A New Analytical Solution for Diaphragm Deflection and its

Application to a Surface-Micromachined Pressure Sensor

William P. Eaton, Fernando Bitsie, James H. Smith, David W. Plummer
Sandia National Laboratories
P. O. Box 5800
Albuquerque, NM 87185-1081
ABSTRACT
An analytical solution Ior large deIlections oI a clamped
circular diaphragm with built-in stress is presented. The
solution is directly applicable to micromachined pressure
sensors. The solution is compared to Iinite element analysis
results and experimental data Irom a surIace-
micromachined pressure sensor.

: MEMS, microsystems, pressure sensor,
diaphragm
INTRODUCTION
Several papers have recently been published regarding
the deIlection characteristics oI micromachined diaphragms
Ior application to pressure sensors and other acoustic
devices |1-5|. For these devices, the applied load is
assumed to be constant over the diaphragm surIace. Optical
and capacitive based devices measure diaphragm deIlection
directly, and deIlected shape and amplitude is oI
importance. Piezoresistive devices, which are the Iocus oI
this work, measure deIlection indirectly. Diaphragm stress
is oI importance in these devices.
Micromachined pressure sensors are currently a
multibillion dollar industry. More than 16 million
disposable blood pressure sensors and 25 million maniIold
absolute pressure sensors are produced annually |6|. Since
pressure sensors are primarily diaphragm-based, it is
important to have models Ior their stresses and strains, as
well as diaphragm deIlection.
In this work, we present an analytical solution Ior a
clamped circular diaphragm with built-in stress and large
deIlections. In general, analytical and exact variational
solutions Ior diaphragm behavior are desirable because oI
their ease oI use and the insight they provide to the
designer. SpeciIic geometric eIIects can be ascertained
Irom these solutions. However, these solutions are
generally only applicable Ior small deIlections. Numerical
techniques, such as Finite Element Analysis, Boundary
Element Analysis, and Finite DiIIerence Analysis, can be
more accurate in predicting deIlection behavior, especially
Ior large deIlections. UnIortunately, these techniques
generally require more eIIort to use and may not supply the
same insight as analytical or exact variational solutions.
The use oI plate theory is appropriate Ior the analysis oI
micromachined thin-Iilm diaphragms. Thin plate or small
deflection theory is oIten used, and is appropriate Ior
deIlections less than 1/5 oI the diaphragm thickness |7|.
Large deflection or thick plate theory is used Ior deIlections
up to three times the diaphragm thickness |7|.
The work in this paper is unique, since it combines the
eIIects oI large deIlections and built-in stresses, thereby
uniIying work reported by Suhir |8| and Voorthuyzen and
Bergveld |1|. Both large deIlections and built-in stresses
are apt to be present in real-world micromachined devices,
especially surIace-micromachined devices. The analytical
expression is compared to both Iinite element analysis
results and experimental results Irom a circular, planar,
surIace-micromachined pressure sensor.
THEORY
Small Deflection/Membrane Theory
Solutions Ior small deIlection theory and membrane theory
are well known, but are repeated here Ior convenience. For
small deIlection, thin plate theory deIlection is said to be
dominated by the resistance oI the diaphragm to bending.
DeIlection, w, oI a clamped circular plate under a uniIorm
applied pressure P is given by |7|
, 1
64
) (
2
2
4
(
(

|
.
|

\
|
=
a
r
D
Pa
r w (1)
where r, a, are the radial coordinate and diaphragm radius,
respectively. D is the Ilexural rigidity, which is a
measurement oI stiIIness, and is given by
)
=
2
1 ( 12
3
Eh
D , (2)
where E, h, and are Young`s modulus, plate thickness,
and Poisson`s ratio, respectively.
In contrast to small deIlections theory, deIlection
in membrane theory is dominated by stresses in the plate
|2|:

(
(

|
.
|

\
|
1 =
2
a
r
h
Pa
r w
i
4
) (
2
, (3)
where
i
is the intrinsic built-in stress oI the plate.
Large Deflections with built in stress
The governing diIIerential equations Ior the bending oI
a circular plate are |7, 8|

2
2
4
r
w
r Dr
h
D
P
w

+ =

(4)
This is a corrected version oI the manuscript originally published presented at International Conference on Modeling and
Simulation of Microsvstems, MSM 99. The only correction is Eq 17.

