A

Xa

b

a

ARRAA

KOMFE

1

tiacva[

tt[cdtswlswpc

h0

Sensors and Actuators A 212 (2014) 93–101

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

j ourna l h o mepage: www.elsev ier .com/ locate /sna

novel MEMS omnidirectional inertial switch with flexible electrodes

i Zhanwena,∗, Zhang Pinga, Nie Weironga, Du Liqunb, Cao Yuna

School of Mechanical Engineering, Nanjing University of Science and Technology, Nanjing 210094, People’s Republic of ChinaSchool of Mechanical Engineering, Dalian University of Technology, Dalian 116024, People’s Republic of China

r t i c l e i n f o

rticle history:eceived 24 October 2013eceived in revised form 17 February 2014ccepted 24 February 2014vailable online 14 March 2014

a b s t r a c t

A novel MEMS omnidirectional inertial switch was designed, simulated and fabricated. The switch is com-posed of three main parts, the proof mass, the axial flexible electrode and four radial flexible electrodes.It introduces the flexible electrodes to form a dual mass–spring system. The switch has omnidirectionalsensitivities in a half sphere. When any acceleration, over the threshold value in radial or/and axial direc-tion, acts to the switch, the switch will turn on. Also, the switch has the enhanced contact effect. Dynamic

eywords:mnidirectional inertial switchEMS

lexible electrodenhanced contact effect

simulation results based on FEM confirm that the contact effect is improved by this new design com-pared to that of traditional inertial switch. The contact duration is prolonged under the shock loading,and the bouncing effect is alleviated. The switch has a 6-layer structure, which is manufactured based onnon-silicon surface micromachining technology. The tests have been done and the results coincide withthat of the simulation.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

MEMS inertial switches, also known as to shock sensors orhreshold accelerometers, have great potential to be widely usedn toys, accessories, automotive, military weapons and industrialpplications. That is due to their smaller size, lower cost, less poweronsumption, more functionality and better performance than con-entional mechanical ones. Furthermore, the inertial switches havebility of avoiding electromagnetic interference in applications1,2].

The MEMS omnidirectional inertial switch usually has a struc-ure with a mass–spring system, where the proof mass served ashe movable electrode and is suspended by surrounding springs3] .There is also other structure with a proof mass suspended byentral springs [4–6]. The schematic diagram of these conventionalesigns is shown in Fig. 1. In Fig. 1, when the switch responseso a shock acceleration, the movable electrode will contact withtationary electrode in rigid mode. The contact-bouncing effectould be inevitable and the switch-on time is transient (usually

ess than 10 �s)[7,8]. The poor contact-bouncing effect and thehort switch-on time make it difficult for signal processing and

eaken the reliability of the switch [9,10]. The movable contactoint [1,7], squeeze film effect [1], the carbon nanotube (CNT)-ontact pad [11], electrostatic force [12], have been adopted to

∗ Corresponding author. Tel.: +86 2584303068.E-mail address: njniewr@hotmail.com (X. Zhanwen).

ttp://dx.doi.org/10.1016/j.sna.2014.02.035924-4247/© 2014 Elsevier B.V. All rights reserved.

the inertial micro-switches to prolong the contact duration andeliminate contact-bouncing effect, but these inertial switches werelimited to a single axis of sensing.

The present work proposed a novel design of the MEMSomnidirectional inertial switch with four radial and an axialflexible electrodes shown in Fig. 2. By introducing this flexiblecontact mechanism, the switch forms a dual mass–spring sys-tem, which gives the improved switch characteristics such asomnidirectional sensitivities, reduced contact-bouncing effect andprolonged switch-on time.

2. The inertial switch design and simulation

2.1. Design and working principle

The structure of the omnidirectional inertial switch is illustratedin Fig. 3. It senses acceleration in hemisphere (both in-plane andout-of-plane) with the single proof mass. The proof mass thicknesst is designed to be much larger than the spring thickness t1 in orderto minimize the area coverage while enabling the desired radial andaxial direction sensitivity. Four radial flexible electrodes supportedby spring are symmetrical around the proof mass. Each one hasthe gap of d1 from the proof mass. The circle loop supported by anelastic cross beam is used as the axial flexible electrode. It has a

distance of d2 from the proof mass, which also helps to enhancethe contact effect in axial direction. The use of flexible electrodes isto prolong the switch-on time and make the switch-on state morestable.

