Exchange biasing of magnetoelectric composites - Nature

ARTICLES PUBLISHED ONLINE: 22 APRIL 2012 | DOI: 10.1038/NMAT3306

Exchange biasing of magnetoelectric composites Enno Lage1 , Christine Kirchhof1 , Viktor Hrkac1 , Lorenz Kienle1 , Robert Jahns2 , Reinhard Knöchel2 , Eckhard Quandt1 and Dirk Meyners1 * Magnetoelectric composite materials are promising candidates for highly sensitive magnetic-field sensors. However, the composites showing the highest reported magnetoelectric coefficients require the presence of external d.c. magnetic bias fields, which is detrimental to their use as sensitive high-resolution magnetic-field sensors. Here, we report magnetoelectric composite materials that instead rely on intrinsic magnetic fields arising from exchange bias in the device. Thin-film magnetoelectric two–two composites were fabricated by magnetron sputtering on silicon-cantilever substrates. The composites consist of piezoelectric AlN and multilayers with the sequence Ta/Cu/Mn70 Ir30 /Fe50 Co50 or Ta/Cu/Mn70 Ir30 /Fe70.2 Co7.8 Si12 B10 serving as the magnetostrictive component. The thickness of the ferromagnetic layers and angle dependency of the exchange bias field are used to adjust the shift of the magnetostriction curve in such a way that the maximum piezomagnetic coefficient occurs at zero magnetic bias field. These self-biased composites show high sensitivity to a.c. magnetic fields with a maximum magnetoelectric coefficient of 96 V cm−1 Oe−1 at mechanical resonance.

M

agnetoelectric composites, that is composites of magnetostrictive and piezoelectric materials, are very attractive in comparison with single-phase magnetoelectric materials owing to their much higher magnetoelectric coefficient and their much higher maximum operation temperature. In such composites, the magnetoelectric effect is a product property of the piezomagnetism, the piezoelectricity of the corresponding phases and their coupling1 . On one hand, the overall magnetoelectric properties can be tailored by choosing suited materials for both phases, but on the other hand, they are restricted by the individual properties of the used components. The most severe limitation arises from the magnetic-field dependency of the piezomagnetic coefficient dm , which is the first derivative of the magnetostrictive strain λ with respect to the magnetic field H . Accordingly, the magnetoelectric voltage coefficient αME is directly proportional to dm . As this piezomagnetic coefficient is almost zero at zero magnetic field for all magnetostrictive materials with small magnetic hysteresis, magnetoelectric composites used for the direct magnetoelectric effect up to now require a magnetic bias field such as 20 mT for Terfenol-D (ref. 2), 8 mT for Ni (ref. 3), 7 mT for nickel ferrite (ref. 4) or 0.6 mT for Fe70.2 Co7.8 Si12 B10 (ref. 5) depending on the used magnetostrictive material and its magnetic anisotropy field. For any application such as magnetic sensing or imaging for biological systems requiring very high sensitivity to a.c. magnetic fields and high spatial resolution, this need for an applied external magnetic bias field is of significant disadvantage as it increases the required space, thus decreasing the possible spatial resolution. It also adds a potentially supplemental noise source, thus lowering the signal-to-noise ratio, and it may interfere with neighbouring sensors, which again limits the spatial resolution or any vector-field approaches based on the combination of individual sensors. To overcome these severe limitations arising from an external magnetic bias field, an internal magnetic biasing of the magnetostrictive component within a magnetoelectric composite is in demand. In principle, five different approaches are feasible on the basis of the integration of hard magnetic layers with a permanent moment6 , the use of remanence magnetization7 or field-dependent

