Note: Descriptions are shown in the official language in which they were submitted.
6~3
Field of the Invention
This application is a divisional application of Serial
~o. 402,975 filed May 14, 1982 for an APPARATUS FOR PROJECTING A
SERIES OF IMAGES ONTO DIES OF A SEMICONDUCTOR WAFER, and although
the whole apparatus is disclosed in this application, this in~ention
relates specifically to a focussing system for maintaining a wafer
at the focal plane of a projected image.
Background
The fabrication of integrated circuits requires a
method for accurately formin~ patterns on a semiconductor
waer. A photoengraving process known as photolithography,
or simply masking, is widely employed for this purpose. The
microelectronic circuit is built up layer by layer, each
layer being based on a pattern received from a
photolithographic mask. Such masks typically comprise a
glass plate approximately the size of a wafer, the plate
having a single pattern repeated many times over its
surface. Each repeated pattern corresponds to a pattern to
be imposed upon a layer of a wafer.
The mask patterns are derived from an optical reticle
having a primary pattern which may be generated by a
co~puter controlled light spot or electron beam which is
scanned across a photosensitive plate. The reticle pattern
is typically ten times the final size of the pattern to be
imposed on the wafer An image one-tenth the size of the
reticle pattern is projected optically on the final mask.
The reticle pattern is reproduced side by side many times on
the mask, in a step-and-repeat process. Recent advances in
reticle production have made it possible to produce reticles
having patterns the same size as the final pattern. If such
a reticle pattern could be aligned and focused onto a wafer,
the mask fabrication could be substantially simplified or
entirely eliminated thereby achieving a substantial savings.
~ 7'~
The photolithographic process requires that each
pattern on the mask be positioned accurately with respect to
the layers already formed on the surface of the wafer. One
technique is to hold the mask just off the surface of the
wafer and to visually align the mask with the patterns in
the wafer. After alignment is achieved, the mask is pressed
into contact with the wafer. The mask is then flooded with
ultraviolet radiation to expose photoresist on the surface
of the wafer. The space between the wafer and the mask is
often evacuated to achieve intimate contact; atmospheric
pressure squeezes the wafer and the mask together. The
latter apparatus is typically known as a contact printer.
One defect of contact printers is that the masks quickly
~ecome abraded and useless. Since mask fabrication is
- 15 expensive, it would be desirable to have another method that
did not wear out the mask.
In view of the foregoing, a recent trend has been
toward a technique known as projection alignment, in which
an image of the mask pattern is projected onto the wafer
through an optical system. In this case, mask life is
virtually unlimited. However, one drawback has been that
wafer sizes have been increasing, and the task of designing
optics capable of projecting an accurate image over the
larger area is becoming more difficult. Another drawback is
the moveable projection optical system used in some machines
for focusing a projected image onto a wafer. It is often
difficult to focus such moveable optical systems and to ho]d
the system in focus.
Recent prolection aligners have attempted to circumvent
the extreme difficul~y of constructing a lens capable of
resolving micrometer-sized features over an area of many
square inches. A much smaller area, on the order of one
square centimeter, is exposed, and the exposure is repeated
by stepping or scanning the projected image of the mask
pattern over the wafer. Such machines are known as
projection steppers. So far, all of the efforts to provide
commercial]y acceptable projection steppers have been less
than satisfactory. It would be desirable to have a projection stepping
machine capable of using the now available, smQller reticles for directly
forming patterns on wafers, thereby eliminating the need for a large,
multiple pattern mask.
Summary of the Invention
The invention provides an apparatus for projecting an image of a
reticle pattern onto a ~afer, with one-to-one magnification. The apparatus
includes means for holding a reticle containin~ a pattern corresponding to
~ the size of the desired wsfer pattern. A one-to-one stationary projection
optical system projects, during illumination, an image of the reticle
pattern onto a predetermined focal plane. Suitable means such as,
preferably, a vacuum chuck holds the wafer. An alignment system preferably
steps and orients the wafer chuck to register markings on the individual
dies of the wafer with correspondins markings on the reticle. A fluid servo
system acts on the chuck to hold at least a portion of the wafer in the
predetermined focal plane of the projection optical system.
As part of the focusing system, the varuum chuck platform,
preferably, has three arms e~tending radially outwardly from the chuck.
Attached to each arm is a piston that can be moved up or do~n in a cylinder
- by means such as a pair of oppositely actiYe diaphragms.~ The diaphragms and
the upper and lower end walls of the cylinder define upper and lower fluid
chambers. The lower chamber of each cylinder is supplied with fluid at a
predetermined fixed pressure. Three fluid probes, each preferably
comprisin~ an orifice disposed adjacent to the wafer, are connected to a
source of fluid pressure. Each probe may also be connected by a fluid
signal line to a corresponding upper chamber of an associated cylinder.
Thus, the fluid pressure in the air gap between each probe and the wafer is
the same as that acting on the
corresponding upper dlaphragmO That upper diaphra~m
pressure is balanced by the fixed lower diaphragm pressure.
