**1. Introduction**

290 Microelectromechanical Systems and Devices

Rebeiz, G. M. (2003). *RF MEMS : Theory Design and Technology,* John Wiley and Sons Ltd,

Varadan, V. K.; Vinoy, K. J. & Jose, K. A. (2003). *RF MEMS and Their Applications*, John

Wiley and Sons Ltd, ISBN 0-470-84308-X, New York USA

ISBN 0-471-20169-3, New York USA

In 1989, a group of scientists and engineers in Salt Lake City started a workshop called Micro-Tele-Operated Robotics Workshop. There, the acronym for Microelectromechanical systems (MEMS) was officially adopted. However, MEMS technology has already had a head start since at least 7 years ago from the classic work published by Petersen in 1982 [1]. Twenty years later, MEMS technology has started major novel innovations in several scientific fields and created highly promising market potential. In 2003, the most conservative market studies predict a world MEMS market in excess of \$8 billion.

Optics and photonics are among these research fields impacted by MEMS techniques. Optical MEMS has created a new fabrication paradigm for optical devices and systems. These micro optical devices and systems are inherently suited for cost effective wafer scale manufacturing as the processes are derived from the semiconductor industry. The ability to steer or direct light is one of the key requirements in optical MEMS. In the past two decades since Petersen published his silicon scanner [1], the field of optical MEMS has experienced explosive growth [2,3]. In the 80's and early 90's, displays were the main driving force for the development of micromirror arrays. Portable digital displays are commonplaces and head mount displays are now commercially available. In the past decade, telecommunications have become the market driver for Optical MEMS. The demand for routing internet traffic through fiber optic networks pushes the development of both digital and scanning micromirror systems for large port-count all-optical switches with the ability to directly manipulate an optical signal, Optical MEMS systems eliminate unnecessary optical-electrical-optical (O-E-O) conversions. In the biomedical arena, micro-optical scanners promise low-cost endoscopic three-dimensional imaging systems for *in vivo* diagnostics.

This chapter summarizes the state of the art of Optical MEMS technology by describing basic fabrication processes to derive with actuation mechanisms and select examples of devices that are either commercially available, or show great promise of becoming products in the near term. The chapter is organized into the following sections: Section 2 describes the generic actuation mechanisms commonly used for MEMS devices. Section 3 discusses the applications.

Optical MEMS 293

**Movable electrode**

maximum. In addition to the staggered and angular vertical combdrive actuators, in-plane configuration vertical combdrive actuator is the third approach firstly demonstrated by Harald [8] at Fraunhofer IPMS Dresden, Germany. The in-plane configuration vertical combdrive (Figure 2-4) means that the moving combdrive is on the same plane of the static combdrive, which is based on the parametric resonant working principle. The resonantly excited actuator has several advantages as below: (1) it can achieve very large tilting angle with low driving voltage; (2) only one-time DRIE is needed for combdrive and gimbal frame structure during micro-machining which eliminates wafer bonding or accurate alignment. However, the vertical combdrive actuators with in-plane configuration only work at resonant mode. Recently, researchers at IPMS also have made effort to realize quasi-static

rotation mode with vertical out-of-plane combdrive using leverage structures [9].

**Movable Comb**

**+** 

**-** 

Fig. 2.1. Schematic of a parallel-plate actuator

**Fixed electrode**

**Fixed Comb**

Fig. 2.2. Schematic of a vertical combdrive actuator

#### **2. Mechanisms of actuations**

Since the actuators are important engines for Optical MEMS, we will introduce different kinds of actuation mechanisms first and explain the working principles in detail. Furthermore, for more specific applications, there are many variants or requirements based on these basic actuation mechanisms.

#### **2.1 Electrostatic actuation**

Electrostatic MEMS devices work at different motion modes. Here we mainly introduce the MEMS devices with torsional rotation. Electrostatic MEMS devices with torsional rotation can be described as follows: when voltage is applied between the movable and the fixed electrodes, the moving part rotates about the torsion axis until the restoring torque and the electrostatic torque are equal. The torques can be expressed as:

$$T\_e(\theta) = \frac{V^2}{2} \frac{\partial \mathbf{C}}{\partial \theta} \tag{1}$$

$$T\_r(\theta) = k\theta \tag{2}$$

where V is the applied voltage across the fixed and movable electrodes, C is the capacitance of the actuator, is the rotation angle, and k is the spring constant. The capacitance is determined by the area of the electrode overlap and the gap between the electrodes. For simple parallel plate geometry, the capacitance can be expressed by

$$\mathbf{C} = \frac{\mathbf{c}\_0 \mathbf{A}}{\mathbf{g}} \tag{3}$$

where 0 is the permittivity of free space, A is the area of electrode overlap, and g is the gap between fixed and moving electrodes.

There are two major types of electrostatic actuators. The first is based on parallel-plate capacitance, and the other is based on comb-drive capacitance. For the parallel-plate type devices (Figure 2-1), the area of the electrode overlap is essentially the area of the fixed electrode. The gap for the parallel-plate actuator is a function of the rotation angle and there is a tradeoff as the initial gap spacing needs to be large enough to accommodate the scan angle, but small enough for reasonable actuation voltage. The stable scan range is further limited by a pull-in phenomenon to 34-40% of the maximum mechanical scan angle [4,5].

Another type of actuator is based on the vertical combdrive. The vertical combdrive offers several advantages: (1) the structure and the actuator are decoupled; (2) the gap between the interdigitated fingers of the combdrive is typically quite small, on the order of a couple of microns [6,7]. Large rotation angle and low actuation voltage can be achieved simultaneously. In the combdrive, the gap is constant and the area of the electrode overlap is a function of the rotation angle. Typically, there are three approaches to realize the vertical combdrive actuators. The first and popular one is the staggered vertical combdrive actuators [46]. Figure 2-2 shows the schematic of a vertical combdrive actuator. The second type is the angular vertical combdrive actuators [8,58,59] (Figure 2-3). Both staggered and angular vertical combdrive actuators can work at either resonant tilting mode or quasi-static (or called DC tilting) mode. At quasi-static mode, the maximum rotation angle usually is limited because it is typically the point where the overlap area of electrodes reaches the

Since the actuators are important engines for Optical MEMS, we will introduce different kinds of actuation mechanisms first and explain the working principles in detail. Furthermore, for more specific applications, there are many variants or requirements based

Electrostatic MEMS devices work at different motion modes. Here we mainly introduce the MEMS devices with torsional rotation. Electrostatic MEMS devices with torsional rotation can be described as follows: when voltage is applied between the movable and the fixed electrodes, the moving part rotates about the torsion axis until the restoring torque and the

2

(1)

(3)

(2)

( ) <sup>2</sup> *<sup>e</sup> V C <sup>T</sup>* 

> ( ) *T k <sup>r</sup>*

where V is the applied voltage across the fixed and movable electrodes, C is the capacitance of the actuator, is the rotation angle, and k is the spring constant. The capacitance is determined by the area of the electrode overlap and the gap between the electrodes. For

> *g*

where 0 is the permittivity of free space, A is the area of electrode overlap, and g is the gap

There are two major types of electrostatic actuators. The first is based on parallel-plate capacitance, and the other is based on comb-drive capacitance. For the parallel-plate type devices (Figure 2-1), the area of the electrode overlap is essentially the area of the fixed electrode. The gap for the parallel-plate actuator is a function of the rotation angle and there is a tradeoff as the initial gap spacing needs to be large enough to accommodate the scan angle, but small enough for reasonable actuation voltage. The stable scan range is further limited by a pull-in phenomenon to 34-40% of the maximum mechanical scan angle [4,5]. Another type of actuator is based on the vertical combdrive. The vertical combdrive offers several advantages: (1) the structure and the actuator are decoupled; (2) the gap between the interdigitated fingers of the combdrive is typically quite small, on the order of a couple of microns [6,7]. Large rotation angle and low actuation voltage can be achieved simultaneously. In the combdrive, the gap is constant and the area of the electrode overlap is a function of the rotation angle. Typically, there are three approaches to realize the vertical combdrive actuators. The first and popular one is the staggered vertical combdrive actuators [46]. Figure 2-2 shows the schematic of a vertical combdrive actuator. The second type is the angular vertical combdrive actuators [8,58,59] (Figure 2-3). Both staggered and angular vertical combdrive actuators can work at either resonant tilting mode or quasi-static (or called DC tilting) mode. At quasi-static mode, the maximum rotation angle usually is limited because it is typically the point where the overlap area of electrodes reaches the

**2. Mechanisms of actuations** 

on these basic actuation mechanisms.

electrostatic torque are equal. The torques can be expressed as:

simple parallel plate geometry, the capacitance can be expressed by

<sup>0</sup>*<sup>A</sup> <sup>C</sup>*

between fixed and moving electrodes.

**2.1 Electrostatic actuation** 

Fig. 2.1. Schematic of a parallel-plate actuator

maximum. In addition to the staggered and angular vertical combdrive actuators, in-plane configuration vertical combdrive actuator is the third approach firstly demonstrated by Harald [8] at Fraunhofer IPMS Dresden, Germany. The in-plane configuration vertical combdrive (Figure 2-4) means that the moving combdrive is on the same plane of the static combdrive, which is based on the parametric resonant working principle. The resonantly excited actuator has several advantages as below: (1) it can achieve very large tilting angle with low driving voltage; (2) only one-time DRIE is needed for combdrive and gimbal frame structure during micro-machining which eliminates wafer bonding or accurate alignment. However, the vertical combdrive actuators with in-plane configuration only work at resonant mode. Recently, researchers at IPMS also have made effort to realize quasi-static rotation mode with vertical out-of-plane combdrive using leverage structures [9].

Fig. 2.2. Schematic of a vertical combdrive actuator

Optical MEMS 295

Magnetic field can be induced by electrical current. This current-induced magnetic field can generate the force exerted on the moving magnetic material [10]. While the moving structure is not made of magnetic material, electromagnetic coils can be integrated on the movable part, making it quasi-magnetic by current injection [11]. Figure 2-5 shows an example of the electromagnetic scanner that is being used in table-top confocal microscopes. Researchers at the University of Michigan have demonstrated a miniature magnetic 3D scanner for optical

Fig. 2.5. (a) Schematic and (b) photograph of packaged electromagnetic 1D scanner in [11]

alignment [12] shown in Figure 2-6.

(Picture courtesy of Hiroshi Miyajima).

