27

Mask Inspection: Theories and Principles

Anja Rosenbusch and Shirley Hemar

CONTENTS

27.1 Introduction
27.2 New Challenges: Sub-Wavelength Lithography and Defect Printability
27.3 Mask Defect Types

27.3.1 Types of Hard Defects
27.3.2 Different Types of PSM Defects
27.3.3 Minimum Defect Requirements

27.4 Inspection of Defects

27.4.1 Basic Principles of Mask Inspection
27.4.2 Aerial Imaging versus Mask Imaging
27.4.3 Aerial Image-Based Mask Inspection

Acknowledgments

27.1

Introduction

Photolithographic masks are the stencils or templates that are used to replicate the
integrated circuit patterns of VLSI devices on silicon wafers. Masks consist of a patterned
chrome film over a quartz substrate, or when more advanced, phase masks are produced
with MoSi or etched quartz. The desired pattern is transferred to the semiconductor wafer
by projecting the mask image at 4:1 or 5:1 reduction onto a photosensitive resist coating
with a ‘‘step and scan’’ lithography tool (stepper), developing the resist, and processing
the film through the resultant pattern.

A single mask can be used in the production of hundreds or even thousands of wafers

(each wafer containing from tens to hundreds of dies, each of which is processed to
become a fully functional device); therefore, an undetected error or defect in the mask can
cause significant loss in yield. To minimize this loss, all reticles are inspected for quality
control several times during their manufacturing process. Currently, there are three
inspection methods being used in the industry: die-to-database inspection, where the
mask image is compared to the design data; die-to-die inspection, where the images of
nominally identical dies within a mask are compared to each other; and contamination
inspection, where the mask is checked for non-pattern-related defects, e.g., particles.
When defects are found, the mask is repaired (if possible) or cleaned and reinspected.
Conventional mask inspection systems employ short wavelengths and high magnification
optics to detect defects down to 60 nm in size.

Rizvi / Handbook of Photomask Manufacturing Technology DK2192_c027 Final Proof page 589 7.3.2005 6:38pm

© 2005 by Taylor & Francis Group.

27.2

New Challenges: Sub-Wavelength Lithography and Defect
Printability

As the design rules shrink, designers must turn to sub-wavelength lithography that
involves printing features as small as 65 nm (or even smaller) with 193-nm illumination,
working below the diffraction limit; wherein the optics cannot faithfully reproduce the
desired pattern without the application of resolution enhancement techniques (RETs).
RETs include optical proximity correction (OPC), e.g., assist features that are sub-
wavelength non-printable features added to enhance the contrast, and phase shifting
masks (PSM) where the light from adjacent features undergoes phase inversion to in-
crease contrast and enhance resolution. Manufacturing high-quality RET masks becomes
increasingly difficult. That makes making masks fabrication as one of the primary chal-
lenges in the 65-nm technology node.

An ambiguous question in mask making is when to call a mask anomaly — pattern or

substrate related — an actual defect. Due to the cost awareness and the fact that mask
repair holds risk of damage not every anomaly is classified as defect. The current trend is
to classify defects on a mask based on their predicted impact onto the wafer. A defect is
classified as printable if it causes a CD variation larger than a certain specification (usually
ranging from 6–10%).

The actual printability of sub-wavelength defects is a complex issue: some may not

print at all causing no effect on yield, while others may produce large pattern errors,
effectively killing the device. The relative dimensional scaling between a mask defect and
the printed defect is termed mask error enhancement factor (MEEF), calculated as:

MEEF ¼

D

CD

wafer

4

D

CD

reticle

where the factor of 4 accounts for the nominal 4:1 stepper reduction ratio, and DCD refers
to the change in critical dimension (CD), e.g., the width of a line or the diameter of a
contact. In high MEEF pattern densities (>1), a small defect on the critical features of the
reticle might be enlarged when transferred to the wafer and cause a printable defect.
Detecting such a defect is crucial for guarantying ‘‘defect-free’’ masks.

Figure 27.1 presents two different scenarios. The first example consists of a set of three

smaller figures belonging to the left half of Figure 27.1. This set addresses the case of a
defective contact. The defect is located at the corner of the contact as shown in the CAD
design in the left figure of the set. The mask image (middle image of the set) shows the
defect clearly developed at the bottom right corner. The wafer CD SEM image (the right
figure from the set) shows the impact of the corner defect. Here the contact itself is slightly

FIGURE 27.1
Defects and their impact on the actual result on wafer: The final wafer result depends on the defect type and the
feature density around the defect location. On the left half of the set of three figures the result due to localized
defect is shown. It affects only the defect contact. On the right half, consisting of three figures, a high-MEEF area
is displayed. Here an isolated pinhole causes two neighboring contacts to bridge.

