1074

IEEE SENSORS JOURNAL, VOL. 13, NO. 3, MARCH 2013

Piezoelectric Vibratory-Cantilever Force Sensors

and Axial Sensitivity Analysis for Individual
Triaxial Tactile Sensing
Kaoru Yamashita, Member, IEEE, Yi Yang, Takanori Nishimoto, Kazuya Furukawa,
and Minoru Noda, Member, IEEE

Abstract Vibratory force sensors are fabricated using piezoelectric capacitors on microcantilever structures for triaxial
sensitivity by the individual sensor element. The cantilevers have
been formed into a 3-D curved shape by controlling residual
stress combination of the multilayered structure. Triaxial tactile
sensitivity of the cantilever sensor is analyzed under a tactile
load application onto the surface of an elastomer in which the
cantilever is embedded, mimicking human skin structure. The
cantilever is converse-piezoelectrically excited by an external ac
voltage and three resonant modes are developed to detect the
applied load vector components by the single sensor element.
Resonant frequency shifts of each mode are investigated upon
load applications. The results show that the frequencies vary
to the three axial tactile loads independently and they can be
superposed with corresponding to the superposition of the load
components. The applied load vectors are estimated by resonant
frequencies of the single cantilever sensor with compensating
nonlinearities of the sensor response. The estimated error is less
than 1.1% to the full scale of the load 4 kPa.
Index Terms Frequency shift,
sensitivity, vibratory cantilever.

tactile

sensor,

triaxial

I. I NTRODUCTION

INIATURIZED tactile sensors have increasingly investigated for precise touch sensing on robots especially
for nursing care applications against the recent rapid increase
of the aged population. The human-caring robots need sophisticated multiaxial tactile sensitivity to not only pressure but
also slippage to prevent the robots from damaging the touching
object. Silicon-micromachined multiaxial tactile sensors have
been developed using strain-gauge cantilevers embedded in
an elastomer material mimicking human skin structure. They
have sensitivity to stress along with normal and two orthogonal
shear directions caused by applied tactile load on the elastomer
surface, by using a combination of flat cantilevers [1], horizontal and vertical cantilevers [2], or slanted cantilevers [3].
Manuscript received June 1, 2012; revised December 21, 2012; accepted
December 27, 2012. Date of publication January 3, 2013; date of current
version February 4, 2013. This work was supported in part by Scientific Grant
2001017 from Tateishi Science and Technology Foundation, and Grant-inAide for Challenging Exploratory Research 23656239 from Japan Society for
the Promotion of Science. The associate editor coordinating the review of this
paper and approving it for publication was Dr. Anna G. Mignani.
The authors are with the Kyoto Institute of Technology, Kyoto 6068585, Japan (e-mail: yamashita.kaoru@kit.ac.jp; m1621036@edu.kit.ac.jp;
shine6m24d@hotmail.co.jp; k0712kazuya@yahoo.co.jp; noda@kit.ac.jp).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JSEN.2012.2237547

Although the developed cantilever sensors measure a tactile

vector by using three or four sensor elements, multiaxial
sensitivity by a single sensor element is required for integration
into a sensor array in a high density and for reduction of the
number of wirings. Moreover, the tactile vector information
might be distorted and measure errors would be caused if the
components of the single vector are measured by the sensor
elements located at different positions. One sensor element of
the strain gauge-type sensors, however, can detect only one
direction of the force vector due to the limitation of static
measurement principle.
On the other hand, piezoelectric vibratory tactile sensors
have been developed for position and distribution measurement
of tactile objects [4], or for elasticity and viscosity measurement [5]. Using the dynamical measurement technique, one
sensor element of vibratory force sensors has possibility to
detect plural components of the force information through
plural vibration frequencies or vibration modes. That is, they
have potentiality to measure the vector components of the
tactile information by individual sensor element [6].
The authors propose vibratory cantilever force sensors
embedded in the elastomer material for three-dimensional
vector tactile sensing by individual sensor element. The
cantilevers have three-dimensionally curved shape and can
respond to stress along with all directions. The vibration state
changes by the external tactile load applied on the elastomer
surface through the internal stress distribution change in the
elastomer-cantilever system. The tactile load is consequently
detected as a resonant frequency change in a resonant mode.
The vibration of the three-dimensionally curved cantilever
embedded in the elastomer should be investigated under a
simulation to obtain the design method for optimal tactile
sensing, while vibrating arms were characterized with active
constrained layer damping [8] and optimum control of multiarm robotic system was investigated in a simulation [9].
In this paper, a fabricated sensor device and its fabrication
process are firstly introduced. Triaxial tactile load sensitivity
by means of resonant frequency shift is then investigated
theoretically by using finite element method on the cantilever
sensor embedded in the elastomer. Three simple and basic
vibration modes are selected in the analysis to easily illustrate the measurement principle with the vibratory cantilever.
Finally the tactile load estimation from the resonant frequency
shift is discussed.

