U.S. patent application number 09/290785 was filed with the patent office on 2002-05-16 for target locking system for electron beam lithography.
Invention is credited to HARTLEY, JOHN GEORGE, KENDALL, RODNEY A..
Application Number | 20020056813 09/290785 |
Document ID | / |
Family ID | 23117558 |
Filed Date | 2002-05-16 |
United States Patent
Application |
20020056813 |
Kind Code |
A1 |
HARTLEY, JOHN GEORGE ; et
al. |
May 16, 2002 |
TARGET LOCKING SYSTEM FOR ELECTRON BEAM LITHOGRAPHY
Abstract
An e-beam lithographic system capable of in situ registration.
The system has an optics section such as a VAIL lens. A
controllable stage moves a substrate with respect to the beam axis
to place substrate writing fields beneath the beam. A field locking
target between the optics section and the stage has an aperture
sized to permit the beam to write a target field on the substrate.
The field locking target includes alignment or registration marks
around the aperture. A differential interferometric system measures
the relative positions of the field locking target and the stage
and controls stage position. The beam patterns the substrate on a
field by field basis. As the stage is moving into position for each
field, the beam is swept until it hits the alignment marks, thereby
checking system alignment. The beam control data, i.e., coil
currents necessary to hit the marks are stored, and drift
correction values calculated from the beam control data. Meanwhile,
pattern beam control is compensated by the drift correction
values.
Inventors: |
HARTLEY, JOHN GEORGE;
(FISHKILL, NY) ; KENDALL, RODNEY A.; (RIDGEFIELD,
CT) |
Correspondence
Address: |
Eric W. Petraske
IBM Corporation
Intellectual Property Law
Bldg. 300-482, 2070 Route 52
Hopewell Junction,
NY
12533-6531
US
|
Family ID: |
23117558 |
Appl. No.: |
09/290785 |
Filed: |
April 13, 1999 |
Current U.S.
Class: |
250/491.1 |
Current CPC
Class: |
B82Y 10/00 20130101;
H01J 2237/31754 20130101; B82Y 40/00 20130101; H01J 37/3045
20130101; H01J 37/3174 20130101; Y10S 977/887 20130101; H01J
2237/31766 20130101 |
Class at
Publication: |
250/491.1 |
International
Class: |
G01J 001/00 |
Claims
We claim:
1. A particle beam lithography system comprising: an optics section
containing beam deflection means for controlling the path of travel
of an electron beam along a system axis; a controllable stage at
one end of said path of travel, a target substrate held on said
controllable stage being positioned with respect to a system axis;
a field locking target spaced above said controllable stage; an
aperture through said field locking target; alignment marks
disposed around the aperture; and a substrate alignment system
measuring the relative positions of the field locking target and
the substrate.
2. The system of claim 1, wherein the substrate alignment system
comprises: a laser; a first target attached to said field locking
target and reflecting laser energy from said laser; and a second
target attached to said substrate and reflecting laser energy from
said laser.
3. The system of claim 2, wherein the field locking target includes
a grid on one surface, periodicity of said grid being the same as
beam subfield periodicity.
4. The system of claim 3, wherein the field locking target further
includes at least one subfield scan area.
5. The system of claim 4, wherein the field locking target is
carbon and the grid is gold.
6. The system of claim 5, wherein the field locking target further
includes a gold layer on a surface opposite said one surface.
7. The system of claim 5, wherein the field locking target is
positioned such that the grid is 4 mm above the target
substrate.
8. The system of claim 5, wherein said aperture is sized to permit
the beam to write a field on the substrate.
9. The system of claim 8, wherein the aperture is 0.2 mm larger
than the field size.
10. The particle beam system of claim 8 wherein said particle beam
is an electron beam.
11. The particle beam system of claim 1 further comprising:
adjustment means for adjusting the field locking plate
position.
12. The particle beam system of claim 11 wherein the adjustment
means mechanically adjusts the field locking plate position.
13. The particle beam system of claim 1 wherein the optics section
is a Variable Axis Immersion Lens (VAIL).
14. The particle beam system of claim 13 wherein the optics section
comprises: beam deflection plates for deflecting the beam to a
precise location; registration focus means for refocusing the beam
onto a reference target; autofocus means for adjusting beam focus
for target height variations; magnetic beam deflection means for
magnetically deflecting the beam to a subfield; and a projection
lens axis shifting yoke shifting the lens variable axis to follow
the deflected beam.
15. The particle beam system of claim 14 wherein the registration
focus means and the autofocus means are a pair of concentric
coils.