,
2
2
4
r
w
r
w
r
E

= (5)
where is the Airy stress Iunction and the biharmonic
operator
4
is given by (Ior azimuthal symmetry)

2
2
2
4
1
|
|
.
|

\
|

=
r r r
. (6)
Most micromachined diaphragms are considered to have
clamped boundary conditions, that is
0 , 0 ) ( =

= =
=edge r
r
w
edge r w . (7)
An additional boundary condition is required to solve the
problem. That is, that the amount oI stretch, u, at the edge
oI the diaphragm is zero. This is Iurther cast in terms oI the
circumIerential strain through the center thickness oI the
diaphragm,

0
, and the built-in residual strain,
i
i.e. |1|
( ) 0
0
= =
=
=
a r
i
a r
r u

. (8)
This is Iurther developed to become |1|

( )
.
1
2
2
0 0 0
E
r r r
E
i
i
a r
r
a r

= =
=
=
(9)
The deIlection and Airy stress Iunctions have the
presumed solutions
( )
4
2
2
1
1 ) ( r C r C f r w + + = (10)
( ) ,
8
4
6
3
4
2
2
1
2
r B r B r B r B f + + + = (11)
where f is the maximum deIlection oI the plate. Applying
the boundary conditions oI Equation (7) to Equation (10)
yields

2
2
1 ) (
(
(

|
.
|

\
|
=
a
r
f r w , (12)
which has the same dependence on r as the small deIlection
case. This solution, when substituted into Equation (5),
simpliIies to

(
(

|
.
|

\
|
48
1
|
.
|

\
|
9
1
+ |
.
|

\
|
4
1
=
8 6 4
2
a
r
a
r
a
r
r B f r
1
2
) (
. (13)
II the boundary condition oI Equation (9) is invoked, the
solution Ior becomes

(
(
(
(
(

|
.
|

\
|
48
1
|
.
|

\
|
9
1
+ |
.
|

\
|
4
1

|
.
|

\
|
) 2(1
+ |
.
|

\
|
|
.
|

\
|
1

=
8 6
2 2
a
r
a
r
a
r
a
r
a
r
E f
r
i
4
2
3 5
12
) (

. (14)
The Iinal step in the derivation is to solve Ior f. This is
done by applying the Bubnov Galerkin method. But instead
oI using the energy Iunctional as the coordinate, we
minimize the governing diIIerential equation oI (4) and
assume that w(r) is orthogonal with respect to all other
coordinate Iunctions, i.e.
0
4
=
(

|
.
|

\
|

|
.
|

\
|

}
A
dA P
r
w
r r r
h
w D w

. (15)
Integrating and collecting terms gives
0 =
3

) (1 9
16
+
) (1 63
) 9 )(23 +
2 2
3
2 2
2
3
1 ( 2 Pa
f
a
Eh
f
a
Eh


. (16)
This is a cubic equation with three roots. The real root oI f
is given by
( )
.
1 3 4
14
2 2
2
3 3
) 9 )(23 + (1 8
7
=
) 9 )(23 + (1
+ +
=
4
+
27
= + + =
2 4 2
2 3

D
h Pa a h
f
i
(17)
The Iunction w(r) is plotted Ior both small and large
deIlection theory in Figure 1. The curves are nearly the
same Ior small deIlections and diverge Ior larger
deIlections, with thin plate theory overestimating the actual
deIlection. A marked reduction oI deIlection is predicted by
the case oI large deIlections with stress.


