Journal of Micromechanics and

Microengineering

PAPER You may also like

- Multiferroic BaTiO3–BiFeO3 composite thin
MEMS magnetic field sensor based on films and multilayers: strain engineering
and magnetoelectric coupling
magnetoelectric composites Michael Lorenz, Vera Lazenka, Peter
Schwinkendorf et al.

- Magnetoelectric coupling in iron oxide

To cite this article: Stephan Marauska et al 2012 J. Micromech. Microeng. 22 065024 nanoparticle—barium titanate composites
Li-Ming Wang, Oleg Petracic, Stefan
Mattauch et al.

- A review of wireless power transfer using

magnetoelectric structures
View the article online for updates and enhancements. Orpita Saha, Binh Duc Truong and Shad
Roundy

This content was downloaded from IP address 193.231.11.68 on 04/10/2024 at 11:00

IOP PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 22 (2012) 065024 (6pp) doi:10.1088/0960-1317/22/6/065024

MEMS magnetic field sensor based on

magnetoelectric composites
Stephan Marauska 1,4 , Robert Jahns 2 , Henry Greve 3 , Eckhard Quandt 3 ,
Reinhard Knöchel 2 and Bernhard Wagner 1
1
Microsystems Technology, Fraunhofer Institute for Silicon Technology ISIT, Fraunhoferstrasse 1,
D-25524 Itzehoe, Germany
2
High-frequency Laboratory, Institute of Electrical Engineering, University of Kiel, Kaiserstrasse 2,
D-24143 Kiel, Germany
3
Inorganic Functional Materials, Institute for Materials Science, University of Kiel, Kaiserstrasse 2,
D-24143 Kiel, Germany
E-mail: [email protected]

Received 18 January 2012, in final form 20 April 2012

Published 22 May 2012
Online at stacks.iop.org/JMM/22/065024

Abstract
For the measurement of biomagnetic signals in the pico- and femtotesla regime
superconducting interference devices (SQUIDs) are commonly used. Their major limitation
comes from helium cooling which makes these sensors bulky and expensive. We show that
MEMS sensors based on magnetoelectric (ME) composites could be capable as a replacement
for biomagnetic measurements. Using surface micromachining processes a cantilever beam
with a stack composed of SiO2/Ti/Pt/AlN/Cr/FeCoSiB was fabricated on a 150 mm Si
(1 0 0) wafer. First measurements of a rectangular micro cantilever with a thickness of
4 μm and lateral dimensions of 0.2 mm × 1.12 mm showed a giant ME coefficient
α ME = 1000 (V m−1)/(A m−1) in resonance at 2.4 kHz. The resulting static ME coefficient is
α ME = 14 (V m−1)/(A m−1). In resonance operation a sensitivity of 780 V T−1 and noise levels
as low as 100 pT Hz−1/2 have been reached.
(Some figures may appear in colour only in the online journal)

1. Introduction integration of ME composites in microelectromechanical

systems (MEMS) promises miniaturized low-cost sensors with
Magnetoelectric (ME) composites have ME coefficients which high sensitivity and high spatial resolution. In the following
are several magnitudes higher than natural multiferroics the fabrication and characterization of a magnetoelectric AlN-
[1–3]. Due to the enhancement of detecting magnetic fields (Fe90Co10)78Si12B10 MEMS sensor will be presented.
at mechanical resonance and their passive nature they
have a high potential as ultrasensitive ac magnetic field 2. Experimental methods
sensors in the picotesla regime [2–8]. Of high interest
is the application for noninvasive medical imaging like 2.1. Sample fabrication
magneto-encephalography or cardiography (MEG, MCG)
Using surface micromachining processes a cantilever beam
as a replacement for state-of-the-art sensors based on
with a stack of SiO2/Ti/Pt/AlN/Cr/FeCoSiB, was fabricated
superconducting interference devices. Despite the high
on 150 mm silicon wafers. A schematic cross section and a
sensitivity of about 10−15 Tesla Hz−1/2 for detecting magnetic
photograph of the sensor device are shown in figure 1. The
fields, SQUIDs have to operate at a low temperature of
process flow is given in figure 2.
4.2 K [9]. Therefore the use of these bulky and expensive
In the first step silicon (1 0 0)-oriented wafers were
sensors especially for sensor arrays is strongly limited. The
thermally oxidized followed by coatings of Ti and Pt films
which were structured by a lift-off process to define the bottom
4 Author to whom any correspondence should be addressed. electrode (steps 2 and 3 in figure 2). Subsequently, AlN and Mo

