U.S. patent application number 11/173665 was filed with the patent office on 2005-12-01 for calibration as well as measurement on the same workpiece during fabrication.
Invention is credited to Borden, Peter G., Li, Jiping, Madsen, Jon.
Application Number | 20050264806 11/173665 |
Document ID | / |
Family ID | 25522194 |
Filed Date | 2005-12-01 |
United States Patent
Application |
20050264806 |
Kind Code |
A1 |
Borden, Peter G. ; et
al. |
December 1, 2005 |
Calibration as well as measurement on the same workpiece during
fabrication
Abstract
A method of fabricating a wafer includes forming a portion of
the wafer, making a first measurement in the wafer using a first
process, making a second measurement in the wafer using a second
process each time the first measurement is made, using one of the
first measurement and the second measurement to calibrate the other
of the first measurement and the second measurement, and changing a
process control parameter used in forming the portion of the wafer
depending on the first measurement and on the second
measurement.
Inventors: |
Borden, Peter G.; (San
Mateo, CA) ; Li, Jiping; (Fremont, CA) ;
Madsen, Jon; (Mountain View, CA) |
Correspondence
Address: |
PATENT COUNSEL, LEGAL AFFAIRS
APPLIED MATERIALS, INC
M/S 2061
P.O. BOX 450A
SANTA CLARA
CA
95052
US
|
Family ID: |
25522194 |
Appl. No.: |
11/173665 |
Filed: |
July 2, 2005 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
11173665 |
Jul 2, 2005 |
|
|
|
09974571 |
Oct 9, 2001 |
|
|
|
6940592 |
|
|
|
|
Current U.S.
Class: |
356/326 |
Current CPC
Class: |
G01N 21/95607 20130101;
G01N 21/9501 20130101; G01N 21/274 20130101; G01N 21/55
20130101 |
Class at
Publication: |
356/326 |
International
Class: |
G01B 009/02 |
Claims
1. A method of fabricating a wafer, the method comprising: forming
a portion of a wafer; making a first measurement in the wafer using
a first process; making a second measurement in the wafer using a
second process each time said first measurement is made; using one
of the first measurement and the second measurement to calibrate
the other of the first measurement and the second measurement; and
changing a process control parameter used in forming the portion of
the wafer depending on the first measurement and on the second
measurement.
2. The method of claim 1 wherein: said second measurement is used
to calibrate said first measurement; said using comprises: based on
the second measurement, generating a model of a property of the
portion of the workpiece as a function of the first measurement;
and looking up the model to determine a value of the property,
based on the first measurement.
3. The method of claim 1 wherein: said changing of process control
parameter is done only if the value of the property exceeds or
falls below a predetermined limit.
4. The method of claim 1 further comprising: repeating said first
measurement and said second measurement in said wafer a plurality
of times.
Description
CROSS-REFERENCE TO PARENT APPLICATION
[0001] This application is a divisional application of U.S.
application Ser. No. 09/974,571 filed Oct. 9, 2001, by Peter G.
Borden et al that is incorporated by reference herein in its
entirety.
CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application is related to and incorporates by reference
herein in their entirety, the following commonly owned U.S. patent
applications:
[0003] Ser. No. 09/799,481, entitled "USE OF A COEFFICIENT OF A
POWER CURVE TO EVALUATE A SEMICONDUCTOR WAFER" filed Mar. 5, 2001,
by Peter G. Borden et al;
[0004] Ser. No. 09/544,280, entitled "APPARATUS AND METHOD FOR
EVALUATING A WAFER OF SEMICONDUCTOR MATERIAL," filed Apr. 6, 2000,
by Peter G. Borden et al which is a continuation of Ser. No.
09/095,804 filed Jun. 10, 1998;
[0005] Ser. No. 09/274,821, entitled "APPARATUS AND METHOD FOR
DETERMINING THE ACTIVE DOPANT PROFILE IN A SEMICONDUCTOR WAFER,"
filed Mar. 22, 1999, by Peter G. Borden et al.;
[0006] Ser. No. 09/521,232, entitled "EVALUATING A PROPERTY OF A
MULTILAYERED STRUCTURE" filed Mar. 8, 2000, by Peter G. Borden et
al; and
[0007] Ser. No. 09/788,273, entitled "EVALUATING SIDEWALL COVERAGE
IN A SEMICONDUCTOR WAFER" Feb. 16, 2001, by Peter G. Borden et
al.
BACKGROUND
[0008] In a damascene process, a stack of dielectric layers is
first laid down on a semiconductor substrate that includes
underlying layers of devices and interconnects, to form a structure
that is eventually cut into multiple dies. The dielectric layers
serve various functions, such as anti-reflection coating,
insulating and etch-stopping. Grooves are then etched in the
dielectric stack. The grooves are then filled with a conductive
metal such as copper using a process such as plating. Finally, an
exposed surface of the resulting structure is polished, to leave
metal lines inlaid within the grooves.
[0009] There are several methods for measuring thickness, or change
in thickness, of an upper-most layer in which the metal lines are
formed. For example, in a method called "stylus profilometry" a
stylus is run along the exposed surface and the height of the
stylus is measured. This method requires contact to the surface of
the structure and also requires that the substrate be precisely
level prior to the damascene process. In addition, the stylus tips
are fragile and require frequent replacement. As another example,
focused ion beam scanning electron micrography (FIB-SEM) uses a
focused ion beam to cut a hole in the structure. A scanning
electron micrograph is then used to image the exposed
cross-section. In a related method, the focused ion beam is used to
cut out a section, which is then viewed with a transmission
electron microscope (TEM). Such methods are slow, destructive, and
not suited for monitoring the fabrication process.
[0010] Dielectric film thickness is also measured using
ellipsometry. In one such method, light at multiple wavelengths is
shone on a surface at an angle and the reflection is measured as
function of incident polarization angle. Applicants note that a
measurement made by this method uses a large spot size because a
source of white light cannot be focused to the diffraction limit of
a single wavelength. Applicants further note that ellipsometry does
not provide the resolution required to measure profiles that vary
across a distance (e.g. 20 .mu.m) that is of the same order of
magnitude as the spot size. Applicants also note that use of
polarization as part of ellipsometry means polarization cannot be
used to obtain measurements of dielectric properties within an
array of metal lines.
[0011] U.S. Pat. No. 5,978,074 (which is incorporated by reference
herein in its entirety) discloses an apparatus for characterizing
multilayer samples. The apparatus focuses an intensity modulated
pump beam onto the sample surface to periodically excite the
sample, and also focuses a probe beam onto the sample surface
within the periodically excited area. The power of the reflected
probe beam is measured by a photodetector. The output of the
photodetector is filtered and processed to derive the modulated
optical reflectivity of the sample. Measurements are taken at a
plurality of pump beam modulation frequencies. In addition,
measurements are taken as the lateral separation between the pump
and probe beam spots on the sample surface is varied. The
measurements at multiple modulation frequencies and at different
lateral beam spot spacings are used to help characterize complex
multilayer samples. In the preferred embodiment, a spectrometer is
also included to provide additional data for characterizing the
sample.