resonant frequency8 in a hysteretic magnetostrictive material, the use of stresses by means of the inverse magnetostriction and the exchange biasing of the magnetostrictive material9 . The last option is considered to be the most promising approach, as it reduces stray fields present in the hard-magnetic-layer approach, it is less history dependent than the use of the hysteresis and it is expected to be much easier to adjust when compared with any stress biasing. The exchange-bias effect refers to a shift HEB of the magnetization curve due to antiferromagnetic–ferromagnetic exchange coupling10 . It is maximum for magnetic fields applied along the induced pinning direction and vanishes perpendicular to it, where it contributes to an enhanced total anisotropy field HK,tot (Fig. 1a). Exchange biasing is well established in read-head technology based on anisotropic, giant or tunnel magnetoresistive effects. The introduction of exchange-bias effect has also been suggested for magnetoelectric read-heads9 . However, in contrast to magnetoresistive sensors, where strong pinning of thin ferromagnetic layers with thicknesses in the nanometre range is required, magnetoelectric composites need moderate and precise biasing of magnetostrictive layers holding thicknesses of several micrometres5 . The interfacial character of the exchange-bias effect results in a reciprocal dependency of the exchange bias with the thickness tFM of the coupled ferromagnetic layer11 . This is confirmed in Fig. 1b, where the exchange bias fields of highly magnetostrictive materials Fe50 Co50 and Fe70.2 Co7.8 Si12 B10 coupled to Mn70 Ir30 as a natural antiferromagnet are shown for different tFM ranging from 6 to 200 nm. An upper limit for the ferromagnetic layer thickness is set by the domain-wall width. Consequently, only a multilayer stack design will be capable of exchange biasing of many individual magnetostrictive layers holding an accumulated thickness in the micrometre range. In this work, different sets of samples were investigated (Supplementary Table S1). The magnetostriction was investigated on two different exchange-biased systems: (7 nm Ta/3 nm Cu/7 nm Mn70 Ir30 /tFM nm Fe50 Co50 ) and (7 nm Ta/3 nm Cu/7 nm Mn70 Ir30 /tFM nm Fe70.2 Co7.8 Si12 B10 ), both as single layers and as multilayers. The magnetoelectric effect was measured on composites comprising the former magnetostrictive multilayers

1 Kiel

University, Institute for Materials Science, Kaiserstr. 2, 24143 Kiel, Germany, 2 Kiel University, Institute of Electrical Engineering, Kaiserstr. 2, 24143 Kiel, Germany. *e-mail: [email protected]. NATURE MATERIALS | VOL 11 | JUNE 2012 | www.nature.com/naturematerials

© 2012 Macmillan Publishers Limited. All rights reserved

523

NATURE MATERIALS DOI: 10.1038/NMAT3306

ARTICLES a

|| pinning direction ⊥ pinning direction

M

1

Hcl

0

Hcl

¬1 HEB ¬15

b 0.2 25

¬10

¬5

0 μ 0H (mT)

HK,tot 5

10

Reciprocal layer thickness (nm¬1) 0.1 0.05

15

0.01

μ 0HEB (mT)

20 15 10 FeCo single layer

5

FeCoSiB single layer FeCoSiB multilayer

0 5

10 tFM (nm)

20

100

Figure 1 | Exchange biasing of ferromagnetic single-layer and multilayer systems. a, Typical normalized magnetization curves of an exchange-biased ferromagnet (here Fe70.2 Co7.2 Si12 B10 ) measured along (black) and perpendicular to (red) the induced pinning direction. The exchange bias field HEB is defined as the midpoint of both coercive fields HC | and HC || recorded along the pinning direction. The total anisotropy field HK,tot is determined by the extrapolation of the linear magnetization regime obtained along the hard axis to the value of the saturation magnetization. b, Thickness dependency of exchange bias for Fe50 Co50 (squares) and Fe70.2 Co7.8 Si12 B10 (circles). A reciprocal illustration was used to confirm the 1/tFM dependency of the exchange bias. For the multilayer case, tFM represents the thickness of the magnetostrictive layer incorporated in a single sequence. By repetition of the sequence the total magnetostrictive layer thickness was adjusted to roughly 1 µm.