94 X. Zhanwen et al. / Sensors and Act

Fig. 1. A schematic diagram of conventional inertial switch.

Fig. 2. A schematic diagram of the novel inertial switch.

Fig. 3. Structural sketch map of the proposed inertial switch. (a) A conventionalinertial MEMS switch with the fixed electrodes (b) A inertial MEMS switch withflexible electrodes.

Fig. 4. Comparison of the dynamic process

uators A 212 (2014) 93–101

Apart from the proof mass–spring system (m1, k1), the radialflexible electrodes are suspended by spring instead of being rigidlyfixed on the substrate. The axial flexible electrode is supported byelastic cross beam. The radial and axial flexible electrodes have con-tact gap d1 and d2 from the proof mass, respectively. For example,as for an acceleration acting along the +Z axis, the working principleis demonstrated by the dynamic process and corresponding switchstate, as shown in Fig. 4. Fig. 4(a) shows the action process of a con-ventional inertial MEMS switch with the fixed electrode. Fig. 4(b)explains the action process of the novel inertial MEMS switch withflexible electrodes. In Fig. 4(b), (1) The proof mass moves towardsone of flexible electrodes due to the acceleration. (2) When theacceleration exceeds the threshold value, the displacement reachesd2, the proof mass contacts the flexible electrode and the switchis turned on. (3) The proof mass keeps moving on with the flexi-ble electrode, therefore the switch-on state is held on for a longertime. (4) After the disappearance of the acceleration, the proof massand the flexible electrode rebound, the switch is turned off, untilthe proof mass is separated from the flexible electrode. (5) Finally,the proof mass and the flexible electrode restore to the equilibriumposition after all the energy is dissipated by free vibration.

As for an acceleration acting along other directions in hemi-sphere, the proof mass moves towards the flexible electrodes inradial and/or axial direction, the switch state is similar to afore-mentioned cases. The inertial switch has radial and axial flexibleelectrodes, and it is able to provide identifiable direction informa-tion according to the identified electrode position.

When acceleration is applied on the switch in the sensitivedirections, the proof mass will be subjected to the inertial acceler-ation a(t) in the opposite direction. Take radial direction motion asexample, the responding motion can be expressed with the motionequation as:

m1x + cx + k1x + m1a(t) = 0 (1)

where m1 is the mass of the proof mass, x is the displacement inradial direction, c presents the damping coefficient, k1 is the systemstiffness and a(t) is the applied acceleration component in radialdirection and is a function of time t. The acceleration a(t) appliedto the switch in practical work is similar to a half-sine pulse withamplitude a0 and duration t0 (about 1 ms). To simplify the solvingprocess, c is neglected [9]. By solving Eq. (1), we have the displace-

ment x(t) of the proof mass.

x(t) = a0

ω2n − ω2

0

(sin ω0t − ω0

ωnsin ωnt

)(2)

and switch state between (a) and (b).

X. Zhanwen et al. / Sensors and Actuators A 212 (2014) 93–101 95

Fv

w

ω

wωa

a

Tωopdbidaoatcta

sfid

Table 1Geometric parameters of the switch.

Component Geometric parameter Value (�m)

Proof mass Radius r 455Thickness t 100

Proof massspring/radial flexibleelectrode spring

Width w1/w2 18/20

Thickness t1/t2 40/40Meander width b1/b2 600/330Connector radius r1/r2 31/20

Axial flexible electrode Circle loop outer radius R3 94Inner radius r3 80Thickness t3 30Cross beam length l4 330Width w4 50Thickness t4 30

Distance to radialflexible electrodes

d1 25

Distance to axialflexible electrode

d2 30

Table 2The displacement of proof mass in radial and axial direction (unit: �m).

