I

NSTITUTE OF

P

HYSICS

P

UBLISHING

J

OURNAL OF

M

ICROMECHANICS AND

M

ICROENGINEERING

J. Micromech. Microeng. 11 (2001) 1–6

www.iop.org/Journals/jm

PII: S0960-1317(01)12880-9

Fabrication of an electrostatic
track-following micro actuator for hard
disk drives using SOI wafer

Bong-Hwan Kim and Kukjin Chun

Inter-university Semiconductor Research Center (ISRC) and
School of Electrical Engineering, Seoul National University, SOEE #038,
Kwanak PO Box 34, Seoul 151-742, Korea

E-mail: [e-mail]

Received 28 March 2000

Abstract
This paper presents track-following control using an electrostatic
microactuator for super-high density hard disk drives (HDDs). The
electrostatic microactuator, a high aspect ratio track-following microactuator
(TFMA) which is capable of driving 0

.3 µg magnetic head for HDDs, is

designed and fabricated by a microelectromechanical systems process. It
was fabricated on a silicon on insulator wafer with a 20

µm thick active

silicon layer and a 2

µm thick thermally grown silicon dioxide layer; a

piggyback electrostatic principle was used for driving the TFMA. The first
vibration mode frequency of the TFMA was 18.5 kHz, which is enough for a
recording density of higher than 10 Gb in

−2

. Its displacement was 1

.4 µm

when a 15 V dc bias plus a 15 V ac sinusoidal driving input was applied and
its electrostatic force was 50

.4 µN when the input voltage was 30.7 V. A

track-following feedback controller is designed using a feedback nonlinear
compensator, which is derived from the feedforward nonlinear compensator.
The fabricated actuator shows 7.51 dB of gain margin and 50

.98

◦

of phase

margin for a 2.21 kHz servo bandwidth.

(Some figures in this article are in colour only in the electronic version; see

www.iop.org)

1. Introduction

Conventionally, most hard disk drives (HDDs) have the
characteristics of a recording density of a few gigabits per
square inch, a track density of 5 kTPI (track per inch)
and a servo bandwidth of 500–600 Hz. These are realized
by the technology of a 4

µm track width and ±0.5 µm

tracking accuracy. HDDs commonly utilize a voice coil motor
(VCM) for track seeking and track following to satisfy these
characteristics. In the future, however, the specifications for
HDDs will be a recording density of 10 Gb in

−2

, a track density

of 25 kTPI and the servo bandwidth of 2 kHz, which requires
a 1

µm track width, ±0.1 µm tracking accuracy and a greater

than 2 kHz servo bandwidth [1–4].

Since conventional servo actuators cannot provide this

level of track accuracy, a new microactuator to realize a
system with the recording density of 10 Gb in

−2

is needed.

Several types of microactuator are being developed based on
electrostatic [2, 3, 5], electromagnetic [6, 7], piezoelectric [8],

thermal and shape memory principles.

Among these, the

electrostatic, electromagnetic and piezoelectric types are
frequently used as microactuators for HDDs. Most recently
Naniwa et al [9] reported the comparisons of piggyback
actuators.

A disadvantage of the electromagnetic and

piezoelectric types is that they locate the actuator far from the
read/write elements and therefore have limited bandwidth due
to suspension vibration. The electrostatic actuator proposed in
this paper can be assembled very near to the head, increasing
the bandwidth for servo control.

However, almost all

electrostatic microactuators require higher driving voltage than
electromagnetic microactuators and are troublesome because
of additional fabrication processes of the head. However, the
proposed microactuator only needs an easy fabrication process
using a silicon on insulator (SOI) wafer. The actuator presented
in this paper differs from other electrostatic actuators which use
copper [2] or nickel [5, 10] as the structural layer.

In order to achieve fast tracking for a track pitch below

1

µm, an additional microactuator is needed for a conventional

0960-1317/01/010001+06$30.00

© 2001 IOP Publishing Ltd

Printed in the UK

1

B-H Kim and K Chun

Microactuator & R/W Head

Disk

Slider

Suspension

V C M

VCM : Track-Seeking
Microactuator : Track-Following

Track-Seeking

(a)

Track-Fo

llowing

Microactuator

Slider

Protection

Cover

R/W Head Plate

R / W H e a d

(b)

Figure 1. Design concept of dual-stage servo system and
head/slider/TFMA assembly. (a) Dual-stage servo system.
(b) Head-assembly.

microactuator such as VCM. Therefore, a dual-stage servo
system [1] is essential for the fast tracking of the suspension,
which is composed of a VCM and a microactuator.

