On the Equivalence of Acoustic

Impedance and Squeeze Film

Impedance in Micromechanical
Charanjeet Kaur Malhi1
Centre for Ultrasonic Engineering,
Resonators
Department of Electronic
In this work, we address the issue of modeling squeeze film damping in nontrivial geome-
and Electrical Engineering,
tries that are not amenable to analytical solutions. The design and analysis of microelectro-
University of Strathclyde,
mechanical systems (MEMS) resonators, especially those that use platelike two-dimensional
Glasgow G1 1XW, UK
structures, require structural dynamic response over the entire range of frequencies of inter-
e-mail: [email protected]
est. This response calculation typically involves the analysis of squeeze film effects and
acoustic radiation losses. The acoustic analysis of vibrating plates is a very well understood
Rudra Pratap problem that is routinely carried out using the equivalent electrical circuits that employ
Professor lumped parameters (LP) for acoustic impedance. Here, we present a method to use the same
Department of Mechanical Engineering circuit with the same elements to account for the squeeze film effects as well by establishing
and Centre for Nano Science and Engineering, an equivalence between the parameters of the two domains through a rescaled equivalent
Indian Institute of Science, relationship between the acoustic impedance and the squeeze film impedance. Our analysis
Bangalore 560012, India is based on a simple observation that the squeeze film impedance rescaled by a factor of jx,
e-mail: [email protected] where x is the frequency of oscillation, qualitatively mimics the acoustic impedance over a
large frequency range. We present a method to curvefit the numerically simulated stiffness
and damping coefficients which are obtained using finite element analysis (FEA) analysis. A
significant advantage of the proposed method is that it is applicable to any trivial/nontrivial
geometry. It requires very limited finite element method (FEM) runs within the frequency
range of interest, hence reducing the computational cost, yet modeling the behavior in the
entire range accurately. We demonstrate the method using one trivial and one nontrivial
geometry. [DOI: 10.1115/1.4031438]

1 Introduction conditions can be found in the works of Darling et al. [5], Pandey
et al. [6], and Mohite et al. [7]. A comprehensive summary of the
A typical MEMS resonator comprises a top plate vibrating
progress in this field can be found in works of Bao and Yang [1]
transversely against a fixed substrate separated by an air gap (refer
and Pratap et al. [8]. If the geometry is complicated, the solution
to Fig. 1). Topologically, both the resonator as well as the bottom
can be obtained using FEM [9]. In the case of complex geome-
substrate can be with or without opening/cuts depending on the
tries, a numerical solution of Reynolds’ equation or Navier
type of application. The resonator has dominantly two modes of
Stokes’ equation can be carried out using commercial packages
dissipation, viz., the squeeze film damping due to the viscous
like ANSYS, COMSOL, etc. Numerical solutions lead to the determina-
shearing of the trapped air in between the vibrating member and
tion of pressure distribution on the face of the vibrating surface,
the fixed substrate, and the acoustic damping which is the energy
which is subsequently integrated to determine the damping and
carried away due to sound propagation from the top of the plate.
the stiffness components of the force. The so-obtained numerical
Both these effects play a significant role in modulating the
solution leads to discrete values of these forces at discrete fre-
dynamic response of an MEMS resonator. The degree of their
quencies. In order to convert these values into suitable expressions
individual influence depends on the geometry of the device as
(which may or may not be frequency dependent), it is required to
well as the frequency of vibration of the resonator.
do some form of curve fitting. The so-obtained expressions are
The movement of the air (generated due to the resonator push-
ing and pulling the trapped air) contributes to damping the motion
of the moving member as well as increasing the stiffness of the
system. The evaluation of this fluid damping and the spring char-
acteristics is critical in the design and analysis of dynamic MEMS
devices. The motion of a fluid trapped in narrow gaps can be ana-
lyzed using Navier Stokes’ or Reynolds’ equation [1]. Relatively
simple geometries like rectangular/circular plates, with or without
holes, are amenable to analytical solutions of Reynolds’ equation.
Significant contribution to this field in terms of reduction of com-
plex Navier Stokes’ equation to Reynolds’ equation and its solu-
tion can be found in Refs. [2–4] among many others. Analytical
solution of linear Reynolds’ equation for various boundary

1
Corresponding author.
Contributed by the Noise Control and Acoustics Division of ASME for Fig. 1 MEMS resonator with various zones of dissipation—
publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received April squeeze film action in between the resonator and fixed sub-
4, 2015; final manuscript received August 14, 2015; published online October 15, strate and energy dissipation in the infinite fluid on top of the
2015. Assoc. Editor: Sheryl M. Grace. resonator due to the acoustic waves