˛

0◦ 15◦ 30◦ 45◦

Radial Axial Radial Axial Radial Axial Radial Axial

90◦ 32.0 0.0 32.0 0.0 32.2 0.0 32.1 0.075◦ 31.0 10.0 31.0 10.0 30.9 10.0 30.9 10.160◦ 27.8 19.4 27.8 19.4 27.8 19.5 27.8 19.545◦ 22.7 27.5 22.7 27.5 22.7 27.6 22.7 27.630◦ 16.0 33.8 16.1 33.8 16.0 33.8 16.0 33.915◦ 8.3 37.8 8.3 37.8 8.2 37.8 8.3 37.8

ig. 5. The theoretical curve and simulation results of threshold acceleration ath

ersus ωn .

here

0 = �

t0, ωn =

√k1

m1(3)

hen the maximum value of x(t) is equal to d1(d1 = x0) and0 < ωn < 5ω0, the corresponding applied acceleration is thresholdcceleration ath.

th = x0ω2

n − ω20

(sin(2�ω0/(ωn + ω0)) − (ω0/ωn) sin(2�ωn/(ωn + ω0)))(4)

he threshold acceleration ath is related to gap d1 (d1 = x0) and

n =√

k1/m1. When ω0 < ωn < 5ω0, as ωn increases, the thresh-ld acceleration increases gradually. By changing the mass of theroof mass, different values of ωn are obtained, so the switch withifferent ωn can be simulated. The threshold acceleration obtainedy simulation is consistent with the theoretical results, as shown

n Fig. 5. When the displacement of the proof mass in radial or axialirection is equal to the gap d1 or d2, respectively, the thresholdcceleration ath of the switch in radial or axial direction would bebtained. The mass–spring system stiffness is different in radial andxial direction, so it is needed to adjust gaps d1 and d2. Thereforehe consistent threshold acceleration in radial and axial directionould be obtained. The switch has only axial and radial flexible elec-rodes, so the exactly same threshold acceleration is impossible atny direction in hemisphere for the omnidirectional switch.

The threshold acceleration of the proposed omnidirectional

witch can be calculated using the parameters. And it can be con-rmed by applying different amplitude loads in radial and axialirection.

Fig. 6. The first four natural frequenc

0◦ 0.0 39.3 0.0 39.3 0.0 39.3 0.0 39.3

2.2. Dynamic simulation on the inertial switch

The geometric parameters used in the simulation are listed inTable 1.

The switch model is built in ANSYS. A full simulation model isemployed. The material is nickel, where the Young’s modulus is180 GPa and Poisson’s ratio is 0.3 [13]. Then the model is meshedand the grid of contact area is made to have high quality. Finally,the displacements of outer end of the springs and cross beams areconstrained to be zero at all degrees of freedom.

2.2.1. Modal analysisThe spring–mass systems of designed device structures are

selected for the finite element modal analysis. And simulationmodal results of the switch are shown in Fig. 6. The first four modalfrequencies are 1961.7 Hz, 2101.1 Hz, 2101.1 Hz, 5281 Hz, respec-tively. The first mode shape is in the vertical direction, the second

and third mode shape is in the horizontal direction. The fourthmode is far away from the first three modes.

ies of the spring-mass system.

96 X. Zhanwen et al. / Sensors and Actuators A 212 (2014) 93–101

2

tdmTp

dtum3

tittis

2

a

Fig. 7. Pulse acceleration curve and different direction in hemisphere.

.2.2. The displacement response of proof massBy applying half sine acceleration pulse with the 450 g ampli-

ude and duration 1.0 ms, transient simulation has been done atifferent direction in hemisphere, as shown in Fig. 7. The displace-ents of the proof mass in radial and axial direction are listed in

able 2. Fig. 8 shows a simulated displacement distribution of theroof mass in the hemisphere.

Table 2 shows that the displacements of proof mass in radialirection are related to the included angle between the accelera-ion and the Z-axis. As a result, the omnidirectional switch will haveniform threshold acceleration in the radial directions. The maxi-um displacements of proof mass in radial and axial direction are

2 �m and 39 �m, respectively.When d1 = 25 �m, the threshold acceleration in radial direc-

ion is 355 g, and when d2 = 30 �m, the threshold accelerationn the axial direction is 350 g. Meanwhile, the proof mass con-acts slightly the flexible electrodes, but it is not enough torigger the switch on-state signal. Only if the input accelerations greater than the threshold we can make sure of switch on-tate.