[0012] Regarding use of a spectrometer, U.S. Pat. No. 5,978,074
states (at column 9, line 58 to column 10, line 10) "In the
preferred embodiment, the subject apparatus further includes a
spectrometer for providing additional data. As noted above, a white
light source 120 is necessary for illuminating the sample for
tracking on a TV monitor. This same light source can be used to
provide spectral reflectivity data. As seen in FIG. 1, a beam
splitter can be used to redirect a portion of the reflected white
light to a spectrometer 142. The spectrometer can be of any type
commonly known and used in the prior art. FIG. 4 illustrates one
form of a spectrometer. As seen therein, the white light beam 122
strikes a curved grating 242 which functions to angularly spread
the beam as a function of wavelength. A photodetector 244 is
provided for measuring the beam. Detector 244 is typically a
photodiode array with different wavelengths or colors falling on
each element 246 in the array. The outputs of the diode array are
sent to the processor for determining the reflectivity of the
sample as a function of wavelength. This information can be used by
the processor during the modeling steps to help further
characterize the sample."
[0013] Use of the white light for aligning the sample implies that
the spectrometer is shown through the measurement objective lens.
This is because the view for alignment must be the same as the view
for measurement. Therefore, the spectrometer must be combined with
the two laser measurement, adding complexity.
[0014] Applicants note that U.S. Pat. No. 5,978,074 is silent on
how to "further characterize the sample," other than to describe
determining the sample's reflectivity as a function of wavelength
as noted above. Applicants further note that U.S. Pat. No.
5,978,074 is also silent on what is done after a sample has been
"further" characterized.
[0015] U.S. Pat. No. 5,978,074 also cites U.S. Pat. No. 5,074,669
granted to Opsal, which discloses using the combination of
modulated optical reflectance plus the non-modulated reflectance of
the two lasers to evaluate the implant dosage level in the
semiconductor sample or to measure the thickness of a layer created
by implantation.
[0016] According to an article entitled "Modules Are In, But
Supertools Endure" by Alexander E. Braun in Semiconductor
International November 1999 available on the Internet at
www.semiconductor.net/semiconductor/iss-
ues/issues/1999/nov99/docs/imt.asp describes use of an ellipsometer
in combination with a spectroscopic reflectometer. Specifically,
the article states "The recently introduced Rudolph S200 system,
for example, uses a proprietary multi-angle laser ellipsometer for
thin films. The ellipsometer also is very sensitive to etch-to-zero
applications, and yet tolerant of refractive index changes in
underlying materials. The ellipsometric measurements also can be
combined with a spectroscopic reflectometer and provide a better
capability to measure overetch and characterize polish rate across
the entire process window at more than 100 wafers per hour (five
sites per wafer) . . . . Rudolph has eliminated the primary cause
of long-term difficulty in obtaining ellipsometer repeatability. If
a microspot lens ellipsometer is used to measure on-product or a
small spot, and high repeatability is attempted, stress
birefringence in the lens--slight temperature changes over
time--causes thickness measurements to vary by a few tenths of an
.ANG.ngstrom. The new capability circumvents this, providing 0.01
.ANG. repeatability."
SUMMARY
[0017] In accordance with the invention, two or more measurements
are made on the same workpiece, during fabrication of the
workpiece, and one of the measurements is used to calibrate another
of the measurements. In one embodiment, each measurement is made
employing a different process, and the measurements are used
together to determine a property (also called "property of
interest") of the workpiece. The multiple measurements may be made
at two or more locations on the workpiece that are separated from
one another, or alternatively even at the same location as long as
different measurement processes are used. If the same is used, the
measurements are made at different locations.
[0018] In one embodiment, multiple measurements from a first
process are used with a predetermined value of the property of
interest in a simulator to generate a simulated value of a signal
to be measured in a second process. One or more such simulated
values and a measured value are used together, to identify a value
of the property of interest.
[0019] If the workpiece's property value is found to not match the
specification, a process control parameter used in fabrication of
the in the workpiece is adjusted, thereby to implement process
control. If the workpiece's property value does match the
specification, fabrication of the workpiece is continued (i.e.
workpiece is not rejected) and the process control parameter is
left unchanged. In some fabrication processes, only one parameter
varies with the process (such as the thickness of the top layer of
a film). In such fabrication processes, in one embodiment, a
calibration measurement is made as described above to determine the
properties of a structure, and then another measurement (such as a
single wavelength measurement) is used to determine the thickness
change of only the top layer.
BRIEF DESCRIPTION OF THE DRAWINGS
[0020] FIG. 1A illustrates, in a flow chart, use of multiple
measurement devices to determine a property, and to control
fabrication of a workpiece, in accordance with the invention.
[0021] FIG. 1B illustrates, in a block diagram, an apparatus that
implements the method of FIG. 1A for a wafer of semiconductor
material as the workpiece.
[0022] FIGS. 2A and 2B illustrate, in flow charts, two alternative
embodiments of the method of FIG. 1A.
[0023] FIG. 3A illustrates locations on a patterned wafer at which
measurements of the type illustrated in FIGS. 2A and 2B are
made.
[0024] FIG. 3B illustrates a cross-section of a two level damascene
structure in the patterned wafer of FIG. 3A, showing measurement
locations.
[0025] FIG. 3C illustrates one embodiment of the apparatus of FIG.
1B used with the patterned wafer of FIGS. 3B and 3C.
[0026] FIG. 3D illustrates, in a flow chart, process control using
the apparatus of FIG. 3C.
[0027] FIG. 4 illustrates fitting measurements from a single
measurement device of the apparatus of FIG. 3C to a curve, to
obtain thickness and refractive index values of two layers (namely
an antireflective coating on an oxide coating) in one example.
[0028] FIG. 5 illustrates a model of thickness of an uppermost
layer as a function of reflectance, and use of the model to
determine actual thickness using a single-wavelength (e.g. 830 nm)
measurement of reflectance.
[0029] FIG. 6 illustrates a graph of dielectric thickness
(including the antireflective coating and the oxide coating) as a
function of distance from a line array measured in one example.
[0030] FIG. 7 illustrates, in a detailed block diagram, a
spectroscopic reflectometer of the prior art, which is used as item
310 in the apparatus of FIG. 3C.
[0031] FIG. 8 illustrates, in a detailed block diagram, a laser
reflectometer of the prior art which is used as item 300 in the
apparatus of FIG. 3C.
DETAILED DESCRIPTION
[0032] In accordance with the invention, a portion of a workpiece
is fabricated (see act 10 in FIG. 1A), at least two measurements
are made on the same workpiece (see acts 20 and 30 in FIG. 1A). The
measurements are combined (e.g. one measurement used to calibrate
another measurement) to determine a value of a property of the
workpiece. If the property value falls outside of specifications
for the workpiece, a process control parameter used in fabrication
is changed in real time.
[0033] Any kind of workpiece may be fabricated in the manner
described herein, including, for example, a wafer (also called
"semiconductor substrate") of semiconductor material that is
fabricated in a wafer fabrication system 100 (FIG. 1B) to form a
"patterned wafer", and that is eventually cut into integrated
circuit dies. Other workpieces that may be fabricated as described
herein include silicon-on-insulator wafers, printed circuits,
multi-layer ceramics, hybrid microcircuits.
[0034] Any fabrication process and/or device well known in the art
may be used to prepare a patterned wafer prior to measurement,
including, e.g. layer formation by chemical vapor deposition (CVD)
and/or etching of metal, rapid thermal annealing (RTA), and/or
chemical mechanical polishing (CMP). For example, FIG. 1B
illustrates use of a layer formation apparatus 101F, dielectric
etching apparatus 101E, and metal deposition apparatus 101I that
are used with a chemical mechanical polisher 102 to form the
patterned wafer. Although not specifically illustrated in FIG. 1B,
any conventional devices used in fabrication of a wafer, e.g. for
photoresist application/exposure/resist strip may be used instead
of or in addition to device 101 in a system 100 as described
herein.