The radial and circumIerential stresses in the diaphragm,

r
and

are given by |8|

,
1 1 6
1 6
2
2
2
2
2
2
2
2
r r r
w
r
w
r h
D
r r r
w
r r
w
h
D
r

+
|
|
.
|

\
|

+
|
|
.
|

\
|

(18)
and the strains can be calculated Irom the plane strain
condition |8|:

( )
( ) .
r
r r

+
+

1
=

1
=
(19)
Output characteristics oI a piezoresistive pressure
0
0.3
0.6
0.9
1.2
1.5
1.8
0 2 4 6 8
Pressure [atm]
D
e
f
I
e
c
t
i
o
n

[

m
]
Small Def lection, Eq. (1)
Large def lection,Eq 12 & 17, no strain
Large def lection, Eq. 12 & 17, with strain
Figure 1. DeIlection vs. Pressure Ior 100 m
diameter, Si
3
N
4
diaphragm. E 300 GPa, 0.24, h
1.2 m,
i
5.510
4
.

similar to that shown in Figure 4 can be calculated using
Equations 19 and examples are shown in Figure 2 Ior
polysilicon and silicon nitride based diaphragms. The
relatively high stress present in micromachined silicon
nitride diaphragms has a proIound eIIect on the output.
SpeciIically, the 100, 150, and 200 m diameter
diaphragms all have similar sensitivity. This result is
unexpected Irom small deIlection theory.

FINITE ELEMENT ANALYSIS
Finite element analysis (FEA) is a valuable design tool
Ior a diaphragm or sensor designer, since it generally gives
more accurate results than analytical solutions. Perhaps
even more important is the ability to model complex
structures, which is diIIicult with analytical solutions. A
circular diaphragm with Iixed edges and a constant residual
stress was modeled in Abaqus. The results are summarized
in Figure 3 and are compared to the analytical solution Ior
several diaphragm sizes. The results agree well, with the
best agreement Ior the 100 m diameter diaphragm.
EXPERIMENTAL DATA
A surIace micromachined pressure sensor |9-12| is
shown in Figure 4. It is a piezoresistive sensor and hence
does not detect diaphragm deIlection directly, but rather by
changes in the diaphragm stress state. The output signal oI
the sensor is the output voltage oI a Wheatstone bridge.
DeIlection data Ior pressure sensors at a constant
external pressure oI 12 psi was extracted with a Wyko
white light interIerometer and is plotted in Figure 5, along
with the analytical expression. The data agree Iairly well,
but discrepancies between the experimental data and the
analytical expression oI Equations 12 and 17 may be due to
uncertainties in physical constants oI the diaphragm (i.e. E
and ), residual stress, and diaphragm thickness. All oI
these parameters are diIIicult to measure with precision on
micron scale structures.
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
-100 -75 -50 -25 0 25 50 75 100
RadiaI position [m]
D
e
f
I
e
c
t
i
o
n

[

m
]
Analytical
FEA
Figure 3. Comparison oI FEA and analytical results
Ior Iour silicon nitride diaphragm diameters: 50 m,
100 m, 150 m, and 200 m. P 12 psi, E 300
GPa, 0.24, h 1.4 m,
i
6.5810
4
.

Figure 4. Scanning electron micrograph oI surIace
micromachined pressure sensor diaphragm |9-12|.
Diaphragm is 100 m in diameter and has six radial
and one circumIerential piezoresistor(s).
0
2
4
6
8
10
0 20 40 60 80 100
AppIied Pressure
O
u
t
p
u
t

v
o
I
t
a
g
e
50 m
100 m
150 m
200 m
PoIysiIicon Diaphragm
0
2
4
6
8
10
12
14
16
0 20 40 60 80 100 AppIied Pressure
O
u
t
p
u
t

v
o
I
t
a
g
e
50 m
100 m
150 m
200 m
SiIicon Nitride Diaphragm

Figure 2. Prediction oI piezoresistive pressure sensor
characteristics Ior polysilicon (top) and silicon nitride
(bottom) diaphragms. For polysilicon: E 180 GPa, h 2
m,
i
510
6
. For nitride: E 300 GPa, h 1.4 m,
i

510
6
.