The system operates to maintain the wafer at a
predetermined distance from the orifices so that the surface
S of the wafer remains at the focal plane of the projection
system. If the air gap between the wafer and an individual
probe changes due to irregularities in the surface of the
wafer or other reasons, then the pressure acting on the
upper diaphragm will also change. If the wafer moves below
the focal plane, the air gap pressure drops and the pressure
in the upper chamber is reduced. As a result, the pressure
on the lower diaphragm acts upon the piston and the
connected support arm to raise the chuck and return the
wafer to its desired position at the focal plane~ Likewise,
if the wafer rises above the focal plane, thereby reducing
the air gap, then the pressure acting on the upper diaphragm
increases, thus forcing the piston and support arm in the
opposite direction to increase the air gap until the wafer
is returned to its desired position at the focal plane. In
the preferred embodiment, three sets of pistons and probes
are used to def.ine planar surfaces which can be positioned
accurately relative to the focal plan~ of the projection
sy~tem.
Brief _escription of the Drawlngs
Figure 1 is a plan view of a typical wafer having a
plurality of dies formed thereon.
Figure 2 is an enlarged fragmentary view of the wafer
of Fig. 1, showing the dies on the wafer and the fiducial
markers on the dies.
Figure 3 is a perspective, schematic view of a
projection stepper.
Figure 4 is an optical schematic view of the
illumination system of the projection stepper.
Figure 5 is an optical schematic view of the proiection
system and fiducial marker detection syste~ of the
projection stepper.
Figure 6 is a simplified plan view of the wafer
platform of the projection stepper, taken on line 6-6 of
Fig. 15.
Figure 7 is a fragmentar~ s~ctional and partly
schematic view of a portion of the wafer platform and
focusing svstem, taken on line 7 7 of Fig. 6.
Figure 8 is a fluid schematic of the focusing system.
Figure 9 is a front elevational view of the reticle and
apparatus for holding and advancing the reticle, taken on
line 9-9 of Fig. 15.
Figure 10 is a partial front elevational view of the
projection stepper showing the illumination system and
portions of the projectlon system in phantom lines.
Figure 11 is a partial sectional view taken along the
15 line 11-11 of Figure 10 showing the optica~ projection and
alignment systems in phantom lines.
Figure 12 is a simplified, enlarged, partial
perspective view of the illumination and projection optical
system.
Figure 13, located on the same sheet as Figure 9, is an
elevation sectional view of the optical illumination system, taken
on line 13-13 of Fig. 15.
Figure 14 is an elevation, partially sectional view
taken on line 14-14 of Fig. 10 of the optical projection system and
portions of the wafer platform, the photomultiplier assembly, and
the illumination system~
Figure 15, located on the same sheet as Figure 12, is a
partial sectional view of the illumination and projection optical
sys tems .
Figure 16, located on the same sheet as Figure 1, is a
detailed plan view of the wafer platform with portions of the focus-
ing system shown in phantom, taken on line 16-16 of Fig. 14.
Figure 17, located on the same sheet as Figure 11, is a
dual plot showing the relative intensity of illuminator output and
the sensitivity of a positive resist as a function of wavelength
between 400-450 mm.
Figure 18, located on the same sheet as Figure 12, is a
planar view of the exposure area of the projection optical system.
Figure 19 is an exploded view of the photomultiplier stage.
Best ~ode for Carryin~ Out the Invention
General
Figures 3 and 10 show perspective and front elevation
views of a projection stepping machine according to the
invention. A shelf 3 supports a wafer positioning system 79
including a chuck 32 shown in Figs. 6 ~ 8, 10, 12, 14 and
16. Underneath the shelf 3 is space 4 to hold power
supplies and a computer (not shown). Above the shelf 3 are
the illumination system 34, projection system 50, a dark
field automatic alignment system 60, and a cathode ray tube
display S for monitoring the alignment system 60.
In general operation a reticle 20 shown in Figs. 9 and
12 - lS is disposed between illumination system 34 and
projection system 50. Alignment system 60 controls the
movement of the wafer positionin~ system 79 to align the
dies 12 of a wafer 10 shown in Figs. 1 and 2 with the
; projected image of reticle 20. A focusing system 100 shown
in Figs. 6 - 8, 10, 11, 14 and 16 maintains the projected
image of the reticle pattern in optimal focus on the wafer.
The power output of illumination system 34 is increased to
develop the exposed (non-imaged) areas of the dies 12.
After exposure, the wafer positioning system is moved or
stepped to bring another portion of the wafer 10 into
alignment and focus with the projected reticle image.
~afer
Figures 1 and 2 show a wafer 10 provided with a
pluxality of dies 12 arranqed in rows and columns. Each die
12 has a pair of fiducial markers 14 and 16 at opposite
; corners of the die. The markers 14 and 16 may be in the
form of small "+" signs. ~s will be described in detail
hereinafter, the markers 14 and 16 are used to align the
dies with the pro~ected imaqe of the reticle pattern.
'
Reticle
Stepping machine 2 also includes a reticle 20 shown in
Figs 9 and 12 - ]5. Reticle 20 is mounted in a frame 22 an-d
has a plurality of patterns 24 arranged in a row within the
frame. Frame 22 in turn is disposed between a pair of
oppositely opening reticle guides 26. A pellicle (not
shown) covers the reticle 20. A pellicle is a thin,
transparent membrane which seals off the reticle surface
Erom dust and other contaminants. The pellicle is held in
frame 22 a predetermined distance from the surface of
patterns 24 so that the projected reticle image is
practically unaffected by contaminants adhering to the
pellicle.
