Micromachining for Optical and
`Optoelectronic Systems
`
`MING C. WU, MEMBER, IEEE
`
`Invited Paper
`
`Micromachining technology opens up many new opportunities
`for optical and optoelectronic systems. It offers unprecedented
`capabilities in extending the functionality of optical devices and the
`miniaturization of optical systems. Movable structures, microactu-
`ators, and microoptical elements can be monolithically integrated
`on the same substrate using batch processing technologies. In this
`paper, we review the recent advances in this fast-emerging field.
`The basic bulk- and surface-micromachining technologies appli-
`cable to optical systems are reviewed. The free-space microoptical
`bench and the concept of optical prealignment are introduced.
`Examples of micromachined optical devices are described, in-
`cluding optical switches with low loss and high contract ratio,
`low-cost modulators, micromechanical scanners, and the X Y Z
`micropositioners with large travel distance and fine positioning
`accuracy. Monolithically integrated systems such as single-chip
`optical disk pickup heads and a femtosecond autocorrelator have
`also been demonstrated.
`Keywords—Integrated optics, integrated optoelectronics, micro-
`electromechanical devices, optical switches.
`
`I.
`
`INTRODUCTION
`The miniaturization and integration of electronics have
`created a far-reaching technological revolution. The inven-
`tion of integrated circuits not only allows a large number of
`transistors to be fabricated on the same silicon chip but also
`enables them to be interconnected into functional circuits.
`Today,
`the optics is at
`the same stage that electronics
`was a couple of decades ago: though high-performance
`optoelectronic devices have been developed, most of the
`optical systems are still assembled piece by piece. In 1969,
`Miller proposed the concept of “integrated optics” [1], in
`which he envisioned active optical devices interconnected
`by optical waveguides, similar to the way transistors are
`interconnected by wires in integrated circuits. Though there
`has been significant development of waveguide-based inte-
`
`Manuscript received July 13, 1997; revised August 5, 1997. This work
`was supported in part by the Defense Advanced Research Project Agency
`under Grant DABT63-95-C-0050 and in part by the Packard Foundation
`under Grant 92-5208.
`The author is with the Electrical Engineering Department, University of
`California, Los Angeles, CA 90095-1594 USA (e-mail: wu@ee.ucla.edu).
`Publisher Item Identifier S 0018-9219(97)08236-4.
`
`grated optics (also known as photonic integrated circuit),
`many free-space optical systems cannot be integrated by
`such technology. Free-space optics can perform optical
`imaging and generate diffraction-limited focused spots, and
`is widely used in optical display, data storage, switching,
`and sensing systems.
`The micromachining, or microelectromechanical systems
`(MEMS) [2], technology has opened up many new possibil-
`ities for free-space optical systems. Movable micromechan-
`ical structures as well as precision optomechanical parts
`can be made by micromachining—a batch-fabrication tech-
`nology similar to the microfabrication process for making
`very large scale integrated (VLSI) circuits. The movable
`structures are attractive for optical applications because
`small mechanical displacement can often produce physical
`effects that are stronger than the conventional electrooptic
`or free carrier effects. For example, a displacement of
`one-quarter wavelength in an interferometer can produce
`an ON/OFF switching. Many new optical devices and sys-
`tems based on movable structures have been reported.
`Compared with macroscale optomechanical devices, the
`micromechanical devices are smaller, lighter, faster (higher
`resonant frequencies), and more rugged. Very efficient light
`modulators, switches, broadly tunable semiconductor lasers,
`detectors, and filters can be realized by the optical MEMS
`technology [3], [4]. The optical MEMS technology is
`sometimes also called microoptoelectromechanical systems
`(MOEMS) or microoptomechanical systems (MOMS). In
`this paper, we will use these terms interchangeably.
