U.S. patent application number 11/781047 was filed with the patent office on 2008-01-24 for focusing method and apparatus.
Invention is credited to David Vaughnn.
Application Number | 20080021665 11/781047 |
Document ID | / |
Family ID | 38972494 |
Filed Date | 2008-01-24 |
United States Patent
Application |
20080021665 |
Kind Code |
A1 |
Vaughnn; David |
January 24, 2008 |
FOCUSING METHOD AND APPARATUS
Abstract
Methods and apparatus for placing wafers axially in an optical
inspection system. A "best worst" focus method includes a series of
through-focus images of a test wafer acquired using full field of
view of the inspection optics. The value of the worst quality in
each image is associated with the respective axial location,
forming a locus of "worst" values as a function of axial location.
The axial location is chosen which optimizes the locus, giving an
axial location that provides the "best-worst" image quality. A
"video focus" method includes a series of through-focus images
generated using reduced field of view. A figure of merit is
associated with each image, providing through-focus information.
The "video focus" can be calibrated against the "best worst" focus.
Further, a point sensor can be used to generate a single z-value
for one (x,y) location that can be calibrated with "video
focus".
Inventors: |
Vaughnn; David; (Edina,
MN) |
Correspondence
Address: |
DICKE BILLIG & CZAJA, PLLC;ATTN: CHRISTOPHER MCLAUGHLIN
100 SOUTH FIFTH STREET, SUITE 2250
MINNEAPOLIS
MN
55402
US
|
Family ID: |
38972494 |
Appl. No.: |
11/781047 |
Filed: |
July 20, 2007 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60832196 |
Jul 20, 2006 |
|
|
|
Current U.S.
Class: |
702/84 |
Current CPC
Class: |
G01N 21/8851 20130101;
G01N 21/9501 20130101 |
Class at
Publication: |
702/84 |
International
Class: |
G06F 19/00 20060101
G06F019/00 |
Claims
1. A method of determining a desired axial location for placement
of an imageable object in an inspection optical system, comprising:
placing an axially-translatable object within a field of view of
the inspection optical system; axially translating the object
sequentially to a plurality of axial locations; acquiring a
plurality of images of the translated object, each of the plurality
of images being associated with a respective axial location;
determining a plurality of lateral-location-dependent figures of
merit for the plurality of images, each of the plurality of
lateral-location-dependent figures of merit being associated with a
respective axial location; identifying a plurality of worst image
quality values from the plurality of lateral-location-dependent
figures of merit, each of the plurality of worst image quality
values being associated with a respective axial location;
determining a best worst image quality value from the plurality of
worst image quality values; associating the best worst image
quality value with a desired axial location, based on the
associations between the worst image quality values and the axial
locations; and selecting the desired axial location.
2. The method of claim 1, wherein each of the plurality of
lateral-location-dependent figures of merit is an edge transition
width.
3. The method of claim 2, wherein each of the plurality of worst
image quality values is a maximum edge transition width.
4. The method of claim 1, wherein each of the plurality of
lateral-location-dependent figures of merit is a contrast.
5. The method of claim 4, wherein each of the plurality of worst
image quality values is a minimum contrast.
6. The method of claim 1, wherein the best worst image quality
value is one of the plurality of worst image quality values.
7. The method of claim 6, wherein the desired axial location is the
axial location associated with the best worst image quality
value.
8. The method of claim 1, wherein the best worst image quality
value is interpolated between two of the plurality of worst image
quality values.
9. The method of claim 8, wherein the desired axial location is
interpolated between two of the plurality of axial locations.
10. The method of claim 1, wherein the axially-translatable object
has at least one imageable feature within the field of view of the
inspection optical system.
11. The method of claim 1, further comprising: placing a second
axially-translatable object within the field of view of the
inspection optical system; defining a reduced field of view of the
inspection optical system; axially translating the second object
sequentially to a second plurality of axial locations; acquiring a
second plurality of images of the second translated object, each of
the second plurality of images being associated with a respective
axial location, each image in the second plurality having the
reduced field of view; and selecting a second axial location based
on the second plurality of images and based on a predetermined
axial offset.
12. The method of claim 11, wherein the predetermined axial offset
accounts for effects within the full field of view of the
inspection optical system but outside the reduced field of view of
the inspection optical system.
13. The method of claim 11, wherein the second axially-translatable
object is different from the first axially-translatable object.
14. The method of claim 11, further comprising: measuring a third
axial location of a predetermined lateral location on the second
axially-translatable object with a point sensor; and calibrating
the point sensor to account for a difference between the second and
third axial locations.
15. A method of calibrating an optical inspection system,
comprising: calculating a "best worst" axial position for a first
objective lens using a full field of view of the first objective
lens; acquiring a series of through-focus images using a reduced
field of view of the first objective lens; and calibrating the
series of through-focus images to the "best worst" axial
position.
16. The method of claim 15, further comprising: axially translating
a wafer through focus; acquiring a series of through-focus images
of the wafer using a reduced field of view of the first objective
lens; and placing the wafer at the "best worst" axial position,
based on the series of through-focus images, and not based on
acquisition of any images using the full field of view of the first
objective lens.
