You are on page 1of 12

Microsyst Technol (2016) 22:1067–1078

DOI 10.1007/s00542-015-2469-8

TECHNICAL PAPER

Effect of creep in RF MEMS static and dynamic behavior


Aurelio Somà · Muhammad Mubasher Saleem ·
Giorgio de Pasquale 

Received: 28 January 2015 / Accepted: 10 February 2015 / Published online: 21 February 2015
© Springer-Verlag Berlin Heidelberg 2015

Abstract  This paper presents the experimental charac- mechanical and electrical failure modes. The mechanical
terization of the creep effect in electrostatically actuated failure modes include fracture (due to overload, shock),
gold microstructures. The tested specimens follow the typi- fatigue (due to repeated loading over a long period of time),
cal configuration of the microbridge based radio frequency wear (related to the surface adhesion, abrasion) and creep
microelectromechanical systems switches and varactors. (due to viscoelastic and plastic time dependent deforma-
Initially, the plastic creep strain accumulation with time tion of material under the influence of the applied mechani-
is measured for the specimens with different geometric cal stress and temperature). The electrical failure modes in
dimensions and at different actuation voltages and tem- MEMS are stiction (due to electrostatic charging), charge
peratures. To avoid the size and cumulative heating effects, accumulation (due to electric stress and radiation) and
three specimens with the same geometric dimensions, electrical shorts (due to dielectric material degradation,
actuation voltages and constant temperatures are tested. electromigration and oxidation) (Huang et al. 2012). In the
The test results allowed decoupling the permanent plas- literature several experimental studies have been reported
tic strains due to the creep effect and reversible anelastic to predict and explain the MEMS failure modes (Younis
strains due to the viscoelastic behavior. The pull-in voltage et al. 2007; Subhash et al. 2011; Li et al. 2013; Brusa et al.
and natural frequency values measured before and after the 2013; Langfelder et al. 2009). The authors have previously
creep tests are compared, revealing the mechanical stiffness presented the effects of the alternate loads on the gold
decrease caused by creep. microstructures fatigue behavior and observed some rela-
tions to the well established fatigue mechanism at macro-
scale (Somà and De Pasquale 2009; De Pasquale and Somà
1 Introduction 2011). The experimental methodologies for the investiga-
tion of different parameters effect on the dynamic response
The accurate characterization of microelectromechanical of the microsystems strictly related to the microscale are
systems (MEMS) failure modes has been a major concern presented by the authors in De Pasquale and Somà (2009,
for the MEMS reliability studies in last years. A number 2010).
of potential failure modes have been identified and decou- Highly conductive metal thin films of Al, Ni, Au and Pt
pled and can be categorized into two major groups i.e. have been widely used in radio frequency microelectro-
mechanical systems (RF MEMS) for improved electrical
performance. However, reduced thickness of thin films and
A. Somà · M. M. Saleem (*) · G. de Pasquale 
fabrication techniques such as sputter deposition, electro-
Department of Mechanical and Aerospace Engineering,
Politecnico di Torino, Corso Duca degli Abruzzi 24, Turin, Italy deposition and evaporation result in a considerable differ-
e-mail: mubasher.saleem@polito.it ence of the electrical (conductivity, resistivity), mechanical
A. Somà (intrinsic stress, elasticity, ductility, adhesion) and thermal
e-mail: aurelio.soma@polito.it (expansion coefficient, thermal conductivity, melting point)
G. de Pasquale properties compared to those of bulk material (Wetzig
e-mail: giorgio.depasquale@polito.it 2003). The performance and life of the metal structure

