See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/345814615

Feasibility study of a micro-electro-mechanical-systems threshold-pressure

sensor based on parametric resonance: experimental and theoretical
investigations

Article in Journal of Micromechanics and Microengineering · December 2020

DOI: 10.1088/1361-6439/abce9c

CITATIONS READS

8 355

3 authors:

Mark Pallay Meysam Daeichin

Binghamton University Binghamton University
14 PUBLICATIONS 215 CITATIONS 11 PUBLICATIONS 223 CITATIONS

SEE PROFILE SEE PROFILE

Shahrzad Towfighian
Binghamton University
97 PUBLICATIONS 1,626 CITATIONS

SEE PROFILE

All content following this page was uploaded by Shahrzad Towfighian on 13 November 2020.

The user has requested enhancement of the downloaded file.

Page 1 of 10 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 1

1
2
3 Feasibility study of a MEMS threshold-pressure
4
5
6
sensor based on parametric resonance: experimental
7
8
and theoretical investigations
9
Mark Pallay, Meysam Daeichin, and Shahrzad Towfighian
10
11
12
13
14 Abstract—A tunable threshold pressure sensor based on para- Parametric excitation is a phenomenon that happens in
metric resonance of a microbeam subjected to electrostatic the response of physical oscillators that have at least one of
15
levitation is proposed. Parametric excitation can trigger a large their properties varying with time [1]–[3]. For example, a
16 amplitude vibration at twice the natural frequency if the mag-
17 nitude of the driving force is large enough to overcome energy simple mass-spring-damper oscillator could exhibit parametric
18 loss mechanisms in the system such as squeeze film damping. excitation if the spring stiffness changes with time. As a
19 This causes a temporarily unstable response with a significant result, the system will experience a dramatic change in its
gain in oscillation amplitude over time until it is eventually vibration amplitude if the magnitude and frequency of the
20 capped by nonlinearities in the force or material or geometric
21 varying stiffness satisfy the conditions needed to trigger
properties. The instability divides the frequency region into two
22 regions: distinct responses bounded by the system nonlinearity, parametric resonance. For any given system, these conditions
23 and trivial responses with very low oscillation amplitudes. It can be found by mapping curves called transition curves in
24 is shown experimentally that the appearance of parametric a 2D space of the magnitude and frequency of the varying
resonance depends on the pressure, which influences the amount parameter. If the magnitude and frequency are inside a
25 of energy loss from squeeze film damping. Therefore, the distinct
26 transition curve, the system is said to be unstable and will
difference in the vibration amplitude can be used to detect when
27 the pressure passes a threshold level. The activation of parametric trigger parametric resonance. These transition curves are
28 resonance also depends on the amplitude of the driving force found at numerous frequency intervals in this space, with the
29 (Vac ). This voltage amplitude can be set to trigger parametric primary transition curve located at twice the natural frequency.
resonance when the pressure drops below a predetermined
30 threshold. A reduced-order model is developed using the Euler-
31 In electrostatic systems, the electrostatic force has a
Bernoulli beam theory to elucidate the nonlinear dynamics of the
32 system. The simulation results from the mathematical model are linear component that is proportional to the displacement
33 in good agreement with the experimental data. The advantages of of the movable electrode. This linear term acts as a second
34 the proposed sensor over pull-in based sensors are its reliability mechanical stiffness that is related to the square of the
and improved resolution from a large signal-to-noise ratio. applied voltage. By applying a voltage between a fixed
35
36 and movable electrode, in addition to the electrostatic
37 I. I NTRODUCTION force that is generated, the total effective stiffness of the
38 Monitoring air (or other gas) pressure in many applications movable electrode is affected. If an AC voltage with a
39 requires periodic checking of the pressure level to see if frequency twice that of the natural frequency is applied,
40 it has dropped below a threshold value. An example is the system can be driven into parametric resonance. This
41 detecting sudden changes in the air pressure inside a plane can happen in any electrostatic system, even one that does
42 cabin to decide if breathing masks should be released for the not use a traditional parallel-plate configuration [4]. The
43 passengers. For these applications, it would be advantageous sudden change in amplitude from parametric resonance
44 to use a smart sensor that only triggers when the pressure has been exploited for sensing and actuation applications
45 drops below the specified level, eliminating the need to [5]–[7]. Mass and gas sensors have been developed based on
46 constantly monitor the pressure and reducing the amount of parametric resonance in microstructures. [8], [9]. In a previous
47 computational overhead required. One method of developing work, The parametric resonance for a circular micromirror
48 this type of smart sensor is to rely on the hysteresis in is used to achieve large displacement at higher frequencies [6].
49 nonlinear systems, which can cause sudden changes in system
50 dynamics when a parameter such as excitation amplitude, Electrostatic levitation is generated by a particular
51 frequency, or damping changes. Nonlinear oscillators that are configuration of electrodes that offers several advantageous
52 susceptible to parametric resonance are one of such systems features over conventional parallel-plate-based sensors [10]–
53 that can trigger a significant jump in oscillation amplitude if [12]. Figure 1 (a) shows the schematic of the electrode
54 the system damping drops below a threshold value. configuration in the levitation scheme. In this approach,
55 the bias voltage is placed on the side electrodes, which are
56 fixed to the substrate. The bottom electrode (also fixed) and
M. Pallay, M. Daeichin, and S. Towfighian are with the Department the moving electrode (beam) above it are grounded. This
57 of Mechanical Engineering, State University of New York at Binghamton,
58 Binghamton, NY, 13902 USA, Corresponding author: S. Towfighian (stow- voltage distribution creates an electrostatic force that pulls
59 [email protected]) the microstructure away from the bottom electrode [11], [12].
60
 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2 Page 2 of 10
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 2

