IOP PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING

J. Micromech. Microeng. 17 (2007) 1788–1795 doi:10.1088/0960-1317/17/9/006

A new test facility for efficient evaluationof MEMS contact materialsZ Yang1, D Lichtenwalner1, A Morris2, S Menzel2,C Nauenheim2, A Gruverman1, J Krim1 and A I Kingon1

1 North Carolina State University, Raleigh, NC, USA2 wiSpry, Inc. 20 Fairbanks, Suite 198, Irvine, CA, USA

E-mail: zyang3@ncsu.edu

Received 2 April 2007, in final form 9 July 2007Published 3 August 2007Online at stacks.iop.org/JMM/17/1788

AbstractA novel test facility for the efficient evaluation of microelectromechanicalsystem (MEMS) switches and the development of alternative contactmaterials is described. The facility utilizes the upper cantilever fromcommercial MEMS contact switches, and tests these against alternativebottom contact materials within a modified atomic force microscope (AFM).The test closely approximates the real switch, but can accommodate a widerrange of test conditions and contact materials. The facility allows alternativecontact materials to be easily and quickly incorporated, and thereforeevaluated by measuring the number of cycles to failure. The evolution of thewear surfaces of the switch contact materials under test can also be easilyexamined. In order to demonstrate the facility, the evolution of the contactresistance and wear of a commercial RF MEMS cantilever with Au contactswas monitored under accelerated test conditions, comparing the behavior ofAu bottom contacts to an alternative Au–Ni alloy contact material. TheAu–Ni (20 at.%) alloy displayed reduced wear rates and improved switchcycle lifetimes compared to pure Au, while retaining acceptable values ofcontact resistance.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

Since the first MEMS switch, which was specially designed formicrowave applications in the early 1990s [1], this technologyhas developed to the point that the devices are currently insmall-scale production. Typical MEMS switches are micron-scale switching devices. Because of their low fabricationcost, low power consumption, wide operational bandand exceptional switching performance, MEMS switchingtechnology holds great promise for implementing agile radiofrequency (RF) systems. Generally, MEMS switches can becategorized into non-contact type (capacitive) and contact type(metal contacts). Figure 1 shows the cross-section of a metal-contact-type MEMS switch [2]. The switch is driven to closurewith an electrostatic force by applying a voltage across parallelelectrodes located on the cantilever and bottom electrode. Thisstudy utilizes this type of MEMS cantilever.

Gold is often utilized as a contact material for metal-contact MEMS switches due to its excellent electrical

conductivity and corrosion resistance. Gold-contact switcheshave been shown to achieve >109 switching cycles underfavorable operating conditions (e.g. low current, optimumcontact force, dry nitrogen environment and ‘cold switching’)[2, 3]. ‘Cold switching’ refers to switch operation underconditions in which there is no field across the contactsas the switch opens or closes. However, there is a strongdesire to significantly broaden the range of conditions underwhich a given switch would operate, including the rangeof currents (i.e. power handling). However, as operationalconditions become more severe in mechanical force orelectrical power handling, Au contact degradation increases,drastically shortening the MEMS switch lifetime. Oneapproach toward increasing MEMS switch robustness is toinvestigate alternative contact materials, and a few havebeen investigated [4–7]. However, there are some seriousimpediments to the broad investigation of alternative contactmaterials. In general, convenient contact wear test facilitiesdo not closely mimic real contact switches. In particular,

0960-1317/07/091788+08$30.00 © 2007 IOP Publishing Ltd Printed in the UK 1788

A new test facility for efficient evaluation of MEMS contact materials

Figure 1. Schematic of the cross-section of a contact-type MEMSswitch [2].

the tests have difficulty duplicating the switch geometry, thecontact geometry and the contact force. Contact geometry hasa strong influence on the heat distribution across the contacts,and thus the switch performance and failure mechanism [8, 9].One clear alternative is to test candidate materials on actualMEMS switches, and such studies have been reported byresearchers from universities and industry [3, 5, 6, 10].However, the approach has a serious difficulty: the switchesare fabricated in Si foundries, and only a limited range ofmaterials may enter the fabrication facility. Secondly, thefabrication process must be optimized for each material, andit may take many months to fabricate a set of switches to testa single candidate contact material. Materials’ compatibilityand process integration issues must be addressed in advance forevery material to be tested. These factors have severely limitedthe range of contact materials that have been investigated forMEMS switches.

