(12) United States Patent
`(10) Patent N0.:
`US 6,442,307 B1
`
`Carr et al.
`(45) Date of Patent:
`Aug. 27, 2002
`
`USOO6442307B1
`
`(54) SOLDER-PACKAGED OPTICAL MEMS
`211310111ch AND METHOD FOR MAKING THE
`
`(75)
`
`Inventors: Dustin W. Carr, Pittstown; Dennis S.
`Greywall, Whitehouse Station; Sungho
`Jin, Millington; Flavio Pardo, New
`Providence; Hyongsok Soh, Basking
`Rldge, all Of NJ (US)
`(73) Assignees: Lucent Technologies Inc., Murray Hill,
`NJ (US); Agere Systems Guardian
`Cor ., Orlando, FL US
`p
`(
`)
`Subject to any disclaimer, the term of this
`patent is extended or adjusted under 35
`U.S.C. 154(b) by 68 days.
`
`( * ) Notice:
`
`21 A 1. N .2 09 705 350
`')
`pp
`0
`/
`’
`(
`(22)
`Filed:
`NOV. 3, 2000
`
`7
`{'11: (3] “““““““““““““““““‘3‘85/18385/120529/62/45
`E23
`(58) Field of Search ...................... 385/14, 18; 359/245
`
`(56)
`
`References Cited
`US. PATENT DOCUMENTS
`5,444,520 A *
`8/1995 Murano ...................... 347/244
`
`..
`6,002,507 A * 12/1999 Floyd et a1.
`359/201
`
`3/2001 Greywall .............. 335/222
`6,201,631 B1 *
`............... 372/20
`6,351,577 B1 *
`2/2002 Aksyuk et al.
`* cited by examiner
`Primary ExaminergRobert H. Kim
`Assistant Examiner—George Wang
`(57)
`ABSTRACT
`In accordance with the invention, a MEMS mirror device
`comprises a mirror layer including a frame structure and at
`least one mirror movably coupled to the frame and an
`actuator la er includin at least one conductive
`ath and at
`y
`g
`.
`.
`P.
`least one electrode for mov1ng the mirror. The mirror layer
`and the actuator layer are provided With metallization pads
`and are bonded together in lateral alignment and with
`predetermined vertical gap spacing by solder bonds between
`the pads. The device has utility in optical cross connection,
`variable attenuation and power gain equalization.
`32 Claims, 12 Drawing Sheets
`
`
`
`FNC 1005
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 1 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 2 0f 12
`
`US 6,442,307 B1
`
`
`
` INPUT/ OUTPUT
`FIBER BUNDLES
`
`
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 3 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 4 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 5 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 6 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 7 0f 12
`
`US 6,442,307 B1
`
`FIG.
`
`5A
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 8 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 9 0f 12
`
`US 6,442,307 B1
`
`FIG.
`
`7A
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 10 0f 12
`
`US 6,442,307 B1
`
`
`
` '
`
`
`"um—wwn.-.,..,.,
`gag—h_
`__
`__
`
`
`8080‘25 50
`
`
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 11 0f 12
`
`US 6,442,307 B1
`
`
`
`

`

`US. Patent
`
`Aug. 27, 2002
`
`Sheet 12 0f 12
`
`US 6,442,307 B1
`
`158
`
`155
`
`
`
`

`

`US 6,442,307 B1
`
`1
`SOLDER-PACKAGED OPTICAL MEMS
`DEVICE AND METHOD FOR MAKING THE
`SAME
`
`FIELD OF THE INVENTION
`
`The present invention relates to optical micro-electro-
`mechanical devices (“MEMs devices”) and, in particular, to
`MEMs devices constructed using solder bond assembly. The
`inventive MEMs devices are particularly useful as movable
`mirrors in optical communication systems.
`BACKGROUND OF THE INVENTION
`
`Optical MEMs devices are of considerable importance in
`optical communication systems.