[0035] System 100 also includes a metrology tool 103 that measures
various material properties in a patterned wafer 106. Although
wafer 106 is illustrated as having been polished, metrology tool
103 may be used even with unpolished wafers, as illustrated by path
109 in FIG. 1B).
[0036] Depending on the embodiment, the multiple measurements may
all be made employing a common device and/or process, or each such
measurement may be made by a different device/process. If a common
measurement device or a common measurement process is used, such
measurements are made with different resolutions and/or precision,
and/or in different locations of the same workpiece. Resolution
relates to the size of features that can be resolved, while
precision determines how accurate a measurement is (the standard
deviation of the measurement result). For example, it may be
necessary to resolve two spatial features separated by 5 .mu.m.
Each feature may have to be measured to a precision of .+-.10 nm in
depth.
[0037] Alternatively, different measurement devices and different
measurement processes may be used for each of the measurements, and
if so the measurements may have the same resolution or different
resolutions. For example, an area over which the measurements are
made may be of different sizes in the measurements. Moreover, if
the measurement devices and processes are different, some (but not
all) of the measurements may be made in a destructive manner, if
made in a test area (also called "test pattern") of the patterned
wafer.
[0038] Therefore, a low resolution (and possibly destructive)
method may be used to make a multi-variable (several film)
measurement in a test area, and the results of such measurement
applied to an area of interest in which only one property changes,
and the change in property is measured in the area of interest
using a high resolution method. Even though the high resolution
method is capable of measuring only one (or a small subset) of the
properties measured by the low resolution method, the high
resolution method is used in combination with the low resolution
method as described herein. In one embodiment of the invention the
reference measurement is done through a separate low powered
measurement device (e.g. a spectrometer) that is not part of the
optical train for the final measurement. In this embodiment, a low
power objective may be used for coarse alignment (for example, to
find a die or large feature), but not for final alignment (for
example, to find a measurement site within a die).
[0039] As noted above, instead of low and high resolution methods,
destructive and non-destructive methods may be used in the test
area and area of interest respectively. Also, as would be apparent
to the skilled artisan in view of this disclosure, any two methods
and/or devices that are not suitable for use in one of the two
kinds of areas (e.g. test area and area of interest) may be used in
combination, when the two areas are sufficiently near one another
for the properties that affect the measurements to remain
substantially the same (e.g. change by no more than 1%). Examples
of such combinations of methods include:
1 Reference Site Active area measurement Thickness of stack Test
pattern Thickness of top layer Dielectric tkns Test pattern Doping,
junction depth SIMS doping Test pattern Junction depth profile
uniformity Linewidth Test array Width of single line
[0040] Furthermore, instead of combining measurements from just two
methods and/or devices, measurements from any number of methods
and/or devices may be combined as described herein. For example,
doping level of an ion implant may be characterized with a rapid
non-destructive, high-resolution method, as follows. The
measurement region has a silicon dioxide coating (known as a screen
oxide). A test area is measured with a reflectometer to determine
the screen oxide thickness. A SIMS measurement (slow method allows
measurement at only one site--say at the wafer center) is also made
in a test area to determine a reference doping concentration. The
screen oxide thickness and reference concentration are used to
calibrate the third measurement that has high speed that is then
applied at a large number e.g. 49 sites on the wafer to determine
uniformity of the doping.
[0041] In the embodiment illustrated in FIG. 1B, metrology tool 103
includes an aligner 103D that positions a wafer to be measured in
one of measurement devices 103A and 103B, and a programmed computer
103C determines a value of a property of the wafer based on
multiple measurements by one or more of devices 103A and 103B. Note
that a single aligner 103D is used in one embodiment to position
the wafer in all measurement devices of tool 103, so that
properties at locations that are coincident with one another or
adjacent to one another within a predetermined distance are
measured.
[0042] However, a reference measurement may be made on a separate
thin-film measuring system and then the high resolution measurement
on a second system that is separate and distinct. Therefore, in an
alternative embodiment, each measurement device has its own
aligner.
[0043] Measurements as described herein may be made by any of a
number of measurement processes and/or devices well known in the
prior art. For example, one measurement may be made by a
spectroscopic reflectometer (well known in the prior art) to
determine a number of properties (such as thicknesses, refractive
indices and absorption constants of one or more films of, e.g.
oxides, nitrides, poly and a-silicon and polyimide), and another
measurement may be made by a laser reflectometer (also well known
in the prior art), to more precisely measure a property of
interest.
[0044] In the just-described example, the spectroscopic
reflectometer has a larger spot size than the laser reflectometer.
Specifically, a single wavelength laser (of the type normally used
in a laser reflectometer) is collimated and coherent and can be
focused to a much smaller spot (on the order of 1 .mu.m) as
compared to a beam of white light (of the type normally produced by
an incandescent bulb which is commonly used in a spectroscopic
reflectometer).
[0045] Instead of a spectroscopic reflectometer, other tools (such
as PQ Ruby and PQ Emerald available from Philips Analytical Inc. 12
Michigan Drive, Natick, Mass. 01760) may be used. For example, an
acoustic tool, a 4-point probe, a scanning electron microscope, a
stylus profilometer or an X-Ray machine may be used to perform a
calibration measurement. Calibration measurements may be made using
the Opti-Probe film thickness measurement tool, available from
Therma-Wave, followed by a second measurement using Therma-Probe
ion implant dose measurement system. Alternatively, measurement of
film thickness may be made by use of KLA UV-1050, or Rudolph
SpectraLaser.
[0046] Commercially available devices that can be used for a first
measurement include:
2 1. KLA-Tencor ASET-F5x, UV-1280SE, UV-1080 2. Rudolph S 200, S
300, SpectraLASER 200 and 300 3. Nanometrics 8300X, 9300 4.
Therma-Wave Opti-Probe 3290, 3290DUV, 5240 5. Cameca IMS 6f (SIMS)
6. Physical Electronics Adept 1010 (SIMS)
[0047] Any of these devices 1-6 can be used for a calibration
measurement. In addition, commercially available devices that can
be used for a second measurement include:
3 7. Therma-Wave Therma-Probe TP-500, TP-630 8. Boxer Cross BX-10
9. Boxer Cross BX-30
[0048] Measurements from the above-described devices can be
combined as follows: (A) measurements from any of 1-4 can be
combined with 9 (dielectric film thickness measurement can be
combined with high resolution dielectric film thickness of top
layer); (B) measurements from any of 1-4 combined with 7 or 8
(overlaying dielectric film measurement with measurement of ion
implant dose); and (C) measurements from any of 5-6 combined with 8
(doping profile measurement combined with junction depth
uniformity).
[0049] In several different embodiments, the following combination
of methods are used:
4 First measurement Second measurement Reflectometry, ellipsometry
Laser reflectance SIMS Dose, junction depth measurement (2 laser
reflectance) Four point probe Dose measurement (2 laser
reflectance) Acoustic metal thickness Thickness of whole stack as
per U.S. Pat. No. 6,054,868 Scatterometry CD Linewidth dependence
as described in U.S. Pat. No. 6,054,868.