Fig. 2.3. Schematic of an angular combdrive actuator

Fig. 2.4. Schematic of an in-plane configuration combdrive actuator

#### **2.2 Magnetic actuation**

Magnetic actuation is practical when the structural dimensions are on the millimeter scale since the magnetic torque (generated by the magnetic device interacting with an external magnetic field) scales with volume for permanent magnetic materials and with total coil area for electromagnets. For an analysis of magnetic torque see Judy and Muller [10]. The overall system size must accommodate the magnets (permanent or electric coils) used to generate the external magnetic field. Therefore, the motivations for this type of scanner are usually cost reduction through batch fabrication and lower power consumption rather than miniaturization. In addition, magnetic actuation also has the advantage of operating in liquid environment.

**Fixed Comb**

**Movable Comb**

**2.2 Magnetic actuation** 

liquid environment.

Fig. 2.3. Schematic of an angular combdrive actuator

Fig. 2.4. Schematic of an in-plane configuration combdrive actuator

Magnetic actuation is practical when the structural dimensions are on the millimeter scale since the magnetic torque (generated by the magnetic device interacting with an external magnetic field) scales with volume for permanent magnetic materials and with total coil area for electromagnets. For an analysis of magnetic torque see Judy and Muller [10]. The overall system size must accommodate the magnets (permanent or electric coils) used to generate the external magnetic field. Therefore, the motivations for this type of scanner are usually cost reduction through batch fabrication and lower power consumption rather than miniaturization. In addition, magnetic actuation also has the advantage of operating in Magnetic field can be induced by electrical current. This current-induced magnetic field can generate the force exerted on the moving magnetic material [10]. While the moving structure is not made of magnetic material, electromagnetic coils can be integrated on the movable part, making it quasi-magnetic by current injection [11]. Figure 2-5 shows an example of the electromagnetic scanner that is being used in table-top confocal microscopes. Researchers at the University of Michigan have demonstrated a miniature magnetic 3D scanner for optical alignment [12] shown in Figure 2-6.

Fig. 2.5. (a) Schematic and (b) photograph of packaged electromagnetic 1D scanner in [11] (Picture courtesy of Hiroshi Miyajima).

Optical MEMS 297

for both pure z-axis displacement actuator and large angle tilting mirror at low speed

Fig. 2.7. Schematic show actuation principle of (a) a single bimorph and (b) SEM of a

Fig. 2.8. Electrothermal bimorph microactuator with initial elevation in [16] (Picture

Piezoelectric material deforms when electric field is applied across the structure. This property can be used as the driving mechanism in MEMS and NEMS. Different kinds of piezoelectric scanners and actuators have been demonstrated [19,20,21,22]. Using piezoelectric thin films, researchers at the University of Michigan have developed a novel thin-film lead zirconate titanate (PZT) based large displacement (around 120um) vertical translational microactuator [19]. The microactuator consists of four compound bendup/bend-down unimorphs to generate z-axis motion of a moving stage. Figure 2-9 and Figure 2-10 show an example of the thin-film PZT based microactuator with large displacement. The large displacement within small footprint and high bandwidth (fast response) of the actuators at low-voltage and low-power levels should make them useful to

bimorph-based micromirror [17] (Picture courtesy of Huikai Xie).

application.

courtesy of Huikai Xie)

**2.4 Piezoelectric actuation mechanisms** 

a variety of optical applications, like endoscopic microscopy.

Fig. 2.6. SEM of the miniature electromagnetic 3D scanner in [12] (Picture courtesy of Il-Joo Cho).

#### **2.3 Thermal actuation**

Thermal actuation utilizes the mismatch between thermal expansion coefficients of materials, which yields structural stress after temperature change. The structure deforms due to this built-in stress. The major advantage of thermal actuation is its ability to generate large deflection. Electrical current injection is one of the common mechanisms used for heating up the structure. However, temperature control and power consumption are issues for this type of actuators. Several new electro-thermal micromirror and actuators have been reported [13,14,15,16,17,18] by researchers from University of Florida. For example, shown in Figure 2-7, the electrothermal bimorph structure in [17] is based on two thin-film materials with different coefficients of thermal expansion(CTE), like Aluminum (high CTE) and SiO2 (low CTE). The heater material can be polysilicon or Pt. The induced thermal bending arc angle (*<sup>T</sup>* ) of the bimorph beam can be expressed as below

$$
\theta\_T = \theta\_0 - \theta\_1 = \beta \cdot \frac{l\_b}{t\_b} \cdot A\alpha \cdot \overline{A\Gamma}
$$

Where 0 and 1 are the arc angles before and after the temperature increase, *bt* and *<sup>b</sup> l* are the bimorph thickness and length, is the CTE difference of the two bimorph materials, is the actuation coefficient related to the thickness and biaxial Young's Modulus ratios of the two materials, and *T* is the average temperature change on the bimorph.

The electrothermal microactuator in [16], shown in Figure 2-8, using three-bimorph actuation mechanism can achieve large lateral-shift-free (LSF) piston motion at low driving voltage (5.3V) with thermal response time around 25ms. The design is potentially suitable

Thermal actuation utilizes the mismatch between thermal expansion coefficients of materials, which yields structural stress after temperature change. The structure deforms due to this built-in stress. The major advantage of thermal actuation is its ability to generate large deflection. Electrical current injection is one of the common mechanisms used for heating up the structure. However, temperature control and power consumption are issues for this type of actuators. Several new electro-thermal micromirror and actuators have been reported [13,14,15,16,17,18] by researchers from University of Florida. For example, shown in Figure 2-7, the electrothermal bimorph structure in [17] is based on two thin-film materials with different coefficients of thermal expansion(CTE), like Aluminum (high CTE) and SiO2 (low CTE). The heater material can be polysilicon or Pt. The induced thermal

*<sup>T</sup>* ) of the bimorph beam can be expressed as below

*b*

*<sup>l</sup> <sup>T</sup>*

 

1 are the arc angles before and after the temperature increase, *bt* and *<sup>b</sup>*

*T* is the average temperature change on the bimorph.

is the CTE difference of the two bimorph materials,

*l* are

*b*

*t*

is the actuation coefficient related to the thickness and biaxial Young's Modulus ratios of

The electrothermal microactuator in [16], shown in Figure 2-8, using three-bimorph actuation mechanism can achieve large lateral-shift-free (LSF) piston motion at low driving voltage (5.3V) with thermal response time around 25ms. The design is potentially suitable

0 1

*T*

 

Fig. 2.6. SEM of the miniature electromagnetic 3D scanner in [12]

(Picture courtesy of Il-Joo Cho).

**2.3 Thermal actuation** 

bending arc angle (

0 and 

the two materials, and

Where

the bimorph thickness and length,

for both pure z-axis displacement actuator and large angle tilting mirror at low speed application.

Fig. 2.7. Schematic show actuation principle of (a) a single bimorph and (b) SEM of a bimorph-based micromirror [17] (Picture courtesy of Huikai Xie).

Fig. 2.8. Electrothermal bimorph microactuator with initial elevation in [16] (Picture courtesy of Huikai Xie)

#### **2.4 Piezoelectric actuation mechanisms**

Piezoelectric material deforms when electric field is applied across the structure. This property can be used as the driving mechanism in MEMS and NEMS. Different kinds of piezoelectric scanners and actuators have been demonstrated [19,20,21,22]. Using piezoelectric thin films, researchers at the University of Michigan have developed a novel thin-film lead zirconate titanate (PZT) based large displacement (around 120um) vertical translational microactuator [19]. The microactuator consists of four compound bendup/bend-down unimorphs to generate z-axis motion of a moving stage. Figure 2-9 and Figure 2-10 show an example of the thin-film PZT based microactuator with large displacement. The large displacement within small footprint and high bandwidth (fast response) of the actuators at low-voltage and low-power levels should make them useful to a variety of optical applications, like endoscopic microscopy.

Optical MEMS 299

provides a transduction capability that is used for both actuation and sensing devices. It has the advantage of remote actuation by magnetic fields. 2D optical scanners using

The Digital Micromirror Device (DMD) started in 1977 by Texas Instruments. The DMD technology has helped revolutionize projector systems by dramatically decreasing costs and increasing performance from traditional projector systems which are based on LCD technology. The research initially focused on deformable mirror device. Eventually DMD becomes the preferred device. TI uses Digital Light ProcessingTM (DLP) to denote optical

The DMD is a reflective spatial light modulator (SLM) which consists of millions of digitally actuated micromirrors. Each micromirror is controlled by underlying complementary metaloxide-semiconductor (CMOS) electronics, as shown in Figure 3-1. A DMD panel's micromirrors are mounted on tiny hinges that enable them to tilt either toward the light source (ON) or away from it (OFF) depending on the state of the static random access memory (SRAM) cell below each micromirror. The SRAM voltage is applied to the address electrodes, creating an electrostatic attraction to rotate the mirror to one side or the other. The details of operating principle, design, fabrication, and testing of DMD have been

Fig. 3.1. Schematic drawing of two DMD mirror with underlying structures (Picture

In projection systems, brightness and contrast are the two primary attributes that impact the quality of the projected image. The DMD has a light modulator efficiency in the range of 65%, and enables the contrast ratio ranging from 1000:1 to 2000:1. Because of the fast switching speed of the mirror, it enables the DLP to have a wide range of applications in video and data projectors, HDTVs, and digital cinema. Though, DMD was developed

magnetostrictive actuators have been reported [23].

**3.1.1 Texas Instruments' Digital Micromirror Device (DMD)** 

projection displays enabled by the DMD technologies [24].

discussed in [25] and will not be repeated here.

courtesy of TI. Reprinted from [24] with permission).

**3.1 Display, imaging, and microscopy** 

**3. Applications** 

Fig. 2.9. (a) Top-view, piezoelectrically actuated vertical stage, (b) Schematic cross-section of a vertical translational stage in [19] (Picture courtesy of Kenn Oldham).

Fig. 2.10. (a) Top view and oblique view of the corner of a prototype vertical actuator under 0 and 20 V. (b) Scanning-electron microscope (SEM) image of end of an actuation leg (full actuator inset), showing the XeF2 etching profile below the device layer

Aluminium nitride (AlN) is another promising potential piezoelectric material for future Optical MEMS applications [21,22]. One of the important advantages of AlN is that it is very suitable for full CMOS compatible MEMS processes. Currently, researchers from IPMS Germany are developing quasistatic deformable mirrors by actively coupling lateral strain in micro machined AlN based membranes.

### **2.5 Other actuation mechanisms**

Magnetostrictive materials transduce or convert magnetic energy to mechanical energy and vice versa. As a magnetostrictive material is magnetized, it strains; i.e., it exhibits a change in length per unit length. Conversely, if an external force produces a strain in a magnetostrictive material, the material's magnetic state will change. This bidirectional coupling between the magnetic and mechanical states of a magnetostrictive material

(a) (b)

a vertical translational stage in [19] (Picture courtesy of Kenn Oldham).

Fig. 2.9. (a) Top-view, piezoelectrically actuated vertical stage, (b) Schematic cross-section of

(a) (b)

actuator inset), showing the XeF2 etching profile below the device layer

in micro machined AlN based membranes.