`Optics is an ideal application domain for the MEMS
`technology: photons have no mass and are much easier
`to actuate than other macroscale objects. Microactuators
`with small force and medium travel distance are useful
`for many optical applications. Packaging of optical MEMS
`devices may also be easier than that of other MEMS devices
`since optics provides a noncontact, nonintrusive access to
`the MEMS devices. Its applications include projection and
`head-mount display, optical data storage, printing, optical
`scanners, switches, modulators, sensors, and packaging of
`
`PROCEEDINGS OF THE IEEE, VOL. 85, NO. 11, NOVEMBER 1997
`
`1833
`
`0018–9219/97$10.00 © 1997 IEEE
`
`Exhibit 1045, Page 1
`
`

`
`shown in Fig. 1(a). The depth of the V-groove can be
`very well controlled by lithography because {111} planes
`are effective stop-etching planes. Pyramidal-shaped holes
`that are ideal for holding ball lenses can also be formed
`by etching through square openings. These V-grooves and
`the pyramidal-shaped holes form the basis of conventional
`microoptical benches. As shown in Fig. 2(a), bulk optical
`components are dropped onto the etched silicon substrates
`and precisely positioned by holes of various geometry.
`Vertical micromirrors can be formed by anisotropic etching
`on a (110) silicon substrate [5], [6], as shown in Fig. 1(b).
`The atomically smooth {111} planes are perpendicular to
`the surface of the substrate and provide large-area, optical-
`quality surfaces. The selectivity of {110} over {111} planes
`is as high as 500 for some KOH solutions. Therefore,
`high-aspect-ratio microstructures can be produced this way.
`Micromirrors with 2- m thickness and 200- m height
`have been reported [9]. These vertical micromirrors are
`semitransparent and can be used as beam splitters. Fig. 2(b)
`illustrates the optical circuits consisting of such thin mi-
`cromirrors, including Fabry–P´erot and Michaelson interfer-
`ometer [9]. In addition to the {111} stop etch planes, some
`silicon etchants exhibit reduced etch rate in regions that
`are heavily doped with boron, adding more flexibility in
`defining the final shapes of the structures [5]. The boron
`diffusion therefore can be utilized to pattern membranes,
`suspended beams, or support beams for the vertical mirrors
`etched on (110) silicon substrate [10].
`The vertical micromirrors created by anisotropic etch-
`ing of (110) silicon substrates are, however, orientation
`dependent. Recently, there has been significant interest in
`three-dimensional structures created by deep reactive ion
`etching (DRIE) [11], [12]. DRIE allows etching of highly
`anisotropic, randomly shaped and located features into a
`single crystal silicon wafer, with only photoresist as an
`etch mask. Fig. 3 shows the scanning electron micrograph
`(SEM) of three-dimensional structures etched by DRIE
`[12]. Though the etched surface is rougher than the wet
`chemically etched (110) wafer, it is still smooth enough
`for many micromirror applications. Marxer et al. [13]
`and Juan et al. [14] have used the DRIE technique to
`fabricate a vertical micromirror for fiber-optic switches. The
`microactuators for the switch are also integrated through the
`same DRIE process. The mirror roughness is estimated to
`be 36 nm, which corresponds to a scattering loss of 6%
`[13].
`
`B. Surface Micromachining
`In contrast to bulk micromachining, in which substrate
`materials are removed to create three-dimensional struc-
`tures, surface-micromachined structures are constructed en-
`tirely from deposited thin films. Alternating layers of struc-
`tural and sacrificial materials are deposited and patterned
`on the substrate. The sacrificial materials can be selectively
`removed by an etchant
`that attacks only the sacrificial
`materials. Suspended beams, cantilevers, diaphragms, and
`cavities can be made this way. The use of sacrificial
`material to free micromechanical devices from the silicon
`
`(a)
`
`(b)
`Fig. 1. Schematic diagrams illustrating the profiles of anisotrop-
`ically etched silicon substrates. (a) (100). (b) (110).
`
`optoelectronic components. The marriage of optics and
`MEMS has created a new class of microoptoelectromechan-
`ical devices and integrated circuits that are more efficient
`than macroscale devices.
