 |
Description  |
|
|
TECHNICAL FIELD
This invention relates to a method and apparatus for making highly precise
measurements of objects within the field of view of instruments altering
the apparent size of such objects and, more particularly to a method and
an apparatus for measuring objects in the field of view of non-optical
microscopic instruments.
BACKGROUND ART
Many industries which utilize microscopic analysis rely upon highly
accurate and repeatable measurements of microscopic structures. For
example, in semiconductor technology measurements of structures on silicon
wafers or mask designs for fabrication of integrated circuits or devices
is often necessary. In response to precise measurement needs, various
direct measurement techniques utilizing both optical and electron beam
scanning devices have been developed. However, due to factors inherent in
such systems, these approaches become subject to increasing measurement
error as the size of the object being studied decreases. Since projected
technological advances, particularly in the semiconductor industry, are
expected to lead to even further reduced geometries, alternative
techniques are necessary to correct deficiencies found in the known art.
Possibly the most familiar among direct measurement approaches known to the
art are optical techniques which employ a laser beam, or other
monochromatic ray shaped by a mechanical slit, either of which are scanned
across the subject structure in discrete, known steps. The number of
sequential steps taken during the scan is simply counted and, for example,
when applied in the semiconductor industry, a linewidth can thereby be
calculated. Recent improvements to optical systems have provided an
increased capability to control measurements. Computers may conform the
mechanical slit to the basic shape of the subject structure and then
position the slit directly over it resulting in an increase in accuracy.
Also, video displays may be used to provide an image on a cathode ray tube
(CRT) which is divided into areas of known dimension. Such areas may be
counted either manually at the CRT or automatically by computer. Despite
these improvements, resolution capabilities ultimately limit all optical
systems to geometries no smaller than about 0.4-0.5 micrometers.
To make measurements of linewidths in the half micrometer category and
below, scanning electron microscope (SEM) approaches are necessary. SEM
systems known to the art operate by scanning a low energy (1-2 keV)
electron beam across the structure mounted on a stage. Backscattered
electrons and secondary electrons resulting from the physical interaction
between the electron beam and the subject structure are sensed by a
detector so that signals having a waveform characteristic of the
measurement subject may be generated. The signal data is digitized and
processed as a video signal to produce an image on a video display screen.
A conventional method of SEM linewidth measurement then entails measuring
the distance of a reference object whose distance is known and comparing
that measurement data with data obtained through performing an identical
routine with the subject of interest. Dimensions of the subject are
thereby estimates relative to the dimensions of the reference object.
The SEM linewidth measurement method as disclosed in U.S. Pat. No.
4,221,965, also allows correction for tilting of the semiconductor wafer,
including the linewidth to be measured, an unknown angle .theta. with
respect to the image plane. In that method, two cursors that define the
distance of measurement interest are manually aligned with two wafer marks
of interest, and a distance measuring circuit or computer then calculates
(using simple geometric relationships) the actual distance between the
wafer marks (the unknown linewidth) based on those cursor positions, the
azimuthal angle .DELTA..theta., and a preset magnification factor.
All of these prior art SEM methods are characterized by their heavy
dependency on careful adjustment and monitoring of several parameters
influencing magnification. These parameters include the SEM acceleration
voltage, focusing coil excitation, and working distance. Typically
calibration before each measurement may be necessary due to insufficient
monitoring and control of the acceleration voltage or focusing coil
excitation. The exact working distance may never be precisely known.
Moreover, the prior methods are further subject to error introduced by
operator misjudgments regarding focus of either the reference or subject
objects and error attributed to misjudgment of distance or alignment on
visual displays.
Thus, in order to eliminate the dependence of linewidth measurement
techniques on magnification and human factors introducing inaccuracy and
nonrepeatability, an improved method and compatible apparatus is
disclosed.
DISCLOSURE OF THE INVENTION
It is a primary object of the present invention to provide a method and
apparatus which allows precise and accurate measurements of objects of
small dimensions.