0960-1317/12/065024+06$33.00 1 © 2012 IOP Publishing Ltd Printed in the UK & the USA
J. Micromech. Microeng. 22 (2012) 065024 S Marauska et al

(a) (Fe90Co10)78Si12B10 with a thickness of about 2.2 μm under

the conditions: 2 × 10−6 bar Ar and 200 W rf power. During
deposition of the magnetostrictive layer a bias field of 10 mT
was applied in the film plane perpendicular to the long axis of
the cantilevers to induce an in-plane anisotropy.
Mo was patterned via photo resist using a phosphoric acid-
based solution (PWS) of 80% H3PO4, 16% H2O and 4% HNO3.
(b)
The AlN film was etched with 25 wt% TMAH heated at 82 ◦ C.
For stripping the Mo and etching the Cr a 50 wt% cerium (IV)
ammonium nitrate (CAN) solution was used. Au was etched
with a solution of Na2S2O3, CH4N2S and C6FeK3N6 with a
ratio of 3:3:1. The lift-off process was performed with Du
Pont EKC830 heated at 85 ◦ C. Silicon dioxide was etched
(c) using an Applied Material P5000 MxP+ system with CF-
based chemistry and the silicon release etch was done in a
memsstar XeF2 system.

2.2. Measurement setup

The piezoelectric properties d33, f and e31, f of the AlN films
were characterized using a double-beam laser interferometer
and a 4-point bending setup from aixACCT, respectively.
For more details about the characterization of AlN we refer
to [10]. Measurements of the magnetostrictive properties of
FeCoSiB were performed using a cantilever deflection setup
Figure 1. Schematic cross section (a), photograph (b), SEM
by Klokholm [11].
micrograph (c) of a micro ME sensor device with (1) bond pad, The characterization of the ME coefficients was done
(2) strip line, (3) ME cantilever and (4) etch groove. in a shielded measurement setup (figure 3). To minimize
mechanical vibrations the setup is decoupled using vibration
thin films were sputtered (steps 4 and 5). In the next step Mo dampers (figure 3(1)). Furthermore the whole box is lined with
was patterned by wet chemical etching (step 6) and using Mo as acoustic absorbers (figure 3(2)). Other problems are disturbing
a hard mask wet etching of AlN was performed (step 7). After magnetic and electric fields which couple into the sensor and
removing the Mo hard mask a Cr–Au plating base (pb) was amplifier. Therefore the sensor and the field generating coils
sputtered and Au conduction lines, bond pads and bond frame are placed in a magnetic shielding cylinder made from ten
were electroplated (steps 8 and 9). After that the plating base layers of mu-metal (figure 3(3)). Additionally, the whole box
was wet etched and a dry etch process was performed to open is electrically shielded by a copper fleece (figure 3(4)). For the
the silicon dioxide around the later cantilever beams (step 10). same reason the charge amplifier is placed in a metal cylinder
The magnetic layer FeCoSiB was sputtered and structured (figure 3(5)).
with a lift-off process (step 11). Finally, a bulk silicon release Driving coils and sensor position are displayed in
etch process with XeF2 gas was used to create free standing figure 4 and a simplified schematic of the measurement setup is
rectangular cantilever beams with a width of 0.2 mm and a shown in figure 5. The sensor is incorporated in the lower tube.
length between 0.3 mm and 1.12 mm (step 12). The driving coils for dc bias and alternating magnetic fields
The growth of the 650 nm SiO2 by a thermal wet oxidation are wound around the upper tube. Both are stacked together
process was performed in an ASM 400 vertical furnace. Using and placed in the magnetic shielding cylinder (figure 3(3)).
a Balzers BAK760 system the AlN seed layer of 20 nm The coils are calibrated in the shielding cylinder because
Ti and 100 nm Pt was evaporated. The 1 μm AlN and the high μr increases the magnetic flux of the coils. During
300 nm Mo thin films were sputtered in an Oerlikon Clusterline measurement the whole box is closed.
200 II. AlN was grown by reactive pulsed-dc magnetron The sensor is driven with a magnetic bias field
sputtering using a 99.9995% pure aluminum target. The sputter Hbias parallel to the long axis of the cantilever in order to
process was done under the following conditions: 3 × 10−6 bar maximize the ME coefficient. The bias field shifts the operating
process pressure, 20 sccm Ar flow, 75 sccm N2 flow, chuck point of the magnetostrictive layer to the inflection point of
temperature of 400 ◦ C, 8 kW pulsed dc power and 65 W rf the magnetostriction curve λ. Therefore, the derivative of the
bias power. The plating base of 40 nm Cr and 250 nm Au magnetostriction with respect to the magnetic field strength H
was sputtered in an Oerlikon EVO II and the electroplating and thus the ME coefficient reaches its maximum.
of Au was performed using a RENA x-EPT. A Von Ardenne A lock-in amplifier SR830 from Stanford Research
CS 730 S sputtering system was used to deposit a 20 nm Systems was used to measure the generated voltage from
thick Cr adhesion layer and then the magnetostrictive layer a charge preamplifier Analog Devices AD745 which was

2
J. Micromech. Microeng. 22 (2012) 065024 S Marauska et al

Figure 2. Process flow of the MEMS ME sensor.