**2.5 Other actuation mechanisms** 

Fig. 2.10. (a) Top view and oblique view of the corner of a prototype vertical actuator under 0 and 20 V. (b) Scanning-electron microscope (SEM) image of end of an actuation leg (full

Aluminium nitride (AlN) is another promising potential piezoelectric material for future Optical MEMS applications [21,22]. One of the important advantages of AlN is that it is very suitable for full CMOS compatible MEMS processes. Currently, researchers from IPMS Germany are developing quasistatic deformable mirrors by actively coupling lateral strain

Magnetostrictive materials transduce or convert magnetic energy to mechanical energy and vice versa. As a magnetostrictive material is magnetized, it strains; i.e., it exhibits a change in length per unit length. Conversely, if an external force produces a strain in a magnetostrictive material, the material's magnetic state will change. This bidirectional coupling between the magnetic and mechanical states of a magnetostrictive material provides a transduction capability that is used for both actuation and sensing devices. It has the advantage of remote actuation by magnetic fields. 2D optical scanners using magnetostrictive actuators have been reported [23].

#### **3. Applications**

#### **3.1 Display, imaging, and microscopy**

#### **3.1.1 Texas Instruments' Digital Micromirror Device (DMD)**

The Digital Micromirror Device (DMD) started in 1977 by Texas Instruments. The DMD technology has helped revolutionize projector systems by dramatically decreasing costs and increasing performance from traditional projector systems which are based on LCD technology. The research initially focused on deformable mirror device. Eventually DMD becomes the preferred device. TI uses Digital Light ProcessingTM (DLP) to denote optical projection displays enabled by the DMD technologies [24].

The DMD is a reflective spatial light modulator (SLM) which consists of millions of digitally actuated micromirrors. Each micromirror is controlled by underlying complementary metaloxide-semiconductor (CMOS) electronics, as shown in Figure 3-1. A DMD panel's micromirrors are mounted on tiny hinges that enable them to tilt either toward the light source (ON) or away from it (OFF) depending on the state of the static random access memory (SRAM) cell below each micromirror. The SRAM voltage is applied to the address electrodes, creating an electrostatic attraction to rotate the mirror to one side or the other. The details of operating principle, design, fabrication, and testing of DMD have been discussed in [25] and will not be repeated here.

Fig. 3.1. Schematic drawing of two DMD mirror with underlying structures (Picture courtesy of TI. Reprinted from [24] with permission).

In projection systems, brightness and contrast are the two primary attributes that impact the quality of the projected image. The DMD has a light modulator efficiency in the range of 65%, and enables the contrast ratio ranging from 1000:1 to 2000:1. Because of the fast switching speed of the mirror, it enables the DLP to have a wide range of applications in video and data projectors, HDTVs, and digital cinema. Though, DMD was developed

Optical MEMS 301

applying an electrical bias between the ribbons and the substrate. When all the ribbons are in the same plane, the GLV device acts like a mirror and incident light is reflected from their surfaces. When alternate ribbons are deflected, the angular direction in which incident light is steered from the GLV device is dictated by the spatial frequency of the diffraction grating formed by the ribbons. As this spatial frequency is determined by the photolithographic mask used to form the GLV device in the fabrication process, the departure angles can be very accurately controlled, which is useful for optical switching applications. The linear deflection of the GLV is quite small, with no physical contact between moving elements, thus avoiding wear and tear as well as stiction problems. There are also no physical boundaries between the pixel elements in the GLV array. When using as a spatial light modulator in imaging applications, this seamless characteristic provides a virtual 100% fill-

The ribbons are made of suspended silicon nitride films with aluminum coating to increase its reflectivity. The silicon nitride film is under tensile stress to make them optically flat. The tension also reduces the risk of stiction and increases their frequency response. The GLV materials are compatible with standard CMOS foundry processes. GLV can be made into one-dimensional or two-dimensional arrays for projection display applications. Today, the GLV technology is used in high resolution display, digital imaging systems and WDM

Retinal scanning display (RSD) uses a different approach from other microdisplays. Rather than a matrix array of individual modulator or source for each pixel as seen in liquid crystal display (LCD), organic light-emitting diodes (OLED), and DMD microdisplays, a RSD optimizes a low power light source to create a single pixel and scans this pixel with a single mirror to paint the displayed image directly onto the viewers' retina. With this technique, it offers high spatial and color resolution and very high luminance. There are several papers that provide an overview of the RSD and its applications [35,36]. This technology is developed by Microvision. The RSD systems typically employ two uniaxial scanners or one biaxial scanner. The combinations of two actuation mechanisms, electrostatic (for faster response) and electromagnetic (for larger force) actuations, were selected for a MEMS scanner [37]. Figure 3-3

The horizontal scanner (the inner mirror axis) is operated at resonance by using electrostatic actuation. The drive plates are located on the substrate below the MEMS mirror. The inner mirror axis has the resonant frequency of 19.5 kHz with the maximum mechanical scan excursion of 13.4 degrees. The vertical scanner (the outer mirror axis) is magnetically driven by means of permanent magnets within the package and cols with a 60 Hz linear ramp waveform. The magnets need to be positioned carefully and provide sufficient magnetic field to move the mirror to the desired angular deflection. The maximum mechanical scan excursion of 9.6 degrees was achieved on the outer mirror axis. The devices were bulk micromachined utilizing both wet and dry anisotropic etching and electroplating was used to form electromagnetic coils on the outer frame. These scanners must be stiff to remain flat and withstand the forces developed in resonant scanning mode. The dynamic mirror flatness of 0.05 microns rms was measured. The scanners also incorporate piezoresistive strain sensors on the torsion flexures for closed loop control. The scanners are designed to meet SVGA video standards that require 800 x 600 resolution. The design, fabrication, and

factor in the image.

telecommunications [34].

**3.1.3 Microvision retinal display and pico-projector** 

shows a schematic drawing of the MEMS scanner.

control details of this bi-axial scanner can be found in [37] and [38].

primarily for projection display applications, there are some interesting non-display applications. An emerging new DMD application is volumetric display, in which DMDs are used to render three-dimensional images that appear to float in space without the use of encumbering stereo glasses or headsets. It is realized by using 3 DMD's to create 3D images viewed without glasses or headsets [26,27]. DMD also has applications in maskless lithography and telecom. Traditionally, the patterns in lithographic applications, such as print settings, printed circuit board (PCB) and semiconductor manufacturing, have been provided via film or photomasks. However, it is desirable to directly write on the UVsensitive photoresist directly from digital files. DMD can be used as the spatial light modulator to generate the designed patterns [28]. For maskless lithography in sub-100nm semiconductor manufacturing, analog micromirror arrays with either tilting or piston motions are needed. Smaller mirror size is also desired.

DMD also has interesting applications in microscopy and spectroscopy. In microscopy application, DMD is used as a spatial modulator of the incident or collected light rays. It replaces the aperture in conventional optical system. The DMD can shape or scan either the illumination or collection aperture of an optical microscope thus to provide a dynamic optical system that can switch between bright field, dark field and confocal microscopy [29,30,31,32]. In spectroscopy application, the DMD is used as an adaptive slit selectively routing the wavelength of interest to a detector. It can also chop the light reaching the detector to improve detection sensitivity [33].

## **3.1.2 GLV display**

The schematic of the Grating Light ValveTM (GLVTM) shown in Figure 3-2 is a diffractive spatial light modulator [34]. The GLV device switches and modulates light intensities via diffraction rather than by reflection. Distinct advantages of GLV include high speed modulation, fine gray-scale attenuation, and scalability to small pixel dimensions.

Fig. 3.2. Cross-section of the GLV device showing the specular and diffraction states (after [34]).

The GLV device is built on a silicon wafer and is comprised of many parallel micro-ribbons that are suspended over an air gap above the substrate. Alternative rows of ribbons can be pulled down approximately ¼ wavelength to create diffraction effects on incident light by

primarily for projection display applications, there are some interesting non-display applications. An emerging new DMD application is volumetric display, in which DMDs are used to render three-dimensional images that appear to float in space without the use of encumbering stereo glasses or headsets. It is realized by using 3 DMD's to create 3D images viewed without glasses or headsets [26,27]. DMD also has applications in maskless lithography and telecom. Traditionally, the patterns in lithographic applications, such as print settings, printed circuit board (PCB) and semiconductor manufacturing, have been provided via film or photomasks. However, it is desirable to directly write on the UVsensitive photoresist directly from digital files. DMD can be used as the spatial light modulator to generate the designed patterns [28]. For maskless lithography in sub-100nm semiconductor manufacturing, analog micromirror arrays with either tilting or piston

DMD also has interesting applications in microscopy and spectroscopy. In microscopy application, DMD is used as a spatial modulator of the incident or collected light rays. It replaces the aperture in conventional optical system. The DMD can shape or scan either the illumination or collection aperture of an optical microscope thus to provide a dynamic optical system that can switch between bright field, dark field and confocal microscopy [29,30,31,32]. In spectroscopy application, the DMD is used as an adaptive slit selectively routing the wavelength of interest to a detector. It can also chop the light reaching the

The schematic of the Grating Light ValveTM (GLVTM) shown in Figure 3-2 is a diffractive spatial light modulator [34]. The GLV device switches and modulates light intensities via diffraction rather than by reflection. Distinct advantages of GLV include high speed

**Specular State Diffraction State**

modulation, fine gray-scale attenuation, and scalability to small pixel dimensions.

**Nitride with Al coating**

Fig. 3.2. Cross-section of the GLV device showing the specular and diffraction states

**Incident Light Reflected Incident Light Diffracted**

The GLV device is built on a silicon wafer and is comprised of many parallel micro-ribbons that are suspended over an air gap above the substrate. Alternative rows of ribbons can be pulled down approximately ¼ wavelength to create diffraction effects on incident light by

motions are needed. Smaller mirror size is also desired.

detector to improve detection sensitivity [33].

**3.1.2 GLV display** 

(after [34]).

applying an electrical bias between the ribbons and the substrate. When all the ribbons are in the same plane, the GLV device acts like a mirror and incident light is reflected from their surfaces. When alternate ribbons are deflected, the angular direction in which incident light is steered from the GLV device is dictated by the spatial frequency of the diffraction grating formed by the ribbons. As this spatial frequency is determined by the photolithographic mask used to form the GLV device in the fabrication process, the departure angles can be very accurately controlled, which is useful for optical switching applications. The linear deflection of the GLV is quite small, with no physical contact between moving elements, thus avoiding wear and tear as well as stiction problems. There are also no physical boundaries between the pixel elements in the GLV array. When using as a spatial light modulator in imaging applications, this seamless characteristic provides a virtual 100% fillfactor in the image.