1
2 The nature of the electrostatic levitation force is completely electrical circuitry and requires less processing power, which
3 different from the electrostatic force in parallel-plate is very desirable. The proposed sensor is tunable meaning
4 capacitors, where the moving electrode is pulled toward the that the threshold at which the sensor is triggered can be
5 bottom electrode [13]. The levitation electrode configuration adjusted by changing the amplitude of time-varying (AC)
6 allows for a large vibration of the moving electrode [14] voltage on the side electrode. Because the beam and the
7 and does not suffer from the pull-in instability between the bottom electrode are grounded, even if the cantilever comes
8 moving and bottom electrodes [15]–[17]. Furthermore, it into contact with the bottom electrode, it will not experience
9 provides the flexibility to increase the bias voltage to increase pull-in. If humidity and other environmental effects cause
10 sensitivity in sensing applications [18]. The use of high bias stiction, the sensor can be reset by applying side electrode
11 voltages may be counted as a drawback of the levitation voltage [32], [33], making the sensor robust and reusable.
12 scheme though. However, advances in Application Specific
13 Integrated Circuits (ASIC) designs has made handling The rest of this paper is organized as follows. Section II
14 high bias voltages possible. For some applications such presents the working principle of the proposed sensor. In
15 as instrumentation devices, the use of high bias voltage is section III, a mathematical model is developed to investigate
16 not of concern. In applications where the use of high bias the nonlinear dynamics of the microbeam under levitation
17 voltages is of concern, the dimensions of the electrodes could electrostatic force. The model is turned into a Mathieu’s
18 be changed to make the use of low bias voltages feasible [14]. equation to demonstrate the system will experience parametric
19 resonance. Section IV presents and discusses the experimental
20 All MEMS (Micro-Electro-Mechanical-Systems) pressure results. In Section V, the highlights of the work is summarized
21 sensors output a continuous measurement of pressure. They and some concluding remarks are presented.
22 take advantage of linear features of MEMS such as resonance
23 frequency of resonators [19] or deflection of a diaphragm
24 [20] in the linear regime. This is because a linear relationship
II. P RINCIPLE OF O PERATION
25 between input and output of the sensor makes the electrical
26 circuitry and data processing much simpler [13]. Therefore, The sensor is built from a micro cantilever that is subjected
27 nonlinearity is undesired in pressure sensors. However, for to the levitation electrostatic force. The microbeam is built
28 threshold pressure sensors, exploiting nonlinear features of with the POLYMUMPs fabrication standard (Fig. 1 (b)). The
29 MEMS such as pull-in [21], buckling [22], and snap-through dimensions of microbeam and other electrodes are given in
30 [23] could be very beneficial because they can improve Table I. We have used QP-QFN44-7MM-.5MM package from
31 the signal-to-noise ratio. For example, the pull-in instability Quik-pak Company for the MEMS chip.
32 of parallel-plate capacitors has been studied for threshold
33 sensing applications [24], [25]. The dynamic pull-in of a The sensor is placed in an environment where it will be
34 parallel-plate capacitor depends strongly on the squeeze measuring pressure. A time-varying voltage that has DC
35 film damping, which is a function of pressure [26]. The and AC components is applied to the side electrodes and
36 main drawback of this approach is that pull-in may cause the microbeam is excited at twice its natural frequency.
37 permanent failure of the device with a detrimental effect on its It is important to note that the natural frequency of the
38 reliability [27]. In previous work by our research group [21], microstructure depends on the stiffness, which in this case,
39 [28], parallel-plate excitation is combined with electrostatic consists of the structural stiffness as well as the electrostatic
40 levitation to make a robust pressure switch. Snap-through field induced stiffness (electrical stiffness). The structural
41 bi-stability of a clamped-clamped MEMS arch is another stiffness of the microstructure comes from the elasticity of the
42 nonlinear phenomenon that could be used as a mechanism microbeam, which depends on the material and the geometry
43 to establish a threshold pressure [23], [29]. The caveat to of the microbeam. The electrical stiffness is introduced
44 this approach is that the fabrication of a clamped-clamped mainly from the DC voltage on the side electrodes and
45 arch beam is more challenging compared to a cantilever depends on the electrode dimensions and their configuration,
46 beam. This is because controlling residual stresses during the especially the gap between the side and the bottom electrodes.
47 fabrication process is a cumbersome task [30].
48 In general, the electrical stiffness in the levitation approach
49 In this study, we extend on previous studies [2], [31] to increases with the bias voltage on the side electrodes.
50 exploit another interesting feature of the levitation approach, Depending on the geometry of the electrodes and their
51 which is parametric resonance-based threshold pressure configuration, it is possible for the natural frequency to
52 sensing. A time-varying voltage on the side electrode initially decrease with an increase in the bias voltage up to a
53 modulates the equivalent stiffness of the system, and, under certain voltage and then to increase. The increase of linear
54 the right damping circumstance, the microbeam exhibits natural frequency with the DC voltage is a characteristic of
55 parametric resonance. It is shown experimentally that the the levitation approach that is in contrast to the parallel-plate
56 dependence of parametric resonance on the damping can configuration. A comprehensive study of the effect of bias
57 be used to build a threshold pressure sensor. A parametric voltage on the resonance frequency of the microcantilever is
58 resonance-based sensor can generate a distinct electrical signal given in [11].
59 that is easy to detect and process. This helps avoid complex
60
Page 3 of 10 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 3