Journal of Vibration and Acoustics Copyright V

C 2016 by ASME FEBRUARY 2016, Vol. 138 / 011005-1

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 useful in modeling the squeeze film curves for the particular ge- and the characteristic dimension of the radiating surface (a)). In
ometry. These expressions can also be used in simulating the sqd, the behavior is dependent on the squeeze number (r), which
dynamic response of the device. is a function of the frequency of oscillation and the characteristic
The other means of dissipation in the structure shown in Fig. 1 flow length. It is evident from Figs. 2(a) and 2(c) that the curves
is through the radiation of the sound energy into the infinite half- representing the acoustic and the modified squeeze film imped-
space on the top side and is known as the radiation resistance. ance look similar. The original squeeze film impedance needs to
Additionally, there is a mass loading effect due to the surrounding be scaled by a factor so that it mimics the acoustic impedance
fluid onto the plate that is known as the radiation reactance curve.
[10,11]. Both of these quantities are collectively known as radia- It can also be noted from Fig. 2 that in both domains, the hori-
tion impedance which can be approximated with an analogous zontal axes constitute of variables dependent on frequency. But
electrical circuit as discussed by Beranek [12]. The entire effect the vertical axes in both domains are different. In the acoustics
can be represented using four elements comprising of two resis- domain, the vertical axis represents impedance but in the rescaled
tors, one inductor, and a capacitor [12]. Doody et al. [13] used this sqd, it is the impedance multiplied by ðjxÞ. Since the vertical axes
circuit for modeling the radiation impedance of a circular plate are different in both domains, the hypothetical assumption we
vibrating in the first mode. Their target application was a capaci- make is that Za ¼ jxZsqd . This assumption allows us to model the
tive micromachined ultrasonic transducer (CMUT) wherein they sqd using a circuit valid in the acoustics domain. The key to our
determined the expressions for the above four parameters using analysis is that since the squeeze film impedance, by itself, does
curve fitting on the numerically obtained radiation impedance not mimic the variation shown by the acoustic impedance, we
data. Krause [14] used the same representation for modeling the rescale it suitably so that the two responses are reasonably close.
radiation impedance of a circular plate with holes uniformly dis- In order to model the squeeze film behavior using the acoustic
tributed over its area. Here, the geometry was modeled in ANSYS circuit elements, a suitable mapping of the constitutive elements
and the target application was a microphone array meant for tur- is needed. The following mathematical description explains the
bulent boundary layer measurements. needed relation between the elements of these two domains:
In the present work, we demonstrate the use of this circuit,  
meant for modeling radiation impedance in the acoustic domain, Ka
Za ¼ Ra þ j Ma x  (1)
for modeling the squeeze film impedance. We adapt the methodol- x
ogy used in the acoustics domain analysis and use it in the squeeze
film domain (sqd). We have demonstrated its use in acoustic On dividing the Za by the factor jx, we obtain
domain along with a detailed explanation of the curve-fitting pro-
cedure in Malhi and Pratap [15]. The present analysis is motivated Za Ra Ka
¼ þ Ma  2 (2)
by the observation that the curve of the radiation impedance is jx jx x
similar to the curve for the squeeze film impedance multiplied by
angular frequency in the frequency range of interest. The basic Here, Za is the notation for impedance which is a ratio of the force
idea is to use the equivalent circuit used for the acoustics domain and the velocity in the acoustic domain, Ra is the resistance which
into the sqd. would be represented by a resistor in its electrical analogous form,
A significant advantage of this method is that it facilitates rapid Ma is the mass represented by an inductor, Ka is the stiffness
analysis of the acoustic and/or the squeeze film damping that in which is the inverse of compliance and is represented by the
turn affects the dynamic characteristics of resonators. Another capacitor, and x is the angular frequency. Similar notations are
important point to be noted is that this method requires very lim- used for the sqd impedance represented as Zsqd in the following
ited FEM runs or experimentally obtained data within the fre- equation:
quency range of interest. The procedure demands concentrated  
data only in the extreme high- and low-frequency regions and can Ks
Zsqd ¼ Rs þ j Ms x  (3)
still be used to predict the response accurately in the entire fre- x
quency range. This reduces the computational cost and modeling
effort significantly. On relating Eqs. (1) and (2) element by element, we can obtain
We demonstrate the proposed method using two test cases. In relations between the two domains as explained below. This is
the first case, we consider a circular plate open along its edge for summarized in Table 1. The two columns in the table relate the
fluid flow and clamped at the edge as a structure. For this struc- impedance in the acoustic domain and the modified acoustic
ture, the analytical solution is already available in the literature domain Za =jx.
[5]. In the second case, we consider a circular plate with slits Referring to the first element of both columns, it can be seen that
near its circumference with the same fluid flow and structural the resistance in the acoustic domain is same as the stiffness in the
boundary conditions as the first case. We carry out a FEA using modified acoustic domain. Now, if we assume that the modified
built-in squeeze film elements in ANSYS for this structure. The nu- acoustic domain impedance is equal to a fictitious squeeze film im-
merical solution leads to the determination of pressure distribu- pedance, i.e., ðZa =jxÞ ¼ Zsqd , then to match the units of the second
tion on the face of the plate which is subsequently integrated to column elements to be the units of an impedance, the expression
determine the damping and the stiffness components of the force. Ra =jx should have the units of impedance. This is possible only if
The extracted damping and stiffness coefficients are curve this Ra has the units of stiffness Ks. This also confirms the known
fitted to determine suitable expressions valid over a range of fact that the high-frequency region in the acoustic domain is resist-
frequencies. ance dominated while in the case of squeeze film damping, the
high-frequency regime is stiffness dominated.
Similarly, referring to the second element in both columns, it
2 Equivalence Between Acoustic and sqds can be seen that the mass loading in the acoustic domain is analo-
Both acoustic and sqds deal with pressure distribution due to gous to the resistor loading in the modified acoustic domain.
the motion of the fluid. In both the domains, the reaction force Again, to match the units of the expression in the second column
that the fluid applies on the vibrating structure is of interest and is to that of an impedance, the Ma in the acoustic domain is analo-
analyzed. Figure 2 shows a comparison between the acoustic gous to Rs in the sqd. This again confirms a fact that the low-
impedance, original squeeze film impedance, and the rescaled frequency regime is mass dominated in acoustic domain while it
squeeze film impedance that the fluid offers to the vibrating struc- is resistance dominated in the sqd.
ture. In acoustic domain, the impedance offered by the fluid is From the entries of the third element in each column, it can be
dependent on the frequency (represented by the wave number (k) inferred that the stiffness in the acoustic domain is analogous to