The ribbons are made of suspended silicon nitride films with aluminum coating to increase its reflectivity. The silicon nitride film is under tensile stress to make them optically flat. The tension also reduces the risk of stiction and increases their frequency response. The GLV materials are compatible with standard CMOS foundry processes. GLV can be made into one-dimensional or two-dimensional arrays for projection display applications. Today, the GLV technology is used in high resolution display, digital imaging systems and WDM telecommunications [34].

#### **3.1.3 Microvision retinal display and pico-projector**

Retinal scanning display (RSD) uses a different approach from other microdisplays. Rather than a matrix array of individual modulator or source for each pixel as seen in liquid crystal display (LCD), organic light-emitting diodes (OLED), and DMD microdisplays, a RSD optimizes a low power light source to create a single pixel and scans this pixel with a single mirror to paint the displayed image directly onto the viewers' retina. With this technique, it offers high spatial and color resolution and very high luminance. There are several papers that provide an overview of the RSD and its applications [35,36]. This technology is developed by Microvision. The RSD systems typically employ two uniaxial scanners or one biaxial scanner. The combinations of two actuation mechanisms, electrostatic (for faster response) and electromagnetic (for larger force) actuations, were selected for a MEMS scanner [37]. Figure 3-3 shows a schematic drawing of the MEMS scanner.

The horizontal scanner (the inner mirror axis) is operated at resonance by using electrostatic actuation. The drive plates are located on the substrate below the MEMS mirror. The inner mirror axis has the resonant frequency of 19.5 kHz with the maximum mechanical scan excursion of 13.4 degrees. The vertical scanner (the outer mirror axis) is magnetically driven by means of permanent magnets within the package and cols with a 60 Hz linear ramp waveform. The magnets need to be positioned carefully and provide sufficient magnetic field to move the mirror to the desired angular deflection. The maximum mechanical scan excursion of 9.6 degrees was achieved on the outer mirror axis. The devices were bulk micromachined utilizing both wet and dry anisotropic etching and electroplating was used to form electromagnetic coils on the outer frame. These scanners must be stiff to remain flat and withstand the forces developed in resonant scanning mode. The dynamic mirror flatness of 0.05 microns rms was measured. The scanners also incorporate piezoresistive strain sensors on the torsion flexures for closed loop control. The scanners are designed to meet SVGA video standards that require 800 x 600 resolution. The design, fabrication, and control details of this bi-axial scanner can be found in [37] and [38].

Optical MEMS 303

Confocal microscopy offers several advantages over conventional optical microscopy, including controllable depth of field, the elimination of image degrading out-of-focus information, and the ability to collect serial optical sections from thick specimens. The concept was introduced by Marvin Minsky in the 1950's when he was a postdoctoral fellow at Harvard University. In 1957, he patented his "double-focusing stage-scanning microscope" in 1957 [40]

In a conventional widefield microscope, the entire specimen is bathed in light from a mercury or xenon source, and the image can be viewed directly by eye or projected onto an image capture device or photographic film. In contrast, the method of image formation in a confocal microscope is fundamentally different. Figure 3-5 shows the schematic drawing of the confocal imaging system. Illumination is achieved by scanning one or more focused beams of light, usually from a laser or arc-discharge source, across the specimen. This point of illumination is brought to focus in the specimen by the objective lens, and laterally scanned using some form of scanning device under computer control. The sequences of points of light from the specimen are detected by a photomultiplier tube (PMT) through a pinhole (or in some cases, a slit), and the output from the PMT is built into an image and

The scanning confocal optical microscope has been recognized for its unique ability to create clear images within thick, light scattering objects. This capability allows the confocal microscope to make high resolution images of living, intact tissues and has led to the

> **Imaging lens**

> > **Point Detector**

**Pin hole**

expectation that confocal microscopy has become a useful tool for in vivo imaging.

**Focal Plane**

**Objective lens**

Fig. 3.5. Schematic drawing shows the concept of the confocal imaging system.

The first compact rectangle shape endoscope (2.5 mm (w) x 6.5 mm (l) x 1.2 mm (t)) based on MEMS scanning mirrors was developed by D.L. Dickensheets, et al. [41]. The architecture of the micromachined confocal optical scanning microscope, illustrated in Figure 3-6, consists of a single-mode optical fiber for illumination and detection, two cascaded one-dimensional bulk micromachined electrostatic scanners with orthogonal axes of rotation to accomplish *x*–*y* scanning, and a binary transmission grating as the objective lens. The maximum mechanical scanned angle is ±2 degrees. The resonant frequencies of both axes are over

**3.1.4 Confocal microscopy and OCT (Optical Coherence Tomography)** 

which is the basis for the confocal microscope.

displayed by the computer.

1 kHz.

**Point Source**

Fig. 3.3. Schematic drawing of the electrostatic/electromagnetic scanner (Picture courtesy of H. Urey. Reprinted from [37] with permission).

Fig. 3.4. MEMS scanning mirror die and close up of PZR strain sensor used for vertical scan position feedback. (Picture courtesy of Davis, W.O.. Reprinted from [39] with permission).

A MEMS scanner based pico-projector system has been commercialized by Microvision [39]. Figure 3-4 shows a real magnetic dual-axis scanner die with piezo-resistor(PZR) strain sensor for 2D raster scanning engine system.

Fig. 3.3. Schematic drawing of the electrostatic/electromagnetic scanner (Picture courtesy of H. Urey. Reprinted from [37] with permission).

Fig. 3.4. MEMS scanning mirror die and close up of PZR strain sensor used for vertical scan position feedback. (Picture courtesy of Davis, W.O.. Reprinted from [39] with permission). A MEMS scanner based pico-projector system has been commercialized by Microvision [39]. Figure 3-4 shows a real magnetic dual-axis scanner die with piezo-resistor(PZR) strain

sensor for 2D raster scanning engine system.

#### **3.1.4 Confocal microscopy and OCT (Optical Coherence Tomography)**

Confocal microscopy offers several advantages over conventional optical microscopy, including controllable depth of field, the elimination of image degrading out-of-focus information, and the ability to collect serial optical sections from thick specimens. The concept was introduced by Marvin Minsky in the 1950's when he was a postdoctoral fellow at Harvard University. In 1957, he patented his "double-focusing stage-scanning microscope" in 1957 [40] which is the basis for the confocal microscope.

In a conventional widefield microscope, the entire specimen is bathed in light from a mercury or xenon source, and the image can be viewed directly by eye or projected onto an image capture device or photographic film. In contrast, the method of image formation in a confocal microscope is fundamentally different. Figure 3-5 shows the schematic drawing of the confocal imaging system. Illumination is achieved by scanning one or more focused beams of light, usually from a laser or arc-discharge source, across the specimen. This point of illumination is brought to focus in the specimen by the objective lens, and laterally scanned using some form of scanning device under computer control. The sequences of points of light from the specimen are detected by a photomultiplier tube (PMT) through a pinhole (or in some cases, a slit), and the output from the PMT is built into an image and displayed by the computer.

The scanning confocal optical microscope has been recognized for its unique ability to create clear images within thick, light scattering objects. This capability allows the confocal microscope to make high resolution images of living, intact tissues and has led to the expectation that confocal microscopy has become a useful tool for in vivo imaging.

Fig. 3.5. Schematic drawing shows the concept of the confocal imaging system.

The first compact rectangle shape endoscope (2.5 mm (w) x 6.5 mm (l) x 1.2 mm (t)) based on MEMS scanning mirrors was developed by D.L. Dickensheets, et al. [41]. The architecture of the micromachined confocal optical scanning microscope, illustrated in Figure 3-6, consists of a single-mode optical fiber for illumination and detection, two cascaded one-dimensional bulk micromachined electrostatic scanners with orthogonal axes of rotation to accomplish *x*–*y* scanning, and a binary transmission grating as the objective lens. The maximum mechanical scanned angle is ±2 degrees. The resonant frequencies of both axes are over 1 kHz.

Optical MEMS 305

Fig. 3.8. DAC microscope architecture (Reprinted from [49] with permission).

**MEMS scanner**

Fig. 3.9. A 5.5-mm diameter DAC microendoscope scanhead. (a) Two collimated beams are focused by a parabolic mirror. Real-time en face scanning is performed by a 2D MEMS scanner. (b) A photograph of the microendoscope without its cap shows a 2D MEMS scanner mounted on the axial translation stage, scale bar 3mm (Reprinted from [49] with permission).

**risley prisms**

**parabolic mirror**

**(a) (b)**

**optical collimator**

**overlapped focal spot**

**axial translation stage**

Fig. 3.6. Schematic drawing of the endoscope head showing various components of the assembly (Picture courtesy of D.L. Dickensheets. Reprinted from [41] with permission).

Later, Olympus Optical Company, Ltd. developed the first commercialized cylindrical shape confocal endoscope with an outside diameter of 3.3 mm and a length of 8 mm [42]. Figure 3-7 shows a cross sectional drawing of the endoscope head. The scanner is a gimbal based two-dimensional bulk micromachined electrostatic scanner [43] with the size of 1.3 mm x 1.3 mm. The mirror has the diameter of 500 um and resonant frequency of 3 kHz. The maximum mechanical scanned angle is ±3 degrees.

Fig. 3.7. A cross sectional drawing of the endoscope head (Picture courtesy of Olympus Optical Company, Ltd. Reprinted from [42] with permission).

Recently, researchers developed the dual-axis confocal (DAC) configuration,in Figure 3-8, to overcome conventional confocal microscope's limitations for endoscope compatibility and in vivo imaging by utilizing two optical fibers oriented along the intersecting optical axes of

Fig. 3.6. Schematic drawing of the endoscope head showing various components of the assembly (Picture courtesy of D.L. Dickensheets. Reprinted from [41] with permission).

Fig. 3.7. A cross sectional drawing of the endoscope head (Picture courtesy of Olympus

Recently, researchers developed the dual-axis confocal (DAC) configuration,in Figure 3-8, to overcome conventional confocal microscope's limitations for endoscope compatibility and in vivo imaging by utilizing two optical fibers oriented along the intersecting optical axes of

Optical Company, Ltd. Reprinted from [42] with permission).

maximum mechanical scanned angle is ±3 degrees.

Later, Olympus Optical Company, Ltd. developed the first commercialized cylindrical shape confocal endoscope with an outside diameter of 3.3 mm and a length of 8 mm [42]. Figure 3-7 shows a cross sectional drawing of the endoscope head. The scanner is a gimbal based two-dimensional bulk micromachined electrostatic scanner [43] with the size of 1.3 mm x 1.3 mm. The mirror has the diameter of 500 um and resonant frequency of 3 kHz. The

Fig. 3.8. DAC microscope architecture (Reprinted from [49] with permission).