deposited on trenched cantilever substrates holding a 2-µm-thick AlN layer, acting as the piezoelectric phase. It is known that a (111) texture of the Mn70 Ir30 layer is mandatory for reliable exchange biasing12,13 . By the introduction of the Ta/Cu bilayer and subsequent annealing, a (111) texture of the Mn70 Ir30 layer was promoted even in the multilayered stack. In Fig. 1b, the data prove that this leads to an effective exchange bias also for total magnetostrictive layer thicknesses up to the micrometre range. The magnitude of the effect is reduced probably owing to an increased roughness and reduction of the degree of texture of these multilayers. This was studied for the Fe50 Co50 samples by transmission electron microscopy (TEM) investigations. The influence of AlN on the multilayer stacking is presented in the high-angle annular dark-field scanning TEM (HAADF-STEM) Z contrast image of Fig. 2a. The columnar growth of AlN initiates a high degree of roughness of the multilayer system next to 524

their common interface. Coarsening phenomena occur also in this specimen. A notable decrease of the roughness is observed from substrate to surface. To validate the periodicity of the respective sequences, chemical analysis profiles were conducted. A representative area is presented as the inset of Fig. 2a. The line scans of Fig. 2b (line in Fig. 2a inset) show a clear separation of the layers and thus the chemical integrity of the material. An exception seems to be Ir. However, the spatial overlap with the Cu and Ta peaks represents an artefact resulting from narrow energy differences of the Ir–L lines (α9.173 keV) and for example Ta–L lines (β9.34 keV). The dashed lines in the plot mark the boundaries for single sequences. These multilayers vary from 36.5 to 37.8 nm. Investigations over a high number of sequences yield an average value of 36.6 nm for a sequence. The lower contrast in the cross-sectional view (Fig. 2a inset) originates from the surface roughness caused by the AlN substrate. A second sample was fabricated on silicon as the substrate, because planar stacking of the multilayer system is ensured and allows the influence-free determination of the orientation relationship between respective layers, as demonstrated by the cross-sectional bright-field image in Fig. 2c. At the substrate/multilayer interface region, selected area electron diffraction (SAED) examinations indicate a highly textured structure. The diaphragm for area selection limits the diffraction to approximately six multilayer sequences. The texture within the multilayer system decreases with increasing distance from the substrate/multilayer interface as a result of coarsening effects (see Supplementary Information). The SAED patterns show the different degrees of the texture by a series of concentric diffraction rings (Fig. 2c) and well-aligned Bragg intensities along [111], respectively. The high-resolution micrograph recorded at the multilayer next to the substrate exhibits a well-defined orientational relationship between the respective layers (Fig. 2d). Epitaxial growth between the seed layers (Ta, Cu) and the antiferromagnetic Mn70 Ir30 layer on the (111) plane is observed using fast Fourier transformation. The corresponding zone axis has been identified as [110] (ref. 14). Furthermore, Fe50 Co50 layers are polycrystalline with preferred orientation of the grains along [111] (ref. 15). To establish exchange biasing of all magnetostrictive layers, the textured stack was cooled down in the presence of a magnetic field µ0 Hcool = 0.3 T from temperatures above the Néel temperature of the incorporated antiferromagnet. By this procedure a collinear alignment of the unidirectional anisotropy due to the exchangebias effect and the uniaxial anisotropy of the ferromagnetic magnetostrictive phase is approached. Magnetization loops in the easy- and hard-axis orientation for different layer thicknesses of the coupled ferromagnetic material do not show only the exchangebias dependency, but also a corresponding influence on the total magnetic anisotropy field HK,tot (Fig. 3). In the present simple case of collinear alignment of both anisotropy contributions, the measured anisotropy field equals the sum of HEB and HK , with HK being the anisotropy field of the ferromagnet. However, as long as HK is small when compared with HEB , the course of the total anisotropy field is dominated by the 1/tFM dependency of HEB . Assuming a constant saturation magnetostriction, the anisotropy field should be approximately reciprocally proportional to the piezomagnetic coefficient and thus has to be minimized for an optimum magnetoelectric effect. As Fig. 3 indicates, this favours comparable thick magnetostrictive layers as long as the resulting exchange bias is large enough to shift the magnetization curve by a value close to half of the anisotropy field to achieve a maximum of the piezomagnetic coefficient in zero field. As a consequence of the collinear alignment of the aforementioned anisotropies, exchange bias is always observed along the pinning direction of the material whereas magnetostriction is maximum perpendicular to it (Fig. 4). Accordingly, an adjustment of the exchange bias of the ferromagnetic layers parallel to the NATURE MATERIALS | VOL 11 | JUNE 2012 | www.nature.com/naturematerials



