4
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 5, NO. 1, JANUARY/FEBRUARY 1999
`
`Free-Space Micromachined Optical
`Switches for Optical Networking
`
`L. Y. Lin, Member, IEEE, E. L. Goldstein, Member, IEEE, and R. W. Tkach, Senior Member, IEEE
`
`Abstract— Fiber-optic switches with high port count have
`emerged as leading candidates for deployment in future optical
`transport networks, where restoration and provisioning in the
`optical
`layer will become increasingly important. This paper
`reviews the principle and performance of free-space microma-
`chined optical switches (FS-MOS) featuring free-rotating hinged
`micromirrors. A single-chip FS-MOS that implements the critical
`function of bridging—essential for restoration in core optical
`networks—is also proposed and demonstrated. The scalability of
`FS-MOS devices, and the dependence of their insertion losses
`on mirror-angle, are estimated theoretically. Simulation results
`suggest that the FS-MOS approach holds considerable promise
`for being expandable to the port-count values that will be needed
`in future core-transport lightwave networks.
`
`Index Terms— Free-space, micromachined, optical crosscon-
`nect, optical networking, optical switch, provisioning, restoration.
`
`I. INTRODUCTION
`
`DUE TO VERY rapid increases in core-transport-network
`
`demand, in the bit rates of individual services, and in the
`number of wavelength-channels being built into WDM trans-
`port systems, fiber-optic switches with large port count have
`quickly emerged as perhaps the most important yet-unrealized
`technological need in future high-capacity lightwave networks.
`These network elements will be chiefly used for network
`restoration, to begin with, with substantial provisioning value
`likely emerging thereafter. Optical switching technologies of-
`fer the potential advantages of bit-rate transparency, low power
`consumption, small volume, and low cost. Nevertheless, the
`requirements in port count (on the order of 512
`512 in three
`to five years) and loss budget represent deep challenges that
`have not yet been met by any current photonic switching
`technology. Although conventional mechanical switches can
`achieve high optical quality, they are large in size and mass,
`and are thus relatively slow in switching speed. On the
`other hand, guided-wave solid-state switches, though compact,
`generally have high loss and high cross talk. The inherent
`disadvantages of these technologies thus, appear to limit their
`expandability to the port counts mentioned above.
`By contrast, micromachined free-space optical-switching
`technology holds particular appeal
`in this application be-
`cause it combines the advantages of free-space intercon-
`nection—low loss and high optical quality—with those of
`monolithic integrated optics, namely, compactness. Various
`small-scale (2
`2) micromachined switches [1], [2] utilizing
`
`Fig. 1. Schematic drawing of the matrix free-space micromachined optical
`switch (FS-MOS).
`
`sliding micromirrors have been demonstrated. In addition,
`collimating optics and rotating micromirrors have also been
`proposed as a means of achieving high-density optical switches
`[3]–[5]. Given the fertility that
`the field of micro-optical
`systems is beginning to show, and the considerable variety
`of switching devices that has already emerged, it is likely that
`diverse applications will be best suited to diverse switching
`technologies. However, for the application of restoration and
`provisioning in core-transport lightwave networks, free-space
`micromachined optical switches (FS-MOS) with free-rotating
`hinged micromirrors are particularly attractive; this is because
`such applications do not require frequent switching, but do
`require very high reliability even for switch mirrors that remain
`in one switching state for extended periods on the order
`of years. Furthermore,
`the submillisecond switching times
`exhibited by FS-MOS devices are well matched to the needs
`of restoration and provisioning in core-transport lightwave-
`communications networks.
`In Section II, we will review the design and performance
`of FS-MOS devices. Their application to optical-network
`restoration will be proposed and demonstrated in Section III.
`In Section IV, we theoretically analyze the scalability of FS-
`MOS devices.
`
`II. DEVICE DESIGN AND PERFORMANCE
`The working principle of a matrix FS-MOS is shown in
`Fig. 1. The microactuated free-rotating mirrors are mono-
`lithically integrated on a silicon chip by means of surface-
`micromachining fabrication techniques. The collimated light
`is switched to the desired output port by rotating a selected
`mirror. Fig. 2 contains a schematic drawing of the actuated
`1077–260X/99$10.00 ª
`
`Manuscript received June 2, 1998; revised September 2, 1998.