If progress is to occur, the testing and development ofalternative contact materials must be more efficient whileremaining representative of typical MEMS switch operation.In this manner, the material comparison and failure analysisdata can be collected consistently and efficiently. In ourexperimental setup, to address these two issues, the switchinglifetime test is performed with an actual cantilever (includingupper contacts) extracted from a commercial RF MEMSswitch (as shown in figure 1) that is contacted against avibrating thin film bottom electrode containing the bottomcontacts. This design enables the dynamic switching test tobe performed under realistic conditions (e.g. contact force,

Figure 2. Configuration for the switching tests.

contact geometry and switch geometry) comparable to those inactual devices. The bottom electrodes are separately fabricatedfrom candidate contact materials and can be easily loaded into,and unloaded from, the test system. In this way, the test anddevelopment efficiency is dramatically improved.

In this initial paper we have selected one material forcomparison with gold in order to demonstrate the new facility.Given that the ‘softness’ of gold under both normal andaggressive operating conditions is considered to be a limitationto the contact lifetime, we have selected a set of Au–Ni alloysfor study. The results for the Au–Ni (20 at.% Ni) alloy will bereported here. The full results for the alloy system are reportedelsewhere [17].

2. Experimental details

We have designed a unique setup for monitoring MEMSswitching behavior by integrating a MEMS cantilever withseparate thin film bottom contact materials into an atomicforce microscope (AFM). The test configuration is shown infigure 2. The cantilever is first dismounted from an array ofthe RF MEMS switches chip, and then transferred to a cleanAFM tip carrier.

The transfer of the cantilever from the MEMS chip tothe AFM carrier is performed at a Rucker & Rolls 260 probestation. In preparation for the transfer, two identical MEMSchips are fixed on a glass slide next to each other. A cleanAFM tip carrier is then mounted and evenly supported bythe two MEMS chips. An xyz-axis movable XYZ 500 TRSmicromanipulator with a probe is first used to apply glue tothe surface of the AFM tip carrier, with the probe temporarilyattached to the AFM tip carrier to facilitate its positioning.Thereafter, another micromanipulator with a micro-probe (tipdiameter = 25 µm) is used to apply an extremely small portionof glue to the MEMS cantilever’s surface. The AFM tip carrieris then aligned and lowered to make contact with the MEMScantilever on the glue area. The transfer is achieved when theAFM tip carrier is lifted again after a 24 h cure. A typicalAFM carrier with a MEMS cantilever attached is shown infigure 3.

1789

Z Yang et al

Figure 3. Optical microscope image showing the MEMS cantileverattached to an AFM tip carrier. The normal AFM tip has beenremoved. The two micro-contacts are at the end of the cantilever.

The tip carrier is next inserted into the carrier hold ofthe AFM and lowered by means of the z-direction positioningmotor. The initial contact occurs between a silicon dioxideinsulating tip at the front edge of the cantilever and the bottomsample (the lower electrode structure). The initial contactis monitored with an optical microscope, and the positionof the carrier is then recorded. The carrier is further lowereduntil the two gold upper micro-contacts make electrical contactwith the patterned bottom contacts, as monitored through theelectrical circuit across the bottom contacts. When the upperand lower electrodes are properly in contact, the deflection ofthe cantilever ensures that these are the only contact points(the lift of the insulating tip can be observed through the insitu optical microscope). The position of the cantilever isagain recorded. From the two positions of the carrier, thetotal deflection of the cantilever is known (∼20 µm), and thecontact force can be approximated using the known geometryand the material properties.

A detailed view of the contact configuration is shownin figure 4. As is shown in figure 4(a), the cantilever iscomposed of three layers: a 2 µm SiO2 layer with 0.5 µm goldlayers on both sides. The width of the cantilever is 75 µm.The effective length of the cantilever can be adjusted whileattaching it to the cantilever carrier. For the tests reportedhere, the effective length was fixed at 150 µm. There are twoelectrically connected micro-contact bumps on each cantilever(top contacts). The diameter of each contact is 5 µm. Thebottom contact is patterned with the thin films of candidatematerials deposited on oxidized Si (0 0 1), with the patternshown schematically in figure 4(b). Details of the fabricationof the bottom electrode structure are discussed later in thisexperimental section.