`In one important
`application, a two-dimensional array of MEMs devices is
`used to provide an optical cross connect between input
`optical fibers and output optical fibers. Each MEMs device
`in the array is a movable mirror disposed to receive the
`optical signal from an input fiber. The mirror can be elec-
`trically moved to reflect
`the received optical
`input
`to a
`desired output fiber.
`A typical MEMs mirror comprises a metal-coated silicon
`mirror movably coupled to a surrounding silicon frame via
`a gimbal. Two torsional members on opposite sides of the
`mirror connect the mirror to the gimbal and define the
`mirror’s axis of rotation. The gimbal, in turn, is coupled to
`the surrounding silicon frame via two torsional members
`defining a second axis of rotation orthagonal to that of the
`mirror. A light beam can therefore be steered in any direc-
`tion.
`
`Electrodes are disposed in a cavity underlying the mirror
`and the gimbal. Voltages applied between the mirror and an
`underlying electrode and between the gimbal and an elec-
`trode control the orientation of the mirror. Alternatively, in
`slightly modified arrangements, an electrical signal can
`control the position of the mirror magnetically or piezoelec-
`trically.
`FIGS. 1(a) and 1(b) illustrate conventional optical MEMs
`devices and their application. FIG. 1(a) shows a typical prior
`art optical MEMs mirror structure. The device comprises a
`mirror 10 coupled to a gimbal 11 on a polysilicon frame 12.
`The components are fabricated on a substrate (not shown) by
`micromachining processes such as multilayer deposition and
`selective etching. After etching, mirror assembly (10, 11, 12)
`is raised above the substrate by upward bending lift arms 13
`during a release process. The mirror 10 in this example is
`double—gimbal cantilevered and attached onto the frame
`structure 12 by springs 14. The mirrors can be tilted to any
`desired orientation for optical signal routing via electrostatic
`or other actuation with electrical voltage or current supplied
`from outside. The light-reflecting surface of mirror 10
`comprises a metal coated polysilicon membrane which is
`typically of circular shape. The metal layers are deposited by
`known thin film deposition methods such as evaporation,
`sputtering, electrochemical deposition, or chemical vapor
`deposition.
`FIG. 1(b) schematically illustrates an important applica-
`tion of optical MEMs mirrors as controllable mirror arrays
`for optical signal routing. The cross connect system shown
`here comprises optical input fibers 120, optical output fibers
`121 and an array of MEMs mirrors 122 on a substrate 123.
`The optical signals from the input fibers 120 are incident on
`aligned mirrors 122. The mirrors 122, with the aid of a fixed
`auxilliary mirror 124 and appropriate imaging lenses 125,
`are electrically controlled to reflect
`the incident optical
`signals to respective output
`fibers 121.
`In alternative
`
`2
`schemes, the input fibers and the output fibers are in separate
`arrays, and a pair of MEMs mirror arrays are used to perform
`the cross connect function.
`
`The tilting of each mirror is controlled by applying
`specific electric fields to one or more of the electrodes (not
`shown) beneath the mirror. Undesirable variations in the gap
`spacing between the mirror layer and the electrode layer,
`symmetric or nonsymmetric, alter the electric field for the
`applied field, which affects the degree of electrostatic actua-
`tion and hence the degree of mirror tilting. This in turn alters
`the path or coherency of light signals reaching the receiving
`fibers, thus increasing the signal loss during beam steering.
`An array of such MEMs mirrors is essentially composed
`of two layers: a mirror layer comprising the array of mirror
`elements movably coupled to a surrounding frame and an
`actuator layer comprising the electrodes and conductive
`paths needed for electrical control of the mirrors. One
`approach to fabricating the array is to fabricate the actuator
`layer and the mirror layer as successive layers on the same
`workpiece and then to lift up the mirror layer above the
`actuator layer using vertical thermal actuators or stresses in
`thin films. This lift-up process is described in US. patent
`application Ser. No. 09/415,178 filed by V. A. Aksyuk et al.
`on Nov. 8, 1999 and assigned to applicant’s assignee.