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19 Fig. 1. (a) Schematic of the electrodes in the electrostatic levitation configuration. This particular voltage distribution creates an electric field that causes an
20 upward electrostatic force on the moving electrode, pulling it away from the bottom electrode.(b) Top view of a fabricated microbeam. The fabrication is
done using POLYMUMPs standard.
21
22
23 III. M ATHEMATICAL M ODEL
24 1
A. Governing Equation of Motion
25
26 In this section, a mathematical model is developed to
investigate the nonlinear dynamics of the microbeam at the
27 0.5
change in parametric resonance. The governing equation of motion for
28
amplitude the beam in the z direction (ŵ) is given in Eq. (1) using
29
the Euler-Bernoulli beam theory. The microbeam is made of
30 0
polysilicon, and its modulus of elasticity and density are given
31 No parametric
in Table I.
32 resonance
33 -0.5
34 ∂ 2 ŵ ∂ ŵ ∂ 4 ŵ
ρA + ĉ + EI = V 2 fˆe (ŵ) (1)
35 Parametric resonance ∂ t̂2 ∂ t̂ ∂ x̂4
36 -1
In Eq. (1), I is the second moment of inertia of the y-z
0 0.2 0.4 0.6 0.8
37 section of the beam about the y axis, V is the electric voltage
38 on the side electrode, and fˆe is the electrostatic force per
39 Fig. 2. Appearance of parametric resonance in the microstructure’s response unit length for when V = 1V olt. The electrostatic force
40 that leads to a distinct difference in the vibration amplitude of the microbeam.
profile (fˆe ) is obtained from a finite element simulation in
This change in the amplitude can be used to build a threshold pressure sensor.
41 COMSOL for a 2D cross section of the beam and electrodes.
42 A polynomial function is fit to the numerical results for force
43 profile from COMSOL [11].
44
45 Using non-dimensional parameters given in Table II, Eq. (1)
46 can be rewritten as:
47 As it will be shown experimentally, at a given DC voltage,
48 the AC component of the voltage determines the threshold
pressure at which the parametric resonance is triggered. 9
49 ∂2w ∂w ∂4w L4 2 X
As the microbeam is resonating, if the pressure inside the + c + = V An hn wn (2)
50 ∂t2 ∂t ∂x4 EIh
environment drops below a certain level, the parametric n=0
51
52 resonance is triggered. As a result, the amplitude of vibration where An are the coefficients of the polynomial for the force
will increase dramatically, and the sensor goes to the ON profile given in Table I. The value of c is given in Table II
53
state. Figure 2 shows the velocity of the microbeam at its as well. The voltage on the side electrode is represented by
54
tip point before and after the parametric resonance. The V in Eq. (2), which has a DC component and a harmonic
55
increase in the amplitude is distinct and could be detected by time-varying component as (3).
56
57 a piezoelectric transducer or capacitive sensing between the
V = Vdc + Vac cos(ΩT t) (3)
58 beam and side electrodes.
59 where Ω is the excitation frequency.
60
 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2 Page 4 of 10
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 4

1
2 TABLE I for the first mode are given in Table I. The mode shapes are
B EAM MATERIAL AND GEOMETRIC PROPERTIES
3 normalized such that the displacement at the free end is unity.
4 Parameter Symbol Value
Substituting Eq. (4) into Eq. (2), multiplying both sides of
5 Eq. (2) by φ1 , and integrating between 0 and 1 using the
Beam Length (µm) L 504.25
6 Beam Width (µm) b 18
orthogonality of the mode shapes yields a nonlinear ordinary
7 Beam thickness (µm) h 2 differential equation on the time-dependent coefficient q(t) as
8 Beam Electrode Gap (µm) g 2
Electrode Gap (µm) s 20.5
9 Bottom electrode Width (µm) b1 31 9
10
X
Side electrode Width (µm) b2 28 q̈(t)+cq̇(t)+β14 q(t) = r1 (Vdc +Vac cos(ΩT t)) 2
pn (q(t))n
11 Electrode Thickness (µm) h1 0.5
n=0
Elastic Modulus (GPa) E 157
12 kg
Density ( m ρ 2330
(6)
3)
13 Moment of inertial (m4 ) I bh3
where
14 Poisson’s Ratio ν
12
0.22
15 Force Constant A0 1.1629 ×10−7
16 Force Constant A1 5.6299 ×10−3
1 1
-1.6118 ×103 L4
Z Z
Force Constant A2
17 Force Constant A3 1.2729 ×108 r1 = I1 = φ21 (x)dx pn = An h n
φn+1
1 dx
18 Force Constant A4 -3.0638 ×1012
EIhI1 0 0
(7)
19 Force Constant A5 -3.4096 ×1017
Force Constant A6 3.6121 ×1022 The shooting method is used to solve Eq. (6) and obtain the
20
Force Constant A7 -1.5228 ×1027 frequency response. This method shoots for periodic solutions
21 Force Constant A8 3.1435 ×1031 of the equation, and it is more computationally efficient
22 Force Constant A9 -2.6096 ×1035
compared to integrating Eq. (6) over time. The details of this
23
method for the electrostatically levitating system are outlined
24
in [11]. By performing multiple frequency sweeps at different
25 TABLE II
N ON -D IMENSIONALIZATION Vac and damping conditions, the transition curves will be
26
extracted. The damping coefficient, c in Eq. (6), will be
27 Parameter Substitution identified from the experimental results presented in Section
28
x̂ IV.
29 x-direction position x= L
z-direction position w = ŵ
30 h
Time t = Tt̂
31 ĉL4
B. Analytical Expression for Transition Curves
Damping c=
32 EIT
β12 In addition to the numerical results from the shooting-
Q =q
33 Quality factor c method solution of Eq. (6), an analytical expression of the
ρAL4
34 Time Constant T = EI transition curves is also extracted. The transition curves
35 β1 1.875
separate the stable regions from unstable regions where
σ1 0.7341
36 D1 0.5 the parametric resonance occurs. As it will be shown, the
37 Q(P = 590mT orr) 300 transition curves depend on the amplitude of the time-varying
Q(P = 990mT orr) 136
38 Q(P = 2000mT orr) 67
voltage (Vac ), the excitation frequency (Ω), and the damping
39 coefficient (c)(or the quality factor (Q)). Although numerical
40 methods can be used to solve Eq. (6) to obtain the transition
41 curves, they are computationally cumbersome and time
42 A reduced-order model can be developed using separation consuming. To obtain the transition curves, the frequency
43 of variables as response of Eq. (6) should be calculated at different electrical
44 excitations (Vdc and Vac ) and different damping conditions,
n
45 X which is a cumbersome task. On the other hand, an analytical
w(x, t) = qi (t)φi (x) (4)
46 formula for the transition curves not only yields the results
i=1
47 with much less computation, but it provides insight on how
48 where qi (t) are the time-dependent coefficients and φi (x) are the dynamics of the system is affected by the intertwined
49 the linear mode shapes of the cantilever given as follows relationship between various parameters, and therefore it is
50 very useful.
51 φi (x) = Di (cosh(βi x) − cos(βi x)−
52 σi (sinh(βi x) − sin(βi x))) (5) To obtain the transition curves, one can rewrite Eq. (6) into
53 the form of Mathieu’s equation given below.
54 where βi are the square roots of the non-dimensional mechan-
55 ical natural frequencies, and σi are constants determined from q̈d (t) + C q̇d (t) + [δ +  cos τ ]qd (t) = 0 (8)
56 the clamped-free boundary conditions for the microcantilever.
The first mode approximation is used for the analysis as it Introducing a change of variable given as
57
58 has shown good agreement with experimental results for this
59 system in a previous study [11]. Values of β1 , σ1 , and D1 q(t) = qst + qd (t), (9)
60
Page 5 of 10 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 5

