SOI Digital Accelerometer Based on Pull-in Time Configuration

Proceedings of the 2009 4th IEEE International Conference on Nano/Micro Engineered and Molecular Systems

January 5-8, 2009, Shenzhen, China

SOl

Digital Accelerometer Based on Pull-in Time Configuration

Lukasz S. Pakula, Vijayekumar Rajaraman and Patrick

J.

French Electronic Instrumentation Laboratory, Delft University o/Technology, The Netherlands

(1 )

(2) Abstract - We aim to present the principle, design, fabrication

and measurement results of a quasi digital accelerometer fabricated on thin silicon-on-insulator (SOl) substrate. The presented device features quasi-digital output, therefore eliminating the need for analogue signal conditioning. The accelerometer can be directly interfaced to digital electronic circuitry. The measurements showed a pull-in voltage of 2.7V and a pull-in time from 0 to 1G to be 3.2fJs.

Keywords - Thin-SO, MEMS, digital accelerometer

I. INTRODUCTION

Among a range of silicon micromachining technologies, micromachining on thin-Sal substrate provides a good platform for monolithic integration of MEMS devices with the Ie readout circuitry. Moreover, thin-Sal micromachining technology enables chip-scale packaging of small footprint MEMS devices on wafer-level, by sealing the MEMS devices with additional deposited layers. On the other hand, large footprint MEMS structures can readily be packaged using wafer bonding techniques. Hence, thin-Sal micromachining technology can be effectively leveraged to realise low-cost accelerometers for automotive, consumer and medical applications.

It is known that the pull-in voltage and pull-in time are functions of the applied acceleration [1, 2]. Previously an accelerometer using the pull-in time mode was fabricated in aluminium by surface micromachining using two masks. However the reported fabrication process using aluminium as a structural material suffered from several limitations such as poor yield and asymmetry in the pull-in time output due to bending of the structure during the releasing process. Although asymmetry issue could be compensated by the driving circuitry, this approach will increase the complexity. In order to overcome these problems, in this work, a quasi-digital accelerometer with pull-in time-mode configuration [2] has been fabricated on thin-Sal substrate with its sensitive axis in the wafer plane. This fabrication is quite straight forward and uses only one mask simplifying the process.

II. OPERATION PRINCIPLE

The principle of the device is shown in Fig. 1. By applying the pulse voltages

tPI

and

th

to the electrodes 1 and 2 alternately, the mass is pull-in at the stoppers 1 and 2 alternately. T]is the pull-in time from the stopper 1 to 2, andT2

is pull-in tome from the stopper 2 to 1. When there is no acceleration in the x direction, T1=T2=To. If there is an

acceleration in the x direction, the differential pull-in time,

i1T=T2-T}, is proportional to the acceleration. i1T is a

pulse-width-modulated signal and can be measured with a digital

L.S. Pakula, V. Rajaraman and Prof Dr. P.J. French are with the Electronic Instrumentation Laboratory (EI), DIMES, Dept. oj Microelectronics, Faculty EEMCS, Delft University of Technology, The Netherlands (phone: +31-15-27-81602; fax: +31-15-27-85755; e-mail: l.pakula@tudelft·nl).

circuit. The sensitivity and the non-linearity are similar to that ofthe differential capacitive sensing.

~a A.

t

i

r---r1

i

r---n

'1'1 t A.

t-H

I

r--t1

i 'l'2~t Lfu II fu i. t X Pi! W!!