17. The method of claim 15, wherein the "best worst" axial position
is obtained from analyzing a series of full field of view
through-focus images using the full field of view of the first
objective lens.
18. The method of claim 15, wherein calibrating a "best worst"
axial position comprises: assigning a video focus figure of merit
to each of the series of through-focus images; generating an
association between the video focus figure of merit and axial
position based on the series of through-focus images; and selecting
a video focus figure of merit corresponding to the "best worst"
axial position.
19. The method of claim 15, further comprising: measuring a point
sensor axial location with a point sensor; and calibrating the
series of through-focus images to the point sensor axial
position.
20. The method of claim 15, further comprising: repeating for a
predetermined length of time: axially positioning and inspecting
each of a series of wafers, the positioning being performed with a
point sensor, the inspecting being performed using the full field
of view of the objective lens; and calibrating the point sensor to
the series of through-focus images.
21. The method of claim 15, further comprising: repeating for a
predetermined number of wafers: axially positioning and inspecting
each of a series of wafers, the positioning being performed with a
point sensor, the inspecting being performed using the full field
of view of the objective lens; and calibrating the point sensor to
the series of through-focus images.
22. The method of claim 15, further comprising: repeating for a
predetermined number of inspections: axially positioning and
inspecting each of a series of wafers, the positioning being
performed with a point sensor, the inspecting being performed using
the full field of view of the objective lens; and calibrating the
point sensor to the series of through-focus images.
23. The method of claim 15, further comprising: repeating when an
accelerometer disposed on the inspection system is triggered:
axially positioning and inspecting each of a series of wafers, the
positioning being performed with a point sensor, the inspecting
being performed using the full field of view of the objective lens;
and calibrating the point sensor to the series of through-focus
images.
24. A method of inspecting a substrate, comprising: determining a
"best worst" axial location; acquiring a series of
reduced-field-of-view, through-focus images of a substrate;
correlating the reduced-field-of-view, through-focus images to the
"best worst" axial location; and placing the substrate at the "best
worst" axial location.
25. The method of claim 24, wherein determining a "best worst"
axial location comprises: acquiring a series of full-field-of-view,
through-focus images of a test object; determining a through-focus
worst image quality from the full-field-of-view, through-focus
images; selecting the best worst image quality from the
through-focus worst image quality; and selecting an axial location
corresponding to the best worst image quality to be the "best
worst" axial location.
26. The method of claim 24, further comprising: measuring a point
sensor axial location on the substrate with a point sensor; and
calibrating the point sensor axial position to the
reduced-field-of-view, through-focus images.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority under 35 U.S.C.
.sctn.119(e)(1) to U.S. Provisional Patent Application Ser. No.
60/832,196, filed Jul. 20, 2006, entitled "Method of Designing
Optical Systems," and bearing Attorney Docket No. A126.197.101; the
teachings of which are hereby incorporated by reference.
TECHNICAL FIELD
[0002] The present invention relates generally to a focusing
technique for visual inspection systems.
BACKGROUND
[0003] In a typical visual inspection system for semiconductor
wafers, all or a portion of a wafer is imaged onto a camera. The
image captured by the camera is typically processed by software to
determine the presence, absence, size and/or orientation of one or
more particular features on the wafer. The particular features may
vary, depending on the particular application of the inspection
system. As used herein, the term "wafer" is to be construed to be
inclusive of at least the following: an entire semiconductor wafer,
a portion of a single die formed on a semiconductor wafer, a single
die formed on a semiconductor wafer, multiple die formed on a
semiconductor wafer, and portions or pieces of an entire
semiconductor wafer (broken or sawn). The term "wafer" may also be
construed as any substrate suitable for visual inspection. Further,
the term "visual inspection" shall be construed to include any
process or apparatus that involves the focusing of electromagnetic
radiation of any wavelength at a given focal point or plane.
[0004] In general, it is desirable that the wafers be inspected
quickly, so that a relatively large number of wafers may be
inspected in a relatively short amount of time. There is continual
effort to improve the efficiency of the inspection system
algorithms and reduce the time required to inspect each wafer.
[0005] One step in a typical inspection system algorithm involves
axially placing each wafer at the desired imaging plane, with
respect to the imaging optics. Once the wafer is placed, the optics
image the wafer or a portion of the wafer. In general, this axial
location helps determine the quality of the image captured by the
camera. It is a continual challenge to balance high throughput of
the wafers through the inspection system (speed) versus accurate
placement of each wafer at the most desirable imaging plane
(accuracy).
[0006] Accordingly, there exists a need for increased efficiency in
ensuring that each wafer is placed at a desired imaging plane, with
respect to the imaging optics.