13

1068 Microsyst Technol (2016) 22:1067–1078

layer based RF MEMS is significantly affected by the tem- in MEMS accelerometers made of Si, Ni and Au and
perature induced plastic deformation, especially for the demonstrated Au to be most affected, considering the
space applications where the devices are kept actuated for decrease in device turn-on voltage over time period of
long time under severe temperature fluctuations. 30 h. Tuck et al. (2005) observed plastic deformation in
Creep is the time dependent accumulation of plas- 2  μm thick polysilicon film at a temperature of 1300 K
tic strain in the material under the influence of an applied and stress level of 89 MPa for a short, 1 min, time dura-
stress. Figure 1 shows the typical creep-strain curve associ- tion. van Gils et al. (2007) measured permanent deforma-
ated to creep. The shape and the duration of creep stages tion in the springs of the 3–5 μm thick top Al electrode
depend on the applied temperature and stress. In the pri- switch at 75 and 100 °C for 15 h. Yan et al. (2009) showed
mary stage, creep strain rate decreases due to the increasing 25 % change in the time dependent modulus of elastic-
of material strength as a result of the dislocation movement ity for 1.2 um electroplated Au MEMS switch over 72 h
in the crystal lattice of the material (work hardening). In under constant strain at room temperature. Hsu et al.
the secondary stage, creep strain rate reaches its minimum (2011) demonstrated the time dependent change in the
due to the balance between the work hardening and thermal capacitance of an electroplated Ni based RF-MEMS var-
softening. The creep life of the material strongly depends actor at room temperature under bias voltage of 40 V.
on the secondary stage steady-state creep strain. The vari- Bergers et al. (2011) observed time-dependent deforma-
ables affecting the secondary stage steady-state creep strain tions in free-standing aluminum-alloy cantilever beam of
are generally identified to model the dominant creep mech- 4.8  μm thickness. Jain et al. (2012) presented a physics-
anism in the material. In the tertiary stage, creep strain rate based theoretical modeling framework to predict the time
accelerates as a result of the accumulated damage, followed dependent capacitance change of a varactor due to the vis-
by the fracture of the specimen. The creep strains are the coelastict behavior. Recently, Scott et al. (2013) presented
combination of the permanent strains (i.e. creep effect) the reliability experimental studies on bimorphs for tem-
and the fully recoverable strains, often termed as anelastic perature sensors including creep effects.
strains due to the viscoelastic behavior. The main objectives of this work are: (a) to investigate
In MEMS, the effect of the creep was originally the creep mechanism in metal thin film based RF-MEMS
observed in digital micromirror devices (DMD) fabri- devices operating under constant actuation and temperature
cated of Al (Douglass 1998). The unloaded position of conditions for a long time (b) to present the applicability
the micromirror was tilted by some degrees with respect and verification of the creep theory at the macromechan-
to the original position for a long time bias voltage and ics in the microscale (c) to decouple the permanent creep
65 °C temperature. Some of the significant initial exper- strains due to plastic deformation and recoverable anelastic
imental investigations into the creep phenomenon in strains due to viscoelastic behavior and finally (d) to study
MEMS are presented in Vickers-Kirby et al. (2000), the effect of the creep mechanism on the mechanical stiff-
Cho et al. (2002), Allameh (2003) and Modlinski et al. ness of the microstructures and hence on the static and the
(2004, 2005). Vickers-Kirby et al. (2000) reported creep dynamic response of the RF-MEMS devices.

2 Specimen design and microfabrication

2.1 Fabrication technology

The specimens used for the experimental investigation of


the creep phenomenon in gold thin films are fabricated
through the gold electrodeposition process, referred as RF
Switch Surface Micromachining (RFS) process, developed
by the Bruno Kessler Foundation (FBK) (Margesin et al.
2003; Kal et al. 2006). The moving structures are built
through two steps electroplating process over a photore-
sist of 3 μm. A seed layer of chromium/gold is deposited,
followed by the first and second electroplated gold layers,
with different thickness. This method allows the selective
superimposition of the two gold layers. The seed layer of
chromium/gold is used to improve the adhesion of the gold
Fig. 1  Typical creep strain–time relationship curve of pure metal with substrate.

13
Microsyst Technol (2016) 22:1067–1078 1069

2.2 Specimen geometry and actuation for creep tests dimensions are measured optically by the interferometric
profilometer ZoomSurf 3D (Fogale Nanotech). The maxi-
The schematic of the electrostatically actuated microbridge mum lateral and vertical resolution of the optical micro-
specimens, used in this study is shown in Fig. 2. The central scope are ±0.3  μm and ±0.5  × 10−4  μm, respectively.
plate is constrained at both sides, using four linear clamped The thickness values for the central plate and the beam
supports, resembling the typical configuration of the micro- supports are measured at pull-in when the plate was com-
bridge based RF-MEMS switch. This particular shape with pletely deflected. The gap between the central plate and
the lateral supports allows an out-of-plane deflection when the bottom electrode is measured indirectly by subtract-
a bias voltage is applied between the central plate and the ing the central plate thickness and the vertical deflection at
lower polysilicon electrode. The microbridge deflection pull-into the height of the central plate at zero bias voltage.
controls the capacitance change in the RF-MEMS varac- An average value of the gap is calculated by considering
tor and the short circuit induced current flow in the corre- the initial central plate curvature in the center. The dimen-
sponding lines of the RF-MEMS switches. The specimen sions reported in Table 1 have been averaged over all the
dimensions are designed to be reasonably close to some of specimens used for the experimental investigation. Figure 3
the RF-MEMS devices developed by using the same fabri- shows the optical image and 3D profile measurement of the
cation process. Table 1 shows the actual dimensions of the test specimen with the central perforated plate supported by
nine specimens tested for the creep characterization. The the elastic beams. The presence of the holes facilitates the