\l:

tiLJtiLl

i

i II iii

V : : : : 1 ,: :, ,: : t

v.

t

i~! i~

2~t

Figure1. The principle of the pull-in operation mode

III. ANALYSIS

The movement of the mass between the stoppers can be described by the equation:

. . . 88 Av2

mx+cx+kx=

(0

y

+ma

2do-x

where a is the acceleration, V the driving voltage, A the area of the electrode, do the initial electrode gap and x the displacement. To keep the device working in the pull-in mode, the driving voltage must be higher than the minimum pull-in voltage and the previous pulse voltage should be held long enough to make sure that the initial condition is

x

=O. To discuss the characteristics in general, a dimensionless equation is preferred. The corresponding dimensionless equation of (1) is

. . . F

x

+

2SX

+

x

= ( ,....,)2

+

a

I-x

where

x

=

x/do,

t

=

dX/d,

,

f

=

d 2

x/

d,2

,

,=

O1ot,

01₀ the circular resonant frequency,

t;

the damping ratio,

F

= &&0

A V

2 /

~kd~

)

the dimensionless driving force and

a

=

mal

kd

0 the dimensionless acceleration. The driving

force must fulfil

F

>

4/27 ,

which corresponds to the minimum pull-in voltage and

a

is always smaller than 1 according to the defmition.

When the damping is zero, the analytical pull-in time can be obtained [2]. The differential pull-in time

Ll,

=

'2 - '1

can be written as:

(3)

where:

3

S=

r

(A+xl~((A-X)+

( ::

)J-2

eli (4)

A I-x 1+.,1

TABLE!. MECHANICAL AND ELECTRICAL DEVICE PARAMETERS

Resonance Frequency 1810 Hz Pull-in Voltage 2.36V Rest Capacitance 165iF Thermal Noise 11.2 Jlg/"Hz Sensitivity 3.2 Jls/g (7) (6) 4 Acceleration [g] LlT[,ls] 20 10 15 -20 -10 -15 -4 -6

The resonant frequency is calculated to be 1.8 kHz. When the displacement is 0, the damping ratio is calculated to be ?104. The pull-inv~lta~eis calculated to be 2.36V. The pull-m voltage of fpull-mgers IS higher than 10V when the pull-mass is at the pull-in position. The static and dynamic properties of the accelerometer in pull-in mode are analysed numerically with Matlab. When the driving voltage is 8V, the pull-in time is

c~lculated to be 77f.!s. Fig.3 shows the differential pull-in time with respect to the acceleration, which was obtained by simulation.

For the device with light damping, the maximum acceleration is larger than (6) due to over-shooting [1]. The maximum acceleration ofthe device without damping is given by:

lal=&-.!.

₂

where,±Ais the positions ofthe stoppers.

It can be obtained from (3) that LlT is proportional to the acceleration. When the acceleration is very large, the structure does not pull-in. The maximum acceleration that can be

meas~edis afun~tionof the operating voltage. To simplify the analysIs, the maXImum acceleration is estimated with the initial conditions

X

=

0

and

i'

=

0 .

When the damping is very high, the maximum acceleration is:

lal<V

2:

F

-1

IV. DEVICE DESIGN

Fig. 2 presents the structure ofthe designed accelerometer.

Figure 2. Accelerometer structure

Based on the above analysis the quasi-digital accelerometer has been designed. The device structure is made on a 4f.!m thick silicon. The mass is supported by four folded beams. The mass is 400f.!mx800f.!m. Each fold of the beam is 2.5f.!mxI50f.!m. Interdigitated electrodes are used to drive the mass. There are 62 fmgers on each side of the mass. Each fmger is 3f.!mxl00f.!m. The fmger-electrode gap is 2f.!m. The mass-stopper gap is 1f.!m. The mass-substrate gap is 400nm. The electrical and mechanical properties have been calculated and presented in Table 1.

Figure 3. Pull-in time vs. acceleration

V. ISSUES RELATEDTo ALMICROMACHINING

Previous fabrication approach using two masks and aluminium as a structural material is explained as follows. At fITst, the If.!m PECVD 4% PSG (4% of phosphorus in PSG) was deposited on oxidised silicon wafer in Novellus system and patterned to form sacrificial layer. Then the 3f.!m low temperature Al was sputtered in Trikon Sigma sputter coater. After lithography the layer was patterned in Trikon Omega 201 plasma etcher, forming the device. Finally the structure was released by removing PSG sacrificial layer in 73%HF (hydrofluoric acid) with addition of iso-propanol. To prevent problem of stiction the freeze drying process was employed after sacrificial step.