`The authors are with AT&T Labs-Research, Red Bank, NJ 07701 USA.
`Publisher Item Identifier S 1077-260X(99)00644-9.
`
`1999 IEEE
`
`FNC 1029
`
`

`

`LIN et al.: FREE-SPACE MICROMACHINED OPTICAL SWITCHES FOR OPTICAL NETWORKING
`
`5
`
`Fig. 2. Schematic drawing of the microactuated free-rotating switch mirror.
`
`mirror. The microfabricated hinges [6] anchor the mirror on
`the Si substrate. The modified interleaved hinges constitute
`the hinge joints, through which the mirror is connected to the
`translation stage by push rods. The translation stage is actuated
`by arrays of scratch-drive actuators (SDA’s) [7]. The design
`efficiently converts the translation of the microactuators to
`rotation of the mirror.
`FS-MOS chips utilizing mirrors of the above-described
`structure have been fabricated using the MCNC MUMP’s
`(multiuser MEMS processes) fabrication process.1 The epi-
`taxial layers consist of one Si N layer for insulation; three
`polysilicon layers for ground-plane purposes (poly-0) and for
`mechanical structures (poly-1 and poly-2); two phosphosilicate
`glass layers for use as sacrificial material; and one gold
`layer for use in fabricating electrical contacts and mirror
`coatings. The mirror and the translation plate are built on the
`second polysilicon (poly-1) layer, and anchored to the substrate
`through hinge and guiding rail structures built on the third
`polysilicon layer (poly-2). The SDA’s are L-shaped polysilicon
`plates formed on poly-2. The hinge joints consist of interleaved
`poly-1/poly-2 structures, and are connected to the push rods
`built on poly-2. After the epitaxial growth,
`the sacrificial
`material
`is selectively etched by immersing the device in
`hydrofluoric acid,
`thus releasing the mechanical structures
`from the Si substrate. Fig. 3 shows the top-view photograph
`8 switch. The whole switch fabric occupies a chip
`of an 8
`1 cm .
`area of 1
`The free-rotating micromirror structures described above of-
`fer precisely the kinds of switching speeds that core-transport-
`network cross connects demand. Device measurements have
`shown submillisecond switching times when the SDA’s are
`100-V square wave at 500 kHz. The maximum
`actuated with
`mirror-rotation angle is defined by the dimension of the
`ground electrode under the translation stage. Fig. 4 shows
`the measured results for the switching-time response. The
`s for rotating the mirror from the
`switching time is 500
`OFF position to the ON position. To this, one must add a
`200- s delay between the application of the switching voltage
`and the onset of measurable optical-switching action. For the
`first demonstration, the mirror is rotated down by pulling the
`translation plate back with polysilicon springs, resulting in an
`
`1 Available www: http://www.mcnc.org.
`
`Fig. 3. Top view photograph of an 8  8 FS-MOS.
`
`Fig. 4. Switching response of the FS-MOS.
`
`OFF switching time of 560 s. In practical systems, the
`ON
`springs would likely be replaced by bidirectional SDA’s. The
`device’s switching curve shows an extinction ratio of more
`than 60 dB.
`4 FS-MOS [4], inte-
`In the previously demonstrated 4
`grated binary-amplitude Fresnel lenses were used for collimat-
`ing optical beams. Binary-amplitude Fresnel lenses possess
`the advantage of fabrication compatibility with MUMPS.
`Nevertheless, they resulted in high, nonuniform insertion loss.
`By employing fiber collimators for input and output coupling,
`losses as low as 3.1–3.5 dB are achieved for the shortest and
`8 switch, respectively. The
`longest optical paths in an 8
`fiber collimators consist of Grin lenses attached to single-mode
`fibers. The optical beam diameter at the fiber facet is 300 m.