Figure 3. Photograph of the shielded measurement setup with (1) Figure 4. Photograph of driving coils with ME device. The inset
vibration dampers, (2) acoustic absorbers, (3) mu-metal cylinder, (4) shows the sensor location in detail.
copper fleece and (5) charge amplifier in metal cylinder.
HAC is generated by a HP33120A function generator from
connected to the sensor. Using the measured voltage Uout the Hewlett Packard. For the intrinsic noise voltage density
ME voltage UME could be calculated with the equation measurement a battery-driven dc bias-field of μ0Hbias = 1 mT
Cf was applied to reach the optimal working point of the sensor.
UME = Uout (1)
CS In the setup a low noise charge preamplifier Analog Devices
where C f is the feedback capacitance and Cs is the capacitance AD745 and a dynamic signal analyzer SR785 from Stanford
of the sensor. The current for the alternating magnetic field Research Systems was used. For detailed information see [12].

3
J. Micromech. Microeng. 22 (2012) 065024 S Marauska et al

Table 1. List of layer thicknesses, Young’s moduli and mass densities to calculate the resonance frequency of the cantilever.
Material FeCoSiB Cr AlN Pt Ti SiO2
d (μm) 2.2 0.01 1.0 0.1 0.02 0.65
E (GPa) 100 140 400 170 110 70
ρ (g cm−3) 7.25 7.19 3.30 21.44 4.51 2.20
Ref. [6] [14, 15] [14, 16] [14, 17] [14, 15] [14, 18]

Figure 5. Schematic of the simplified measurement setup.

Figure 7. Measured dependence of the ME coefficient and ME

voltage on the applied bias field.

the ME coefficient α ME was calculated from the measured

ME voltage UME and the thickness of the piezoelectric layer
dAlN. The ME coefficient at the mechanical resonance is
α ME = 1000 (V m−1)/(A m−1). A static ME coefficient of
α ME = 14 (V m−1)/(A m−1) is calculated using the quality
factor of the resonance curve at ambient air pressure of
Q = 73.
The mechanical resonance at 2.4 kHz can easily be
Figure 6. ME coefficient and ME voltage in dependence on the
observed. This is in good agreement with the calculated value
drive frequency.
of 2.3 kHz for the first resonance frequency of a cantilever
beam using the expression [13]
The following measurements of the ME properties were 
performed on a cantilever with a length of 1.12 mm. 1 λ21 En In
fr = 2
 , (3)
2π L mn
3. Results and discussion where L is the length of the cantilever beam, En is Young’s
modulus, In is the moment of inertia and mn is the mass
3.1. Characterization of the ME layers
per unit length of the individual layers of the material stack.
The sputtered AlN films with a thickness of 1.0 μm are highly The constant λ1 for a single-side rigidly clamped rectangular
c-textured and showed an effective piezoelectric coefficient of cantilever beam is 1.875. The used Young’s moduli, layer
e31, f = −1.1 C m−2 and an effective longitudinal coefficient thicknesses and mass densities are given in table 1.
of d33, f = 4.7 pm V−1 [10]. The saturation magnetostriction Figure 7 shows the ME voltage UME and ME coefficient
λs of the (Fe90Co10)78Si12B10 layer was determined to α ME in dependence on the applied bias field. The maximum
64 ppm [6]. ME coefficient of α ME = 1000 (V m−1)/(A m−1) is achieved
Despite the fact that AlN possesses moderate piezoelectric for a bias field of μ0Hbias = 1 mT in resonance.
coefficients it has a low dielectric coefficient and low dielectric The highest achieved resonant value of 2260 (V m−1)/
losses. This makes AlN a favorable material for sensing (A m−1) at 330 Hz so far was measured by Greve
applications where a high signal-to-noise ratio is required. et al for centimeter sized ME sensors [19]. They further showed
that the product of resonance frequency f 0 and corresponding
3.2. ME coefficient ME coefficients in resonance α ME( f 0) is constant. Using the
measured 1000 (V m−1)/(A m−1) at the resonance frequency
The frequency dependence of the ME coefficient α ME of the of 2.4 kHz leads to a product of 2400 (Hz kV m−1)/(A m−1)
MEMS ME sensor by applying a constant magnetic field of and results in a ME coefficient of 7200 (V m−1)/(A m−1)
μ0Hbias = 1 mT and an ac magnetic field of μ0HAC = 0.01 mT at 330 Hz. The reason for this higher value certainly lies
is shown in figure 6. Using the equation in the nonexistent substrate clamping compared to the larger
 
