CA2066261C

CA2066261C - Micromechanical switch

Info

Publication number: CA2066261C
Application number: CA002066261A
Authority: CA
Inventors: Anthony David Welbourn; Judith Clare Mclaughlin
Original assignee: British Telecommunications PLC
Current assignee: British Telecommunications PLC
Priority date: 1989-09-26
Filing date: 1990-09-07
Publication date: 1998-04-21
Anticipated expiration: 2010-09-07
Also published as: CA2066261A1; GB8921722D0; WO1991005284A1; DE69019530D1; DE69019530T2; US5262000A; JPH05501615A; EP0493425A1; EP0493425B1; ATE122799T1; JP2987198B2

Abstract

In order to make a micromechanical switch a first sacrificial layer is formed on a substrate (1). A second sacrificial layer (4) is then formed as an island on the first sacrificial layer (3). A switch ele ment layer (7) of resilient material is then formed on the second sacrificial layer (4), and the outline (5) of a switch element (8) is defined on the switch element layer (7). The outline (12) of a window is then defined, and the second sacrificial layer (4) is etched through the window using an etchant which laterally undercuts that portion of the switch element layer (7) which is to form the swit ch element (8). The first sacrificial layer (8) is then etched through the window defined by the etched second sacrificial layer to define a cavity (13) beneath said portion, thereby defining the switch element (8).

Description

W O 91/05284 1 ~ PC~r/GB90/01391 U~ h~ANICAL ~WITCH

This invention relates to a micromech~nir~1 switch, and to a method of making such a s~itch. In particular, the invention is concerned with the fabrication of micromechAnic~1 beams, bridges and torsion eiemen~s for use in opticai switches and modulators.
Silicon based micromech~nic~1 switches, which incorporate microm~hined deflecting beams, bridges or torsion elements (swltcn elements) are known. Typically, this ~ype of device is formed by etching a switch element from monocrystalline silicon, the etching process being such as to form a cavity or well beneath tne switch element. lectrodes are then added (or formed in the monocrystalline silicon) for controlling the switch element. '~hen making mirror switches or modulators, the surface of the switch element _s provided with a coating of reflective metal. Such switches offer significant advantages compared with conventionai switches, these aavantaaes arlsing from their small size, fast response -nd negligible ageing effects. ~oreover, ~hey can be manufactured by techniaues that are compatible with standard integrated circuit (IC) processing methoas, ana ;o offer Lhe ~otential of batch processing and of integration with associated electronic circuitry.
~ or some cptical switching applica~ions, it is advantageous to use larger switch elements than those which can be made bv known microm~ining techniques (typically ~3-75um). Jnfortunately, .abrication problems arise as the size of the switch element s .

SUBSTlfUTE SHEEl-W O 91/05284 2 0 6 6 2 61 - 2 - P ~ /GB90/01 increased. mhus, if the switch element has a dimension ~reater Ihan ~oOum, a very high degree of etching selectivity is re~uired. ~his is because the region tO be protected from etching is coated with an etch-resistant ~asRina layer àS thin as 0.~ to 0.2 um, and the material 3eneath the switch element (which is to be undercut) mus~
be completely eroded. This calls for a long over-etch, which needs a selectivity of more than 3000:1 between the lndercul region and the etch mask. ~o give some idea of -his constraint, Si~ ~4 ls usually used 2S a ?rotective maSR agalns~ the usual oxide e~chant (buffered ~F), but the SiO2: Si~ N4 etch ratio is only 50:1.
Similarly, where an anistroplc Si etchant is used to elch ~100~ planes auickly, whilst l'etch-stoppingll on {111~
oianes, selectivity is typically 50:1 for EDP and between oo: 1 and 300: 1 for KOH.
~ oreover, lt is important that the edqes of the cavity -ormed beneath the switch element are well defined during he undercut etching process. mhus, f an undercut of several hundred ~m is ~o be formed (which is necessary -or arqe area switch elements), a iateral etch s~op layer ~s required around the cavity edges to p~even~ the cavity 3eing enlarged by several hundred ~m per side. Since some enlargement :s tolerable, a selectivity of greater than 100:1 is adequate here. However, even this degree of etch selectivity is difficult to achieve.
Another problem that can arise with the manufacture of mirror switches/modulators is that the metal reflectivity can be affected by the etching process. Tn order to overcome this problem, the metal may be passivated (i.e.
coated with non-etching material) prior to the deep ~naercut etch step. mhis passivating layer must be formed by a low temperature process to avoid thermal damage to the metal ~ilm and stress damage to the switch element.