Fig. 2  Schematic of the speci-


men as viewed from above (top)
and from the side (bottom)

Table 1  Measured specimen dimensions with testing actuation voltage and temperature values
Plate Plate Support Support Number Hole Plate Support Air gap Actuation Temperature
length width length (ls) width (ws) of holes size thickness thickness voltage (V) (°C)
(lb) (wb) (tb) (ts)

Specimen 1 150 150 94 12.2 49 10.4 5.99 1.99 3.2 60 60


Specimen 2 131.1 131.5 93.6 12.1 36 7.5 6.3 2.22 3.4 60 60
Specimen 3 132 132 93.4 12 36 7.8 6.3 2.20 3 30 70
Specimen 4 132.1 132 93.4 12 36 7.8 6.32 2.1 3 30 70
Specimen 5 132 132.1 93 12 36 7.9 6.28 2.21 3.1 40 70
Specimen 6 132 132 93.5 12 36 7.8 6.3 2.20 2.9 40 70
Specimen 7 132.1 132.1 93.5 12 36 13 6.3 2.35 3.62 60 60
Specimen 8 132.1 132 93.2 12.2 36 13 6.3 2.32 3.60 60 60
Specimen 9 132 132 93 12.2 36 13 6.3 2.3 3.46 60 60

All dimensions are in μm

13

1070 Microsyst Technol (2016) 22:1067–1078

chemical removal of the sacrificial layer, used to obtain the the applied voltage and the equivalent stress distribution at
suspended parts and achieve a final gap of 3 μm. different actuation voltages is obtained using FEM models.
The non-linear electromechanical coupling between the
2.3 FEM modeling electrical and the mechanical domains is modeled by using
the 1-D multiphysics transducer elements trans126. These
The equivalent stress distribution in the test specimens elements take on the form of a lumped element with volt-
is estimated by developing 3D non-linear FEM mod- age and structural degree-of-freedoms (DOFs) as cross var-
els in ANSYS. The specimens are modeled with actual iables and current and force as through variables. Figure 5
dimensions, shown in Table 1. The adopted values for the shows the contour plots of the voltage induced equivalent
Young’s modulus, Poisson ratio and density of the gold are stress distribution and the corresponding vertical down-
98.5 GPa, 0.42 and 19.32 × 10−15 kg/μm3 respectively. The ward displacement of the central plate of the specimen at
effect of the applied temperature on the stress distribution the actuation voltage of 60 V. The combined effect of the
and deflection profile of the specimens is simulated with thermal heating and actuation voltage on the stress distribu-
coupled thermal-structural analysis. Figure 4 shows that tion in the test specimens at the actuation voltage of 60 V
at the applied temperature of 60 °C the maximum stress and temperature of 60 °C is shown in Fig. 6. The maximum
is concentrated in the connecting supports and the central stress is 104 MPa and is extremely localized at the connec-
plate is deflected upwards. The static relationship between tion of the connecting supports and central plate.

