Adapting the sampling strategy to inspect a sparser fraction
of the wafer area.
The high cost of 300 mm wafers also exerts economic pressure for
devices to be built all the way to the edge of the wafer, and thus inspection
closer to the wafer edge is necessary to protect yield.

250 Thin-Film Deposition Processes and Technologies
3.3

Trends in Inspection Strategies

Shift of Process Monitors from Bare Wafers to Production
Wafers. Monitor wafers have been used widely for tool qualification and
process monitoring throughout the IC manufacturing process. Bare wafers
have been run on every shift to qualify the equipment for use. In recent years,
focused contamination reduction efforts on the equipment have enabled a
better understanding and control of the contribution of the process equipment
to contamination of product wafers. Defect types and mechanisms are better
characterized, and programs for cleaning or preventive maintenance of the
equipment are in place. Thus the need for such rigorous equipment monitoring is predicted to decrease.
The cost of monitor wafers has always been significant, and has
become more so with the introduction of 200 mm wafers and the imminent
move to 300 mm. Thus, more fabs have begun to relegate the use of monitor
wafers to bringing up new tools and diagnosing specific contamination
problems, while using product wafers to monitor their processes.
Interestingly, the predicted decline in use of monitor wafers has not
been reflected in declining sales of unpatterned wafer inspection equipment.
The rapid expansion in the semiconductor industry during the 1990s has
driven strong growth in unpatterned wafer inspection system sales, dominating any effects of decreasing monitor wafer use.
Looking Ahead to In Situ Inspection and Adaptive Process
Control. A logical extension to process monitoring is to incorporate the
inspection system into the process tool itself. An in situ inspection system
could provide information to the process tool, so that when defect excursions
are detected, the process tool could be flagged and shut down, or perhaps
even adjusted to eliminate the defect mechanism. Having the inspection tool
provide closed-loop control of the defectivity of the process is an example of
adaptive process control.
The barriers that currently exist for achieving this scenario are significant. At present, the capability of a stand-alone inspection system would be
difficult to reproduce inside the economic and physical constraints of a
process chamber. Also, understanding how to adapt a process to eliminate
the defects detected by the inspection system is nontrivial for a team of
experienced engineers. Designing an expert system to replace that body of
knowledge would be a significant challenge. However, strong economic
pressures exist to reduce the cost of the defect reduction process, and part
of the solution may involve meeting the challenges of in situ inspection.

Chapter 6: Metrology and Inspection

251

An alternative approach—and another example of adaptive process
control—is to provide tighter control of the process parameters through
improvements to in situ environmental sensors. This more direct, causal
means of addressing process control issues involves rapid measurement and
feedback based on process parameters, e.g., temperature and pressure. In
contrast, an in situ inspection system measures the effects of out-of-control
process parameters: defects incurred on the product wafer.
Trends in Automation of Wafer Inspection Equipment. An ongoing trend for IC inspection has been a growing emphasis on automation of the
inspection process: automating the inspection and defect review equipment
itself, and integrating it with yield data using defect data management
systems. The cost of ownership of the inspection process decreases as
automation is introduced, because trained engineers and operators can focus
elsewhere in the fab. Repeatability and accuracy increase as the subjective
nature of human judgment is replaced by standard algorithms. Automation
facilitates the de-localization of a given manufacturing process, allowing the
process to be copied exactly from fab to fab around the world. Automation
can also support a more rapid return on investment by helping ramp a process
to yield in a shorter time.
Automation began with the introduction of automated wafer handling
using mechanical robots, and has expanded its scope to include other
subsystems within the wafer inspection and defect review tools. In an effort
to reduce operator error and increase throughput, Optical Character Recognition (OCR) and Bar Code Reading (BCR) were incorporated into these
tools to read lot and wafer information during automatic inspection or review
sequences. Signal towers communicate the status of the systems through
colored or flashing lights, and mini-environments and pods enclose the wafer
cassette or the inspection or review tool to enhance the cleanliness of the local
environment.
The development of the Semiconductor Equipment Communication
Standard (SECS) protocol allows a host computer to operate the inspection
and review systems remotely, initiating automatic inspection and controlling
the flow of data from the inspection tool to the review tool to the fab
database. One benefit of this automation is the reduction of operator error in
the selection of measurement recipes and data entry of basic lot information.
Automated defect data management systems were introduced to
deal with the tremendous amount of data generated by automatic defect
inspection systems. Yield correlation is one of the primary tasks of the
defect data management system. The newest systems automate yield

252 Thin-Film Deposition Processes and Technologies
correlation by overlaying maps of electrical test results with defect maps,
bringing in defect type information from review stations, then delivering
yield statistics by process level and defect types. This information has
given the defect reduction engineer the ability to focus quickly on the
defect types and process layers that most affect yield.
Defect data management systems have also adopted the use of
statistical process control (SPC) monitors that flag out-of-control defect data.
In a typical fab today, the inspection tools automatically feed data (defect
count, type, intensity, spatial signature) to the data management system. The
data management system constructs SPC charts from the incoming data and
checks for out-of-control status (using conditions predetermined by the
engineer). If the data are out-of-control, the system alerts the engineer by
e-mail or pager.
Automated transfer of data from the inspection system to the review
station represented a tremendous step forward in automating the inspection
process. Review stations—traditional white-light, laser confocal or SEM;
stand-alone or incorporated into the inspection system—take the coordinates
of the defects reported by the inspection system and automatically drive to
those locations on the wafer. The defects are then quickly reviewed, and
classified either manually or using automatic defect classification (discussed
in next section). This capability led to faster and better understanding of
defect origins and mechanisms and their impact on yield.
Trends in Automation through Advanced Algorithms. A further
step in automation is the use of algorithms to replace human operators for
optimization of system parameters to create “recipes” to inspect a given level
for a given set of defect types, and automatic defect classification (ADC).
Advancements in these algorithms are likely to reduce cost of ownership of
the inspection process substantially over the next few years.
Automated Recipe Creation Procedures. When a new device and/or
process level is inspected for the first time, a recipe has to be generated that
contains optimized measurement parameters such as optical and signal
processing configurations. The procedures for creating these recipes—and
the number of parameters involved—have become more complex as inspection technology has advanced. Thus, automation of the procedure has
escalated in importance. Automated recipe creation is particularly important
in an ASIC fab in which many different products are manufactured, and
during an excursion when quick recipe creation at a nonstandard inspection
point may be needed.

Chapter 6: Metrology and Inspection

253

Simply stated, automated recipe creation works by evaluating changes
in the signal-to-noise ratio as the different optical and signal-processing
parameters are systematically varied. A brute-force approach would try
every combination of every parameter. For today’s complex systems, the
number of variables would make this a cost-prohibitive task. A more elegant
approach would make use of existing knowledge, at least to eliminate certain
combinations of parameters, and perhaps even to find an efficient path
through the multivariable space that arrives at a unique, repeatable solution.
Today, automated recipe creation results in a recipe that sometimes
can benefit from further optimization by a well-trained engineer. Improving
automated recipe-creation algorithms is an area of focus for developers of
wafer inspection tools, due to its importance in increasing the overall
effectiveness of the equipment.
Automatic Defect Classification (ADC). One of the biggest bottlenecks in the inspection process is classification of defects, a necessary step
to determining and eliminating their source. At present, most defect classification is still manual, requiring a trained operator to judge a microscope image
and sort defects into categories based on the operator’s experience (often
using a reference book containing pictures of “typical” defects). This process
is limited in speed, accuracy and repeatability, and thus does not fit well into
an industry driven by time to market and cost control.
For these reasons automatic defect classification (ADC) has gained
tremendous attention of late. ADC has begun to reduce significantly the
amount of manual classification needed, increasing the throughput of the
classification process, and reducing subjectivity and error from operator
classification.[5] ADC thus enables more and better analysis of defect
information.
The current implementation of ADC begins first by teaching the
system to recognize the defect types by providing it with clear example
images. Then during actual automatic classification a review microscope
(off-line or built into the inspection tool) drives to the coordinates of the
detected event on the wafer and re-detects the event within the field of view.
The review station generates a digital image of the event using traditional
optical, confocal or SEM-based techniques. The ADC algorithm then
extracts features from the event and compares those features statistically to
the example images. The output includes a classification for the event along
with a goodness-of-fit value that describes the image’s similarity to the
images from the example defects in its class.