Each pattern 24 has a pair of fiducial markers 28 and
- 15 30 at adiacent or opposite corners of the pattern in a
manner similar to the markers 14 and 16 on dies 12. The
markers 14 and 16 on each individual die are respectively
aligned with the markers 28 and 30 of the projected images
of reticle 20 before the image of that reticlé is printed on
each individual die.
l~lafer Positionincl System
Holding means, such as a vacuum chuck 32 shown in Figs
~ - 8, 10 - 12, 14, 15 and 16, is disposed below projection
system 50. Chuck 32 is moveable rectilinearly in two
coordinate directions, such as X and Y directions, to align
one of markers 14, 16 on the dies 12 with one markers 28, 30
on the pro~ected images of reticle 20. The chuck is also
rotatable in the same plane as that defined by the X and Y
directions, to align the other of markers 14, 16 Oll the dies
12 with the other of markers 20, 30 on the projected images
of reticle 20. Chuck 32 is also moveable vertically to
provide an optimal focusing of the projected images on the
dies 12, as will be discussed subsequently. The chuck 32 is
provided on its upper surface with a pluralitv of concentric
narrow lapped lands 302 shown in Fig. 8. Relatively wide
grooves 303 separate lands 302 for wringing in the wafer 20
7~
to lie substantially flat on lands 302 as described
hereinafter.
Illumination System
A reticle illumination system 34, shown in Figs. 3 - 5,
- 15 and 19, comprises a light source 35 such as a
mercury short arc lamp having a rating of 200 watts. The
mercury lamp is pulsed at 500 ~atts during wafer exposure
and held at a standby power of 100 watts during alignment
and other operations. Thus, the average power consumption
of the lamp during a typical wafer stepping operation is
approximately 200 watts.
An elliptical reflector 36 focuses the arc image of the
lamp onto one end of a light pipe 40. A dichroic mirror 37
- reflects only a selected wavelength band of light, thereby
preventing the infrared and ultraviolet portions of the lamp
spectrum from reaching the reticle. Hemispheric lenses 38,
39 are cemented to opposite ends of the light pipe 40 which
aid the coupling of the light in and out of the pipe 40 as
well as protect the end faces thereof. Light leaves the
light pipe 40, passes through lens 39 and a shutter stator
43 having moveable shutter 44, and a lens and mirror
arrangement 47 for illuminating a reticle 20.
The function oE the light pipe 40 is to efficiently
convert the nonuniform intensity distribution of light at
the lamp end to a uniform distribution of light at the
reticle end. Internal reflections within the light pipe are
essentially lossless. The incoming light is folded and
integrated with each internal reflection, thereby reducing
nonuniformities. A main advantage of the light pipe 40 is
that misalignment of the lamp or light source 35 merely
reduces the total output intensity without noticably
affecting the uniformity.
'7~
Optical Projection Svstem
Ali~nment Shutter
After a predetermined exposure, monitored by a detector
(not shown) located near the output of illumination system
34, the lamp power is dropped to 100 w and slmultaneously
shutter 44 is movecl into the aperture plane. A small
fraction of the light from source 35 passes through a cross
opening 45 in shutter 44 and illuminates the marker 28, 30
on the reticle 20. A high pass dielectric filter (not
shown) covers the opening 45 to prevent the g and h lines
from exposing the wafer duriny alignmentO On certain wafer
levels, it may be necessarv to use the mercury g line to
enhance the alignment signal. In this case, it can be shown
that the relative exposure ~alue of the intensity reaching
the wafer is 2% during normal exposure.
Broadband Illumination
At 500 w, the output intensity of the illumination
system 34 between 400-450 nm has been measured at .5w/cm2.
As seen in Fig. 17, this spectral distribution is
characterized by a high continuum with strong lines at 405
nm and 436 nm. Given the sensitivity of positive resist
shown in Fig. 17, approximately a 3-fold reduction in
exposure time is realized using the entire 400-450 nm band
as compared with using only the 436 nm line. Furthermore,
broadband illumination reduces the effects of standing
waves, resulting in improved linewidth control over oxide
steps.
Optical projection svstem 50, shown in Figs. 3 - 5, 11
- 15 and 19, projects an image identical in size and form to
reticle pattern 24 (i.e., without magnification or
reduction) onto a predetermined focal plane. Projection
system 50 comprises two components: a four inch front
surface spherical mirror 52 and a cemented achromat-prism
assembly 54. Assembly 54 comprises a cemented miniscus
element 53 and plano-convex element 55 which correct any
astigmatism of the concave mirror 52 at one-to-one for the g
and h mercury lines. A pair of prisms 56, 57 are part of
the optical design and also separate the reticle pattern
plane R from the wafer image plane W, as seen in Fig. 5. To
provide adequate clearance between vacuum chuck 32 and
reticle 20, the optical axis 51 is tilted at 15 degrees from
the horizontal, thereby placing reticle 20 at 30 degrees to
the X-Y plane of movement of vacuum chuck 32. Thus, light
passing through pattern 24 is reflected by the prism 56
through the lenses 55, 53 onto the mirror 52, back throu~h
the lenses 53, 55 and prism 57 and onto a wafer 10
positioned on vacuum chuck 32.
It will be noted that the mirror 52 includes a conical
aperture 58 which is part nf alignment system 60. Automatic
- 15 alignment of each die with the projected reticle image is
accomplished through the projection system 50 using a type
of dark field imaging to produce an alignment signal. The
desicJn of the projection system 50 is simplified by
providing for independent movement of the wafer to achieve
proper focus, so that the optical members may remain
stationary.