1
2 -8
10
3 18 4.8
4
5 16 4.6
6
4.4
7 14
8 4.2
9 12
10 4
11 10
12 3.8
13 8
3.6
14
15 6
3.4
16
4
17 3.2
18
2
19 3
20
0 2.8
21 0 50 100 150 200 -6 -3 0 3 6
22
23
24 Fig. 3. (a) The solution of Eq. (11) gives the static displacement at the tip of the beam for different DC voltages on the side electrodes (b) The linearization
25 of the electrostatic force about qd = 0 given in Eq. (12). According to this figure, the linearization of the electrostatic force is a reasonable assumption.
26
27
one can rewrite Eq. (6) as
28
29 β12
30 q̈d (t) +
Q
q̇d (t)+
31 q̈d (t) + cq̇d (t) + β14 qd (t) =
[(β14 − δ1 ) − 11 cos(ΩT t) − 21 cos(2ΩT t)]qd (t) =
32 9
X
33 r1 (Vdc + Vac cos(ΩT t))2 pn (qst + qd (t))n − β14 qst (10) [(δ0 − β14 qst ) + 10 cos(ΩT t) + 20 cos(2ΩT t)]
34 n=0 + [δ2 + 12 cos(ΩT t) + 22 cos(2ΩT t)]O(qd2 ) (14)
35
where qst is the solution of the algebraic static equation, given
36 where Q is the quality factor given in Table II. All the other
in Eq. (11). Figure 3 (a) shows the static displacement of
37 coefficients in Eq. (14) are given in the Appendix. In Eq. (14),
the microcantilever at the tip point as the voltage on the side the higher-order terms are ignored, (O(qd2 )). That serves as a
38
electrodes increases.
39 good approximation because the electrostatic force (f (qd )) is
40 9
mostly linear about the static solutions corresponding to high
41 β14 qst = r1 (Vdc )2
X
pn (qst )n (11)
DC voltages as shown in Fig. 3 (b). This figure illustrates that
42 n=0
at the qst corresponding to 165(V ) on the side electrodes,
43 the electrostatic force can be approximated with a line. All
44 The summation in Eq. (10), which is a function of qd , can of the experiments are performed at this voltage. To obtain
45 be expanded about qd = 0 using Taylor series expansion as the transition curves, the homogeneous version of Eq. (14)
46 is considered. The last assumption to make is to ignore the
47 21 cos(2ΩT t) term in Eq. (14) compared to the 11 cos(ΩT t)
48 9
X term. This is a reasonable approximation when Vac is much
49 f (qd ) = pn (qst + qd (t))n = α0 + α1 qd + O(qd2 ) (12) smaller than Vdc as shown in the equation below.
50 n=0

51 21 Vac
where = (15)
52 11 4Vdc
53 Using another change of variable given in Eq. (16), and
54 9 ˙(0) X 9
applying the assumptions explained above, Eq. (14) can be
X f
55 α0 = f (0) = pn (qst )n α1 = = npn (qst )n−1 rewritten in the form of Mathieu’s equation as given in Eq.
1!
56 n=0 n=0
57 (13) (17).
58 Substituting Eq. (12) into Eq. (10) and rearranging the terms,
59 one can write ΩT t = τ (16)
60
 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2 Page 6 of 10
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 6