1530437X/$31.00 2013 IEEE

YAMASHITA et al.: PIEZOELECTRIC VIBRATORY-CANTILEVER FORCE SENSORS AND AXIAL SENSITIVITY ANALYSIS

1075

(a)
(f)

(b)

(c)

Fig. 1. Fabricated L-shape cantilever sensor element with 400 + 400-m

long by 50-m wide. (a) Photograph focused on the substrate. (b) Photograph
focused on the tip of the cantilever. (c) Schematic illustration of the cross
section of the layered structure of the root part of the cantilever.

(g)

(d)

(h)

II. S TRUCTURE AND FABRICATION OF S ENSOR D EVICE

A. Structure of the Sensor Device
Figure 1 shows photographs of a fabricated cantilever sensor and a schematic illustration of the cross section of the
layered structure of the cantilever. A lead-zirconate-titanate
(PZT) capacitor is formed on an L-shape SiO2 cantilever
and underneath silicon is removed by deep etching from the
backside. Combination of compressive stress by the SiO2
layer and tensile stress by the PZT layer makes the whole
structure into curved shape when the cantilever is released. The
cantilever is patterned into L-shape on the substrate in 50 m
wide by 400 m long each for the root part and the tip part.
The both parts of the L-shape cantilever curl from the SiO2
layer to the PZT layer and make up the three-dimensionally
curved structure. Figure 1 (a) and (b) show the photograph
focused on the substrate and that focused on the tip of the
cantilever, respectively. These photographs indicate the double
curled structure curves upward from the substrate and then to
the right in 90 each. The curvature radius of the both part is
255 m, which corresponds well to formerly designed value
from the preliminary study for residual stress of the layers [7].
B. Fabrication Process of the Sensor Device
Figure 2 shows the fabrication process of the curved cantilever structure. Note that the substrate is put upside-down
in steps (d), (e) and (g) in Fig. 2 due to the etching process
from the backside of the substrate. The released cantilevers
spontaneously stand up, and the fabrication process has been
designed delicately to prevent the cantilevers from collapsing.
(a) The start substrate is a silicon wafer oxidized thermally
on the both sides up to 1 m thick. (b) Bottom electrode
of platinum and titanium thin films are deposited by using
rf magnetron sputtering, piezoelectric PZT film is deposed
by using conventional sol-gel method, and top electrode of
gold thin film is deposited by using rf magnetron sputtering.
(c) The gold thin film is patterned into the top electrode
by using iodine and potassium iodide solution, the PZT film
is patterned into the cantilever shape and for contact holes

(e)
Fig. 2. Fabrication process of the curved piezoelectric cantilever sensor
structure. The substrate is put upside-down in steps (d), (e), and (g) due to
the etching processes from the backside.