While the upper micro-contacts make electric contact withthe bottom sample contact pad area, the contact resistance ismeasured by a four-point probe setup (figure 4(b)). Fromthe figure, the top two probes supply 1 mA current, whilethe bottom two probes sense the voltage. The resistancefrom the four-point probe measurement consists of the contactresistance of the two pairs of micro-contacts and the resistanceof the metal connection on the cantilever. To obtain contactresistance for a single pair of micro-contacts, we deduct the

(a)

(b)

Figure 4. Schematic view of the contact configuration.(a) Cross-sectional view of the upper cantilever and bottomelectrode. (b) Top view of the top and bottom contact configurationand the four-point probe setup.

metal connection resistance (∼0.2 �) and divide the remainingresistance by 2. A piezoelectric actuator (Physik Instrumente,P-802.00) is used to vibrate the patterned bottom contactsvertically with an amplitude of 1.8 µm, which is sufficientto bring the upper micro-contacts and bottom contact padsinto and out of contact. This vertical vibration does notchange the cantilever’s deflection status (∼20 µm deflection)significantly. It only switches the contacting points of the topcantilever between the insulating SiO2 tip and the two uppermicro-contacts. It also results in the upper micro-contactsbeing slightly tilted with respect to the plane of the bottomelectrodes when contact is first initiated (in the test device, thisangle is about ∼2.5◦ and it is constant for every test). Whencontact is made between the upper micro-contacts and thebottom sample, 1 mA current flows through the two contacts,and the contact resistance is measured by the four-point probesetup. When contact is made between the insulating tip andthe bottom sample, no current flows through the upper micro-contacts. The insulating tip assures a clear ‘off’ state wherethe electrical contact is broken between the micro-contacts andthe bottom electrode.

The drive signal for the actuator and the voltage dropacross the micro-contacts are both input into an oscilloscopeso that the dynamic contact process is monitored in situ. Thesample is cycled until a contact resistance failure occurs.The failure criterion that is utilized is a factor 2 increase inthe measured contact resistance, although a rapid increase to amuch higher resistance is normally observed. Thereafter, theworn contacts are examined by AFM and scanning electronmicroscope (SEM). Typical resistance switching data areshown in figure 5 for low frequency switching (0.5 Hz) with theopen state resistance (∼1150 �) set by a test system parallelresistor. This resistor also determines the open-circuit voltage.With sufficiently high open-circuit voltage, additional material

1790

A new test facility for efficient evaluation of MEMS contact materials

Figure 5. An example of low frequency dynamic switchingresistance response.

damage may be induced by possible arc energy dissipation ortransient current heating other than mechanical wear. In orderto address this acceleration factor, both hot switching and coldswitching test conditions were investigated in the switchingtest for pure gold. Quantitative volume change analysis (usingGwyddion software) was conducted to evaluate this effect.

Pure Au and Au–Ni (20 at.% Ni) bottom contact sampleswere prepared using ion-beam sputter deposition. Thesubstrate utilized was typically a Si(0 0 1) wafer coated with1000 A silicon dioxide (thermal oxide). Xenon was usedas the working gas due to its relatively large mass number,thus lower reflective ion energy bombarding the film duringgrowth. The base pressure was 3 × 10−7 torr and the depositionpressure was 4 × 10−4 torr. A 5 nm Cr layer and 10 nm Molayer were deposited as adhesion and diffusion barrier layers,respectively. 250 nm Au or Au-20% Ni thin films were thendeposited. The acceleration voltage for the ion beam was600 V. The chamber was designed to allow the three metallayers to be deposited in situ. Rotation of the target throughoutthe deposition ensured uniformity of the composition for goldalloy depositions. The as-deposited thin film samples werepatterned using photolithography and ion etching, to facilitatein situ four-point probe measurement of the contact resistancein the lifetime tests.

Electrical resistivity data were obtained on as-depositedthin films (before patterning) using a four-point probetechnique (MAGNE-TRON Model-700). A four-point sheetresistance correction [16] was applied. Micro-hardness datawere obtained using a Hysitron Triboscope nano-indentor(Berkovich indenter tip with a 142.5◦ included angle).

3. Effect of material properties on contact resistance

Pure Au as a contact material has the advantage of achievingvery low resistance in the closed state, due to its low resistivityand softness. With the use of alternate contact materials,closed-state resistance is expected to increase, depending onthe material’s resistivity and hardness. This section comparesthe measured material-dependent closed-state resistance tothat expected from a simple theoretical model.

Knowing the dimensions of the MEMS cantilever, itsYoung’s modulus (93.3 GPa) [11] and the deflection distance,

Table 1. Properties of Au and Au–Ni samples.