254 Thin-Film Deposition Processes and Technologies
The ultimate implementation of ADC would be for classification to
happen in parallel with inspection, without having an impact on inspection
throughput. Partial accomplishment of the goal of real-time ADC is available
today using the techniques of real-time grouping and/or spatial signature
analysis.
Real-time grouping (also called real-time defect classification or RTDC)
makes maximum use of whatever descriptive information can be recorded
during inspection. The intensity of the scattering signal; the difference in
intensity seen by collectors spanning different solid angles of the scattering
hemisphere; the gray scale intensity or perimeter of the pixels comprising the
image—these are examples of types of information that might be captured in
real time by an optical inspection system. Separating defects into coarse
groups such as “large,” “small,” “bright,” “dark,” and so forth can now be
accomplished by many inspection systems during inspection.
The spatial relationship between detected events (also called the
signature) can be used to discriminate between typical extended defects
found on wafer surfaces. Novel clustering algorithms are used to capture the
spatial signature of extended defect types such as polishing scratches,
handling damage, crystallographic defects, voids, foreign particles and stress
related defects.
Real-time grouping reduces the number of detected events that need to
be classified manually or using high-resolution ADC. Every step towards
improving time to results provides value for cost-effective IC manufacturing.
The Growing Emphasis on Cost of Ownership, Overall Equipment Effectiveness, and Ease of Use. Many of the trends in inspection
technology discussed above are driven by the need for integrated circuits to
be brought to market quickly and at contained cost. The challenge to the
inspection part of the process is to provide more defect information in an
increasingly cost-effective manner. The concept of lowering cost of ownership (COO) has been replaced recently with a new concept: overall equipment effectiveness (OEE). As applied to inspection equipment, COO includes the purchase price of the system and its throughput, whereas OEE also
weights heavily those characteristics that enable a system to provide the best
defect information in the fastest time, with maximum ease of use and
minimum cost, in a production environment.
For inspection equipment, key elements of high overall equipment
effectiveness include:

Chapter 6: Metrology and Inspection

255

• High defect capture probability, i.e., high probability of
detecting a representative fraction of defects from the
population of all defect types present on the wafer.
• High correlation of detected events with yield-limiting
defects: low false or nuisance counts.
• High throughput.
• Low system purchase price.
• High reliability: high repeatability of measurements;
minimum downtime; minimum maintenance.
• System matching: highly correlated results using systems of
same model number; easy transfer of recipes between tools
and between fabs.
• Ease of use: automated recipe learns and operation with
minimum intervention of highly paid personnel.
• Maximum integration of defect inspection, review,
classification and analysis process.
Not all of these elements may be captured in equations measuring
OEE, but all of them are key elements for operating the defect reduction
process at highest possible efficiency and lowest cost.
The final element of successful yield enhancement through defect
reduction is close communication between defect metrology and process
engineering groups. A highly effective defect reduction program seamlessly
integrated with a strong process engineering program is well positioned for
success in the semiconductor market.
4.0

THEORY OF OPERATION, EQUIPMENT DESIGN
PRINCIPLES, MAIN APPLICATIONS, AND STRENGTHS
AND LIMITATIONS OF METROLOGY AND INSPECTION
SYSTEMS

This section briefly discusses the theory of operation, main applications, and the strengths and limitations of several thin film metrology
systems. For a more detailed theoretical discussion, the reader is encouraged
to consult Ref. 6.

256 Thin-Film Deposition Processes and Technologies
4.1

Film Thickness Measurement Systems

Theory of Operation. The common optical measurement techniques
include reflectometry (using unpolarized or polarized light) and ellipsometry.
System implementations use multiple wavelength or multiple angles of
incidence. Regardless of the type of system, the data analysis methods that
transform the directly measured quantities to the parameters of interest such
as thickness and refractive index are similar. The measurement recipe
contains information about the film stack to be measured, such as the type of
material, approximate thickness of the material, and (implicitly) the refractive
index of the material. The Fresnel reflectance equations are used to calculate
theoretical spectra for the film stack. The calculated spectra are compared
with the measured spectra, and regression analysis is performed by varying
the parameters of interest until the best fit is obtained. The best-fit values are
reported, along with a figure of merit referred to as the goodness-of-fit
(GOF).
The capability of a system to report parameters of interest such as
thickness and refractive index values derives from the amount and type of
raw data measured by the system. Generally, the information content of the
raw measured values (for one wavelength and one angle of incidence)
increases in the following order: unpolarized reflectometry (R); polarized
reflectometry (Rp and Rs); and ellipsometry (Ψ and ∆). Although both
polarized reflectometry and ellipsometry measure two values at each wavelength and angle, ellipsometry is unquestionably the more powerful technique
for a number of reasons.[7] Ellipsometry measures the phase of the reflected
light (not just amplitude). Ellipsometric measurements are relatively insensitive to intensity fluctuations of the illumination source, temperature drifts of
electronic components, and macroscopic roughness. Macroscopic roughness
causes light loss by scattering incident light away from the detector, which
can be a serious problem in reflectometry but not in ellipsometry, for which
absolute intensity measurements are not required. Ellipsometry is inherently
a double beam technique, because one polarization component serves as
amplitude and phase reference for the other.[8]
The other system implementation choice is to collect data using
multiple angles of incidence or using multiple wavelengths. Data collected
using multiple wavelengths generally have higher information content than
data collected using multiple angles of incidence. An examination of the phase
term φ of the Fresnel reflectance equations shows a first order change with
wavelength, while a change with angle is contained within a sine term:

Chapter 6: Metrology and Inspection

Eq. (1)

φ = 4π

257

d
d
2
n1 cosθ 1 = 4π
n1 − sin 2 θ 0
λ
λ

where θ1 is the refracted angle of light in the film, θ0 is the incident angle (in
ambient), d is film thickness, and n is refractive index.
The consequence is higher sensitivity to film changes with wavelength
than angle. (For oxide, the percentage change in the phase term with
wavelength from 200 to 800 nm is an order of magnitude higher than with a
change in angle from 0 to 90°). Additionally, multiple wavelength systems
take advantage of the fact that a physical property of a film, the refractive
index, is a function of the illumination wavelength (refractive index dispersion) as illustrated in Fig. 2. Refractive index is not a function of the angle of
incidence of the illumination. Therefore, little information about the physical
properties of the material can be deduced by measurements using multiple
angles. Conversely, the RI dispersion difference between different materials
can be exploited by multiple wavelength systems. The consequence is that a
change in the thickness (or RI) of one film in a multiple-layer film stack will
change the measured spectra differently than a change in a different layer.
The different optical penetration depth as a function of wavelength provides
additional information to resolve the thicknesses of several layers in multiple
layer structures.[8] This implies that a multiple wavelength metrology system
will be sensitive to, and be able to differentiate between, thickness and index
changes in multiple-layer film stacks. Spectroscopic ellipsometry provides
the highest level of capability of the optical thin film measurement techniques, by collecting ellipsometric data over a large range of wavelengths.
Only spectroscopic ellipsometry uses all properties of polarized light: amplitude, phase, and wavelength.
Unpolarized reflectometers measure reflectance (R) versus wavelength at a single angle of incidence (normal to the wafer). Light reflecting
from the top surface of the film interferes with light reflecting from the film
to substrate interface, resulting in a periodic variation in reflectance as a
function of wavelength spectrum.
Polarized reflectometers measure polarized reflectance (Rp, Rs)
versus angle of incidence at a single wavelength. A high numerical
aperture objective lens is used to achieve a small spot size, resulting in
multiple angles of incidence. Reflectometers, whether using polarized or
unpolarized light, require calibration to known standard wafers (referred

258 Thin-Film Deposition Processes and Technologies
to as a referenced technique). These wafers can be either bare silicon or
wafers of a “known thickness.” This is because reflectometers need to
know very accurately the value of absolute reflectance, and to be able to
compensate for intensity losses through aging of the illumination source,
efficiency of the metrology system optics, etc. (referred to as optical
throughput).
Ellipsometers measure the amplitude ratio (Ψ) and phase change (∆)
of polarized light upon reflection from a sample. Single wavelength
ellipsometers are available with a high numerical aperture objective lens to
achieve a small spot size, resulting in multiple angles of incidence. The use of
multiple angles eliminates the problem of thickness order ambiguity.