16. The particle beam system of claim 14 wherein the registration
focus means and the autofocus means are provided by a single
coil.
17. The particle beam system of claim 14 wherein the magnetic beam
deflection means is a pair of coils.
18. A method of calculating drift correction values for a particle
beam alignment system, said method comprising: a) initiating
movement of a stage to center a writing field on a substrate on
said stage; b) while the stage is moving into position, sweeping a
particle beam to hit marks adjacent to a centered field position;
c) recording beam control data when said marks are hit and
calculating beam correction values; d) measuring the relative
positions of said stage and said centered field position and
adjusting said centered field position as needed; and e) writing
the centered writing field with the beam corrected by the
calculated beam correction values.
19. The method of calculating drift correction values of claim 18,
wherein the centered field position is within an aperture in a
field locking stage and said marks are on said field locking stage,
adjacent said aperture.
20. The method of calculating drift correction values of claim 19,
before the step (a) of initiating stage movement, further
comprising the steps of: a1) placing a calibration target at said
writing field within said aperture; a2) sweeping said beam to hit
said marks adjacent said aperture; a3) centering said marks within
an expected location; a4) centering a first writing field within
said aperture; and a5) writing a pattern on said first writing
field.
21. The method of calculating drift correction values of claim 20,
wherein the step (a3) of centering the marks comprises mechanically
adjusting the location of said aperture.
22. The method of calculating drift correction values of claim 18,
wherein the centered field position is within said particle beam
alignment system's beam feedback capture range.
23. A method of writing patterns with an electron beam alignment
system, said method comprising: a) centering a field locking target
aperture; b) moving a stage to center a writing field on a
substrate on said stage within said aperture; c) writing a pattern
in said centered writing field; d) initiating centering of a next
writing field within said aperture; e) checking beam registration
while said next writing field is being centered; and f) repeating
steps (c) through (e) until all field on said substrate have been
written.
24. The method of writing patterns with an electron beam alignment
system of claim 23, wherein the step (a) of centering the field
locking target aperture comprises the steps of: 1) placing a
calibration target within said field locking target aperture; 2)
sweeping said electron beam to hit said calibration target and
marks adjacent said field locking target aperture; 3) adjusting
said beam to place said marks within an expected area; and 4)
mechanically adjusting said marks to place said marks at an
expected location.
25. The method of writing patterns with an electron beam alignment
system of claim 23, wherein the step (e) of checking beam
registration comprises the steps of: 1) scanning registration marks
adjacent said field locking target aperture; and 2) determining
beam correction values.
26. The method of writing patterns with an electron beam alignment
system of claim 25, wherein in the step (c) of writing the pattern,
the beam is adjusted by said determined correction values.
Description
BACKGROUND OF THE INVENTION
[0001] 1. Field of the Invention
[0002] The present invention generally relates to particle beam
lithography systems and, more particularly, to a particle beam
lithography system with in situ calibration and calibration methods
therefor.
[0003] 2. Background Description
[0004] Electron beam (e-beam) lithography tools are commonly used
in semiconductor manufacturing to form sub-micron shapes on a
semiconductor wafer. Shapes are formed by directing a beam of
electrons from a source at one end of a column onto a
photoresistive layer on a substrate at an opposite end of the
column. A typical substrate may be 200-300 mm in diameter or
larger. These submicron shapes may be formed either by writing the
shape directly onto a photoresistive layer on the substrate,
wherein the substrate is a semiconductor wafer; or, by writing the
shape onto a photoresistive layer on a substrate which is used as a
mask, subsequently, to print the shape onto the semiconductor
wafer.
[0005] Further, there are two broad types of writing modes used in
electron beam lithography. The first type is referred to as "blind
mode" or a "dead reckoning mode" and is commonly used in mask
making. In the blind mode, the substrate is a featureless blank
coated with resist and all of the patterns are placed by dead
reckoning. The second mode, which may be referred to as the
"registered write mode" or a "direct write mode," is commonly used
in direct write applications, i.e. writing directly onto a
semiconductor wafer, in what are referred to as device fabrication
runs. In the registered write mode case, the patterns must be
precisely located relative to previous levels which requires
registration targets built into each level and the substrate as
well. Regardless of the mode employed, accurately placing or
repeating sub-micron shapes at precise locations across a distance
of 200-300 mm demands precise beam registration.
[0006] However, even if the beam is registered adequately when
pattern printing begins, during the course of writing the pattern,
the e-beam may exhibit what is referred to as drift, i.e.,
exhibiting increasing inaccuracy in one direction as time passes.