Figs. 4 and 5 show the SEM images of the fully released Al device. However this fabrication approach suffers from several limitations such as poor yield and asymmetry in the pull-in time output due to bending of the structure during the releasing process, Fig. 6. The corresponding measured CV curves are presented in Fig. 7. Although this asymmetry could be compensated by the driving circuitry, this approach will increase the complexity. Besides, micromachining using Al as

4

2

6

8

10

12

14

Voltage (V)

Figure 7. Asymmetry in pull-in voltage and measured capacitance due to bending problem

~

0.28

~

50.26

Q) Q

§

0.24

~ .(3

a

0.22

~

u

0.20

Figure 8. Thin-Sal MEMS Fabrication Process

VI. THIN-SOIFABRICATION

In order to overcome these problems, in this work, the accelerometer was fabricated on a thin-Sal substrate. This post-IC fabrication is quite straight forward and is illustrated in Fig. 8. This process uses a single-mask and a lOOmm,400Jlm

thick sal substrate as the starting material (Fig. 8a). The active silicon layer is4Jlmthick and the buried oxide thickness is 400nm. The initial step starts with the deposition of photoresist mask followed by lithographic patterning of the microaccelerometer structure (Fig. 8b). The active silicon layer is patterned using RIE, stopping on the buried oxide (Fig. 8c). In this step, the proof-mass is also perforated to assist the releasing process. The buried oxide layer is then sacrificially etched using a HF-based etchant to release the microstructure (Fig. 8d). This step is immediately followed by a Cyclohexane-based freeze-drying process in order to avoid stiction of the released microstructures. Finally, a thin layer of aluminium is sputtered onto the bond pads (Fig. 8e), completing the device processing.

Figure 6. Al structure bent during release process causing the asymmetry in the output Figure 5. Properly released Al structure Figure 4. Fully released Al quasi-digital accelerometer

a structural material also restricts the thickness of the

proof-mass which oses limitation on the achievable sensitivi .

0.30

The corresponding measured CV curves are presented in Fig. 6. Although this asymmetry could be compensated by the driving circuitry, this approach will increase the complexity. Besides, micromachining using Al as a structural material also restricts the thickness of the proof-mass which poses limitation on the achievable sensitivity.

VII. RESULTS

Fig. 9 presents the SEM image of a free-standing SOl microstructure. A detailed SEM image of a fully released quasi-digital accelerometer is presented in Fig. 10 that shows the actuating comb-drives, perforated proof-mass, the spring and the stopper.

Fig. 11 presents top view of fabricated device which was measured. The capacitance-voltage results are shown in Fig. 12. The pull-in voltage was around 2.7 V which can be explained due to small undercut occurring during the RIE etching (Fig 9)

Figure 9. SEM image of a freestanding thin-SOl layer

245 Li:' 225 :!:. QI u c 205

"'

+' 'u

"'

c. 185

"'

u 165 145

a

0.5 1.5 Voltage [V] Figure 12. CV measurement 2.5 3.5

Figure 10. SEM image of the quasi-digital accelerometer

~.

Figure Photo of the quasi-digital accelerometer

CONCLUSIONS

The principle, design, fabrication and measurement results of a thin SOl quasi digital accelerometer fabricated were presented. The presented device features quasi-digital output, therefore eliminating the need for analogue signal conditioning circuitry. The use of thin-SOl substrate overcomes the limitations of the Al surface micromachining. The pull-in voltage was 2.7 V. The pull-in time from 0 to IG was 3.2f.!s.

ACKNOWLEDGMENT

The authors wish to thank the IC processing group of DIMES for technical assistance.

REFERENCES

[1] W. C. Tang, Digital Capacitive Accelerometer, US Patent 5353641, 1994.

[2] H. Yang et al. A novel operation mode for accelerometers, Pacific rim workshop on transducers and micro/nano technologies, July 22-x24, 2002, Xiamen, China, pp.303-306.

[3] R. E. Ziemer, W. H. Tranter, Principles of communications systems, modulation and noise. Houghton Mifflin Company, 1985.