by using nitric acid and hydrofluoric acid, the platinum and

titanium thin films are patterned into the cantilever shape by
using ICP (inductively coupled plasma)-assisted sputter etching with argon plasma, and the SiO2 layer is patterned into the
cantilever shape by using BHF (buffered hydrofluoric acid).
(d) The backside SiO2 layer is patterned to make etching mask
for the following silicon bulk etching. (e) The bulk silicon of
the substrate underneath the cantilever patterns is etched from
the backside vertically by using ICP-RIE (reactive ion etching)
with conventional Bosch process, and the etching is stopped
before the silicon is completely removed. The thickness of the
remained silicon is around 10 m. (f) The front side of the
substrate is covered with photoresist to prevent the cantilevers
from standing up during the following silicon etching. (g) The
remained silicon is completely removed by using normal RIE
with SF6 plasma. Note that at this step the substrate does
not need to be cooled and we can make a gap between the
substrate and the wafer chuck of the etcher. The gap prevents
the cantilevers from collapsing when they stand up partially
from the substrate towards the wafer chuck. (h) The photoresist
is removed by using normal RIE with oxygen plasma, and
the cantilevers completely stand up to be freed from the
photoresist.
III. T RIAXIAL S ENSITIVITY A NALYSIS
Figure 3 illustrates the analytical model; the L-shape curved
PZT/SiO2 -bilayer structure is embedded in PDMS (polydimethylsiloxane) elastomer. The cantilever is fixed at the
root and the elastomer is fixed at the bottom. Tactile loads
Px , Py and Pz are applied on the surface of the elastomer,
and the cantilever is excited by 5 V ac applied on the PZT

1076

IEEE SENSORS JOURNAL, VOL. 13, NO. 3, MARCH 2013

Py
2 mm
PDMS
Elastomer

R
=
PZT
SiO2

Fixed at 50 m
from the bottom
d
Magnifie

ed
ge

(b)

y
z

Relative frequency shift [104]

(c)
(a)

Top view
x

2
3

3rd mode
62.8 kHz
(d)

F(0, Py, 0)

2
1
2nd mode

3rd mode

od
e

1
2
3

F(0, 0, Px)

2
d
2n

de
mo

1st mode

3rd m
ode

2
3
6

(d)

Fig. 4. 3-D curved cantilever and its vibration modes. (a) Birds-eye view
of the L-shape cantilever. (b) Vibration mode at the first resonant frequency.
(c) Vibration mode at the second resonant frequency. (d) Vibration mode at
the third resonant frequency.

Relative frequency shift [104]

xe
d

3rd mode

2nd
mo
de

1st mode
21.5 kHz

z y
x

Fi

F(Px, 0, 0)+F(0, Py, 0)+F(0, 0, Px)

4
2

od
2nd m

0
2

m
5
25

400

(c)

ode

m
1st

(b)

Top view
x
z

2nd mode
45.0 kHz

1s
t

1 mm

Fig. 3. Schematic illustration of an analytical model for the piezoelectric

cantilever embedded in PDMS elastomer.

Front view
x

Fixed at the bottom

3rd m
ode

od

1 mm

F(Px, 0, 0)

tm

m
5
25
5 Vac

(a)

1s

2 mm

2 mm

Relative frequency shift [104]

Px

Relative frequency shift [104]

Pz

4
6
4
2

od
2nd m

3rd m
o

de

tm

od
e

2
1s

(e)

Relative frequency shift [104]

F(Px, Py, Px)

layer. Resonant frequency shift upon a load application was

investigated using finite element analysis software ANSYS
Academic Teaching Introductory version 12. Static analysis
was carried out upon a load application on the elastomer surface followed by frequency response analysis under the stress
caused by the applied load. Vibration damping of the cantilever
by the elastomer was estimated by using the preliminary result
of damping oscillation measurement of a simple cantilevershaped PDMS. The damping coefficient of the PDMS was
measured as 50 s1 and this value was used for the analysis.
Figure 4 shows the birds-eye view of the cantilever and three
vibration modes used for the tactile sensing. The tip part of
the cantilever mainly vibrates at the first mode whereas the
root part mainly vibrates at the second mode, and the both
parts vibrate at the third mode. The combination of the three
modes was selected just because each mode shows a simple
and basic vibration of the L-shape curved cantilever and helps
us with understanding the tactile measurement principle.