As shown in Fig. 5, mirror 52 and composite
achromat-prism assembly 54 are disposed symmetrically about
optical axis 51,. The reticle pattern plane R lies on one
side of the axis 51 and the wafer image or object plane W
lies on the opposite side. Projection system 50 is best
described with reference to the following Table I. It will
be appreciated by those skilled in the art that the Table
describes the optical system in accordance with the optical
surfaces and materials through which light passes along one
half of the optical path. Column 1 identifies the
successive surfaces. Column 2 lists the thicknesses in
millimeters of material behind the surface. Column 3 lists
geometric data and Column ~l lists materials. The materials
for surfaces B, C, D (prism 56 or 57, plano-convex element
55 and miniscus element 53, respectively) are identified by
ll
the names used by Schott Company, a well-known supplier of
optical glass.
TABLE I
1 ~ 3 4
Surface Thickness Radius of Material
in mm Curvature in mm __ _
A 1.79 ~ (flat) ~IR
B 26.80~ ~flat) LAKN7
C 10.02 35.00 KF6
10 D 37.60 74.95 SF~
E 189.37 264.00 AIR
-
Those skilled in the art will recognize that the Schott
material LAKN7 is a lanthium long crown glass; the Schott
m~terial KF6 is a light flint glass; the Schott SF2 is a
dense flint glass.
The prisms 56, 57 perform a plurality of functions.
The apex angle ~ of both prisms 55, 57 is 75; the opposite
equal interior angles ~ are 52.5. See Fig. 15. Small
notches 59 at apex angle ~ provide relief for thermal
stresses that develop in the prisms 56, 57 during full
illur~ination. Prisms 56, 57 couple light into and out of
projectlon system 50. In addition, the prisms are fashioned
to provide a 1.78 mm air gap between surface 56a and reticle
20 and between surface 57a and wafer 10. Such an air gap is
required for the needed mechanical clearances to move the
wafar 10 and the reticle 20 into and out of the respective
wafer image plane W and reticle pattern plane R. The air
gap is also sufficiently large enough that dust particles as
large as 200 microns will not adversely affect the system.
Such particles will not be focused in a 1. 78 mm air gap. It
is the unique com~ination of prism material and angular
configuration which efficiently couples the light through
the system 50 and provides the large air gap.
12
Another advantage of the optical system 50 is that all
the optical lens elements 53, 55, 56, 57 are all fashioned
from preferred glasses. Such glasses are more easily and
consistently manufactured than are other kinds of glasses.
Lens Fabrication
Three of the 10 optical surfaces are spherical, two of
which require fabricakion to better than ~/10. Surfaces A
and s are polished flat to ~/4. The prism diagonals 56d
and 57d are specified to ~/20 to minimize lens-to-lens
distortion. Autoalignment system 60 can be used to align
the concave mirror 52 to the prism assembly 54 by adjusting
for zer~ lateral 55, 53 color. With this alignment
procedure, the decentering tolerance for plano-convex
element 55 and meniscus element 53 is large by most design
standards, approximately 125 ~m. Cementing the prisms 56,
57 to the planar side of meniscus element 53 requires some
care to avoid vignetting of rays close to the edges of the
image field.
Optical coatings are applied to the air-glass surfaces
A and D to maintain spuxious reflections and ghost images to
less than 1~ of peak exposure. This is accomplished with a
single ~/4 coating of MgF2 on the glass-air surfaces. Due
to the prism design, total internal reflection occurs at all
ray angles, thus avoiding the requirement for metallic or
dielectric coatings on the prism diagonals 56d, 57d, which
would introduce polarization and phase disturbances with
possible adverse effects on the imaae quality. The concave
mirror 52 is coated with protected aluminum with
approximately 90% reflectivity from 500~600 nm. Including
absorption and surface reflections, the overall transmission
through the projection lens is 80~ in the 400-450 nm
spectral band of resist sensivity.
13
~o i~avelength Correction
_
The design performance of an actual projection system
is summarized in Table II. A Strehl ratio of 1.0
signifies a perfect lens whose performance is limited only
by fabrication errors and defocus. A dedicated lens design
program was developed to maximize the Strehl ratio at two
wavelengths over a specified field height. The design was
optimized at the g and h mercury lines (436 nm and 405 nm
respectively), achieving a minimum Strehl ratio of .99 over
a 16.8 mm field radius. The residual astigmatism is held to
within -.65 ~m of the focal plane which corresponds to a
peak-to-peak wavefront error over the aperture of ~/15.
Table II. Lens Performance
Numerical aperture O30
15 Field Height 16.8 mm
Corrected bandpass 400-450 nm
Alignment bandpass 400-600 nrn
Strehl ratio >.99
Min usable linewidth .80 m
Depth of focus ~1 um lines) 3.5 m
Telecentricity <1.0 mrad
Resolution and Depth of Focus
. _
It is estimated that a 4.8 ~m depth of focus is
required to maintain linewidth control to better than 0.125
~m. This estimate assumes a partia] coherence of ~= 0.4
~hich results from using an f/4 illuminator with an f/1.6
projection lens. The estimated linewidth variation is based
on a ~ 40% variation in actual resist exposure caused by
changes in wafer reflectivitv and topography. Subtracting
out the residual astigmatism, the usable depth of focus
becomes 3.5 ~m. The minimum geometry a~tainable in
production is estimated at 0.8 ~m, based on a l~m depth of
focus. Achieving this resolution, therefore, depends on the
1~
underlying topography and reflectivity associated with a
given wafer level.
Telecentric Design
An importan~ consideration in designing a one~to-one
pro~ection system is the requirement to loca~e telecentric
stops at reticle pattern plane R and wafer image plane W.
