 |
Description  |
|
|
The present invention relates, in general, to a process for producing
microdynamic structures and to the structures produced by that process,
and more particularly is directed to the fabrication of three-dimensional
tungsten cantilever beams on a substrate.
The field of micro-electromechanical systems is a new, emerging technology
which has as its goal the integration of electronic circuits, sensors, and
electromechanical motion devices to build complete electromechanical
systems on a micrometer scale. Recent research interest in such system has
focused on the fabrication of microactuators and micromotors which have
applications as micromechanical positioners, robotic actuators, and
microprobes. Recent research has shown that rotating and sliding
structures can be fabricated using modified silicon processing. To date,
research emphasis has been directed toward the fabrication of movable
microstructures using polycrystalline silicon with sacrificial layers
which, upon removal, releases the microstructures for motion. One
disadvantage of the present technology is that the deposition of only
relatively thin layers of polysilicon are practical, and thus the silicon
micromechanical structures are usually planar structures that are not
easily extended to three dimensions. Furthermore, these structures are
fragile and require many process steps to create a movable, free
structure.
Silicon-based electrostatic actuators and electrostatic motors are crucial
to the construction of integrated micro-electromechanical systems.
However, attempts to achieve motion in microdevices have been impeded by
the complexity of the processes required to build three dimensional
structures, and by the forces that make materials stick together upon
contact, to impede or prevent relative motion. Four major challenges to
the production of a working micro motor have been identified as being the
control of friction and wear; the control of surface charges and
interfacial forces; the development of a process that produces movable
parts; and the control of stress, especially in the movable parts. The
first two challenges require considerable experimental research to measure
and identify the controlling parameters, while the third and fourth
challenges are very closely related since a movable structure must conform
to a designed shape, whereas stress plays a major role in distorting a
structure when it is released from a mold.
SUMMARY OF THE INVENTION
The present invention is directed, in general, to a chemical vapor
deposition (CVD) tungsten/silicon process for fabricating movable
structures for electrostatic actuators and motors. The movable actuator is
fabricated on a suitable substrate such as a silicon chip, with careful
attention being given to production of a structure which permits
measurement of friction, stress, and electrostatic forces to enable the
production of a stress-free device capable of mechanical motion under the
control of applied electrostatic voltages.
Thin films several micrometers in thickness can be deposited using chemical
vapor deposition processes. Both CVD silicon dioxide and CVD tungsten
processes have been developed with deposition rates of greater than 1,000
angstroms per minute. As described in U.S. Pat. No. 4,746,621 of David C.
Thomas et al, assigned to the assignee of the present application, the
disclosure of which is incorporated herein by reference, a selective
tungsten on silicon process has been developed for producing patterned
integrated circuit metal layers. This process uses patterned CVD silicon
dioxide trenches which are ion implanted with silicon to make patterned
tungsten microstructures. In accordance with that patent, the CVD tungsten
seeds only on the silicon implanted at the bottom of the trenches, and
continues to grow vertically while filling the trenches from wall to wall.
No anisotropic tungsten etch is required to produce high aspect ratio
tungsten structures in accordance with that process.
The present invention is an extension of the selective tungsten CVD process
described in the aforesaid U.S. Pat. No. 4,746,621 to produce stress-free
cantilever beams that can be deflected using applied electric fields. In
accordance with this process, a pair of beams form the arms of
microtweezers which can be deflected in both the X-Y plane, which is the
plane of the two beams and which preferably is parallel to the plane of
the substrate, and in the Z direction perpendicular to the substrate
plane, by the application of potential differences between the tweezer
arms or between the tweezer arms and the substrate, respectively. The
application of such potential differences produces motion in the tweezer
arms which is a function of the time during which the potentials are
applied and which is a function of the magnitude of such potentials.
In accordance with a preferred form of the process of the present
invention, a layer of low pressure CVD silicon dioxide is applied to a
wafer coated with a thin silicon nitride layer, the latter layer serving
to ensure insulation between the substrate and the tungsten beams produced
by the process. A thin silicon nitride layer is deposited as an
implantation mask on the top of the silicon dioxide layer and the mask is
patterned to obtain vertical resist profiles. The channels for the
tungsten beams are then formed in the silicon dioxide layer by reactive
ion etching. After the photoresist is removed, silicon atoms are implanted
in the bottom of the channels. Thereafter, the silicon nitride mask is
removed and a tungsten film is selectively deposited to fill the implanted
oxide channels. An isolation mask is applied to etch the oxide surrounding
the tungsten channels to thereby free the tungsten beams. In a preferred
form of the invention, the beams are then covered with a conformal coating
of CVD silicon dioxide to provide a thin insulating layer on the beams.
Although the substrate preferably is silicon or gallium arsenide, it will
be apparent that any substrate that can receive the thin silicon nitride
and the CVD silicon dioxide layers can be used. The silicon or gallium
arsenide substrates are preferred, however, when the microtweezers are to
be incorporated into electronic circuitry such as VLSI circuits.