It is another object of the present invention to provide a method and
apparatus, as above, which allows precise and accurate measurements using
an electron device irrespective of substantial changes in magnification
due to acceleration voltage uncertainty, scan system drift and imprecise
knowledge of working distance.
It is another object of the present invention to provide a method and
apparatus, as above, in which an instrument for performing a linewidth
measurement is calibrated at the same time that a measurement is carried
out, using a calibration procedure independent of unreliable magnification
parameters.
It is a further object of the present invention to provide a method and
apparatus, as above, capable of providing accurate and repeatable
measurements within a range of approximately 0.1 micrometers to at least
100 micrometers.
It is yet another object of the present invention to provide a method and
apparatus, as above, which improves accuracy of measurement systems which
utilize scanning electron microscopes or other imaging devices.
It is still another object of the present invention to provide a method and
apparatus, as above, which is substantially insensitive to human error in
video cursor alignment.
These and other objects and advantages of the present invention over
existing prior art forms will become more apparent and fully understood
from the following description in conjunction with the accompanying
drawings.
In general, the method and apparatus of the present invention includes a
movable stage which carries the object or specimen to be measured. An
instrument, such as an SEM, is used to obtain a first scan representation
of the object. The stage is then displaced a known distance along a known
reference direction resulting in a displacement of the object in the
direction of the distance to be measured. Then the instrument is used to
obtain a second scan representation. Both the first and second scan
representations are stored by a processor means which compares the
representations to determine the apparent magnitude of the displacement of
the object in arbitrary units. The apparent magnitude of the displacement
is then equated with the known stage displacement to determine the
distance represented by arbitrary units. The processor means can then use
either the first or second scan representation to calculate the precise
measurement of any portion of the object.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram of a system for measuring objects according to
the concept of the present invention.
FIG. 2a is a representation of the field of view as would be shown on a
video display utilized in the present invention prior to the measurement
of the object.
FIG. 2b is a representation of the field of view as would be shown on a
video display utilized in the present invention midway through the
measurement of the object.
FIG. 3a is a representation of a typical signal to the video display which
creates the image appearing in FIG. 2a.
FIG. 3b is a representation of a typical signal to the video display which
creates the image appearing in FIG. 2b.
FIG. 4 is a partial block diagram of a system for measuring objects in
which an interferometer is utilized with a conventional mechanical stage
to displace the sample a precisely measured distance.
PREFERRED EMBODIMENT OF THE INVENTION
FIG. 1 shows, in block schematic form, an apparatus embodying the concepts
of the present invention. Though the description of the preferred
embodiment will only be directed to an X-axis scan, it would be
appreciated by one skilled in the art that the actual device used may have
identical capabilities for performing Y-axis and Z-axis scans.
A conventional SEM is generally indicated by the numeral 10. SEM 10
includes a focusing coil or lens 11 which focuses the emitted electron
beam probe 12, the electron beam being directed toward sample 13. Sample
13 may be any material suitable for scanning electron microscopy
examination, such as a semiconductor silicon wafer. Scan coils 14 control
the deflection of the electron beam 12 such that the beam scans across the
surface of sample 13 along a direction determined by the excitation of
scan coils 14. An electron detector 15 collects the backscattered and/or
secondary electrons 16 emitted from the sample when the electron beam 12
impinges on the sample. These primary elements of SEM 10 are all wholly
contained within a vacuum chamber 17 suitable to allow formation of an
electron probe.
Sample 13 is mounted on a stage 18. Stage 18 may be any conventional
device, such as the piezoelectrically driven stage shown in FIG. 1, which
provides precise movement over distances as short as 1 micrometer or less,
or as long as 10 micrometers. Stage 18 is operated by a conventional
piezoelectric driver circuit 19 upon receipt of a digital signal
representative of the desired travel distance from a conventional
microprocessor 20 programmable by one of ordinary skill in the art.