[0050] The acoustic metal thickness measurement is used in e.g. the
Rudolph MetaPULSE to measure the thickness of each layer in a metal
stack, with a spot size of 7-10 .mu.m. This could be coupled with
the method of U.S. Pat. No. 6,054,868 to measure changes in the
thickness of the stack as a whole (without resolving individual
layers) but in fine patterns or with higher spatial resolution.
[0051] Scatterometry measures critical dimensions (CD) but needs
about a 50 .mu.m spot to do so. The method described in U.S. Pat.
No. 6,054,868 is sensitive to line width, where the spot may be
only 2 .mu.m wide.
[0052] Regardless of the process and/or device used in measurement,
multiple measurements of the type described herein are made on the
same workpiece in accordance with the invention, and are used
together (see act 40 in FIG. 1A) to determine a value of a property
(also called "property of interest") of the workpiece. Various
kinds of measurements from the same workpiece may be combined in
any manner to obtain the property value.
[0053] If a workpiece's property value, which has been determined
by combination of the measurements as described above, is found to
not match the specification (see act 60 in FIG. 1A), a process
control parameter used in the workpiece's fabrication is adjusted
(see act 61 in FIG. 1A), thereby to implement process control. Such
a workpiece may be discarded (because property value doesn't match
specification). If the workpiece's property value matches
specification (e.g. falls within a specified tolerance around a
specified value), the workpiece is processed further, i.e. another
portion of the workpiece is fabricated (e.g. by returning to act 10
in FIG. 1A).
[0054] Depending on the embodiment, the above-described acts 20-61
for measuring the property of interest may be repeated (as
illustrated by act 70 in FIG. 1A) at a number of locations, e.g. to
determine a profile of the property of interest radially across the
workpiece or around the circumference if the workpiece has a
circular shape. In such embodiments, the profile is used in
determining whether or not a workpiece under fabrication matches
the specification.
[0055] In a first example all measurements are of the same
property, and a first measurement on the in-fabrication-workpiece
is used to calibrate a second measurement on the same workpiece. In
such an example, identical measurement methods may be used, at
different sites of the workpiece. In a second example, the
following measurements are made on the same workpiece: a first
measurement is of a different property from a second measurement,
and the first measurement is used in a simulator (e.g. a personal
computer programmed with simulation software) to generate a
simulated value for the second measurement, and the simulated value
is then used to calibrate the second measurement.
[0056] In a third example, the following measurements are made on
the same workpiece: a first measurement is used to generate a set
of simulated values for a second measurement, based on a set of
predetermined values of the property that the workpiece is likely
to have, and the second measurement is used to identify the closest
simulated value which is then used to determine the property. For
convenience, in the following description a first measurement of
any of the just-described three examples is referred to as
"calibration measurement" (even though calibration is not performed
in the third example), and the second measurement is referred to as
"actual measurement".
[0057] The multiple measurements on a workpiece being fabricated as
described above can be made at the same location, or at different
locations. When the measurements are made at different locations,
the locations are selected to ensure that one or more properties
that may affect any of the measurements, other than a property of
interest, are substantially identical (e.g. differ by no more than
1) between the locations. Under such conditions, the multiple
measurements when combined as described herein identify any local
variations (i.e. variations between the locations) that are caused
by the process being used to fabricate the portion of the
workpiece.
[0058] As noted above, in one embodiment, a property of a
semiconductor substrate is measured by system 100 (FIG. 1B) as
described herein. Specifically, system 100 uses a first measurement
device 103A to measure (see act 110 in FIG. 2A) a number of
properties of the semiconductor substrate. Next, system 100 uses
(see act 111 in FIG. 2A) a simulator 103C to generate a simulated
value of a to-be-measured signal, based on a predetermined value of
the property of interest (e.g. a value identified in the
specification), and also based on values of properties (other than
the property of interest) that can affect the to-be-measured
signal. The values of properties (other than the property of
interest) are measured by the first measurement device 103A, for
use by the simulator. In this embodiment, the simulator is
repeatedly operated (see a loop formed by acts 112, 113 and 110),
so that a number of simulated values are generated for a
corresponding number of predetermined values (which may be selected
to cover a range of values for the property of interest permitted
by the specification).
[0059] Next, system 100 uses (see act 141 in FIG. 2A) a second
measurement device 103B (FIG. 1B) to obtain a measured value of a
signal indicative of the property of interest. Next, in one
implementation, programmed computer 103C compares the measured
value (see act 142 in FIG. 2A) with one or more simulated values
(which may be held in, for example, a table) to identify the
closest simulated value. Thereafter, programmed computer 103C
determines (see act 143 in FIG. 2A) the value of the property of
interest, based on a predetermined value for the property of
interest that generated the closest simulated value.
[0060] For example, if the measured value is same as one of the
simulated values, then computer 103C determines the property value
to be the corresponding predetermined value. If the measured value
differs from the simulated value by a certain percentage then
computer 103C determines the property value of the semiconductor
substrate to have the same percentage difference relative to the
corresponding predetermined property value i.e. performs an
interpolation.
[0061] Thereafter, computer 103C checks if the property value
matches the specifications (see act 160 in FIG. 2A), and if so, the
semiconductor substrate is processed further. If the property value
does not match the specifications, computer 103C drives a control
signal to, for example, layer formation apparatus 101F and/or to
chemical mechanical polisher 102, for process control. Note that
even when a property value matches the specifications, if the
property value falls within a predetermined range, process control
may be performed (although the semiconductor substrate is not
discarded) e.g. to correct an upcoming problem.
[0062] The just-described interpolation may be linear or nonlinear,
depending on the embodiment (e.g. depending on the dependence of
the property of interest on the signal being measured). Instead of
using simulated values directly, computer 103C may determine a
curve to which the simulated values fit, and then use the curve to
look up the property value.
[0063] An alternative embodiment includes use of a look-up table
based on externally generated values by another computer (when
computation time is long) or based on empirical values or fits to
empirical values. In another embodiment illustrated in FIG. 2B, the
above-described simulation is not automatically repeated (in the
loop formed by acts 112, 113 and 110), and instead, a measured
signal from the second process is compared (see act 173 in FIG. 2B)
with a simulated value (generated by the simulator), which is based
on a predetermined value for the property as defined by the
specification. If there is no match, the simulator is operated
again, with another predetermined value of the property of interest
(see act 173 in FIG. 2B), until a match is found.
[0064] In one implementation, first and second measurements are
performed at different locations: the first measurement in a test
area, and the second measurement in an area of interest (such as a
region containing a number of metal lines) of a wafer 200 (FIG. 3A)
that is under fabrication. As noted above, such a workpiece may be
a wafer having a number of areas (also called "die areas") that
eventually form dice, such as area 201 shown in a circle 205 that
is an enlarged view of a corresponding circle on wafer 200 (FIG.
3A). As illustrated in FIG. 3A, area 201 is surrounded by streets
201a-201d. Each of streets 201a-201d is, for example, 100 .mu.m
wide, and forms the area in which a saw is to be run, to separate
the wafer into individual dies after fabrication is complete.
Various areas in streets 201a-201d may be used for test patterns
because such areas are better controlled than die areas in which
integrated circuits are formed.
[0065] In this implementation, test area 203 (present inside circle
206 which is an enlarged view of a portion of wafer 200 illustrated
in circle 205 in FIG. 3A) is chosen to be a box that is devoid of
patterning, and is located in street 201a. Test area 203 is chosen
to be within the nearest street to the area of interest (e.g. line
array 202). Note that any other test area may be selected, e.g. if
properties of the test area are well controlled, and if the area is
arbitrarily near to (and preferably but not necessarily separated
from) the area of interest. The properties that must be well
controlled and that are measured in test area 203 are all of the
properties that affect a measurement in the area of interest. In
one embodiment it is assumed that these properties remain the same
for both areas (i.e. the test area and the area of interest).