`An alternative approach is to fabricate the mirror layer on
`one substrate, the actuator layer on a separate substrate and
`then to assemble the mating parts with accurate alignment
`and spacing. The two-part assembly process is described in
`US. Pat. No. 5,629,790 issued to Neukermans et al. on May
`13, 1997 and US. patent application Ser. No. 09/559,216
`filed by Greywall on Apr. 26, 2000, both of which are
`incorporated herein by reference. This two-part assembly
`provides a more robust structure, greater mirror packing
`density and permits larger mirror sizes and rotation angles as
`well as scalability to larger arrays.
`In the two—part assembly process, the mirror layer—and the
`actuator layer should be bonded for mechanical sturdiness
`and long-term reliability. Neukermans et al. and Greywall
`suggest anodic bonding, solder glass bonding, and epoxy
`bonding. The gap spacing between the mirror layer and the
`actuator layer determines the electric field for the given
`magnitude of applied voltage (or the magnetic field for the
`given electrical current level). Therefore, an accurate and
`reliable establishment of the gap spacing during the assem-
`bly and bonding of the two parts, as well as the dimensional
`stability of the gap during device handling, shipping and
`operation are important. The accurate lateral alignment of
`the mating parts of the mirrors and electrodes is also
`desirable for reliable operation.
`SUMMARY OF THE INVENTION
`
`In accordance with the invention, a MEMs mirror device
`comprises a mirror layer including a frame structure and at
`least one mirror movably coupled to the frame and an
`actuator layer including at least one conductive path and at
`least one electrode for moving the mirror. The mirror layer
`and the actuator layer are provided with metallization pads
`and are bonded together in lateral alignment and with
`predetermined vertical gap spacing by solder bonds between
`the pads. The device has utility in optical cross connection,
`variable attenuation and power gain equalization.
`BRIEF DESCRIPTION OF THE DRAWINGS
`
`10
`
`15
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`The nature, advantages and various additional features of
`the invention will appear more fully upon consideration of
`the illustrative embodiments now to be described in detail
`
`with the accompanying drawings. In the drawings:
`
`

`

`US 6,442,307 B1
`
`3
`illustrate conventional MEMs
`FIGS. 1(a) and 1(b)
`devices and their application;
`FIGS. 2(a) and 2(b) schematically illustrate a first
`embodiment of a MEMs device in accordance with the
`invention;
`FIGS. 3(a) and 3(b) illustrate steps in a first method of
`fabricating the FIG. 2 device;
`FIGS. 4(a), 4(b) and 4(6) illustrate steps in a second
`method of fabricating the device using a removable spacer;
`FIGS. 5(a) and 5(b) illustrate exemplary, removable spac-
`ers suitable for solder bond height control;
`FIGS. 6(a) and 6(b) illustrate fabricating a MEMs device
`incorporating a permanent spacer;
`FIGS. 7(a) and (b) schematically illustrate two exemplary
`embodiments of the MEMS device according to the inven-
`tion;
`FIGS. 8(a) and 8(b) illustrate additional examples of
`MEMS devices using spacers;
`FIG. 9 schematically illustrates a variable attenuator com-
`prising the optical MEMS structure according to the inven-
`tion; and
`FIG. 10 schematically illustrates an optical communica-
`tion system comprising a dynamic gain equalizer based on
`optical MEMS device according to the invention.
`It is to be understood that the drawings are for purposes
`of illustrating the concepts of the invention and are not to
`scale.
`
`DETAILED DESCRIPTION OF THE
`INVENTION
`
`Referring to the drawings, FIG. 2(a) illustrates an exem-
`plary optical MEMs device in accordance with the invention
`comprising a mirror layer 20 including at least one gim-
`balled mirror 21 movably coupled to the surrounding frame
`22 and a mating actuator layer 23 including at least one
`actuating electrode 24 and leads (not shown) for applying
`voltage between electrode 24 and gimballed mirror 21.
`Layers 20 and 23 are advantageously polysilicon or single
`crystal silicon. The mating surfaces of the layers are each
`provided with a plurality of wettable metallization regions
`25 and solder ball bonds 26 between metallization regions
`25. The layers are spaced apart by the solder balls 26. The
`gimballed mirror 21 can be a double-gimballed, cantilevered
`mirror and is coupled to the frame structure 22 of layer 20
`by torsion bars or springs (not shown). Hence the mirror 21
`can be tilted to any desired orientation.