Fig. 3.9. A 5.5-mm diameter DAC microendoscope scanhead. (a) Two collimated beams are focused by a parabolic mirror. Real-time en face scanning is performed by a 2D MEMS scanner. (b) A photograph of the microendoscope without its cap shows a 2D MEMS scanner mounted on the axial translation stage, scale bar 3mm (Reprinted from [49] with permission).

Optical MEMS 307

**Reference**

Since its initial use for imaging the transparent and low-scattering tissue of eyes [51], OCT has become attractive for noninvasive medical imaging. Real time in vivo endoscope based OCT imaging systems [52] of the gastrointestinal and respiratory tracts of a rabbit were demonstrated with an axial resolution of 10 um and sensitivity of more than 100 dB. The catheter-endoscope consisted of an encased, rotating hollow cable carrying a single-mode

**Beam Splitter**

**Detector**

**Sample**

Previously, the scanning element inside the OCT probe head used in clinical trials uses a spinning reflective element to scan the light beam across the tissue in circumferential scan geometry [52, 53]. This scanning arrangement allows the imaging probe to view only targets that are directly adjacent to the probe. The scan control of the probe is located outside the probe (proximal actuation). This type of actuation has some drawbacks such as a nonuniform and slow speed scanning. In addition, by applying a rotating torque on the optical

By using MEMS scanning mirrors, the scan control is located inside the probe head (distal actuation) which can reduce the complexity of scan control and potentially have a lower cost. Because of the scanner's miniature size, the overall diameter of the endoscope can be very small (< 5 mm). High speed and large transverse scan can also be achieved which

Therefore, a need for compact, robust, and low cost scanning devices for endoscopic applications has fueled the development of MEMS scanning mirrors for OCT applications. Y. Pan, et al. developed a one-axis electro-thermal CMOS MEMS scanner for endoscopic OCT [54]. The mirror size is 1 mm by 1 mm. The SEM is shown in Figure 3-12. The bimorph beams are composed of a 0.7-um-thick Al layer coated on top of a 1.2-um-thick SiO2 layer embedded with a 0.2-um-thick poly-Si layer. The mirror is coated with a 0.7-um-thick Al layer, and the underlying 40-um-thick single-crystal Si makes the mirror flat. The maximum

Later, J.M. Zara, et. al fabricated one dimensional bulk micromachined MEMS scanner [55]. The scanner (1.5 mm long) is a gold-plated silicon mirror bonded on a 30-um-thick flat polyimide surface (2 mm long and 2.5 mm wide) that pivots on 3-um-thick polyimide

fiber, it can cause unwanted polarization effects that can degrade image quality.

enables real time in vivo imaging and large field of view, respectively.

optical scanned angle is 37 degrees (only in one direction).

Fig. 3.11. Schematic drawing of a Michelson-type interferometer.

**Low-coherence Light Source**

optical fiber.

two low-NA objectives to spatially separate the light paths for illumination and collection [44,45,46,47,48]. A state-of-the-art miniature (OD 5mm) 2D MEMS scanner based nearinfrared dual-axis confocal microscopy system with z-axis focusing has been demonstrated, using 2D MEMS scanner. Figure 3-9 shows the schematic drawing of the whole system. It is the first time that MEMS device based endoscopy system, shown in Figure 3-10, has been used for in vivo imaging in human patients [49].

Fig. 3.10. (a) A DAC microendoscope is passed through the instrument channel of an Olympus XT-160 therapeutic endoscope that has a 6-mm diameter instrument channel. (b) Distal end of an endoscope shows the protruding DAC microendoscope, scale bar 5 mm. (c) Cropped reflectance image of a 1951 United States Air Force resolution test chart collected with the DAC microendoscope shows a transverse resolution of 5 µm, scale bar 20 µm (Reprinted from [49] with permission).

Optical coherence tomography (OCT) is an optical imaging technique that is analogous to Bmode medical ultrasound except that it uses low coherent light (low coherence interferometry) instead of sound. Generally, OCT imaging is performed using a fiber-optic Michelson interferometer with a low-coherence-length light source. Figure 3-11 shows the schematic drawing of the Michelson-type interferometer. One interferometer arm contains a modular probe that focuses and scans the light onto the sample, also collecting the backscattered light. The second interferometer arm is a reference path with a translating mirror or scanning delay line. Optical interference between the light from the sample and reference paths occurs only when the distance traveled by the light in both paths matches to within the coherence length of the light [50]. The interference fringes are detected and demodulated to produce a measurement of the magnitude and echo delay time of light backscattered from structures inside the tissue. The obtained data constitute a two-dimensional map of the backscattering or back reflection from internal architectural morphology and cellular structures in the tissue. Image formation is obtained by perform repeated axial measurement at different transverse positions as the optical beam is scanned across the tissue.

two low-NA objectives to spatially separate the light paths for illumination and collection [44,45,46,47,48]. A state-of-the-art miniature (OD 5mm) 2D MEMS scanner based nearinfrared dual-axis confocal microscopy system with z-axis focusing has been demonstrated, using 2D MEMS scanner. Figure 3-9 shows the schematic drawing of the whole system. It is the first time that MEMS device based endoscopy system, shown in Figure 3-10, has been

**(a) (c)**

**therapeutic upper GI scope**

**parabolic mirror cap**

Optical coherence tomography (OCT) is an optical imaging technique that is analogous to Bmode medical ultrasound except that it uses low coherent light (low coherence interferometry) instead of sound. Generally, OCT imaging is performed using a fiber-optic Michelson interferometer with a low-coherence-length light source. Figure 3-11 shows the schematic drawing of the Michelson-type interferometer. One interferometer arm contains a modular probe that focuses and scans the light onto the sample, also collecting the backscattered light. The second interferometer arm is a reference path with a translating mirror or scanning delay line. Optical interference between the light from the sample and reference paths occurs only when the distance traveled by the light in both paths matches to within the coherence length of the light [50]. The interference fringes are detected and demodulated to produce a measurement of the magnitude and echo delay time of light backscattered from structures inside the tissue. The obtained data constitute a two-dimensional map of the backscattering or back reflection from internal architectural morphology and cellular structures in the tissue. Image formation is obtained by perform repeated axial measurement at different transverse

Fig. 3.10. (a) A DAC microendoscope is passed through the instrument channel of an Olympus XT-160 therapeutic endoscope that has a 6-mm diameter instrument channel. (b) Distal end of an endoscope shows the protruding DAC microendoscope, scale bar 5 mm. (c) Cropped reflectance image of a 1951 United States Air Force resolution test chart collected with the DAC microendoscope shows a transverse resolution of 5 µm, scale bar 20 µm

**hemispherical lens**

used for in vivo imaging in human patients [49].

**(b)**

**DAC microendoscope**

(Reprinted from [49] with permission).

positions as the optical beam is scanned across the tissue.

Fig. 3.11. Schematic drawing of a Michelson-type interferometer.

Since its initial use for imaging the transparent and low-scattering tissue of eyes [51], OCT has become attractive for noninvasive medical imaging. Real time in vivo endoscope based OCT imaging systems [52] of the gastrointestinal and respiratory tracts of a rabbit were demonstrated with an axial resolution of 10 um and sensitivity of more than 100 dB. The catheter-endoscope consisted of an encased, rotating hollow cable carrying a single-mode optical fiber.

Previously, the scanning element inside the OCT probe head used in clinical trials uses a spinning reflective element to scan the light beam across the tissue in circumferential scan geometry [52, 53]. This scanning arrangement allows the imaging probe to view only targets that are directly adjacent to the probe. The scan control of the probe is located outside the probe (proximal actuation). This type of actuation has some drawbacks such as a nonuniform and slow speed scanning. In addition, by applying a rotating torque on the optical fiber, it can cause unwanted polarization effects that can degrade image quality.

By using MEMS scanning mirrors, the scan control is located inside the probe head (distal actuation) which can reduce the complexity of scan control and potentially have a lower cost. Because of the scanner's miniature size, the overall diameter of the endoscope can be very small (< 5 mm). High speed and large transverse scan can also be achieved which enables real time in vivo imaging and large field of view, respectively.

Therefore, a need for compact, robust, and low cost scanning devices for endoscopic applications has fueled the development of MEMS scanning mirrors for OCT applications. Y. Pan, et al. developed a one-axis electro-thermal CMOS MEMS scanner for endoscopic OCT [54]. The mirror size is 1 mm by 1 mm. The SEM is shown in Figure 3-12. The bimorph beams are composed of a 0.7-um-thick Al layer coated on top of a 1.2-um-thick SiO2 layer embedded with a 0.2-um-thick poly-Si layer. The mirror is coated with a 0.7-um-thick Al layer, and the underlying 40-um-thick single-crystal Si makes the mirror flat. The maximum optical scanned angle is 37 degrees (only in one direction).

Later, J.M. Zara, et. al fabricated one dimensional bulk micromachined MEMS scanner [55]. The scanner (1.5 mm long) is a gold-plated silicon mirror bonded on a 30-um-thick flat polyimide surface (2 mm long and 2.5 mm wide) that pivots on 3-um-thick polyimide

Optical MEMS 309

micromachining techniques [57]. The angular vertical comb (AVC) bank actuators provide high-angle scanning at low applied voltage [58]. The combination of both fabrication techniques enables high actuation force, large flat micromirrors, flexible electrical interconnect, and tightly-controlled spring constants [58,59]. The schematic drawing of the 2D scanner is illustrated in Figure 3-14. An single-crystalline silicon (SCS) micromirror is suspended inside a gimbal frame by a pair of polysilicon torsion springs. The gimbal frame is supported by two pairs of polysilicon torsion springs. The four electrically isolated torsion beams also provide three independent voltages (V1 to V3) to inner gimbals and mirrors. The torsion spring is 345 µm long, 10 or 12 µm wide, and 3.5 µm thick. The scanner has 8 comb banks with 10 movable fingers each. The finger is 4.6 µm wide, 242 µm long, and 35 µm thick. The gap spacing between comb fingers is 4.4 µm. The mirror is 1000 µm in diameter and 35 µm thick. The AVC banks are fabricated on SCS. The movable and fixed comb banks

Fig. 3.14. Schematic of 2D AVC gimbal scanner (Reprinted from [58] with permission).

The endoscope head is 5-mm in diameter and 2.5-cm long, which is compatible with requirements for minimally invasive endoscopic procedures. Figure 3-15 shows a schematic of the fiber coupled MEMS scanning endoscope. The compact aluminum housing can be machined for low cost and allows precise adjustment of optical alignment using tiny set screws. The optics consists of a graded-index fiber collimator followed by an anti-reflection coated achromatic focusing lens producing a beam diameter of

The 2D MEMS scanner is mounted at 45 degrees and directs the beam orthogonal to the endoscope axis in a side-scanning configuration similar to those typically used for endoscopic OCT procedures. Post-objective scanning eliminates off-axis optical aberration encountered with pre-objective scanning. Figure 3-16 shows a scanning electron micrograph of the 2D AVC scanner located inside the endoscope package. The large 1-mm diameter

are completely self-aligned [58].