© 2005 by Taylor & Francis Group.

distorted at the bottom right. The contacts located next to the defective one are not or only
slighly affected. The scenario on the set of right half shows a contact array with an isolated
pinhole located in-between two neighboring contacts. The wafer result (right figure of this
set) actually shows a bridge between the two contacts resulting from the pinhole defect.
The impact of the pinhole is severe. The wafer/die might have to be scraped.

27.3

Mask Defect Types

Defects can be found on four different surfaces of the mask: the patterned area, the
backside of the mask, and the front and backside of the pellicle. The defects of most
interest to mask maker and mask user are the defects on the patterned surface, as they are
in focus while under stepper exposure and will most likely be transferred onto the wafer.
Figure 27.2 shows examples of defects on the four different surfaces. Defect type 1a is a
chrome defect at the patterned surface. Defect type 1b is a defect on the quartz surface of
the mask. A particle or additional chrome falls into this category. Defect type 1c shows a
transmission defect in the quartz surface, which is caused by the different quartz thick-
nesses or a thin residue of a solvent such as cleaning solution. Defect type 2 represents a
defect at the backside of the quartz. The defect specifications for this type of defect are
usually more relaxed, as they should not affect wafer-printing results. Defect type 3
shows a defect at the front side of the pellicle. Defects on the backside of the pellicle are
illustrated in number 4. The defect requirements for defects types 3 and 4 are less
aggressive as the pellicle is not in focus during lithography exposure; hence, defects on
these surfaces do not affect wafer results. Nevertheless, particles on the back side of the
pellicle may fall on the pattern during the lifetime of the mask thus detecting them is still
important.

Mask defects are commonly divided into two different categories: hard and soft defects.

A defect is called a hard defect if it is not possible to remove it by a cleaning process.
Added or missing features in the chrome, phase shifter, or absorber area fall into this
category. Pindots, as well as pinholes are called hard defects as well. A defect is called soft
defect if it is possible to remove it by a cleaning process. Particles, stains, contaminations,
such as crystals, and residue materials are soft defects.

27.3.1 Types of Hard Defects

Figure 27.3

illustrates some of the most common hard defects such as chrome extensions

(1), clear intrusions (2), and corner defects (3 and 4). In addition to that missing or added
features are called hard defects as well.

4

3

2

1a

1b

Pellicle

Chrome

Quartz

1c

1d

FIGURE 27.2
Defects on the four different surfaces of a
mask. Defect types 1 (a, b, c, and, d) are very
critical as they are in focus while being ex-
posed. These defects are quite likely to print
on the wafer. Defect type 2 located on the
backside of the glass and the defects of types
3 and 4 on the pellicle surfaces are less likely
to print.

Rizvi / Handbook of Photomask Manufacturing Technology DK2192_c027 Final Proof page 591 7.3.2005 6:38pm

© 2005 by Taylor & Francis Group.

Defects in the clear quartz area of the mask, such as pindots (5), scratches, and bubbles,

are classified as hard defects as well. Defects in the opaque areas of the mask, such as
pinholes (6), are in this category of defects. In addition to that there are defects related to
the transmission of the mask. Defects like transmission defects in the opaque (7) or
semitransparent defects on clear (8) are counted as hard defects.

In addition to these types of defects, any kind of feature mis-sizing and misplacements

are called hard defects. Figure 27.4 shows some more examples. Feature misplacement
(1) can be caused by errors in the original data preparation of a mask, as well as by mask
writer errors. Feature mis-sizing is another hard defect type. A feature can be mis-sized
in either direction, x or y (2). Missing and additional features are classified as hard
defects (3).

Another type of hard defects is global CD and/or quality change over the entire mask.

Edge roughness is one example, as well as global CD uniformity changes due to heating
or etching effects. As these changes usually occur gradually over the span of the mask
area, state-of-the-art mask inspection hardly detects them.

FIGURE 27.3
Classical hard defects illustrated on a binary mask.
Defect types are chrome extensions on edges (1) and
in corners (3), clear intrusions on edges (2) and corners
(4), as well as chrome (5), clear (6) and semitransparent
(7,8) spots known as pinholes and pindots.

FIGURE 27.4
Examples of hard defects: feature misplacement
(1) feature mis-sizing (2) and missing feature. (3)

1

2

3

2

2

Rizvi / Handbook of Photomask Manufacturing Technology DK2192_c027 Final Proof page 592 7.3.2005 6:38pm

© 2005 by Taylor & Francis Group.