.2.3. The omnidirectional sensitivity of the switchFig. 3 shows that the MEMS omnidirectional switch has radial

nd axial flexible electrodes, the radial and axial flexible electrodes

Fig. 9. The states of the switch contacted betwe

Fig. 8. The maximum displacement of the mass under 450 g in 3-D space.

have a contact gap d1 and d2 from the proof mass, respectively.Ideally, the movement of the proof mass is translational motion,the contact of the proof mass with the flexible electrode occurswhen the displacement component in the radial or axial directionis greater than the gap. In XOZ plane, the maximum displacementcurve of proof mass approximates arc circular (Fig. 8) under thesame acceleration. Therefore, it is impossible to use the exactlysame acceleration value as a switch threshold acceleration in anydirection.

The acceleration threshold in X, Z direction are 355 g and 350 g,respectively, corresponding to the fabricated inertial switch withthe gap values of d1 = 25 �m and d2 = 30 �m. Only the accelerationfrom the Z direction with the value of 350 g can make the switchclosed in the Z direction. When the acceleration is 400 g respectivelywithin the scope of X, Z axis angle of 29◦, the switch can be closedwith a closed blind angle of 32◦. The switch will close under theacceleration of more than 499 g in any direction in the XOZ plane,and there is no closed blind angle.

Fig. 9(a) shows that the switch is closed only in Z-direction, as350 g acting in different directions. Fig. 9(b) shows that it has closedblind corner in XOZ plane, as 400 g acting in different directions.Fig. 9(c) shows that it can be closed in any direction in the xozplane, as 499 g acting in different directions.

For the design of MEMS omnidirectional inertial switch, thethreshold acceleration can be defined as radial and axial direc-tional threshold acceleration respectively. When any acceleration

en the proof mass and flexible electrodes.

X. Zhanwen et al. / Sensors and Actuators A 212 (2014) 93–101 97

he sw

cdt

2

oniacr

oaddflm

Fig. 10. Simulated displacement–time curves of the proof mass in t

omponent in radial or/and axial direction applied to the omni-irectional switch is over the threshold acceleration, the switch willurn on. It indicates that the switch has omnidirectional sensitivity.

.2.4. The switch on-state durationWhen the proof mass is induced by the acceleration, the pair

f switch electrodes move to contact, forming the electrical con-ection. As the two electrodes get closer, the flexible electrode

s moving with the proof mass until the two electrodes are sep-rated. Resultantly, the electrical on-time becomes much longerompared to the conventional inertial micro switches based on twoigid electrodes depicted in Fig. 4(a) and (b).

In order to clearly illustrate the dynamic impact process, thever-threshold acceleration, 450 g, was applied to the switchlong its radial and axial direction. Fig. 10 shows the simulated

isplacement–time curves of the proof mass in radial and axialirection. As a comparison, the contact duration of the radialexible electrodes can prolong to 45 �s in Fig. 10(b), which isuch longer than the electrode being rigidly fixed on substrate in

Fig. 11. The contact duration in different st

itch under 450 g half-sine acceleration in radial and axial direction.

Fig. 10(a). The contact duration of the axial flexible electrode wasprolonged to 33 �s in Fig. 10(d), which is much longer than that inFig. 10(c).

The stiffness coefficient k2 of the radial flexible electrode alsoinfluences the contact-effect, either larger or smaller value will leadto undesirable results. Theoretical formula of the stiffness of microspring was deduced, which has been proven that this analytical cal-culation method and finite element simulation can match to betterthan 1% [14,15]. By reducing the width of the spring, we can geta range of different spring stiffness coefficient. When the stiffnesscoefficient k2 is too smaller, such as 147.98 N/m (the width is 6 �m),the proof mass and the flexible electrodes will have contact effect,the flexible electrodes will bounce, causing discontinuous contactbetween the two electrodes, shown in Fig. 11(a). When the stiffnesscoefficient k2 becomes larger, such as 15312.35 N/m (the width is

26 �m), the displacement of the flexible electrode becomes smallerand the contact duration is shorter, shown in Fig. 11 (c). Takinginto account the fabrication process, the spring width designedis 20 �m, as shown in Fig. 11 (b), which eliminates the bounce

iffness of the radial flexible electrode.