Nickel Electrical Micro- Calculated Measuredcomposition resistivity hardness resistance resistance(at.%) (µ� cm) (GPa) (�) (�)

0 3.8 2.7 0.14 0.16–0.220.0 20.5 6.1 0.6 0.9–1

the total contact force can be approximated using a cantileverbeam model [12]. The model is described by

F =[E · w · t3

4 · l3

]d, (1)

where E is the Young’s modulus of the cantilever, w is thewidth, t is the thickness, l is the length and d is the deflectiondistance of the cantilever beam. For a 150 µm long cantilever,the deflection distance is close to 20 µm in our case. The totalcontact force is then calculated to be ∼300 µN, which is twoto three times higher than the contact force of typical MEMSswitches.

A simple model has been tested to predict the contactresistance for the two contact materials. The relationshipcorrelating contact resistance to the contact force has beendeveloped by previous researchers for gold-contact MEMSswitches [9, 13, 14]. For a contact radius greater than theelectron mean free path, the Maxwell spreading resistancemodel can be used to calculate the contact resistance. In ourtest, the nominal contact radius is much larger than the meanfree path in gold (∼36 nm). The contact resistance Rc isdescribed by

Rc = ρ

2rc, (2)

where ρ is the resistivity and rc is the contact radius [15].While two different materials are in contact, the resistivity isestimated as the average of the two. The contact radius can beobtained using an asperity deformation model [14]:

rc =√

F

ξπH, (3)

where F is the applied force, H is the hardness of the materialand ξ is the coefficient of the material deformation mode.Different modes are given as ξ < 0.3, elastic deformation;0.3 < ξ < 0.75, elastoplastic deformation; and 0.75 < ξ < 1,plastic deformation.

For the contact between different materials, which is thecase in our experiments, the averaged electrical resistivityand hardness value are used for the calculation. Plasticdeformation is also assumed (ξ = 1) based on the relativelylarge contact force.

The measured electrical resistivity, micro-hardness andcontact resistance, along with the calculated contact resistancebetween upper pure gold micro-contacts and either the puregold or gold–nickel bottom contacts are shown in table 1. Forthe pure gold, the experimental data fit the calculated contactresistance data well. However, for the gold–nickel case, themeasured contact resistance is higher than the calculated value.It should be noted that the roughness of the Au–Ni alloy asmeasured by AFM is greater than that of Au. Roughness isnot taken into account in the above simple model, but shouldincrease the contact resistance by decreasing the contact area.

1791

Z Yang et al

(a)

(b)

Figure 6. (a) Contact resistance evolution for sputtered gold againstgold micro-contacts in the present test facility under acceleratedconditions. (b) Lifetime test of a similar switch under (moderate)normal service conditions (number of switch cycles plotted on a logscale) [11].

These results point to the need for a more sophisticated modelfor the relationship between contact material properties andthe contact resistance. The effect of topography on contactresistance will be discussed in a separate paper [17], and amore complete multiscale model has been developed [18].

4. Accelerated lifetime test: a unique electricalfailure detection mechanism

In order to evaluate contact performance of candidate materialsin our test system, we chose the ‘cycle number to electricalfailure’ as the indicator of the degree of micro-contact damage.Since the damage is accelerated by a relatively large contactforce under hot switching conditions, and the wear to failureis very small (determined by the wear required to producecontact only between the bottom electrode and the cantilevertip), such tests can be termed an ‘accelerated lifetime test’.During the test, the open-circuit voltage is fixed at ∼1 V,measurement current is 1 mA and total contact force is∼300 µN. Figure 6 shows switching tests for sputtered puregold bottom electrodes against the upper gold micro-contacts.Figure 6(a) shows accelerated test data for the switch in thistest facility under hot switching, in comparison with resistancedata for a similar switch in the actual MEMS switch productconfiguration (figure 6(b)) subjected to purely cold switching[11]. In both cases, the usual three resistance regimes can beobserved. The initial contact resistance falls during the ‘burn-in’ period, which may be as short as a few cycles. The switchthen appears to operate in a steady state, prior to the onset ofinstability and resistance increase that corresponds to contactdegradation.

(a)

(b)

(c)

Figure 7. AFM analysis of upper gold micro-contacts afterswitching failure (the bottom electrode was pure gold). (a) AFMimage of the whole micro-contact. (b, c) Height profilescorresponding to X1–X2 and Y1–Y2.