4
6

2
0
2
Load Px, Py, Pz [kPa]

Fig. 5. Relative resonant frequency shifts of the three modes versus applied
loads. (a) x-axis, F(Px , 0, 0). (b) y-axis, F(0, Py , 0). (c) z-axis, F(0, 0, Pz ).
(d) Sum of the shifts on the three axes, F(Px , 0, 0)+F(0, Py , 0)+F(0, 0, Pz ).
(e) Triaxial, F(Px , Py , Pz ).

Figure 5 shows the frequency shifts in the three modes

versus the loads Px , Py and Pz . Here the relative frequency

YAMASHITA et al.: PIEZOELECTRIC VIBRATORY-CANTILEVER FORCE SENSORS AND AXIAL SENSITIVITY ANALYSIS

1077

TABLE I
C OEFFICIENTS OF THE R ELATIVE R ESONANT F REQUENCY

TABLE III
S UPERPOSITION E RROR OF THE R ELATIVE F REQUENCY

S HIFT TO A PPLIED L OAD [106 kPa1 ]

S HIFT TO A PPLIED L OAD [%]

F
P
F1
P
F2
P
F3
P

F
Px

F
Py

F
Pz

a1x = 38.1

a1y = 89.9

a1z = 10.2

a2x = 33.1

a2y = 7.4

a2z = 57.2

a3x = 5.6

a3y = 10.6

a3z = 18.5

Mode 1

Mode 2

Mode 3

0.23

4.76

1.50

S HIFT TO A PPLIED L OAD [%]

Mode 2

0.16

9.33

5.95

F(0, Py , 0)

0.04

0.43

8.49

F(0, 0, Pz )

9.60

3.50

6.56

Mode 3

F(Px , 0, 0)
+F(0, Py , 0)
+F(0, 0, Pz )

0.59

10.75

8.28

F(Px , Py , Pz )

0.81

14.78

6.88

shift F is defined as
F(Px , Py , Pz ) =

f (Px , Py , Pz ) f (0, 0, 0)
f (0, 0, 0)

where f (0, 0, 0) and f (Px , Py , Pz ) are resonant frequencies

on application of no load and of a certain load, respectively.
Figure 5 (a)(c) show that each frequency shift depends
somewhat linearly on the load along with each axis. Moreover,
Fig. 5 (d) and (e) indicate that the frequency shifts can
be superposed according to the superposition of the load
components. Thus the forward problem of the system from the
applied load vector P to the relative resonant frequency shift
vector F can be approximately expressed as F = A P using
the conversion coefficient matrix A where their components
are


a1x a1y a1z
Px
F1
F2 = a2x a2y a2z Py .
(1)
F3
a3x a3y a3z
Pz
The components of A are calculated from the forward system
analysis shown in Fig. 5. The calculated values are shown in
Table I. Each component is the incline coefficient in linear
approximation of each graph shown in Fig. 5 (a)(c).
The nonlinearity of the response along with each load axis,
which is the maximum deflection from the linear approximation, is shown in Table II. Totally the second mode shows a
rather large nonlinearity. Each mode show the tendency that
the response shows larger nonlinearity on the load axis parallel
to the vibration direction of the cantilever. The superposition
error, which is the maximum deflection between the sum of
three responses along with each axis and the triaxial response,
is shown in Table III. The superposition error is relatively
small compared to the response nonlinearity. The nonlinearity

4
4
0

]
kPa
Py [

Mode 1
F(Px , 0, 0)

Pz [kPa]

TABLE II
N ONLINEARITY OF THE R ELATIVE F REQUENCY

4
4

0
Px [kPa]

Fig. 6. Load estimation results at 4 and 0 kPa from the relative resonant
frequency shift vector F through the inverse system P = A1 F. Circles
indicate the estimated load vectors by their positions in Px -Py -Pz space. The
dashed curves indicate quadratic interpolation of the estimation results on the
edges.

might be an issue for the estimation of the tactile load from

the resonant frequency shift.
IV. L OAD V ECTOR E STIMATION FROM R ESONANT
F REQUENCY S HIFT OF A S INGLE S ENSOR E LEMENT
Tactile sensing system is realized by obtaining the estimated
input load vector P through solving the inverse problem
P = A1 F