Fig.  3  a Optical image of the electroplated gold sample and b 3D profile measurement detected by the interferometric microscope

Fig. 4  Thermal heating at 60 °C only a Von-Mises stress distribution, b specimen vertical deflection

13
Microsyst Technol (2016) 22:1067–1078 1071

Fig. 5  Electrostatic actuation at 60 V only a Von-Mises stress distribution, b specimen vertical deflection

Fig. 6  Both thermal heating at 60 °C and electrostatic actuation at 60 V. a Von-Mises stress distribution, b specimen vertical deflection

3 Experimental setup specimens is applied by using a voltage generator capable


of generating voltage in the range of 0–200 V, with up to
The tests for the creep evaluation in the specimens are 2 MHz frequency. The connection of the power supply to
conducted by using an interferometric microscope Fogale the specimens is made through adjustable needles mounted
ZoomSurface 3D with a voltage generator and a Peltier cell on the Probe Head PH100 SUSS, which is magnetically
with a temperature sensor, as shown in Fig. 7. The Fogale fixed to the workplane of the profiling system.
optical profiling system is based on non-contact optical The thermal heating is provided with the thermoelectric
interferometry. The profile of the specimen and its vertical Peltier cell that is supplied by a dedicated voltage genera-
position is measured by detecting the recorded light inten- tor. A PT100 temperature sensor is applied on the Peltier
sity after the interference, as a function of the specimen cell surface to provide the temperature feedback measure-
height, by a CCD camera. The image processing consists of ment. The precise profile measurement of the specimens
eliminating the specimen tilting and applying median filter with increasing time, under constant actuation voltage and
to remove the artifact peaks. The actuation voltage for the temperature, is obtained by placing the Peltier cell with

13

1072 Microsyst Technol (2016) 22:1067–1078

Fig.  7  a Interferometric microscope Fogale ZoomSurface 3D. b Peltier cell with temperature sensor for heating the specimens

temperature sensor on the movable stage of the profilom- show the creep strain accumulation, in the form of cen-
eter. The specimens are tested in the temperature range of tral plate displacement, for the two specimens tested at
40–70 °C, according to some of the operating conditions of 30 and 70 °C. The specimens showed a sharp decreasing
the RF-MEMS devices mentioned in the literature (Rebeiz strain rate for the first 4 h followed by a relatively steady
2003). strain rate for the next 4 h. The central plate deflection
after 8 h of creep tests for both the specimens is 0.16 and
0.18 um for the specimen 3 and specimen 4 respectively
4 Results and discussion (Fig. 9). In the case of the two specimens tested at actua-
tion voltage of 40 V and 70 °C the central plate deflection
4.1 Actuation voltage of 60 V and applied temperature is higher as compared to the specimens tested at 30 V and
of 60 °C 70 °C as shown in Fig. 10. The plate deflection after 8 h
of test is 0.19 μm and 0.21 μm for the specimen 5 and 6
Initially, the progressive creep strain is measured for the respectively.
first two specimens with different geometric dimension The creep strain curves for the first six specimens
under same test conditions i.e. at the actuation voltage of showed in Figs. 8, 9 and 10 are consistent with the expected
60 V and applied temperature of 60 °C. The creep strain trend of accumulated creep strain accumulation curve
accumulation is measured by monitoring the central plate shown in Fig. 1. For the first 3–4 h the specimens showed
deflection at the fixed intervals of time. Figure 8 shows the a behavior similar to the primary creep with a decrease in
creep strain curve for the two tested specimens. The speci- strain rate with increasing strain. For the next 4 h the creep
men 1 showed an increasing strain for the first 3 h, followed curves approximate a constant trend resulting in a behav-
by steady strain for the next 5 h. The displacement in the ior consistent with secondary creep. The snap down of the
specimen 2 increased linearly for the 7 h without entering specimen 2 in Fig. 8, as a result of excessive creep strain
the steady state region and eventually snapped down after exceeding the instability threshold, represents a behavior
7 h (snap down is indicated by the cross mark). The plate similar to the tertiary creep.
displacement behavior over time is quite different for the
two specimens. The plausible reason for this difference is 4.3 Repeatability tests and decoupling of anelastic
the cumulative heating of the specimen 2: both specimens and permanent creep
are on the same die and specimen 1 has been tested first.
The creep strain accumulation in the specimen 1 after 8 h The creep strain curves are generally the combination of
of testing is 0.3 μm. the time dependent permanent strains (creep effect) and
fully recoverable strains (anelastic creep effect). The accu-
4.2 Actuation voltage of 30–40 V and applied temperature rate characterization and decoupling of the permanent and
of 70 °C anelastic creep strains is very important for the long term
reliability of MEMS devices. To decouple these mecha-
Four specimens are tested at the same applied tempera- nisms and avoid the sequential heating and size effect on
ture and low actuation voltages of 30 and 40 V. The results the creep strain accumulation, three specimens with the