The marked reduction in the number of cycles to failurecan be observed in the accelerated test in the new test facility.After only ∼100 cycles, the closed contact resistance started torise and reached an open-circuit condition. The same test wasrepeated using different pure gold samples with new MEMScantilevers. The results are reproducible, with the variation ofthe number of cycles to failure being less than ∼50%.

The AFM analysis of the corresponding failed uppermicro-contact is shown in figure 7. The switching weardamage is visible at the front edge of the micro-contact (nearthe top of image (a)), which corresponds to the region ofcontact or impact between the micro-contact on the cantileverand the lower electrode. Particularly revealing are the two lineprofiles taken across the micro-contact. Figure 7(b) shows theheight profile of the ‘X’ scan taken across the contact alongthe cantilever width, showing no wear. In comparison, the ‘Y’line scan of figure 7(c) runs parallel to the cantilever directionand includes the impact area. It can clearly be seen that thecontact is eroded in the impact region nearest the cantilever tip,

1792

A new test facility for efficient evaluation of MEMS contact materials

0.08

0.04

X1-X2

(c)

0.00

X1 X2

(a) (b)

Figure 8. Atomic force microscope images of the pure Au bottomelectrode after failure (1 mA). (a, b) The damaged bottom contactarea. (c) The height profile corresponding to X1–X2.

with material having been transported away from the impactregion.

The corresponding AFM images of the pure gold bottomcontact area after the failure test are shown in figure 8. Asmall crater or depression approximately 30 nm in depth and∼2 µm in diameter was formed during the switch cycling.Furthermore, there is a build-up of material around this crater.

As is mentioned in section 2, the upper cantilever remainsin bent with almost unchanged curvature during switchingto exert constant contact force. The vibration of the bottomelectrode by the piezoelectric actuator alternates the cantilevercontact spots between the insulating tip and bottom electrode(open), and the micro-contacts and bottom electrode (closed).As a result of the slightly tilted micro-contacts (∼2.5◦), thecontact area (or damaged area) is always located at the frontedge of the micro-contacts. Upon closure, all the contactforce is exerted on that small area. Possible transient currentheating or arc energy dissipation may also occur at the samearea under certain hot switching conditions. Thus duringcycling, this part of the micro-contact is more susceptible topotential damage (as is shown in figure 7). After an amountof material volume change occurs on the front edge of themicro-contacts, the small vertical lift of the bottom electrodeduring switching (∼1.8 µm) will not be sufficient to allowtouching between the worn bottom electrode and worn micro-contacts. Consequently, no electrical contact is made betweenthe upper micro-contacts and the bottom sample, leading to anopen-circuit situation.

It must be emphasized that this electrical failure is notan absolute indicator of the lifetime of the MEMS switches.Instead, it serves as an indicator that a certain amount ofsurface modification occurred on the micro-contacts and thebottom contacts during a switching test (enough that thepreset piezoelectric actuator displacement distance is no longersufficient to produce physical contact of the micro-contactsand bottom electrode; only the insulating tip touches the

Table 2. Effect of Voc on material transfer (pure gold).

Voc Cycle Transferred materialSample (dc) (V) number volume (µm3)

1 6 1 000 0.0462 6 2 000 0.1163 6 3 600 0.4304 0 3 600 0.0075 0 40 000 0.090

bottom contact area). Thus, it allows a direct comparison ofmaterial wear rates of different contact materials. These samelifetime limiting factors such as mechanical wear, transientcurrent heating and arc energy dissipation are expected in theoperation process of real MEMS switches. The difference forreal MEMS switches is such that the tilt angle is normally lessthan 1◦, the cantilever is not pre-bent and the insulating tipdoes not touch the bottom electrode. Thus in actual MEMSdevices, it takes a substantially larger number of cycles toaccumulate damage over the whole micro-contact to reachan electrical failure. We are not presently able to absolutelycorrelate this accelerated lifetime to an actual device lifetime;however, the relative difference between measured lifetimes inthis test facility should be proportional to differences in actualdevice lifetime under similar operation conditions. Thus, ourunique electrical failure mechanism may be used as an efficienttool to more quickly evaluate the switching performance ofdifferent materials under a variety of different test conditions.