When this requirement has been met, rays entering parallel
to the optical axis on the reticle side exit parallel on the
wafer side. This feature ensures that no error in the size
of the projected image results from small changes in the
con~ugate planes. As seen in Table II, the projection
sys~em 50 departs on by 1 mrad from perfect telecentricity.
As a result, the axial position of reticle 20 can vary as
- much as -2 mils; and yet the magnification error will be
less than .05~rn over the entire exposure area.
Thermal Gradients
; Another factor to be considered with this type of lens
is the degradation of the image quality due to absorption of
near uv radiation within the lens elements. This problem
has been particularly troublesome in lenses designed to work
in the 365 nm region. With the proper choice of glasses
used in prism assembly 54 and complete rejection of
wavelengths helow 400 nm by illumination system 3~, the
optical effects due to absorption can be made negligible.
A computer simulation of thermal gradients produced in
the meniscus element 53 was performed using the
manufacturer's values for absorption and thermal
conductivity of the glass. The maximum time-averaged flux
through the recticle was estimated at 200 mw/cm2, ~stepping
one exposure per second. Assuming the ~orst case of a clear
reticle pattern with an area of 1.5cm2, the computer
simulation yielded a maximum temperature gradient of
.07C/cm within the meniscus lens 53. Estimates of the
inhomogeneity in the index of refraction, derived from the
simulated temperature profiles, ne~er exceeded 1.5xlO 6
'73
which is comparable to the best available optical cut
blanks.
Exposure Area
The exposure area of Fig. 18 has a circular perimeter
with a 16.3 mm radius and a cord 5.5 mm from the center. To
insure uniform resolution out to the corners of the exposure
area, the perimeter radius is purposely chosen 0.5 mm
smaller than the design field height H to provide a margin
of safety for errors in fabrication and reticle placement.
The constraint imposed by the cord insures clear passage
through the prisms of all rays originating from the lower
edge of the reticle field. As shown in Figure 18, the
lar~est squar~ area permitted with the above constraints is
lOxlO m~l. Also shown is the largest available aspect ratio
LS of 3:1, corresponding to a 7~21 mm exposure area. The user
can choose from a continuous selection of aspect ratios
between these two extremes with the total area per exposure
ranging from 1 cm2 to slightly under 1.5 cm2.
Choosing the larger areas can substantially reduce the
number of exposure steps required to cover a 4" wafer. For
exa~iple, a 2x8 clie arra~r with a 103~138 mil pitch will fit
into the 7x21 mm exposure area, requiring 51 steps to cover
a 4" wafer. If the pitch were changes to 158x180 mils, one
could fit a 2x4 array into an 8x18.3 mm exposure area,
requiring only 48 steps to cover a 4" wafer. In both
examples, over 90 exposures per 4" wafer would be required
with the 14-14.5 mm diameter fields a~railab]e with current
10:1 projection lenses.
Reticle Alignment
Reticle 20 in one actual embodiment comprised a 3" x 5"
x .090" L.E. 30 AR Cr plate 21, two reticle guides 26, and
pellicle frame 22, with pellicle membrane attached. The
reticle plate included four lX pattern fields 24 wi-th
fiducial marks 28, 30 in the corners of each field. ~o
larger a]ignment kevs were provided at each end of the
16
reticle to permit reticle frame alignment and three fiducial
marks were provided to permit reticle guide alignment, none
of which are illustrated in Fig. 9. The guides and pellicle
frame were attached with adhesives.
The various reticle alignment marks and keys may be
generated on plate 21, usually with an electron beam pattern
generator, at the same time as the pattern fields are
generated, to provide the necessary alignment accuracy. In
practice, three complete sets of pattern fields and
alignment marks and keys have been written one inch apart
across a 5"x5" plate. Such a plate can be rotated 90
degrees and its patterns compared in an automatic inspection
machine to permit selection of the best row. Large clear
windows (not shown) may be provided at top and bottom and
- 15 left and right to facilitate aligning pellicle frame 22 over
the best row of patterns. The frame may be glued on with a
die-cut double-sided adhesive ring. Such windows permit
inspection of the bond.
After the pellicle is attached, the plate is cut to
size in a suitable glass cutting fixture, care being taken
to protect the delicate pellicle. The plate is then placed
in a fi~ture that clamps guides 26 in position relative to
the fiducial marks provided to aid reticle guide alignment.
Guides 26 are glued in thi.s position. Careful positioning
of guides 26 ensures that the alignment mechanism of the
stepping machine will be within its operating range. See
Figs. 8, 14 and 19. The alignment kevs of reticle 20
preferably are placed a standard X-distance from a selected
origin so the machine can scan the image of the kevs through
the cross masks of alignment .system 60 by moving reticle 20
along the reticle stage and "know" where to place reticle 20
so that the fiducial marks 28, 30 will appear in cross masks
68, 69 when scanning a wafer. The alignment keys of reticle
20 preferably are placed at standard Y-distances from the
same origin (4500 ~ 3500 microns, left & right,
respectively) so the machine can scan the image of the keys
through cross masks 68, 69 of the alignment syste~m 60 bv
shifting the image wi-th the tilting window 64 of the
alignment system. See Fig. l9. The machine will then
"know" how much to tilt the window so that fiducial marks
28, 30 will appear in the cross masks when scanning a wafer.
The amount of tilt can be varied slightly for each reticle
pattern to compensate for small errors in mounting reticle
guides 26 to reticle plate 21.