It will be appreciated that since the accuracy of the method described
herein is ultimately dependent upon the precision with which is known the
travel distance of stage 18, it is greatly desirable and may be essential
in determining this distance to employ an additional stage-shift
calibration. One such stage-shift calibration may simply involve the
independent measurement of travel distance by a highly precise optical
means such as the laser interferometer 32 shown in FIG. 4. Alternately, a
diffraction grating replica may be utilized in the manner explained
hereinbelow.
In addition to providing digital control signals for movement of the stage
18, microprocessor 20 provides control of the electron beam, acquisition
and storage of the video signal, and processing of the video signal to
calculate dimensional parameters of sample 13.
In controlling electron beam scan, microprocessor 20 clocks a counter 21
whose digital counter signal output is received by a digital-to-analog
(hereinafter called "D/A") converter 22 compatible in bit size with
counter 21, which transforms the digital counter signal into a linear
analog scan signal that determines the scan pattern of the electron beam.
After alignment correction noted below, the scan signal is attenuated by a
12-bit multiplying D/A converter 23 to provide for continuous
magnification control. Such control, in the preferred form, is adjustably
variable up to a factor of 20 to 1. The attenuated scan signal is
amplified by scan amplifier 24 and then applied to scan coils 14. Scan
coils 14 respond to the scan signal to generate a straight line scan by
the electron beam. Combined with scan amplifier 24, the magnification
control unit produces a range of line scans from 20 micrometers
(.times.5,000) down to 1.00 micrometer (.times.100,000). Counter 21 can be
a conventional 10-bit counter, in which instance 1024 discreet pixels
would comprise each line scan. It will be apparent to the skilled artisan
that the microprocessor clock rate signal used to clock counter 21 will
determine the pixel dwell time and thus the amount of time used to
generate a single line scan. A pixel dwell time of 50 microseconds will
result in approximately a 50 millisecond line scan time.
Often the direction of line scan is not aligned with the axis of travel of
stage 18, called the X-axis of stage 18. A digital rotate circuit 28,
interposed between D/A converter 22 and multiplying D/A converter 23,
provides a +90 degree rotation of the electron beam scan as observed from
the field of view shown by a display 27. Electron detector 15 generates an
analog video image signal of each line scan by methods well known to one
skilled in the art. This video signal is amplified by a conventional video
amplifier 25 and then digitized by an 8-bit analog-to-digital (hereinafter
called "A/D") converter 26. Also, the video image signal is received by
display 27 and by conventional methods provides a visual image of the line
scan. In order to enable the operator to identify the area of sample 13 to
be measured, display 27 should include at least one, but preferably two
potentiometers or other controls 27A for manually positioning a cursor at
the boundary of interest.
Microprocessor 20 initiates each line scan by directly clocking counter 21,
as previously described. After each clock pulse output, the microprocessor
inputs, after a predetermined delay, the corresponding 8-bit video signal
from A/D converter 26 and sequentially stores this value in a memory
address which directly represents the horizontal distance (in pixels)
along the line scan. The delay is desirable to allow sufficient pixel
dwell time to generate a stable signal from electron detector 15. It will
be apparent to one skilled in the art that each of the 1024 memory
addresses corresponds to a particular pixel location along the line scan.
The remaining features and elements of the invention can best be described
by an explanation of a typical calibration and measurement process, such
as a linewidth measurement on a semiconductor wafer.
Simply stated, the method herein disclosed involves correlating a precisely
known stage 18 travel distance to the observed video shift to determine
what shall be called the calibration factor. The calibration factor yields
a precise knowledge of instrument magnification, which in turn permits
precise distance measurements between points on objects on stage 18 by
measurement of their observed video separation, followed by a simple
algebraic calculation.