SUB~ 1 1 I'UTE SHEET

3 .' 2066261 W O 91/05284 P ~ /GB90/01391 Unfortunately, low temperature passivating layers have a poor etch resistance, so it is difficult to prevent ~he metal reflectivity being affected.
When the switch element of such a switch is controlled electro~lr~11y, very high voltages may be required, in wnich case it is important to obtain good isolation/breakdown strength between the control electrodes. If the switch element is a torsion element, this is very difficult to achieve monolithic~lly.
Furthermore, lt is difficult to ~L-e~ent the torsion bars of such a torsion element from deflecting under the forces used to control the element.
The aim of the invention is to provide a method of making micromech~nir~l switches which does not suffer from these problems.
The present invention provides a method of making a ~icro~ech~nic~1 switch, the method comprising the steps of:-a~ forming a first sacrificial layer on a substrate;
b) forming a second sacrificial layer as an island on~he first sacrificial layer;
r) forming a switch eiement layer of resilient material on the second sacrificial layer;
d) defining Ihe outline of a switch element on the switch element layer;
e) defining the outline of a window;
) etching an aperture in the second sacrificial layer through the window using an etchant which laterally undercuts that portion of the switch eiement layer which is to form the switch element; and g) etching the first sacrificial layer through the a~e.LuLe in the etched second sacrificial layer to define a cavity beneath said portion.

SUBSTITUTE SI~

2 ~ G ~ 2 6 1 ~ 4 ~
W O 91/05284 P ~ /GB90/01 Advantageously, the first sacrificial layer s an oxide layer formed by deposition or thermal growth, and the second sacrificial layer is a polysilicon layer formed by deposition. Preferably, ~he second sacrificial layer has a thir~ness which is small compared with that of the .irst sacrificial layer. ~onveniently, the thicKness of the first sacrificial layer is substantially lO~m, and -he thickness of the second sacrificial layer is between 0.5 and 2um.
The method may further comprise the step of .orming a nitride layer cn the second sacrificial layer prior to the formation of the switch element layer, the nitride layer having a thic~ness cf between O.l and 0.4~m.
Preferably, the switch eiement layer is doped polysilicon which is formed by deposition, ~he thicKness of the switch element layer Deing between 0.5 and 2um.
~ dvantageously, the outline of the switch element is defined photolithogr~phicAlly, and the switch element is formed by etching the switch eiement iayer selectively to 'he nitride layer. Conveniently , the switch element ayer s etched using a plasma etch.
~ further nitride layer may be deposited following the step of etching the switch eiement layer, the further nitride layer having a thickness of between 0.1 and 0.3um. The method may further comprise the c~ep of orming a layer of metal on that portion cf the further nitride layer which covers the switch element.
The method may further comprise the step of forming a ?assivating layer over the iayer of metal.
The outline of the window may be formed ?hotolithogr~ph;c~lly on the passivating layer.
Preferably, ~he passivating layer is etched, through the window, using a buffered HF etchant. Both nitride layers SU~S~ JTE S~