SUMMARY
[0007] One embodiment of a method of determining a desired axial
location for placement of an imageable object in an inspection
optical system, involves placing an axially-translatable object
within a field of view of the inspection optical system; axially
translating the object sequentially to a plurality of axial
locations; acquiring a plurality of images of the translated
object, each of the plurality of images being associated with a
respective axial location; determining a plurality of
lateral-location-dependent figures of merit for the plurality of
mages, each of the plurality of lateral-location-dependent figures
of merit being associated with a respective axial location;
identifying a plurality of worst image quality values from the
plurality of lateral-location-dependent figures of merit, each of
the plurality of worst image quality values being associated with a
respective axial location; determining a best worst image quality
value from the plurality of worst image quality values; associating
the best worst image quality value with a desired axial location,
based on the associations between the worst image quality values
and the axial locations; and selecting the desired axial
location.
[0008] Another embodiment concerns a method of calibrating an
optical inspection system, which may involve calculating a "best
worst" axial position for a first objective lens using a full field
of view of the first objective lens; acquiring a series of
through-focus images using a reduced field of view of the first
objective lens; and calibrating the series of through-focus images
to the "best worst" axial position.
[0009] A further embodiment is a method of inspecting a substrate
that may include determining a "best worst" axial location;
acquiring a series of reduced-field-of-view, through-focus images
of a substrate; correlating the reduced-field-of-view,
through-focus images to the "best worst" axial location; and
placing the substrate at the "best worst" axial location.
BRIEF DESCRIPTION OF THE DRAWINGS
[0010] FIG. 1 is a perspective drawing of an exemplary inspection
system.
[0011] FIG. 2 is a schematic drawing of an exemplary optical system
having field curvature, with three exemplary through-focus image
quality maps.
[0012] FIG. 3 is an exemplary plot of through-focus image quality,
and the locus of "worst" image quality.
[0013] FIG. 4 is a diagram illustrating a method of identifying a
"best-worst" position for a wafer.
[0014] FIG. 5 is a diagram illustrating a calibration process for
an optical inspection system.
[0015] FIG. 6 is a diagram illustrating a method of a inspecting a
wafer at a "best-worst" position of an optical system.
DETAILED DESCRIPTION
[0016] In the following detailed description of the invention,
reference is made to the accompanying drawings that form a part
hereof, and in which is shown, by way of illustration, specific
embodiments in which the invention may be practiced. In the
drawings, like numerals describe substantially similar components
throughout the several views. These embodiments are described in
sufficient detail to enable those skilled in the art to practice
the invention. Other embodiments may be utilized and structural,
logical, and electrical changes may be made without departing from
the scope of the present invention. The following detailed
description is, therefore, not to be taken in a limiting sense, and
the scope of the present invention is defined only by the appended
claims and equivalents thereof.
[0017] An exemplary inspection system is illustrated in FIG. 1. The
inspection system may include an optical sensor or camera 20, one
or more selectable focusing mechanisms 18, such as objectives or
lenses, an auxiliary sensor 14, and an inspection platform 12
coupled to a wafer alignment device 16 for moving the platform 12
relative to the camera 20.
[0018] The inspection platform 12 is in some embodiments a rotary
stage that may be equipped with a universal interface platform with
vacuum to provide a flexible interface for wafer and die package
fixturing. The platform 12 is defined such that it quickly mounts
and inspects substrates such as whole wafers, portions of wafers,
sawn wafers on film frame, die in gel pak, die in waffle-pak, MCM,
JEDEC trays, Auer boats, and other wafer and die package
arrangements and configurations. Note that as used herein, the term
"wafer" may be used interchangeably with the term "substrate"
and/or with any of the alternative items named in the preceding
sentence, among others.
[0019] The wafer alignment device 16, which aligns each and every
wafer at the same x, y, z, and .theta. location, may be a precision
system of rotary motors, ball screws, direct or belt driven motors,
worm or other gears, actuators, hydraulics, push rods, vacuums, or
other mechanical or electrical equipment for moving the rotary
stage either linearly or angularly to a precise desired
location.
[0020] The focusing mechanism 18 may be an optical imaging
mechanism with multiple optics therein for using different
inspection resolutions. A motorized microscopic turret allows for
selecting of the imaging optics from the multiple choices. For
instance, a number of optics, such as three or five optics, may be
supplied and typical choices include 1.25.times., 2.5.times.,
5.times., 10.times., 20.times., 50.times. and 10.times. objectives.
The motorized microscopic turret and discrete objectives provide
the means to select the optical magnification.
[0021] The camera system 20 or other visual inspection device is
for visual inspection of wafers and/or die and portions thereof.
The camera system may be any type of camera capable of high
resolution inspection. An example of one part of such a camera
system is a 3-CCD inspection camera used to capture die or other
images during defect analysis.
[0022] Focusing of the camera 20 is accomplished by facilitating
relative movement between a wafer positioned on inspection platform
12 and the camera 20. Relative motion between the inspection
platform 12 and the camera system 20 is along an axis defined by
the optical axis of the camera system 20, commonly designated the
"z" axis (the "x" and "y" axes commonly being arranged in the plane
of a wafer on the inspection platform 12). Relative motion between
the camera 20 and the inspection platform 12 is achieved by moving
the wafer alignment device 16 (to which is mounted the inspection
platform 12) with respect to the camera 20, the camera 20 with
respect to the inspection platform 12, or both, the one with
respect to the other.