[0066] A first measurement may be made in a region 207 (e.g. of
diameter 50 .mu.m (in the example, the street is 100 .mu.m, so the
test area must be smaller)) in test pattern 203 (see FIG. 3A),
which is located in a field region 201a as illustrated in FIG. 3B.
If a destructive process (such as Secondary Ion Mass Spectrometry,
or SIMS) is used to perform the first measurement, a pit is formed
therein at the end of measurement. Moreover, one or more second
measurements may be made at locations 204a-204e (FIG. 3A) that are
between test pattern 203 and conductive lines 202. Lines 202 are
embedded within a damascene structure (see FIG. 3B) that is formed
in wafer 200, e.g. by chemical mechanical polishing.
[0067] Specifically, in one example, wafer 200 has a new level of
metal interconnect formed (e.g. by apparatus 101 and polisher 102)
over a preexisting structure 210 which is an underlying level of
metal interconnect. Structure 210 includes a dielectric matrix 213,
inlaid metal lines 211, and a pad (a large area feature) 212.
Structure 220 has a dielectric stack and a set of inlaid lines. The
dielectric stack may include etch stop layers 221 and 223,
dielectric layers 222 and 224, and anti-reflection coating layer
225. The inlaid lines 202a-202K are formed in dielectric layer 224
and one or more via interconnects, such as structure 227, are used
to connect the conductors in structure 220 to the conductors in
structure 210. The function of etch stop layers 221 and 223 is to
provide a material to stop the groove etching used to form the
grooves in which lines 202a-202k are formed, and in which via
interconnect 227 is formed. The function of the anti-reflection
coating layer 225 is to control the optical properties of the stack
for photolithography exposure.
[0068] Structures 210 and 220 of wafer 200 (FIG. 3B) may be
fabricated in any manner well known in the art. In one example,
structure 220 is formed in the following manner. First, on
structure 210, a dielectric stack of layers 221-225 (FIG. 3B) is
formed. Thereafter, grooves are etched through layers 225 and 224,
stopping at layer 223. Next, layer 223 is removed at the point
where via interconnect 227 is to be formed. Thereafter, a hole is
etched for via interconnect 227 through layer 222, stopping at
layer 221. Next, layer 221 is removed at the bottom of the hole,
exposing the metal line in structure 210. Thereafter, the grooves
are filled with metal such as copper. Such filling leaves copper
over the top of the entire structure. The resulting structure is
polished by polisher 102 (FIG. 1B), removing the blanket copper
film coating the top of structure 220 and leaving lines 202a-202k
remaining, thereby to form structure 220.
[0069] In addition to removing the excess copper, polisher 102 may
also remove some dielectric material over lines 202a-202k, so that
layer 225 in structure 200 becomes increasingly thin when going
from street 201a towards a first conductive line 202a (FIGS. 3B and
3A), i.e. layer 225 has the shape of a wedge when viewed in
cross-section. The slope (which is thickness T of the sloping
surface divided by width W of layer 225) can be small as compared
to other slopes e.g. width W may be 10 .mu.m (100,000 .ANG.) and
thickness T may be 200 .ANG., so slope is {fraction (1/500)}, or
about 0.12 degrees. However, the slope can be much larger if
polisher 102 is not well controlled. The sloped region 228 (also
called "erosion edge") may occur within 10 .mu.m of metal lines
202a-202k. Such an erosion edge 228 may also extend into line array
202.
[0070] The slope of an erosion edge of wafer 200 can be measured
with high depth resolution (e.g. 10 .ANG.) and spatial resolution
(e.g. 1 .mu.m) based on Multiple measurements of the type described
herein. Specifically, wafer 200 is loaded by an aligner 330 onto a
stage (not shown) and moved under system 300. In this example,
wafer 200 has a two layer coating consisting of an anti-reflection
coating (ARC) 225 (e.g. of thickness 603 A) over silicon dioxide
layer 224 (e.g. of thickness 3479 A). In one example, there are
actually 5 dielectric layers over the reflecting surface of pad
212: ARC layer 225, silicon dioxide layers 222 and 224, and
etch-stop layers 221 and 223. The index of refraction and thickness
of these layers are measured over pad 212 in the reference
measurement. Wafer 200 is placed in system 300 and aligned by
aligner 330 so that a measurement by a first measurement device 310
is done in test pattern 203 (FIG. 3A).
[0071] Specifically, system 300 has two measurement devices 310 and
320, with a site to be measured in wafer 200 located at position
301a under the first measurement device 310 and after the
measurement is completed the same site in the same wafer 200 is
located (by wafer aligner 330) at position 301b under the second
measurement device 320.
[0072] First measurement device 310 includes a broadband
spectroscopic reflectometer well known in the art. Such systems
consist of a white light source that creates optical beam 311,
which is focused onto wafer 200 at position 301a with lens 312. The
numerical aperture of lens 312 may be, for example, 0.2 and the
spot size may be on the order of, e.g. 20 .mu.m. Such a
spectroscopic reflectometer is available commercially from, for
example, Ocean Optics of Dunedin, Fla.
[0073] Optical beam 311 is of white light (also called
"polychromatic white light"), and includes light from multiple
parts of the color spectrum (e.g. the presence of multiple colors),
such as the light produced by an incandescent bulb or a halogen
lamp. Such white light typically covers a spectral range of 300 to
800 nm. Measurements of reflection are taken at 40-80 equally
spaced wavelengths over the full spectral range using the
spectroscopic reflectometer 310.
[0074] White light is used in first measurement device 310 so that
multiple parameters (e.g. refractive index and thickness of each of
four layers in a stack) may be measured by sensing reflectivity of
light at a large number of different wavelengths (e.g. at
wavelength starting at 300 nm and incrementing by 10 nm until 800
nm), in the normal manner of a spectrometer. As noted above, such
measurements are used by a computer 340 to generate a function to
be used with a measurement from the second measurement device 320
to determine a value of the property of interest.
[0075] In this particular embodiment, computer 340 is programmed
with analysis software to take the reflection vs. wavelength signal
from measurement device 310 and convert the signal into a table of
thickness and index of refraction of each layer. Such software is
available commercially, for example, as WVASE32 Analysis software
sold by J. A. Woollam Company, Inc. of Lincoln, Nebr., and is
described in the user's manual entitled "Guide to Using WVASE32,"
1995, available from J. A. Woollam Company, Inc. Instead of WVASE32
Analysis software any other ellipsometric analysis program may be
used.
[0076] When programmed with such software, computer 340 employs
curve fitting methods to fit measurements 401 to a curve 402 (see
FIG. 4). In one example, computer 340 is informed of the number of
layers that are present in wafer 200 (based on the fabrication
process), and is also informed of nominal values for thicknesses of
the layers, the materials used to form the individual layers and
the position of the layers relative to one another. For example,
for wafer 200 illustrated in FIG. 3B computer 340 is provided with
information in the following table.