98 X. Zhanwen et al. / Sensors and Actuators A 212 (2014) 93–101

Ft

eek

3

nSmr

ig. 12. Dependence of the dynamic properties of the switch on contact time versushe stiffness.

ffect when the proof mass touches the flexible electrode and hasxtended contact duration. Fig. 12 shows the relationship between2 and contact duration.

. The fabrication of the inertial switch

The omnidirectional inertial switch was fabricated based on

on-silicon surface micromachining technology with multipleU-8 photoresist sacrificial layers on a single wafer. The basicicromachining technologies used herein were six times lithog-

aphy of thick SU-8 photoresist, six times Ni electroforming of

Fig. 13. The fabrication details of the

Fig. 14. SEM image of the omnidirectional inertial switch.

proof mass, axial flexible electrodes, radial flexible electrodes,support springs and anchors, three times sputtering of Cu seedlayers and releasing of the three-dimensional suspended structure.The fabrication details of the omnidirectional inertial switch areshown in Fig. 13. As electrical conductive layer, Cu seed layers weresputtered with the thickness about 200 nm. In order to releasethe three-dimensional suspended structure, inorganic acid boilingwas used to remove the SU-8 sacrificial layer. The SEM image of

the switch with a size of 2.8 mm × 2.8 mm × 210 �m and minimumlinewidth 15 �m is shown in Fig. 14. Because the omnidirectionalinertial switch is a six-layer device and was fabricated by six times

omnidirectional inertial switch.

X. Zhanwen et al. / Sensors and Actuators A 212 (2014) 93–101 99

Fig. 15. Shock test devices for micro-switch prototypes.

Fig. 16. Shock test results of micro-switch under the acceleration (a) 380 g, (b) 450 g in X-axis direction.

Fig. 17. Shock test results of micro-switch in the direction of (a) 45◦with the X-axis direction in the XOY plane, (b) 45◦with the Z-axis direction in the XOZ plane.

100 X. Zhanwen et al. / Sensors and Actuators A 212 (2014) 93–101

the ac

Nw

4

w5ataaeb

eWdi

1aaa

otatttosmittdt

mdat

Fig. 18. Shock test results of micro-switch under

i electroplating, the main difficulty met in fabrication processas low bonding strength among electroforming deposit layers.

. Testing and results analysis

The fabricated inertial switch prototype was fixed on test boardith a standard accelerometer (CA-YD-180) with sensitivity of

.292 mV/g and a voltage division circuit with DC supply voltage 3 Vnd load resistance 200 �. A data acquisition system was adoptedo capture the acceleration data when the switch had output volt-ge, as shown in Fig. 15. The different testing angle between switchnd acceleration was achieved by a special fixture. Half-sine Accel-rations with different amplitude and pulse width can be obtainedy adjusting the shock table parameters.

Fig. 16 shows the results of shock test, where different accel-ration amplitudes in X-axis direction are applied to the switch.hen the acceleration amplitude is about 380 g, 450 g, the contact

uration of the switch is about 40 �s, 60 �s, respectively, as shown Fig. 16.

Also, a half-sine acceleration with amplitude 450 g and duration ms was applied to the switch in different shock directions, suchs 45◦with the X-axis direction in the XOY plane, Z-axis directionnd 45◦with the Z-axis direction in the XOZ plane. The test resultsre shown in Fig. 17.

It is shown in Figs. 16–18 that the switch can be reliably turnedn in radial direction, axial direction and a certain direction withhe axial direction under the applied half-sine acceleration withmplitude 450 g and duration 1 ms, the switch with flexible elec-rode is very beneficial for improving contact effect and prolonginghe contact duration. However, there is a little deviation on its con-act duration compared with the simulation results. This is becausef fabrication etching technology error. When the spring width ismaller, the stiffness decreases obviously and the mass displace-ent increases, which causes increase of the contact duration. So it

s considered as the main possible reason for the deviation betweenhe test and simulation results. When higher acceleration is appliedo the switch in Z-axis direction, the contact duration of the switchecreases inversely as shown in Fig. 18(a) and (b), which is due tohe bounce phenomenon previously mentioned.