[0023] It is instructive to provide a simplistic example of the
issues that affect the axial placement of each wafer with respect
to the optics in the inspection system. This simplistic example is
intended merely to clarify these issues, and should not be
construed as in any way limiting the scope of the present
disclosure.
[0024] FIG. 2 is a schematic drawing of the image quality at three
different through-focus planes, for a generic inspection system 10
in which the optics have a finite field curvature. This is a
simplified example; in practice, the optics may have other
aberrations in addition to field curvature, but the case of simple
field curvature is sufficiently illustrative for this example.
[0025] The optics 21 of the inspection system 10 are exemplified as
a simple single-element lens in FIG. 2, but may include any or all
of a light source, a fiber, a beam homogenizer, a waveplate, a
beamsplitter, a collimator lens, a camera, a detector and/or an
objective lens. Light emerges from the optics 21 and comes to a
focus, preferably at or near the wafer to be inspected. Light
reflects off the wafer (not shown) and returns through the optics
21 to a camera (not shown) or other imaging device that produces an
image of the wafer under inspection. The illumination optical
system may operate in bright field and/or in dark field, and the
collection optical system may also operate in bright field and/or
in dark field.
[0026] The optics 21 in this example have a finite field curvature,
so that rays 22 near the center of the field of view come to focus
at a particular plane 26, rays 23 away from the center of the field
of view come to focus at a plane 27 closer to the optics 21 than
plane 26, and rays 24 near the edge of the field of view come to
focus at a plane 28 still closer to the optics 21 than either plane
26 or plane 27. The locus of these foci forms a curved field 25. In
practice, the field 25 may have a much more complex shape than a
sphere or a paraboloid, but this simple shape of the curved field
25 is sufficient to show the issues associated with axial
positioning of the wafer.
[0027] Planes 26, 27 and 28 represent potential axial positions at
which the wafer may be placed for inspection. Elements 11, 13 and
15 are graphic representations of the image quality at planes 26,
27 and 28, respectively, and may be referred to as image quality
maps.
[0028] The notations in FIG. 2 of "excellent", "very good", "good",
"fair" and "poor" may correspond to a figure of merit suitable for
inspection systems, such as RMS spot size, RMS wavefront error,
edge transition width, encircled energy, Modulation Transfer
Function (MTF), Strehl Ratio, or any other suitable single-valued
figure of merit. The precise correlation between these terms and
numerical values is unimportant for the simplistic example of FIG.
2, but we may generalize that an "excellent" location in the image
(or, equivalently, in the field of view) has a smaller RMS spot
size, or a higher Strehl Ratio, etc., than a "very good" location,
and so forth. The notations should be interpreted literally, in
that "poor" may not truly be unacceptable, and so forth; these
notations exist only for relative comparison.
[0029] For image 11, at plane 26, the center of the image may have
"excellent" quality, so that features in this central portion are
clearly represented in the image. Away from this central portion,
the image quality erodes, becoming "good" and then "poor" at the
edge of the field of view. Features near the edge of the field of
view may be blurred or poorly resolved.
[0030] For image 13, at plane 27, the central portion of the image
may have "fair" quality. The portion surrounding the center may
have "very good" image quality, and the edge of the image may also
have "fair" quality.
[0031] For image 15, at plane 28, the central portion of the image
may have "poor" quality, the portion surrounding the center may
have "fair" image quality, and the edge of the image may have
"good" quality.
[0032] Note also that the image quality maps 11, 13 and 15 are
determined only by the optics 21 of the inspection system 10, and
are not affected by the wafers that are to be inspected. These
image quality maps generally do not vary significantly over time
unless changes are made to the inspection system optics.
[0033] Based in part on the image quality maps or on other analysis
of the image quality, the inspection system designer and/or
operator decides at which plane the wafers under inspection should
be placed. For instance, if it is decided that the wafers should be
located at plane 26, then for each captured image of a wafer, the
center of the field of view may look "better" than the edge of the
field of view. Similarly, if it is decided that the wafers should
be located at plane 28, then the edge of the field of view may look
"better" than the center.
[0034] Once a decision is made as to which plane the wafers should
be placed, it becomes an engineering challenge to repeatably place
each wafer at the chosen plane, to do it within a certain
positional tolerance and to do it quickly. Accordingly, much
attention is devoted in the following paragraphs to these issues.
First, the text describes a scheme for determining at which plane
the wafer should be placed. Following that, the text describes a
scheme for placing the wafer at the desired plane.
[0035] The following paragraphs describe a scheme for determining
at which plane the wafer under inspection should be placed, which
may be referred to as the "best worst" scheme.
[0036] For each of the three axial positions (or, equivalently,
focus positions), shown in image quality maps 11, 13 and 15,
certain portions of the image (or, equivalent, the field of view)
may have higher quality than other portions. For each axial
position 26, 27 and 28, one may find the "worst" portion of each
image. For instance, for position 26, the "worst" portion may be in
the corners of the image. For position 27, the "worst" portion may
be either in the center or in the corners. For position 28, the
"worst" portion may be in the center of the image. In general, for
any and all of the axial positions, there exists a value for the
"worst" quality in the image.