(2)

from the output frequency shift vector F. Figure 6 shows the

load estimation result by the inverse system on the applied
loads at 4 and 0 kPa. The circles indicate the estimated load
vectors by their positions in Px -Py -Pz space. Although the
circles should be located at the cross points of the lattice,
the positions of the circles have rather large deflections with
the maximum estimation error of 33%. This large estimation
error is caused by the nonlinearity of the forward system
described in Section III. Since the condition number of the
system matrix A is 3.76, the estimation error might be up to
the product of the nonlinearity by the condition number [10].
In order to reduce the large estimation errors, the results
are used for
in Fig. 6, the first direct estimation vectors P,
compensation of the nonlinearity. The dashed curves in Fig. 6
indicate quadratic interpolation of the results P on the edges of
the curved, distorted hexahedron in Px -Py -Pz space. We can
interpolate the inside of the hexahedron by using transfinite

interpolation technique [11] with the twelve point data P.

Defining a projection T as the conversion from the lattice
a refined version of the
cube to the distorted hexahedron P,

estimated load vector P is theoretically calculated by

P = T 1 { A1 F}.

(3)

1078

IEEE SENSORS JOURNAL, VOL. 13, NO. 3, MARCH 2013

P001

L0y1

Py [

]
kPa

P101

Pz [kPa]

L1y1

L00z
L11z
Sx1z

P01-

2
4
0

Lx01

P-11 Lx11 P111 P1-1

L01z

P-01
Sxy1

P0-1
P011

P11Sx0z

P00-

S0yz

S1yz

P10L10z

0
Px [kPa]

P010
P0-0

Fig. 7. Load estimation results in the range of 4 kPa for each direction in
2 kPa step. The results are indicated by using refined estimation error vectors
E P from the exact applied load vectors, which correspond the lattice points
in Px -Py -Pz space. The circles indicate the estimated load vectors by their
positions with the exaggerated errors by ten times.

Details of the transfinite interpolation are described in Appendix. Here we evaluate the refined estimation error E P as
E P = A1 F T { P}.

P110

P-10 Lx10

(4)

Figure 7 shows the refined estimation results by the errors E P

with exaggeration by ten times. The distortion is effectively
compensated and the estimation error is drastically reduced.
The maximum estimation error is less than 43 Pa in the load
range of 4 kPa, which corresponds to the load pressure holding
an adult.
V. C ONCLUSION
Vibratory force sensors for individual vector tactile sensing have been fabricated using piezoelectric capacitors on
microcantilever structures for triaxial sensitivity by a single
sensor element. The cantilevers have been formed into threedimensionally curved shape by controlling residual stress
combination of the multilayered structure and the fabrication
process has been delicately designed not to damage the fragile
structure.
Triaxial tactile sensitivity of the cantilever sensor by means
of resonant frequency shift has been analyzed under a load
application onto the surface of an elastomer in which the
cantilever is embedded. Resonant frequency shift of each
mode is investigated upon load applications and the results
show that the frequencies vary somewhat linearly to the three
axial loads and they can be superposed with corresponding
to the superposition of the load components. The applied
load vectors are estimated by resonant frequency shifts of
a single cantilever sensor in three vibration modes, through
a nonlinearity compensation technique using the transfinite
mapping procedure, in an error less than 1.1% to the full scale
of the load 4 kPa.
These results show a promising potentiality and basic
mechanism of the individual vector tactile sensing by the
vibratory cantilever. The superposition of the resonant frequency shift by superposing the tactile load vectors has been

P1-0 L1y0

Sxy0
L0y0
P000

Lx00

P100

P-00

Fig. 8. First direct estimation results P and symbols for the description of
the transfinite interpolation.

well demonstrated. The next issue of the sensing system is

a sophisticated design of the sensor structure including the
elastomer geometry and vibration mode selection to reduce
the nonlinearity of the frequency response to the tactile loads.
A PPENDIX
This appendix describes the transfinite interpolation of the
Figure 8 is the same data points of
first direct estimation P.
P shown in Fig. 6 with symbols for the description. Twenty
points P000 P0-1 correspond the first direct estimation P
including eight corner points,
P000 :
P110 :
P001 :
P111 :