In

the dual-stage servo system, a track-following microactuator
(TFMA) is used as a fine and high-bandwidth actuator while
a conventional VCM is used as a coarse and low-bandwidth
actuator. Figure 1(a) shows the design concept of the dual-
stage servo system and the head/slider/TFMA assembly. The
assembly consists of a slider, a TFMA, a head and a protection
cover. Fusion bonding is applied to attach the TFMA on the top
of the slider. Then, a magnetic read/write head is integrated on
the head plate of the fabricated TFMA. After processing for a
slider, a TFMA and a read/write head is ended, fusion bonding
is also used between the slider and the protection cover as
shown in figure 1(b).

2. Design and fabrication

2.1. Design

In order to implement high bandwidth, high flexure stiffness
and high driving force, a SOI wafer and a comb-drive-type
principle were used. The parameters that should be considered
in the design are the natural frequency, easy fabrication,
operating stability, cost and the protection from data damage.
The specifications of the TFMA are defined as follows.

• Size of microactuator: 1000 µm×300 µm.

• Area of head plate: 300 µm×100 µm.

• Applied voltage for 1 µm stroke: less than 20 V.

• First natural frequency of microactuator: greater than

10 kHz.

For a 2 kHz servo bandwidth, the natural frequency of the

system should be greater than 15 kHz, resulting in a comb drive
of a 2

µm flexure width, 3 µm comb width and 2.2 µm finger

gap. The flexure stiffness was calculated by a finite-element

l

b

l

f

g

1

g

3

g

2

y

Figure 2. The schematic of the TFMA.

method (FEM) using ANSYS software and the structure was
designed to be symmetric to reduce any residual stress. The
flexure stiffness and natural frequency of the TFMA were
calculated from the following equations:

K ∝ EI

(1)

I =

1

12

(width)

3

(thickness)

(2)

w

n

=

K

M

(3)

where

K is the flexure stiffness, E is Young’s modulus of

silicon and

M is the mass of the moving part of a TFMA.

To drive a 1

µm displacement within a 20 V applied voltage,

the flexure and finger gaps were defined using a MATLAB
simulation. As shown in figure 2, we derived an equation to
calculate electrostatic force of the TFMA as follows:

F

es

=

1

2

n

1

ε

0

V

2

t

g

1

+

1

2

n

2

ε

0

V

2

tl

f

(g

2

− y)

2

−

1

2

n

3

ε

0

V

2

tl

b

(g

3

+

y)

2

(4)

where

V is the applied voltage, ε

0

is the permittivity of free

space,

t is the thickness of the TFMA, n

1

,

n

2

and

n

3

are the

number of finger pairs,

l

f

and

l

b

are the length of the finger

and finger bar,

y is the displacement, g

1

,

g

2

and

g

3

are lengths

(see figure 2 for details). The values of

g

1

,

g

2

,

g

3

,

l

b

and

l

f

,

which are results of the MATLAB simulation, are 2.2–3

µm,

7

µm, 8 µm, 135 µm and 3 µm, respectively.

2.2. Fabrication

A TFMA with a metal signal line to access the recording was
fabricated with the following procedure, as shown in figure 3.
The starting material is a 100 mm diameter SOI wafer with
a 20

µm thick active silicon layer and a 2 µm thick silicon

dioxide intermediate layer on a 525

µm thick silicon substrate.

In order to simplify the fabrication process, we used only
three masks: a silicon nitride mask, a metal mask and a CVD
oxide mask. First, we doped the active silicon at 1000

◦

C

for 60 min after a pre-deposition with POCl

3

at 1000

◦

C for

30 min. The junction depth was 2

µm and the sheet resistance

was 2

.6 / . Silicon nitride, 2500 Å thick, was deposited by

low-pressure CVD and was patterned and etched by reactive
ion etching for the contact between metal and silicon. A 1

µm

thick molybdenum (Mo) layer was sputtered and patterned.
A 6

µm thick tetraethylorthosilicate (TEOS) deposited by

2

Fabrication of an electrostatic TFMA

Figure 3. Process flow of the TFMA.

plasma-enhanced CVD was used as a hard mask layer to deep
etch the silicon deep. This process used only one mask to
etch 1

µm Mo and 20 µm silicon. Finally, sacrificial oxide

was etched in 7:1 BHF. Figure 4(a) shows a SEM image of
the fabricated TFMA with 2

µm wide flexures, 3 µm combs

and 2

.2 µm gap. The aspect ratio is 10:1 for the silicon depth

versus width and the anisotropy is 0.997, which is adequate for
many applications.