In a modified version of the invention, the channels are formed in an oxide
layer which is formed over a polysilicon layer, which in turn is formed on
the thin silicon nitride layer on the wafer. The channels expose the
polysilicon layer, which then acts to seed the CVD tungsten which is
deposited to form the tungsten beams. The cavity formed around the beam
exposes the nitride layer under the beams to free them, as before, with
the nitride layer remaining as an insulator on the top surface of the
substrate.
In a still further modification of the invention, the substrate may be a
silicon wafer formed by a Simox process, wherein a silicon substrate is
covered by a thin layer of silicon dioxide which in turn is covered by a
thin layer of silicon. This commercially available wafer is then covered
by an oxide layer such as CVD silicon dioxide, and trenches are formed
therein, as discussed above. These trenches extend down to the thin layer
of silicon, so taht CVD tungsten will seed in the trenches without the
need for ion implantation. A cavity is then formed around the beams to
release them, as previously described.
The tungsten beams are extensions of the tungsten microcircuit conductors
formed in a silicon dioxide layer, and extend in cantilever form from the
silicon dioxide layer into the formed cavity. The beams are elongated and
are substantially square in cross-section, and extend substantially
parallel to the floor of the cavity. Upon application of a potential
across the beams, the beams move toward or away from each other, while
application of a potential difference between a beam and the substrate
will cause the beam to move toward or away from the substrate.
In carrying out the foregoing process, tungsten beams have been fabricated
having a length of 200 micrometers, a height of 2.7 micrometers and a
width of 2.5 micrometers. Such beams have been moved in a controlled
manner by the application of selected voltages.
BRIEF DESCRIPTION OF THE DRAWINGS
The foregoing and additional objects, features and advantages of the
invention will become apparent to those of skill in the art from a
consideration of the following detailed description of a preferred
embodiment, taken in conjunction with the accompanying drawings, in which:
FIG. 1 is a perspective view of tungsten beams forming microtweezers in
accordance with the present invention;
FIG. 2 is a diagrammatic top view of the tweezers of FIG. 1;
FIG. 3 is a side view of the tweezers of FIG. 1;
FIG. 4 is an end view of one of the tweezer beams of FIG. 2;
FIG. 5 is a graphical depiction of the free end deflection of one tweezer
arm as a function of the applied voltage between the tweezer arms of FIG.
1;
FIG. 6 is a graphical depiction of forced density distributions along the
length of one tweezer arm;
FIG. 7 is a graphical depiction of the dynamic beam deflection as a
function of time for one tweezer arm;
FIG. 8 is graphical depiction of the dynamic beam deflection as a function
of position for one tweezer arm and for time;
FIGS. 9 through 15 illustrate the steps in a process of fabricating the
tweezer of FIG. 1; and
FIGS. 16-20 and 21-25 illustrate the steps in alternative fabrication
processes.
DESCRIPTION OF PREFERRED EMBODIMENT
Turning now to a more detailed consideration of the present invention there
is illustrated in FIG. 1 a segment 10 of a microcircuit incorporating a
pair of three-dimensional cantilever beams 14 and 16 fabricated from a
material such as tungsten in accordance with the present invention. The
microcircuit is formed on a substrate or base 18 which in a preferred form
of the invention is a silicon or a gallium arsenide wafer, but which may
be any material capable of receiving an insulating layer 20 of, for
example, silicon nitride. A layer of CVD silicon dioxide 22 covers the
layer 20, with the beams 14 and 16 being formed in a cavity 24 created in
the silicon dioxide layer. The beams are embedded in the silicon dioxide
layer 22 at their near ends 26 and 28, respectively, and extend outwardly
in a cantilever fashion from one face 29 of the cavity 24 with their far,
or free ends 30 and 32, respectively, being suspended in the cavity and
above the bottom surface 34 thereof. The beams 14 and 16 lie in a common
X-Y plane which may be parallel to the surface 34 and are free to move in
this plane and in a Z direction perpendicular to the X-Y plane. The near
ends of the beams 26 and 28 are formed as a part of, and thus are
connected to, metal circuit connector lines 36 and 38, respectively, in
the illustrated embodiments. These circuit lines may lead, for example, to
external contacts or other circuit components for supplying voltages to
the respective beams 14 and 16 to thereby actuate the beams in a manner to
be described. The two beams 14 and 16 in the illustrated embodiment can be
moved toward and away from each other upon application of appropriate
voltages between the beams, and accordingly can perform the function of
microtweezers. In one form of the invention, the beams may be formed of
material exhibiting piezoelectric characteristics, so that motion of the
beams by externally applied forces results in the generation of electrical
signals which can be detected by the microcircuit to which the beams are
connected.