P(4,
4, 4),

P(4,
4, 4),

P(4,
4, 4),

P(4,
4, 4),

P100 :
P010 :
P101 :
P011 :

P(4,
4, 4),

P(4,
4, 4),

P(4,
4, 4),

P(4,
4, 4)

and twelve middle points on the edges, indicated by white

circles,
P-00 :
P-10 :
P00- :
P11- :
P-01 :
P-11 :

P(0,
4, 4),

P(0, 4, 4),

P(4,
4, 0),

P(4,
4, 0),

P(0,
4, 4),

P(0,
4, 4),

P1-0 :
P0-0 :
P10- :
P01- :
P1-1 :
P0-1 :

P(4,
0, 4),

P(4, 0, 4),

P(4,
4, 0),

P(4,
4, 0),

P(4,
0, 4),

P(4,
0, 4).

Twelve edges L x00 L 0y1 are defined by quadratic interpolation of two corner points through one middle point, indicated
by blue dashed curves, as
L x00 (x):
L x10 (x):
L 00z (z):
L 11z (z):
L x01 (x):
L x11 (x):

P000 P-00 P100 ,

P110 P-10 P010 ,
P000 P00- P001 ,
P110 P11- P111 ,
P001 P-01 P101 ,
P011 P-11 P111 ,

L 1y0 (y):
L 0y0 (y):
L 10z (z):
L 01z (z):
L 1y1 (y):
L 0y1 (y):

P100 P1-0 P110 ,

P010 P0-0 P000 ,
P100 P10- P101 ,
P010 P01- P011 ,
P101 P1-1 P111 ,
P001 P0-1 P011 .

YAMASHITA et al.: PIEZOELECTRIC VIBRATORY-CANTILEVER FORCE SENSORS AND AXIAL SENSITIVITY ANALYSIS

Six surfaces Sxy0 Sxy1 are defined by surrounding four

edges as
Sxy0 (x, y) = (1 y)L x00(x) + y L x10 (x)

1079

The projection T of a vector U = (Ux , U y , Uz ) is expressed

using the interpolation V (x, y, z) as
T {U} = V (Ux , U y , Uz ).

(5)

+(1 x)L 0y0(y) + x L 1y0 (y)

(1 x)(1 y) P000 (1 x)y P010
x(1 y) P100 x y P110,
Sxy1 (x, y) = (1 y)L x01(x) + y L x11 (x)
+(1 x)L 0y1(y) + x L 1y1 (y)
(1 x)(1 y) P001 (1 x)y P011
x(1 y) P101 x y P111,
S0yz (y, z) = (1 z)L 0y0 (y) + z L 0y1(y)
+(1 y)L 00z (z) + y L 01z (z)
(1 y)(1 z) P000 (1 y)z P001
y(1 z) P010 yz P011,
S1yz (y, z) = (1 z)L 1y0 (y) + z L 1y1(y)
+(1 y)L 10z (z) + y L 11z (z)
(1 y)(1 z) P100 (1 y)z P101
y(1 z) P110 yz P111,
Sx0z (x, z) = (1 x)L 00z (z) + x L 10z (z)
+(1 z)L x00(x) + z L x01(x)
(1 x)(1 z) P000 (1 x)z P001
x(1 z) P100 x z P101,
Sx1z (x, z) = (1 x)L 01z (z) + x L 11z (z)
+(1 z)L x10(x) + z L x11(x)
(1 x)(1 z) P010 (1 x)z P011
x(1 z) P110 x z P111.
Then the interpolation V (x, y, z) throughout the volume is
finally defined as
V (x, y, z) = (1 z)Sxy0 (x, y) + z Sxy1 (x, y)
+(1 x)S0yz (y, z) + x S1yz (y, z)
+(1 y)Sx0z (x, z) + y Sx1z (x, z)
(1 y)(1 z)Sx0z (x, 0) (1 y)z Sx0z (x, 1)
y(1 z)Sx1z (x, 0) yz Sx1z (x, 1)
(1 x)(1 z)Sxy0(0, y) (1 x)z Sxy0(1, y)
x(1 z)Sxy1(0, y) x z Sxy1(1, y)
(1 x)(1 y)S0yz (0, z) (1 x)y S0yz (1, z)
x(1 y)S1yz (0, z) x y S1yz(1, z)
+(1 x)(1 y)(1 z)S0yz (0, 0)