So, in order to maintain adequate precision, pattern writing may be
interrupted periodically, depending on the particular tool's
inherent e-beam drift, to check tool registration and, whenever
registration error exceeds an acceptable tolerance, to adjust the
beam.
[0007] Normally, the substrate is held on a stage opposite
(beneath) the beam source and this registration measurement
involves diverting the stage to position a registration target
under the beam. Then, the beam is scanned over the registration
target, the target's location is measured and the target's measured
location is compared against an expected result. Any measured
errors are corrected by adjusting the beam or adjusting stage
positional controls. Then, the stage is returned to its former
position to resume writing the mask pattern. This measurement and
reregistration can be time consuming.
[0008] Furthermore, for this e-beam registration approach, the
registration measurement takes place at a stage location other than
where the pattern is actually written. Consequently, even after
measuring and correcting errors, moving the stage back into
position from the measurement area may actually introduce errors,
such as from the stage slipping or from other move related
stresses. Also, to assure complete accuracy, the beam should be
reregistered, frequently, preferably at each field. However, when
throughput is a consideration, as it nearly always is, it is
impractical to correct the beam registration prior to printing each
field.
[0009] Consequently, attempts have been made to perform
registration in place while writing the pattern, i.e., in situ. One
in situ approach, suggested by MIT as reported in
"Spatial-phase-locked Electron-beam Lithography:Inital Test
Results", pp2342-5 J. Vac. Sci. Technol. B. Vol. 11, No 6 November
December 1993, is referred to as the Spatial-Phase-Locked E-Beam
Lithography (SPLEBL) system. Implementing a SPLEBL type system
would require including special registration patterns on every
substrate and a blanket exposure of the substrate for the
registration cycle. However, such a blanket exposure is unusable
for high sensitivity resists. Further, a SPLEBL type system also
would require a sophisticated method of extracting positioning
information during exposure. While such a sophisticated method may
be feasible with a Gaussian or fixed beam shape; it is highly
improbable that such a method could be developed for a variable
shape beam such as a described in U.S. Pat. No. 4,243,866 entitled
"Method and Apparatus for Forming a Variable Size Electron Beam" to
Pfeiffer et al.
[0010] Another in situ registration technique has been proposed for
x-ray membrane exposure by Scientists at Naval Research
Laboratories (NRL) as described, for example, by Perkins et al. in
"Improving Pattern Placement Using Through the Membrane Signal
Monitoring," J. Vac. Sci. Technol. B, Microelectron. Nanometer
Struct. (USA) Vol. 16, No. 6 November December 1998 pp3567-71. This
technique requires fixing a registration grid to the carrier,
directly under the x-ray membrane. The registration grid forms a
Schottky diode junction with the carrier, allowing it to function,
simultaneously, as a high gain detector of incident electrons.
However, this registration technique is limited to high
transmission membranes and its resolution capability has not yet
been evaluated. Further, below the membrane, the beam diverges and
scatters rapidly, which limits the usefulness of this
technique.
[0011] U.S. Pat. No. 4,119,854 entitled "Electron Beam Exposure
System" to Tanaka et al. teaches an e-beam exposure system that may
be compensated for e-beam drift and workpiece drift. The system of
Tanaka et al. uses a pair of x and y lines as a reference target.
The relative position of the stage with respect to the x and y
lines is determined using differential interferometry. A coil is
included for refocusing the beam onto the x,y reference target
lines. However, refocusing the beam introduces hysteresis into a
system such as that taught in Tanaka et al. The hysteresis, itself
reduces beam accuracy.
[0012] Thus, there is a need for in situ registration methods for
e-beam lithography system and more particularly for VAIL e-beam
lithography systems.
SUMMARY OF THE INVENTION
[0013] It is therefore a purpose of the present invention to
improve e-beam lithography system registration;
[0014] It is another purpose of the present invention to reduce the
time required for reregistering e-beam lithography systems;
[0015] It is yet another purpose of the present invention to
improve VAIL e-beam lithography system accuracy;
[0016] It is yet another purpose of the present invention to
improve VAIL e-beam lithography accuracy without impacting e-beam
tool throughput.