`The measurement results show slightly higher loss than the
`theoretical calculations in Section IV. This may be attributed
`to the imperfect alignment of the fiber collimators and slight
`curving of the mirror surfaces. The flatness of the mirror
`surface can be improved by depositing thinner Au (currently
`5000 ˚A) on the polysilicon plate.
`
`

`

`6
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 5, NO. 1, JANUARY/FEBRUARY 1999
`
`Fig. 5. Polarization-dependent loss (PDL) of FS-MOS versus wavelength.
`The device’s PDL is below the noise floor of the measurement system.
`
`Fig. 7. Schematic of the FS-MOS incorporating the bridging function.
`
`Fig. 6. Schematic illustration of the bridging operation in core-transport
`networks.
`
`Unlike guided-wave approaches, these devices tend to ex-
`60 dB) between adjacent
`hibit very low cross talk values (
`channels without resorting to dilated architectures. More-
`over, one substantial advantage of the FS-MOS is its low
`polarizations-dependent loss (PDL). Fig. 5 shows PDL as a
`function of the wavelength used to illuminate the switch ports.
`The device’s PDL is seen to be smaller than the measurement
`system’s noise floor of about 0.3 dB. In general, the very
`high optical qualities of these switching structures highlight
`the inherent virtues of free-space interconnection approaches.
`
`III. APPLICATION IN NETWORK RESTORATION
`Any optical switching or cross connection network element
`used for core-transport network restoration must provide the
`critical function of bridging. This is illustrated in Fig. 6. Under
`normal operation [Fig. 6(a)], traffic traverses the service link.
`Upon link failure [Fig. 6(b)], it is switched to the restoration
`. Upon repair,
`path, sustaining an interruption of duration
`traffic must be bridged to both service and restoration links
`[Fig. 6(c)], in order to avoid a second interruption due to
`imperfectly synchronized head- and tail-end switching. What
`is critical is that the above bridging capability be incorporated
`into the optical switch with sufficient simplicity to avoid exac-
`
`Fig. 8. Top-view photograph of the FS-MOS configured to execute bridging.
`
`erbating the already-challenging scaling demands incumbent
`on it.
`We demonstrate here a means of accomplishing this for
`devices of the FS-MOS type. By integrating into the FS-MOS a
`few simple additional micro-optical elements, advanced func-
`tions like bridging can be incorporated at little cost in insertion
`loss or, ultimately, in scalability. Fig. 7 contains a schematic
`drawing of the FS-MOS with bridging. An additional row of
`beam-splitters, an additional column of free-rotating mirrors,
`and a fixed mirror are incorporated into the switch fabric to
`achieve the any-one-to-any-two bridging function. To accom-
`modate multiple simultaneous bridging operations, multiple
`copies of these elements could be integrated on a substrate.
`The beam-splitters are fabricated using the same process as the
`switch mirrors, except that they consist of a 1.5- m polysilicon
`plate without any reflective gold coating.
`The device’s operation is most readily apparent from Fig. 8.
`This is a top-view photograph of the FS-MOS, with array
`elements configured to execute bridging. The corner mirror (A)
`has been assembled and is held perpendicular to the substrate
`by spring latches [6] and precision side-latches [8]. Switch
`mirrors B and C and beam-splitter D have also been rotated
`
`

`

`LIN et al.: FREE-SPACE MICROMACHINED OPTICAL SWITCHES FOR OPTICAL NETWORKING
`
`7
`
`Fig. 9. Measured loss at 1.55 m through the beam-splitter and mirrors of
`the FS-MOS.
`
`up, perpendicular to the substrate. During normal operation,
`only switch mirror C is rotated up, providing a light path from
`the input to the output service link. Upon service-link failure,
`B and D rotate up, thus directing traffic to the restoration link.
`With C upright as well, traffic is bridged onto both service
`and restoration links. Finally, to carry out recovery, B and D
`are lowered, with C left upright, so that traffic again flows on
`the service link only.
`Among the chief determinants of scalability in these devices
`is loss. Here, the effect upon loss of the bridging structure is
`characterized. The insertion losses through beam-splitters and
`mirrors, for both service and restoration paths, are measured
`by coupling light at 1550 nm into the switch fabric using
`a fiber collimator. Losses along the optical paths were then
`measured with a large-area detector. Fig. 9 shows the results.