`Each mirror 21 can be electrically grounded and tilted for
`optical signal routing via electrostatic actuation by one or
`more electrodes 24 placed underneath the mirror. An exem-
`plary desired size (effective diameter) of mirrors suitable for
`optical MEMs applications is in the range of 50—10000 Mm,
`preferably in the range of 200—2000/1m. The mirrors can be
`coated with a reflecting metal layer comprising Au, Ag or Al.
`The use of other metals or alloys is not precluded.
`The mirror layer 20 and the actuator layer 21 are solder
`bonded together with accurate lateral alignment and with
`accurate gap spacing. The mirror layer 20 typically com-
`prises an array of mirrors 21. The two-part structure is
`conveniently fabricated using a single-crystal silicon wafer
`fabrication process, although a polysilicon surface micro-
`machining can also be used for providing the separate mirror
`and the electrode parts. See the aforementioned U.S. patent
`application Ser. No. 09/559,216 to Greywall.
`FIG. 2(b) illustrates a two-dimensional array of such
`movable gimballed mirrors 21. The mirror elements 30 are
`
`10
`
`15
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`4
`attached to gimballs 31 by torsion bars 32. The gimballs, in
`turn, are attached by torsion bars 33 to a frame 34 The
`mirrors can be patterned on a single layer of substrate, for
`example, an SOI (silicon-on-insulator) single crystal silicon
`wafer using conventional silicon fabrication processes such
`as lithographic patterning, chemical etching, or reactive ion
`etching. The open areas 35 between the gimballed mirrors
`can be utilized to place solder bonds as shown in FIG. 2(a).
`This solder bonded MEMs device has many processing
`and performance advantages as compared with conventional
`devices. The solder bonding, as shown herein, can provide
`highly accurate spacing of the gap between the layers 20, 23
`and facilitate highly accurate lateral alignment. Such a
`structure permits precise control of the mirrors, a more
`robust structure, greater packing density, larger mirror sizes,
`and larger mirror rotation angles than are conventionally
`obtained and easier electrical connection to the mirrors.
`
`FIGS. 3(a) and 3(b) illustrate steps in assembly of the
`optical MEMS device shown in FIG. 2(a). The mirror layer
`20 and the actuator layer 23 can be accurately laterally
`aligned, e.g., by using a micro aligner (not shown) for x-y
`positioning. The vertical gap can be accurately controlled,
`e.g., by using a z-axis position sensor and micro-positioner
`(not shown), optionally guided by camera vision sensor, or
`capacitive gap sensor. The solder bonding operation may be
`carried out while the pre-aligned lateral and vertical posi-
`tions are held in position until the solder solidification is
`completed. The thermal contraction of solder joint height is
`desirably considered in determining the z-axis gap position-
`mg.
`
`The layers 20, 23 are typically comprised of solder
`non-wettable surfaces such as silicon, silicon oxide. In order
`to make these surfaces solderable, underbump metallization
`(UBM) coatings 25 are desirable. The UBM layers for the
`optical MEMS assembly advantageously comprise at least
`one adhesion layer such as Ti or Cr with an optional CriCu
`gradient alloy layer, at least one solderable metal layer such
`as Cu or Ni, and at least one surface oxidation-resistant layer
`such as Au, Pd, Pt or Ag. The use of one or more diffusion
`barrier layers to prevent or minimize undesirable inter-layer
`metallurgical reactions is not precluded.
`The preferred thickness of the adhesion layer is in the
`range of 2—500 nm, preferably 20—200 nm. The preferred
`thickness of the solderable layer is in the range of 50—5000
`nm, preferably 100—1000 nm. The desired thickness of the
`oxidation-resistant protective layer is in the range of 10—500
`nm, preferably 20—200 nm. The UBM coatings 25 are
`typically applied to the surfaces of both the mirror layer and
`the actuation layer facing each other, at locations between
`the mirrors, Patterning of the UBM coating 25 can be done
`using known techniques such as lithography or deposition
`through shadow masks with desired pattern geometry.