27.3.2 Different Types of PSM Defects

Phase shifting masks (PSMs) suffer from the conventional hard defects as described in the
previous section. In addition to that, this type of mask is affected by phase-specific
defects. Figure 27.5 presents some of the more common phase shifting defects in embed-
ded attenuated PSM (EPSM). A mask might have additional phase material (types 1 and
2) or missing shifter material (type 3). In an embedded attenuated PSM, the absorber
material is slightly transmitive; hence, a shifter defect is also transmitive. In other words,
its lithography behavior is different from a conventional opaque defect. This presents
additional challenge to mask inspection systems, as the system has to support a different
rejection mechanism and specification for this kind of defects.

Another type of masks is alternating PSM. In alternating PSM (APSM), an additional

type of defect occurs due to the manufacturing process involved in generating an APSM.
In an APSM, the shifter area is generated, for example, by an additional etch step. This
step might produce defect as shown in Figure 27.6. Defect type 1 is an additional partial
shifter in the phase area. This defect might change the phase behavior of the mask. Type 2
presents additional full shifter; hence, the target CD (generated by the interference of non-
shifter and shifter areas) might not be guaranteed by the mask. Types 3 and 4 are more
conventional. Additional absorber is located in the shifter or non-shifter area. The inspec-
tion support of these types of mask is today still the biggest challenge for mask inspection.
Phase errors might only be seen at the stepper exposure wavelength. As the inspection
wavelength might differ from the stepper exposure wavelength, the inspection system
might not be able to detect a phase defect.

27.3.3 Minimum Defect Requirements

At each generation of semiconductor lithography a minimum defect size, based on
minimum gate width, has been defined. These are listed in the ITRS roadmap.

Figure

27.7

shows the minimum defect requirement as defined in the latest International

Technology Roadmap for Semiconductor (ITRS) 2003. In addition, more and more often
the minimum defect size criteria are being extended by defect capture criteria based on

180û,6%T

1

2

3

1808, 6%T

FIGURE 27.5
PSM defect types on an EAPSM: partially added shifter material (1), additional shifter (2), and missing shifter
material (3).

1

3

4

2

1

4

FIGURE 27.6
PSM defect types on APSM: additional partial shifter (1), additional full shifter (2), and additional chrome (3 and 4).

Rizvi / Handbook of Photomask Manufacturing Technology DK2192_c027 Final Proof page 593 7.3.2005 6:38pm

© 2005 by Taylor & Francis Group.

printability. Defects are classified based on their impact on CD variation in the aerial
image or wafer result.

27.4

Inspection of Defects

In this section, basic principles of mask inspection are explained. Mask makers have to
provide defect-free masks; hence masks must meet the minimum defect criteria as
defined by their customer. It is a usual practice for a mask to be inspected with sensitivity
settings below the actual sensitivity requirement in order to have a safety margin against
eventual inspection system flaws.

27.4.1 Basic Principles of Mask Inspection

Mask inspection is designed to verify the integrity of a mask. It could be compared to an
insurance policy. It ensures that the mask will meet customer requirements regarding
defectability. In simpler words, mask inspection verifies that the light is transmitted
through a mask as defined in the mask layout. There are two different methods to ensure
that.

The first method compares the printed mask features to their actual designs. This

method is called die-to-database inspection. Assuming that the mask making process,
which transfers the mask design layout into actual mask (chrome) features, had no errors,
a die-to-database inspection will detect all differences between the database and the mask
itself. If a difference violates the defect criteria given, it will then be classified as a defect.
Differences that are detected during an inspection, but do not violate the defect criteria,
are usually called nuisance. One of the system requirements for mask inspection systems
is to keep the number of these nuisances as small as possible (a usual number is 20).
Otherwise defect review becomes too time-intensive.

Figure 27.8 shows the die-to-database comparison of a gate feature. On the left is

the database image. The feature is well defined with straight edges and sharp corners.

Year

2003

2004

2005

2006

2007

2008

Minimum defect size (nm)

80

72

64

56

52

45.6

FIGURE 27.7
Minimum defect size criteria as defined in ITRS 2003.

FIGURE 27.8
Die-to-database inspection compares the feature on the
actual mask to the database. This inspection method is
applicable to all types of masks.

Database
pixels

Image
pixels

Rizvi / Handbook of Photomask Manufacturing Technology DK2192_c027 Final Proof page 594 7.3.2005 6:38pm

© 2005 by Taylor & Francis Group.

The image on the right side shows the binarized image of the actual mask feature. The
challenge of die-to-database inspection is to identify real mask defects distinguishing
them from systematic errors, such as edge roughness (etch step), or corner rounding
(mask writer) introduced by the mask manufacturing itself.