Figure 2. RI dispersion of materials commonly used in the semiconductor industry. The
refractive index n is the upper plot; the extinction coefficient k is the lower plot.

Chapter 6: Metrology and Inspection

259

Spectroscopic ellipsometers measure amplitude ratio and phase
change versus wavelength at a single angle of incidence. (Research grade
spectroscopic ellipsometer systems also offer multiple angles of incidence
for material characterization). The angle of incidence is typically 70–75°,
near the Brewster angle, selected for its maximum sensitivity to changes in
film thickness on silicon substrates (Fig. 3). Attributes of commercially
available instruments are summarized in Table 2.

Figure 3. Sensitivity of SE technology (phase change term, cos ∆) to very thin oxide
thickness changes when operating near the Brewster angle.

Table 2. Typical Instrument Configurations Used in Semiconductor Process Monitoring
Property of Light Used
Amplitude Phase Wavelength

Angle of
Incidence

Referenced
Measurement

Unpolarized
Reflectometer

yes

no

yes

single
(normal)

yes

Polarized
Reflectometer

yes

no

no

multiple
(~normal)

yes

yes

yes

no

multiple
(oblique)

no

yes

yes

yes

single
(oblique)

no

Focused
Beam
Ellipsometer
Spectroscopic
Ellipsometer

260 Thin-Film Deposition Processes and Technologies
Main Applications. Thin film metrology systems are used in every
process module in semiconductor fabs; to monitor thickness and/or refractive index uniformity in deposition and diffusion areas, for removal rate and
uniformity in etch and planarization areas, and to monitor reflectivity in metal
deposition or photolithography areas. Depending on the measurement technique, semitransparent films from several angstroms to several microns in
thickness can be measured. Historically, ellipsometry was used in diffusion
areas for very thin films, and reflectometry was a general thin film tool. These
distinctions are diminishing as advances are made in measurement technology, and many systems now incorporate more than one type of technology.
Strengths. The ability of technologies such as spectroscopic ellipsometry to simultaneously and independently measure multiple film thicknesses
and refractive indices offers opportunities to semiconductor manufacturers
for reduced cost and enhanced process control. Wafer fabrication costs can
be reduced through decreased use of monitor wafers. This is especially
important for 200 mm and 300 mm wafers. As an example, the ability to
measure nitride over amorphous silicon over oxide in a DRAM structure
means that a nitride on silicon monitor wafer can be eliminated. Additionally,
the increasing use of multi-process chamber cluster deposition tools requires
the ability to measure multiple-layer film stacks, since there is no opportunity
to measure each layer after it is deposited.
Semiconductor fabs are becoming more interested in monitoring and
controlling film quality, as opposed to merely film thickness. The refractive
index of a film is a key indicator of film composition. For films such as
amorphous silicon and polycrystalline silicon, the extinction coefficient k
~
(part of the complex refractive index N = n + ik ) is directly related to the
crystallinity of the film. Process temperature (of the deposition or anneal
step) determines the crystallinity. New materials such as silicon-rich oxynitrides
and nitrides are increasingly used as anti-reflective layers. The stoichiometry
of these films can also be monitored by UV spectroscopic ellipsometry
measurement of refractive index.[1] Therefore, the ability to measure refractive index, not only of a single-layer film, but also of a film in a multiple-layer
film stack, is beneficial.
Limitations. Reflectometers require calibration using wafers of known
reflectance, and can be subject to measurement drift over time. Single
wavelength ellipsometry is known for limitations such as thickness order
uncertainty, and thickness regions where refractive index cannot be calculated. Multiple angle systems eliminate thickness order ambiguity for single

Chapter 6: Metrology and Inspection

261

layer films. Use of a small spot size and a single wavelength can be
problematic when measuring rough films that scatter light.
All optical tools require a film stack model for their regression
analysis. Accuracy can be affected by selection of an appropriate refractive
index dispersion model. Materials which do not have a consistent spatial
composition, e.g., their refractive index changes with depth, require special
models to measure correctly.
Future gate dielectric process control will be a difficult challenge, if
composition information is required as well as thickness uniformity. Increasing attention will need to be paid to the interface between the gate dielectric
and the silicon substrate for proper measurement accuracy.[9] There is also an
inherent difficulty in monitoring metrology system stability at a 4 nm gate
dielectric thickness, as these wafers tend to change thickness over time due
to environmental effects.
Measurement spot size considerations result in the use of test structures, typically located in the scribe lines, since no metrology system can
currently measure inside of submicron features.
4.2

Resistivity Measurement Systems

Four-Point Probes. Theory of Operation. Four-point probe (4PP)
systems measure sheet resistance (Rs), the local resistance of a sheet of
material, in units of ohms/square. Sheet resistance is expressed in the
equation Rs = Rb /t, where t is the thickness of the conductive layer, and Rb
is the bulk resistivity (ohm-cm) of the layer. For a material with constant bulk
resistivity, the sheet resistance is only a function of thickness. A four-point
probe consists of four spring-loaded conductive probes (usually in a linear
array) which are placed in contact with the material whose sheet resistance
is to be measured. (The four-point probe technique requires some isolating
junction or blocking layer to the DC current used).
Typically, a known current is forced between the outer probes, and the
resulting voltage across the inner two is measured. Ohm’s Law (V = IR) is
then used, with a geometrical correction factor, to calculate the sheet
resistance of the material. To compensate for geometric errors arising from
variations in probe tip spacing and proximity to the wafer edge, the dual
configuration technique was developed. A second measurement is made,
with current forced through pins 1 and 3, and voltage measured between pins
2 and 4 (Fig. 4). The geometrical correction factor can then be calculated
based on the ratio of the measured resistivities.[10]

262 Thin-Film Deposition Processes and Technologies

Figure 4. Dual configuration four-point probe measurement setup.

Main Applications. Resistivity systems are used throughout the fab
to measure any conductive semiconductor layer; from incoming silicon
wafer inspection, to metal deposition and etch/planarization for removal
rate and uniformity, to dopant uniformity in diffusion and ion implant
operations. Ion implantation was the first process to extensively use 4PP
for process control, and also the first to exploit the power of full wafer
uniformity mapping as a process diagnostic. The normal range for sheet
resistance in semiconductor processing is from less than 0.02 ohm/square
for aluminum films, to about 1 mega-ohm/square for low dose implants into
silicon.
Strengths. The four-point probe is the most common tool used to
measure sheet resistance, due to its accuracy, repeatability (~ 0.2% 1σ),
and relatively low cost. The most accurate systems employ temperature
compensation, to account for variations in the resistance that occur with
variations in ambient temperature. Temperature compensation is especially important to achieve system to system accuracy matching, an increasingly common requirement not only within a fab, but also for process
transfers between fabs.