To fabricate the integrated signal line of a TFMA,

sequential

etching

is

an

important

fabrication

issue.

We developed the process to etch a 1

µm thick Mo and

20

µm thick silicon membrane using the only the TEOS hard

mask. Another significant issue is the stiction in the actuator.
We developed a new anti-stiction coating procedure and used
it for this structure [11]. Figures 4(b) and (c) shows SEM
images of the TFMA with an integrated Mo signal line. This
figure indicates that an integrated Mo signal line of TFMA was
successfully made.

3. Results and discussion

3.1. Experiment

The characteristics of the TFMA were measured using the
measurement apparatus as shown in figure 5. The apparatus
consisted of a dynamic signal analyzer (HP 35670A), a
voltage amplifier (HP 6826A), a digital oscilloscope (Tektronix
TDS 754A), a fiber interferometer (Polytec OFV 512), a micro
spot head (Polytec OFV 130) and a vibrometer controller
(Polytec OFV 3001).

The first natural frequency of the TFMA is 18.5 kHz

and its displacement is 1

.4 µm when a 15 V dc bias plus

a 15 V ac sinusoidal driving input is applied. In particular,
its electrostatic force is 50

.4 µN when the input voltage is

30.7 V. The electrostatic force is calculated using equation (4)
and the measured parameters in table 1. Figure 6 shows the
frequency response of the TFMA with 2

.2 µm finger gaps,

2

.2 µm wide flexures and 3 µm wide combs. Figure 7

shows the driving voltage against the electrostatic force and
the displacement characteristics for the same dimensions as in
figure 6. When the voltage is applied to the microactuator,

(a)

stator

stator

stator

spring

spring

spring

spring

signal

line

+ y

R / W

h e a d

plate

m o v e r

(b)

stator

m o v e r

t

g

3

g

2

g

1

l

f

l

b

(c)

Figure 4. Fabricated microactuator. (a) SEM image of the
fabricated TFMA with 2

µm wide flexures, 3 µm wide combs and

2

.2 µm comb gaps. (b) SEM image of the microactuator. (c) Design

parameters of the microactuator.

Digital

Oscilloscope

Digital Signal

Processor

Monitor

Vibrometer Controller

Dynamic Signal Analyzer

Voltage Amplifier

Optical Table

system

output

system

input

source

out

amplified voltage

to

F

A

B

C

D

E

F

G

A

B

C

D

E

F

G

: CCD Camera

: Fiber Interferometer

: Micro Spot Head

: Laser Vibrometer Stand

: Microscope

: XY Stage

: Probe

: Microactuator Wafer

Figure 5. Experimental apparatus for microactuator measurements.

an electrostatic force results in the measured nonlinear part of
the microactuator. The modeling of the nonlinear part where
the displacement (

y) is measured by a laser vibrometer and

3

B-H Kim and K Chun

Table 1. Measured parameter values.

ε

0

8

.854 × 10

−12

F m

−1

g

1

1

.82 µm

t

20

µm

g

2

4

.55 µm

n

1

800

g

3

7

.01 µm

n

2

832

l

f

3

.34 µm

n

3

30

l

b

134

µm

Figure 6. Frequency response of the TFMA: 2

.2 µm finger gaps,

2

.2 µm wide flexures and 3 µm wide combs.

Figure 7. Characteristic curve of voltage against electrostatic force
and displacement: 2

.2 µm finger gaps, 2.2 µm wide flexures and

3

µm wide combs.

theflexure stiffness is calculated using FEM after measuring the
real value by SEM. The modeling and experiment values of the
electrostatic force of the microactuator are almost the same.