011005-2 / Vol. 138, FEBRUARY 2016 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 2 Comparative between acoustics and sqds: (a) acoustic impedance, (b) squeeze film
impedance (original), and (c) modified squeeze film impedance

the mass in the sqd. In order to match the units of Ka =x2 to that Here, Ma, Ra1 ; Ra2 , and Ca are the elements which are derivable
of an impedance, Ka should be equal to jMs x3 . from the analysis for simple geometries [12], and s is the Laplace
Thus, the equivalence between the two domains, viz., the variable. For complex geometries, these expressions are obtained
acoustic and the squeeze film, is described by the following from curvefits on data generated from FEM analysis [13–15].
equations: In order to model suitable expressions for the four elements, the
behavior of the circuit is evaluated in the extreme high- and low-
R a ¼ Ks (4) frequency regimes. This is done to eliminate some of the elements
which become ineffective in the extreme high- or low-frequency
Ma ¼ Rs (5) regimes. The elements which remain active in these regimes are
shown in Fig. 3. It can be reasoned out as follows. At very low-
Ka ¼ jMs x3 (6) frequency, the capacitor’s impedance, 1=jxC, is very high. So
it offers a high resistance path and hence is removed from the

3 Acoustic Modeling With Equivalent Circuit Table 1 Equivalence between the acoustic (Za) and the modi-
fied acoustic domain ðZa =jxÞ LP and the corresponding units
Figure 3 shows the electrical circuit that approximates the
behavior of the radiation impedance over a frequency range [12]. Acoustics (units) ðAcoustics=jxÞ (units)
It uses four electrical elements in the acoustic domain, viz., an
inductor representing the mass loading of the air, two resistors Ra (N s/m) Ra
(N/m)
representing the dissipative behavior of the air in the low- and jx
high-frequency domains, and a capacitor that represents the com- Resistance Stiffness
pliance behavior of the air in the high-frequency domain. jxMa (N s2/m) Ma (N s/m)
In the Laplace domain, the radiation impedance of the acoustic Inductance Resistance
circuit as seen at the end terminal can be calculated as
jKa Ka
(N/m) (N s2/m)
x x2
Ma sðRa1 þ Ra2 þ Ra1 Ra2 Ca sÞ
Za ¼ (7) Stiffness Inductance
ðCa Ma Ra1 Þs2 þ ðMa þ Ca Ra1 Ra2 Þs þ Ra1 þ Ra2