-~- 2066261 W O 91/0~284 ~ PC~r/GB90/01391 may then be etched, using the etched passivating layer as ~ masR, using orthophosphoric acid as the etchant.
Alternatively, the passivating layer and both the nitride ayers are etched, through the window, by a singie plasma etching step. Advantageously, the plasma etching step uses CHF~ as the etchant.
mhe invention may further comprise the step of formina a further iayer of metal over par~ cf the passivating iayer and part of the first sacrificial layer, the further layer of metal being formed prior to etching the ~assivating layer. Preferably, the further layer of material is formed by electroplating and has a thickness of ~etween 3 and 4um.
Advantageously, the second sacrificial layer is etched ~sinq EDP as the etchant, and the first sacrificial layer is etched using a buffered HF etchant.
mhe method may further comprise the step of forming one or more eiectrodes in the substrate prior .o the -ormation of the first sacrificial layer.
~ he invention also ?rovides a micromech~nir~l switch ~henever made DV the method aescribed a~ove.
The invention will now be described is greater detail, oy way of example, with reference to ~he accompanying ~rawings, in wnich:-Figs to ~ are schematic longitudinal cross-sections which illustrate the basic process sequence of the fabrication method of ~he invention;
~igs 4 to 6 are diagrammatic longitudinal ~ross-sections through a practical device constructed in accordance with the invention at various stages of fabrication;

SUBSTITUTE S~

' r . ~ 6 WO 91/05284 2 0 6 6 2 61 P ~ /GB90/0:

Fig 7 is a diagrammalic plan view of the device at an intermediale stage of fabrication;
r ig 8 is a diagrammatic transverse cross-section through the finished version of a device which is a modified version cf the device of Figs 1 to 4;
Fig 9 1s a diagrammatic cross-section, similar to that of Fig 1, of a modified form of device; and Figs 10 to 12 are diagrammatic iongitudinal cross-sections throu~h an electrical switch device constructed in accordance with the invention at various stages of ~abrlcation.
~ eferring to the arawings, Fig. 1 shows an intermediate stage in the basic process for fabricating a micromech~nic~l switch. A first relatively-thick sacrificial layer Sl is formed on a silicon substrate S, the substrate containing one or more lower electrodes tnot shownj. A second, relatively-thin sacrificial layer S2 _s then formed on the first sacrlficial layer Sl. ~
cantilever beam iayer S~ is then formed on the layer S2.
This stage of the process is snown in Fig. 1. The iayer S2 is then removed (using a suitable etchant) from beneath that portion of the layer S3 which is to form a cantilever ~eam B by a long undercut (see Fig. 2). A cavity C is then etched in the layer Sl through the cavity formed ln ~he layer S2. The aeep cavity C ls, therefore, etched in a time related to the thickness of the layer Sl. rather than to the depth of undercut. The requirements for very high etch-rate selec;ivity and very thick layers are, ~hus, separated into two different materials.
Figs . to 6 show various stages of the fabrication ?rocess of a practical device . Thus, Fig. 4 shows an SUt~a 1 1 1 UTE SI~EET

_ 7 _ ~ 2066X61 W O 91/05284 P ~ /GB90/0139 ntermediate stage in the fabrication process of an optical micromechAnic~l switch. The starting point for the process according to the invention is a monocrystalline silicon subs~rate i. A lower electrode 2 is formed in the s~bstrate 1 by a diffusion process in which a heavily-doped region defines the ~ower electrode.
The heavily-doped region may be p (in which case boron s used as the dopant) or n (in which case arsenic or phosphorus is used as the dopant).
A thic~ (_lOum) sacrificial oxide layer 3 is thermally grown on the substrate 1, after which a thin (0.5 to 2.0~m) sacrificial polysilicon layer 4 is deposited by a chemical vapour deposition (CYD) process.
Using a photolithographic technique, the outline ~ of a cavity (to be described below) is then printed onto the thin sacrificial layer ~ (see Fig. 4). The outline of the cavity is then defined by etching the polysilicon layer 4 using a plasma etching or a wet etching process. The layer 4 is then covered with a protective nitride layer 6, this iayer being deposited by a CVD process - either ?lasma enhanced CVD (PECYD) or low pressure CVD (LPCVD) -~o a depth of G.l to 0.4um. The layer 6 acts to protect -he underside of a layer 7 which is subsequently added.
The layer 7 is a doped polysilicon layer which is deposited by CVD to a depth of 0.5 to 2.0~m. The layer 7 is doped either p or n , so that this layer can act as an electrode (as described below). ~he doping of the layer 7 also anneals the polysilicon, ~hereby relieving stress in this layer. The Dattern of a cantilever beam 8 (see Fig.4) is then printed on the layer 7 by a photolithographic ?rocess, and the polysilicon is etched away using 2 plasma etcn to define the shape of the beam.
The plasma etch is a selective etch which terminates at the nitride layer 6.