[0017] The present invention is an e-beam lithographic system
capable of in situ registration. The preferred system is a Variable
Axis Immersion Lens (VAIL) e-beam system and is a double hierarchy
deflection system. A controllable stage moves a substrate with
respect to the beam axis placing the intended substrate writing
field within an aperture on a field locking target. The field
locking target is located between the optics section and the
substrate and the aperture is sized to permit the beam to write the
field. The field locking target includes alignment marks around the
aperture. A differential interferometric system measures the
relative positions of the field locking target and the stage. As
the stage is moving into position for writing a field, the beam is
swept to hit the alignment marks, checking system alignment. The
beam control data (coil currents and electrostatic deflection plate
voltages) required to hit the marks are stored, and drift
correction values calculated and the field beam control data is
compensated. Writing resumes on the newly positioned field with the
beam control data corrected by the calculated drift correction
values.
[0018] The field locking target may include a mechanical adjustment
for fine tuning aperture location. After an initial calibration of
the system, the field locking target is adjusted to position the
alignment marks at their expected location.
BRIEF DESCRIPTION OF THE DRAWINGS
[0019] The foregoing and other objects, aspects and advantages will
be better understood from the following detailed preferred
embodiment description with reference to the drawings, in
which:
[0020] FIGS. 1A-D show the pattern relationship between spots,
rectangles, subfields, fields and a skeleton;
[0021] FIG. 2 is a cross-sectional diagram of the preferred
embodiment e-beam lithography system;
[0022] FIG. 3 shows a sectional view of one corner of the field
locking target;
[0023] FIG. 4 shows a diagram of operation of the preferred
embodiment system;
[0024] FIG. 5 is a flow diagram of target mark acquisition and
centering;
[0025] FIG. 6 is a flow diagram for target locking and calibration
of the preferred embodiment system;
[0026] FIG. 7 is a flow diagram of a default A1 cycle;
[0027] FIG. 8 is a flow diagram of an A1 cycle in which the locking
target is scanned;
[0028] FIG. 9 is a flow diagram of the A1 cycle wherein target
locking is performed in conjunction with pattern writing.
DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT OF THE INVENTION
[0029] Referring now to the drawings, and more particularly, FIGS.
1A-D represent the hierarchical relationship between the individual
components printed to form a pattern. FIG. 1A represents a typical
substrate pattern that is divided into an array of fields 40 to
accommodate a pattern which is much larger than the preferred
e-beam lithography system's writing field. The collection of fields
40 is referred to as a skeleton 42. Each field 40 is identified by
the x,y address of its center point. In FIG. 1B, each field 40 is
divided into multiple sub-fields 44, each identified by the x,y
address of its center point. In FIG. 1C, each subfield 44 may
include one or more rectangles 46, 48, 50 and 52. Rectangles are
logical groupings and are not specifically addressed. In FIG. 1D,
each rectangle 46, 48, 50 and 52 is formed from groups of
individual spots 54.
[0030] The preferred e-beam lithography system is a Variable Axis
Immersion Lens (VAIL) e-beam system and is a double hierarchy
deflection system. U.S. Pat. No. 4,376,249 entitled "Variable Axis
Electron Beam Projection System" to Pfeiffer et al. and U.S. Pat.
No. 4,544,846 entitled "Variable Axis Immersion Lens Electron Beam
Projection System" to Langer et al., both assigned to the assignee
of the present invention and incorporated herein by reference, are
examples of VAIL e-beam systems. Although the preferred embodiment
system is a VAIL lens system, the present invention is not intended
to be limited to VAIL lens e-beam systems and other deflection
systems may employ the present invention as well with appropriate
modification. Furthermore, as used herein, substrate is not
intended as a limitation, but merely refers to material bearing the
surface upon which the pattern is being printed.
[0031] FIG. 2 shows a cross-sectional diagram of the preferred
embodiment e-beam lithography system 100. The preferred embodiment
system includes an optics section 102 with a registration focus
coil 104a, an autofocus coil 104b, beam deflection coils 106, 108,
a projection lens axis shifting yoke 110 and beam deflection plates
111. Although registration focus coil 104a and autofocus coil 104b
are represented as two concentric coils in FIG. 2, these two coils
could be combined into a single shared common coil with a summing
circuit combining their individual currents into a single driving
current. An e-beam source 90 emits a beam represented by arrow 112,
which, during writing, travels to a target field on a substrate
held on carrier 114. Autofocus coil 104b adjusts beam focus for
target height variations resulting from substrate imperfections,
thickness variations, etc. In the preferred Vail lens system,
double deflection yokes 106, 108 magnetically deflect the beam 112;
and axis shifting coil 110 shifts the variable axis of the
projection lens to follow the deflected beam 112. The relatively
slow magnetic deflection from coils 106, 108 determines the
subfield location, while within the subfield, the beam 112 is
deflected by the high speed electrostatic deflection plates
111.