R EFERENCES
[1] C.-C. Wen, C.-M. Lin, and W. Fang, 3-Axes tactile sensor with tunable
sensing range, in Proc. Trans., Jun. 2007, pp. 14571460.
[2] Y. Tanaka, A. Nakai, E. Iwase, T. Goto, K. Matsumoto, and I. Shimoyama, Triaxial tactile sensor chips with piezoresistive cantilevers
mountable on curved surface, in Proc. APCOT, Jun. 2008, p. 1.
[3] Y. Huang, M. Sohgawa, K. Yamashita, T. Kanashima, M. Noda,
M. Okuyama, and H. Noma, Fabrication and normal/shear stress
responses of tactile sensors of polymer/Si cantilevers embedded in
PDMS and urethane gel elastomers, IEE Jpn. Trans. Sens. Micromach.,
vol. 128, no. 5, pp. 193197, May 2008.
[4] G. M. Krishna and K. Rajanna, Tactile sensor based on piezoelectric
resonance, IEEE Sens. J., vol. 5, no. 5, pp. 691697, Oct. 2004.
[5] K. Motoo, F. Arai, and T. Fukuda, Piezoelectric vibration-type tactile
sensor using elasticity and viscosity change of structure, IEEE Sens. J.,
vol. 7, no. 7, pp. 10441051, Jul. 2007.
[6] K. Yamashita, K. Yoshikawa, T. Kawata, F. Inoue, Y. Teranishi, and
M. Noda, Piezoelectric resonant microcantilever sensors and multi-axis
force detection, in Proc. APCOT, Jul. 2010, pp. 235237.
[7] F. Inoue, K. Yoshikawa, K. Yamashita, and M. Noda, Fabrication of
piezoelectric microcantilever structures for force and tactile sensing, in
Proc. IMFEDK, May 2009, pp. 100101.
[8] E. H. K. Fung and D. T. W. Yau, Vibration characteristics of a rotating
flexible arm with ACLD treatment, J. Sound Vibrat., vol. 269, nos. 12,
pp. 165182, Jan. 2004.
[9] W. S. Owen, E. A. Croft, and B. Benhabib, A multi-arm robotic
system for optimal sculpting, Robot. Comput.-Integr. Manuf., vol. 24,
pp. 92104, 2008.
[10] G. Strang, Linear Algebra and Its Applications, 2nd ed. London, U.K.:
Academic, 1976.
[11] W. J. Gordon and C. A. Hall, Transfinite element methods: Blendingfunction interpolation over arbitrary curved element domains, Numer.
Math., vol. 21, no. 2, pp. 109129, 1973
Kaoru Yamashita (M11) received the B.E. degree in control engineering
from Osaka University, Osaka, Japan, in 1990, where he received the M.E.
degree in control engineering and the D.E. degree in electrical engineering
from the Graduate School of Engineering Science in 1992 and 2002, respectively.
He was an Assistant Professor with Osaka University from 1994 to 2007,
and was involved in research on piezoelectric and ferroelectric thin films,
their preparation technologies and their application to sensors and actuators,
wireless driving and controlling technologies of microrobots using solar cells
and piezoelectric thin films, and wavelength differential imaging technology
for spectral characteristic information detection. He has been an Associate
Professor with the Kyoto Institute of Technology, Kyoto, Japan, since 2007,
where he is involved in research on piezoelectric and ferroelectric thin films
and their applications, especially expanding to ultrasonic and tactile sensors
and sensing systems, including fabrication technologies of three-dimensional
microstructures and new measurement principle using multiple frequencies in
resonance.
Prof. Yamashita is a Senior Member, an Officer of Editorial Affairs with
the board society, a Vice Chairperson of the Editorial Committee, a Secretary
of the Program Committee, a Member of the Steering Committee and the
Chairperson of the Program Committee of Society Conference, and a Secretary
of the Technical Committee of Physical Sensors, Sensors and Micromachines
Society, The Institute of Electrical Engineers of Japan, a member and a
Secretary of The Society of Sensing Technology of Japan, and a member
of the Japan Society of Applied Physics.