ARTICLES

a

b

90

Co

90

Fe

90

Mn

Counts

40 nm 60

Ir

120

Cu

90

Ta

20

200 nm

c

40

60 80 Distance (nm)

d

100

120

FeCo

[111] Mn3Ir

[110] Cu

∗

Ta 500 nm

[110]

4 nm

Figure 2 | TEM of an exchange-biased magnetostrictive multilayer system. a, Z-contrast image of the Fe50 Co50 multilayer system on a AlN substrate (right-hand side of the image, columnar grown) and a magnified view (inset) with a line indicating the selected area for chemical analysis. b, Associated energy-dispersive X-ray spectra for several multilayer sequences. The single sequences (dashed lines) vary from 36.5 to 37.8 nm. c, Cross sectional TEM bright-field image of the Fe50 Co50 multilayer system on a silicon substrate with SAED patterns of different specimen areas. Asterisk: area next to multilayer/substrate interface. d, Associated high-resolution TEM micrograph (left) and corresponding fast Fourier transformation patterns (right) of the multilayer sequence next to the multilayer/substrate interface. The marked spots correspond to (111) planes for Mn3 Ir and Cu layers, respectively.

direction of the measurement direction would result in an almost complete extinction of the magnetostrictive response of the sensors discussed here. To overcome this antagonism, it was necessary to induce the exchange bias in a certain inclination angle φ to the measurement direction. Figure 5a shows measured magnetization loops of magnetostrictive multilayers for φ ranging from 18◦ to 72◦ . The apparent shift along the measurement direction HEB,long results from the projection of the exchange bias field. Its continuous decrease following a cos φ proportionality is consistent with theory and experiments on simple antiferromagnet–ferromagnet bilayers16,17 (Fig. 5d red circles). Figure 5b shows magnetostriction curves being shifted to the same extent as the corresponding magnetization curves of Fig. 5a. The maximum magnetostriction amounts to about 50 ppm for inclination angles φ between 72◦ and 90◦ . A severe reduction of the magnetostriction is observed only for values of φ below 36◦ (Fig. 5b,d black squares). Focusing on the application as a magnetic-field sensor, the slope of the magnetostriction curve has to be maximized at zero field. The magnetostriction curves

of Fig. 5b indicate that there are two effects acting on that slope, that is, the shift of the curve and the overall magnetostriction as a function of φ. The course of the piezomagnetic coefficient with φ is given in Fig. 5d. For the extremes (φ = 0◦ and φ = 90◦ ), the piezomagnetic coefficient reduces to small values of less than 0.5 ppm mT−1 reflecting the above antagonism. Variation of φ leads to a pronounced increase of dm . Finally, the measurement for φ = 66◦ (Fig. 5c) proves that angular and thickness dependency of HEB allow for tuning its projection HEB,long in a way that a maximum of the piezomagnetic coefficient dm = 4.6 ppm mT−1 is adjusted to zero magnetic field. Using this result, magnetoelectric cantilevers were fabricated featuring the self-biased magnetostrictive multilayer and a piezoelectric AlN layer. Figure 6a shows the dependency of the magnetoelectric coefficient on the external magnetic bias field, which was measured at the mechanical resonance (f = 1,011.5 Hz) of the cantilever, because operation at resonance frequency is known to enhance the sensitivity of the sensor.

NATURE MATERIALS | VOL 11 | JUNE 2012 | www.nature.com/naturematerials


525


ARTICLES Reciprocal layer thickness (nm¬1) 0.1 0.05

0.2 30

0.01

20 15 10

FeCo HK,tot single layer FeCoSiB HK,tot single layer

5

FeCoSiB HK,tot multilayer 0

5

10 Layer thickness tFM (nm)

20

100

Figure 3 | Total anisotropy field versus magnetostrictive layer thickness tFM for the same samples as in Fig. 1b. As the magnitude of HK,tot is determined by HEB and HK of the Fe50 Co50 or Fe70.2 Co7.8 Si12 B10 layers, respectively, the 1/tFM dependency according to the contribution of HEB is distorted by the change of HK with layer thickness.