3.2. Micromachining using SOI

Recently, as surface micromachining has appeared to realize
mechanical structures, commercially available SOI wafers
have been attractive for smart power sensors and actuators,
as the etch component is dielectrically isolated [12, 13].
In particular, moving elements such as the cantilever,
microbridge, micropump, microvalve and diaphragm can
be fabricated by SOI. Many various SOI techniques for
microelectromechanical systems (MEMS) are shown in
table 2. The silicon dioxide layer of SOI for sensors and
actuators is used as the etch stop, for dielectric isolation and
as a sacrificial layer. In general, SOI wafer can be classified
into three groups concerning the fabrication process [14].

One is SIMOX (separation by implanted oxygen) [15]. A
buried oxide layer is formed by high dose (

∼10

18

cm

−2

)

oxygen implantation followed by high-temperature (1300

◦

C)

annealing. Buried oxide and thin-film silicon have already
been used as piezoresistors and high-temperature silicon
sensors [16].

Another fabrication process is ZMR (zone

melting recrystallization). A ZMR structure is fabricated by
the deposition of a polysilicon film on an oxidized Si wafer
[17]. The final process is bonded SOI [18]: SDB (silicon direct
bonding), SFB (silicon fusion bonding) and DWB (direct wafer
bonding) all have the same meaning in the SOI techniques.
Wafer bonding using thermal oxide was initially proposed
by Lasky in 1986 [19]. The basic process sequence for the
fabrication of the bonded SOI wafer is as follows.

First,

an active wafer is bonded to the handle wafer, which has a
layer of thermally grown silicon dioxide upon it. Then, the
bonded pair wafer is annealed at a high temperature of about
1100

◦

C to increase the bonding strength and to remove the

interfacial voids. In this paper, we used bonded SOI wafers
for fabrication because the bonded SOI technique can freely
change the thickness of silicon dioxide.

Most of the SOI applications are CMOS and SOI–MEMS

devices with thin active silicon layers and less than 1

µm thick

silicon dioxide layers. However, Moore et al [20] showed an
accelerometer with a 3

µm Si layer on a 4 µm SiO

2

layer on a

thick 400

µm Si substrate. The accelerometer was fabricated

by HF and KOH wet etching. Timothy et al [21] showed a
new technique for providing both electrical isolation and an
embedded interconnect to a SOI based inertial sensor. This
process is similar to the SCREAM process [22].

3.3. Sacrificial oxide etching with metal line

Since Nathanson et al [30] etched sacrificial oxide to fabricate
a free-standing gold microbridge, used as a resonant gate
transistor, silicon dioxide sacrificial layer etching has become
a major surface micromachining method of fabrication of
microsensors and microactuators, which are often made of
polysilicon. The various wet etchants, such as HF (typically
50%), BHF, NH

4

F/HF solutions and HNO

3

/HF solutions have

been used to etch silicon dioxide. In particular, high selectivity
between the Al and the sacrificial oxide is crucial for the
mechanical, electrical or optical properties of the MEMS
structures because Al is widely used as a metal signal line
in the IC industry [31]. Gennissen [32] reported sacrificial
oxide etching compatible with Al metallization using 73%
HF and several mixed HF solutions. For most of the CMOS
compatible processes, Al is a useful material because of easy
sputtering, good quality and appropriate resistivity. Therefore,
the characteristics of the Al are reported in many papers [31–
34], but Al is also easily attacked by HF solutions. Although
some researchers showed that pad etch had a higher selectivity
to Al when compared to standard BHF [34], this etchant has
lower etch rate to sacrificial oxide than HF solutions.

To overcome this limitation, we had to find some kind

of material for the metal line on the TFMA because our
structure required longer etch times. In this paper, since it
was also critical that metal signal line should be passivated
during sacrificial oxide etching, several metals such as Au, Al,
Mo and Ti were recommended. All of these were good as

4

Fabrication of an electrostatic TFMA

Table 2. Comparison of SOI techniques.

Thickness of SOI

SOI

SiO

2

Application

type

Active Si

SiO

2

etching

Reference

Pressure sensor

SIMOX

0

.2 µm

0

.4 µm

[23]

Pressure sensor

SIMOX

0

.2 µm

0.37–0

.4 µm

[24]

Accelerometer

SDB

3

µm

4

µm

HF & H

2

O

[20]

Inertial sensor

SDB

45

µm

1

µm

HF

[21]

Microactive probe

SDB

0

.2 µm

1

.5 µm

BHF

[25]

Resonator

SDB

Various

1

µm

BHF

[26]

thicknesses

Optical chopper

No data

1

µm

1

µm

50% HF

[27]

Tunneling probe

No data

20–50

µm

1

µm

HF

[28]

Capacitive accelerometer

SDB

No data

No data

No data

[29]

TFMA (Our work)

SDB

20

µm

2

µm

BHF

Table 3. Etch rates of several solutions [Å/min.].