Output data Irom pressure sensors is shown in Figure 6.
The results are qualitatively similar to the analytical model
oI Figure 2. The bending over oI the 150 and 200 m
diameter nitride sensors is due to the diaphragms contacting
the substrate. The 200 m diameter curve is missing Irom
the polysilicon diaphragm because oI an incomplete release
etch during device Iabrication. As the analytical solution
predicts, sensitivity clustering does occur in the silicon
nitride diaphragm.
DISCUSSION & CONCLUSIONS
A new analytical solution Ior large deIlections oI a
clamped, circular diaphragm with built-in strain has been
presented. The solution agrees well with Iinite element
analysis, but less well with interIerometry Irom an actual
surIace-micromachined pressure sensor diaphragm.
Imprecise data on Young`s modulus and residual stress are
likely contributors to this discrepancy. Nevertheless, the
solution predicts the phenomenon oI sensitivity clustering
in nitride-based diaphragms. This solution is a powerIul,
easy to use tool Ior micromachinists.
ACKNOWLEDGMENTS
Sandia is a multiprogram laboratory operated by the
Sandia Corporation, a Lockheed Martin Company, Ior the
United States Department oI Energy under contract DE-
AC04-94-AL85000.
REFERENCES
|1| J. A. Voorthuyzen and P. Bergveld, Sensors and
Actuators, 6, pp. 201-213, (1984).
|2| R. Schellin, G. Hess, W. Khnel, C. Thielemann, D.
Trost, J. Wacker, and R. Steinmann, Sensors and Actuators
A, 41-42, pp. 287-292, (1994).
|3| D. Maier-Schneider, J. Maibach, and E. Obermeier,
Journal of Microelectromechanical Svstems, 4(4), pp. 238-
241, (1995).
|4| H. E. Elgamel, Sensors and Actuators A, 50, pp. 17-22,
(1995).
|5| R. Steinmann, H. Friemann, C. Prescher, and R.
Schellin, Sensors and Actuators A, 48, pp. 37-46, (1995).
|6| R. H. Grace, Guest editorial, Sensors Maga:ine, 16(2),
(1999).
|7| S. Timoshenko and S. Woinosky-Krieger, Theorv of
Plates and Shells, McGraw Hill Classic Textbook Reissue
(1987).
|8| E. Suhir, Structural Analvsis in Microelectronic and
Fiber-Optic Svstems, Volume I, Van Nostrand Reinhold
(1991).
|9| W. P. Eaton and J. H. Smith, Proceedings of Smart
Structures and Materials 1995, Jol. 2448, pp. 258-265
1995.
|10|W. P. Eaton and J. H. Smith, Micro-machined Devices
and Components II, Jol. 2882, Austin, TX pp. 259-265
1996.
|11|W. P. Eaton and J. H. Smith, Smart Materials and
Structures, 6, pp. 530-539, (1997).
|12|W. P. Eaton, Surface Micromachined Pressure Sensors,
Ph.D. Dissertation, University oI New Mexico, 1997.
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
-100 -75 -50 -25 0 25 50 75 100
RadiaI position [m]
D
e
f
I
e
c
t
i
o
n

[

m
]
Analytical
nterf erometer

Figure 5. Comparison oI FEA and analytical results
Ior Iour silicon nitride diaphragm diameters: 50 m,
100 m, 150 m, and 200 m. P 12 psi, E 300
GPa, 0.24, h 1.4 m,
i
6.5810
4
.
0
1
2
3
4
5
0 20 40 60 80 100 120
AppIied Pressure [psi]
O
u
t
p
u
t

v
o
I
t
a
g
e

[
m
V
/
V
]
50 m
100 m
150 m
PoIysiIicon Diaphragm


SiIicon Nitride Diaphragm
0
1
2
3
4
5
6
0 20 40 60 80 100 120
AppIied pressure [psia]
O
u
t
p
u
t

V
o
I
t
a
g
e

[
m
V
/
V
]
50 m
100 m
150 m
200 m

Figure 6. Experimental output voltage characteristics
Irom devices similar to Figure 4 Ior polysilicon
diaphragms (top) and silicon nitride diaphragms
(bottom).