`The solder alloys may be applied onto the UBM surface
`as a paste, preform or deposited film, e.g., prepared by
`sputtering, evaporation, or electrodeposition. The solder
`material is then reflowed (melted by heating) so as to form
`well-defined, near-spherical solder balls 26 as shown in FIG.
`3(a) solder bumps. Either the mirror layer, the actuator layer,
`or both contain these solder bumps.
`The mirror and the actuator layers are brought together
`into contact and heated to produce solder joints and
`mechanically attach the two parts as shown in FIG. 3(b). A
`solder flux is optionally used to aid the soldering process.
`For the optical MEMS assembly, a vapor flux such as formic
`acid carried by nitrogen or argon gas is preferred over liquid
`flux which may get trapped in tiny gaps near the mirror or
`
`

`

`US 6,442,307 B1
`
`5
`spring hinge structure. Examples of solder alloys suitable for
`the inventive assembly of the two-part optical MEMS
`devices include 37% Pb-63% Sn (by weight) eutectic solder
`(with a melting point (m.p.) of ~183° C.), high-lead 95Pb-
`5Sn solder (m.p. ~308° C.), Sn-3.5Ag (m.p. ~221° C.),
`57Bi-43Sn (mp. ~139° C.), 95$n-58b (mp. ~245° C.),
`49Sn-51In (mp. ~137° C.), and 80Au-20Sn (mp. ~280° C.).
`The gap spacing between the mirror layer 20 and the
`actuator layer 23 is important for reproducible actuation of
`the optical MEMS mirrors. As an alternative to mechanical
`positioning and soldering, the spacing may be controlled by
`careful selection of solder volume applied onto each UBM
`pad, the area of the pad, the weight applied over the top of
`the assembly during the soldering, and the temperature and
`time of soldering. Larger volume, smaller pad area, smaller
`weight applied, and lower soldering temperature tend to
`produce taller solder joints. The solder joints so produced
`typically exhibit a truncated sphere shape.
`One may approximate the relation between vertical gap
`spacing and solder volume by assuming that the solder joint
`has the form of a sphere of radius R truncated at the planes
`of contact with the metallization pads. In an exemplary
`device we may assume that the diameter of the circle of
`contact (plane of truncation) is approximately the height of
`the joint, then the joint height h=V2R. The joint height h can
`then be correlated to the volume of the truncated solder
`
`(0.884X4/3TI3R3) with a ratio V2R/(3.701R3). Thus by con—
`trolling the volume of the solder on the metallization pads,
`the joint height can be controlled. The soldering process
`onto mating pad areas has a self aligning effect due to the
`surface tension of molten solder and the driving force to
`reduce the surface energy of the molten solder at the joint.
`The mirror layer above and the actuator layer below are thus
`pulled together while the mating pads are automatically
`laterally aligned. For even more accurate lateral alignment,
`supplemental mechanical alignment may optionally be
`employed, e.g., by utilizing alignment of notches, edges or
`corners against fixed-positioned notches or steps in the
`assembly.
`The attainable accuracy in the vertical gap spacing in the
`inventive solder assembly is with less than 110% variation
`from the preferred nominal spacing, preferably with less
`than 15% variation, and even more preferably with less than
`12.5% variation. For example, 600 um diameter mirrors may
`operate with a mirror-electrode spacing of 60 um The
`attainable variation from this 60 gm spacing is less than 16
`,urn, preferable less than 13 um, even more preferably less
`than 11.5 ,um. The lateral alignment of mirror position vs
`electrode position is also critical as a deviation from desired
`alignment may lead to an undesirable deviation in the
`intensity of electrostatic (or magnetic) actuation of the
`MEMS device. The attainable lateral alignment
`in the
`inventive, solder-packaged optical MEMS device is with an
`accuracy of within 15%, preferably within 11% of the
`effective mirror diameter. The effective mirror diameter of a
`
`non-circular mirror is the diameter of a circle having the
`same area as the mirror.