UME V m−1 bulk silicon-based sensors from Greve et al. This is supported
αME = (2)
dAlN · HAC A m−1 by the fact that the static ME coefficient is about five times

4
J. Micromech. Microeng. 22 (2012) 065024 S Marauska et al

Figure 9. Sensitivity and linearity of the MEMS ME sensor system.

Figure 8. Frequency dependence of noise voltage density Eo.
3.4. Sensitivity and linearity
higher. Comparing the ME coefficient with other micro ME
sensors the measured value in resonance is at least ten timers The sensitivity and linearity of the sensor are shown in
higher [20]. figure 9. Starting from a magnetic field BAC = 100 pT the
An increase of the coefficient could be expected by measured values begin to scatter because of low SNR. For
removing the SiO2 from the backside of the cantilever due values below BAC = 100 pT the output ME voltage no longer
to a reduced clamping effect. Moreover an optimization of depends on the applied magnetic field strength. This shows
the thickness ratio between the magnetostrictive layer and that MEMS ME sensors could be capable of biomagnetic
piezoelectric layer could lead to a higher magnetoelectric measurements. The sensitivity of the sensor in resonance is
coefficient. In the future a vacuum encapsulation of the sensor 780 V T−1 and the calculated static sensitivity is 11 V T−1,
at the wafer level is planned. A strong increase of the coefficient respectively. Improvements are expected by the above-
is expected due to enhanced mechanical quality factor of the described advancements.
cantilever.
4. Conclusion
3.3. Noise level of the sensor system
Despite the fact that the investigated micro ME cantilever
The calculated and measured noise voltage density Eo of the beam is significantly smaller than state-of-the-art centimeter-
ME sensor device is shown in figure 8. In the calculation are sized ME sensors the measured ME coefficient α ME =
included the noise source of the amplifier, the thermal voltage 1000 (V m−1)/(A m−1), sensitivity of 780 V T−1 and noise
noise of the sensor and the thermal noise of the feedback level above 100 pT Hz−1/2 is quite competitive [8, 12,
resistor. Additionally, the measured noise voltage density from 19, 21]. The fact that the product of the ME coefficient times
the signal analyzer SR785 is also depicted in figure 8. For the frequency is a constant leads to the conclusion that the
detailed information about the noise sources we refer to [12]. ME coefficient of 1000 (V m−1)/(A m−1) at 2.4 kHz measured
The corresponding magnetic field noise density Bnoise can be in this work is considerably higher than the coefficient of
calculated from the measured noise voltage density Eo using 2260 (V m−1)/(A m−1) at 333 Hz measured with larger
the equation bulk silicon-based sensors [19]. It is expected that further
Eo improvements in sensor design and vacuum encapsulation
Bnoise = . (4)
dAlN · αME /μ0 will enhance the sensor characteristics. This demonstrates
In resonance and with a bias field of 1 mT a noise voltage that MEMS-based ME sensors have potential as ultrahigh
density level of Eo = 90 nV Hz−1/2 at the amplifier output sensitivity and high spatial resolution magnetic field sensors.
yields Bnoise = 110 pT Hz−1/2.
This is higher than the reported magnetic noise density in Acknowledgments
the lower pT Hz−1/2 regime for centimeter sized ME sensors
[4, 21]. The calculated noise voltage density revealed that the The authors would like to thank the German Science
dominating noise sources are the voltage and current noise Foundation (DFG) for financial funding through the
from the charge amplifier. Improvements of the measurement Collaborative Research Center SFB 855 ‘Magnetoelectric
setup and sensor design could provide a remedy by decreasing Composite Materials—Biomagnetic Interfaces of the Future’.
parasitic capacities, which results in higher sensor currents and
voltages. Further, an increase of the AlN thickness will reduce
the noise due to a higher voltage output (see (4)). Moreover References
this will lead to a higher resonance frequency and a further [1] Eerenstein W, Mathur N D and Scott J F 2006 Multiferroic
reduction of the 1/ f noise. It is expected that an increase of and magnetoelectric materials Nature 442 759–65
the ME coefficient due to vacuum encapsulation will lead to [2] Fiebig M 2005 Revival of the magnetoelectric effect J. Phys.
ultralow noise levels. D: Appl. Phys. 38 123–52