SU~a I I I ~JTE S3 IEET

2066261 - ~-W O 91/05284 P ~ /GB90/0l A second nitride layer 9 is then deposited conformally by CVD, this layer having a thirknPss of 0.1 to 0.3~m and acting to protect the top and sides of the beam 8.
The conformal deposition of this iaye~ 9 is necessary to ensure adequate protection for the sides of the beam 8. A
thin layer 10 of aluminium lS then aeposited over 'he layer 9 by evaporation or sputtering. This layer 10 is to constitute both an upper electrode and a mirror surface.
The layer '0 must be a thin layer (50 nm to 0.2~m) to ~educe the stress on the beam 8 which is to be formed OUt of the polysilicon layer 7. ~owever, a minimum thickness of 0.5~m of metal is needed in those regions wnere contact wires for the upper electrode are to be ho~ded to the layer 10. In practice, therefore, a thick (0.5~m) layer of aluminium is deposited over the entire area of the device, and aluminium is subsequently removed by plasma etching, following a photolithographic printing step, eve~ywhere except those contact regions and any necessary wiring tracks. A passivating layer 11, which acts to ?rotect the layer 10 from subsequent processing steps, is then deposited. This layer 11 is a 0.5~m conformal, iow temperature oxide coating which is formed by PECYD. Fig.
4 illustrates the device at .his stage of the fabrication ?rocess .
The outline 12 (see Fig. 7) of an etch window, which is tO open up a cavity 13 (to be described below) whilst protecting the cantilever beam 8, ls then formed by a photolithographic technique on the passivating layer 11.
The passivating layer '1 ls then etched away within the outline 12 using a buffered HF etchant. The nitride !ayers 6 and 9 are subsequently etched, within the outline '2, using the passivating layer as a mask and orthophosphoric acid as the etchant. The thin sacrificial layer 4 is then etched usin~ EDP as the etchant. This SU~ 111 IJTE S~EET

2û66261 W O 91/05284 PC~r/GB90/01391 e~ching step undercuts the polysilicon of the layer 4 3eneath the cantilever beam 8. It should be notea that, n EDP, the etching seiec~ivlty of polysilicon tO nltr'de ls ~ 6000:1, and the etching seiectivity of polysiiicon to ~assivating oxide is > 2500:1. Fig. 2 illustrates ~he device at this s~age of the fabrication process.
The cavity i3 is then formed by etching the thick sacrificial layer 3 using buffered HF. This etching step also strips away the passivating layer ll. In buffered RF, the etching selectivity of oxide to nitride is > 50:1. The etching time is virtually independent of he area of the cavity 13, depending only on ~he thic~ness of the layer ~, because the etchant can immediately penetrate fully under the cantilever beam 8. The sidewells of cavity 13 are undercut by a distance approximately equal to the depth of the cavity.
If a switch having a torsion element is required, the fabrication process described above with reference to Figs 4 to 7 could be modified. Thus, in order to define a ;orsion element rather than a cantilever beam, 'he outline of the etch window would be modified to define a pair of torsion bars 14 (see Fis. 8) exten~;ng tranversely, n ~pposite directions, from the central region of the orsion element 8'. ~oreover, an additional nitride layer 15 would be formed (between the steps, forming the two sacrificial layers 3 and 4) beneath the regions in which 'he torsion ~ars 14 are to be formed. In this region, only the thin sacrificial layer 4 is removed in the etching steps which forms the cavity 13, and the resultant pillar (see Fig. 8) forms a support for the torsion element 8', thereby minimising bending of the element.
Fig. 9 shows an intermediate stage (equivalent to ~ig. 4) in the fabrication of another form of modified device, ~hich incorporates a cantilever beam 8". The beam 8n is a SU~ JTE S~EET