[0037] A preferable axial position is at or near the point at which
this "worst" quality in the image is optimized--literally the
"best-worst" point. For the simplistic example of FIG. 2, a
preferable axial position may be position 27, at which the worst
quality anywhere in the image is merely "fair"; on either side of
position 27, at positions 26 and 28, the worst quality anywhere in
the image is "poor".
[0038] FIG. 3 is an exemplary plot of "quality" versus focus
position, for all the locations in an image. Each curve represents
a particular location in the field of view, such as a pixel or
group of pixels. Each curve has its own local minimum (or maximum,
depending on the figure of merit), but the local minima (or maxima)
for all the curves do not necessarily lie at the same axial
location.
[0039] From these curves, one may form a locus of the "worst"
quality anywhere in the image, for a particular focus position.
Graphically, this locus may be obtained by examining a vertical
slice of the curves and using the maximum (or minimum, depending on
the figure of merit) value in the slice as a point in the
locus.
[0040] Once the locus of the "worst" quality is formed, one may
find the minimum value of the locus, which is literally the "best"
value of the locus of "worst" values. The focus position that
provides this "best worst" value is the desired focus position. In
other words, it is desirable that each wafer to be inspected be
placed at or near this "best worst" focus position.
[0041] Note that "quality", as used in the plot of FIG. 3, may
refer to any suitable single-valued figure of merit, such as RMS or
peak-to-valley spot size, RMS or peak-to-valley wavefront error,
Strehl Ratio, MTF at a particular spatial frequency, edge
transition width, and so forth. Note that this figure of merit may
correspond to the low-number-is-good condition shown in FIG. 3, as
is the case for spot size, wavefront error, edge transition width,
and so forth. Alternatively, the figure of merit may correspond to
a high-number-is-good condition, such as Strehl Ratio or MTF, in
which case the y-axis in FIG. 3 would be inverted. It is desirable
to use a figure of merit that may be measured, and even more
desirable to use a figure of merit that may be measured using the
actual optics of the inspection system and a test wafer. For this
reason, the edge transition width is a particularly desirable
figure of merit, which can be measured directly by the inspection
system when light reflected from a test wafer having sufficiently
small features is collected and detected by the camera/imaging
optics.
[0042] Note also that the curves in FIG. 3 may be much more
numerous in practice. For instance, if a detector or camera has a
sensor configuration of 640 pixels by 480 pixels, the number of
curves may be as large as 307,200. In practice, the point on the
"worst" locus corresponding to a given axial location is found by
taking the value of the "worst" image quality of these 307,200
pixels.
[0043] In addition, the actual curves may have much more complex
shapes than those in FIG. 3. Depending on the optical system, the
curves may have local maxima or minima, or may curve more steeply
on one side than the other.
[0044] In practice, the precise number and shape of the curves are
relatively unimportant. The point on the locus of "worst" quality
may be found for each focus location by selecting the "worst" value
of image quality for each image's pixels or subgroup of pixels, and
the "best" value on this locus provides the desired axial location
for the wafer to be inspected.
[0045] Note that the images are taken at discrete planes,
through-focus. The "worst" value of image quality may be
interpolated between these discrete planes to form a continuous
function. In this manner, the "best-worst" value may be associated
with a plane that is not necessarily one of the planes at which the
images are taken.
[0046] Note that the focus position shown in FIG. 3 and described
above is typically fairly close to the position at which the
"quality" variation (best-to-worst, or peak-to-valley) is
minimized. In many camera-based inspection systems, it is desirable
to minimize or reduce the variation in quality across the
field-of-view, rather than optimize one location in the field at
the expense of another location in the field. In other words, it
may be acceptable in many inspection applications to have a
slightly degraded image, as long as the degradation is roughly
uniform over the field-of-view. The variation in such a degradation
across the field of view may be sufficiently reduced at or near the
"best worst" focus position described above.
[0047] In some embodiments, it is desirable that the axial
placement of the wafer be within the depth of focus of the imaging
system. The depth of focus may be defined in several ways, such as
an axial displacement that degrades the image by a particular
amount. This degradation due to defocus may be an increase in RMS
spot size by a particular percentage such as 10%, or an increase in
the RMS wavefront error to a particular threshold such as 0.07
waves, or a decrease in the Strehl Ratio to a particular threshold
such as 80%. In some embodiments, the depth of focus may depend
linearly on the central wavelength of the illuminating light, and
may depend inversely as the numerical aperture of the collection
optics, squared.
[0048] The "best worst" focus condition described above may be
found experimentally, by stepping a test wafer through focus, and
acquiring and processing the through-focus images of the test
wafer. Note that this "best worst" focus condition does not depend
on the wafer under inspection, but depends only on the
illumination, collection and/or imaging optics in the inspection
system. For instance, if the "best worst" focus condition in a
particular system is such that the wafer should be placed 1.0 mm
away from a particular feature on the objective lens, then as long
as the optics remain aligned and thermally stable, the 1.0 mm
criterion will hold, regardless of any properties of the wafer
under test. In other words, once the "best worst" focus condition
is established for a given optical system, it generally need not be
re-established until something in the system changes.