`
`II. MICROMACHINING TECHNOLOGY FOR
`OPTICAL APPLICATIONS
`
`A. Bulk Micromachining
`Bulk micromachining has long been employed to cre-
`ate three-dimensional optomechanical structures on silicon
`substrate for aligning optical fibers or forming microoptical
`elements [5]. Single crystal silicon has excellent mechanical
`properties, and silicon substrates with high purity are read-
`ily available at low cost for semiconductor manufacturing.
`Silicon can be machined precisely by anisotropic etchants,
`whose etching rates depend on the crystallographic orien-
`tations [5], [6]. The etching rate of anisotropic etchants,
`such as ethylene diamine pyrocatechol (EDP), potassium
`hydroxide (KOH), and tetramethylammonium hydroxide
`(TMAH), is much slower in the <111> directions than
`in the <100> and <110> directions. Selectivity, defined
`as the ratio of the etch rates of the desired direction to
`those of the undersized one, for such anisotropic etchants
`can be higher than 100. This is a very powerful technique
`to create three-dimensional optomechanical structures with
`high precision.
`Silicon V-grooves are now widely used in optical in-
`struments and packaging of fiber and optoelectronic com-
`ponents [7], [8]. They are created by anisotropic etching
`of a (100) silicon substrate with stripe openings along the
`<110> or <110> directions. The exposed {111} planes
`form a 54.74
`slope with the surface of the wafer, as
`
`1834
`
`PROCEEDINGS OF THE IEEE, VOL. 85, NO. 11, NOVEMBER 1997
`
`Exhibit 1045, Page 2
`
`

`
`(a)
`
`(b)
`Fig. 2. Silicon optical benches on silicon substrates. (a) (100) [7]. (b) (110) (after [9]).
`
`substrate was first demonstrated by Nathanson et al. [15]
`for fabricating a field effect transistor with a suspended
`resonant gate. In 1983, Howe and Muller [16] described
`the use of polysilicon as the structural material and silicon
`dioxide as the sacrificial material. Because of the excellent
`mechanical properties of polysilicon material and the high
`selectivity of sacrificial etching with hydrofluoric acid,
`this combination has become the most popular choice for
`surface micromachining.
`Fig. 4 illustrates the surface-micromachining process for
`making cantilevers. This process requires one layer of
`sacrificial material and one layer of structural material.
`The complexity of the surface-micromachining process can
`be quantified by the number of structural and sacrifi-
`cial materials. With two structural polysilicon layers, free-
`moving mechanical gears, springs, and sliders have been
`demonstrated [17], [18]. Micromotors [19], [20] and other
`
`microactuators were later demonstrated using similar fabri-
`cation processes. One of the main features that distinguishes
`the surface micromachining from the bulk micromachining
`is that many different devices can be fabricated using a
`common fabrication process. By changing the patterns on
`the photomask layouts, different devices such as cantilever
`resonators, sliders, micromotors, or comb drive actuators
`are fabricated simultaneously on the same substrate. This
`methodology is similar to that used in today’s VLSI circuits.
`For this reason, the surface-micromachining process is often
`referred to as an integrated circuit (IC) process or VLSI-
`like process. Today,
`there already are two commercial
`foundries offering such polysilicon surface-micromachining
`processes.1 Some of these processes can also be integrated
`1MEMS Technology Applications Center, Microelectronics Center at
`North Carolina (MCNC), Research Triangle Park, NC, and Integrated Mi-
`cro Electro Mechanical Systems, offered at Analog Devices, Cambridge,
`MA.
`
`WU: MICROMACHINING
`
`1835
`
`Exhibit 1045, Page 3
`
`

`
`of the substrate plane and become perpendicular to the
`substrate. It is also possible to achieve other angles for the
`polysilicon plates. The microhinge technology allows the
`three-dimensional structures to be monolithically integrated
`with surface-micromachined actuators. It
`is particularly
`useful for fabricating integrable free-space microoptical
`elements, as will be shown in the next section.