W O 91/05284 2 0 6 6 2 61 P ~ /GB90/01 _ _ multi-layer ~eam which incorporates a piezoelectric layer i6. The beam 8" thus constitutes a piezoelectric bimorph ac~uator, the deflection of which is controlled by the electric field within ~he piezoelectric layer 16.
?igs 10 to 12 show :row the fabrication process of the nvention can be modified to make an electric micromech~nic~1 switch. As most of the process steps are sililar to those of the process described above with reference to Figs. 4 to 6, only the modified steps of the ?rocess of Figs. 10 to 12 will be described in detail and similar reference numerals (with the addition of 100) will be used for similar parts. Thus, a lower electrode 102 is deposited (for example by SDuttering a refractory metal~
on a monocrystalline silicon substrate 101. Thick and thin sacrificial layers 103 and 104 respectively are then formed, after which a switch element layer 107 is formed.
This layer 107 is a conducting layer, which is to form a cantilever beam switch element 108 (see Fig.12), made for example of doped polysilicon or metal. This layer could also be an insulating layer coated top and bottom with metal. A third sacrificial (oxide) layer 111 is deposited by PECVD, this layer acting as a passiviating layer and having a thickess of between 1 and 2~m. A 3 to 4~m thick metal layer 120 is then formed, by electroplating, over part of the layer 111 and over part of the layer 103. This stage of the process is shown in Fig. iO.
The sacrificial layer 111 is then etched out using a buffered HF etchant to leave a gap between the layer 107 and the layer 120 (see Fig. 11). The remaining steps of the process are similar to the final process steps described above with reference to Figs 4 to 6, and relate -o the formation of the beam 108 and the cavity 113. Fig.
2 shows the finished electric micromerh~ni~l switch, in which the beam 108 curves naturally upwards so as to ma~e SUBSTITUTE: S~EET

W O 91/05284 ~ ~ PC~r/GB90/01391 electrical contact with the iayer 120. This curve ls formed as a result of an upward bending movement imparted lo the beam 108 due to intrinslc stresses formed in the beam by the use of two different materials. ~he beam 108 can then be controlled electrostatically, using the lower electrode 102, 'o make cr break electrical contact between ~he beam and the layer 120, hus forming a one-way switch. A two-way switch could be formed by providing a .urther electrode within the cavity 113. Alternatively, the beam 108 would be a piezoelectric bimorph element, which could be bent both upwardly and downwardly to make contact with the lower electrode 102 and the layer 120 (upper electrode).
The micromech~nlc~l switch fabrication process described above has the following advantages:-l. It offers complete flPxih;lity in the placement of electrodes on the base of the cavity and on the switch element, so it is adaptable to torsion elements and bridges, as well as cantilever beams.
2. It allows the enlargement of the cavity 13 to be easily controlled comoared with methods using a single sacrificial layer (which requires a very deep etch which would give problems with later step coverage).
. t does not require an epitaxial growth step.
4. The reauired etch selectivities can all be achieved with materials which are readily available.
5. It could be used for both normally-open and normally-closed electrical switches.
6. It intrinsically gives a multi-layer cantilever allowing stressed to be compensated.
-t will be apparent that the fabrication process described above could be modified in a number of ways.
For e s mple, the lower electrode could be formed by the deposition of a refractory metal or by ion implantation.

SUt~ lTE S~EET

W O 91/05284 2 0 6 6 2 61 PC~r/GB90/01 Where a single lower electrode 2 is required, the en~ire substrate l could be doped (either n cr p ). t would also be possible, ror example where a torsion element constitutes the switching element, to provide a pair of lower electrodes. Tt would also be possible to deposit the thick sacrificial oxide layer ~ by, for example, PECVD, particularly if a me~al electrode 2 is utilised. Instead of .orming the beam 8 (switch element) .rom a polysilicon layer " .his iayer would be SiO2 deposited by, for example, PECVD. Moreover, the metal used for the electrode/mirror lO would be gold instead if aluminium. Tn this case, Ihere would be no need for the subsequent formation of the passivating layer ll. Where a passivating layer ll s necessary, this could alternatively be a nitride layer. As an alternative to using two etching steps to etch the passivating layer ll and the two nitride layers 6 and 9, these layers would be etched in a sing~e plasma etching step using, for example, CHF3.