The result demonstrates the success of the exchange-bias approach as a value of 21.2 V cm−1 Oe−1 is obtained without external biasing. A magnetoelectric coefficient of approximately 550 V cm−1 Oe−1 would have been expected at a resonance frequency of 1 kHz for state-of-the-art Fe70.2 Co7.8 Si12 B10 thinfilm composites18 . This reduction by a factor of 26 originates from three factors: the lower effective magnetostrictive film thickness (0.8 µm when compared with 1.75 µm); the enhanced total anisotropy field of 11 mT (Fig. 4b) instead of 0.9 mT measured for 1.75 µm Fe70.2 Co7.8 Si12 B10 by vibrating sample magnetometry; and the lower piezomagnetic coefficient of crystalline Fe50 Co50 . Combining the findings on Fe50 Co50 , an optimized magnetoelectric composite was fabricated on a 2-µm-thick piezoelectric AlN layer, coated with an eight times repeating layer sequence of 7 nm Ta/3 nm Cu/7 nm Mn70 Ir30 /130 nm Fe70.2 Co7.8 Si12 B10 . Here, the main improvement was the reduction of the five times lower total anisotropy field by using much thicker magnetostrictive layers (130 nm, compared with 20 nm; see Fig. 3). For this layer thickness, the interplay of HEB and HK,tot causes an optimum inclination angle φ = 70◦ . The corresponding magnetoelectric measurement is shown in Fig. 6b. A maximum magnetoelectric coefficient of

b 50

1.0

50

0.5

40

0.5

40

0

30

0

30

20

¬0.5

20

¬0.5

10 ¬1.0

λ (ppm)

1.0

λ (ppm)

M

a

M

µ0HK,tot (mT)

25

96.7 V cm−1 Oe−1 at the sensors’ resonance frequency of 1,197 Hz was obtained at zero external bias field, which is an enhancement by a factor of 4.5. For such sensors, the noise voltage density as a function of frequency is measured at the charge amplifier output and plotted in Fig. 7a. It shows two pronounced noise regimes indicated by their lines of best fit. On the one hand, a 1/f noise behaviour is observed for frequencies√below 250 Hz. On the other hand, a constant noise level of 66 nV/ Hz appears for frequencies above 250 Hz. Considering the voltage gain of 30.16 of the charge amplifier and the sensor’s magnetoelectric coefficient of 96.7 V cm−1 Oe−1 , a theoretical √ limit of detection of 11 pT/ Hz at a signal-to-noise ratio of 1 can be calculated. To verify this value, the limit of detection was measured in an electromagnetically and acoustically shielded set-up. Figure 7b shows the sensor response at resonant frequency as a function of the applied a.c. magnetic field. The graph shows a constant responsivity of 5.7 V mT−1 prevailing for several orders of magnitude down to approximately 100 pT where data scattering begins. The measured signal is dominated by √noise at about 10 pT coinciding√with a limit of detection of 10 pT/ Hz and the noise level of 66 nV/ Hz derived from Fig. 7a. The achieved limit of detection demonstrates the high capability of exchange-biased thin-film magnetoelectric sensors. In biomagnetic sensing, for instance, most signals are confined to frequencies below 100 Hz and a high spatial resolution is favoured19 . With an adapted sensor design the resonant frequency and sensor size can be matched to these specific demands. To further demonstrate the effectiveness of the exchange-bias approach, a comparison of the sensor response of systems with and without exchange bias is made in Fig. 8. In the latter system, tantalum was substituted for the antiferromagnetic layer. This provides virtually identical multilayer stacks differing in the absence of exchange biasing. The sensor without exchange bias shows the highest response provided that the magnetic bias is set to the optimum. In absence of any bias field, the sensor response almost vanishes. In contrast, the sensor exhibiting exchange bias provides a response at zero bias field being reduced by only a factor of 2.4 when compared with the sensor without exchange bias at its optimum bias field. This effective reduction originates in the enhancement of Hk,tot , which measures 28 Oe and 8 Oe for the systems with and without exchange bias, respectively. In principle, our concept of exchange-biased magnetoelectric composites for magnetic-field-sensor applications could be transferred to multiferroic antiferromagnets. In that case, the multiferroic antiferromagnet would be used to establish exchange biasing of a magnetostrictive material and to produce the sensor output voltage by means of the piezoelectric effect. Such a composite of