[0077] In this example, computer 340 uses such information as an
initial model for wafer 200, and uses look-up tables supplied with
the software, for the index of refraction of the ARC and silicon
dioxide layers, to determine any changes to be made to the initial
model. Specifically, computer 340 varies the index of refraction
and thickness of each of the layers until a model is found that
generates a good fit for the experimentally measured reflectance
across the electromagnetic spectrum. In FIG. 4, curve 402 is the
best fit for measurements 401. After such calibration, computer 340
determines that in spot 207 the thickness of the ARC layer 225 is
603 .ANG. and the thickness of the silicon dioxide layer 224 is
3479 .ANG..
[0078] In addition to generating a model of wafer 200 based on
measurements from first measurement device 310, computer 340 in
accordance with the invention also uses the model to generate
reflectance and/or transmission data for the uppermost layer 225
(FIG. 3B). Specifically, a reflectance generation function of
computer 340 is used to determine reflectance of wafer 200 as a
function of thickness of the uppermost layer 225, at the frequency
(e.g. 830 nm) of a laser beam 321 used in second measurement device
320 (discussed below). This function is illustrated in FIG. 5, by
curve 501.
[0079] Note that a laser beam 321 is used in a second measurement
device 320, and therefore the laser beam itself is not used in the
first measurement device 310. Moreover, although a sensor at the
wavelength (e.g. 830 nm) of laser beam 321 may have been used
during the first measurement, this is not necessary. In the example
illustrated in FIGS. 4 and 5, there may not be sampling of
reflectivity at the specific laser wavelength (e.g. if the laser
wavelength is 827 nm), and instead the function represented by
curve 501 is determined from the model which is constructed from
measurements by device 310.
[0080] Depending on the embodiment, at the time computer 340
generates the reflectance and/or transmission data for layer 225,
laser beam 321 may or may not have been applied to wafer 200. So,
variation of a material property (in this example thickness) as a
function of a to-be-measured signal (in this example reflectance)
is determined from the measurements by device 310. When plotted on
a graph, this function is also called "calibration curve", and is
illustrated in FIG. 5 by curve 501.
[0081] Second measurement device 320 of this embodiment includes a
laser reflectometer that produces a collimated laser beam 321. Lens
322 focuses beam 321 onto a site (e.g. of diameter 1 .mu.m) on
wafer 200 when located at position 301b. The reflected laser light
is sent to detector 324 with beam splitter 323. Detector 324
provides a measurement, which is used as described above, with a
function obtained from simulation (represented by curve 501 in FIG.
5), to look up a value of the property of interest.
[0082] Depending on the embodiment, more than one laser beam may be
used by device 320 to illuminate the same site, e.g. two
monochromatic songs may be modulated at two corresponding
frequencies and light reflected by the site filtered at these
frequencies obtain reflectances of the site at the respective
wavelength of the two sources.
[0083] In one embodiment, the slope of erosion edge 228 (FIG. 3B)
is measured by scanning wafer 200 under the spot of a beam 321 in
small increments, say 1 .mu.m steps towards array 202 (FIG. 3A),
and a measurement is taken after each step. For example, five
measurements are made at the corresponding locations 204a-204e,
spaced 2.5 .mu.m apart, providing reflectance values 402a-402e. The
first value 402a yields a thickness of 603 .ANG. for layer 225,
which is consistent with the measurement in test pattern 203. Note
that the first location 204a may be at a point prior to the erosion
edge, so that the thickness of ARC layer 225 is the same as for the
reference measurement. The value of the ARC thickness at this point
may be used to verify that the first and second measurements are
properly calibrated.
[0084] Progressive measurements 204b-204e map the erosion edge 228.
In the above-described example, the thickness of the silicon
dioxide layer 224 is added to the measured thickness of the
antireflective coating 225, to obtain total thickness of the stack.
The data points are then plotted as points 502a-502e (FIG. 6)
corresponding to measurement points 402a-402e, thereby to yield a
line 502. Line 502 provides a graph of the thickness as a function
of distance from the left edge of array 202.
[0085] Use of first measurement device 310 during fabrication of a
wafer as described herein, calibrates a laser reflectometer for
each wafer being evaluated. Combination of the measurements from
devices 310 and 320 as described above assumes that initially all
thickness variation is due solely to a diminishing thickness of the
topmost layer, for example, anti-reflection coating 225. When the
measured thickness change becomes greater than the thickness of
anti-reflection coating 225, computer 340 assumes that the
thickness change comes solely from a diminishing thickness of the
uppermost layer which is the next layer in the stack, for example,
dielectric layer 224. In this manner, computer 340 of this
embodiment always determines only one parameter--the thickness of
the uppermost layer--using one measurement--reflectance at a single
laser wavelength.
[0086] In some cases, a laser at a second wavelength (which is
different from the first wavelength) may be used to make an
additional measurement in device 320, and computer 340 uses the
additional measurement instead of the measurement at the first
wavelength, because the reflection signal is periodic in thickness,
so that certain thickness values may be at a maximum or minimum
where the derivative with respect to thickness is zero (thereby
leading to ambiguity). In such cases the measurement at the first
wavelength provides less resolution than the additional measurement
at the second wavelength.
[0087] FIG. 3D illustrates a decision flow chart associated with
measurement of a property of interest in one particular
implementation of the invention. In step 351 a wafer 200 is loaded
into the measurement system 300. In step 352 wafer 200 is moved to
a position so that in step 353 the spectroscopic reflectometer can
measure the thickness of each layer in the stack at a location near
the site of the high-resolution measurement and create a model for
reflectance at the wavelength (e.g. 830 nm) of a laser in the laser
reflectometer. In step 354 the reference site is moved under the
high-resolution laser spot and the reflectance is measured. This
provides a reference model to be used in calibrating the
reflectance signal measured at the laser wavelength.
[0088] A loop of steps 355-358 is now entered where measurements
are performed at a series of sites, for instance as a line scan
with points spaced by a fixed distance to scan over an erosion
edge. In step 355 the system moves to a measurement site, a focus
is performed, and the reflectance is measured. The reflectance is
then converted to thickness of the top layer and, by adding the
thickness of the underlying layers, the thickness of the stack is
found in step 356.
[0089] If thickness measurement based on the current model for
stack 220 indicates the entire topmost layer is removed, a second
model is created (e.g. by omitting layer 225 from the current
model), and used to measure the thickness of the next layer 224
(which is now the uppermost layer) assuming the absence of layer
225. In step 357 (FIG. 3D) the data is optionally further analyzed
(although, in the simplest form, the thickness is reported and the
measurement process is done).
[0090] If the reflectance vs. thickness curve is near an inflection
point in a model the thickness resolution is poor, (i.e. the
resolution is less the process tolerance; for example, if the
process can tolerate .+-.50 .ANG. and the resolution is 50 .ANG.,
then the resolution is poor; a stricter definition is based on 3
standard deviations, so the resolution must be 3 times the
tolerance), and in such an event a second measurement may be made
with a second laser at a second wavelength to obtain a higher
resolution thickness measurement. Finally, in step 358 (FIG. 3D) a
decision is made as to whether all sites have been measured and the
scan is complete, or whether the next site should be measured.
[0091] After the measurements are complete, the thickness profile
is analyzed in step 359 to determine, for example, if the erosion
step is too deep. Examples of erosion values could be from 100 to
1000 .ANG.. Too deep is a matter of process tolerance, but values
on the order of 1000 .ANG. may be considered unacceptable. If the
profile is ok (either it matches a profile that is obtained on
devices that perform properly, or the erosion depth is within
tolerance, as determined by the process) the loop is complete and a
new wafer is measured, or a new site on the same wafer is measured.