Chapter 6: Metrology and Inspection

263

Limitations. Because the 4PP technique is based on physical contact with the wafer, its use is limited to monitor wafers. Care must be taken
to ensure low contact resistance between the probe tips and the conductive
layer. Probe tip “conditioning” and qualification routines have been developed to manage this concern.
The most challenging applications for 4PP are low energy (ultrashallow junction), low dose (high sheet resistance) ion implants. These
processes require enough probe tip pressure to penetrate down to a more
conductive depth but not so far as to penetrate past the junction. Reductions
in measurement edge exclusion are also required.
Eddy Current (Mutual Inductance). Theory of Operation. Eddy
current or mutual inductance techniques determine sheet resistance of a film
by creating a time-varying magnetic field from a coil (“probe head”). The coil
radiates energy, is placed close to the conductive layer, and eddy currents are
generated by the absorbed energy in the conductive layer. The eddy currents
in turn create a magnetic field that induces a reverse current in the coil (hence
the term mutual inductance). The sheet resistance of the conductive layer
determines how much current is induced in the Rs sensing circuit. The more
conductive the film, the more energy is trapped in the film to create eddy
currents. The magnetic field drops off with distance (depth into the film).
Various configurations exist: “single-side,” where the wafer is placed
under the inductive coil (Fig. 5), and “dual-side,” where the wafer is placed
inside the turns of the inductive coil. The dual-side approach, developed first,
typically has a larger spot size and greater measurement range. The measurement is made in a transmission mode. Decreasing the spot size requires
decreasing the distance between the coils. A practical limit of spot size results
from clearance between the sample and the coils. The single-side approach,
essentially a reflection measurement, offers advantages of higher sensitivity
to the top layer than the underlying layer/substrate, smaller measurement
spot size, and reduced edge effect.
Main Applications. Historically, eddy current systems have been
used for incoming inspection of resistivity of prime silicon wafers. More
recently, as the RC time delay of interconnect layers has become a gating
item to increasing device speeds, the technique has been applied to
measurement of blanket metal layers prior to patterning.[11] Since the
metal layer is unpatterned, it supports eddy current formation. Underlying
metal films are patterned, which prevents the formation of eddy currents.
The typical resistance range covered by eddy current systems is 0.001~50
ohms/square for single-side and 0.01~5000 ohms/square for double-side.

264 Thin-Film Deposition Processes and Technologies

Figure 5. Eddy current measurements (single-side configuration).

Strengths. Because it is a non-contact technique, it can be used on
product wafers. This is desirable for several reasons: significant cost savings
by reducing monitor wafers, better equipment utilization by processing fewer
monitors, and improved process control by measuring actual product wafers.
Limitations. Since the magnetic field penetrates through the top metal
layer of interest, the technique works best when the resistivity of the
underlying layers/substrate is high relative to that of the conductive metal
layer (e.g., typical memory applications). In the case of low resistivity
substrates, the technique can be extended somewhat by calibration or
software correlation.[12] Eddy current systems are typically calibrated using
four-point probes. Spot size can be an issue as smaller measurement edge
exclusions are required.
4.3

Stress Measurement Systems

Although properties like film thickness, density, and resistivity can be
immediately related to device performance, the level of stress in a thin film
is more important to long term device reliability and lifetime.
Stress in a thin film develops primarily during the deposition process
and consists of two components: the intrinsic stress and the extrinsic stress.
The intrinsic stress is the component of stress in the film caused by the
deposition process itself. For example, a slowly sputtered metal film deposited onto a heated substrate will grow with near zero stress, although the film
structure may be metastable. By contrast, a chemical vapor deposition
(CVD) oxide film can have a highly variable stress level depending on

Chapter 6: Metrology and Inspection

265

film density, moisture content or residual reactants such as hydrogen.
Within a wide range, the intrinsic film stress for most materials can be
controlled by the process parameters of the deposition system.
The extrinsic stress is the component of stress caused by a change in
the external conditions on the wafer. For example, most films have a thermal
expansion coefficient different from the silicon substrate, so when the
temperature changes, the film and substrate try to expand or contract by
different amounts. However, because they are bound together, a stress will
develop in both the film and the substrate. Since films are typically deposited
above room temperature, the process of cooling after deposition will introduce a thermal component of stress into most films. The total film stress will
be the sum of the intrinsic and extrinsic components.
Theory of Operation. To maintain static equilibrium, the forces and
moments in the film must balance the forces and moments in the substrate,
which requires a shape change of the wafer when a stressed film is deposited.
For the geometry of a thin uniform film deposited on a much thicker, but
platelike wafer, the shape change caused by the addition of a film will be a
uniform bowing of the wafer like a spherical bowl or dome, but several
assumptions are involved:
• The substrate can be treated as a plate which simplifies the
elasticity equations. The basic requirement for meeting the
plate geometry is that the characteristic length dimension be
more than about 10 times the thickness. The approximation
errors can be significant unless the ratio is notably higher, and
in the semiconductor industry, a typical 200 mm wafer is
under 1 mm thick. The ratio is well over 100:1, so the plate
approximation is quite accurate.
• The film is uniform and homogeneous in any feature that can
influence stress, primarily the thickness. The stress in a film
deposited only on one face of a wafer will give rise to forces
and bending moments that change the shape of the wafer. If
the film is uniform, the forces will be evenly distributed
causing an even shape change. The stress is biaxial, so the
elasticity equations can be solved to relate radius of curvature
of the resulting shape change to the film stress. Any significant
variations in the film thickness, chemical composition or
internal structure of the film can lead to nonuniform bending
and an inaccurate average stress calculation.

266 Thin-Film Deposition Processes and Technologies
• The film is much thinner than the substrate. The assumption
leads to the simplest equation to relate film stress to change in
curvature of the wafer which is called the Stoney equation (Eq.
2). Most researchers and equipment companies use the Stoney
equation to calculate stress. The equation relies only on the
elastic properties of the substrate, so the stress in any film
material regardless of quality can be determined. However, if
the film thickness reaches about 5% of the substrate thickness,
the calculation error will be about 10%. Again, in the
semiconductor industry, films are typically less than a few
microns on substrates over 500 microns, so the ratio is under
1% and the error is insignificant.
An important point about the typical commercial equipment available
to determine stress is that all systems measure curvature or shape. The raw
data must be analyzed to yield a radius of curvature before and after film
deposition. The change in radius is then used to calculate stress. A tool to
accurately and conveniently measure film stress directly does not exist. The
stress calculation is based on the Stoney equation,

Eq. (2)

σf =

1 E s t s2  1
1
− 

6 1 − Vs t f  R f Rs 



The subscripts s and f refer to the substrate and film, respectively,
while E is Young’s modulus, v is Poisson’s ratio, t is thickness and R is the
measured radius of curvature.
Main Applications. Film stress can give rise to a number of problems
that can lead to failure in the operation of an integrated circuit, so determining
film stress is important for maintaining a reliable process. Two common
issues are cracks forming in highly stressed, brittle passivation layers, and
voids forming and growing in aluminum lines. Stress also contributes to
reliability failures such as electromigration. Other issues can include debonding
of high stress metal films like tantalum, or sorption of volatiles like water and
organic solvents from porous films.
Stress measurement can be divided into two categories with distinctly
different goals: testing done at room temperature and testing done while
thermally cycling a wafer. Room temperature testing is typically used for
monitoring an established deposition process for an SPC style control.

Chapter 6: Metrology and Inspection

267

Thermal cycling is typically used for process development and materials
characterization.
Room Temperature Testing. Room temperature testing provides
basic process control data. The stress level in a film can be influenced by
controlling the deposition parameters such as temperature, pressure, reactant
flow rates and input power, so film stress can be used as part of the process
development. Once an acceptable set of deposition conditions are established, continual monitoring of the resultant film stress will give a measure of
the long-term stability of the deposition system. The results are ideally suited
to a simple SPC control on the film deposition process.
Advanced stress measurement systems include stress mapping capabilities over the surface of the wafer. Information about the stress distribution
throughout the film is especially valuable in determining the uniformity of a
deposition process beyond basic film thickness uniformity results.
Thermal Cycle Testing. Thermal cycling of a film-wafer sample
provides data that delves more deeply into the mechanisms of stress
generation and evolution. Generally, the thermal expansion coefficient of the
film and substrate will be different, so changing the sample temperature will
impose a thermal component of strain that can lead to high stresses in the
film.
Optical techniques are required in which a laser can be aimed through
a transparent window to measure curvature and stress continuously during
thermal cycling. Numerous thin film effects have been observed including
yield behavior in metals, effusion of volatiles from porous films, phase
changes, and hillock formation.
Thermal testing allows for a more fundamental examination of the
mechanical behavior of thin film materials. The stress data obtained during
thermal cycling can be used to approximately determine thermal expansion
coefficient, modulus of elasticity and some activation energies. The technique has also been used on polymeric films to determine glass transition
temperatures.
Deflection Measurement Techniques. Several techniques relying on
differing technologies have been developed to measure film stress, but all
basically measure the average radius of curvature of a wafer before and after
the film deposition. An overview of four measurement techniques will be
given: one directly measures bow optically, and the other three scan the shape
of the wafer surface using either a two plate capacitor, a contact stylus
profiler, or a laser lever.