[0032] A passive field locking target 116 permits the beam 112 to
write the pattern in the substrate's target field through an
aperture 118. The preferred aperture is rectangular and is large
enough to permit writing an entire field. During normal pattern
writing, substrate subfields are placed within the field locking
target aperture 118 and electrostatic deflection is used to write
spots which form the pattern shapes. During registration, the
subfield is defined as being over marks on the field locking target
116 adjacent to the aperture 118; and, the beam is deflected
accordingly, as represented by arrows 112'. Then, the marks on the
field locking target 116 are scanned, in situ, with the
electrostatic deflection, to provide near real time positional
feedback information. Typically, the marks on the field locking
target 116 are located 4 mm above the normal writing plane of the
substrate.
[0033] So, within the main field 40, magnetic deflection is used to
raster the beam to an x,y address for a selected subfield 44. Then,
the sub-field 44 is patterned using vectored electrostatic
deflection. Rectangles are not specifically addressed by any
element of deflection. Instead, the shaped beam is
electrostatically positioned at each spot 54 for sufficient time to
allow deflection settling and resist exposure time. So, groups of
spots 54 are written to form a rectangle 46, 48, 50 or 52.
Rectangles 46, 48, 50 and 52 are written to form a subfield 44.
Subfields 44 are written to form a field 40. Fields 40 are written
to form a skeleton 42 which completes the pattern.
[0034] After forming all of the rectangles 46, 48, 50 and 52 of the
selected subfield 44, the beam is magnetically deflected to the
next selected subfield 44. After forming all of the subfields 44 in
the selected field 40, the stage moved to position the next
selected field 40 in the aperture 118, and so on, until the
skeleton 42 is complete.
[0035] As indicated hereinabove, while some e-beam systems, e.g.,
Tanaka et al., may include a coil for refocusing the beam onto a
reference target, as with any electromagnet, hysteresis is
introduced into the system each time the electromagnetic lens
refocuses the beam. This focusing hysteresis effect of the
electromagnetic lens occurs as a result of superposition of the
field from the current flowing through the lens windings with the
fields induced in the surrounding magnetic medium. The first time a
magnetic lens is energized, a field is created which is the vector
sum of these two superimposed fields. When the coil current is
turned off, a residual magnetic field remains from magnetization of
the surrounding medium. If, prior to the cycling of this lens, the
electron beam had been in focus (as happens whenever the beam is
moved onto the field locking target), the beam will be out of focus
when it is returned to the substrate.
[0036] However, in the e-beam system of the present invention, the
registration focus coil 104a is continuously cycled between these
two locations. This continuous cycling occurs even if the reference
target is not being written or scanned and insures that the off
state magnetization stabilizes. With the off state magnetization
stabilized, an correction offset may be applied to the projection
lens. Thus, the focusing hysteresis effect is compensated for in a
preferred embodiment system.
[0037] For tracking and selecting stage location, the preferred
embodiment e-beam system 100 includes a differential
interferometric system 120. The interferometric system 120 directs
a laser, represented by arrows 122, to laser targets 124 and 124'
to measure the relative position of the field locking target 116 to
the stage mirror assembly 126. Laser target 124 is mechanically
coupled to field locking target 116 and laser target 124' is
attached to a stage mirror assembly 126. The carrier 114 is
kinematically clamped to the stage mirror assembly 126 at points
128. The stage mirror assembly 126, in turn, is flexure mounted to
a stage base 130 at points 132. An x or y drive 134 is attached to
an appropriate side of the stage base 130 to drive the stage,
typically under computer control, in the x or y direction; and,
once in place, to lock the stage in place. A mechanical centering
adjustment 136 provides a fine adjustment for the field locking
target 116 to precisely place it with respect to the beam.
Alternatively, instead of including mechanical centering adjustment
136, the beam could be centered by increasing the magnetic
deflection of the beam, but at the expense of increasing system
noise.
[0038] A pattern is formed by first positioning the stage, under
laser control 120, 122, at a corresponding x,y address of a
selected field 40. The stage pauses at the selected field 40 and
beam exposure begins and continues until the field pattern is
complete. It should be noted that the mechanical centering
adjustment 136 allows the field locking target 116 to be centered
both on an off center field. Mechanically centering the field
locking target avoids the noise that would be otherwise induced by
providing the deflection field with a sufficient range to
electrically compensate for misalignment.