5. Effect of open-circuit voltage (hot switching) onmaterial transfer

The (accelerated) switching lifetime tests reported hereare typically undertaken under aggressive test conditions,including hot switching, ambient air and greater contact force(up to two to three times the contact force of the commercialdevice). Under such hot switching conditions, the open-circuitvoltage (Voc) may produce arc discharges and substantiallyhigh transient currents (decays to 1 mA within a few µs)during closing, and these are expected to accelerate the failureprocess. In order to assess the effect of Voc, both hot switchingand cold switching test conditions were investigated in theswitching test for pure gold. In the test, each MEMS cantileverwas cycled upon a pure gold bottom electrode for a certainnumber of cycles. A stable contact resistance of 0.6–0.7 � wasachieved during all the tests. After characterization, the AFMdata of each micro-contact and bottom electrode were analyzedusing Gwyddion software. The sum of the material volumechange of the two micro-contacts for each sample is listed intable 2. The material transfer is observed to increase withthe cycle number under both hot switching and cold switchingconditions, although the measured material transfer is muchless in the case of cold switching. These results reveal thecapability of this test facility to effectively evaluate materialwear under various test conditions.

Differentiating electrically induced wear effects (such asthe open-circuit voltage effect) from the purely mechanicalwear effects in this new test system is an important steptoward our research into novel alternative contact materials.Since arc energy dissipation, transient current heating and

1793

Z Yang et al

Figure 9. Contact resistance evolution for Au contacts againstAu–Ni (20%) alloy contacts. The gold against gold test result is alsoshown as a comparison.

mechanical wear are typical lifetime-limiting factors for hotswitching MEMS switches, it is valuable to test differentcontact materials under various conditions to distinguish theeffects of these factors.

6. Effect of contact material on MEMS switchlifetime

Au–Ni alloys were investigated as an alternative contactmaterial for MEMS switches. The enhanced hardness isexpected to improve the lifetime by reducing material wearduring the switching operation. The closed contact resistanceevolution for gold micro-contacts against either Au or Au–Ni (20 at.% Ni) bottom electrodes is shown in figure 9. Thenumber of cycles to electrical failure is about 1200 for the Au–Ni bottom electrode. Compared to the pure gold electrode, the‘lifetime’ is dramatically improved. Figure 10 shows the SEMand AFM analyses of upper gold contacts after failure. Thedamaged area is again located on the front edge of the micro-contact, where impact occurs. Also similar to the previouscase, material loss in the contact area is confirmed by the heightprofile image comparison. Material built-up surrounding theimpact area on the micro-contact occurs in drop-like shapesthat are reminiscent of local melting. The melt-like golddroplets are also observed at the foot of the micro-contacton the upper electrode of the cantilever.

AFM and SEM analyses of Au–Ni (20%) bottomelectrodes after failure are shown in figure 11. Material built-up on the bottom electrode is apparent. A few small indentedholes can also be identified. The diameter of the holes istypically less than 0.5 µm and the depth is less than ∼20 nm.Thus the damage volume is greatly reduced in comparisonwith that of the pure gold bottom electrode, even though thenumber of cycles before failure is more than ten times higher.

Compared to the pure gold against pure gold test, thegold against gold–nickel test shows a much longer ‘lifetime’under the same test conditions (1 mA, Voc = ∼1 V, a totalof 300 µN on two contacts). Also, the SEM and AFMfailure analyses show that the damage is less severe, eventhough the total number of cycles is much larger for Au–Ni contacts. It is shown that material transfer on the uppermicro-contacts changes the topography and the shape of themicro-contacts. The damage accumulates until the contactgeometry is changed. In this way, an open-circuit situation is

(a)

(b)

(c)

(d )

Distance (µm)

Distance (µm)

Figure 10. Analysis of failed upper gold contacts (against Au–Ni20 at.%) after the failure. (a, b) SEM image and AFM image of themicro-contact; (c, d) Height profiles corresponding to X1–X2 andY1–Y2.

0.08

0.04

0.00

X1 X2

X1-X2

(a) (b)

(c)

Figure 11. Gold–nickel bottom electrode after failure (1 mA).(a) The SEM image and (b) the AFM image of the micro-contact.(c) The height profile along X1–X2

then formed because the insulating lip of the cantilever touchesthe bottom electrode instead of the micro-contacts. Thepure gold against pure gold contact seems more susceptibleto hot switching degradation and mechanical wear, leadingto much more rapid failure while achieving lower contactresistance (∼0.2 �/contact). Detailed discussion of the effectof material properties on hot switching energy dissipation isbeyond the scope of this facility paper and will be presentedelsewhere.