If the profile is not ok, in step 360 the fabrication process is
corrected before measuring the next wafer.
[0092] Sometimes it is desired to make a single measurement rather
than a scan over a region. The advantage of a scan is that the
dielectric stack structure at the starting point in the second
measurement device 320 is equal to the stack as measured by the
first measurement device 310. The top layer 225 then diminishes in
thickness from that point. As noted above, at some point, the top
layer 225 may be fully removed, and a second model of the stack
that does not incorporate the original top layer is used.
[0093] If device 320 is to measure only at a specific location of
interest (also called "measurement site") instead of the
above-described scan, it may be ambiguous from such a measurement
as to whether the topmost layer in the region measured by device
310 is or is not present. In this case, an alternate procedure
described next is followed using two lasers at the specific
location of interest, and the lasers have two different wavelengths
.lambda.1 and .lambda.2.
[0094] 1. A number of reflectance measurements at a corresponding
number of wavelengths are made by device 310, to determine the
optical constants and thicknesses of the layers in the stack in
test pattern 203 (also called "reference site") in the manner
described above.
[0095] 2. Computer 340 is programmed to create a number of models
that predict thickness as a function of reflectance measured by
device 320, for four cases: namely for each of two wavelengths
.lambda.1 and .lambda.2 and with and without the topmost layer in
test pattern 203.
[0096] 3. Two reflectance measurements are made by device 320 at
the reference site to calibrate the reflectance of lasers with
wavelength .lambda.1 and .lambda.2.
[0097] 4. Two additional reflectance measurements are made at the
measurement site at wavelengths .lambda.1 and .lambda.2.
[0098] 5. Computer 340 is programmed to determine thickness for the
four cases (two wavelengths, with and without the topmost
layer)
[0099] 6. Computer 340 is further programmed to compare thickness
values for the two wavelengths .lambda.1 and .lambda.2, for each of
two cases (with and without the topmost layer). Computer 340 is
also programmed to select the case where the thickness values most
closely match, as the correct case.
[0100] In one example, a first model including the topmost layer
gives a thickness of 985 nm for .lambda.1 and 755 nm for .lambda.2,
and a second model without the topmost layer gives a thickness of
694 nm for .lambda.1 and 696 nm for .lambda.2. The difference for
the case of a topmost layer being present is 230 nm and the
difference for the case of the topmost layer being removed is 2 nm.
On comparison of the two differences, the 2 nm difference is
smaller than the 230 nm difference, and therefore the case of the
topmost layer being removed is selected as the correct case, and
the second model is used to determine the property of interest. In
this example, computer 340 determines that the thickness is an
average of the two readings, i.e. 695 nm.
[0101] FIG. 7 shows a schematic of the reflectometer apparatus,
which provides the low resolution characterization of the
multiple-film stack. Note that the apparatus of FIG. 7 is not a new
invention by itself, and is constructed with commercially available
components. White light source 701 emits light 702 that is
collimated with collection optics 703 to form a beam of white light
that is nearly focused on the wafer 705 with objective lens 704 (a
near focus point is used to prevent imaging the lamp filaments,
thereby providing uniform illumination in the spot).
[0102] Specifically, lens 704 collects light reflected from wafer
705. The reflected light is diverted using beam splitter 706. The
reflected light passes through beam splitter 707 and is focused
into fiber 709 with lens 708. Fiber 709 couples the reflected light
to spectrometer 710 (Ocean Optics), which provides an output signal
giving the reflection as a function of wavelength. The reflection
vs. wavelength data is sent to computer 711.
[0103] In one implementation, computer 711 is programmed with
WVASE32 Analysis software (available from J. A. Woollam Company,
Inc. Lincoln, Nebr.) that includes dispersion models (models of
index of refraction vs. wavelength) for the materials in the stack
and determines the thickness of each layer by optimizing the fit
between a model and the measured data. The WVASE32 Analysis
software also calculates a table of thickness of the top-most layer
as a function of reflectivity at the measurement laser wavelength.
A similar table may also be calculated in the absence of the top
layer for cases where the top layer has been completely removed and
the erosion step 28 occurs in the layer underneath the top
layer.
[0104] Beam splitter 707 diverts a portion of the reflected beam to
a low-power vision system used to find the measurement site on the
wafer 705. This consists of a microscope formed by the combination
of objective lens 704 and lens 712. Video camera 713 provides an
image to a pattern recognition system (Cognex Corp.) to perform
site alignment.
[0105] FIG. 8 shows a schematic of the laser reflectometer
hardware, providing a high resolution measurement. Laser 801 (830
nm laser with 100 mW emitted power from Spectra Diode Labs of San
Jose, Calif., for example) emits beam 802 that is collimated using
lens 803 to a beam diameter of 2.3 mm. Beam 802 is then focused on
a site on wafer 805 with objective lens 804, which is a 100.times.,
NA=0.9 lens from Olympus of Tokyo, Japan. The spot size is about 1
.mu.m at the surface of wafer 805.
[0106] Before a measurement is taken, the wafer is moved to the
measurement site and focused. Movement to the measurement site is
accomplished by illuminating the sample with white light for the
purpose of imaging. The white light source 818 is collimated with
lens 817 and is injected using beam splitter 816. The reflected
white light is imaged with a microscope formed with the combination
of lens 804, beam splitter 815, and lens 814. Camera 809 images the
site and the wafer is aligned according to stage motions initiated
by the pattern recognition software that runs in computer 811.
During the site alignment laser 801 may be turned off or shuttered
to prevent the bright laser light from swamping camera 809.
Alternately, the laser power may be turned to a minimum value so
the laser spot may be viewed overlaid on the wafer pattern to
confirm the exact site of measurement.
[0107] The focus at the site is then accomplished using the
auto-focus consisting of lens 812 and auto-focus, element 813
including a pinhole and a detector. Laser 801 is used as the light
source for the auto-focus.
[0108] Following focus at the site, the reflectance is measured.
Laser beam 802 is collimated with lens 803 and focused onto the
wafer 805 with objective lens 804. The return reflection is split
off with beam splitter 806, passes through beam splitter 807 and is
focused onto detector 810 with lens 808. The signal is digitized
and sent to computer 811, which compares the reflection to the
model generated earlier to determine the thickness of the topmost
layer according to the procedure previously described.
[0109] In the event that a second laser wavelength is desired (e.g.
to eliminate ambiguity in depth measurement caused by an inflection
point as discussed elsewhere herein), such a laser may be coupled
into the system in FIG. 8 in a manner analogous to the white light
source 818. In this case, the second laser beam is collimated and
injected collinear with the beam from laser 801 using a beam
splitter similar to element 816.
[0110] It may be necessary to calibrate the laser reflectance. This
is done either by measurement on a reference sample of known
reflectance, or by measurement at the site used with the low
resolution reflectometer measurement. Alternately, the first
measurement site of a scan may be chosen outside of the eroded
area, and the thickness at this site may be assumed to correspond
to the thickness measured with the low resolution reflectometer in
the test pattern.
[0111] In one application of a measurement as described herein, a
wafer requires a next level of metal interconnection. The wafer
goes into process module 101 (FIG. 18), where an interconnect layer
is formed, including dielectric stack deposition, groove etching,
backfill and polishing. The wafer, with the completed additional
interconnect layer, is measured in system 103 that applies the high
resolution dielectric measurement to determine that the polishing
process has been successfully completed. Results of measurements
performed by system 103 are transferred to computer 103C. In the
event that results are judged unacceptable (for example, because
erosion is too deep), signals are sent to process module 101 to
alter or halt the process to enable correction of the problem.