[0039] FIG. 3 shows a perspective view of a corner of a field
locking target 116. Preferably, the field locking target 116 is a
carbon substrate with an aperture 118 tailored to be, preferably,
0.2 mm larger than the field size in each direction, e.g., for an
0.9.times.0.9 mm field size the aperture is 1.1.times.1.1 mm.
Aperture sidewalls 160 and the bottom of the field locking target
116 are coated with gold to provide a radiation barrier. A
perpendicular grid (not drawn to scale) or mesh is formed on the
upper surface by crisscrossing lines 162, 164 spaced at the
subfield periodicity. The lines 162, 164 are of a material selected
as having a much higher atomic number (high Z) and, therefore,
scatters electrons much more than the carbon background material.
So, for the e-beam, selecting contrasting materials with a
difference in atomic number of at least one order of magnitude
provides a strong material contrast to the scanning electron beam.
Thus, the carbon background material has a relatively low atomic
number (low Z) providing a high contrast to the high Z gold lines
of the preferred embodiment, which results in a high contrast back
scatter signal. Then, back scattered electron detection is used to
determine the position of the electron beam relative to marks.
[0040] Although the preferred field locking target 116 is gold on
carbon, other material combinations may be substituted, provided
they do not degrade the resulting back scatter effects by reducing
the signal to noise ratio or by adversely affecting the e-beam
environment. Further, the preferred field locking target 116
individual material thicknesses are not critical, but are selected
to be thick enough to block stray electrons or x-rays from reaching
the photosensitive pattern layer on the substrate, which would
inadvertently expose resist. Thus, the carbon thickness, the bottom
gold layer thickness, and the sidewall thickness must be considered
together in determining an appropriate thickness for each.
[0041] At each corner of aperture 118 are two A1 cycle subfields
represented by overlapping dotted line boxes 166 and 168. Subfield
166 includes four vertical scan rectangles 170 that identify the y
position of the locking target's horizontal bars 164. Subfield 168
includes four horizontal scan rectangles that identify the x
position of the locking target's vertical bars 162. Also, the
average of the scan rectangles 170 positions in A1 cycle subfields
166 and 168 serve to identify the location of the field corners
relative to the field locking target bars 162 and 164. This
positional relationship serves as the basis of computing the field
distortion. Individually, each of the four subfield scans 170
establish the x and y components of subfield distortion.
[0042] There are several advantages to providing multiple alignment
bars 162 and 164. First, with a single mark as in, for example, the
Tanaka et al. system, the target field must be located, precisely,
with respect to the single mark. This requirement is relaxed or
eliminated for a preferred embodiment system by making multiple
alignment marks (bars) 162 and 164 available at each corner for
registration. Including multiple bars 162 and 164 assures that a
registration mark is always located to within a grid period of the
edge of the aperture 118. Second, in the preferred embodiment
e-beam lithography system 100, only drift errors are measured at
the four corners of the aperture 118. This measurement provides
sufficient information to correct X and Y translation, X and Y skew
(rotation), X and Y magnification and X and Y trapezoidal
distortion. Optionally, more precise corrections for even higher
order terms, such as barrel-pincushion, symmetric and
anti-symmetric quadratic distortion etc. may be generated by
including additional target locations for measuring additional
error information. Third, at each corner of the preferred
embodiment, four vertical 170 and horizontal 172 scans may be made
at each calibration location, with each scan located at the nominal
subfield limits. These scans provide data to monitor and correct
subfield level distortions.
[0043] Having described the preferred embodiment system 100, system
100 operation can be understood with reference to the simple flow
diagram FIG. 4. Unlike prior art systems, the preferred embodiment
system 100 continuously cycles through 4 distinct phases, an A1
phase 180, an A2 phase 182, a B phase 184 and a C phase 186. During
the A1 phase 180, the beam is scanned across reference targets 162,
164 at the four corners of aperture 118. In the direct write mode,
the registration marks on the substrate are scanned during the A2
phase 182. The pattern is printed onto the main field of the
substrate during the B phase 184. The C phase 186 is preparation
for the next field, i.e., the table is moved to shift between
fields and, if necessary, beam servos may be adjusted, autofocus
height data may be collected, autofocus and field locking coil
digital to analog converters (DACs) may be loaded and the pattern
may be setup.
[0044] The preferred stage 128 includes a permanently mounted
system calibration grid. Prior to beginning normal operation, the
aperture 118 must be centered, mechanically, with respect to the
exposure field. To that end, the beam is scanned over it's full
deflection, in focus on the target grid. Back scattered electrons
from the calibration grid and the field locking target 116 are
detected and displayed. The field locking target's x/y position
relative to the deflection field is delineated by the change in
focus between the back scattered image of the calibration target
and the field locking target. The image from the field locking
target 116 is centered in the deflection field using mechanical
adjustment 136.