10 ¬1.0

¬40

¬20

0 μ 0H (mT)

20

40

0

¬40

¬20

0 μ 0H (mT)

20

40

0

Figure 4 | The antagonism between exchange bias and magnetostriction. Normalized magnetization curves (black) and magnetostriction curves (red) of Fe50 Co50 multilayers deposited on cantilevers. a,b, The pinning direction is induced either parallel (a) or perpendicular (b) to the longitudinal cantilever axis. Whereas the exchange bias shows a maximum value when measured parallel to the pinning direction, the magnetostrictive response almost vanishes. In the case of the measurement perpendicular to the pinning direction, the projection of the exchange bias vanishes; however, the magnetostrictive response is maximal. 526




ARTICLES

a

a 20

αME (V cm¬1 Oe¬1)

M

1

0 18° 36° 66°

¬1

72°

10

0

¬10

b 45

¬20

λ (ppm)

¬30

¬15

0 μ 0H (mT)

30

15

30

b 100

18°

15

36° 50

αME (V cm¬1 Oe¬1)

66° 0

72°

c d m (ppm mT¬1)

5 0 ¬5

d

¬15

0 μ 0H (mT)

15

6

40

4

30 20

2

(ppm mT¬1) HEB, long (mT) d m 0

50

λmax (ppm)

¬6

30

60

10 0

0

30

60

90

¬50

¬100

66°

¬30

0

0

φ (°)

Figure 5 | Solution to the antagonism by introducing the inclination angle φ. a,b Normalized magnetization curves (a) and corresponding magnetostriction curves (b) of Fe50 Co50 multilayer stacks on cantilever samples for selected values of the inclination angle φ. Magnetization curves and corresponding magnetostriction curves hold the same shift HEB,long . The curves confirm a continuous decrease of HEB,long with increasing φ. c, Piezomagnetic coefficient versus magnetic field for φ = 66◦ . It demonstrates the successful shift of a maximum of dm to zero field. d, Maximum magnetostrictive strain λmax obtained for magnetic fields along the longitudinal cantilever axis, HEB,long and piezomagnetic coefficient d0m for zero magnetic field as a function of φ.

the multiferroic antiferromagnet and a magnetostrictive material would also produce a magnetoelectric response without the need for a d.c. magnetic bias field. Exchange bias with multiferroic antiferromagnets was shown for various materials such as YMnO3 (ref. 20), Cr2 O3 (ref. 21) and BiFeO3 (refs 22,23). The last material

¬4

¬2

0 μ 0H (mT)

2

4

6

Figure 6 | Magnetoelectric voltage coefficients αME of exchange-biased magnetoelectric composite sensors as a function of an externally applied magnetic field H. a, Plot of αME for a Fe50 Co50 -based sensor (FeCo-ME in Supplementary Table S1) with the pinning direction induced at φ = 66◦ . It holds a value of 21.2 V cm−1 Oe−1 at zero field at mechanical resonance frequency fres = 1,011.5 Hz. b, The Fe70.2 Co7.8 Si12 B10 -based sensors (FeCoSiB-ME in Supplementary Table S1) show a maximum magnetoelectric coefficient of 96.7 V cm−1 Oe−1 for φ = 70◦ and fres = 1,197 Hz.