13
Microsyst Technol (2016) 22:1067–1078 1073

Fig. 8  Measured displacement
of the central plate of specimen
1 and specimen 2 with time, at
60 V and 60 °C

Fig. 9  Measured displacement
of the central plate of specimen
3 and specimen 4 with time, at
30 V and 70 °C

Fig. 10  Measured displacement
of the central plate of specimen
5 and specimen 6 with time, at
40 V and 70 °C

13

1074 Microsyst Technol (2016) 22:1067–1078

Fig. 11  Measured central plate


displacement for specimen 7, 8
and 9 tested at 60 V and 60 °C.
A permanent measured central
plate deflection after 3 days of
relaxation

Fig. 12  Longitudinal optical
profile of the specimen 9 before
and after the creep test

same geometric dimensions and test conditions but on the deflection after 8 h is 0.28 μm for the specimen 7, 0.26 μm
separate dies are tested. Before the creep tests the pull- for the specimen 8 and 0.3 μm for the specimen 9. These
in voltage and the natural frequency values are measured deflection values are comparable to the plate deflection
experimentally for each specimen. The biasing of the test after 8 h for the specimen 1 tested at actuation voltage of
structures for the resonance frequency and pull-in volt- 60 V and temperature of 60 °C shown in Fig. 8. After 25 h
age extraction is obtained by applying a voltage difference the specimen 7 snapped down while for the specimen 8 and
between the upper suspended plate and the ground fixed 9 a central plate deflection of 0.48 and 0.56 μm is meas-
plate. The ground plate fully covers the area under the cen- ured respectively. The specimens are then allowed to relax
tral plate. The specimens are tested at the actuation voltage for three days by removing the applied actuation voltage
of 60 V and constant temperature of 60 °C for 25 h. and temperature. A permanent central plate deflection of
Figure  11 shows the creep strain curves for the three 0.14, 0.15 and 0.12 μm is measured for the specimen 7, 8
specimens. The specimens show a decreasing strain rate and 9 respectively. Figure 12 shows the longitudinal optical
with increasing plate deflection for the first 2 h followed profiles, including both the anchor and central plate, of the
by a constant strain rate. The maximum central plate specimen 9 before and after the creep test.

13
Microsyst Technol (2016) 22:1067–1078 1075

Table 2  Measured natural frequency values for the specimens before The mechanical natural frequency is directly propor-
and after the creep test tional to the effective spring constant. The spring constant k
Before creep test After creep test ffinal for the microspecimens tested in this study is given by:
finitial
natural frequency natural frequency  3
kHz (finitial ) kHz (ffinal ) t
k = 4Ew
l (1)
Specimen 7 33.4 32.4 0.97
Specimen 8 33.6 32.3 0.96 where E is the Young’s modulus, t is the support thickness,
Specimen 9 32.9 30.5 0.93 w is the support width and l is the support length. Assum-
ing that all the other parameters remain unchanged, the
decrease in the natural frequency values for the specimens
can be attributed to the permanent decrease in the mechani-
Table 3  Measured pull-in voltage values for the specimens before cal stiffness due to the developed creep strains. This shows
and after the creep test
that the creep phenomenon results in the permanent change
Before creep test After creep test VPI final in the Young’s modulus values. For the verification of this
VPI initial
pull-in voltage (V) pull-in voltage (V) phenomenon 3D FEM models are developed in ANSYS and
(VPI initial ) (VPI final )
the corresponding natural frequency values are obtained for
Specimen 7 97 89 0.92 the three specimens with the measured dimensions reported
Specimen 8 98 88 0.90 in Table 1. Initially the natural frequency values measured
Specimen 9 90 79 0.88 before the creep the test are compared with the FEM model
frequency values. The standard value of the Young’s mod-
ulus i.e. 98.5 GPa is assumed in the simulations. Table 4
The measured profile of the specimens after three days of shows the comparison of the natural frequency values
relaxation shows that almost one-fourth of the creep strains obtained experimentally with the FEM simulation values
are plastic and permanent strains while the remaining meas- before the creep test. The consistency of the experimental
ured strains are fully recoverable and can be attributed to and FEM model natural frequency allows the parametric
the anelasticity. It shows that at low stress and temperatures identification and verification of the permanent change in
dominant mechanism for the strains accumulation is anelas- the mechanical stiffness of thethree specimens due to the
creep effect. The ffinal finitial ratios for the three speci-