[0049] The preceding paragraphs describe a particular "best worst"
axial position. An image of a wafer placed at the "best worst"
axial position generally has a reduced variation in image quality
across the field-of-view (or, equivalently, across the spatial
extent of the image). It is an ongoing engineering challenge to
place wafers at or near that axial position, and to place them
rapidly and accurately. The following paragraphs describe several
focusing methods and their use together to address this
challenge.
[0050] The following paragraphs describe use of a so-called "point
sensor", which is shown schematically in FIG. 1 as an "auxiliary"
sensor 14. The point sensor takes data in one (x,y) location and
provides z-data, which may be the distance between the point sensor
and the wafer under inspection or a scaled or shifted version
thereof. In some embodiments, the point sensor is a largely
self-contained unit that is attached to the mechanical supports
that house the camera optics. Alternatively, the point sensor may
be integrated into the camera optics or into the mechanical
supports themselves. The point sensor may be arranged so as to
measure a z-height at an (x,y) location located within the field of
view of the optics, or, alternatively, at an (x,y) location outside
the field of view. Point sensors may use triangulation, confocal
techniques, or any suitable technique that provides z-height
data.
[0051] An advantage to using point sensors is that they operate
quickly. A typical point sensor can return a z-value in about 1
millisecond, and can run in a repeat mode with a frequency on the
order of 1 kHz. The response of a point sensor may be fast enough
to produce measurements on moving wafers, as the wafers are
translated past the (relatively stationary) point sensor location
by the handling robotics of the inspection system.
[0052] For instance, a point sensor may be used to measure a global
object position (analogous to measuring piston error in an
interferometer) by making a z-measurement at the center of the
wafer under inspection. The point sensor may make a tip-tilt
measurement by measuring z at 4 or 5 points along the circumference
(and optionally the center) of the wafer (analogous to measuring
tilt in an interferometer). Furthermore, the point sensor may map
out a focus map of the wafer by taking z measurements in a loose
grid along the wafer surface (analogous to measuring low-order
aberrations in an interferometer).
[0053] For all of the above measurements, the point sensor
typically remains fixed or stationary, and the wafer is typically
translated and/or rotated past the (x,y) measurement spot.
Alternatively, the wafer may remain relatively stationary while the
point sensor moves, or both the point sensor and the wafer may
move.
[0054] In addition, a point sensor may operate with a featureless
wafer. Because no wafer features are needed to produce a
z-measurement, there is no need to use a test wafer with a point
sensor.
[0055] Despite the numerous advantages of using a point sensor, a
point sensor alone is generally insufficient to fully inspect a
wafer, and a camera or imaging system is typically used in
conjunction with the point sensor. Because the point sensor and
imaging optics may be in separate mechanical housings, there may be
a small amount of z-drift over time between the point sensor and
the imaging optics. The effect of this drift may be that when the
point sensor z-measurement indicates that a wafer is in the desired
plane for high-quality imaging, the wafer may actually be axially
displaced away from the desired plane by the amount of the drift.
This z-drift may be caused by temperature changes, excessive
building vibration or any other perturbing feature from the
environment.
[0056] In order to correct for the drift between the point sensor
and the camera or imaging optics, a calibration may be done using a
series of through-focus images of the wafer, taken by the camera or
imaging optics. This calibration may be referred to as "video
focus".
[0057] The through-focus images may use a limited portion of the
full field of view, such as the central portion, so that they may
be acquired and processed more quickly. The term "video focus field
of view" may be used to refer to the limited field of view, and the
term "inspection field of view" may be used to refer to the full
field of view.
[0058] Each of the through-focus images may be processed by a
so-called "fast" algorithm that can find a desirable focus. For
instance, an algorithm that measures the (point-by-point) contrast
of an image may be used on each of the through-focus images, with
the most desirable focus occurring at the z-location where contrast
is maximized. Similarly, an algorithm using the edge transition
width may also be used, with the most desirable focus occurring
when the edge transition width is minimized. Many of these
algorithms require one or more features on the wafer, so that if an
inspection system is configured to test only featureless wafers, a
test wafer should be used for "video focus". The test wafer may be
stored in or with the inspection system. The test wafer may have
one or more imageable features on it, such as a line, an edge, a
corner, a point, an interface, a fiducial mark or index, and so
forth.
[0059] Despite being referred to as "fast", the so-called "fast"
algorithm still takes significantly longer to perform than a
measurement from a point sensor. With current computing power, a
typically video focus procedure (acquiring and processing a series
of reduced-field-of-view images through focus) may take on the
order of 3 seconds.
[0060] The "video focus" procedure may be used to calibrate the
z-drift between the point sensor and the imaging optics as follows.