[0045] Then, as shown in the target mark acquisition and centering
flow diagram of FIG. 5, during an initial A1 phase 180, the beam is
scanned to find bars 162 and 164 in each of the four corner A1
subfields 166 and 168. The field is centered and in step 190, the
four expected locking target 166, 168 locations, i.e., at the
corners of the aperture 118, are scanned to determine if beam to
target alignment is sufficiently accurate. If the locking targets
are found at their expected location, no further adjustment is
necessary, and so, in step 192, target acquisition is complete. If,
however, the locking targets are not found at their expected
locations in step 190, then, proceeding to step 194, a check is
made to determine whether a mark 170 was detected in each scan. If
not, then in step 196, for each locking target 166, 168 where a
mark 170 was not detected, the mark deflection location is
adjusted; and, using an appropriate search delta, such as half the
length/width of a subfield, the area in the proximity of the
expected mark location is searched. If a mark is initially detected
in each scan in step 194 or, after finding the marks in step 196,
the individual centering of each mark 170 is checked in step 198.
If the marks are not properly centered, then, in step 200, for each
off center mark, the mark deflection location is adjusted by an
appropriate amount to center the mark. After either finding that
all of the marks are centered in step 198 or, after centering off
center marks in step 200, the A1 deflection is loaded in step 202
and, the A1 cycle is begun.
[0046] This is immediately followed by an A2 phase 182 and B cycle
184 where, as represented in the target locking and calibration
flow diagram of FIG. 6. First in step 204, the system calibration
grid, which is at the writing plane, is scanned and subfield center
and corner coordinates are measured and adjusted to bring the
deflection field and subfields into alignment with the grid. Then,
the deflection control parameters are perturbed by a preselected
amount and, the A1, A2 and B phases are re-executed. using the
writing plane calibration grid located on the stage, the A2 and B
deflections are calibrated to generate correction factors. Then, in
step 206, a relationship is established between the affect of the
correction factors in the writing plane and at the field locking
target.
[0047] Immediately after locating the center subfield marks, during
A2 and B cycle, the system calibration grid, mounted on the stage
at the writing plane height is scanned. Then, the subfield center
and corner coordinates are adjusted to bring the deflection field
and subfields into alignment with the grid. Finally, the deflection
control parameters are varied, intentionally, an identical amount
during each of an A1, A2 and B cycle. Any relative displacement
between the A1 and A2, B phases is measured and translated into
writing plane corrections. After this initial calibration,
substrate patterns are formed normally, cycling through all four
phases A1, A2, B and C.
[0048] As noted above, in the preferred embodiment system 100,
field locking is done regardless of the system's operating mode,
whether in INACTIVE, STANDBY or ACTIVE mode. Further, whether in
STANDBY or ACTIVE mode, the beam is scanned over the reference
targets 166, 168 during every A1 cycle 180 to maintain the targets
166, 168 in thermal equilibrium. Without continually scanning the
reference targets 166, 168, target temperature will vary
unpredictably due to the irregular power input. This temperature
variation will cause the reference targets 166, 168 to exhibit
unacceptable thermal drift. Although the reference targets 166, 168
are scanned every cycle, processing scan results is optional in
STANDBY mode, but processing is mandatory in ACTIVE mode to
recalibrate the system.
[0049] Further, in a direct write operation, the A2 correction
normally overrides any corrections derived from an A1 cycle, making
the A1 cycle is superfluous. However, both cycles may be used to
perform a field locking diagnostic in a field locking diagnostic
mode. In this diagnostic mode, a registration target that is fixed
onto the XY stage 114, is moved into the deflection field and the
A2 target scan is performed repeatedly on a the registration
target. The A1 scans are not processed. The A2 scans are not for
generating corrections; but instead, for monitoring field
translation drift and distortion. Then, drifts are induced,
deliberately, by perturbing the upper column, followed by A1 field
locking scans. Corrections are calculated from the A1 field locking
scans and applied to the tool. If the system is functioning
properly, the A2 scans should not measure any drift because, the
drift will have been detected in A1 cycle and applied as a
correction for the A2 and B cycle.