`Insertion losses along the beam-splitter’s transmission and
`reflection paths were 4.15 and 2.25 dB, respectively. The
`service path’s switch mirror adds an additional 0.1-dB loss.
`Along the restoration path,
`the loss from beam-splitter to
`output is 0.3 dB. The total excess loss for the two paths is 0.31
`dB, of which 0.09 dB results from scattering and absorption
`in the beam-splitter itself.
`
`IV. SCALABILITY ANALYSIS
`In this section, we analyze loss versus port count expected
`for FS-MOS-type devices, emphasizing the dependence of
`loss on optical beam size. We also analyze the dependence
`of loss on imperfect angular alignment in the switch mirrors
`and describe means of optimizing fiber-coupling efficiency.
`Gaussian-beam approaches are employed [9]. The analysis
`and simulation results described here are generally applicable
`not only to the FS-MOS, but also to optical matrix switches
`with free-space beam propagation. The detailed theoretical
`algorithm will be reported elsewhere.
`First, we assume that the optical beam that emerges from the
`input fiber has a minimum Gaussian-beam waist at the point of
`half-beam-width
`emergence. We characterize this by the
`at the facet of the input fiber. Due to its significance, it
`is useful to define a new variable given by the ratio of the
`
`(a)
`
`(b)
`
`Fig. 10. Loss versus number of pitches traveled for various mirror radii,
`assuming (a) a = 1:5, (b) a = 1, where a is the ratio of mirror radius to
`Gaussian beam waist.
`
`. Thus,
`divided by the beam waist
`switch-mirror radius
`, is a variable that can
`the dimensionless quantity
`.
`be optimized depending on the value of
`The coupling efficiency through the FS-MOS is obtained
`by calculating the overlap integral of the wave functions of
`the optical beam and the fiber mode at the receiving fiber
`facet. The optical beam diverges as it propagates through free
`space due to the finite beam waist, thus limiting the coupling
`efficiency. It is clear that the optimal size of the Gaussian
`beam will represent a compromise between competing forces.
`Although Gaussian beams with larger beam diameters ex-
`hibit smaller divergence angles, the mirror size that would
`be required increases with beam diameter. As a result, the
`propagation distance of the optical beam also increases for
`between
`a given device port-count. In particular, the pitch
`adjacent mirrors is geometrically related to the mirror radius
`by
`(in m) for the FS-MOS design described
`here. Fig. 10(a) shows the simulation results for various mirror
`, i.e., that the
`radii and beam sizes, assuming that
`mirror is 1.5 times as large as the beam waist. The variable
`labeling the abscissa in Fig. 10 represents the total propagation
`distance of the optical beam through the switch, measured
`
`

`

`8
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 5, NO. 1, JANUARY/FEBRUARY 1999
`
`(a)
`
`(b)
`
`Fig. 11. Highest loss in a (a) 32  32 and (b) 45  45 switch fabric versus
`a for various mirror radii.
`
`in number of mirror pitches or, roughly, rows plus columns.
`The distance between the input/output fiber facet and the
`. Therefore,
`is
`closest mirror is assumed to be
`the number of pitches corresponding to the longest path in
`32 switch. The coupling loss in Fig. 10 is seen to
`a 32
`decrease rapidly as the mirror/beam size increases, suggesting
`that beam divergence is the dominating factor in this region.
`Nevertheless, the effect saturates as the mirror radius increases
`beyond 400 m.
`unchanged, the Gaussian-
`While keeping the mirror radius
`can be increased to reduce the divergence
`beam waist
`effect. The crosstalk induced by scattering of the optical beam
`from the mirror edge is expected to be negligible in most
`cases where separation between adjacent mirrors is sufficient.
`, that is, for
`Fig. 10(b) shows the simulation results for
`mirror radius equal to beam waist. Compared to Fig. 10(a),
`the loss is greatly reduced for small mirrors. As the optical
`beam cannot be captured completely by the reflecting mirror
`when it arrives at the mirror surface due to beam divergence,
`1.3 dB appears in the results.