`
`FIGS. 4(a), 4(b) and 4(6) schematically illustrate the use
`of a removable spacer 40 to further enhance the accuracy of
`the gap spacing in a solder-assembled,
`two-part optical
`MEMS device. With spacer 40, an accuracy of within 15%,
`preferably within 12.5% of a desired gap spacing can be
`obtained. In FIG. 4(a),
`the UBMs and solder bumps are
`provided on selected sites of between-mirror spaces with
`certain regular or irregular intervals. A removable spacer
`assembly is prepared with a pre-determined height
`(or
`thickness), desirably with less than 11% and preferably with
`less than 10.5% deviation from the target spacing.
`
`10
`
`15
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`6
`FIG. 4(1)) shows the mirror layer 20 laterally aligned with
`the actuator layer 23 and placed on the movable spacers 40
`to provide accurate vertical spacing. The assembly is then
`heated to solder the layers 20, 23 together.
`After soldering, the removable spacers 40 are removed
`(physically or by etching), leaving the structure shown in
`FIG. 4(6).
`The removable spacer is preferably an assembly 50 of a
`periodic and parallel array of spacers 40. FIGS. 5(a) and 5(b)
`illustrate exemplary assemblies 50 with connecting ends 51.
`With such assemblies, a connecting end 51 can be cut and
`using the other connecting end, the spacers 40 can be pulled
`out
`in a linear
`fashion without disturbing any of the
`mechanically fragile mirrors.
`The spacer assembly 50 is placed over the actuator layer
`23 which already contains solder bumps at specific sites. The
`location of each of the spacers 40 is selected in such a way
`that they do not overlap with the existing solder bumps. The
`mirror layer 20 is then placed above the removable spacer
`assembly, laterally aligned with respect to the actuator layer.
`Then the entire assembly is heated and solder bonded while
`a weight or pressure is optionally being applied so that the
`solder height turns out to be exactly the same as the spacer
`height. Even in the absence of the applied pressure, the
`solder surface tension can be utilized to vertically pull the
`mirror and the electrode layer closer until they both touch
`the removable spacer.
`After the soldering operation, one edge of the spacer
`assembly is cut off, and the spacers are mechanically pulled
`out,
`leaving the final configuration of FIG. 4(c).
`Alternatively, although less preferable, the spacers can be
`chemically etched away or burned away if the spacer mate-
`rial is chosen such that the etching or burning process does
`not damage the mirror and associated components.
`Solder alloys generally exhibit very high coeffecient of
`thermal expansion (CTE), especially those containing a
`large percentage of Pb or In. The CTE values of solder alloys
`can be as high as 23—28X10'6/degree C as compared to
`~4x10'6/degree C for Si, ~13x10‘6/degree C for nickel, and
`~18x10’6/degree C for copper alloys. If the spacers of FIG.
`4 have a much lower CTE value than the solder contracts,
`then the spacers advantageously should be removed after the
`soldering operation is completed. Repeated changes in tem-
`perature environment during service, storage or shipping of
`the packaged optical MEMS devices might otherwise result
`in a premature fatigue failure of the solder joints.
`The solidification shrinkage combined with the generally
`greater thermal contraction of the solder joint material as
`compared to the thermal contraction of the spacer material
`may result in a trapping of the spacer. In order to allow an
`easy removal of the spacer by horizontal pull out without
`seriously disturbing or damaging the delicate mirror and
`associated structures, one or more of the following
`approaches can be utilized.
`1. The spacer can be made up of round or oval crosssec-
`tioned wires with minimal contact with the upper mirror
`layer and the lower electrode layer, and be coated with
`low-friction-coeffecient material such as a fluorocarbon, a
`diamond or diamond-like-carbon film so that the mechanical
`pulling out of the spacer becomes easier.
`2. The spacer can be made of a round tube material with
`a hollow core instead of solid wire material so that the spacer
`material has some vertical compliancy and is more easily
`removed by horizontal pulling.