~ 12 um [56,57].

mirror allows high-numerical-aperture focusing.

torsion hinges. Figure 3-13 shows an optical image of the endoscope head. The actuator used to tilt the mirror, the integrated force array (IFA), is a network of hundreds of thousands of micrometer-scale deformable capacitors. The capacitive cells contract because of the presence of electrostatic forces produced by a differential voltage applied across the capacitor electrodes.

Fig. 3.12. SEM of electro-thermal CMOS MEMS scanner with an inset shows a close-up view of the bending springs (Picture courtesy of T. Xie. Reprinted from [54] with permission).

Fig. 3.13. An optical image of the endoscope head (Picture courtesy of J.M. Zara. Reprinted from [55] with permission).

Researchers at MIT and UCLA [56] have developed the first two-dimensional endoscopic MEMS scanner for high resolution optical coherence tomography. The two dimensional scanner with angular vertical comb actuators (AVC) is fabricated by using surface and bulk

torsion hinges. Figure 3-13 shows an optical image of the endoscope head. The actuator used to tilt the mirror, the integrated force array (IFA), is a network of hundreds of thousands of micrometer-scale deformable capacitors. The capacitive cells contract because of the presence of electrostatic forces produced by a differential voltage applied across the

Fig. 3.12. SEM of electro-thermal CMOS MEMS scanner with an inset shows a close-up view of the bending springs (Picture courtesy of T. Xie. Reprinted from [54] with permission).

Fig. 3.13. An optical image of the endoscope head (Picture courtesy of J.M. Zara. Reprinted

Researchers at MIT and UCLA [56] have developed the first two-dimensional endoscopic MEMS scanner for high resolution optical coherence tomography. The two dimensional scanner with angular vertical comb actuators (AVC) is fabricated by using surface and bulk

capacitor electrodes.

from [55] with permission).

micromachining techniques [57]. The angular vertical comb (AVC) bank actuators provide high-angle scanning at low applied voltage [58]. The combination of both fabrication techniques enables high actuation force, large flat micromirrors, flexible electrical interconnect, and tightly-controlled spring constants [58,59]. The schematic drawing of the 2D scanner is illustrated in Figure 3-14. An single-crystalline silicon (SCS) micromirror is suspended inside a gimbal frame by a pair of polysilicon torsion springs. The gimbal frame is supported by two pairs of polysilicon torsion springs. The four electrically isolated torsion beams also provide three independent voltages (V1 to V3) to inner gimbals and mirrors. The torsion spring is 345 µm long, 10 or 12 µm wide, and 3.5 µm thick. The scanner has 8 comb banks with 10 movable fingers each. The finger is 4.6 µm wide, 242 µm long, and 35 µm thick. The gap spacing between comb fingers is 4.4 µm. The mirror is 1000 µm in diameter and 35 µm thick. The AVC banks are fabricated on SCS. The movable and fixed comb banks are completely self-aligned [58].

Fig. 3.14. Schematic of 2D AVC gimbal scanner (Reprinted from [58] with permission).

The endoscope head is 5-mm in diameter and 2.5-cm long, which is compatible with requirements for minimally invasive endoscopic procedures. Figure 3-15 shows a schematic of the fiber coupled MEMS scanning endoscope. The compact aluminum housing can be machined for low cost and allows precise adjustment of optical alignment using tiny set screws. The optics consists of a graded-index fiber collimator followed by an anti-reflection coated achromatic focusing lens producing a beam diameter of ~ 12 um [56,57].

The 2D MEMS scanner is mounted at 45 degrees and directs the beam orthogonal to the endoscope axis in a side-scanning configuration similar to those typically used for endoscopic OCT procedures. Post-objective scanning eliminates off-axis optical aberration encountered with pre-objective scanning. Figure 3-16 shows a scanning electron micrograph of the 2D AVC scanner located inside the endoscope package. The large 1-mm diameter mirror allows high-numerical-aperture focusing.

Optical MEMS 311

(a) (b)

Optical MEMS devices offer a promising alternative to piezoelectric and other deformable mirror types used in adaptive optics applications. Adaptive optics refers to optical systems which adapt to compensate for undesirable optical effects introduced by the medium between the object and its image. It provides a means of compensating for these effects, leading to appreciably sharper images approaching the theoretical diffraction limit. These efforts include wave front correction, aberration cancellation etc. While sharper images come with an additional gain in contrast for astronomy, where light levels are often very low, this means fainter objects can be detected and studied [60, 61]. Other interesting applications used adaptive optics are confocal microscopy [62], adaptive laser wavefront correction [63], cryogenic adaptive optics [64], and human vision [65]. Several groups have

A two-level silicon surface micromachining approach was employed by researchers at Boston University to produce MEMS deformable mirrors using an original architecture described in Figure 3-18 [66,67]. The kilo-pixel spatial light modulator is made up of 1024 individually addressable surface-normal electrostatic actuators with center posts that support individual optical mirror segments. Each electrostatic actuator consists of a silicon membrane anchored to the substrate on two sides above a silicon electrode. These devices were manufactured at a commercial MEMS foundry [68]. A post centered on each actuator supports a 338μm x 338µm x 3µm optically coated mirror segment. The spatial light modulator has an aperture of 10 mm, an actuator stroke of 2 µm, and a position repeatability

At Jet Propulsion Laboratory (JPL), researchers developed a single crystal silicon continuous membrane deformable mirror with underlying piezoelectric unimorph actuators as shown in Figure 3-19 [69]. A PZT unimorph actuator of 2.5 mm in diameter with optimized PZT/Si thickness and design showed a deflection of 5 µm at 50V. Deformable mirrors consisting of 10µm thick single-crystal silicon membranes supported by 4×4 actuator arrays were

Fig. 3.17. Electrothermal MEMS OCT reported by Liu et al. (a) SEM and schematic of through vias electrothermal MEMS mirror, (b) 2.6mm probe 3D model and assembled

probe, (Picture courtesy of Huikai Xie. Reprinted from [18] with permission).

**3.1.5 Adaptive optics** 

developed MEMS deformable mirrors.

of 3 nm. The resonant frequency of the mirror is around 60 kHz.

Fig. 3.15. Schematic drawing of the endoscope head (Reprinted from [56] with permission).

Fig. 3.16. SEM of 2D AVC gimbal scanner (Reprinted from [58] with permission).

Recently, Huikai Xie et. al developed a miniature endoscopic optical coherence tomography probe system employing a two-axis microelectromechanical scanning mirror with throughsilicon vias (TSV). The new scanner's design (Figure 3-17) improves the assembly and package of the probe.

Fig. 3.15. Schematic drawing of the endoscope head (Reprinted from [56] with permission).

Fig. 3.16. SEM of 2D AVC gimbal scanner (Reprinted from [58] with permission).

package of the probe.

Recently, Huikai Xie et. al developed a miniature endoscopic optical coherence tomography probe system employing a two-axis microelectromechanical scanning mirror with throughsilicon vias (TSV). The new scanner's design (Figure 3-17) improves the assembly and

Fig. 3.17. Electrothermal MEMS OCT reported by Liu et al. (a) SEM and schematic of through vias electrothermal MEMS mirror, (b) 2.6mm probe 3D model and assembled probe, (Picture courtesy of Huikai Xie. Reprinted from [18] with permission).

#### **3.1.5 Adaptive optics**

Optical MEMS devices offer a promising alternative to piezoelectric and other deformable mirror types used in adaptive optics applications. Adaptive optics refers to optical systems which adapt to compensate for undesirable optical effects introduced by the medium between the object and its image. It provides a means of compensating for these effects, leading to appreciably sharper images approaching the theoretical diffraction limit. These efforts include wave front correction, aberration cancellation etc. While sharper images come with an additional gain in contrast for astronomy, where light levels are often very low, this means fainter objects can be detected and studied [60, 61]. Other interesting applications used adaptive optics are confocal microscopy [62], adaptive laser wavefront correction [63], cryogenic adaptive optics [64], and human vision [65]. Several groups have developed MEMS deformable mirrors.

A two-level silicon surface micromachining approach was employed by researchers at Boston University to produce MEMS deformable mirrors using an original architecture described in Figure 3-18 [66,67]. The kilo-pixel spatial light modulator is made up of 1024 individually addressable surface-normal electrostatic actuators with center posts that support individual optical mirror segments. Each electrostatic actuator consists of a silicon membrane anchored to the substrate on two sides above a silicon electrode. These devices were manufactured at a commercial MEMS foundry [68]. A post centered on each actuator supports a 338μm x 338µm x 3µm optically coated mirror segment. The spatial light modulator has an aperture of 10 mm, an actuator stroke of 2 µm, and a position repeatability of 3 nm. The resonant frequency of the mirror is around 60 kHz.

At Jet Propulsion Laboratory (JPL), researchers developed a single crystal silicon continuous membrane deformable mirror with underlying piezoelectric unimorph actuators as shown in Figure 3-19 [69]. A PZT unimorph actuator of 2.5 mm in diameter with optimized PZT/Si thickness and design showed a deflection of 5 µm at 50V. Deformable mirrors consisting of 10µm thick single-crystal silicon membranes supported by 4×4 actuator arrays were

Optical MEMS 313

(a) (b)

Researchers at UCLA fabricated two cascaded two dimensional scanners for optical surveying instruments [71]. Currently, cascaded acousto-optic deflectors are used the optical surveying instruments. MEMS scanners are very attractive candidates for replacing those scanners. They offer many advantages, including lower power consumption, smaller size, and potentially lower cost. Optical surveying instruments require mirrors with reasonably large scan range (~ ±6° mechanical), high resonant frequencies (5-10 kHz for fast axis), large

Typically, the target, a highly reflective surface consisting of corner cubes, is located several meters to several kilometers away from the instruments. The required scanning angular range is relatively small, on the order of a few degrees. The angular divergence of the measurement laser beam is typically a few milliradians or narrower. Hence the target search system needs to resolve several tens of spots in the entire scan range. Raster scanning has been used because the laser beam needs to search the entire area within the field of view to find the target. A combination of fast and slow scanning scanners has been employed. Both are fabricated by using bulk micromachining technique with 25-μm-thick SOI wafer. The fast scanner has a circular shape and achieved a resonant frequency of 7.5 kHz with the maximum mechanical rotation in DC mode of 3.2 degrees. The slow scanner has an elongated circular shape and achieved a resonant frequency of 1.2 kHz with the maximum mechanical rotation in DC mode of 0.74 degree. Figure 3-21 shows the SEM of the fast

Fig. 3.20. a) Schematic diagram of a 700 μm diameter (vertex-to-vertex) mirror segment. Scaling is highly exaggerated in the vertical direction. b) Die photograph of a 489-actuator 163-piston/tip/tilt-segment DM with 7.7 mm inscribed aperture (Product name: PTT489-5).