Initially it is desirable, and for some types of stages 18 may be
essential, to independently verify that the distance traveled by stage 18
is precisely known. In the present example, where a
piezoelectrically-driven stage is utilized, a diffraction grating replica
(hereinafter called the "grating") may be utilized. In this instance the
grating is placed on stage 18 and the two cursors aligned by the operator
to coincide with two grating lines at least two grating periods apart. The
actual separation between these two periods may be entered into
microprocessor 20 by the operator via keyboard. Microprocessor 20
determines the video separation (in pixels) between the periods defined by
the cursors by the use of correlation techniques. Grating edge roughness
is accounted for by averaging several measurements along the grating
lines. Microprocessor 20 next determines a scale calibration factor by
equating the video separation (in pixels) with the actual separation (in
micrometers). Several nominal 1 micrometer stage shifts are then actuated
by microprocessor 20 and the average apparent video shift (in pixels) is
calculated. The scale calibration factor is applied to the video shift
distance to yield a stage shift calibration factor in micrometers which is
used for all subsequent measurements. Where desired, additional
stage-shift calibrations may be done for stage shifts of other lengths.
After calibration is complete, the grating is removed from stage 18 and
sample 13, which may be any semiconductor wafer which has been processed
to the point of including circuit lines, photoresist lines, or oxide
lines, is placed thereon. SEM 10 is energized and the desired area to be
scanned is located in the field of view on display 27, by methods well
known to the skilled electron microscopist.
A typical raster image 31 of the field of view is presented on display 27
and can be seen in FIG. 2. The specific linewidth 30 to be measured is
shown as being in the right half of the raster 31 in FIG. 2a. The
magnification control provided by D/A converter 23 is preferably used to
maximize the area of the linewidth in the right half of the raster in
order to maximize resolution. The digital scan rotate circuit 28 can be
used to orient the linewidth as in FIG. 2a such that the electron beam
horizontal line scan will be orthogonal to the linewidth in this example.
Cursor control 27A is adjusted by the operator to align a cursor over the
point in linewidth 30 at which the line scan is to be taken.
Microprocessor 20 then initiates the line scan as previously described.
When the first line scan is made, 1024 digitized video samples from A/D
converter 26 are stored by the microprocessor in a temporary memory. Nine
more line scans are made and each time 1024 video signals are added into
the temporary memory. When 10 scans are completed each of the 1024 memory
addresses in the temporary memory contains 10 readings summed together.
Each value in the temporary memory is then divided by 10 to obtain an
average video signal, a conventional video signal processing technique to
reduce noise. The 1024 average values are then sequentially loaded by the
microprocessor into a video memory.
When the initial scan (the ten scan average) is completed microprocessor 20
produces a digital signal to piezo driver 19 to cause stage 18 to shift.
The shift must be such that the linewidth being measured remains within
the raster 31 as shown in FIG. 2b. In the example being described, for a
magnification range of .times.5000 up to 10,000 the stage shift is
nominally 4 micrometers; for a magnification range of 10,000 up to 20,000
the stage shift is nominally 2 micrometers; for a magnification range of
20,000 up to 40,000 the stage shift is nominally 1 micrometer; and for a
magnification range of 40,000 up to 100,000 the stage shift is nominally
0.5 micrometer.
After the stage has moved, microprocessor 20 initiates the second set of 10
line scans. The 1024 averaged video samples are generated as previously
described and are stored in a second video memory in microprocessor 20. An
analysis is then performed by microprocessor 20 on the data in the two
video memories in order to determine the apparent video line shift. FIGS.
3a and 3b illustrate typical video signals which would be stored before
and after the shift of stage 18, respectively. The apparent video line
shift may be determined by conventional threshold techniques where the
contents of the video memories are searched for a preselected trigger
amplitude C, and the apparent line shift is found by taking the difference
between the video memory locations A and B containing this trigger
amplitude. We have found that by correlating the video signals in video
memories, the apparent video line shift can be determined substantially
independent of deleterious processing factors.
This preferred autocorrelation technique involves comparing at least a
portion of the first video memory (which will be called the "window")
containing at least one edge of the linewidth being measured in its
pre-shift position with a window of comparable width in the second video
memory containing the post-shaft video image of the same linewidth edge.