A "video focus" procedure is run on a particular feature on a
wafer, producing a series of through-focus images, each image
associated with a particular z-location. An algorithm is applied to
the images, sensing contrast, edge transition width or any other
suitable metric. The algorithm images generate the metric as a
function of z-location, and a desired z-location is chosen based on
the metric having a maximum value, a minimum value, a predetermined
value, or entering a particular range of values. The particular
feature is then translated and/or rotated to lie at or near the
(x,y) location of the point sensor, if required. The point sensor
produces its own value of z-location of the wafer at the particular
feature. The z-locations of the video focus and point sensor may
then be compared, with a difference in reported z-values
corresponding to a drift between the point sensor and the imaging
optics. Alternatively, the point sensor procedure may be performed
before the video focus procedure.
[0061] Ultimately, the calibration procedure between the video
focus and point sensor ensures that drift between the point sensor
and the imaging optics is accounted for. In this manner, the point
sensor may be used (either in a closed-loop feedback loop or in
open-loop) to quickly move a particular wafer under inspection to
the proper axial location.
[0062] Acquiring the readings from the point sensor to find the
proper axial location for the wafer is practically instantaneous,
requiring only about 1 millisecond per measurement. Many point
sensor measurements may be taken as needed as the wafer under
inspection is moved into the desired axial location. When a point
sensor is used, the positioning of the wafer at the desired axial
location is limited only by the speed of the positional system in
the inspection system. In contrast, if only the video focus is used
to find the proper axial location, it may take about 3 seconds per
lateral location on a wafer, which is significantly slower.
[0063] This calibration between the point sensor and the video
focus may be performed at periodic intervals, such as once an hour,
or any other suitable regular or irregular interval. Running a
3-second-long procedure once an hour is certainly preferable to
running the 3-second-long procedure for every lateral location on
each wafer under inspection.
[0064] Note that the desired axial location for the wafer may
satisfy the "best worst" condition described above, or may be any
suitable axial location.
[0065] Once the wafer is brought to the desired axial location, a
full-field-of-view image is taken, which may be then processed and
analyzed for inspection. Note that the full-field-of-view image may
be taken using the same optical system as the video focus, where
the video focus uses a reduced field-of-view.
[0066] It is worth noting that the "video focus", which uses a
reduced field of view, may optionally include an axial offset with
respect to the "best worst", which uses the full field of view. For
instance, if the "video focus" uses only the center of the field of
view, and the optical system has the characteristics shown in FIG.
2, then the "video focus" used without an axial offset may
erroneously indicate that plane 11 is the most suitable plane for
the wafers, rather than plane 13. In other words, the "video focus"
may not be sensitive to effects outside its limited field of view.
Because of this, there may be a calibrated axial offset built into
the "video focus" routine, which can account for effects that lie
outside the limited field of view. This predetermined axial offset
may be set up during construction of the optical system, or may be
manually adjustable by an operator.
[0067] It is worth noting that the axial placement of an object
typically has a tolerance associated with it, which depends on the
mechanical positioning element(s) in the inspection system. The
tolerance may vary with the specific application, and may
optionally vary with the desired magnification of the optical
system. In general, placement "at" a particular plane may refer to
placement at the plane to within the tolerance.
[0068] It is instructive to compare several of the focus methods
described above to each other, in terms of the frequency at what
they may be performed and the duration of such a technique.
[0069] The "best worst" technique may be performed quite
infrequently, since there is essentially no drift between the "best
worst" optics and the "video focus" optics; they both use the
identical optical path. Finding the "best worst" axial position
(or, equivalently, focus position) should be done once, when the
optical system is first assembled. It may be done on a periodic but
infrequent basis, such as once a year, if there is a noticeable
degradation in the performance of the optics. For production, one
would not want to perform the "best worst" technique for every
wafer; such a technique may take up to four hours or more, which is
unsuitable for a high-throughput production environment.
[0070] The "video focus" technique may be performed periodically in
a production environment. For instance, the video focus may be
performed once for a predetermined length of time, such as an hour
or a week. Or, the video focus may be performed once for every few
wafers inspected, such as 10 wafers, 100 wafers, or any suitable
number. Alternatively, the video focus may be performed for a given
number of inspections, or actual number of images taken. As a
further alternative, it may be performed whenever the inspection
system is bumped, such as when an accelerometer with the inspection
system is triggered. In general, a video focus may be performed as
needed to keep the throughput high, and may be performed to address
fairly fast drift changes, such as a change in temperature or when
the inspection system is powered up or down. It may be performed in
production as needed, since it takes only about up to three seconds
or so to run.
[0071] The "point sensor" technique may be performed for every
wafer, and may be performed at multiple locations on every wafer if
needed. The configuration of the point sensor measurement points
may depend on resolution requirements, and may vary depending on
the specific application. The time required for a point sensor
measurement is minimal, on the order of about 1 millisecond.
[0072] In general, it would be desirable to use the point sensor
technique exclusively to place a wafer at the desired axial
location, due to the great speed of the technique. However, the
point sensor will inevitably begin to drift axially away from the
imaging optics. When this drift occurs, a video focus may be
performed to re-establish calibration between the point sensor and
the imaging optics. This re-calibration may be performed
periodically or as needed, but need be performed only as often as
required. Between these recalibrations, the point sensor may be
used exclusively for placing wafers axially.