+(1 x)(1 y)z S0yz(0, 1)

+(1 x)y(1 z)S0yz (1, 0) + (1 x)yz
S0yz (1, 1)
+x(1 y)(1 z)S1yz (0, 0) + x(1 y)z
S1yz (0, 1)
+x y(1 z)S1yz (1, 0) + x yz S1yz(1, 1).

Yi Yang received the B.E. degree in mechanoelectronic engineering from

Xidian University, Xian, China, in 2009. He is currently pursuing the
Graduate degree with the Department of Electronics, Graduate School of
Science and Technology, Kyoto Institute of Technology, Kyoto, Japan.
He is currently involved in research on tactile and ultrasonic microsensors using piezoelectric thin films on microstructures fabricated using
silicon-micromachining technology, and new measurement techniques for the
microsensors using multiple frequencies in their resonance.

1080

Takanori Nishimoto received the B.E. degree in electronics with research on

structure design and frequency response simulation for piezoelectric vibratory
tactile microsensors from the Department of Electronics, Faculty of Science
and Technology, Kyoto Institute of Technology, Kyoto, Japan, in 2011.
He has been with Oyo Electric Co., Ltd., Kyoto, Japan, since 2011.

Kazuya Furukawa received the B.E. degree in electronics with research

on three-dimensional microfabrication technology for piezoelectric vibratory
tactile microsensors from the Department of Electronics, Faculty of Science
and Technology, Kyoto Institute of Technology, Kyoto, Japan, in 2011.
He has been with The Institute of Science and Industrial Research, Osaka
University, Osaka, Japan, since 2011.

Minoru Noda (M06) received the B.S. degree in electrical engineering from
Osaka University, Osaka, Japan, in 1981, where he received the M.S. and D.E.
degrees in electrical engineering from the Graduate School of Engineering
Science in 1983 and 1993, respectively.

IEEE SENSORS JOURNAL, VOL. 13, NO. 3, MARCH 2013

He was with Mitsubishi Electric Corporation, from 1983 to 1997, where he

was involved in research on process, device, and circuit design of compound
semiconductor ICs in R&D laboratory and division of high-frequency and
high-speed electron device and ICs, including RF front-end transceivers,
prescalers, etc., for mobile handy-phone system, chipset for 10-Gb/s optical
communication, and digital LSIs of static RA and multiplexer/demultiplexer.
He was an Associate Professor with Osaka University from 1997 to 2007,
where he was involved in research on applications of ferroelectric film to
ICs such as infrared array sensor, nonvolatile ferroelectric FET-type memory
device, and microwave and millimeterwave ferroelectric tunable device. He
has been a Professor with the Graduate School of Science and Technology,
Kyoto Institute of Technology, since 2007. His current research interests
include biosensors and biosensing technologies using artificial cell membrane
of liposome.
Prof. Noda was a recipient of the R&D100 Award in 1989 for GaAs highspeed prescaler IC, the Society President Award in 2003 for sensors from
the Society of Sensing Technology of Japan, the Japanese Journal of Applied
Physics Paper Award in 2006, and the APEX/JJAP Editorial Contribution
Award from The Japan Society of Applied Physics in 2010. He has been a
Member of the Editorial Board of Integrated Ferroelectrics: An International
Journal, an Associate Editor of the Japanese Journal of Applied Physics, a
Committee Member of the IEEE EDS Japan Kansaichapter, a member of the
Japan Society of Applied Physics, Institute of Electrical Engineers of Japan,
The Institute of Electronics, Information and Communication Engineers and
Society of Sensing Technology of Japan.