The comparison is incremented pixel by pixel until the difference between
video amplitude data stored in the two video memories is at a minimum,
whereupon the two profiles will have been correlated and the distance (in
pixels) between the two comparison windows equals the apparent video line
shift. It will be appreciated that since each video amplitude is digitally
represented by a byte of preselected bit length, the differences examined
are most readily determined from the sum of the magnitude of the
differences for each corresponding bit pair in the two windows. At this
point the actual distance moved by the piezo stage is known in micrometers
from the calibration procedure, and the apparent video line shift has been
found by autocorrelation. A calibration factor in micrometers per pixel is
then calculated by dividing the piezo stage distance moved, by the video
line shift distance. Thus, by correlating a precisely known change in
distance provided by a shift in the piezo stage with the apparent video
image shift, a precise calibration of the SEM image is obtained,
independent of working distance, magnification and like parameters which
heretofore made calibration of the SEM image unreliable.
Once the calibration factor has been determined, the linewidth measurement
is next calculated. The data in either the first or second video memory or
a combination thereof can be analyzed to determine the linewidth. In this
example, analysis of data in the first memory will be described. Referring
again to FIG. 3a, the linewidth W is the mathematical difference between
distance D and distance A. These distances may be determined by first
selecting a suitable trigger value which could again but need not be point
C. Where different structures on sample 13 are to be measured the selected
trigger level can be adjusted so as to take into consideration inherent
profile differences between different structures which may be examined.
Such trigger levels may be based on theoretical approximations of the
structure profile under measurement. The values between the cursor
boundaries stored in the first memory are then sequentially scanned from a
first cursor boundary toward the second cursor boundary until the first
memory address is located which has a video signal value equal to or
greater than the trigger value. Next, the values between the cursor
boundaries stored in the first memory are sequentially scanned from the
second cursor boundary toward the first cursor boundary until the memory
address is located which contains a video signal value equal to or greater
than the trigger value. The memory addresses so determined represent
distance A and distance D in pixels. Linewidth W can then be calculated as
the difference between distance A and distance D in pixels. The resulting
difference is then multiplied by the calibration factor which is in units
of micrometers/pixel, to result in a linewidth W in units of micrometers.
A signal representative of length W can be transmitted by microprocessor
20 and received by a length display 29 which may be a conventional printer
or CRT.
Although it is not at all critical to the operation of the present
invention, it should be pointed out that another pre-measurement
calibration routine may be desirable. This provides calibration of the SEM
magnification at a particular working distance and voltage setting, for
example, to yield an estimate of the stage shift distance in pixel units.
The magnification calibration routine, when performed can, for instance,
greatly reduce processing time, since a majority of the video buffer
memory would not have to be searched in order to locate a shifted edge
profile. The SEM system magnification factor determines the appropriate
stage shift distance such that the chosen shift is within the field of
view. The image shift distance in pixel units becomes the calibrated shift
(CSHIFT) at the given magnification when the routine is performed
(CALMAG). Any changes in magnification should track the calibrated shift
and therefore the estimated image shift distance in pixels at any
magnification will be:
##EQU1##
where CALMAG is the magnification used when the magnification calibration
routine was performed;
CSHIFT is the image shift distance in pixel units as measured during the
magnification calibration routine; and,
MAG is the current magnification programming value for the scan.
Thus for any given magnification, the estimated shift distance in pixels
can be supplied to the processor to facilitate its search of the second
video memory.
Several possible modifications to the apparatus described herein would
occur to one of ordinary skill in the art. In addition to measurements in
the X and Y plane, height measurements could be made in the Z-axis by
developing a stage suitable for accurate elevational shifts. Also, memory
capacity could be increased allowing for higher resolution and larger
measurement areas.
Inasmuch as the present invention is subject to many variations,
modifications, and changes in details, a number of which have been
expressly stated herein, it is intended that all matter described
throughout this entire specification or shown in the accompanying drawings
be interpreted as illustrative and not in a limiting sense. It should thus
be evident that a device constructed according to the concept of the
present invention, and reasonably equivalent thereto, will accomplish the
objects of the present invention.
* * * * *
|
|
 |
|
 |
Description |
|