Wet oxide

TEOS

PSG

Nitride

Al

Mo

HF (49%)

17 625

39 690

47 784

148.8

383.3

1.5

7:1 BHF

1326

1068

10 242

10.2

30

5

10:1 HF

484

1572

9216

15

3200

1.5

Solution 1

a

890

1861

13 746

7.8

9

3.3

Solution 2

b

426

1520

39 624

28.8

412.5

0.7

Solution 3

c

168

483

2833

2.7

65.5

0.2

a

Solution 1 NH

4

F(40%):HF(49%):glycerin(C

3

H

8

O

3

) = 4:1:2.

b

Solution 2 HNO

3

:HF(49%):glycerin

= 1:5:10.

c

Solution 3 NH

4

F:CH

3

COOH:C

2

HO

2

:deionized water

= 13.5:31.8:4.2:

50.5.

a mask for etching. Considering only resistivity, Au is very
good, but Au is difficult to dry etch and fine pattern. Ti is
attacked easily during sacrificial oxide etching by HF solutions.
Compared with Al, Mo is a more durable material to HF and
BHF solutions, and has the better adhesion to silicon or silicon
nitride. This is why we selected Mo as a metal signal line.
Table 3 shows the etch rates for several solutions in Ångstroms
per minute.

4. Conclusions

In this paper, we have realized electrostatic comb-drive type
microactuators using bonded SOI wafers.

To achieve a

recording density of more than 10 Gb in

2

, the microactuator

should have a 1

µm track width and a 0.1 µm tracking accuracy.

Using the simplified process, the integrated signal line TFMA
was fabricated by only three masks. As a metal signal line,
Mo is the best choice because of good durability and good
adhesion on silicon nitride. In order to release the structure
layer on thermally grown silicon dioxide, 7:1 BHF and a
new anti-stiction coating method is utilized. The measured
first natural frequency is 18.5 kHz. Excellent track-following
control performance gave a 2.21 kHz servo bandwidth, 7.51 dB
gain margin and a 50

.98

◦

phase margin. The electrostatic force

of the TFMA is 50

.4 µN at 30.7 V.

Acknowledgments

This research was supported by the Ministry of Science and
Technology and the Ministry of Industry and Energy under

the Micromachine Technology Development Program.

In

addition, this work was partly funded by the Ministry of
Education. The authors would like to thank Sangjun Park,
Hyeon-Cheol Kim, Jong-Won Lee, Seung-Han Kim, Hyo-
Jung Lee, Woo-Kyeong Seong, as well as Professor Dong-Il
(Dan) Cho for their invaluable help.

References

[1] Hirano T et al 1999 Micro-actuator for tera-storage Proc.

MEMS ’99 (Orlando, FL) pp 441–6

[2] Fan L-S, Ottesen H H, Reley T C and Wood R W 1995

Magnetic recording head positioning at very high track
densities using a microactuator-based, two-stage servo
system IEEE Trans. Indust. Electron. 42 222–33

[3] Horsely D A, Cohn M B, Singh A, Horowitz R and Pisano A P

1998 Design and fabrication of an angular microactuator for
magnetic disk drives J. MEMS 7 141–8

[4] Fan L-S, Lane L H, Robertson N, Crawforth L, Moser M A,

Reiley T C and Imaino W 1993 Batched-fabricated
milli-actuators Proc. IEEE MEMS Workshop (Fort
Lauderdale, FL) pp 179–83

[5] Hirano T, Fan L-S, Gao J Q and Lee W Y 1998 MEMS

milliactuator for hard-disk-drive tracking servo J. MEMS 7
149–55

[6] Takaishi K et al 1996 Microactuator control for disk drive

IEEE Trans. Magn. 32 1863–6

[7] Koganezawa S et al 1997 Development of an integrated

piggyback milli-actuator for high density magnetic
recording Int. Conf. on Micromechatronics for Information
and Precision Equipment (Tokyo, Japan)