ticity. This anelastic stress relaxation effect is reported by
Yan et al. (2009) for the gold films of 1.2 μm thickness, and mens (shown in the Table 2) and the fact that the mechani-
cal natural frequency ω = k m, depicts a decrease of 6,

a 20 % reduction in the Young’s modulus is observed at con-
stant strain rate. This time dependent change in the modulus 8 and 14 % in the Young’s modulus for the specimen 7, 8
leads to a time dependent change in the effective spring con- and 9 respectively. Table 4 shows the natural frequency val-
stant and eventually results in the change of both the static ues for the three specimens obtained using the FEM models
and dynamic response of the device. with the respective decrease in the elastic modulus depicted
by the experimental measurements.  The closeness of the
4.4 Effect of the creep on dynamic and static response FEM model and experimental ffinal finitial ratios confirms
 

the permanent decrease in the mechanical stiffness due to


To study the effect of the plastic creep strain accumulation the variation in the elastic modulus values.
of with time on the dynamic and static response of the test The permanent decrease in the mechanical stiffness due
specimens the natural frequency and pull-in voltage val- to the creep strains is verified by comparing the measured
ues measured before and after the creep test are compared. pull-in voltage values before and after the creep tests. The
Tables  2 and 3 show the comparison of the natural fre- pull-in voltage for the tested specimens is given as;
quency and pull-in voltage values for the specimen 7, 8 and 
9 measured before and after the creep tests. The difference 3
8 kdave
between the initial frequency before creep test and final fre- VPI = (2)
27 εo A
quency after creep tests and three days of relaxation is 3,
4 and 7 % for the specimen 7, 8 and 9 respectively. The where dave is the average gap height between the central
difference between the static pull-in voltage values for the plate and the bottom electrode, A is the electrode area and
three specimens, before and after the creep test, is higher as εo is the vacuum dielectric permittivity. The pull-in voltage
compared to the natural frequency as shown in Table 3. The value depends on both the mechanical stiffness and elec-
measured difference is 8, 10 and 12 % for the specimen 7, trostatic gap height. After the creep tests the central plate
8 and 9 respectively. is observed to be deflected downwards as discussed earlier.

13

1076 Microsyst Technol (2016) 22:1067–1078

Table 4  Comparison of the experimental and FEM model natural frequency values before and after the creep test

Before creep test natural frequency kHz (finitial ) After creep test natural frequency kHz (ffinal ) ffinal
finitial

Experimental 3D FEM with nominal stiffness Experimental 3D FEM with decreased stiffness Experimental 3D FEM

Specimen 7 33.4 33 32.4 32.1 0.97 0.97


Specimen 8 33.6 33.1 32.3 31.7 0.96 0.96
Specimen 9 32.9 32.6 30.5 30.3 0.93 0.93

Table 5  Comparison of the experimental and FEM model pull-in voltage values before and after the creep test, considering only plate deflection
Before creep test pull-in voltage (V) After creep test pull-in voltage (V) VPI final
VPI initial
(VPI initial ) (VPI final )
Experimental 3D FEM with Experimental 3D FEM with Experimental 3D FEM with
nominal stiffness measured deflection measured deflection

Specimen 7 97 96 89 91.2 0.92 0.95


Specimen 8 98 96.2 88 91 0.90 0.94
Specimen 9 90 89.4 79 81.8 0.88 0.91

Table 6  Comparison of the experimental and FEM model pull-in voltage values before and after the creep test, considering both the plate
deflection and modulus reduction
Before creep test pull-in voltage (V) After creep test pull-in voltage (V) VPI final
VPI initial
(VPI initial ) (VPI final )
Experimental 3D FEM with Experimental 3D FEM with Experimental 3D FEM with
nominal stiffness measured deflection measured deflection and
and decreased stiffness decreased stiffness