Journal of Vibration and Acoustics FEBRUARY 2016, Vol. 138 / 011005-3

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 3 Equivalent circuit for the acoustic domain with its high-
and low-frequency behaviors
Fig. 4 Pressure distribution obtained from the FEM model for
a typical frequency of oscillation of the plate
low-frequency behavior. The entire impedance then is due to the
inductor and the resistors. Therefore, the imaginary part of the used for fluid flow resulting from the oscillations of a structure
impedance is solely due to the inductor which represents the mass that has transverse motion with respect to a fixed substrate.
loading of the air in the vibrating structure. The impedance of the Using harmonic analysis, the pressure distribution over the sur-
inductor representing the mass is jxMa which is very low at low face of the vibrating structure can be calculated. The complex
frequencies. At very high-frequency, the impedance of the mass is pressure distribution can be used to extract the damping coeffi-
very high and is removed from the circuit. cient and the stiffness as well as the mass coefficients. The real
part of the pressure component contributes to the damping of the
system. The imaginary part of the pressure contributes to the stiff-
4 Numerical Modeling and Simulation Using FEA ening of the system. Figure 4 shows resultant pressure distribution
Prior to the extraction of expressions for the LP, we need to at a particular frequency.
determine the damping and the spring forces. For the first case FLUID138 is a two-noded element and it supports a circular
where the plate has all sides open for the fluid flow boundary con- hole configuration as well as a rectangular hole configuration
dition, the analytical solution is available in Darling et al. [5]. But (using KEYOPT(3)).3 The pressure at the center of the hole (flush
for the second case, where the plate has slits along its circumfer- with the plate surface in the thin film region) has the same value
ence, we determine the solution numerically using finite element as that at the nodes located on the periphery of the hole. The nodes
models in ANSYS. Analytical expressions for estimating the of the FLUID136 element at the hole periphery are coupled with
squeeze film damping losses for a structure with periodic slits is the node of the FLUID138 element as shown in Fig. 5.
well presented in the works of Homentcovschi and Miles [16] as Figure 6 shows the geometry of our plate (designed by the pres-
well as in the paper by Sun et al. [17]. In both these references, ent authors and fabricated using PolyMUMPs4 process) and the
the derived analytical formula is used for estimating the damping corresponding scanning electron microscopy (SEM) image. The
caused due to a single slot and the overall viscous damping loss is nontrivial geometry (refer to Fig. 6) comprises a circular plate
determined by multiplying the so-obtained value by the number of with slits near its circumference. The dimensions of the slits con-
slots present in the structure. Hence, this method is applicable sidered are as follows: width of the slit is 5 lm, outer radius of the
only for geometries having periodic openings. For our geometry, slit placement is 290 lm, and slit angle is 70 deg. Additionally,
where the slits are present only on the circumference of the plate, the plate also has 12 holes arranged in two circular rings of radius
such a method is not applicable. This is because we would not be 50 lm and 150 lm from the plate center. Four holes are arranged
able to develop the concept of an imaginary cell around a slit for in the inner circle with an angular spacing of 90 deg and another
our case due to the lack of periodicity in the placement of the slits. eight holes in the outer circle with an angular spacing of 45 deg.
Therefore, we resort to the numerical estimation of damping The radius of these holes is 2 lm. We choose this particular geom-
losses for our structure with holes and slits. We also do not con- etry for potential application as a front vented MEMS micro-
sider the three-dimensional analysis (solution for Navier Stokes’ phone. The pressure drop in the circular holes can be modeled
equation) for our particular case because the air gap that we con- directly using the circular hole option for FLUID138 element as
sider here is 2 lm which is of the same order as the thickness of the shown in Fig. 5.
plate which is also 2 lm. The damping contribution from the Since each slit extends over an arc of 70 deg, the opening so
squeeze film in the gap will be dominant as compared to the damp- formed cannot be approximated as a single slot of either a circular
ing contribution from the flow through the holes and slits in our or a rectangular cross section. Rather it essentially represents a
case. This circumvents the need for a rigorous three-dimensional sector of a circle. Each slit needs multiple FLUID138 elements so
analysis which is going to be computationally exhaustive. that the arc length of a single element can be approximated as a
The entire analysis is done using three different elements out of straight line. We used an angle of half a degree for each element
which two are for the fluid domain, (FLUID136 and FLUID138)2 within the slit and approximated the cross section to be a rectan-
and one is for the structural domain, i.e., SOLID185. FLUID136 gular one. In other words, a single slit is modeled with 140
and FLUID138 solve the linear/nonlinear Reynolds’ equation FLUID138 elements with each element placed side-by-side. One
within the fluid domain. FLUID136 elements are overlaid on the FLUID138 element is needed to model each imaginary
surface of the moving structural domain. FLUID138 elements
model the viscous behavior of the fluid flow in narrow channels/ 3
pipes of circular or rectangular cross section. These elements are Built-in ANSYS command used to define the cross section of fluid flow to be either
circular or rectangular/square.
4
PolyMUMPs is the commercial three-layer polysilicon surface and bulk
2
Inbuilt elements in ANSYS for solving fluid–structure interaction problems. micromachining process by the MEMSCAP Company.

011005-4 / Vol. 138, FEBRUARY 2016 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 5 FLUID138 element modeling

rectangular hole assumed in the plate. Another implicit assump-

tion is that there exists a fictitious boundary between two adjacent
holes. In order to validate the efficacy of this assumption, we per-
form a similar squeeze film analysis for a rectangular plate with
rectangular hole using both options, viz., modeling the central cut
using a single FLUID138 element as well as using four
FLUID138 elements (see Fig. 7). We observe the pressure
response and also determine the corresponding stiffness and
damping coefficients. Table 2 shows the summary of the results
obtained. It can be seen that the error is less than 6%. This ensures
that modeling the hole with multiple FLUID138 elements yields
results similar to the condition of modeling it with a single
FLUID138 element with error percentage less than 10% in the
computed stiffness and damping coefficients.

5 Calculations for the Derived Quantities From the

Numerical Results
The calculations for the various derived quantities such as
forces and impedance are explained in this section. In ANSYS, the
real component of pressure is integrated over the face of the plate
to determine the damping force Fd, and the imaginary component
of the pressure is integrated to give the stiffness force Fs. These
forces can be normalized with the product of area of the plate A
and the atmospheric pressure P. The corresponding real and imag-
inary components of the mechanical impedance can be calculated
as Fd =V and Fs =V, respectively, where V is the velocity of the
vibrating structure. Fx=V gives the curve for which acoustic cir-
cuit is valid. These basic quantities will differ in their definitions Fig. 6 Plate with slits and holes and its SEM image
in the acoustics and sqd because of the scaling Za ¼ jxZsqd . This
can be explained as follows. To obtain the normalized forces in
demarcation of the low- and high-frequency regimes depends
the sqd, Zsqd is to be multiplied by the velocity V of the structure upon the frequency of interest. In this section, we present in detail
and normalized by the product of P and A. Similarly, the normal-
the curvefit procedure we follow for the plate with slits, which
ized force in the acoustic domain is ðZa V=xPAÞ. Table 3 shows requires FEM simulations to determine its impedance. For the sec-
the calculations of various quantities derived from ANSYS data in ond case, we present only the results obtained using the same
the sqd and in the acoustics domain. The sqd and the acoustic
procedure.
domain quantities differ by a factor of x. For the case of plate with slits, we analyzed the behavior till
500 kHz frequency which corresponds to a squeeze number (r) of
6 Flowchart for the LP Extraction—Application 155. This is decided based on the frequency range of interest for
of Acoustic Circuit Into the Sqd which the desired curvefit parameters are needed. Our region of
low frequency is assumed from 10 Hz to 2500 Hz, which corre-
In order to determine the expressions for the lumped elements sponds to the squeeze number (r) between 0.003 and 0.757. The
in the sqd, we can either use the high- and low-frequency limit low-frequency range could be assumed within 0.5–2% of the total
analysis of the circuit valid in sqd or use the acoustic circuit. frequency span. On the other extreme, our high-frequency regime
When using the acoustic circuit, the squeeze film impedance is to
be multiplied by jx and then processed further. The entire proce-
dure is explained in the flowchart shown in Fig. 8. The sequence
and the methodology of parameter extraction are explained in
Sec. 7.