`a loss floor of
`,
`This suggests the existence of an optimal value of
`as is evident from the simulation results shown in Fig. 11.
`
`Fig. 12. Additional loss due to mirror-angle variation for the longest path in
`a 32  32 switch fabric.  is twice the mirror-angle variation. The values of
`a have been optimized for different mirror radii.
`
`and 89,
`Fig. 11(a) and (b) is calculated for the cases of
`that is, for the longest optical paths and, therefore, highest
`32 and 45
`45 switch, respectively. The
`losses in a 32
`increases as the mirror radius increases.
`value of the optimal
`This is because the effect of optical-beam divergence is more
`significant for small beam sizes.
`The minimum loss for the longest optical path through
`512 or 1024
`1024 switch fabric can,
`therefore,
`a 512
`be obtained by simulation. We assume for illustration the
`use of three-stage Clos switch-fabric architecture [10] as a
`means of constructing large-port-count switches using smaller
`constituent switch modules. Assuming the input port count of
`, in order to achieve
`each switch module in the first stage is
`the property of strict nonblocking, the output port count of the
`. This is also
`switch module needs to be greater than
`the required input and output port count for the switch module
`in the second stage. The switch module in the third stage is a
`mirror image of that in the first stage. Therefore, to construct
`512 switch fabric, one can employ 32 layers, with
`a 512
`32, 32
`32, and 32
`16 switch modules comprising the
`16
`1024 switch fabric, 45 layers of
`three stages. For a 1024
`45, 45
`45, and 45
`23 switch modules may be used.
`23
`With these assumptions, using the approach described above,
`it can be shown that a mirror radius of 200 m achieves a
`512 switch
`5.8 dB loss through the longest path of a 512
`fabric. Similarly, a mirror radius of 250 m achieves 5.8-dB
`1024 switch fabric.
`loss through the longest path of a 1024
`In each of these cases, the switch’s second stage imposes loss
`values indicated in Fig. 11; the first and third stages impose
`smaller losses due to their smaller size.
`The results described above assume a perfectly aligned
`optical system. To characterize imperfectly aligned systems,
`the dependence of coupling efficiency on the angular error in
`mirror alignment may be captured by performing a coordinate-
`system transformation on the wave function of the propagating
`Gaussian beam. Fig. 12 shows the additional loss versus twice
`the mirror-angle variation for various mirror radii, assuming
`. The value
`the traversed number of pitches is fixed at
`has been optimized individually for different mirror radii
`of
`
`

`

`LIN et al.: FREE-SPACE MICROMACHINED OPTICAL SWITCHES FOR OPTICAL NETWORKING
`
`9
`
`functions of this type can be incorporated. Theoretical scal-
`ability suggests that FS-MOS-type switches with port counts
`greater than 512 should be achievable with reasonable values
`of insertion loss. In view of these virtues, with compactness
`conferred by integrated micro-fabrication, and high optical
`quality derived from the use of free-space optics, devices of
`the type described here appear to hold substantial promise as
`restoration and provisioning vehicles in high-capacity core-
`transport lightwave networks.
`
`ACKNOWLEDGMENT
`The authors would like to thank A. Saleh and J. Simmons
`for helpful discussions, and JDS Fitel for collaboration on
`packaging issues.
`
`REFERENCES
`
`[1] S. S. Lee, E. Motamedi, and M. C. Wu, “Surface-micromachined free-
`space fiber optic switches with integrated microactuators for optical fiber
`communication systems,” in Transducers’97, Chicago, IL, June 16–19,
`1997.
`[2] C. Marxer, M. A. Cr´etillat, N. F. de Rooij, R. B¨attig, O. Anthamattern, B.
`Valk, and P. Vogel, “Vertical mirror fabricated by reactive ion etching
`for fiber optical switching applications,” in IEEE Workshop on Micro
`Electro Mechanical Systems, Nagoya, Japan, Jan. 26–30, 1997.