Frosty Wafer
In order to produce an image at cross masks 68, 69 of
alignment system 60 in a dark field imaging system, a
special device called a "frosty wafer" is used to scatter
the light from the projected image of the alignment keys of
the reticle back into the dark center cone. The frosty
- wafer comprises a blank wafer with a 0.1 micron layer of
thermally grown silicon dio~ide for thermal coefficient
compensation, followed by a 1 micron laver of evaporated
aluminum to provide a mirror, topped by a 1 micron laYer of
unflowed silicon dioxide applied by chemical vapor
deposition to provide light scattering. The effect is
similar to a beaded movie screen, except that the qrain size
of the top layer is an order of magnitude smaller than the
projected image of the alignment keys. This contributes to
a smooth signal at a photomultiplier tube 66 provided in
alignment syster.l 60. Such a signal is necessary because
waf~r lO is scanned under ~he proiected image as reticle 20
is moved along the reticle stage 92.
Dark Field Automatic Aliqnment System
_ . _ . . . _ . . . _ . . _
Alignment svstem 60 comprises the necessary optical and
mechanical features to enable the stepping machine to ad~ust
itself for different size recticle patterns and for reticle
assembly errors. See Figs. 5, 10, 11, 14 and 19. Starting
at spherical mirror 52, light which was scattered into the
dark centra] cone bv the frosty wafer or a fiducial marker
passes through aperture 58, which formed the dark cone. The
~8
light beam i.s partially focussed by a 147 mmf achromat lens
62 and bounced up toward tube 66 by a folding mirror 63.
The beam passes through a tilt window 64 which will
re~ractively shift the image in the Y-direction if tilted by
a computer controlled window motor 65. The beam then passes
off center through a 200 mmf plano-convex lens 67 to finally
focus at the plane of a pair o~ cross masks 68, 69. The off
center passage corrects for color shift introduced when the
return beam passed off center through achromat-prism
assembly 54.
Before reaching cross masks 68, 69, the beam passes
through one of three apertures 70, 71, 7~ in a shutter 73
operated by a computer controlled shutter motor 74. Shutter
73 may be positioned so that right or left apertures 70, 71
- 15 alternately view reticle fiducial markers 28, 30 or the
additional reticle alignment markers or keys previously
described. Shutter 73 may also be positioned so that larger
: aperture 72 views both types of markers at once. The net
e~ect of this is to split the di~ference in alignment error
when aligning wafers. Large aperture 72 is fitted with a
50% neutral density filter to maintain constant signal
stength at tube 66.
The light beam is focused at the cross masks 68, 69.
The purpose of the cross masks is to blank out all light
coming up the dark cone e~cept that from the small area
around an alignment marker or key. Masks 68, 69 are mounted
in cross sliders 75 guided by a straiqhtedge ~not shown) and
are moved equal distances apart by a wedge 76 driven by a
motor 77. The compute.r can thus select the correct
separation for a given set of alignment markers or keys.
Aft:er the cross masks the beam begins to spread, so a
38 mmf lens 78 is provided to gather the rays enough to hit
the target cathode in photomultiplier tube 66. This tube
converts light beam intensity to an electrical signal which
is amplified and sent to the computer, in which the raw
signal is modified by a ~ero suppression circuit and
amplification gain adjus~ment to present the signal which
7~
19
the operator mav monitor on CRT display 5. By means of a
peak detection circuit an a voltage divider, the computer
selects a sample point. As an optical image is scanned, the
peak is detected and the siqnal drops to the sample point.
The computer collects a position sample from the laser
controlled stage. A scan in opposite directions cancels
phase (time delay1 errors so the computer can take the
simple average as the position of an alignment feature. If
inspection of the final results indicates a consistent
error, the user may enter a compensating offset in the
computer software.
Aperture 58 provides a dark cone or field in whlch
light, scattered by one of die markers 14, 16 is readily
detected by tube 66. As the pro~ected image of one of
- 15 reticle pattern markers 28, 30 is brought into registration
with one of die markers 14, 16, light scattered from the
pattern edges of the die marker passes through central
aperture 58 in mirror 52, as shown schematically in Fig. 5.
Such scattered light is transmitted through lens 62, past
mirror 63 and through an aperture of a shutter 64 ~Fig. 14)
for detection b~ photomultiplier tube 66, constructed in the
conventional manner to convert received light to a
corresponding production of electrons on an amplified basis.
This technique provides a high signal to noise ratio, so
that alignment accuracv is rather insensitive to defocus.
The output of tube 66 is displayed on CRT 5. The signal
waveform resembles a parabola as one of alignment marks 14,
16 is scanned along a given axis while being illuminated.
As shown in Figs. 10, 11, 14 and 16, chuc~ 32 is
disposed on a platform 79 supported on air bearings in a
well known manner. ~otors 80 (Fig. 10) 81 (Fig. 11) and 82
(~ig. 16) are respectively associated with the platform 79.
Motors 80 and 81 are respectively coupled to the platform 79
to move it horizontally in X and Y coordinate directions in
a conventional manner. Motor g2 is coupled to the platform
79 through a lead screw 83 to rotate the platform about a
vertical aYis extending through the center of the chuck 32.