7 Extraction of LP in the Acoustics Domain

Fig. 7 Rectangular plate with a rectangular slot modeled in
Extraction of parameters from the simulated data requires isola- two ways—using a single FLUID138 element and using four
tion of elements in the low- and high-frequency domains. The FLUID138 elements

Journal of Vibration and Acoustics FEBRUARY 2016, Vol. 138 / 011005-5

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 2 Variation of stiffness and damping coefficients at various frequencies

Freq. C one FLUID138 K one FLUID138 C multiple K multiple %error in %error 
(Hz) (N s/m) (1) (N/m) (2) FLUID138 (N s/m) (3) FLUID138 (N/m) (4) C ð1Þð3Þ in K ð2Þð4Þ
1 2

150 1.95  105 1.10  106 1.88  105 1.03  106 3.59 5.85
25,000 1.95  105 3.06  102 1.884  105 2.88  102 3.53 5.84
150,000 1.94  105 1.09 1.87  105 1.03 3.54 5.79
300,000 1.92  105 4.34 1.85  105 4.08 3.56 5.82

Table 3 Calculations of various quantities for ANSYS simulation From the curvefit shown in Fig. 9
results, the modified acoustics, and sqds
xMa ¼ 6:7  102 r þ 0:12 (9)
Quantity ANSYS Acoustics sqd
6:7  102 r þ 0:12
Normalized force F Za V Zsqd V Ma ¼ (10)
P A x P A P A x
F Za In the high-frequency regime, since the resistance dominates, the
Impedance Zsqd
V x entire contribution of the real part of the impedance can be attrib-
F x uted to Ra2 . Hence, the value of Ra2 is determined from the high-
Impedance x Za Zsqd x frequency real impedance values
V

xImagðZsqd Þ ¼ Ra2 (11)

is assumed from 455 kHz to 500 kHz which corresponds to (r) of
138–151. The high-frequency regime can be taken from 90% of Figure 10 shows the curvefit for the determination of resistance
the total frequency span onward. It is in these two frequency in the high-frequency domain. From the fitted values, we find that
regimes that we carry out FEM simulations with finer spacing in
frequency. In the intermediate frequency zone, the FEM simula-
tions are sparsely spaced. Ra2 ¼ 0:0186r2 þ 1:134  104 (12)
The low-frequency behavior can be approximated as a mass
shunted by the total resistance for the acoustics domain [12]. In In the low-frequency regime of operation, the impedance load
our analysis, since we are using the acoustic circuit to analyze the presented by the air on the resonator is that of a mass shunted by a
squeeze film data, the low-frequency real part of the impedance very large resistance [12]. This facilitates the determination of Ra1
will represent only the mass loading. So, the coefficient for Ma is in combination with Ra2 and Ma to match the low-frequency real
determined from the low-frequency real valued impedance. Figure impedance data (see Fig. 11). The vertical axis in this figure repre-
9 shows the curvefit for the determination of mass in the low- sents the total resistance which is a combination of Ra1 and Ra2 .
frequency regime. The horizontal axis is the squeeze number r The mathematical proof for this can be understood as follows:
which is a nondimensional number and is a function of frequency.
It is written as r ¼ ð12la2 =gh20 Pa Þx. Here, l is the viscosity of 1 1 1
the fluid medium, a is the radius of the circular plate, x is the ¼ þ (13)
Zlow Rtotal jxMa
angular frequency, ho is the initial air gap, Pa is the atmospheric
pressure, and g is the index of the process Since at extremely low frequencies, x2 Ma2  R2total , therefore
the expression for the impedance at extremely low frequencies
xRealðZsqd Þ ¼ xMa (8) can approximated as

Fig. 8 Various steps in determining the expressions for the equivalent elements of the sqd
using the circuit valid in acoustic domain

011005-6 / Vol. 138, FEBRUARY 2016 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 9 Low-frequency real impedance for determination of
mass in the acoustics domain which is equivalent to resistance Fig. 11 Low-frequency imaginary impedance for determination
loading in the sqd. The lowest value of r on the X axis is 0.003. of total resistance in the acoustics domain in the low-frequency
regime which is equivalent to stiffness in the sqd. The lowest
value of r2 on the X axis is 9 3 1026 .
jxMa Rtotal ðRtotal  jxMa Þ
Zlow ¼ (14)
R2total coefficient that requires the values of impedance data in the inter-
mediate frequency zones. This coefficient is chosen such that it
It can be seen from Eq. (14) that the imaginary part of the equa- gives a minimum error between the curvefit and the reference val-
tion involves only the mass Ma and the real part of the equation ues (ANSYS or analytical). Another important point is that the simu-
can be used to find Rtotal once Ma is known from the imaginary lated data may be needed to be closely spaced around the cutoff
part point to capture the peak accurately

x2 Ma2 Ka
ImagðZlow Þ ¼ (15) xRealðZsqd Þ ¼ (17)
Rtotal x

Equation (15) can be rearranged to determine the value of Rtotal Ka

¼ 4:5  102  0:5r (18)
from the value of xMa and real impedance values in the low- x
frequency regime. This is shown in Fig. 11.
From the curvefit in Fig. 11, the value of total resistance that 1
Ca ¼ (19)
the air offers is calculated as follows: ð4:5  102  0:5rÞx