5
J. Micromech. Microeng. 22 (2012) 065024 S Marauska et al

[3] Nan C W, Bichurin M I, Dong S, Viehland D and Srinivasan G thin-film magnetoelectric sensors IEEE Trans. Instrum
2008 Multiferroic magnetoelectric composites: historical Meas. 60 2995–01
perspective, status, and future directions J. Appl. Phys. [13] Voiculesco I, Zahgloul M E, McGill R A, Houser E J
103 031101 and Fedder G K 2005 Electrostatically actuated resonant
[4] Jahns R, Greve H, Woltermann E, Lage E, Quandt E microcantilever beam in CMOS technology for the
and Knoechel R 2011 Magnetoelectric sensors for detection of chemical weapons IEEE Sensors 5 641–47
biomagnetic measurements Proc. IEEE Medical [14] Shackelford J F and Alexander W (ed) 2001 CRC Materials
Measurements and Applications (Bari, Italy, 2003) Science and Engineering Handbook 3rd edn (Boca Raton,
pp 107–10 FL: CRC Press)
[5] Srinivasan G, Rasmussen E T, Levin B J and Hayes R 2002 [15] Halg B 1990 On a nonvolatile memory cell based on
Magnetoelectric effects in bilayers and multilayers Phys. micro-electro-mechanics IEEE Proc. Micro Electro
Rev. B 65 134402 Mechanical Systems (Napa Vally (CA) USA, 1990) pp 172–6
[6] Greve H, Woltermann E, Quenzer H J, Wagner B [16] Yoshida M I, Okumiya M, Ichiki R, Khalifa W, Tekmen C,
and Quandt E 2010 Giant magnetoelectric coefficients in Tsunekawa Y, Hara T and Tanaka K 2009 Mechanical
(Fe90Co10)78Si12B10-AlN thin film composites Appl. Phys. properties of aluminum nitride layer formed by duplex
Lett. 96 18250 coating of barrel nitriding and plasma nitriding Plasma
[7] Quandt E, Stein S and Wuttig M 2005 Magnetic vector field Process. Polym. 6 310–13
sensor using magnetoelectric thin-film composites Trans. [17] Delobelle P, Fribourg-Blanc E and Rèmiens D 2006
Magn. 5 3667–69 Mechanical properties determined by nanoindentation tests
[8] Dong S, Li J-F and Viehland D 2003 Ultrahigh magnetic field of [Pb(Zr,Ti)O3] and [Pb(Mg1/3Nb2/3)1−xTixO3] sputtered
sensitivity in laminates of TERFENOL-D and thin films Thin Solid Films 515 1385–93
Pb(Mg1/3Nb2/3)O–PbTiO3 crystals Appl. Phys. Lett. [18] Petersen K E 1978 Dynamic micromechanics on silicon:
83 2265–67 techniques and devices IEEE Trans. Electron Devices
[9] Drung D and Koch H 1994 An integrated DC SQUID 25 1241–49
magnetometer with variable additional positive feedback [19] Greve H, Woltermann E, Jahns R, Marauska S, Wagner B,
Supercond. Sci. Technol. 7 242–45 Knöchel R, Wuttig M and Quandt E 2010 Low damping
[10] Marauska S, Dankwort T, Quenzer H and Wagner B 2011 resonant magnetoelectric sensors Appl. Phys. Lett.
Sputtered thin film piezoelectric aluminium nitride as a 97 152503
functional MEMS material and CMOS compatible process [20] Onuta T, Wang Y, Long C J and Takeuchi I 2011 Energy
integration Proc. Eng. 25 1341–4 harvesting properties of all-thin-film multiferroic
[11] Klokholm E 1976 The measurement of magnetostriction cantilevers Appl. Phys. Lett. 99 203506
ferromagnetic thin films IEEE Trans. Magn. [21] Zhuang X, Sing M L C, Cordier C, Saez S, Dolabdjian C,
12 819–21 Das J, Gao J, Li J and Viehland D 2011 Analysis of noise in
[12] Jahns R, Greve H, Woltermann E, Quandt E and Knöchel R magnetoelectric thin-layer composites used as magnetic
2011 Noise performance of magnetometers with resonant sensors IEEE Sensors J. 11 2183–88