A computer (not shown) processes the si~nals produced
by the tube ~6 to determine the relative coincidence of each
marker 14 on die 12 with the projected image of marker 28 on
reticle 20. The computer uses these signals to operate the
motors 80 and 81 for respectivelv driving the platform 79 in
the X and Y directions to position the image of markers 28
directly in registration with marker 14. Once the wafer 10
is aligned, its position can be subsequently accurately
monitored by any suitable means, such as a laser
interferometer system (not shown). Such a system will
continuously update the computer with signals representative
of the change in position of the platform 79.
When accurate registration has been obtained between
markers 14 and the image of marker 28, the computer
energizes a motor for rotating the shutter 64. Shutter
moves to a position where its aperture is provided for
viewing marker 16 on die 12 to determine its registration
with the projected image of reticle 20. The computer then
causes the motor to drive the platform 79 about the center
of the chuck 32 until marker 16 registers with the projected
image of marker 30.
The adiustments in the X and Y directions by the motors
80 and 81 and in the polar direction by the motor 82 may be
continued until alignment is simultaneously provided between
markers 14, 16 and the projected images of markers 28, 30.
Upon the occurence of such simultaneous alignments, shutter
44 is opened and source 35 is fully energi~ed to expose die
12 to the projected image of the patterns 24 on reticle 20.
The die 12 is thereafter treated tby apparatus not
constituting this invention) to produce electrical circuitry
in accordance with such image.
The pattern 2a on reticle 20 may be reproduced on a
plurality of different dies 12 on the wafer 10. ~uch
reproduction is under the control of the computer. ~lowever,
before such reproduction takes place, the chuck 32 is
repositioned so that the markers 14 and 16 on the next die
register with the projected images of the markers 28 and 30
on reticle 20. Such realignment is provided in the manner
described above.
Reticle Operat _
As previously described, reticle 20 comprises a
transparent glass substrate or plate 21, on which a
plurality of patterns 24 are provided. After one of
- patterns 24 has been reproduced on a particular number of
dies 12 in accordance with the controls provided by the
computer, reticle 20 may be advanced to the next pattern by
a push rod 25 and bell crank mechanism 27 that temporarily
couples the reticle to the X/Y platform 79. ~ee Figs. 9,
10, 12, 14 and 15. Reticle guides 26 are biased by spring
loaded roller 91 to bear against a reticle bearing and
- alignment member 92 shown schematically in Fig. 9. Member
92 has a smooth straight bearing surface on which guides 26
may be moved. This arrangement positi~ely locates reticle
20 in one direction. The reticle may be positively located
and advanced in the orthogonal direction by a push rod 25
havillg two closely spaced pins 226, 227. The leading pin
226 bears against the edqe of one guide 26. The other pin
227 sets into a recess provided in reticle guide 26. A bell
crank mechanism 27 shown schematically in Fig. 10, or other
suitable mechanism, selectivelv couples the push rod 25 to
the platform 79 for moving the reticle 20 from one pattern
to the next.
The controlled advance of reticle 20 through a distance
corresponding to the spacing of patterns 24 is facilitated
by the disposition of a pair of spaced rollers 94 that are
spring loaded to bear against reticle 20. This reticle
advancing feature facilitates the use of test reticle
patterns during printing operation. During movement of
reticle 20 from one pattern to the next, pressuriæed air is
applied through ports 96 beneath reticle 20 to displace the
reticle 20 into engagement with rollers 94. Upon movement
of reticle ~0 to the next pattern 24, the flow of
pressurized air through the ports 96 is discontinued and a
vacuum is applied to the ports. This causes the reticle 20
to become disposed against support surfaces 98 so that
patterns 24 will be in a fixed and proper position in the
optical path.
Focusing System - Construction
A focusing system 100 shown in Figs. 6 -- 8, 10, 11, 14
and 16 maintains the projected image of patterns 24 on
reticle 20 in focus on die 12. System 100 includes a
housing 101 for pro~ection system 50 and a block 102
extending downwardly from the housing 101. The bottom
surface 103 of block 102 extends above the top surface of
the chuck 32. Three pneumatic probe lines 104 a, b, c
extend downward through the block 102 to bottom surface 103,
as seen in Figs. 6, 7, 8 and 16. Pro~e lines 104 a, b, c
communicate with a pressure line 106 extending from a source
108 of pressurized fluid, such as dry nitrogen or clean,
dr~, compressed air.
The focusing system 100 includes an upper spider
assembly 201 and a lower spider assembly 202. The lower
asse~ly has three radial arms 202 a, b, c, each for
supporting a servo cylinder and piston assembly 110 a, b, c.
The piston 111 of each assembly 110 is connected to the
pedestal 33 of chuck 32 by an arm 124 a, b, c of upper
spider 201. Each upper spider arm 124 is connected to the
corresponding lower spider arm 202 a, b, c by one of three
f]exures 112 a, b, c to permit axial and prevent lateral
movement. The three pistons 111 establish three points
needed to define a plane parallel to the plane of wafer lO.
As shown in ~ig. 7, pedestal 33 comprises a housing 304
within which a plunger 305 is forced up against detent balls
306 by a plurality of springs 307 acting on a spring plate
308. A shaft 309 supports chuck 33 on plunger 305.
Microswitches (not shown) are actuated by a plate 308 to
shut down the stepper x/y stage if plunger 305 is knocked
off detellt balls 306 by accident. The lower spider 202 is
supported for rotation in short arcs by pre-loaded vee
~3
bearings 113 fixed to the top of platform 79. ~ ~00 step
motor 82 drives an 80 pitch screw 83 having a travelling nut
116 attached to spider 202 at a radius of ~.25 inches. This
linkag2 provides a theoretical resolution of 0.1 micron at
each end of a maximum 21 mm image.