`3. The spacer can be made of fine metallic tubes with a
`round or preferably an oval crosssection into which finer,
`
`

`

`US 6,442,307 B1
`
`7
`round wires are inserted. When the soldering operation is
`completed, the inner round wires are pulled out first, which
`will then make the tube collapse into an oval shape with less
`height than the solder joint height. This allows easy removal
`of the remaining spacer frame material.
`4. The material used for the spacer can be selected so that
`it has higher CTE value than the solder material used, and
`of course substantially higher melting point than that for the
`solder. On cooling to room temperature (or to even below
`room temperature if necessary) after soldering operation, the
`spacer then thermally contracts more than the solder joint
`height, loosens, and is easily removed. The desired CTE
`value for such thermal-contraction-removable spacer mate-
`rial is at least 10% higher, preferably at least 25% higher
`than the CTE of the particular solder used for the packaging.
`For example, an aluminum based alloy spacer with a CTE of
`~23x10'6/degree C can be utilized if the solder used is 57%
`Bi—43% Sn eutectic solder with the CTE of ~17.5><10‘6/
`degree C, a magnesium-based alloy spacer with CTE values
`of ~26x10‘6/degree C can be utilized if the solder used is
`Sn-3.5% Ag eutectic with a CTE of 23x10‘5/degree C, and
`a Zn-1% Cu alloy spacer with a CTE of ~35x10'5/degree C
`may be utilized if 37Pb-6SSn eutectic solder with a CTE of
`~25x10'fi/degree C is used. The Bi—Sn solders with high Bi
`contents have a unique and advantageous behavior of
`expanding upon solidifacation thus making the removal of
`the spacer easier. The desired high CTE spacer material does
`not have to be restricted to metallic materials. For example,
`a spacer of the type shown in FIG. 5(a) or 5(b) may be made
`of a polyimide, thermosetting epoxy or other plastic material
`which remains stable at the typical soldering temperature of
`100—300° C., and exhibit high CTE of often greater than
`50x10'6/degree C and consequently a large thermal
`contraction, thus allowing easy removal of the spacer after
`soldering operation. 5. The spacer can be made of a shape
`memory alloy such as the equiatomic Ni—Ti alloy, and
`alloys in general which undergo phase transformation and
`associated volume change near room temperature heating
`and cooling. The crosssection of the wire part of the shape
`memory spacer is made to have a collapsed oval shape at
`room temperature, which is converted to a taller crosssection
`by phase transformation when heated to the soldering
`temperature, which then returns to the original lower height
`when cooled back to room temperature. This allows easy
`removal from the solder-bonded optical MEMS package.
`FIGS. 6(a) and 6(1)) illustrate an alternative embodiment
`of the invention where the gap-defining spacer can be a
`permanent drop—in spacer 60 which is permanently left
`inside the solder-bonded optical MEMS package. In this
`case, the coefficients of thermal expansion (CTEs) of the
`spacer material and the solder material are carefully
`matched. Conveniently, the permanent spacer 60 is simply
`dropped-in into the gap and laterally aligned just prior to the
`soldering operation, and the spacer removal step can be
`omitted. Advantageously, the actuator layer 23 can be pro-
`vided with notches 61 to facilitate lateral alignment of the
`spacer 60.
`An important and preferred configuration of a drop-in
`(permanent) spacer 60 simultaneously incorporates aerody-
`namic isolation for of each mirror in the array. The move-
`ment of one mirror, for example, electrostatically actuated
`tilting for optical signal re-routing, can disturb or cause the
`movement of adjacent mirrors. The drop-in spacer is pref-
`erably configured so that the boundary area between adja-
`cent mirrors is compartmented or blocked as much as
`possible. Blockage may be defined as the proportion of the
`360° periphery underlying the mirrors that is blocked off by
`
`8
`the spacer. The preferred blockage is 011 the average, with at
`least 20%, preferably 50%, even more preferably 80% of the
`total boundary wall area. After solidification of the bonding
`solder, the gap spacing between the mirror and the electrode
`layer is dictated by the pre-set
`thickness of the drop-in
`spacer. The relative adjustment of CTE values of the spacer
`and the solder by suitable materials selection in combination
`with the consideration of solidifacation shrinkage or expan—
`sion involved can be utilized to lock the spacer in place as
`shown in FIG. 6(b), yet without excessive stresses, espe-
`cially tensile stresses, induced into the solder joints.