If the mask has more than one die, a second method, known as die-to-die inspec-

tion method can be applied. This method assumes that all dies of a mask are similar.
Die-to-die inspection between all dies will identify differences between different dies. The
drawback of this method is that it cannot identify a systematic error that can occur in all
dies, such as additional feature or a little extension as shown in the bottom of Figure 27.9.

Mask inspection is used more than once during the mask manufacturing process. Most

commonly a mask inspection (die-to-database) is performed after the mask etch step, in
order to identify all mask repair locations. Then another standard inspection is performed
after cleanup and pellicle to qualify the mask for customer shipment. As mask inspection is
used with relatively high frequency, one of the major aspects of tool selection is efficiency.

The efficiency of a system is based on several aspects, such as detection sensitivity,

capture rate, false alarm rate, inspection throughput, and comprehensiveness of the appli-
cation provided. One major factor is the algorithm efficiency and quality on which the
difference (defect) detection is based. The design of these algorithms is based on many
system-specific criteria, such as inspection wavelength, image capture mechanism, pixel
size and data transfer, and conversion (for die-to-database) rate. In addition, mask-related
factors such as mask reflectivity and transmission, mask quality aspects such as uniformity
and linearity (especially for die-to-database), edge roughness and pattern fidelity need to
be taken into consideration. The introduction of RETs like OPC and PSM increases the
complexity of mask inspection drastically. The treatment of subresolution features compli-
cates mask inspection as they usually are in the same intensity range as defects or nuisance.

Another aspect of inspection efficiency is the cost. Inspection costs are based on many

factors. The most important ones are data conversion and preparation times (which
become even more crucial with OPC and PSM due to data feature explosion or multi-
layer processing), number of inspections necessary, time of each inspection (throughout),
defect review, classification and disposition time. Mask inspection time is mainly driven by
the pixel size used. The smaller the pixel size is, the longer are data preparation and
conversion times. In conventional mask inspection systems, smaller pixel sizes are neces-
sary to obtain better detection sensitivity. With smaller pixel size, the probability of
nuisance (or false alarms) becomes higher too. Nuisance defects are differences in the
mask, which are below the defined mask defect specification requirement. Higher nuisance
rate impacts especially review and classification times, hence might impact the inspection
costs.

Besides guaranteeing that a mask arrives defect free at a customer, mask inspection also

serves as process control tool in the mask shop itself. Mask process issues might manifest
themselves by increased etch roughness or particle count, which can be caught by
conventional mask inspection using mask imaging.

Image
pixels

Image
pixels

FIGURE 27.9
Die-to-die inspection compares the dies of a multi-die reticle.
Differences between the different dies of a mask are
detected. Systematic errors like the small extension shown
in the bottom of the features cannot be identified by die-
to-die inspection.

© 2005 by Taylor & Francis Group.

27.4.2 Aerial Imaging versus Mask Imaging

The mask aerial image is the light intensity distribution at the wafer plane, as produced
by the stepper illumination source and projection optics. While the aerial image does not
predict the final pattern in the developed photoresist, it does faithfully reproduce all the
optical physics, including RET and MEEF. Therefore, an aerial imaging inspection tech-
nology will alert the operator of defects, which actually print, ignoring those that may be
present on the mask but have no impact on the final result.

Many reticle defects, such as extrusions, protrusions and proximity pinholes, and

pindots manifest themselves on the wafer as CD variation and contribute to CD non-
uniformity. Figure 27.10 depicts an example of a bump defect between two lines in an
alternating PSM. On the wafer this defect produces a CD variation in the lines and the
bump itself is not printed at all. Figure 27.11 shows an example of a clear protrusion at the
corner of a contact hole. Once again, the defect causes a CD variation on the wafer, as well
as shape asymmetry in the contact hole, but the defect itself is not resolved or printed.

Figure 27.12

illustrates the case of the actual MEEF by comparing the reticle and aerial

images of a contact defect in an attenuated PSM. The defect that happens to be a dark
intrusion at the lower edge of a contact can barely be seen by conventional means. The
error however becomes apparent by the CD variation estimated from image cross sec-
tions, measuring a 33% CD variation in the aerial image versus a 12% CD variation in the
reticle image, implying a MEEF factor of 2.75. The actual CD variation on the wafer
printed from this reticle was 35%, consistent with the aerial image measurement.