268 Thin-Film Deposition Processes and Technologies
Bow Measurements. The most direct measurement technique uses a
fiber optic sensor to determine the bow at the wafer center before and after
film deposition. The wafer is supported by a knife edge ring of diameter D,
so the bow, d, is related to the radius of curvature by the equation
Eq. (3)

R=

D2
8d

The change in radius caused by the film is related to the stress using the
Stoney equation. The fiber optic sensors can measure bow changes with a
resolution in the range of 0.05 microns, but using one central measurement
to represent the entire wafer shape limits the accuracy of the stress measurement. Advanced versions of these systems include multiple probes that allow
deflection and stress mapping over the entire wafer.
Strengths of this technique are that it is simple, non-contact, and
sensitive. Limitations are a limited amount of data (one data point) and poor
thermal performance.
Capacitance Measurements. The capacitance probe technique
measures the capacitance between a small probe and the surface of the
wafer from which the distance to the wafer can be determined. By using
probes simultaneously on the front and back surface of the wafer, the
wafer thickness is determined along with the wafer position between the
two probes. The wafer is automatically moved through the probe to obtain
a map of thickness and shape. Using the Stoney equation and numerically
fitting the shape data, average stress or stress maps can be obtained. The
capacitance probes determine bow with a repeatability (1σ) of about 0.5
microns and wafer thickness repeatability (1σ) of about 0.05 microns.
Strengths of this technique are high measurement speed, so a large
quantity of data can be collected. A limitation is the lack of thermal capability.
Profilometry. Profilometry uses a contact stylus sensor to determine step heights and general surface topography over a wafer. By
scanning over a sufficiently large area of the wafer, a map of the wafer
shape can be obtained before and after film deposition. The data are fit to
determine the change in radius and put into the Stoney equation. Profilers
can achieve exceptional performance with approximately 0.01 micron
repeatability (1σ) in vertical resolution of the surface shape.
The main strength of profilometry is its high sensitivity. Limitations
are slow measurement speed, the probe tip contacts the sample, no thermal
capability, and full wafer data must be “stitched” together.

Chapter 6: Metrology and Inspection

269

Optical Lever Measurements. Optical lever systems aim a laser at
the surface of a wafer and measure the direction of the reflected beam
using a position sensitive light detector. From knowledge of the system
geometry, the wafer surface normal and therefore the tangent are measured. Scanning over the wafer surface provides a map of tangent versus
position which is fit to determine the change in radius and put into the
Stoney equation. Optical systems can determine bow with a repeatability
(1σ) of about 0.5 microns.
Strengths include simplicity, measurement speed, and thermal capability. Limitations are that the laser is diffracted by patterned wafers, and
interference effects occur in some films.
4.4

Defect Inspection Systems

This section covers theory of operation and equipment design principles, main applications, and strengths and limitations for unpatterned
and patterned wafer inspection systems. The section is organized as
follows:
General Theory of Inspection System Operation and Design
Optical Imaging
Optical Scattering
Unpatterned Inspection Systems
Applications
Strengths and Limitations
Patterned Inspection Systems
Applications
Strengths and Limitations
General Theory of Inspection System Operation and Design. Very
generally, an inspection system must be able to detect the presence of the
defects on the wafer and identify their spatial locations. Defect detection
requires some kind of contrast mechanism to distinguish the defect from its
surroundings. Typical contrast mechanisms include those associated with
optical imaging (both dark-field and bright-field); optical scattering (darkfield); and electron imaging. In this section we will focus on the optical
techniques as these are the most commonly used in monitoring defectivity
for IC production today.

270 Thin-Film Deposition Processes and Technologies
In order to discuss the technology of optical inspection, some usage
conventions are helpful. On a perfect mirror surface, the light incident at a
given angle reflects at the equivalent angle in the plane of incidence to form
the specular beam (Fig. 6). On a real surface some of this light will be
scattered: absorbed, diffracted or otherwise directed to an angle outside the
specular beam. Particles, scratches, surface roughness, local device topography or interfaces between different materials can cause light to scatter.

Figure 6. Simplified schematic of optical subsystem for optical defect detection. There are
three beams of light: the incident beam, the specular beam and a ray representing the
scattered light. The plane formed by the wafer surface and the scattered light beam is at an
angle φs from the incident plane, which is formed by the wafer surface and the incident
beam. The specular beam lies in the incident plane.

Optical Imaging. For optical imaging an area of the surface of the
wafer is illuminated uniformly. Features and defects within the illuminated
area scatter the light according to their material properties and topography. A
series of lenses captures the specular beam (bright-field design) or scattered
light (dark-field design), imaging its spatial variation on the wafer surface
onto an area detector such as a CCD camera or TDI detector. The
information conveyed by the image arises from the differences in the way the
defects and features on the surface scatter light. Defects are detected by
comparing the digital image of one part of a die with the image from
equivalent areas in neighboring dies, and identifying differences.

Chapter 6: Metrology and Inspection

271

The resolution of an optical imaging system is determined by the
pixel size of the area detector, the spectrum of the light source and photon
density of the illuminated area, and the optical contrast between the defect
and its surroundings.
Optical Scattering. For optical scattering, a small spot of high photon
density is illuminated on the wafer surface. Features and defects within the
illuminated spot scatter light according to their material properties and
topography. The specular beam is discarded, and the scattered light is
collected over a particular solid angle, then focused onto a point detector such
as a photomultiplier tube (PMT). In essence, the amount of light scattered by
that illuminated area into a given solid angle is recorded. Various signal
processing schemes are then employed to determine whether or not a defect
is present at the location of the illuminated spot. For example, this signal could
be compared with a signal from an equivalent area elsewhere on the wafer.
The entire surface of the wafer is sampled by moving the beam across the
wafer, or the wafer under the beam, or a combination thereof.
The resolution of an optical scattering inspection system is determined
by the ratio of the light scattered by the defect to that scattered by its
surroundings. This is in turn determined by the photon density of the
illuminated spot, the wavelength and incident polarization of the light
source, the solid angle(s) subtended by the collection optics, the polarization
collected in the detector, and details of the optical and topographic properties
of the defect and its local environment.
Thus the probability of capturing a defect of a particular shape,
material and size, residing on a particular film stack and in the presence of a
given local device topography varies with many parameters. The specifics of
the optical design of the inspection system—in particular the angles of
incidence and of collection—influence the capture probability, as does the
signal processing technique. Capture probabilities can be predicted rather
well through mathematical modeling for spherical particles on bare wafers.[13]
For specific defect types on patterned wafers, simple models can be
constructed that also produce useful results.[14]
Unpatterned Wafer Inspection Systems. Typically dark-field optical scattering is used for detection of particles, scratches and crystal defects
of unpatterned wafers. The complexities of the pattern signal are not present,
and the cost of the unpatterned wafer is lower than that of the product wafer.
For cost-effective inspection, throughput is very important in unpatterned
wafer applications.

272 Thin-Film Deposition Processes and Technologies
The original unpatterned wafer inspection systems used a red helium neon laser (633 nm), but in the early 1990s a switch was made to a
blue argon ion laser (488 nm) whose shorter wavelength provides increased sensitivity.[15] Unpatterned wafer inspection systems use either a
normal-incidence or oblique-incidence design (Fig. 7).