Based on the measured parameters of the switch prototype, by

odifying the simulation model with the thickness of circle loop

ecreased to 22 �m, the width of cross beam decreased to 42 �m,nd the spring width of proof mass and radial electrode decreasedo 14 �m and 18 �m, respectively, the test results coincide with the

[

celeration (a) 450 g, (b) 500 g in Z-axis direction.

simulation results. It is necessary to further optimize the structuralparameters to improving performance of the switch.

5. Conclusion

The omnidirectional inertial switch has been realized in thisdesign by integration of single proof mass and flexible electrodes.The flexible electrodes were utilized for omnidirectional inertialswitch to extend the contact duration and therefore to obtainreliable and stable output signals. When the proof mass collideswith the flexible stationary electrode, it keeps moving to the flex-ible electrodes. Therefore the switch-on state is held on for alonger time, significantly prolonging the contact duration. Whenthe acceleration component in radial or/and axial direction appliedto the omnidirectional switch is over the threshold acceleration, theswitch will turn on. It indicates that the switch has omnidirectionalsensitivities. The tests verify the simulation results.

Acknowledgments

This work is supported by National Natural Science Foundationof China (51075057).

References

[1] T. Matsunaga, M. Esashi, Acceleration switch with extended holding time usingsqueeze film effect for side airbag systems, Sens. Actuators A: Phys. 100 (2002)10–17.

[2] J. Zhao, J. Jia, H. Wang, W. Li, A novel threshold accelerometer with post-buckling structures for airbag restraint systems, IEEE Sens. J. 7 (2007)1102–1109.

[3] C.H. Robinson, Omnidirectional microscale impact switch, 2004, US 6,765,160.[4] D.J. Jean, O. MD, MEMS multi-directional shock sensor, 2007, US 7,159,

442.[5] D.S. Greywall, MEMS-based inertial switch, 2007, US 7,218,193.[6] G.L. Smith, O. MD, Triaxial MEMS acceleration switch, 2012, US 8,237,521.[7] H.G. Cai, G.F. Ding, Z.Q. Yang, Z.J. Su, J.S. Zhou, H. Wang, Design,

simulation and fabrication of a novel contact-enhanced MEMS inertialswitch with a movable contact point, J. Micromech. Microeng. 18 (2008)115033.

[8] Z. Yang, G. Ding, H. Cai, X. Xu, H. Wang, X. Zhao, Analysis and elimination ofthe ‘skip contact’ phenomenon in an inertial micro-switch for prolonging itscontact time, J. Micromech. Microeng. 19 (2009) 045017.

[9] W. Ma, Y. Zohar, M. Wong, Design and characterization of inertia-activatedelectrical micro-switches fabricated and packaged using low-temperature pho-

toresist molded metal-electroplating technology, J. Micromech. Microeng. 13(2003) 892–899.

10] Z.Q. Yang, G.F. Ding, H.G. Cai, X.L. Zhao, A MEMS inertial switch with bridge-typeelastic fixed electrode for long duration contact, IEEE Trans. Electron Devices55 (2008) 2492–2497.

nd Act

[

[

[

[

strips on the substrate, IEEE Trans. Compon. Packag. Manufac. Technol. 1 (2011)

X. Zhanwen et al. / Sensors a

11] J. Lee, Y. Song, H. Jung, J. Choi, Y. Eun, J. Kim, Deformable carbon nanotube-contact pads for inertial microswitch to extend contact duration, IEEE Trans.Ind. Electron. 59 (2012) 4914–4920.

12] M. Jia, X. Li, Z. Song, M. Bao, Y. Wang, H. Yang, Micro-cantilever shocking-acceleration switches with threshold adjusting and ‘on’-state latchingfunctions, J. Micromech. Microeng. 17 (2007) 567–575.

13] K.J. Hemker, H. Last, Microsample tensile testing of LIGA nickel for MEMS appli-cations, Mater. Sci. Eng. A 319 (2001) 882–886.

[

uators A 212 (2014) 93–101 101

14] Z.Q. Yang, G.F. Ding, H. Wang, H.G. Cai, X.L. Zhao, Modeling, simulation and char-acterization of a micromachined acceleration switch with anti-stiction raised

1195–1204.15] G.K. Fedder, Simulation of microelectromechanical systems (Ph.D. thesis),

Dept. Electr. Eng. & Comput. Sci., Univ. California at Berkeley, Berkeley,1994.