[8] Mori K, Munemoto T, Otsuki H, Yamaguchi Y and Akagi K

1991 A dual-stage magnetic disk actuator using a
piezoelectric device for a high track density IEEE Trans.
Magn. 27 5298–300

5

B-H Kim and K Chun

[9] Naniwa I et al 1999 Low voltage driven piggy-back actuator of

hard disk drives Proc. MEMS’99 (Orlando, FL) pp 49–52

[10] Hirano T, Furuhatha T, Gabriel K T and Fujita H 1992 Design,

fabrication and operation of submicron gap comb-drive
microactuator J. MEMS 1 52–9

[11] Kim B H et al 1999 A new class of surface modifiers for

stiction reduction Proc. MEMS’99 (Orando, FL) pp 189–93

[12] Auberton-Herve A J 1996 SOI: materials to systems IEDM’96

pp 3–10

[13] Usenko A Y et al 1999 Silicon-on-insulator technology for

microelectromechanical applications Semiconductor Phys.
Quantum Electron. Optoelectron. 1 93–7

[14] Kanda Y 1991 What kinds of SOI wafers are suitable for what

micromachining purposes TRANSDUCERS ’91 pp 452–5

[15] Izumi K 1989 Application of SIMOX technology to device

isolation OYO BUTURI 58 1202–11

[16] Diem B 1990 SIMOX: a technology for high temperature

silicon sensors Sensors Actuators A 21–23 1003–6

[17] Zavracky P M et al 1988 Large diameter SOI wafers by

zone-melting-recrystallization Mater. Res. Soc. Symp. Proc.
107 213–19

[18] Abe T et al 1990 Wafer bonding technique for

silicon-on-insulator technology Solid State Technol. 33
39–40

[19] Lasky J B 1986 Appl. Phys. Lett. 48 76
[20] Moore D F et al 1996 Silicon-on-insulator material for sensors

and accelerometers Silicon Fabricated Instruments (Digest
No:1996/227), IEE Colloquium on 9/1–9/5

[21] Timothy J et al 1997 Embedded interconnect and electrical

isolation for high-aspect-ratio, SOI inertial instruments
Transducers ’97 (Chicago, IL) pp 637–40

[22] Shaw K A et al 1996 Integrating SCREAM micromachined

devices with integrated circuits MEMS’96 pp 44–8

[23] Diem B et al 1995 SOI ‘SIMOX’; from bulk to surface

micromachining, a new age for silicon sensors and actuators
Sensors Actuators A 46–47 8–16

[24] Ziermann R et al 1997 A high temperature pressure with

β-SiC piezoresistors on SOI substrates Transducers’97
(Chicago, IL) pp 1411–14

[25] Yi Y W et al 1997 A micro active probe device compatible

with SOI-CMOS technologies J. MEMS 6 242–8

[26] Noworolski J M et al 1995 Fabrication of SOI wafer with

buried cavities using silicon fusion bonding and
electrochemical etchback Transducers ’95 pp 71–4

[27] Tabata O et al Electrostatic driven optical chopper using SOI

wafer 7th Int. Conf. on Solid-State Sensors and Actuators
pp 124–7

[28] Toshiyoshi H et al 1999 Micromechanical tunneling probe and

actuators on a silicon chip Microprocesses and
Nanotechnology ’99 pp 180–1

[29] Yamamoto T et al 2000 Capacitive accelerometer with high

aspect ratio single crystalline silicon microstructure using
SOI structure with polysilicon-based interconnect technique
MEMS 2000 pp 514–19

[30] Nathanson H C et al 1967 The resonant gate transistor IEEE

Trans. Electron. Devices 14 117–33

[31] Buhler J et al 1997 Review article: Silicon dioxide sacrificial

layer etching in surface micromachining J. Micromech.
Microeng. 7 R1–13

[32] Gennissen P T J et al 1997 Sacrificial oxide etching

compatible with aluminum metallization Transducers’97
(Chicago, IL) pp 225–8

[33] Goosen J F L et al 1997 Problems of sacrificial etching in the

presence of aluminum interconnect Sensors Actuators A 62
692–7

[34] Buhler J et al 1997 Electrostatic aluminum micromirrors using

double-pass metallization J. MEMS 6 126–35

6