(a)

(b)

Figure 7. (a) A normal-incident laser-scattering system avoids the specular beam, but
collects a large portion of the light in the hemisphere above the wafer. (b) An obliqueincident laser-scattering system avoids the specular beam, but also avoids the higher-angle
non-specular light and collects light close to the wafer horizon and to the side of the incident
and reflected specular beam.

Many unpatterned inspection systems in use today scan the laser
spot in the x-direction while the wafer moves in the y-direction. The
scattered light is collected either in a large solid angle near the specular
beam (for normal-incidence systems), or to the side of the wafer close to
the horizon (for oblique-incidence systems). The beam spot is typically
10s or 100s of micrometers across, and every defect is sampled several
times by several sweeps of the laser spot.
Another successful commercial unpatterned inspection system uses a
markedly different optical design. For improved sensitivity and uniformity
the entire incident and collection optics are held stationary, while the wafer
is rotated and translated beneath the optical system (Fig. 8). The laser spot
traverses a spiral path to sample the entire wafer surface. The system has two
collection channels that together span almost the entire scattering hemisphere. The collection optics are axially symmetric, allowing very uniform defect capture even for defects that scatter light highly directionally,

Chapter 6: Metrology and Inspection

273

like scratches. The system also includes a Nomarski differential interference contrast microscope to distinguish concave from convex defects, and
to capture large-scale, low-topography defects such as a surface quality
problem called “orange peel.” This inspection system is described in more
detail in Ref. 16, and Nomarski microscopy is discussed in Ref. 17.

Figure 8. Schematic of unpatterned wafer inspection system showing dynamic wafer stage,
and rotationally symmetric (static) collection optics.

During signal processing the scatter arising from surface roughness
or haze is separated from the defect scatter. The resulting information is
typically divided into three categories: haze, point defects, and large-area
scattering events like scratches. The values reported for the defects include
their coordinates, scattering cross-section (µm2) and/or “diameter” (µm).[13]
The “diameter” values are calibrated from the scattering cross-sections of
populations of polystyrene latex (PSL) spheres of known size, deposited on
the film of interest.
Applications. Unpatterned inspection systems are used for the main
applications shown in Table 3.

274 Thin-Film Deposition Processes and Technologies
Table 3. Applications of Unpatterned Inspection Systems
Application

Defect Types

Wafer

Goal

Silicon wafer
manufacturing

Particles, pits,
scratches, crystal
defects, haze

Bare wafers

Increase quality
Increase yield
Monitor yield

IC manufacturing
Incoming inspection

Same as above

Bare wafers

Quality control

Equipment qualification

Particles

Bare wafers

Ensure tool is not
adding defects, is ready
for production

Equipment monitoring

Particles

Bare wafers

Monitor tool for
change in added defects

Process monitoring

Particles;
scratches (CMP
area)

Blanket films

One monitor wafer in
cassette with product
wafers tracks defectivity
of process

Unpatterned inspection systems are used heavily by silicon manufacturers and IC manufacturers for inspection of bare silicon and blanket
films. Silicon manufacturers use the systems to measure the number of
particles, pits, scratches, and crystal defects[18,19] and to characterize haze
on the wafers.[20][21] In IC manufacturing they are typically used to measure PWP—Particles per Wafer Pass—to characterize particles added by
process or metrology tools during wafer processing. From this measurement a tool is qualified for production use and/or monitored using SPC
charts to understand if the tool is degrading and causing any defect issues.
For example, a process tool may be studied with PWP tests to understand
how often the chamber needs to be cleaned, and then monitored with PWP
tests for any other abnormal excursions.
Strengths and Limitations. See Table 4 for some of the strengths and
limitations of defect detection systems.
Patterned Wafer Inspection Systems. Patterned-wafer inspection
systems may use bright-or dark-field imaging, or dark-field scattering, or any
combination of these as their fundamental defect detection technology. The
challenges specific to patterned-wafer defect inspection come directly from
the presence of pattern near the defects, and from the fact that the types of
possible defects are extended to include defects in the pattern itself—
bridges, opens, missing vias, etc. Several different system designs exist

Chapter 6: Metrology and Inspection

275

among successful commercial inspection systems; these can each be
optimized for sensitivity and throughput using several configurable system parameters. Capture probability for a given defect type on a given
layer for a given IC product will vary according to the configuration of
the system and its overall design. Determination of the optimum inspection system to detect a range of key defects for a given segment of
the process is best determined experimentally.
Table 4. Strengths and Limitations of Defect Detention Systems for
Unpatterned Wafers
Unpatterned Wafer Inspection Systems
Strengths
Limitations
• Can detect defects as small as
80 nm on bare silicon wafers
• Excellent sensitivity to particles
on smooth and rough blanket
films
• Can detect crystal defects
• Can measure haze/surface
roughness
• Low cost per inspection

• Unpatterned wafers only
• Defect must have some
type of light-scattering
signature to be detected:
topography, or change in
material properties
• Sizing calibration is based
on known polystyrene latex spheres

• Automated
• High throughput

Bright-field imaging systems work by flood-illuminating an area of
the wafer, then using an objective lens to construct an optical image of that
area from the specular beam reflecting off the surface. The image is
captured by a digital camera, and that image is compared with another
image from a similar area on the wafer. In periodic areas such as memory
cells the compared area may be within the same memory cell; for nonperiodic areas an equivalent area from a neighboring die is used. Differences between images become candidates for identification as defects.

276 Thin-Film Deposition Processes and Technologies
Dark-field imaging is a similar technique, but instead of utilizing
information from the specular beam, the detector is placed away from the
specular beam to intercept a fraction of the scattered light. Although both
dark-field imaging and dark-field scattering use scattered light to detect
defects, dark-field imaging differs in that it preserves the spatial relationships
among features illuminated within the spot. Both dark-field techniques
should provide similar defect capture probabilities, if angles of incidence and
collection (and other system parameters) are comparable.
Dark-field scattering technology for patterned wafer inspection is
similar in principle to unpatterned wafer inspection technology except that the
scattering signal from the pattern must be managed. In the first patterned
wafer inspection systems all signals having a period similar to that of the die
spacing were discarded, while all aperiodic signals became candidates for
identification as defects. The latest dark-field scattering inspection systems
retain all signals above a threshold, then employ a die-to-die comparison
algorithm similar to that described above for imaging systems.[22]
A critical requirement of all patterned wafer inspection systems is fast
and reliable alignment of the wafer to the scan axes. This is necessary for
proper implementation of the signal processing algorithms, and this requirement has driven the use of highly precise stages and sophisticated alignment
algorithms in these systems.
Applications. Patterned wafer inspection systems were introduced in
the 1980s to drive yield enhancement by providing a direct look at defect
densities on production wafers. These systems replaced the long-standing
tradition of inspection by operators using optical microscopes. Today’s
inspection systems are very automated and offer higher throughput than an
operator, more repeatable and objective results, and significantly higher
detection sensitivity and defect capture.
Patterned wafer inspection systems are used to inspect wafers for
defects at any level in the IC manufacturing process. Two main applications
of these systems have been discussed previously in this chapter: in-line
monitoring and yield learning. Patterned wafer inspection systems are usually
integrated closely with data analysis systems, review stations, and automatic
defect classification (ADC) to provide an entire system for identifying and
eliminating yield-limiting defects, and to monitor the process line to catch
defect excursions and drive continuous improvement.
Increasingly, these systems are being used for process equipment
monitoring with patterned production wafers. The trend towards reducing
the cost of monitor wafers in the fab has led to increased dependence on

Chapter 6: Metrology and Inspection

277

production wafer defect data to help determine the defect contribution of
process equipment.
Strengths and Limitations. See Table 5 for some of the strengths
and limitations of defect detection systems.