`In addition to the microhinges, alternative surface-
`micromachining techniques have been proposed for
`fabricating three-dimensional
`structures. The research
`group at
`the University of Tokyo proposed to use
`the “reshaping” technology to create complex three-
`dimensional structures [25]. The basic concept
`is to
`use thin flexure beams to connect polysilicon plates.
`The beams are then buckled or twisted by integrated
`microactuators to create the desired three-dimensional
`structures. By passing current through the polysilicon beam
`until plastic deformation, the three-dimensional structures
`are permanently fixed. Other techniques have also been
`proposed. Green et al. proposed to use the surface tension
`of molten solder to produce out-of-plane rotation [26].
`Smela et al. used active polymers for controlled folding of
`microstructures in electrolyte liquid solution [27].
`
`III. FREE-SPACE MICROOPTICAL BENCH
`These surface-micromachining techniques have opened
`up the possibility of monolithically integrating free-space
`microoptical elements, micropositioners, and microactua-
`tors on the same substrate. This new technology, called
`free-space microoptical bench (FS-MOB), is illustrated in
`Fig. 6 [28]. In free-space optical systems, photons propa-
`gate between optical elements without being confined in
`physical media. Normally, the free-space optical systems
`are constructed on optical tables, with each optical element
`mounted on an
`micropositioning stage for optical
`alignment. With the micromachining technology, the optical
`system can be miniaturized and batch fabricated on a silicon
`substrate. Unlike the conventional systems, the optical ele-
`ments can be integrally fabricated on translation or rotation
`stages. Microactuators for moving the optical elements can
`also be fabricated by the same micromachining process.
`FS-MOB offers many advantages over conventional op-
`tical systems. First, the FS-MOB is made by a VLSI-like
`batch-fabrication process, which can significantly reduce
`the system cost. Conventional optical systems often need
`custom design and expensive assembly. Second, the optical
`system can be miniaturized by the FS-MOB technology.
`Many optical systems are limited by the sizes of mi-
`cropositioning stages and optomechanical structures. Using
`the MEMS structures and actuators in FS-MOB, the size
`and weight of the optical systems can be greatly reduced.
`Third,
`the entire optical system can be monolithically
`integrated on a single chip. The use of out-of-plane optical
`elements allows multiple elements to be cascaded along the
`optical axes on the same substrate. Therefore, single-chip
`microoptical systems can be achieved. Fourth, the optics in
`FS-MOB can be “prealigned.” Since all the microoptical
`
`Fig. 3. The SEM micrograph of high-aspect-ratio structures cre-
`ated by DRIE [12].
`
`Fig. 4. Schematic of
`the
`micromachined cantilevers.
`
`fabrication process
`
`for
`
`surface-
`
`with complementary metal–oxide–semiconductor circuits
`[21], [22]. Surface micromachining using other combina-
`tions of structural/sacrificial materials has also been demon-
`strated. For example, aluminum structure material and
`organic sacrificial material are used in Texas Instruments’
`digital micromirror devices (DMD’s) [23], which will be
`discussed later.
`
`C. Microhinges
`Out-of-plane structures with high aspect ratios are often
`needed for free-space optical systems. Though they can
`be obtained by anisotropic etching or deep dry etching,
`it is difficult to pattern their side walls, as often required
`for free-space optical elements. In 1991, Pister and his
`coworkers proposed using the microhinges to fabricate a
`variety of three-dimensional structures using the surface-
`micromachining process [24]. This allows the surface-
`micromachined polysilicon plates to be patterned by pho-
`tolithography and then folded into three-dimensional struc-
`tures. The schematic cross section and the fabrication
`processes of the microhinge are illustrated in Fig. 5. It con-
`sists of a hinge pin and a confining staple. After selective
`etching of the sacrificial silicon dioxide, the polysilicon
`plate connected to the hinge pin is free to rotate out
`
`1836
`
`PROCEEDINGS OF THE IEEE, VOL. 85, NO. 11, NOVEMBER 1997
`
`Exhibit 1045, Page 4
`
`

`
`Fig. 5. Schematic of the fabrication process for surface-micromachined microhinges (after [24]).