[0050] FIG. 7 is a flow diagram of an initial setup A1 cycle 180,
wherein the locking target is not scanned. Prior to centering the
deflection on the field locking target 116, stray electrons may
reach the substrate. Therefore, prior to centering the beam should
not be active whenever there is a substrate on the stage. However,
once the locking target has been acquired as described above with
reference to FIG. 5, the beam is active in every A1 cycle.
[0051] Thus, even though the beam 112 is inactive in the initial
setup A1 cycle 180, to maintain a stable duty cycle and to cancel
out hysteresis effects, focus coils 104a, 104b, and deflection
yokes 106, 108 and lens shifting yoke 110 are cycled. In step 210,
upon entering the A1 cycle, status bits are set to indicate that
the A1 cycle is in process (A1IP). Then, for the first time, in
step 212 the status bits are checked to determine if the A1 cycle
is in process as indicated by the A1IP bits being set.
[0052] If the A1IP bits are set then, in step 214, the autofocus
coil 104b is set to the calibration grid focus value grid location
and the registration focus coil 104a is set to its active value.
The beam feedback status (BF_STATUS) is placed in the system data
path and the beam feedback state (BF_STATE) is set to zero to
disable mechanical stage error corrections. During normal
operation, beam feedback information is collected to provide stage
positional information in the nanometer range. The collected
information is used as a fine tune adjustment to compensate beam
placement for stage vibrations and mechanical inaccuracies from the
coarser mechanical placement resolution. Typically, stage placement
mechanical accuracy is to within what is known as the beam feedback
capture range. During registration, when the beam is directed to
the registration target 116, the actual stage position is
irrelevant and, so, beam feedback is disabled.
[0053] Then, in step 216 is the magnetic deflection is driven to
actually deflect the beam to each of the corner subfields 166 and
168 to scan the reference target. Once scanning the reference
target is complete, in step 218, the autofocus coil is set to the
target value (i.e., the field address) and the focus locking coil
is turned off. The BF_STATUS is placed in the system data path, the
BF_STATE is set to one to enable mechanical stage error corrections
and, simultaneously, the A1IP bits are turned off.
[0054] Then, in step 212 the A1IP bits are checked again. However,
the recheck in step 212 indicates that the A1IP bits are off. So,
in step 220, an A2 HOLDOFF signal is checked to determine whether
to continue to the A2 cycle. If, in step 220, the A2 HOLDOFF signal
is active, the system enters a wait cycle in step 222.
Periodically, the A2 HOLDOFF signal is rechecked in step 220 until
it is found to be inactive, at which point the system proceeds to
an A2 cycle 182.
[0055] FIG. 8 is a flow diagram of an A1 cycle in which the locking
target is scanned, but which otherwise is substantially the same as
the flow diagram in FIG. 7. Accordingly, identical steps with
identical function are identified with identical numbers. In the A1
cycle of FIG. 8, when the system enters the A1 cycle, in step 216',
the locking target is scanned, measured and the scan results are
fed back. Additionally, once the preferred e-beam lithography
system enters the A2 cycle from the A1 cycle of FIG. 8, the system
resumes default cycling and begins processing the A1 scan feedback
signal in step 224.
[0056] As described with reference to FIG. 3, the mesh on the upper
surface of the field locking target 116 provides an initial guide
for the preferred system, aiding in initially locating suitable
reference targets. When field locking is initially activated, the
beam is sequentially deflected around the periphery of the field
aperture 118, during an A1 cycle 180, at expected mark locations.
In practice, however, because of deflection field distortions and
mechanical positioning errors, the marks are seldom found at these
initially expected locations. So, a searching algorithm is used to
locate center subfield marks at each of the respective marks and,
the corresponding mark command addresses are modified
accordingly.
[0057] FIG. 9 is a flow diagram of the A1 cycle wherein target
locking is done as a pattern is being written but, which otherwise
is substantially the same as the flow diagram in FIG. 8.
Accordingly, identical steps with identical function are identified
with identical numbers. So, in step 216", as the pattern is being
scanned, the registration focus coil 104a is checked in step 260 to
determine whether correction is needed. This check is done by
perturbing the field locking coil from a current value and
returning for a second A1 iteration in step 216". After the second
A1 iteration, in step 262, correction terms are computed and passed
forward for application to pattern position controls during the A2
cycle 182 and B cycle 184. Also, during pattern writing, in step
264, after entering the A2 cycle 182, the patterns are adjusted
using update corrections generated during the A1 cycle.
[0058] While the invention has been described in terms of preferred
embodiments, those skilled in the art will recognize that the
invention can be practiced with modification within the spirit and
scope of the appended claims.
* * * * *