`[3] H. Toshiyoshi and H. Fujita, “Electrostatic micro torsion mirrors for an
`optical switch matrix,” J. Microelectromech. Syst., vol. 5, pp. 231–237,
`1996.
`[4] L. Y. Lin, E. L. Goldstein, and R. W. Tkach, “Free-space micromachined
`optical switches with submillisecond switching time for large-scale op-
`tical crossconnects,” IEEE Photon. Technol. Lett., vol. 10, pp. 525–527,
`1998.
`[5] L. Y. Lin, E. L. Goldstein, J. M. Simmons, and R. W. Tkach, “High-
`density connection-symmetric free-space micromachined polygon opti-
`cal crossconnects with low loss for WDM networks,” in Conf. Optical
`Fiber Communications, San Jose, CA, Feb. 22–27, 1998, postdeadline
`paper.
`[6] K. S. J. Pister, M. W. Judy, S. R. Burgett, and R. S. Fearing, “Mi-
`crofabricated hinges,” Sensors and Actuators A, vol. 33, pp. 249–256,
`1992.
`[7] T. Akiyama and H. Fujita, “A quantitative analysis of scratch drive
`actuator using buckling motion,” in IEEE Workshop Micro Electro
`Mechanical Systems, Amsterdam, The Netherlands, Jan. 29–Feb. 2,
`1995.
`[8] L. Y. Lin, S. S. Lee, K. S. J. Pister, and M. C. Wu, “Micro-machined
`three-dimensional micro-optics for integrated free-space optical system,”
`IEEE Photon. Technol. Lett., vol. 6, pp. 1445–1447, Dec. 1994.
`[9] J. T. Verdeyen, Laser Electronics, 2nd ed. Prentice Hall, 1989.
`[10] C. Clos, “A study of nonblocking switching networks,” Bell Syst. Tech.
`J., vol. 32, pp. 406–424, 1953.
`
`L. Y. Lin (S’93–M’96), photograph and biography not available at the time
`of publication.
`
`E. L. Goldstein (S’84–M’89), photograph and biography not available at the
`time of publication.
`
`R. W. Tkach (M’84–SM’98), photograph and biography not available at the
`time of publication.
`
`Fig. 13. Loss versus number of pitches traveled for various mirror radii with
`optimized a using focusing fiber collimators.
`
`. The system’s sensitivity of angular misalignment is seen to
`decrease rapidly as the optical beam size decreases. For
`m, for example, the misalignment penalty can be held to
`3 dB if the mirror-angle variation is held smaller than
`values
`0.08 . Experimental studies of both the mirror-angle variation
`of the FS-MOS and of additional mechanical structures for
`achieving enhanced angular precision are currently underway.
`All results so far assume that the system is configured so
`that the input beam waists reside at the facets of the input
`fiber collimators. In practice, however, various modifications
`can be made to the fiber collimators so as to reduce the
`coupling loss suffered by the systems due to beam divergence.
`For example, the focal length of the input fiber collimators
`can be designed such that the minimum Gaussian-beam waist
`is located at the facet of the receiving fiber. The coupling
`efficiency will then be optimized for certain input/output fiber
`pairs. Fig. 13 shows loss versus traveling distance for various
`, assuming that
`the input fiber collimators
`mirror radii
`have again been
`are the focusing variety. The values of
`optimized individually. The fiber collimators are optimized for
`the longest achievable focal length within the diffraction limit
`. By comparing Fig. 13 with Fig. 11(a),
`it is seen that the maximum coupling loss through the switch
`fabric can be substantially reduced by exploiting this approach.
`
`V. CONCLUSION
`FS-MOS featuring free-rotating micromirrors have been
`proposed for use as restoration and provisioning vehicles
`in core-transport lightwave networks. Several of their more
`important properties have been characterized experimentally
`and theoretically. This technology is seen to achieve low
`loss,
`low crosstalk,
`low polarization dependence, and fast
`switching speed. Moreover, a single-chip switch fabric incor-
`porating bridging functionality for optical network has been
`demonstrated, showing the simplicity with which advanced
`
`