(Reprinted from [70] with permission).

radius of curvature, and low supply voltage (< 50V).

**3.1.6 Other examples** 

mirror.

fabricated and optically characterized. An assembled deformable mirror showed a stroke of 2.5µm at 50V with a resonant frequency of 42 kHz.

Fig. 3.18. Schematic cross section of a gold coated spatial light modulator with a central deflected actuator (Picture courtesy of Thomas G. Bifano. Reprinted from [66] with permission).

Fig. 3.19. Cross-sectional drawing of the deformable mirror (Reprinted from [69] with permission).

Another MEMS adaptic optical system has been developed by IRIS AO Inc. [70] using Aucoated dielectric-coated segmented MEMS deformable mirrors (DM) for Optical MEMS application. The AO system (Figure 3-20) consists of piston-tip-tilt (PTT) segmented MEMS deformable mirrors (DM) and adaptive optics controllers for these DMs, shown in Figure 3-20. Researchers from IRIS AO Inc. have made many improvements for mirrors to realize open-loop flatten up to 4 nm rms.

fabricated and optically characterized. An assembled deformable mirror showed a stroke of

Fig. 3.18. Schematic cross section of a gold coated spatial light modulator with a central deflected actuator (Picture courtesy of Thomas G. Bifano. Reprinted from [66] with

Another MEMS adaptic optical system has been developed by IRIS AO Inc. [70] using Aucoated dielectric-coated segmented MEMS deformable mirrors (DM) for Optical MEMS application. The AO system (Figure 3-20) consists of piston-tip-tilt (PTT) segmented MEMS deformable mirrors (DM) and adaptive optics controllers for these DMs, shown in Figure 3-20. Researchers from IRIS AO Inc. have made many improvements for mirrors to realize

Fig. 3.19. Cross-sectional drawing of the deformable mirror

(Reprinted from [69] with permission).

open-loop flatten up to 4 nm rms.

2.5µm at 50V with a resonant frequency of 42 kHz.

permission).

Fig. 3.20. a) Schematic diagram of a 700 μm diameter (vertex-to-vertex) mirror segment. Scaling is highly exaggerated in the vertical direction. b) Die photograph of a 489-actuator 163-piston/tip/tilt-segment DM with 7.7 mm inscribed aperture (Product name: PTT489-5). (Reprinted from [70] with permission).

#### **3.1.6 Other examples**

Researchers at UCLA fabricated two cascaded two dimensional scanners for optical surveying instruments [71]. Currently, cascaded acousto-optic deflectors are used the optical surveying instruments. MEMS scanners are very attractive candidates for replacing those scanners. They offer many advantages, including lower power consumption, smaller size, and potentially lower cost. Optical surveying instruments require mirrors with reasonably large scan range (~ ±6° mechanical), high resonant frequencies (5-10 kHz for fast axis), large radius of curvature, and low supply voltage (< 50V).

Typically, the target, a highly reflective surface consisting of corner cubes, is located several meters to several kilometers away from the instruments. The required scanning angular range is relatively small, on the order of a few degrees. The angular divergence of the measurement laser beam is typically a few milliradians or narrower. Hence the target search system needs to resolve several tens of spots in the entire scan range. Raster scanning has been used because the laser beam needs to search the entire area within the field of view to find the target. A combination of fast and slow scanning scanners has been employed. Both are fabricated by using bulk micromachining technique with 25-μm-thick SOI wafer. The fast scanner has a circular shape and achieved a resonant frequency of 7.5 kHz with the maximum mechanical rotation in DC mode of 3.2 degrees. The slow scanner has an elongated circular shape and achieved a resonant frequency of 1.2 kHz with the maximum mechanical rotation in DC mode of 0.74 degree. Figure 3-21 shows the SEM of the fast mirror.

Optical MEMS 315

**"DROP" Channels**

The first 2D matrix optical switch reported by Toshiyoshi and Fujita employed pop-up-type switching elements, as shown in Figure 3-23 [72]. It consists of two bonded wafers. The mirrors are suspended by a pair of torsion beams in the plane of the top wafer. The biasing electrodes are fabricated on the bottom wafer. When a voltage is applied between the mirror and the bottom electrode, the mirror rotates downward by 90 by electrostatic actuation. The mirror angle in the ON (down) state is controlled by a stopper on the bottom wafer. Since the mirror angle is defined by the relative positions of two wafers, precise alignment is necessary to achieve accurate and uniform mirror angles. A single-chip electrostatic pop-up mirror has recently been reported [76]. The actuation and mechanical stopper are realized between a back-flap and a vertical trench etched in the silicon substrate. The angular accuracy and uniformity of their mirrors depend on the etched sidewall profile and the

**Input Channels**

**"ADD" Channels**

**Collimator Arrays** 

AT&T Lab has reported surface-micromachined 2D switches with free-rotating hinged mirrors [73,77,78]. The schematic drawing and the SEM of the switch are shown in Figure 3-24. It is fabricated using the MUMPs process. The mirror is pivoted on the substrate by microhinges. A pair of pushrods is used to convert in-plane translations into out-of-plane rotations of the mirrors. The switch is powered by scratch drive actuators [79]. Though scratch drive actuators do not move at high speed, fast switching time is achieved (700 sec) because only a short traveling distance (22 m) is needed for the mirror to reach 90. The free-rotating microhinges have an inherent 0.75-m clearance between the hinge pins and the staples, which could result in large variations in mirror angles. Using improved design and mechanical stoppers that are insensitive to lithographic misalignment during fabrication, mirror angular repeatability of better than 0.1 was experimentally demonstrated [80]. The mirror flatness was improved by using a multi-layer structure with phosphosilicate glass (PSG) sandwiched between two polysilicon layers. The largest switch size demonstrated is 8x8 due to the foundry-imposed chip-size limits of 1 cm x 1 cm. One of the potential issues is the constant tear and wear of the free-rotating hinges and actuators.

Fig. 3.22. Schematic of 2D MEMS optical switches.

**Output Channels**

**2D MEMS Switch Chip** 

lithographic alignment accuracy.

Fig. 3.21. SEM of the fast mirror (Reprinted from [71] with permission)

#### **3.2 Optical communication 3.2.1 2D MEMS switch**

The two-dimensional (2D) MEMS optical switch is basically an optical crossbar switch with N2 micromirrors that can selectively reflect the optical beams to orthogonal output ports or pass them to the following mirrors. They are often referred to as "2D switches" because the optical beams are switched in a two-dimensional plane. This is in contrast to the 3D switch (discussed in the following section) in which the optical beams are steered in threedimensional space. A generic configuration of the 2D switch is shown in Figure 3-22. The core of the switch is an NxN array of micromirrors for a switch with N input and N output fibers. The optical beams are collimated to reduce diffraction loss. The micromirrors intersect the optical beams at 45, and can be switched in and out of the optical beam path. The micromirrors are "digital", that is, they are either in the optical beam path (ON) or completely out of the beam path. When the mirror in the i-th row and j-th column (Mij) is ON, the i-th input beam is switched to the j-th output fiber. Generally, only one micromirror in a column or a row is ON. Thus during operation of an NxN switch, only N micromirrors are in the ON position while the rest of the micromirrors are in the OFF position. MEMS 2D switches were first reported by Toshiyoshi and Fujita [72]. Several different ways of switching micromirrors have been reported, including rotating, sliding, chopping, and flipping motions. The switches are usually actuated by electrostatic, electromagnetic, or piezoelectric mechanisms.

2D switches using various types of flip-up (or pop-up) mirrors have been reported [72,73,74,75]. They are realized by either the bulk- [72] or the surface-micromachining [73] technology, or the combination of both [74]. The mirror lies in the plane of the substrate during OFF state and pops up in the ON state. Since the mirror angle is changing continuously during switching, a common challenge for this type of switch is the reproducibility of mirror angle. This is a critical issue as the mirror angles and their uniformity play a critical role in the performance and the scalability of 2D switches. The reproducibility of mirror angles over switching cycles determines the repeatability of insertion loss.

The two-dimensional (2D) MEMS optical switch is basically an optical crossbar switch with N2 micromirrors that can selectively reflect the optical beams to orthogonal output ports or pass them to the following mirrors. They are often referred to as "2D switches" because the optical beams are switched in a two-dimensional plane. This is in contrast to the 3D switch (discussed in the following section) in which the optical beams are steered in threedimensional space. A generic configuration of the 2D switch is shown in Figure 3-22. The core of the switch is an NxN array of micromirrors for a switch with N input and N output fibers. The optical beams are collimated to reduce diffraction loss. The micromirrors intersect the optical beams at 45, and can be switched in and out of the optical beam path. The micromirrors are "digital", that is, they are either in the optical beam path (ON) or completely out of the beam path. When the mirror in the i-th row and j-th column (Mij) is ON, the i-th input beam is switched to the j-th output fiber. Generally, only one micromirror in a column or a row is ON. Thus during operation of an NxN switch, only N micromirrors are in the ON position while the rest of the micromirrors are in the OFF position. MEMS 2D switches were first reported by Toshiyoshi and Fujita [72]. Several different ways of switching micromirrors have been reported, including rotating, sliding, chopping, and flipping motions. The switches are usually actuated by electrostatic, electromagnetic, or

2D switches using various types of flip-up (or pop-up) mirrors have been reported [72,73,74,75]. They are realized by either the bulk- [72] or the surface-micromachining [73] technology, or the combination of both [74]. The mirror lies in the plane of the substrate during OFF state and pops up in the ON state. Since the mirror angle is changing continuously during switching, a common challenge for this type of switch is the reproducibility of mirror angle. This is a critical issue as the mirror angles and their uniformity play a critical role in the performance and the scalability of 2D switches. The reproducibility of mirror angles over switching cycles determines the repeatability of

Fig. 3.21. SEM of the fast mirror (Reprinted from [71] with permission)

**3.2 Optical communication 3.2.1 2D MEMS switch** 

piezoelectric mechanisms.

insertion loss.

Fig. 3.22. Schematic of 2D MEMS optical switches.