SUBSTITUTE SHEEl'

Claims

1. A method of making a micromechanical switch, the method comprising the steps of:
a) forming a first sacrificial layer on a substrate;
b) forming a second sacrificial layer as an island on the first sacrificial layer;
c) forming a switch element layer of resilient material on the second sacrificial layer;
d) defining the outline of a switch element on the switch element layer;
e) defining the outline of a window;
f) etching an aperture in the second sacrificial layer through the window using an etchant which laterally undercuts that portion of the switch element layer which is to form the switch element; and g) etching the first sacrificial layer through the aperture in the etched second sacrificial layer to define a cavity beneath said portion.

2. A method as claimed in claim 1, wherein the first sacrificial layer is an oxide layer formed by deposition or thermal growth.

3. A method as claimed in claim 1 or claim 2, wherein the second sacrificial layer is a polysilicon layer formed by deposition.

4. A method as claimed in any one of claims l to 3, wherein the second sacrificial layer has a thickness which is small compared with that of the first sacrificial layer.

5. A method as claimed in claim 4, wherein the thickness of the first sacrificial layer is substantially 10µm, and the thickness of the second sacrificial layer is between 0.5 and 2µm.

6. A method as claimed in any one of claims 1 to 5, further comprising the step of forming a nitride layer on the second sacrificial layer prior to the formation of the switch element layer.

7. A method as claimed in claim 6, wherein the nitride layer has a thickness of between 0.1 and 0.4µm

8. A method as claimed in any one of claims 1 to 7, wherein the switch element layer is doped polysilicon which is formed by deposition.

9. A method as claimed in claim 8, wherein the thickness of the switch element layer is between 0.5 and 2µm.

10. A method as claimed in any one of claims 1 to 9, wherein the outline of the switch element is defined photolithographically.

11. A method as claimed in claim 6, or in any one of claims 7 to 10 when appendent to claim 6, wherein the switch element is formed by etching the switch element layer selectively to the nitride layer.

12. A method as claimed in claim 11, wherein the switch element layer is etched using a plasma etch.

13. A method as claimed in claim 11 or claim 12, wherein a further nitride layer is deposited following the step of etching the switch element layer.

14. A method as claimed in claim 13, wherein the further nitride layer has a thickness of between 0.1 and 0.3µm.

15. A method as claimed in claim 13 or claim 14, further comprising the step of forming a layer of metal on that portion of the further nitride layer which covers the switch element.

16. A method as claimed in claim 15, further comprising the step of forming a passivating layer over the layer of metal.

17. A method as claimed in any one of claims 1 to 16, wherein the outline of the window is formed photolithographically on the passivating layer.

18. A method as claimed in claim 17, wherein the passivating layer is etched, through the window, using a buffered HF
etchant.

19. A method as claimed in claim 18, wherein both nitride layers are etched, using the etched passivating layer as a mask, using orthophosphoric acid as the etchant.

20. A method as claimed as claim 17, wherein the passivating layer and both the nitride layers are etched, through the window, by a single plasma etching step.

21. A method as claimed in claim 20, wherein the plasma etching step uses CHF3 as the etchant.

22. A method as claimed in any one of claims 18 to 21, further comprising the step of forming a further layer of metal over part of the passivating layer and part of the first sacrificial layer, the further layer of metal being formed prior to etching the passivating layer.

,

23. A method as claimed in claim 22, wherein the further layer of material is formed by electroplating and has a thickness of between 3 and 4µm.

24. A method as claimed in any one of claims 1 to 23, wherein the second sacrificial layer is etched using EDP
as the etchant.

25. A method as claimed in any one of claims 1 to 24, wherein the first sacrificial layer is etched using a buffered HF etchant.

26. A method as claimed in any one of claims 1 to 25, further comprising the step of forming one or more electrodes in the substrate prior to the formation of the first sacrificial layer.

27. A micromechanical switch whenever made by the method of any one of claims 1 to 26.