1
2
3 q̈d (t) + CM q̇d (t) + [δM + M cos τ ]qd (t) = 0 (17)
4
5 where CM , δM , and M are given in the Appendix.
6 For the Mathieu’s equation given in Eq. (17), the transition
7 curves are obtained by using perturbation techniques, and they
8 are given by the following expression [34].
9 p
10 1 2M − c2M
δM = ± + O(2M ) (18)
11 4 2
12 Substituting all the coefficients in the equation for the
13 transition curves results in:
14
15 2
Vac
16 β14 − r1 α1 (Vdc
2
+ 2 )
=
17 (ΩT ) 2

18
q
β2
1 2r1 α1 Vdc Vac ( 2r1 α 1 Vdc Vac 2
) − ( QΩT
1
)2
19 ∓( )
(ΩT )2
+ O(2M )
20 4 (ΩT )2 2
21 (19)
22 which is an implicit algebraic equation relating the Vac , quality
23 factor (Q), and the excitation frequency (Ω). In the next
24 section, the prediction of this equation for the transition curves Fig. 4. The experimental setup. The microbeam is excited by applying voltage
to the side electrodes. The velocity of the microbeam is measured with a
25 is shown and compared with the experimental results. laser vibrometer. The pressure inside the chamber is controlled via a pressure
26 controller set. By reducing the pressure inside the chamber, the microbeam
27 exhibits large amplitude vibrations because of parametric resonance.
IV. E XPERIMENTAL R ESULTS AND D ISCUSSION
28
29 The experimental set up is shown in Fig. 4. The sensor
30 is placed inside a chamber with controlled pressure. All the 7 shows the transition curves for three different pressures.
31 experiments are conducted in the room temperature (25◦ C). The minimum AC voltage required to trigger the parametric
32 The beam is excited by applying a voltage with DC and resonance increases with the pressure :::::::
(Fig. 8).That indicates
33 AC components to the side electrodes. The velocity of the the energy dissipation that system needs to overcome before
34 beam at its free end is measured with a Polytec MSA 500 the parametric resonance can occur.
35 laser vibrometer. To demonstrate the parametric resonance of
36 the microbeam, a forward and backward frequency sweep is As depicted in Fig. 7, the transition curves obtained
37 conducted at a constant pressure of P = 590 mT orr when from the simulations are in good agreement with the
38 the DC voltage on the side electrodes is 165 V . Figure 5 experimental results. The damping coefficient in Eq. (6) is
39 (a) presents the results for when the AC voltage is 10 V . identified from matching the minimum points of the transition
40 As shown in this figure, there is a frequency band (∆f ) curves obtained from the experiment and shooting method.
41 where the beam undergoes parametric excitation. The branch Furthermore, this figure shows that the transition curves
42 associated with parametric excitation expands well beyond that are given by the analytical perturbation method (Eq.
43 this frequency band. However, reaching the points outside (19)) are in excellent agreement with the results from the
44 of this frequency band depends on the initial conditions of shooting methods. For these transition curves, we have used
45 the beam. For reliable and consistent operation of the sensor, the same quality factor that was identified by comparing
46 we only focus on this frequency band, where, independent of the results of the shooting method to the experiment. ::: The
47 initial conditions, the beam is forced to undergo parametric numerical::::::
values:::
of :::
the ::::::
quality:::::::
factors ::
at:::::
each :::::::
pressure:::
are
::::::::
48 excitation.
given:::
in :::
the :::::
Table ::
II.Substituting the values of quality factors
49 ::::

50 The frequency band in Fig. 5 (a) depends on the amplitude into Eq. (19) yields an implicit algebraic equation between
51 of the AC voltage on the side electrode. As shown in Fig. Vac and Ω, which is plotted with MATLAB. Solving these
52 5 (b) and Fig. 6, this frequency band expands with an algebraic equations are computationally more efficient than the
53 increase in the AC voltage. The curve in Fig. 5 (b), which numerical method of shooting that solves the nonlinear
54 is called a transition curve (or instability tongue), separates differential equation. The transition curves from the model
55 the stable and unstable regions depending on the emergence are all symmetric, where as the experimental ones are slightly
56 of parametric resonance. This figure illustrates that there is a bent to the right. This could be attributed to the stiffness and
57 minimum value for the AC voltage below which the beam will damping of the air which is not considered in the model.
58 not experience parametric resonance. This minimum point is Because of the large vibration of the beam at the parametric
59 strongly dependent on the pressure inside the chamber. Figure resonance, the air stiffness because of the squeeze film effect
60 could be considerable and needs further investigations.
Page 7 of 10 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 7

1
2
3 (a) (b)
1 20
4 Forward Sweep
5 0.9
Backward Sweep
18
6
D
7 0.8 16
8
9 0.7 14