Ra2 þ Ra1 ¼ 41r2 þ 1:2  104 (16) We carry out similar curvefit analysis for a simple case of a cir-
cular plate open on all sides for the fluid boundary condition. The
Finally, we determine the coefficient for Ca, which has an analytical solution for this is available in Ref. [5]. For this simple
impact only in the transition region. It is the determination of this case, we analyzed the behavior up to 384 kHz frequency which
corresponds to a squeeze number (r) of 117. Our region of low
frequency is assumed from 10 Hz to 3500 Hz which ends up at the
squeeze number (r) of 1.06. On the other extreme, our high-
frequency regime is assumed from 364 kHz to 384 kHz which
amounts to a squeeze number (r) of 110.7 to 117. These ranges of
simulations are clearly shown in Fig. 12 presented in Sec. 8 for
the case of plate without slits. It is worth mentioning here that the
curvefit process on analytical solution is done only for demonstra-
tion purpose here.
Table 4 shows a summary of the various coefficients obtained
from the analysis for both the cases, i.e., a plate with and without
slits. It must be noted that since acoustic circuit is used for simu-
lating sqd impedance, Ma which is the mass in the acoustic do-
main will represent the damping Rs in the sqd. Similar relations
can be derived for the remaining quantities as per Table 1.

8 Numerical Results for Validation of the Proposed

Model
We first validate our modeling procedure with a standard
benchmark problem for which the analytical results are available
in the literature. We choose a solid circular plate structurally
Fig. 10 High-frequency imaginary impedance for determina- clamped on its outer edge but having all sides open as its fluid
tion of resistance in the acoustics domain which is equivalent boundary condition. This kind of structural configuration is com-
of stiffness in the sqd mon for MEMS devices fabricated using wafer bonding. The

Journal of Vibration and Acoustics FEBRUARY 2016, Vol. 138 / 011005-7

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 12 Acoustic impedance (which is equal to the squeeze Fig. 13 Acoustic impedance (which is equal to the squeeze
film impedance multiplied by jx) obtained using curvefit and film impedance multiplied by jx) obtained using curvefit and
analytical formulation for the plate without slits. The relevant analytical formulation for the plate with slits
dimensions for the calculation of the damping and the stiffness
impedances analytically are: h0 5 2 lm, g 5 1 for the isothermal
process, l 5 1.85 3 1025 Pas, a 5 295 lm, hdash 5 0.1 (displace- calculated as ðImðZa ÞV=xÞ and its normalized value is calculated
ment assumed is one-tenth of the original gap), and V 5 xhdash. as ðImðZa ÞV=xPa AÞ. Here, A is the area of the plate, and Pa is the
atmospheric pressure. Similarly, the stiffness force can be calcu-
lated using the real component of impedance ðReðZa ÞV=xÞ. This
analytical solution for this case is discussed by Darling et al. [5]. is the method for the curvefit data. It must also be noted that the
The match between our curvefit results and the analytical results damping force from ANSYS results is calculated here as ðfre =Pa AÞ.
validates our method of parameter extraction. Figure 12 shows the Figure 15 shows the original squeeze film impedance for the
validation of the numerically obtained results for a solid plate plate with slits. The damping impedance for the sqd is calculated
using our model with the results from Darling et al. Here, we are as ðImðZa Þ=xÞ and the stiffness impedance for the sqd is calcu-
plotting the rescaled squeeze film impedance (which is equal to lated as ðReðZa Þ=xÞ. This is done for the curvefit results. The
the acoustic impedance) obtained using both the analytical expres-
damping impedance from the ANSYS data is calculated as ðfre =VÞ.
sions as well as the curvefit parameters. The shaded zones in the Another point worth mentioning is that after determination of
low- and high-frequency regimes are the areas where the curvefit the curvefit parameters for a particular range of frequency, it can
was carried out. The damping impedance is calculated as be extended accurately beyond the simulation range. In other
xðfim =VÞ, and the stiffness impedance is calculated as xðfre =VÞ. words, the parameters remain valid beyond the range used for
Here, fim is the analytically determined imaginary component of simulation. But this range of validity cannot be generalized and
force (in the original sqd), while fre is the real component (in the varies from case to case. This is illustrated for the first case as
original sqd) and V is the velocity of the vibrating plate. The shown in Fig. 16. The original curvefit was carried out up to a r
damping impedance is equal to ImðZa Þ while the stiffness imped- value of 117 but the curvefit parameters can be readily used up to
ance is calculated as ReðZa Þ for the acoustic domain. a r value of 170.
An important point to be mentioned here is the following. The
analytically obtained impedance in the sqd is Zanalytical
¼ ðfim =VÞ  jðfre =VÞ. We have plotted the absolute values for the 8.1 Simulation of Frequency Response. We demonstrate the
stiffness impedances in all our results. efficacy of our model with the simple illustrator of a resonator
Figure 13 shows the acoustic impedance for the plate with slits with a circular plate clamped at its periphery and open on its sides
determined using ANSYS and the curvefit parameters. While plot- for the squeeze film damping boundary condition. In order to
ting the ANSYS results, it is the real component of the force which
is used to calculate damping impedance. This is different from the
analytical case mentioned above because in ANSYS, the real com-
ponent of force represents the damping. It can be noted from Figs.
12 and 13 that the low-frequency regime is damping dominated
and the high-frequency regime is stiffness dominated for the sqd.
Figure 14 shows the normalized damping and spring forces (for
the original sqd) for the plate with slits. The damping force is