Table 5. Strengths and Limitations of Defect Detection Systems for Patterned Wafers
Optical Patterned Wafer Inspection Systems
Strengths
• Can detect defects as small as
0.1 µm on front-end, patterned
layers, on product wafers
• Techniques exist to reduce
noise effects of film nonuniformity and process variation
• Automated
• Integrated with defect review
and classification systems, defect data analysis systems

4.5

Limitations
• Defect must have some
type of light-scattering
signature to be detected:
topography, or change in
material properties
• Lower throughput than
unpatterned inspection
systems

Automatic Defect Classification

The main goal of Automatic Defect Classification is to reduce the
number of defects that require manual review and classification. ADC
incorporates several levels of operation to achieve this. The first level
involves the use of clustering algorithms that group together defects with
certain spatial signatures, such as scratches. These cluster algorithms are
normally also available on the inspection system, providing a first-pass
defect classification that happens while the wafer is scanned.

278 Thin-Film Deposition Processes and Technologies
After clusters have been removed a smaller group of defects is left
behind to classify. This group can be reduced further by using a data analysis
system to compare the current defect map with maps from previous layers,
and selecting only defects from the current layer. Next, defects can be
removed from the remaining group by considering characteristics stored
during inspection, such as defect size. After the data set has been significantly
reduced by these techniques, the remaining defects are reviewed and
classified automatically or manually, comparing the characteristics of the
reviewed image of the defect to information in a database. Manual classification involves having a trained operator compare the microscope image to
example images in a defect scrapbook. The remainder of this section
describes how this part of the review and classification process can be done
automatically, via high-resolution automatic defect classification.
The database used for high-resolution ADC is built by the user and
requires multiple example images of each type of defect. Typically 5 to 20
examples of each defect type are needed to construct the database. The
system measures the value of each of a number of features describing the
defect. For a given defect category, each feature spans a particular range.
The set of feature ranges distinguishes one defect category from another.
These category features are used as a reference for high-resolution automatic
classification of new defects.
After the system has been trained, new defects are classified as
follows:
If ADC is on board the inspection system the system already
has the wafer aligned and in the system. Otherwise the wafer
must be loaded and automatically aligned. Using the coordinates
determined by the inspection system, the ADC system drives
to a defect’s position and re-detects the defect. The redetection algorithms are similar to the die-to-die processing
discussed previously. Depending on the type of device being
inspected, the ADC system will drive to another location
within the die that has identical pattern (for example, to
another location within the memory array for a DRAM), or
will drive to one or two adjacent die to compare the
corresponding image(s) and subtract the images to determine
the defect image. After the defect has been re-detected, the
system compares the values of the features of the defect
image to those stored in the database.

Chapter 6: Metrology and Inspection

279

The final results include the best match of that defect to a known
category, and a number indicating how good the match is between the
defect’s particular characteristics and that category’s stored characteristics.
Some systems will also show the second-best match for the defect being
classified.
Applications. Early ADC systems in production-level fab focused
on a few specific types of layers in the front-end of the process. Continuing
developments in ADC technology have opened up the entire process for
automatic defect classification. There are several different types of review
stations available with various illumination sources (e.g., optical microscopes, confocal laser-scanning microscopes, scanning electron microscopes),
and ADC is available with many of these. ADC with white light and with
confocal laser-scanning review stations is being used currently in fabs for
classification of defects in both the front end and back end of the process
(back end being defined here as all process steps used for forming the metal
interconnects that wire together all of the transistors formed in front end
processing). Future technology advancements in ADC will see the inclusion
of ADC with scanning electron microscopes (SEMs). This will be particularly
important as the linewidths of devices decrease, and the identification of
defects smaller than 0.2 micron becomes necessary on a regular basis. For
further reading on automatic defect classification, please see Ref. 23.
Strengths and Limitations. Automatic defect classification reduces
the amount of time and resources necessary to do in-line monitoring of the IC
process. ADC is available integrated directly with the inspection tool, or
separated from the inspection tool in an off-line review station. The former
is useful because it minimizes the overall time between when the wafer is
loaded, and when the defects are classified. The latter is useful because the
inspection tool’s time is not occupied with ADC activities (both creating a
database and doing actual ADC), and it can therefore remain dedicated to
wafer inspection. (See Table 6.)
ADC using a white-light source provides excellent results on front-end
layers. Some challenges arise in the back-end of the process when the
topography of the surface of the device is complicated by the presence of
many etched layers. In this case consistent autofocus on the top surface can
be difficult with a white-light microscope, and this can affect ADC by
lowering the rate of successful defect re-detection. Fortunately, ADC results
using confocal laser scanning have demonstrated improvements in redetection rates on back-end layers. These systems have a restricted depth of
focus with variable height positioning that can be used to generate a
digitized 3-D surface image for automatic defect classification.[24]

280 Thin-Film Deposition Processes and Technologies
Table 6. Strengths and Limitations of ADC Systems
Automatic Defect Classification Systems
Strengths
• ADC is faster, more repeatable and more accurate than
classification by operators:
faster time and results
• White light or laser-confocal ADC can be used to reduce number of defects
needing slower SEM review

Limitations
• Images of typical defects in
each class need to be acquired for setup
• Differing defect types must
have different appearance to
be properly classified

• Flexibility in platform:
available off-line from, or
integrated with, the inspection system

4.6

Defect Data Analysis Systems

Data analysis became a more critical part of the defect reduction
process with the adoption of automated defect inspection systems, particularly patterned wafer inspection systems. Patterned inspectors resulted in a
dramatic increase in the amount of defect data generated for analysis. Defect
data analysis systems are now networked multi-user systems that provide
access to and analysis of inspection data throughout the fab, and incorporate
and correlate defect data with other metrology and parametric measurements
such as electrical test results.
The data analysis systems function in several ways:
Defect data analysis systems manage the information flow and
maintain a historical database of inspection, review, and metrology and
parametric data throughout the fab. Data analysis systems usually have a
large amount of memory allocated for data coming from all types of
inspection systems (data include defect density, size, coordinates, etc.) and
from review stations (i.e., defect images and classification information).
Typically this flow of information between the tools and the analysis
system is automated and requires little or no operator intervention.

Chapter 6: Metrology and Inspection

281

Defect data analysis systems provide basic SPC functionality and
real-time feedback for in-line monitoring of the process. The systems
provide basic graphical analysis such as wafer maps and trend charts.
Since the flow of data into the system is automated, the system can be
programmed to alert an operator or engineer automatically, and shut down
the process equipment, if the density of a given defect type at a particular
inspection point in the process exceeds set control limits.
Defect data analysis systems provide capabilities for correlating
defect data to yield data. Clustering algorithms group defects based on their
spatial relationships on the wafer. Identification of these clusters helps track
excursions in defect density—and yield—more intelligently. Clustering algorithms may also help identify the source of the defects, as some process
equipment may create defects having a distinct spatial signature. Partitioning analysis helps determine at what layer the defects first arose. This
information helps significantly to pinpoint the source of defects. Finally, the
capability of correlation of defect information with electrical test sort maps
establishes which defect types are yield-limiting.
The output of the data analysis system can be in a variety of formats:
wafer maps, Pareto charts or histograms, or tabular reports. These systems
are also able to output new wafer map review files with only certain defects
included. This may be useful, for example, for reviewing only defects added
at a given process layer.
For further reading on defect data analysis systems, see Ref. 25.
GLOSSARY
ADC

Automatic Defect Classification, a means of categorizing detected events by comparing their digital optical
images with reference images.

AFM

Atomic Force Microscope, an instrument derived from
a Scanning Tunneling Microscope and related to a stylus profilometer, used to form 3-D high resolution topographic images of solid surfaces.

ASIC

Application Specific Integrated Circuit.

Back end

See BEOL.

BCR

Bar Code Reader, used for wafer identification and
tracking.

282 Thin-Film Deposition Processes and Technologies
BEOL

Back End Of Line; the process steps used for forming
the metal interconnects that wire together the transistors formed in front end processing.

Bright-field

Describes a technique based on collection of the
specularly reflected light from the sample.