[0112] As noted above, two or more lasers can be used to increase
the accuracy of the measurement. In certain cases, the reflectance
is cyclical with the thickness of the layer. For example, the
reflection of a single layer is a cosine function of the
wavelength. Therefore, there are certain values of thickness where
the change in reflectance with respect to thickness is zero (where
the argument of the cosine is 0 or .pi.). At those thickness
values, a second reading can be made with a second wavelength,
where the change in reflectance with respect to thickness is
non-zero. This increases accuracy across the full range of
thickness values. Alternately, the second laser can be used instead
of the first when the reference measurement indicates that the
thickness lies at an inflection point of a model based on a
measurement using the first laser If one or more lasers are used to
make measurements as described herein, one or both lasers can be
modulated for the reflectance measurement. This has various
advantages. For example, a pumping laser may already be modulated,
so it is simple to read the modulated reflectance of this laser
alone by removing a blocking filter from the front of the detector.
Also, measuring the reflectance of a modulated laser beam enables
use of the lock-in amplifier to measure the reflectance signal.
This provides a very accurate, noise-free measurement, much more
accurate than available with a dc measurement.
[0113] In addition, if one laser is modulated and the other is not,
the reflectance of both beams can be measured simultaneously by
measuring the modulated laser reflectance with a lock-in amplifier
and measuring the unmodulated laser reflectance with a dc
amplifier. This increases throughput over measurement of dc
reflectance of both beams, since the measurement of the dc
reflectance of each beam must be done in sequence rather than in
parallel.
[0114] The laser light can also be polarized to enable use of
methods as described in U.S. patent application Ser. No.
09/521,232. For example, if the light is polarized with the
electric field vector perpendicular to the direction of the metal
lines, then the metal lines will be "invisible" (have a very small
cross-section to the laser light). It is then possible to measure
the dielectric thickness within the metal arrays.
[0115] Low resolution values can be used in case of dielectric
measurements, in two ways. First, if the dielectric step to the
array is measured, as with the above methods, then the dielectric
thickness at the array edge is known. This value can be used to
extend the high resolution measurement into the array. Second, a
model could be built to measure dielectric thickness with low
resolution in the array--this would be exactly analogous to the
measurement in the test pattern, only a new model would be
required, since the WVASE32 software will not handle measurement in
an array.
[0116] Another embodiment for the high resolution measurement uses
two high resolution lasers at the same time. This embodiment has a
real advantage in that the reflectance at both wavelengths can be
measured simultaneously, thereby speeding up the measurement. One
high resolution laser is modulated at one frequency and the other
at a second frequency, or one is modulated and the other is at
constant power (dc). The detector signal is then split using either
two bandpass filters or a high-pass and low-pass filter (the latter
can be a capacitor to block dc and pass ac). The signals from both
lasers are read at the same time.
[0117] For example, the 830 nm laser is modulated at 2 kHz and the
980 nm laser is operated at constant amplitude. Both beams are
reflected from the sample and intercepted with the detector. The
detector signal is capacitively coupled to a first amplifier, whose
signal measures the reflection at 830 nm. The detector signal is
also sent through a low-pass filter to a second amplifier, whose
signal measures the reflection at 980 nm.
[0118] In another embodiment, the 830 nm laser is modulated at 2
kHz and the 980 nm laser is modulated at 0.2 kHz. The detector
output is sent to two bandpass filters, one set at 2 kHz and the
second at 0.2 kHz. The output of the first filter provides the
reflectance signal at 830 nm; the output of the second provides the
reflectance signal at 980 nm.
[0119] Numerous modifications and adaptations of the
above-described embodiments, implementations, and examples will
become apparent to a person skilled in the art of measuring
properties of various workpieces in general and semiconductor
wafers in particular. For example, in an alternative embodiment,
instead of using a laser to make the high resolution measurement,
measurement by another method (e.g. near-field optical microscopy)
may be used.
[0120] In certain embodiments, the first measurement and the second
measurement are performed in sequence (in any order), without any
intervening process steps of wafer fabrication. In some
implementations, a number of first measurements are made in
sequence one after another before a corresponding number of second
measurements are made. In other implementations, a pair of
measurements (namely a first measurement and a second measurement)
are made in sequence, followed by another pair of measurements. In
another implementation, a first reference measurement is made and
used as a common reference for a set of near-by high-resolution
measurements.
[0121] In another embodiment, two or more measurements of the type
described above are made employing the same process although the
resolution of each measurement may be different, and the
measurements are used together to determine the property of
interest. For example, a low resolution measurement is used to
calibrate a high resolution measurement.
[0122] Furthermore, acts of the type described herein can be
combined with acts described in any of U.S. patent application Ser.
Nos. 09/799,481, 09/544,280, 09/276,821, 09/521,232 and 09/788,273
incorporated by reference above.
[0123] Also, the above-described method 90 of FIG. 1A (which is
performed when executing acts 10-70) can also be performed in a
manner similar or identical to that described in the related U.S.
patent application Ser. No. 09/274,821, entitled "APPARATUS AND
METHOD FOR DETERMINING THE ACTIVE DOPANT PROFILE IN A SEMICONDUCTOR
WAFER," filed Mar. 22, 1999, by Peter G. Borden et al.
(incorporated by reference above), with a production wafer used in
all acts of the method. The dopant profile can be measured with a
SIMS at a reference site (in this case, the reference site might be
the center of the wafer because SIMS has very low throughput, so no
more than one site is feasible in production. The method of Ser.
No. 09/274,821 is then calibrated at the reference site and used to
measure uniformity over the whole wafer.
[0124] A dose measurement method of the prior art (e.g. practiced
by apparatus sold by Therma-Wave or Boxer Cross) that is sensitive
to both implant dose and energy can be used with a calibration
measurement made by another device. For example, a SIMS profile may
be used to determine the dose and energy at a single point,
followed by fitting the as-implanted profile to models of ion
implantation profiles as a function of dose and energy (energy
effects the depth of the profile, dose the amplitude of the
profile) It can then be assumed that the energy is constant across
the wafer and dose variation across the wafer can be measured.
[0125] Therefore, the invention enables use of a second measurement
that has desirable features (e.g. spatial resolution, throughput,
sensitivity not available from the first measurement. The invention
may also be used in cases where the second measurement requires
additional information available only from the first
measurement).
[0126] Moreover, measurements of the type described herein can be
made at the same location or at different locations, depending on
the implementation. When made in different locations, the locations
may be selected to be optimum for a process that is to be used at
that location. For example, a low-resolution process may be used in
an area that has no pattern and a high-resolution process may be
used in an area having a pattern.
[0127] In one embodiment, the locations are sufficiently close to
one another so that a number of properties of the workpiece, other
than a property of interest (such as thickness of the topmost
layer), remain substantially identical (e.g. vary by less than 1%)
between the locations, while the property of interest is
substantially different (e.g. changes by more than 1%). However,
multiple properties of the workpiece in two locations at which
measurements are performed need not be substantially identical,
e.g. if a rate of change of such properties is known, e.g. from
low-resolution measurements at the two locations.
[0128] Therefore, numerous such modifications and adaptations of
the above-described embodiments, implementations, and examples are
encompassed by the attached claims.
* * * * *
References