`
`Fig. 6. Schematic illustrating the concept of FS-MOB. Microoptical elements, micropositioners,
`and microactuators are monolithically integrated on silicon substrate by surface micromachining.
`
`elements and the optomechanical structures are made at
`the same time by the photolithographic processes, they can
`be aligned during layout of the photomasks. The accuracy
`of the alignment is limited by the misalignment error of
`photolithography and the mechanical clearance between the
`movable structures, which is on the order of a micrometer.
`0.1 m) can be achieved by
`Fine optical alignment (
`on-chip microactuators. The integrated microactuators also
`allow dynamic tracking of alignment. The optical pre-
`alignment enables the “interconnections” between optical
`elements to be fabricated at the same time as the optical
`elements. This allows a functional optical system to be
`monolithically integrated and aligned on a single chip.
`This is similar to the concept of VLSI,
`in which the
`interconnections between transistors are fabricated mono-
`lithically. Combining the large number of transistors and the
`monolithic interconnections, highly functional electronic
`systems such as microprocessors have been produced.
`FS-MOB represents a paradigm shift for the optical
`systems. The conventional optical system is assembly in-
`
`tensive. The optical elements are made separately and then
`assembled into optical systems. FS-MOB resembles more
`the VLSI systems: it is design intensive, and the same
`standard process is used to fabricate different functional
`circuits. It is based on batch processing techniques and is
`more suitable for mass production. In the following, we
`describe the basic building blocks of FS-MOB.
`
`A. Diffractive Microlenses
`Diffractive microlenses are very attractive for integrating
`with FS-MOB because:
`
`1) their focal length can be precisely defined by pho-
`tolithography;
`2) microlenses with a wide range of numerical apertures
`(F/0.3–F/5) can be defined;
`3) microlenses with diameters as small as a few tens of
`micrometers can be made;
`4) their thickness is on the order of an optical wave-
`length [29]–[31].
`
`WU: MICROMACHINING
`
`1837
`
`Exhibit 1045, Page 5
`
`

`
`(a)
`
`(b)
`Fig. 7.
`(a) Schematic of a Fresnel zone plate. (b) Schematic of
`a multiple-level binary microlens.
`
`Fig. 9. The assembly process for the three-dimensional mi-
`cro-Fresnel lens.
`
`Fig. 8. Schematic of the out-of-plane micro-Fresnel lens fabri-
`cated on a hinged polysilicon plate.
`
`The thin construction is particularly suitable for the surface-
`micromachining process because the thicknesses of the
`structural layers are only on the order of 1 m. Fig. 7 shows
`the schematic diagram of two diffractive microlenses: (a)
`Fresnel lens and (b) multiple-level binary microlens. The
`binary-amplitude Fresnel zone plate has alternating trans-
`mission and blocking zones. Though it is very easy to
`fabricate, however, its efficiency (diffraction efficiency into
`the first-order beam) is limited to 10%. The efficiency of a
`binary microlens with
`phase levels is [29]
`
`The efficiency increases with the number of phase levels at
`the expense of more complicated fabrication processes. For
`41% for
`,
`81% for
`, and
`example,
`99% for
`. Fabrication of binary microlenses on
`various substrates has already been demonstrated [29]–[31].
`These microlenses are usually confined to the surface of
`the substrates.
`Microlenses with optical axes parallel to the substrate
`are necessary for single-chip microoptical systems. This
`is achieved by combining the conventional microoptics
`technology with the surface-micromachined microhinges.
`As shown in Fig. 8, the micro-Fresnel lens is fabricated on
`
`Fig. 10. The SEM of the out-of-plane micro-Fresnel lens. The
`lens has a diameter of 280 m, a focal length of 500 m, and an
`optical axis of 254 m above the silicon substrate.