Specimen 7 97 96 89 88 0.92 0.916


Specimen 8 98 96.2 88 88 0.90 0.914
Specimen 9 90 89.4 79 78.4 0.88 0.876

The measured permanent central plate deflection of the The FEM model results show that the central plate
specimen 7, 8 and 9 is 0.14, 0.15 and 0.12 μm respectively. deflection only, does not explain the measured difference
Initially the static pull-in voltage for the three specimens between the pull-in voltage values before and after the
is obtained using the 3D FEM models with the measured creep tests. Next, 3D FEM models are developed and the
dimensions and nominal value of the Young’s modulus. effect of the reduction in the modulus values as predicted by
Table 5 shows the comparison of the experimentally meas- the natural frequency comparison is included in the simula-
ured pull-in voltage values and obtained using the FEM tion along with the measured deflection. Table 6shows the
models before the creep test. Assuming that the mechanical comparison of the pull-in voltage ratio (VPI final VPI initial )
stiffness after the creep test is same as before the test and measured experimentally and by considering the effect of
the measured decrease in the pull-in voltage is only due to both the permanent deflection and reduction in the Young’s
the downward deflection of the central plate, the ratio of the modulus.
final pull-in voltage after the creep test
 and initial pull-in  The  good correspondence of the experimental
voltage before the creep test (VPI final VPI initial ) obtained ffinal finitial and (VPI final VPI initial ) ratio with that
 

using the FEM simulations is 0.95, 0.94 and 0.91 for the obtained using FEM models, considering both the plate
specimen 7, 8 and 9 respectively. This shows that the meas- deflection and stiffness variation, confirms that the creep
ured downward deflection of the central plate, due to the phenomenon in RF microstructures can result in the perma-
creep phenomenon, accounts only 5, 6 and 9 % decrease nent shift in the static and dynamic response of the device.
in the pull-in voltage after the creep test for the three The creep induced plastic strains result in the permanent
specimens. variation in the device profile and stiffness.