`Aspacer usefiil for the structure of FIG. 6 can be made of
`elemental metals, alloys, ceramics, silicon or polymer. The
`CTE matching of the permanent drop-in spacer to that of the
`solder used is important in avoiding fatigue failure of solder
`joints on thermal cycling caused by ambient temperature
`changes or assembly/test process. In the inventive solder—
`packaged optical MEMS device,
`the CTE of the spacer
`material
`is preferably closely matched with that of the
`solder, with the difference being desirably less than 120%,
`preferably less than 110%, even more preferably less than
`15% of the solder CTE. For example, if the solder used is the
`37Pb-635n eutectic solder with a CTE of ~25x10'6/degree
`C, the spacer material can be made of aluminum with a CTE
`of 23.6x10'6/degree C or a magnesium alloy (Mg-9%
`Al-1% Zn) with a CTE of 26.1x10'6/degree C. In general,
`alloy compositions containing Al, Zn, Mg, Pb, In, or Cu can
`be made to exhibit CTE values comparable to those of the
`commonly used solder alloys.
`Alternatively, the spacer can be locked in place before the
`soldering operation by fixturing or attaching onto part of the
`overall device structure. FIG. 7(a) illustrates an embodiment
`of the invention wherein a pre-attaehed spacer 70 is pat-
`terned and bonded onto one of the mating parts, e.g., either
`on the mirror layer or the actuator layer. These spacers are
`to be permanently left in the packaged structure. The pre-
`attached spacers 70 are desirably configured so that the
`boundary area between adjacent mirrors is compartmented
`or blocked as much as possible. The preferred blockage is on
`the average, with at least 20%, preferably 50%, even more
`preferably 80% of the total boundary wall area. The solder
`bumps are separately added for bonding purpose. The spacer
`thickness again accurately defines the gap spacing between
`the mirror and the electrode during the soldering operation.
`The preferred accuracy of these pre—attached spacers is
`within less than 15%, more preferably less than 12.5%
`deviation from the desired gap spacing. The pre-attached
`spacer 70 can be made of a metal, silicon, ceramic, poly—
`imide or plastic material. Plastic materials tend to outgas in
`a hermetically sealed atmosphere, and also have generally
`higher CTE values, thus are less preferred than inorganic
`materials such as metals, ceramics or silicon materials.
`In FIG. 7(a), the mating faces of the mirror layer 20A and
`the actuator layer 23 are both in planar configuration, and the
`spacer is added after the mirror layer 20 with movable
`mirrors 21 is fabricated. The movable mirrors are preferably
`made of silicon, and is typically only several micrometers
`thick. In this embodiment the mirror layer 20A is fabricated
`on a silicon-on-insulator substrate comprising a first layer of
`silicon 71, a layer of SiO2 72 and a second layer of silicon
`73. The thin silicon mirror can be made using the well-
`known silicon-on-insulator (SOI) fabrication process. The
`501 process permits convenient fabrication of the thin
`silicon mirrors 21, and the presence of the buried oxide layer
`72 is useful as an etch-stop barrier. In FIG. 7(a), selected
`patterned areas of the upper portion 73 of the SOI substrate
`20A are etched, e.g., by using chemical etch, reactive-ion
`
`10
`
`15
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`

`

`US 6,442,307 B1
`
`9
`etch, or a combination of these processes to form the mirror
`array. The gimbals and the torsion bars (not shown) are also
`formed around each mirror. The $01 material and process
`are described, for example,
`in Concise Encyclopedia of
`Selnicondacting [Materials and Related Technologies, Edited
`by S. Mahajan and L. C. Kimmerling, Pergamon Press, New
`York, 1992, p. 466.
`Since the movable mirror is typically thin and