`
`a hinged polysilicon plate [32], [33]. Since the polysilicon
`plate lies on the surface of the substrate before assembly,
`conventional planar fabrication technology can be used to
`define the patterns of the Fresnel lens. After releasing etch,
`the micro-Fresnel lens plate is rotated out of the substrate
`plane and becomes perpendicular to the substrate. The
`assembly process is illustrated in Fig. 9. To enhance the
`mechanical strength and to more precisely define the angle
`of the microlens, a pair of side support plates is added. The
`side support plates are also made on polysilicon using the
`microhinge technology. They have a V-shaped opening at
`the top, followed by a long, narrow groove in the center.
`When they are folded onto the microlens plate, the 2- m-
`thick microlens plate is firmly locked into the 2.5- m-wide
`groove. As will be shown later, the side support plates play
`a critical role in the robustness of such three-dimensional
`microstructures. The maximum angular variation of the
`microlens depends on the geometry of the side support
`plates. Angular variation of less than 0.1 is achievable.
`Fig. 10 shows the SEM micrograph of a micro-Fresnel lens
`with a diameter of 280 m, a focal length of 500 m, and
`an optical axis of 254 m above the silicon surface [33].
`For simplicity, only a binary-amplitude Fresnel zone plate
`
`1838
`
`PROCEEDINGS OF THE IEEE, VOL. 85, NO. 11, NOVEMBER 1997
`
`Exhibit 1045, Page 6
`
`

`
`Fig. 12. Schematic of an out-of-plane refractive microlens.
`
`diffractive optical elements. Their focal length is indepen-
`dent of the optical wavelength (except a weak dependence
`due to the dispersion of the lens material). It is also easier
`to make fast refractive microlenses without tight critical
`dimension control. Refractive microlenses do not suffer
`from diffraction loss, and it is easier to make a high-
`efficiency lens at
`low cost. In contrast, high-efficiency
`diffractive microlenses need multiple critical lithography
`steps.
`refractive microlens array
`Fabrication of a planar
`on semiconductor and dielectric substrates has been
`demonstrated using photoresist/polyimide reflow techniques
`[35]–[38]. The lens pattern can also be transferred to
`substrate through reactive ion etching or ion milling. Here,
`we combine the planar refractive microlens fabrication with
`the surface-micromachining process to create low-cost,
`high-quality out-of-plane refractive microlenses. Fig. 12
`shows the schematic drawing of the out-of-plane refractive
`spherical lens [39]. One potential issue for such microlenses
`is the scattering loss of the supporting polysilicon plate.
`The scattering loss is related to the surface roughness by
`[13], [40]
`
`is the
`is the percentage of the scattering loss,
`where
`root-mean-square roughness of the polysilicon surface,
`is the incident angle, and
`is the optical wavelength.
`The surface roughness of as-grown polysilicon is on the
`order of 45 nm (the exact value depends on the deposition
`condition), and the corresponding scattering loss is 17% per
`interface at 1300-nm wavelength. There are two approaches
`to reducing the scattering loss. The first is to smooth the
`surface by chemical-mechanical planarization (CMP). The
`surface roughness can be reduced to 1.7 nm, which will
`improve the scattering loss to 0.02%. Both surfaces need
`to be polished for a transmission device, however, and it is
`difficult to polish the bottom side of the polysilicon plate.
`Alternatively, we can remove the polysilicon material at
`the center of the lens, which forms an aperture for the
`
`Fig. 11. The intensity profile and contour plot of the optical beam
`emitted from an optical fiber and collimated by the micro-Fresnel
`lens.
`
`is demonstrated. Lenses as large as 650 m and as tall as
`1.4 mm also have been demonstrated [32].
`The lens exhibits very good optical performance. The
`output beam from a single-mode fiber is successfully
`collimated by the lens [32]. Fig. 11 shows the intensity
`profile of
`the collimated beam and the contour plot
`of the intensity distribution. Very good agreement with
`Gaussian shape is obtained. The intensity full-width-at-
`half-maximum (FWHM) divergence angle of the collimated
`beam has been reduced from 5.0 to 0.33 . The diffraction
`efficiency of the micro-Fresnel
`lens was measured to
`be 8.6% using the method described by Rastani et al.