[0073] It is instructive to trace through several exemplary
procedures and illustrate them in the flowcharts of FIGS. 4-6.
[0074] FIG. 4 is a flowchart of a procedure 40 to determine the
"best worst" axial position.
[0075] In element 41, the test object is positioned at an initial
axial location. Here the test object may be a test wafer or portion
of a wafer, and may preferably include at least one feature, such
as a line, and edge or a corner.
[0076] In element 42, the test object is translated axially away
from the initial location to an intermediate axial location. In
other words, the test object is stepped through focus. The step
interval may be predetermined.
[0077] In element 43, the optical system acquires a
full-field-of-view image of the test object at the intermediate
axial location.
[0078] In element 44, the inspection system checks to see if the
object is located at a final location, so that the initial and
final axial locations surround the "best worst" axial location. The
initial and final axial locations may be predetermined in an open
loop fashion by simulation, or may be determined experimentally by
measurements, or may be determined in an ad hoc manner during this
process. If the object is not yet at the final location, elements
42 and 43 are repeated.
[0079] In element 45, for each of the images acquired in element
43, a lateral-location-dependent figure of merit is calculated. For
instance, the edge transition width may be calculated for each
pixel in the image, or a predetermined group of pixels in the
image. Once calculated, the inspection system has values for the
edge transition width (or other suitable figure of merit) as a
function of (x,y) within each image, and also as a function of z
for the multiple images. The figure of merit may be left as
discrete values, or alternatively may be smoothed in any or all of
x, y and z.
[0080] In element 46, for each image, or, equivalently, for each
z-location if the z-values are smoothed, the worst figure of merit
for all values of (x,y) is selected for a particular z. In element
47, this worst value is associated with the respective z-value. In
this manner, elements 46 and 47 may form a locus of worst values,
with a z-dependence.
[0081] In element 48, the best value may be selected from the locus
of worst values.
[0082] In element 49, the z-value associated with the best value
from element 48 is selected and is chosen to be the "best worst"
axial location.
[0083] FIG. 5 is a flowchart of a calibration procedure 50 for an
optical inspection system.
[0084] In element 40, the "best worst" axial position is
determined. This is shown in greater detail in FIG. 4. Note that
this procedure 40 is typically performed once, when the optical
system is initially constructed, such as when an objective lens is
installed or replaced. Typically, procedure 40 need not be
performed with any regularity.
[0085] In element 51, the inspection system calibrates the "video
focus" to the "best worst" axial position obtained from element 40.
The "video focus" may use a reduced field of view, compared to the
full field of view of element 43 (FIG. 4). Optionally, the "video
focus" may use a relatively large field of view but may reduce the
resolution or number of pixels in the images.
[0086] Once element 51 is completed, the inspection system will
have a relationship between a "video focus" figure of merit and the
"best worst" axial location. In this manner, the inspection system
can run a "video focus", acquiring and processing a series of
reduced field-of-view through-focus images. The system can then
place a wafer under test at or near the "best worst" axial location
based on these reduced field of view images, rather than using the
full field of view of the optics, which takes more time to
process.
[0087] In element 60, the inspection system can test wafers or
portions of wafers. Element 60, the testing of wafers, is shown in
greater detail in FIG. 6.
[0088] In general, it is desirable to use a point sensor to place
wafers as much as possible, because the point sensor measurement is
very fast. However, the point sensor may drift with respect to the
inspection optics, due to thermal changes, vibration and so forth.
Therefore, the inspection system may calibrate the point sensor to
the "video focus" periodically and/or as needed. In this manner,
the point sensor may be used to place wafers at or near the "best
worst" axial location, until the drift significantly affects the
wafer placement.
[0089] In element 61, a wafer is placed at the "best worst" axial
location using "video focus". In element 62, the point sensor is
calibrated to the "video focus". The point sensor may use wafers
with or without features, and the "video focus" may use only wafers
with features. If the inspection system is configured to test
featureless wafers, then element 62 may be performed with a test
wafer that has features, which may be kept in the inspection
system.
[0090] A wafer under test is loaded in element 63 and is placed at
the "best worst" axial location using the point sensor in element
64. Once suitably placed, the wafer is imaged in element 65,
typically with the full field of view of the optics. The wafer is
then unloaded in element 66.
[0091] If there are no more wafers to be tested, the system stops
in element 68. If there are further wafers, the system
continues.
[0092] If the point sensor needs calibration, then the system
performs element 62. If no calibration is needed, element 63 is
performed.
[0093] Although specific embodiments of the present invention have
been illustrated and described herein, it will be appreciated by
those of ordinary skill in the art that any arrangement that is
calculated to achieve the same purpose may be substituted for the
specific embodiments shown. Many adaptations of the invention will
be apparent to those of ordinary skill in the art. Accordingly,
this application is intended to cover any adaptations or variations
of the invention. It is manifestly intended that this invention be
limited only by the following claims and equivalents thereof.
* * * * *