FIGURE 27.10
Line and space defect. Reading from the left to right: defect on the reticle (SEM image); an aerial image of
the reticle, and finally the last picture on the right shows the image ‘‘as printed’’ on the wafer (SEM image). The
phase bump between two lines causes a CD variation in the lines, as predicted by the aerial image.

FIGURE 27.11
Contact defect. From left to right: defect on the reticle (high-magnification optical image); the aerial image of the
reticle; and finally the image ‘‘as printed’’ on the wafer (SEM image). The clear protrusion from the contact
corner causes a CD variation and asymmetry in the contact on the wafer, as predicted by the aerial image.

© 2005 by Taylor & Francis Group.

These examples clearly illustrate how the defect in the aerial image can be much more

representative of the transferred pattern than the image of the defect on the reticle,
indicating that the aerial image of the mask provides more useful information on the
quality of the mask, compared to the magnified mask image.

All this leads to one of the basic problems of current solutions in mask defect detection

that is the discrepancy between those defects found by the inspection systems and those
of interest to the user. Advanced reticles using RETs are built on physical optical effects to
enhance image quality and reduce printable feature size. Sub-wavelength feature sizes,
especially in dense patterns, may result in high MEEF values. Inspection systems that do
not illuminate the reticle at the stepper wavelength and do not reproduce the physics of
the optical path might not be able to account for these effects in the course of inspection.

27.4.3 Aerial Image-Based Mask Inspection

Figure 27.13

presents a schematic comparison of the optical path of a stepper to that of a

proposed aerial image-based inspection system.

The aerial image-based mask inspection system inspects the aerial image of the reticle

for defects in the aerial image plane. The tool emulates the stepper optics, while looking at
a small area of the reticle with high magnification to enable defect detection.

The numerical aperture and illumination settings (s values) of the stepper can be set on

the inspection tool, and various off-axis illumination schemes (e.g., annular, quadrupole)
can be applied. The illumination used is an excimer pulse laser, which has the same
wavelength and employs the same technology used on steppers. This ensures that

Reticle image

Aerial image

FIGURE 27.12
Edge intrusion defect (bottom edge of contact), MEEF > 1. On the left is an optical image of the defect on the
reticle and on the right is the aerial image of the same defect. The plots show a vertical profile cross-section of the
defect area, where the reference is shown by the dotted lines and the defect is in solid lines. The CD variation of
the defective contact in the reticle is estimated to be 12%, whereas on the aerial image the CD variation is 33%,
implying a MEEF of 2.75.

Rizvi / Handbook of Photomask Manufacturing Technology DK2192_c027 Final Proof page 597 7.3.2005 6:38pm

© 2005 by Taylor & Francis Group.

all wavelength-related phenomena, such as phase shift and OPC, are inherently taken into
consideration. The inspection concept is a classical die-to-die inspection model, where the
identical dies of a reticle are compared with each other to look for defects. The inspection
scheme is straightforward:

.

Laser light is transmitted through the reticle, creating an aerial image of the
reticle by optics, which emulates the stepper optics (in hardware).

.

A 2-D CCD camera grabs the aerial images of the two compared dies.

.

The images are sent to an image-processing module where algorithms are
applied for finding inconsistencies between the two images.

Since the inspection is at-wavelength and using aerial imaging, issues, such as OPC,
MEEF, PSM (alternating or attenuated), are taken into consideration without overhead
or special effort. The reticle is inspected and qualified in a single step. The aerial images
can be later used during the review stage to perform resist modeling simulation and
process window analysis.

Acknowledgments

The authors would like to thank SEMATECH and Taiwan Semiconductor Manufacturing Company,
Ltd., for providing mask and wafer CD images used in this presentation.

Sigma apertures wheel

Reticle on holder

Imaging optics with

variable NA

CCD camera

Light source

Illumination

optics

Reticle

Projection

optics

Wafer

Stepper schematic

Aerial image inspection tool schematic

CCD camera

Imaging

optics

Reticle

Light source

Illumination

optics

Inspection tool uses

similar light source

Inspection tool

optics designed to

emulate stepper

optics with variable

NA and sigma

In inspection tool,

reticle is imaged on

CCD camera instead

of wafer

FIGURE 27.13
Schematic comparison of the optical path of a stepper to a proposed aerial image-based inspection system. The
inspection system uses a similar illumination source as the stepper typically does — an excimer pulse laser at the
same wavelength. The optics of the inspection system are designed to emulate the stepper optical path, including
varying NA and Sigma settings. The main difference between the stepper and the inspection system is that while
in the stepper the reticle image is reduced and imaged on a resist-coated wafer, in the inspection system the
reticle image is magnified and imaged on a CCD camera.

© 2005 by Taylor & Francis Group.