`[30]. This is in agreement with the theoretical limit of
`the binary-amplitude Fresnel zone plate. As mentioned
`earlier, efficiency greater
`than 80% can be achieved
`by multilevel Fresnel
`lenses at
`the expense of more
`complicated fabrication processes. Another potential issue
`of fabricating diffractive optical elements on the surface-
`micromachined polysilicon plates is the surface roughness.
`The plates might need to be smoothened by chemical-
`mechanical planarization, which can reduce the surface
`roughness to 17 ˚A [34]. Another alternative to achieve
`high efficiency is to use a refractive lens, as described in
`the next section.
`
`B. Refractive Microlenses
`The refractive microlenses are complementary to the
`diffractive microlenses. They offer several advantages over
`
`WU: MICROMACHINING
`
`1839
`
`Exhibit 1045, Page 7
`
`

`
`Fig. 13. The fabrication process for the three-dimensional refrac-
`tive microlens. The microlens is fabricated on a hinged polysilicon
`plate by the reflow technique.
`
`Fig. 14. The surface profile of the microlens. The lens is 30 m
`thick at center and 300 m in diameter.
`
`microlens. The scattering as well as reflection loss of the
`polysilicon plate are thus eliminated. We have adopted the
`second approach here.
`The fabrication process is illustrated in Fig. 13 [39]. First,
`a lens mount is fabricated using the surface-micromachining
`technique. This process is similar to that of the micro-
`Fresnel lens except that an aperture is etched in the center
`of the polysilicon plate. Before releasing, a 20- m-thick
`AZ 4620 photoresist cylinder is deposited on the hinged
`polysilicon plate. The lens is realized by heating the pho-
`toresist cylinder to 200 C, causing the photoresist to reflow
`into a spherical shape. Fig. 14 shows the surface profile
`of the reflowed photoresist. The lens is 30 m thick at
`center, and the maximum deviation from spherical is less
`than 0.5 m to within 5 m of the edges. The resulting
`lens has a diameter of 300 m and a focal
`length of
`670 m. Refractive microlenses with an F-number ranging
`from one to five and diameters from 30 to more than 500
`m can be made by this technique. The microlens can be
`made vertical to the substrate by rotating the supporting
`polysilicon plate after the reflow processes. Fig. 15 shows
`the SEM micrographs of both the front and back sides of the
`microlens. The back-side view clearly shows the aperture of
`the microlens. The optical loss of the microlens is measured
`
`Fig. 15. SEM micrographs of the front and back sides of the
`three-dimensional refractive microlens. The aperture of the mi-
`crolens is clearly visible from the backside.
`
`to be 0.7 dB at 632-nm wavelength. Lower loss is expected
`at longer wavelengths. The optical axis of the microlenses
`is fixed to the same height as the rest of the FS-MOB by
`photolithography. Therefore, the refractive microlens can be
`aligned to other optical elements to within the fabrication
`tolerance.
`Very good optical performance has been achieved for the
`refractive microlens. The divergence angle of the light emit-
`ted from a single-mode fiber at 630 nm has been reduced
`from 3.3 to 0.18 after it is collimated by the refractive
`microlens. The intensity contour of the collimated beam is
`also very symmetric, as shown in Fig. 16, indicating the
`high quality of the spherical microlens.
`
`C. Micropositioners
`As described in [17] and [18], various types of movable
`structures can be made by the surface-micromachining
`
`1840
`
`PROCEEDINGS OF THE IEEE, VOL. 85, NO. 11, NOVEMBER 1997
`
`Exhibit 1045, Page 8
`
`

`
`Fig. 16. The intensity contour of the optical beam collimated by
`the spherical microlens.
`
`Fig. 17. Surface-micromachined translation stage with (a) flexure
`springs and (b) sliding guides.