Table 4 Summary of the curvefit values for various LP

LP Curvefit for a plate with slits Curvefit for a solid plate

Ma 6:7  102 r þ 0:12 1:7  103 r þ 10

x x
Ra2 0:0186r2 þ 1:134  104 0:035r2 þ 1:14  104
Ra1 41r2 þ 1:2  104  Ra2 301r2 þ 1:01  104  Ra2
1 1
Ca
ð450  0:5rÞx ð1000  0:09rÞx Fig. 14 Normalized forces in the original sqd obtained using
curvefit and analytical formulation for the plate with slits

011005-8 / Vol. 138, FEBRUARY 2016 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 15 Acoustic impedance divided by x (which is equal to
the squeeze film impedance) obtained using curvefit for the
plate with slits

simulate the frequency response of the resonator, we make a few

assumptions to reduce the number of impedances to be handled.
We assume that the electrical forces are absent and hence we do
not consider the electrostatic stiffness. Further, we also assume
that the only dissipation mechanism for this resonator is due to the
squeeze film damping. Hence, we do not consider the contribution
due to the acoustic radiation impedance to the overall impedance
of the resonator. The entire circuit can be seen as shown in
Fig. 17. Fig. 17 Equivalent circuit for a resonator. The effect due to the
The response of the resonator could be written as the following. radiation impedance, electrostatic stiffness is ignored for analy-
This response is the open circuit response without any external sis purpose.
circuitry connected to the resonator

I 1 acoustic units. The expressions for these two quantities can be

¼ (20) written as follows [18].
Vin Zm The lumped acoustic compliance of a circular diaphragm can
be written as
Here, I is the output current, Vin is the voltage equivalent of the
input signal to the resonator which could be sound in the case of a
microphone, Zm is the mechanical impedance of the whole system
and can be written as Zm ¼ Zplate þ Zsqd  Zplate is the impedance of
the resonator due to its mass and compliance and can be written as
Zplate ¼ jxMdia þ ð1=jxCdia Þ, where Mdia is the equivalent mass
of the resonator and Cdia is its equivalent compliance in the

Fig. 18 Simulated frequency response of the simulator using

the squeeze film parameters derived from our curvefit and
using the squeeze film damping and stiffness forces from Dar-
ling et al. [5]. The relevant dimensions for the calculation of the
acoustic compliance and mass are: a 5 295 lm, q 5 2320 kg=m3 ,
h 5 2 lm, E 5 160 GPa, and m 5 0.22. The relevant dimensions for
Fig. 16 Acoustic impedance divided by x (which is equal to the calculation of the damping and the stiffness impedances
the squeeze film impedance) obtained using curvefit for the analytically are: h0 5 2 lm, g 5 1 for the isothermal process,
solid plate without slits. The range of applicability is extended l 5 1.85 3 1025 Pas, hdash 5 0.1(displacement assumed is one-
beyond the range of values used for curvefit procedure. tenth of the original gap), and V 5 xhdash.