Chuck 32 has a total vertical movement range of 0.090
inches and platform 79 has a 6.25 hy 12 inch travel, which
allow the chuck to load and unload itself and to rise above
wafer image plane W to contact pre-align microswitches (not
shown). A two~stage vacuum source 310 permits skidding the
wafer 10 onto chuck 32 at the loading station under light
vacuum to "wring it in" flatter on the chuck, after which
the pre-align microswitches can be bumped with the ~afer at
full vacuum without skidding. Concentric grooves 303
between narrow lapped lanas 302 on the surface of chuck 32
provide a place for minute particles to settle when the
wafer is wrung on at the loading station. The particles are
scraped of~ the underside of the wafer 10 and fall into the
grooves 303. Preferably, grooves 303 are substantially
wider than lands 302.
Fluid under pressure is introduced into the lines 104
a, b, c through the line 106 from the source 108. The air
flows through the lines 104 a, b, c to the bottom of the
block 102 and is discharged through orifices 103 a, b, and
c. Air flows through the space between bottom surface 103
and the top of the chuck 32. This flow of air provides an
air bearing between the block 102 and the chuck 32 to
maintain the block and the chuck in spaced relationship.
This spacing is in the order of a few thousandths of an inch
such as three thousandths of an inch (0.003").
Cylinder and piston assemblies lln a, b and c are
disposed at equally spaced positions around the periphery of
the chuck. Since the various components associated with
each assembly are identical, in the followin~ discussion
reference numerals wil] be used without alphabetic
desi~nations, except where required to distinguish one
assembly from another~ Each ol pistons 111 is coupled to
6~7~
the chuck 32 by a rigid arm 124. The vertical dispositior.
of the chuck 32 at a position adjacent to a piston 111
accordingly depends upon the vertical position of the
associated arm 124.
The upper end of each assembly ]10 is connected to a
pressure line 105 which is in fluid communication with a
probe line 104. The line 105 com~unica~es with an upper
chamber 117 defined at its upper end by a cover plate 118
and at its lower end by a resilient member or diaphragm 119
engaged with piston 111 and made from a suitable material
such as rubber. Diaphragm 119 is retained in stretched
relationship in chamber 117 by being clamped between the
cover plate 118 and the cylinder 120 of assembly 110.
Diaphraym 119 engages piston 111 which is vertically
- 15 moveable in position. Each piston 111 is connected to a
radiall~ extending arm 124 which is vertically slidable in a
slot 126 through the side of cvlinder 120. The flexure 112
is in turn attached at one end to the arm 124 and at the
other end to a lower spider arm 202.
A resilient member or diaphragm 130 made from a
suitable matexial such as rubber engages the underside of
piston 111 and is retained in a stretched relationship in
the lower chamber 131 by being clamped between cylinder 120
and the upper surface of spider assembly 202. Chamber 131
is preferably provided with a fixed pressure such as
approximately 7.5 pounds, or one half the pressure of the
source 108, ~ria a pressure line 132.
Focusing System
Figure 8 schematically illustrates how fluid from
source 108 is controlled by a pressure regulator 133 to
provide a precise output pressure, typically 15 psi. The
regulator can be adjusted to other pressures by precision
stepping motor 134. R three-way solenoid valve 135 is
operable to turn off air to lines 104 to prevent blowing
particles (dust) out of the chuck 32 when no wafer is
present. Needle valves l36 control the flow of fluid to
7~
orifices (probes) 103. Needle valves 136 are adjusted to
provide the same fluid pressure in the air gap as in the
lower chamber 131. The lower chambers 131 are held to a
predetermined fixed pressure, typically 7.5 psi via line
137, a pressure regulator 138, and feed lines 132. Pressure
lines 105 provide fluid comunication between the upper
chambers ]17 and the pressure in air gap ~ adjacent the
associated orifice 103. A solenoid valve l39 in each
pressure line 105 allows the computer to hold the upper
chamber 117 at a given pressure before the wafer edge steps
from underneath the probe. The air gap Z will develop a
given back pressure as a function of the rate of flow.
Thus, a change in the setting of the upper regulator 133
will simultaneously change the gap Z of all three orifices
103. The step motor 134 controls regulator 133 and is
itself driven by the computer to permit initial focus
adjustments.
Each pressure line 105 receives air at the same
pressure as that in the line 104. ~hen the pressure of the
air in the lines 104 and 103 increases due~ for example, to
a decrease in air gap Z upon placement of a wafer 10 on
chuck 32, an increased pressure is produced in the chamber
117. This pressure is exerted downwardly against diaphragm
119, so that piston 111 and arm 124 are accordingly moved
downwardly against the force exerted hy the pressure in
lower chamber 131 acting on lower diaphragm 130, which acts
as a return spring. The resultant movement downwardly of
the arm 124 produces a corresponding movement downwardly of
the chuck 32.
As the chuck 32 is moved downwardly, air gap Z is
increased. This relieves the pressure of the air in the
line 104 so that the pressure of the air in the line is
regulated at a substantially constant value. ~s will be
seen from the above discussion, the chuck 32 is wobbled
individually by each of the piston and cylinder assemblies
110 in a direction transverse to the X-Y plane of movement
of the chuck. In this way, the gap betweell the block 102
26
and a wafer on chuck 32 is regulated to maintain the die 12
on the w~fer in focus with the projected image of pattern 24
of reticle 20.
'