`
`technique. A linear translation stage and rotation stage can
`be built in a similar way on FS-MOB. The microlenses
`and other out-of-plane microoptical elements can be inte-
`grated with translation and rotation stages by attaching the
`microhinges to a movable polysilicon plate. The movable
`stage consists of a fully released polysilicon plate that is
`attached or confined to the substrate by either 1) flexure
`springs or 2) sliding guides. Fig. 17 illustrates these two
`types of micropositioners. The spring-attached stage is more
`accurate; however, it has limited travel distance. The spring
`restoring force also needs to be constantly balanced by
`the actuators. The sliding type of stage has a long travel
`distance. Since there is no restoring force,
`the friction
`between the stage and the substrate or the sliding guides can
`hold the stage in place without any power consumption. On
`the other hand, the finite clearance between the stage and
`the sliding guides may cause some statistical variation and
`hysteresis. These two types of stages are complementary.
`To integrate the microoptical elements on translation
`stages, the microoptics plate now needs to be constructed
`on the second polysilicon layer and connected to the
`microhinges defined on the first polysilicon layer. The bases
`of the spring latched and staples are now connected to the
`movable plate on the first polysilicon layer instead of the
`substrate. Fig. 18 shows the SEM micrograph of a three-
`dimensional micrograting integrated on a rotation stage
`[41]. The stage in the SEM has been rotated by 20 . The
`rotary grating has been employed to build an integrated-
`wavelength meter for a single-wavelength laser [42] and
`microspectrometer [43].
`
`D. Microactuators
`Perhaps the most interesting feature of micromachined
`optics is the ability to integrate microactuators with the
`
`Fig. 18. SEM micrograph of a three-dimensional micrograting
`integrated on a rotation stage.
`
`optical elements. Many new surface-micromachined mi-
`croactuators have been proposed and demonstrated since
`the invention of micromotors. It is beyond the scope of this
`paper to have a complete review of these microactuators.
`Here, we will describe only a selected set of microactuators
`that are particularly useful for optical applications. Fig. 19
`shows the schematic drawing of these actuators.
`1) Comb Drive Actuator [44], [45]: The comb drive is
`actuated by electrostatic force between a pair of movable
`combs and fixed combs, as shown in Fig. 19(a). When
`a voltage is applied, the movable comb is attracted to
`the stationary comb. The position of the movable comb
`can thus be controlled by voltage. This is perhaps by far
`the most popular surface-micromachined actuator. Typical
`surface-micromachined polysilicon comb drive actuators
`can produce micronewtons of force and a few micrometers
`of displacement. The comb drive actuators have been
`employed to drive optical choppers [46] and bar-code
`scanners [47]. Another electrostatic actuator is gap-closing
`actuators [48], [49]. They have larger force but smaller
`displacement.
`2) Linear Microvibromotor [50], [51]: Impact actuation
`can be used to obtain relatively large motion from small-
`displacement
`resonant structures such as comb drive
`actuators. The linear microvibromotor shown in Fig. 19(b)
`resonant comb drives,
`two
`45
`consists of two 45
`resonant comb drives, and a slider. Upon each impact from
`the comb drives, the slider moves by an average step size
`of 0.27 m [51]. The standard deviation of the step size is
`relatively large (0.17 m) due to the impact actuation and
`wobble of the slider. The position and speed of the slider
`can be controlled by adjusting the number of impacts. A
`maximum speed of 1 mm/s has been achieved. The linear
`microvibromotor has been used to actuate a slide-tilt mirror
`for alignment for optical beams for fiber coupling [52].
`3) Stepper Motor: In addition to impact actuation, the
`long travel distance of the slider can also be achieved by the
`“hold-and-pull” actuation [53], [54]. A large-force, small-
`displacement actuator, such as the gap-closing actuator, first
`
`WU: MICROMACHINING
`
`1841
`
`Exhibit 1045, Page 9
`
`

`
`(a)
`
`(b)
`
`(e)
`(d)
`Fig. 19. Schematic of microactuators commonly used in surface-microm