10 0.6 12
11 C
f
12 0.5 10
13
14 0.4 8

15
0.3 6
16
17 0.2 4
18
19 0.1 A 2
f B
20
0 0
21 22.5 23 23.5 24 24.5 25 23.5 23.6 23.7 23.8 23.9 24 24.1 24.2 24.3
22
23
24 Fig. 5. (a) Forward and backward frequency sweeps at Vdc = 165(V ) − Vac = 10(V ) − P = 590(mT orr). The jump in the amplitude on the left side
25 of the instability region is much more pronounced compared to the jump on the right side. (b) The transition curve at Vdc = 165(V ) − P = 590(mT orr).
26 Inside the transition curve lies the frequency region for which the beam exhibits parametric resonance. Below a certain Vac there will be no parametric
resonance.
27
28
29 stiffness and damping effect, which causes similar deviations
30 12 on the left side of the transition curve in Fig. 7. The best
31 V ac = 4 V
performance of the sensor is obtained when it is operated
32 10
V
ac
=8V
at an AC voltage and a frequency (Vac , f ) corresponding to
V = 16 V
33 ac points on the left side of the transition curves. This is because
34 8 the jump in the amplitude on the left side of the instability
35 region (point A in Fig. 5 (a)) is much larger than the jump
36 6 on the right side (point B in Fig. 5 (a)). Operating the sensor
37 at this (Vac , f ) ensures a large change in the amplitude when
4
38 the parametric resonance is triggered. Figure 10 illustrates
39 how the transition curve grows by decrease in pressure. The
2
40 operating point which is initially outside of the transition
41 0
curve, will be encompassed by the growing transition curve
42 23.6 23.7 23.8 23.9 24 24.1 24.2 24.3 as the pressure drops. That means, if the pressure drops in
43 the environment by a very small amount, the oscillations will
44 significantly increase. This significant rise in the amplitude
45 Fig. 6. Forward frequency sweeps at Vdc = 165(V ) − P = 900(mT orr) can be observed using a charge amplifying circuit, which can
and different Vac . As shown in Fig. 5, the frequency range that is used to
46 obtain the transition curve, can be extracted by performing only the forward detect the capacitance change. This study has characterized
47 frequency sweep. This frequency range enlarges by increasing Vac . the pressure sensor in pressure ranges below 2 Torr. Studying
48 the performance of the pressure sensor at pressures above 2
49 Torr can be the subject of future investigations.
50
51
52 To present the operating voltages of the pressure sensor, In theory, the sensor could be used to detect either drop
53 the minimum AC voltages required to trigger parametric or rise in pressure. However, we have studied the device
54 resonance at 23.8kHz for different pressures are shown in only to detect a drop-in pressure below a certain threshold.
55 Figure 9. The figure compares simulations with experimental The application of the sensor for detecting a rise in pressure
56 results. Based on this figure, the AC voltage can be tuned to above a threshold value needs further investigation and could
57 establish different threshold pressure values. The operating be studied in a future work. This is because for the sensor
58 DC voltage is fixed at 165 V. The difference between the to detect a pressure rise, we need to change the operating
59 experimental and simulation results is related to the air point of the sensor. In Fig. 11, we have demonstrated the
60
 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2 Page 8 of 10
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 8

1
2
3 20 20 20
4
5 18 18 18

6
16 16 16
7
8 14 14 14
9
10 12 12 12
11
VAC (V)

12 10 10 10

13
8 8 8
14
15 6 6 6
16
17 4 4 4
18 Experiment
Experiment
19 2 Experiment 2
Numerical (Q=136)
2 Numerical (Q=67)
Numerical (Q=300) Analytical
20 Analytical
Analytical
0 0 0
21 23.6 23.8 24 24.2 23.6 23.8 24 24.2 23.6 23.8 24 24.2
22 Frequency (kHz)
23
24 Fig. 7. Transition curves for three different pressure values. By increasing the pressure the transition curve shrinks, and therefore the minimum voltage for
25 the appearance of the parametric resonance increases. The center-line indicates the experimental results, while the dashed line shows the analytical solution
26 and the solid line depicts the numerical simulation data obtained from the shooting technique.
27
28
29 9.95 20
30 Experimental data Model
10-3 Experiment
31 Linear fit (Vac = 5.1 P - 0.2395)
7.5
32 15
33
Vac (V)

34 4.8
35 10

36
37 2.35
38 5
0.8
39
40 150 590 990 1500 2000 0 500 1000 1500 2000 2500 3000
41 Pressure (mTorr)
42
43 Fig. 8. The minimum AC voltage required to trigger parametric resonance Fig. 9. The Minimum Vac required for the appearance of parametric resonance
increases with pressure. at different pressures on the left hand side of the transition curve at 23.8kHz
44 (Vdc = 165(V ). Red (Solid-dot) curve: Experimental Measurements. Blue
45 (dash-dot) curve: theoretical predictions. This curve can be used for the op-
46 eration of the sensor to establish different pressure thresholds when detection
47 difference between detecting a pressure drop versus detecting of drop in the pressure is desired. At each pressure, the required Vac for the
a pressure rise by schematically showing the forward and operation of the sensor can be identified from this curve.
48
49 backward frequency sweeps at different pressures. If the
50 sensor is operating at point A shown in the Fig. 11, a drop a pressure rise, the amplitude will drop suddenly to point
51 in pressure will trigger the parametric resonance and the E. This distinct drop in amplitude because of pressure rise
52 amplitude jumps up to point B. But if the pressure increases, can easily be realized in a charge amplifying circuit for
53 the amplitude will not fall back to point A. It rather gradually capacitance detection.
54 sinks down to point C and there would be no sudden jump
55 down similar to the jump up. A better operating point for
56 detecting the pressure rise is point D that exists at the end of
the high branch oscillation for the backward frequency sweep. V. C ONCLUSION
57
58 When the sensor operates at point D at low pressure, the A threshold pressure sensor is proposed using paramet-
59 sensor will have a large amplitude. If the sensor experiences ric resonance in a microbeam subjected to the levitation
60
Page 9 of 10 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 9