13
Microsyst Technol (2016) 22:1067–1078 1077

5 Conclusion De Pasquale G, Somà A (2010) Dynamic identification of electrostati-


cally actuated MEMS in the frequency domain. Mech Syst Signal
Process 24(6):1621–1633
A progressive accumulation of creep strains with time is De Pasquale G, Somà A (2011) MEMS mechanical fatigue: effect of
observed for the multiple electrostatically actuated gold mean stress on gold microbeams. J Microelectromechanical Syst
microspecimens at fixed actuation voltage and tempera- 20(4):1054–1063
ture values. The experimental creep strain curves for all Douglass MR (1998) Lifetime estimates and unique failure mecha-
nisms of the digital micromirror device (DMD). In: Reliability
the tested specimens follow the typical creep strain curve physics symposium proceedings, pp 9–16
for the metals with the decreasing strain rate for the first Hsu HH, Koslowski M, Peroulis D (2011) An experimental and
2–3 h (primary stage) followed by steady state rate for next theoretical investigation of creep in ultrafine crystalline
few hours (secondary stage) and in two specimens a sudden nickel RF-MEMS devices. IEEE Trans Microw Theory Tech
59(10):2655–2664
snap down (tertiary stage) as a result of the excessive creep Huang Y, Vasan ASS, Doraiswami R, Osterman M, Pecht M (2012)
strain exceeding the instability threshold. MEMS reliability review. IEEE Trans Device Mater Reliab
Three specimens with the same geometric dimensions 12(2):482–493
are tested for 25 h at the same actuation voltage of 60 V Jain A, Palit S, Alam MA (2012) A physics-based predictive mode-
ling framework for dielectric charging and creep in RF MEMS
and constant temperature of 60 °C. The central plate deflec- capacitive switches and varactors. J Microelectromechanical Syst
tion in the range of 0.48–0.56 μm is measured after 25 h. 21(2):420–430
The specimens are then allowed to relax for three days and Kal S, Bagolini A, Margesin B, Zen M (2006) Stress and resistivity
a permanent deformation of the central plate in the range analysis of electrodeposited gold films for MEMS application.
Microelectron J 37(11):1329–1334
of 0.12–0.15 μm for the three specimens is measured. Langfelder G, Longoni A, Zaraga F, Corigliano A, Ghisi A, Merassi A
This suggests that almost one-fourth of the creep strains (2009) A new on-chip test structure for real time fatigue analysis
are permanent (creep effect) and the remaining anelas- in polysilicon MEMS. Microelectron Reliab 49(2):120–126
tic strains are fully recovered and can be attributed to the Li G, Zhang W, Li P, Sang S, Hu J, Chen X (2013) Investigation of
charge injection and relaxation in multilayer dielectric stacks for
viscoelasticy. capacitive RF MEMS switch application. IEEE Trans Electron
The variation in the natural frequency values measured Devices 60(7):2379–2387
before and after the creep tests is explained with the perma- Margesin B, Bagolini A, Guarnieri V, Giacomozzi F, Faes A, Pal
nent reduction in the mechanical stiffness. This decrease in R, Decarli M (2003). Stress characterization of electroplated
gold layers for low temperature surface micromachining. In:
the mechanical stiffness is also verified by comparing the Design, test, integration and packaging of MEMS/MOEMS,
pull-in voltage values of the three specimens before and pp 402–405
after the creep tests. The decrease in the mechanical stiff- Modlinski R, Witvrouw A, Ratchev P, Puers R, den Toonder JMJ, De
ness as depicted by the natural frequency comparison and Wolf I (2004) Creep characterization of Al alloy thin films for use
in MEMS applications. Microelectron Eng 76(1):272–278
the permanent central plate downward deflection justifies Modlinski R, Ratchev P, Witvrouw A, Puers R, De Wolf I (2005)
the decrease in the pull-in voltage measured experimen- Creep-resistant aluminum alloys for use in MEMS. J Microme-
tally after the creep test. These results depict that the time chanics Microengineering 15(7):S165
dependent variation in the static and dynamic response of Rebeiz GM (2003) RF MEMS: theory, design and technology. Wiley,
Hoboken
the RF-MEMS devices due to the mechanical creep effect Scott S, Katz J, Sadeghi F, Peroulis D (2013) Highly reliable
must be considered at the design, optimization and fabrica- MEMS temperature sensors for 275 °C applications—part 2:
tion level to achieve long term reliability. creep and cycling performance. J Microelectromechanical Syst
22(1):236–243
Somà A, De Pasquale G (2009) MEMS mechanical fatigue: experi-
mental results on gold microbeams. J Microelectromechanical
References Syst 18(4):828–835
Subhash G, Corwin AD, de Boer MP (2011) Evolution of wear char-
Allameh SM (2003) An introduction to mechanical-properties-related acteristics and frictional behavior in MEMS devices. Tribol Lett
issues in MEMS structures. J Mate Sci 38(20):4115–4123 41(1):177–189
Bergers LIJC, Hoefnagels JPM, Delhey NKR, Geers MGD (2011) Tuck K, Jungen A, Geisberger A, Ellis M, Skidmore G (2005) A study
Measuring time-dependent deformations in metallic MEMS. of creep in polysilicon MEMS devices. J Eng Mater Technol
Microelectron Reliab 51(6):1054–1059 127(1):90–96
Brusa E, De Pasquale G, Somà A (2013) Experimental characteriza- van Gils M, Bielen J, McDonald G (2007). Evaluation of creep in
tion of electro–thermo–mechanical coupling in gold RF micros- RF MEMS devices. In: Thermal, mechanical and multi-physics
witches. J Microelectromechanical Syst 22(4):919–929 simulation experiments in microelectronics and micro-systems,
Cho HS, Hemker KJ, Lian K, Goettert J (2002) Tensile, creep and pp 1–6
fatigue properties of LIGA nickel structures. In: Micro electro Vickers-Kirby DJ, Kubena RL, Stratton FP, Joyce RJ, Chang DT, Kim
mechanical systems, pp 439–442 J (2000) Anelastic creep phenomena in thin metal plated cantile-
De Pasquale G, Somà A (2009) Numerical and experimental validation vers for MEMS. In: MRS proceedings, vol 657. pp EE2-5
of out-of-plane resonance closed formulas for MEMS suspended Wetzig K (2003) Metal based thin films for electronics. Wiley-VCH,
plates with square holes. Microsyst Technol 15(3):391–400 Weinheim

13

1078 Microsyst Technol (2016) 22:1067–1078

Yan X, Brown WL, Li Y, Papapolymerou J, Palego C, Hwang J, Vinci Younis MI, Jordy D, Pitarresi JM (2007) Computationally efficient
RP (2009) Anelastic stress relaxation in gold films and its impact approaches to characterize the dynamic response of microstruc-
on restoring forces in MEMS devices. J Microelectromechanical tures under mechanical shock. J Microelectromechanical Syst
Syst 18(3):570–576 16(3):628–638

13

You might also like