Journal of Vibration and Acoustics FEBRUARY 2016, Vol. 138 / 011005-9

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

   frequencies, the present method would need modifications. This
pa6 1  ð Þ2 is because of the type of curvefitting used in the higher
Cdia ¼ (21)
16Eh3 frequency regime wherein the variation of the acoustic impedance
Za ¼ jxZsqd is assumed to be quadratic in nature with respect to
where a is the radius, h is the thickness, E is the Young’s modulus, the frequency. The method would need a suitable modification in
and  is the Poisson’s ratio of the diaphragm. the curvefit procedure in the high-frequency regime to properly
The lumped acoustic mass of a circular diaphragm is mimic the high-frequency behavior in such a case.
9qh
Mdia ¼ (22)
5pa2
where q is the density of the diaphragm. Acknowledgment
Figure 18 shows the simulated response of the resonator using This work was partially supported by MCIT grant CEN phase
the equivalent circuit parameters derived from curvefit as well as II, NPMASS Grant 3.14. The authors are thankful to the anony-
using the analytical expression available in Darling et al. [5] for mous reviewers for the useful inputs given during the review
the squeeze film damping and stiffness forces. stages of the manuscript.
In both of the simulated cases, the Zplate remains the same and
is determined from the expressions given by Eqs. (21) and (22).
The only difference in the calculation of frequency response as
given by Eq. (20) is the way Zsqd is calculated. The Zsqd References
¼ ðZa =jxÞ in one case (refer to Fig. 17) and is equal to [1] Bao, M., and Yang, H., 2007, “Squeeze Film Air Damping in MEMS,” Sens.
Actuators A, 136(1), pp. 3–27.
ðfim =VÞ  jðfre =VÞ for the other. The contributions from other [2] Langlois, W., 1962, “Isothermal Squeeze Films,” Q. Appl. Math., 20(2),
sources of impedance (e.g., the radiation impedance and the elec- pp. 131–150.
trostatic stiffness) are neglected for analysis purposes. [3] Griffin, W., Richardson, H., and Yamanami, S., 1966, “A Study of Fluid
It can be seen from Fig. 18 that the simulated resonance fre- Squeeze-Film Damping,” ASME J. Basic Eng., 88(2), pp. 451–456.
[4] Blech, J. J., 1983, “On Isothermal Squeeze Films,” ASME J. Lubr., 105(4),
quency as well as the magnitude of output is similar in both the pp. 615–620.
cases and the percentage variation is 0.6% in the resonant fre- [5] Darling, R., Hivick, C., and Xu, J., 1998, “Compact Analytical Modeling of
quency and 1.43% in the magnitude of the output response. Squeeze Film Damping With Arbitrary Venting Conditions Using a Greens
Function Approach,” Sens. Actuators A, 70(1–2), pp. 32–41.
[6] Pandey, A. K., Pratap, R., and Chau, F. S., 2007, “Analytical Solution of Modi-
9 Conclusions fied Reynolds Equation for Squeeze-Film Damping in Perforated MEMS
Structures,” Sens. Actuators A, 135(2), pp. 839–848.
We have demonstrated the equivalent parameter extraction pro- [7] Mohite, S. S., Kesari, H., Sonti, V., and Pratap, R., 2005, “Analytical Solutions
cedure for the sqd using the equivalent electrical circuit elements for the Stiffness and Damping Coefficients of Squeeze Films in MEMS Devices
valid for the acoustics domain. We have established the equiva- With Perforated Back Plates,” J. Micromech. Microeng., 15(2), pp. 2083–2092.
[8] Pratap, R., Mohite, S. S., and Pandey, A. K., 2007, “Squeeze Film Effects in
lence between the two domains mathematically. Two cases are MEMS Devices,” J. Indian Inst. Sci., 87(1), pp. 75–94.
considered to demonstrate the parameter extraction. The first case [9] Mehner, J., Kurth, S., Billep, D., Kaufmann, C., Kehr, K., and Dotzel, W.,
study is that of a structurally edge clamped plate with fluid free to 1998, “Simulation of Gas Damping in Microstructures With Nontrivial Geo-
flow at the outer edge. The second case is that of a plate with slits metries,” 11th Annual International Workshop on Micro Electro Mechanical
Systems (MEMS98, 1998), pp. 172–177.
on its periphery. For the first case, we compare our curvefit values [10] Lamb, H., 1910, The Dynamical Theory of Sound, E. Arnold, London, UK.
with the standard analytical solution. For the second case, we [11] Kinsler, L. E., Frey, A. R., Coppens, A. B., and Sanders, J. V., 1999, Funda-
extract the LP using curvefit on the FEA data in the high- and mentals of Acoustics, Wiley, New York.
low-frequency regimes. The extracted parameters have been used [12] Beranek, L., 1986, Acoustics, AIP, New York.
[13] Doody, C. B., Xiaoyang, C., Rich, C. A., Lemmerhirt, D. F., and White, R. D.,
for dynamic analysis of the plate over the entire frequency range 2011, “Modeling and Characterization of CMOS-Fabricated Capacitive
of interest. We have demonstrated this with a simple illustration Micromachined Ultrasound Transducers,” J. Microelectromech. Syst., 20(1),
of simulating the frequency response of a plate mentioned as the pp. 104–118.
first case above. The squeeze film damping of the plate is modeled [14] Krause, J. S., 2011, “Micromachined Microphone Array on a Chip for Turbu-
lent Boundary Layer Measurements,” Ph.D. thesis, TUFTS University, Med-
using analytical formulae as well as the curvefit parameters. The ford, MA.
resulting overall error in the response values and the resonant fre- [15] Malhi, C. K., and Pratap, R., “On the Method of Extraction of Lumped Parame-
quency is less than 2%. We have shown that the extracted parame- ters for the Radiation Impedance of Complex Radiator Geometries,” Microsyst.
ters can be effectively used to model the effect of fluid trapped Technol. (in press).
[16] Homentcovschi, D., and Miles, R. N., 2004, “Modeling of Viscous Damping of
between a fixed and a moving structure. The proposed method can Perforated Planar Microstructures. Applications in Acoustics,” J. Acoust. Soc.
be readily used for any geometry. A major advantage is the Am., 116(5), pp. 2939–2947.
reduced finite element computations apart from obtaining the [17] Sun, Y., Bao, M., Yang, H., and Huang, Y., 2006, “Modified Reynolds Equa-
frequency-based expressions for the squeeze film phenomenon. tion for Squeeze-Film Air Damping of Slotted Plates in MEMS Devices,” Chin.
J. Semicond., 27(3), pp. 473–477.
The present method is based on rescaling of the squeeze film [18] Martin, D. T., Liu, J., Kadirvel, K., Fox, R., Sheplak, M., and Nishida, T., 2006,
impedance by a factor of jx. When the squeeze film theory breaks “Development of a MEMS Dual Backplate Capacitive Microphone for Aeroa-
down at lower pressure and/or lower gaps as well as at very high coustic Measurements,” AIAA Paper No. 2006-1246.

011005-10 / Vol. 138, FEBRUARY 2016 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 12/15/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use