As illustrated in FIGS. 2, 3 and 4, the beams are generally rectangular in
cross section and are capable of deflection through an angle .theta., with
the deflection occurring over the full length L of the beam, and with the
total motion away from the initial position defined by axis x being
identified in FIG. 4 by u(x). In FIG. 2 this deflection is shown in the
X-Y plane, but it will be apparent that a similar deflection can also be
produced in the Z direction. In one form of the invention, the length L
may be 200 micrometers, the width w of the beam may be about 2.5
micrometers, and the height h of the beam may be about 2.7 micrometers.
The beams 14 and 16 are rigidly fixed at their inner ends at cavity wall
29, which coincides with the axis y diagrammatically illustrated in FIG.
2, but are free to deflect over their entire length L. When a step
function potential difference is applied between the arms 14 and 16, as by
applying potentials to the lines 36 and 38, an attracting or repelling
electrostatic force occurs between the charged arms, depending on the
polarity of the potential. The arms act as electrodes, so the tweezers can
be mathematically modeled as an air gap capacitor, and the distance
between the two arms, d(x, t), is given by the following equation:
d(x,t)=d.sub.0 +2u(x,t), (Eq.1)
where u(x, t) is the deflection of one arm away from the x axis, at any
given position x along the tweezer arm, and where the potential is
supplied to the arms for a time t greater than 0. The spacing between the
tweezer arms 14 and 16 is illustrated at d.sub.0 in FIG. 1 and in FIG. 2.
The total capacitance density, c(x, t), along the two arms 14 and 16 is
the combination of the fringing capacitance and the plate capacitance, and
is given by the following equation:
##EQU1##
where .epsilon..sub.0 is the dielectric constant between the tweezer arms
and h is the height of the tweezer arm, as indicated above. Equation 2 is
an approximate expression to account for the fringing capacitance.
The force density F(x, t), per unit length along one of the arms due the
attraction of the other arm is given by the following equation:
##EQU2##
Since the total deflection of the tweezer arm at the free end thereof is
made much smaller than its length, the Euler-Bernoulli beam equation,
including the damping and external forcing terms for each arm, is given by
the following equation:
##EQU3##
where E is the Young's modulus for tungsten; M=.mu.mw is its mass density
per unit length: .mu. the volume mass density; I=h w.sup.3 /12 is its
moment of inertia; and P is the damping coefficient.
In order to solve the equation for the deflection of the tweezers in the
dynamic case, the partial differential equation 4 is replaced by a set of
difference equations using the central difference scheme. The scheme is
explicit and is stable when the temporal increment is sufficiently small
compared with the spatial increment. In the static case, the time
dependent terms vanish, and the resulting ordinary differential equation
can be replaced by a set of algebraic equations.
In order to determine when the free ends of the tweezer arms 14 and 16 will
touch in response to an applied potential, the deflection of tips 30 and
32 for different applied voltages may be simulated, with the results of
such simulations being illustrated in FIG. 5. In this Figure, curve 40
illustrates the free end deflection of one tweezer arm as a function of
the applied voltage between the two tweezer arms. As illustrated, for an
applied voltage slightly above 112 volts, the deflection of the arm
becomes unstable, and the tweezers close abruptly.
FIG. 6 illustrates the force density distribution along one arm of the
tweezers due to an applied voltage of 112 volts. Curve 42 illustrates the
density distribution at time t=0 upon application of a step function input
of 112 volts, while curve 44 illustrates the force density distribution at
time t=infinity. As illustrated by these curves, initially when the
tweezers arms are fully opened, the force density of attraction decreases
along the length of the arm. However, when the arms are deflected toward
each other, the force density decreases. For an applied voltage greater
than 112 volts, the forced density increases rapidly and the two arms
close.
FIG. 7 illustrates the dynamic beam deflection of one tweezer arm as a
function of time and for positions along the beam upon application of a
step input of 112 volts. Curves 46, 48, 50 and 52 illustrate the response
at four points along the arm, where X=80, 120, 160 and 200 micrometers,
respectively. In this case, the free ends touch at x=L=200 micrometers. At
other positions along the length of the tweezers the arms may collide and
bounce back and forth, exhibiting decaying vibrations. The dynamic
response approaches a steady state after a short time. The damped
responses are slower than the ones for the undamped case. The free ends
for the undamped case exhibit increased vibrational motion.
FIG. 8 shows the dynamic beam deflection as a function of the position for
one tweezer arm over a period of time upon application of a step function
of 112 volts. Curves 54, 56, 58 and 60 show the spatial shapes of the
tweezer arms for times t=0, 8, 15 and 20 microseconds, respectively. As
illustrated in this Figure, the two arms will deflect and touch within 20
microseconds.
The foregoing simulations illustrated in FIGS. 5, 6, 7 and 8 are for
voltages applied between the tweezer arms 14 and 16 or between one arm and
a nearby, horizontally spaced electrode, to produce motion in the X-Y
plane parallel to surface 34. In similar manner, the arms 14 and 16 can be
deflected in a vertical direction by application of a suitable potential
between a tweezer arm and a vertically spaced electrode such as the
substrate 18. It has been found that the voltage required to cau | | |