1
2 method, and the results are compared to the experimental data.
20
3 An analytical expression for the transition curves of the system
4 Parameteric Resonance When is obtained by rewriting the governing equation of motion in
5 the form of Mathieu’s equation. The comparison shows that
15
6 the developed model is capable of capturing the nonlinear
7 dynamics of the system that can serve in designing high
8 signal-to-noise ratio sensors with improved pressure sensing
10
9 Parameteric Resonance When
resolution.
10
11
5 ACKNOWLEDGMENT
12
13 The authors would like to acknowledge the financial support
14 of this study by the National Science Foundation (NSF)
0
15 23.7 23.8 23.9 24 24.1 24.2 24.3 through grant #ECCS1608692.
16
17
Fig. 10. By operating the sensor at a frequency and Vac outside of the VI. A PPENDIX
18 transition curves for a particular pressure (the operation point), the sensor
19 does not experience parametric resonance. With a decrease in pressure, the
20 transition curve grows and encompasses the operation point. Therefore, the
2
2
Vac
sensor will exhibit parametric resonance. δ0 = r1 α0 (Vdc + ) (20)
21 2
22
23
24 10 = 2r1 α0 Vdc Vac (21)
25
26 2
r1 α0 Vac
27 20 = (22)
2
28
29 2
30 2 Vac
δ1 = r1 α1 (Vdc + ) (23)
31 2
32
33 11 = 2r1 α1 Vdc Vac (24)
34
35
36 2
r1 α1 Vac
21 = (25)
37 2
38
39 2
Vac
2
40 δ2 = r1 (Vdc +
2
) (26)
Fig. 11. The forward and backward frequency sweeps are shown at two
41 pressure levels schematically. This figure shows that for detecting a rise in
42 pressure, the sensor should be operated at point D, where an increase in
43 pressure will lead to a dramatic drop in the amplitude of vibration. 12 = 2r1 Vdc Vac (27)
44
45 2
electrostatic force. The experimental data shows the distinct r1 Vac
46 22 = (28)
47 difference in the vibration amplitude of the beam when it 2
48 is parametrically excited compared to when it is not. The
appearance of the parametric resonance depends on the pres- β12
49 cM = (29)
sure and AC voltages. Therefore, the AC voltage can be used QΩT
50
51 as a tuning parameter to establish different values for the
52 threshold pressure. The sensor is reusable, and in theory, it β14 − δ1
δM = (30)
53 can be used to detect when the pressure goes above or drops (ΩT )2
54 below a certain threshold. The operation of the sensor for
55 detecting a drop in pressure is studied experimentally and 11
theoretically. The application of the sensor for detecting rise M = − (31)
56 (ΩT )2
57 in the pressure needs further investigations. A mathematical
58 model is developed using the Euler-Bernoulli beam theory. 2
Vac
2
The governing equation of motion is solved using the shooting δ0 = r1 α1 (Vdc + ) (32)
59 2
60
 AUTHOR SUBMITTED MANUSCRIPT - JMM-104920.R2 Page 10 of 10
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MAY 2020 10