CCD

Charge Coupled Device; in this context describing a
type of digital camera.

CD

Critical Dimension, or smallest linewidth of a conductive line on an IC.

CMP

Chemical-Mechanical Polishing, a global planarization
technique.

COO

Cost Of Ownership, a metric used to evaluate semiconductor equipment.

C-V

Capacitance-Voltage.

CVD

Chemical Vapor Deposition, a technique commonly
used to deposit dielectric layers.

Dark-field

Describes a technique based on collection of the nonspecularly reflected (scattered) light from the sample.

Defectivity

The quality of having defects; the number of defects on
the wafer.

DRAM

Dynamic Random Access Memory, a type of semiconductor memory.

Electromigration The process resulting in current-induced open circuit
failure in metal interconnect lines.
FTIR

Fourier Transform InfraRed spectroscopy, used for
chemical compositional analysis.

GEM

Generic Equipment Model, a communications standard
used for factory automation (newer than SECS).

IC

Integrated Circuit.

In-line

Relating to measurements that occur in a processing
tool, outside of the process chamber.

In situ

Relating to measurements that occur within a process
chamber.

Chapter 6: Metrology and Inspection

283

Junction

The point at which the conductivity changes from ptype to n-type or vice versa.

MOCVD

Metal-Organic CVD, used for depositing certain conductive layers.

OCR

Optical Character Recognition, used for wafer identification and tracking.

OEE

Overall Equipment Effectiveness, a metric used to evaluate semiconductor equipment (newer than COO).

Off-line

Relating to measurements that occur on a stand-alone
metrology system in the manufacturing area.

Nomarski

Differential interference contrast microscope, a technique for detecting the phase difference between two
adjacent points on the sample. This method uses a
birefringent crystal, and polarized light, to separate the
incident light into two adjacent beams, and determines
the phase difference between the beams through interference.

Pareto analysis

A list of items contributing to a problem communicating the order of their importance.

PMT

Photo Multiplier Tube, a type of light detector chosen
for its fast response, that quantifies the amount of light
striking its active area per unit time.

PSL

PolyStyrene Latex sphere, a standard used to characterize the defect capture performance of defect inspection
systems.

PVD

Physical Vapor Deposition, commonly used to form
metal interconnects.

RC

Resistive-Capacitive, a time delay constant that affects
chip operation speed.

Recipe

An electronic file of system parameter values used to
control semiconductor processing or metrology
equipment.

RI

Refractive Index, the ratio of speed of light in a vacuum
to speed of light in a material.

284 Thin-Film Deposition Processes and Technologies
Rp , Rs

The Fresnel reflection coefficients, p and s polarized.

RTDC

Real Time Defect Classification; a first pass defect
classification that uses only information collected during defect inspection, such as intensity, size and spatial
distribution of collections of defects s and p polarization, the perpendicular and parallel components (respectively) of the polarization vector.

SECS

Semiconductor Equipment Communications Standard,
used for factory automation.

SEM

Scanning Electron Microscope; uses an electron beam
to produce very highly magnified images. Used for
surface viewing and cross sectional analysis of device
dimensions.

Slurry

An abrasive suspension of hard particles in a viscous
chemical solution used in chemical-mechanical polishing.

SIMS

Secondary Ion Mass Spectrometry, used for characterizing dopant and impurity distribution with depth profiles.

Solid angle

The 3-dimensional equivalent of angle, often defined
by azimuth and elevation.

SPC

Statistical Process Control, a method of tracking the
variations in process parameters to help identify out of
control situations.

SRP

Spreading Resistance Probe, used for characterizing
dopant distribution with depth profiles.

Stoichiometry

The chemical combination of a material composed of
other materials.

TDI

Time Delay Integration, in this context describing a
type of camera.

Yield

The percentage of wafers or die produced in an operation or process that conform to specifications.

Chapter 6: Metrology and Inspection

285

REFERENCES
1. Tompkins, H. G., et al., The Analysis of PECVD Silicon Nitride with
Spectroscopic Ellipsometry and RBS, International Conference on
Metallurgical Coatings and Thin Films, San Diego (1996)
2. Metrology Roadmap: A Supplement to the National Technology Roadmap
for Semiconductors, Technology Transfer #94102578A-TR, SEMATECH
(1995)
3. National Technology Roadmap for Semiconductors, Semiconductor Industry
Association, San Jose, CA (1994)
4. Hendricks, D., et al., Characterization of a New Automated ElectronBeam Wafer Inspection System, SPIE Integrated Circuit Metrology,
Inspection, and Process Control IX, 2439:1741–83 (1995)
5. Breaux, L., and Kolar, D., Automatic Defect Classification for Effective
Yield Management, Solid State Tech., 39(12):89–96 (1996)
6. Schroder, D., Semiconductor Material and Device Characterization, John
Wiley and Sons, New York (1990)
7. Aspnes, D. E., The Accurate Determination of Optical Properties by
Ellipsometry, Handbook of Optical Constants of Solids, (E. D. Palik,
ed.), Academic Press Inc., San Diego (1985)
8. Pickering, C., Nondestructive Characterization of Semiconductor
Multilayers, JOM, 46(9):60–64 (1994)
9. Nguyen, N. V., et al., High-Accuracy Principle-Angle Scanning
Spectroscopic Ellipsometry of Semiconductor Interfaces, Semiconductor
Characterization: Present Status and Future Needs, pp. 438–442, American
Institute of Physics, Woodbury, NY (1996)
10. Perloff, D. S., Gan, J. N., and Wahl, F. E., Dose Accuracy and Doping
Uniformity of Ion Implantation Equipment, Solid State Tech., 24(2) (1981)
11. Johnson, W., and Brennan, W., Non-destructive Determination of Step
Coverage on Semiconductor Wafers, Thin Solid Films, 270:467–471
(1995)
12. Non Contact Sheet Resistance Measurements on Low Resistivity Wafers,
KLA-Tencor Applications Note RS–18
13. Howland, R. S., et al., Using Calibration Curve Modeling for Scanning
Surface Inspection Systems, Micro, 13(7):61–71 (1995)
14. Swecker, A. L., et al., Comparison of Defect Detection Schemes Using
Rigorous 3-D EM Field Simulations, SPIE Metrology, Inspection and
Process Control for Microlithography XI, 3050:313–321 (1997)
15. van de Hulst, H. C., Light Scattering by Small Particles, Dover, New York
(1957)

286 Thin-Film Deposition Processes and Technologies
16. Altendorfer, H., et al., Unpatterned Surface Inspection for Next-Generation
Devices, Solid State Tech., 39(8):93–99 (1996)
17. Lessor, D. L., et al., Quantitative Surface Topography Determination by
Nomarski Reflection Microscopy: I. Theory, J. Opt. Soc. Amer.,
69(2):357–366 (1979)
18. Morita, E., et al., Distinguishing COPs from Real Particles, Semiconductor
International, 17(8):156–162 (1994)
19. Shen, J. J., and Cook, L. M., Meeting the Challenge of Submicron Defect
Characterization on Final-Polished Wafers, Micro, 15(3):53–62 (1997)
20. Larson, C. T., Measuring Haze on Deposited Metals with Light-Scatteringbased Inspection Systems, Micro, 14(8):31–38 (1996)
21. Bullis, W. M., Characterizing Microroughness and Haze on Silicon Wafers,
Micro, 14(1):47–53 (1996)
22. Howland, R. S., et al., High Speed Detection of Pattern Defects Using Laser
Scattering, Solid State Tech., 38(11):123–126 (1995)
23. Breaux, L., and Kolar, D., Automatic Defect Classification for Effective
Yield Management, Solid State Tech., 39(12):89–96 (1996)
24. Uritsky, Y., et al., An Integrated System for Rapid Process Defect
Evaluation, Solid State Tech., 38(6):61–66 (1995)
25. Burggraaf, P., Yield Analysis Software Solutions, Semiconductor
International, 19(1):79–85 (1996)