The micro-hardness of the gold–nickel bottom electrodeis much greater (∼6.1 GPa) compared to that of the pure gold

1794

A new test facility for efficient evaluation of MEMS contact materials

bottom electrode (∼2.7 GPa). The adhesion between goldand gold–nickel surfaces is also expected to be lower. All ofthese will make the contacts more resistant to force-induceddamage. Pure gold deformation upon gold–nickel bottomcontacts allows stable electrical contact with low contact force(∼150 µN/contact). It is shown that both the upper andlower electrodes are less damaged despite a much larger totalcycled number. Failure appears to arise from material transferfrom the contact area of the upper contacts to the surroundingarea and the bottom electrode. The harder Au–Ni showsmore resistance to deformation, but also leads to a slightlyhigher contact resistance (∼1 �/contact). The AFM and SEMobservations suggest that local melting may occur during thecycling test.

7. Conclusions

In this paper, a new test facility for MEMS contacts isdescribed. The test utilizes cantilevers from commercialswitches and mimics the operation of real RF MEMS contactswitches. It allows efficient tests of different candidate contactmaterials as bottom contacts against upper pure gold micro-contacts. Using this facility, candidate materials for MEMScontacts can be efficiently examined and characterized in anaccelerated manner. To demonstrate the facility, we have runstandard Au–Au tests under different conditions (hot switchingversus cold switching) followed by a volume analysis ofmaterial transfer. It has been verified that hot switchingconditions accelerate the material transfer process. StandardAu-to-Au contacts were compared with Au-to-Au–Ni(20 at.% Ni) contacts under accelerated test conditions. Wehave observed that the harder gold–nickel alloy displays almostone order of magnitude improved ‘lifetime’ or wear resistancecompared to a pure gold bottom contact.

Acknowledgments

This work has been supported by the Extreme Friction MURIprogram, AFOSR grant FA9550-04-1-0381 entitled ‘Multi-

functional Extreme Environment Surfaces: Nanotribology forAir and Space’.

References

[1] Larson L E, Hackett R H, Melendes M A and Lohr R F 1991Micromachined microwave actuator (MIMAC)technology—a new tuning approach for microwaveintegrated circuits Microwave and Millimeter-WaveMonolithic Circuits Symp. Dig. (Boston, MA) pp 27–30

[2] Morris A S, Cunningham S, Dereus D and Schropfer G 2003Proc. 33rd European Microwave Conf. vol 1, pp 21–4

[3] Rebeiz G 2003 RF MEMS Theory, Design, and Technology(New York: Wiley)

[4] Schimkat J 1998 IEEE MEMS, The 11th Ann. Int. Workshopp 200

[5] Hiltmann K, Schumacher A, Guttmann K, Lemp E,Sandmaier H and Lang W 2001 47th IEEE Holm Conf. onElectrical Contacts pp 117–21

[6] Hiltmann K, Keller W and Lang W 1999 Sensors ActuatorsA 74 203–6

[7] Coutu R A 2004 Electrostatic radio frequency (RF)microelectromechnical systems (MEMS) switches withmetal alloy electric contacts PhD Dissertation Air ForceInstitute of Technology

[8] Chen L, McGruer N E and Adams G G 2005 Contact evolutionin a micromechanical switches ASME World TribologyCongress (Washington, DC)

[9] Mehregany D H 1999 IEEE Trans. Compon. Packag. Technol.22 357–64

[10] Majumder S, Lampen J, Morrison R and Maciel J 2003 IEEEInstrum. Meas. Mag. 6 12

[11] Morris A 2006 Technical data on RF MEMS switcheswiSpry Inc. (Irvine, CA)

[12] Senturia S 2001 Microsystem Design (The Netherlands:Kluwer)

[13] Jensen B D 2004 17th IEEE Int. Conf. on MEMS pp 137–40[14] Bromley S C and Nelson B J 2001 47th IEEE Holm Conf. on

Electrical Contacts pp 122–7[15] Timsit R S 1999 Electrical Contacts: Principles and

Applications (New York: Dekker)[16] Perloff D S 1977 Solid-State Electron. 20 681–7[17] Yang Z, Lichtenwalner D J, Morris A S, Krim J and

Kingon A I 2007 J. Microelectromech. Syst. submitted[18] Rezvanian O, Zikry M, Brown C and Krim J 2007

J. Microelectromech. Syst. submitted

1795