1
2 R EFERENCES [23] M. H. Hasan, F. M. Alsaleem, and H. M. Ouakad, “Novel threshold
pressure sensors based on nonlinear dynamics of mems resonators,”
3 Journal of Micromechanics and Microengineering, vol. 28, no. 6, 2018.
[1] K. Moran, C. Burgner, S. Shaw, and K. Turner, “A review of parametric
4 resonance in microelectromechanical systems,” Nonlinear Theory and
[24] R. K. Gupta and S. D. Senturia, “Pull-in time dynamics as a measure of
5 Its Applications, IEICE, vol. 4, no. 3, pp. 198–224, 2013.
absolute pressure,” in Proceedings IEEE The Tenth Annual International
Workshop on Micro Electro Mechanical Systems. An Investigation of
6 [2] M. Pallay and S. Towfighian, “A parametric electrostatic resonator using
Micro Structures, Sensors, Actuators, Machines and Robots, 1997, pp.
7 repulsive force,” Sensors and Actuators A: Physical, vol. 277, pp. 134
290–294.
– 141, 2018.
8 [3] C. Guo, E. Tatar, and G. K. Fedder, “Large-displacement parametric
[25] E. M. Miandoab, H. N. Pishkenari, A. Meghdari, and M. Fathi, “A
general closed-form solution for the static pull-in voltages of electro-
9 resonance using a shaped comb drive,” in 2013 IEEE 26th International statically actuated mems/nems,” Physica E: Low-dimensional Systems
10 Conference on Micro Electro Mechanical Systems (MEMS), 2013, pp. and Nanostructures, vol. 90, pp. 7 – 12, 2017.
173–176.
11 [4] Y. Linzon, B. Ilic, S. Lulinsky, and S. Krylov, “Efficient parametric
[26] F. Najar, M. Ghommem, and A. Abdelkefi, “A double-side electrically-
actuated arch microbeam for pressure sensing applications,” Interna-
12 excitation of silicon-on-insulator microcantilever beams by fringing tional Journal of Mechanical Sciences, vol. 178, p. 105624, 2020.
13 electrostatic fields,” Journal of Applied Physics, vol. 113, no. 16, p. [27] W. Zhang, H. Yan, Z.-K. Peng, and G. Meng, “Electrostatic pull-in
163508, 2013.
14 [5] Wenhua Zhang, R. Baskaran, and K. L. Turner, “Changing the behavior
instability in mems/nems: A review,” Sensors and Actuators A: Physical,
vol. 214, pp. 187 – 218, 2014.
15 of parametric resonance in mems oscillators by tuning the effective cubic [28] M. Pallay and S. Towfighian, “A combined mems threshold pressure
16 stiffness,” in The Sixteenth Annual International Conference on Micro sensor and switch,” in 2019 IEEE SENSORS, 2019, pp. 1–4.
Electro Mechanical Systems, 2003. MEMS-03 Kyoto. IEEE, 2003, pp.
17 173–176.
[29] A Novel Threshold Pressure Sensor Based on Nonlinear Dynamics of
MEMS Arches, ser. International Design Engineering Technical Confer-
18 [6] M. Ozdogan, M. Daeichin, A. Ramini, and S. Towfighian, “Parametric ences and Computers and Information in Engineering Conference, vol.
19 resonance of a repulsive force mems electrostatic mirror,” Sensors and Volume 9: 13th ASME/IEEE International Conference on Mechatronic
Actuators A: Physical, vol. 265, 2017.
20 [7] D. I. Caruntu and E. Juarez, “Voltage effect on amplitude–frequency
and Embedded Systems and Applications, 08 2017.
[30] M. A. A. Hafiz, L. Kosuru, A. Ramini, K. N. Chappanda, , and M. I.
21 response of parametric resonance of electrostatically actuated double- Younis, “In-plane mems shallow arch beam for mechanical memory,”
22 walled carbon nanotube resonators,” Nonlinear Dynamics, vol. 98, 2019. Micromachines, vol. 7, no. 10, 2016.
23 [8] Wenhua Zhang and K. L. Turner, “A mass sensor based on parametric [31] M. N. Hasan, M. Pallay, and S. Towfighian, “Threshold pressure
resonance,” A Solid State Sensor, Actuator and Microsystems Workshop, sensing using parametric resonance in electrostatic mems,” in 2019 IEEE
24 Hilton Head Island, South Carolina, June 6–10, 2004. SENSORS, 2019, pp. 1–4.
25 [9] Wenhua Zhang and K. L. Turner, “Application of parametric resonance [32] M. Pallay, A. I. Ibrahim, R. N. Miles, and S. Towfighian, “Pairing elec-
26 amplification in a single-crystal silicon micro-oscillator based mass trostatic levitation with triboelectric transduction for high-performance
sensor,” Sensors and Actuators A: Physical, vol. 122, no. 1, pp. 23 self-powered mems sensors and actuators,” Applied Physics Letters, vol.
27 – 30, 2005. 115, no. 13, p. 133503, 2019.
28 [10] Ki Bang Lee and Young-Ho Cho, “Laterally driven electrostatic [33] M. Pallay and S. Towfighian, “A reliable mems switch using electrostatic
29 repulsive-force microactuators using asymmetric field distribution,” levitation,” Applied Physics Letter, vol. 113, no. 21, p. 213102, 2018.
Journal of Microelectromechanical Systems, vol. 10, no. 1, pp. 128– [34] R. H. Rand, “Lecture notes on nonlinear vibrations,” Cornell University,
30 136, March 2001. vol. 53, 2012.
31 [11] M. Pallay, M. Daeichin, and S. Towfighian, “Dynamic behavior of
32 an electrostatic mems resonator with repulsive actuation,” Nonlinear
Dynamics, vol. 89, no. 2, p. 1525–1538, 2017.
33 [12] M. Daeichin, R. N. Miles, and S. Towfighian, “Experimental characteri-
34 zation of the electrostatic levitation force in mems transducers,” Journal
35 of Vibration and Acoustics, vol. 142, no. 4, 2020.
36 [13] M. Pallay, R. N. Miles, and S. Towfighian, “Merging parallel-plate and
levitation actuators to enable linearity and tunability in electrostatic
37 mems,” Journal of Applied Physics, vol. 126, no. 1, p. 014501, 2019.
38 [14] S. Towfighian, S. He, and R. Ben Mrad, “A low voltage electrostatic
39 micro actuator for large out-of-plane displacement,” 8th International
Conference on Micro- and Nanosystems, 08 2014.
40 [15] M. Daeichin, M. Ozdogan, S. Towfighian, and R. N. Miles, “Dynamic
41 response of a tunable mems accelerometer based on repulsive force,”
42 Sensors & Actuators: A. Physical, vol. 289, pp. 34–43, 2019.
[16] M. Ozdogan, S. Towfighian, and R. N. Miles, “Fabrication and ex-
43 perimental characterization of a mems microphone using electrostatic
44 levitation,” in 2019 IEEE SENSORS, 2019, pp. 1–4.
45 [17] M. Daeichin, S. Towfighian, and R. N. Miles, “Lateral pull-in instability
of mems transducers employing repulsive force,” Nonlinear Dynamics,
46 pp. 1927–1940, 2020.
47 [18] M. Ozdogan, S. Towfighian, and R. N. Miles, “Modeling and charac-
48 terization of a pull-in free mems microphone,” IEEE Sensors Journal,
49 vol. 20, no. 12, pp. 6314–6323, 2020.
[19] K. Bruckner, V. Cimalla, F. Niebelschutz, R. Stephan, K. Tonisch,
50 O. Ambacher, and M. A. Hein, “Gas pressure sensing based on mems
51 resonators,” in SENSORS, 2007 IEEE, 2007, pp. 1251–1254.
52 [20] Z. Zhang, Yan-Hong Zhang, L. Liu, and Tian-Ling Ren, “A novel mems
pressure sensor with mosfet on chip,” in SENSORS, 2008 IEEE, 2008,
53 pp. 1564–1567.
54 [21] M. Pallay, R. N. Miles, and S. Towfighian, “A tunable electrostatic mems
55 pressure switch,” IEEE Transactions on Industrial Electronics, pp. 1–1,
2019.
56 [22] Y. D. Yukang Yang and Y. Li, “Buckling of piezoelectric sandwich
57 microplates with arbitrary in-plane bcs rested on foundation: effect of
58 hygro-thermo-electro-elastic field,” The European Physical Journal Plus,
vol. 135, no. 1, 2020.
59
60

View publication stats