The first 2D matrix optical switch reported by Toshiyoshi and Fujita employed pop-up-type switching elements, as shown in Figure 3-23 [72]. It consists of two bonded wafers. The mirrors are suspended by a pair of torsion beams in the plane of the top wafer. The biasing electrodes are fabricated on the bottom wafer. When a voltage is applied between the mirror and the bottom electrode, the mirror rotates downward by 90 by electrostatic actuation. The mirror angle in the ON (down) state is controlled by a stopper on the bottom wafer. Since the mirror angle is defined by the relative positions of two wafers, precise alignment is necessary to achieve accurate and uniform mirror angles. A single-chip electrostatic pop-up mirror has recently been reported [76]. The actuation and mechanical stopper are realized between a back-flap and a vertical trench etched in the silicon substrate. The angular accuracy and uniformity of their mirrors depend on the etched sidewall profile and the lithographic alignment accuracy.

AT&T Lab has reported surface-micromachined 2D switches with free-rotating hinged mirrors [73,77,78]. The schematic drawing and the SEM of the switch are shown in Figure 3-24. It is fabricated using the MUMPs process. The mirror is pivoted on the substrate by microhinges. A pair of pushrods is used to convert in-plane translations into out-of-plane rotations of the mirrors. The switch is powered by scratch drive actuators [79]. Though scratch drive actuators do not move at high speed, fast switching time is achieved (700 sec) because only a short traveling distance (22 m) is needed for the mirror to reach 90. The free-rotating microhinges have an inherent 0.75-m clearance between the hinge pins and the staples, which could result in large variations in mirror angles. Using improved design and mechanical stoppers that are insensitive to lithographic misalignment during fabrication, mirror angular repeatability of better than 0.1 was experimentally demonstrated [80]. The mirror flatness was improved by using a multi-layer structure with phosphosilicate glass (PSG) sandwiched between two polysilicon layers. The largest switch size demonstrated is 8x8 due to the foundry-imposed chip-size limits of 1 cm x 1 cm. One of the potential issues is the constant tear and wear of the free-rotating hinges and actuators.

Optical MEMS 317

**Hinge**

**Hinge joint**

**Actuated translation stage**

**Switch mirror**

**Pushrod**

Fig. 3.24. Schematic and SEM of the surface-micromachined free-rotating hinged mirrors reported by AT&T (Picture courtesy of Lih Y. Lin. Reprinted from [73,77] with permission). On the other hand, the "chopper type" 2D switch employs a vertical mirror whose height can be changed by MEMS actuators [81,82]. The mirror angle is fixed during switching, and excellent repeatability of insertion loss has been reported. Figure 3-25 shows the schematic diagram of OMM's 2D switch [81]. The mirror is assembled vertically at the tip of a long actuator plate. The plate is tilted upward and fixed by micro-latches to raise the mirror height. Large traveling distance is achieved by extending the actuator arm. Several hundred microns displacement can be achieved with this configuration. The switch is actuated electrostatically by applying a voltage between the actuator plate and a bottom electrode on the substrate. The mirror moves in the vertical direction and the mirror angle

This might affect the reliability of the switch and the accuracy and uniformity of mirror angles over many switching cycles.

Fig. 3.23. Schematic and SEM of bulk-micromachined 2D switch with free-rotating torsion mirrors (Pictures courtesy of Hiroshi Toshiyoshi. Reprinted from [72] with permission).

This might affect the reliability of the switch and the accuracy and uniformity of mirror

MIRROR is ON

**M4**

**COUNTER ELECTRODE**

MIRROR STOPPER (SHALLOW DIMPLE)

**M2**

SPHERICAL LENS

**M1**

MIRROR is OFF

**M3**

Fig. 3.23. Schematic and SEM of bulk-micromachined 2D switch with free-rotating torsion mirrors (Pictures courtesy of Hiroshi Toshiyoshi. Reprinted from [72] with permission).

angles over many switching cycles.

**3mm**

**TORSION MIRROR CHIP**

Fig. 3.24. Schematic and SEM of the surface-micromachined free-rotating hinged mirrors reported by AT&T (Picture courtesy of Lih Y. Lin. Reprinted from [73,77] with permission).

On the other hand, the "chopper type" 2D switch employs a vertical mirror whose height can be changed by MEMS actuators [81,82]. The mirror angle is fixed during switching, and excellent repeatability of insertion loss has been reported. Figure 3-25 shows the schematic diagram of OMM's 2D switch [81]. The mirror is assembled vertically at the tip of a long actuator plate. The plate is tilted upward and fixed by micro-latches to raise the mirror height. Large traveling distance is achieved by extending the actuator arm. Several hundred microns displacement can be achieved with this configuration. The switch is actuated electrostatically by applying a voltage between the actuator plate and a bottom electrode on the substrate. The mirror moves in the vertical direction and the mirror angle

Optical MEMS 319

Fig. 3.26. (a) SEM of OMM's 16x16 switch. (b) Measured distribution of mirror angles for the

The structure of a 2x2 2D MEMS switch can eventually be simplified so that it requires only one single micromirror between two pairs of orthogonal fibers. A simple, elegant solution for 2x2 switches is using SOI-based Optical MEMS [83,84]. The schematic and SEM picture (only fiber grooves shown) of the 2x2 switch are shown in Figure 3-27. Electrostatic comb drive actuators and vertical micromirrors are fabricated on 75-m SOI wafers. The mirrors can be coated with metal by angle evaporation to increase their reflectivity. The most critical part of the process is the etching of thin (< 2 m) vertical mirrors with smooth sidewalls. Thin vertical mirror is required for such a 2x2 switch because the offset of the reflected optical beams from the opposite sides of the mirror caused by the finite thickness of the mirror will introduce additional optical loss. The Institute of Microtechnology (IMT) at

16x16 switch (Picture courtesy of Li Fan. Reprinted from [81] with permission).

is maintained at 90 during the entire switching cycle. The actuator is basically a gapclosing actuator. A mechanical stopper defines the lower position of the mirror. OMM employs a curved landing bar with a single point contact to minimize stiction and increase reliability (see Figure 3-25). More than 100 million cycles have been demonstrated with repeatable mirror angle and performance. The landing bar also provides a cushion that helps reduce mirror ringing and improve switching time. They have demonstrated a switching time of 12 ms using a square-wave driving voltage without pre-shaping the waveform. OMM's switch is fabricated with polysilicon surface-micromachining technology. The mirrors and the actuators are batch-assembled into the 3D structures. Figure 3-26(a) shows the SEM of a 16x16 switch. The distribution of mirror angles is shown in Figure 3-26(b). The uniformity is better than 0.1 degrees for 256 mirrors. The switch is hermetically packaged with optical collimator arrays. Extensive testing has been performed for the packaged switches. The maximum insertion loss is less than 3.1 dB and the crosstalk is less than – 50 dB. Loss variation over the wavelength range of 1280 – 1650 nm range is less than1 dB. Return loss is greater than 50dB, and maximum temperature variation is < 1 dB over a temperature range of 0 - 60C. Polarization dependent loss (PDL) is < 0.4 dB and polarization mode dispersion (PMD) is < 0.08 ps. Vibration tests show < 0.2 dB change under operation, and 3 axis shock tests confirm no change of operational characteristics under 200 G.

Fig. 3.25. Schematic of a switching element in OMM's 2D switch (Picture courtesy of Li Fan. Reprinted from [81] with permission).

is maintained at 90 during the entire switching cycle. The actuator is basically a gapclosing actuator. A mechanical stopper defines the lower position of the mirror. OMM employs a curved landing bar with a single point contact to minimize stiction and increase reliability (see Figure 3-25). More than 100 million cycles have been demonstrated with repeatable mirror angle and performance. The landing bar also provides a cushion that helps reduce mirror ringing and improve switching time. They have demonstrated a switching time of 12 ms using a square-wave driving voltage without pre-shaping the waveform. OMM's switch is fabricated with polysilicon surface-micromachining technology. The mirrors and the actuators are batch-assembled into the 3D structures. Figure 3-26(a) shows the SEM of a 16x16 switch. The distribution of mirror angles is shown in Figure 3-26(b). The uniformity is better than 0.1 degrees for 256 mirrors. The switch is hermetically packaged with optical collimator arrays. Extensive testing has been performed for the packaged switches. The maximum insertion loss is less than 3.1 dB and the crosstalk is less than – 50 dB. Loss variation over the wavelength range of 1280 – 1650 nm range is less than1 dB. Return loss is greater than 50dB, and maximum temperature variation is < 1 dB over a temperature range of 0 - 60C. Polarization dependent loss (PDL) is < 0.4 dB and polarization mode dispersion (PMD) is < 0.08 ps. Vibration tests show < 0.2 dB change under operation, and 3 axis shock tests confirm no change of

Fig. 3.25. Schematic of a switching element in OMM's 2D switch (Picture courtesy of Li Fan.

operational characteristics under 200 G.

Reprinted from [81] with permission).

Fig. 3.26. (a) SEM of OMM's 16x16 switch. (b) Measured distribution of mirror angles for the 16x16 switch (Picture courtesy of Li Fan. Reprinted from [81] with permission).

The structure of a 2x2 2D MEMS switch can eventually be simplified so that it requires only one single micromirror between two pairs of orthogonal fibers. A simple, elegant solution for 2x2 switches is using SOI-based Optical MEMS [83,84]. The schematic and SEM picture (only fiber grooves shown) of the 2x2 switch are shown in Figure 3-27. Electrostatic comb drive actuators and vertical micromirrors are fabricated on 75-m SOI wafers. The mirrors can be coated with metal by angle evaporation to increase their reflectivity. The most critical part of the process is the etching of thin (< 2 m) vertical mirrors with smooth sidewalls. Thin vertical mirror is required for such a 2x2 switch because the offset of the reflected optical beams from the opposite sides of the mirror caused by the finite thickness of the mirror will introduce additional optical loss. The Institute of Microtechnology (IMT) at

Optical MEMS 321

**Scanner Array**

> **Output Collimator Array**

**Array Reflecting**

Fig. 3.28. Configuration for 3-D optical switch (NxN) with 2N analog scanning mirrors.

Fig. 3.29. SEM of surface-micromachined 2-axis scanners Lucent Technologies (Courtesy of [109] Lucent Technologies Inc. © 2003 Lucent Technologies Inc.

All rights reserved.)

**Reflecting Scanner Array**

**Input Collimator**

Lucent technologies employed a self-assembly technique which was driven by the residual stress in deposited thin films (Cr/Au on polysilicon) to raise two-axis polysilicon scanners (500 m mirror diameter) to a fixed position 50 m above the substrate [106,107]. The scanning electron micrograph (SEM) is shown in Figure 3-29. Two-axis scanning is achieved

Neuchatel has perfected the mirror etching technology by DRIE [85]. A surface roughness of 36 nm has been achieved. The switch has excellent optical performance: 0.3-0.5 dB optical insertion loss and 500 sec switching time, and very low polarization dependence [84].

Fig. 3.27. Schematic and SEM of 2x2 switch fabricated by SOI MEMS (Picture courtesy of Nico de Rooij. Reprinted from [83] with permission).