seems to be the most promising candidate for the aforementioned purpose, because room-temperature exchange bias was recently demonstrated in polycrystalline BiFeO3 /NiFe films that were deposited by radiofrequency sputtering24 . The challenge one has to accept here is multilayering to enhance the volume of the pinned magnetostrictive sensing layer. For the presented composites, the possibility to pin each ferromagnetic layer in a multilayer stack is demonstrated. This results in a defined overall exchange bias of relatively large ferromagnetic volume and allows for the realization of highly sensitive magnetoelectric sensors without the need for externally applied magnetic bias fields. In the case of magnetostrictive actuators, this internal biasing could be used to replace sophisticated pre-stressing for bidirectional motion. The antagonism between exchange bias and high magnetostriction is overcome by inducing the exchange bias under an angle with respect to the longitudinal cantilever axis. As the piezomagnetic coefficient in zero field is dependent on both quantities, a specific optimum can be found for any set of samples. As a result, we prove successful adjustment and implementation of an internal bias in Fe50 Co50 -based magnetoelectric composite sensors with a magnetoelectric voltage coefficient of 21.2 V cm−1 Oe−1 at zero field.



527


ARTICLES a 2.5

10¬4

2.0

UME (a.u.)

Unoise (V/√Hz)

10¬5

10¬6

Line of best fit 100

101

0 102

103

104

1,800

Frequency (Hz) 10¬1

1,900

2,000 Frequency (Hz)

2,100

2,200

Figure 8 | Comparison of magnetoelectric sensors with and without exchange bias. The graph shows the frequency-dependent magnetoelectric voltage response UME of a non-internally-biased sensor for absent (open squares) and optimum external magnetic bias field (filled squares). The response of a sensor holding exchange bias is also plotted (circles). For clarification the signals are normalized to the maximum response of the exchange-biased sensor.

10¬2 10¬3

UME (V)

1.0 0.5

Unoise

10¬4 10¬5 10¬6 10¬7 10¬8

10¬12

fres = 1,259 Hz 10¬11

10¬10

10¬9 10¬8 μ 0HAC (T)

10¬7

10¬6

10¬5

Figure 7 | Performance of the Fe70.2 Co7.8 Si12 B10 -based magnetoelectric sensor operated with a charge amplifier. In both measurements (a,b) the voltage gain of the charge amplifier was 30.16. a, Plot of noise voltage density Unoise and line of best fit for 1/f and constant noise regimes. Above √ 250 Hz, a constant voltage noise of 66 nV/ Hz is measured. b, Sensitivity curve at mechanical resonance fres = 1,259 Hz. For several orders of magnitude the sensors voltage response UME is linear and yields a responsivity of 5.7 V mT−1 at the output of the charge amplifier. Data scattering starts at about 100 pT. Below the limit of detection of about √ 10 pT/ Hz, the signal is covered by noise.

Furthermore, we show possibilities to improve the magnetoelectric voltage coefficient. By the reduction of the total anisotropy field, we obtain a maximum value of 96.7 V cm−1 Oe−1 at zero field for a sensor comprising highly magnetostrictive Fe70.2 Co7.8 Si12 B10 . Combining this sensor with a low-noise charge √ amplifier, a responsivity of 5.7 V mT−1 , a noise level of 66 pT/ Hz and a resulting limit of √ detection of 10 pT/ Hz at resonant frequency were achieved. This proves the high capability of magnetoelectric thin-film sensors for magnetic-sensing applications.

Methods All of the layers were magnetron-sputtered using a von Ardenne CS730 S sputtering system for the magnetostrictive component and an Oerlikon Clusterline CLN200 sputtering tool for the piezoelectric AlN layer. Silicon (100) substrates (2 mm × 20 mm × 300 µm and 5 mm × 5 mm × 300 µm) were used for magnetostriction and magnetization measurements. The magnetoelectric composite sensors were fabricated on silicon (100) cantilevers (3 mm × 25 mm × 650 µm) functionalized with a double layer of Ti and Pt as the bottom electrode followed by a 2 µm layer of AlN. The AlN was structured and partially etched in 85% phosphoric acid to expose the bottom electrode. The magnetostrictive multilayer stack was deposited on top of AlN and structured by means of a lift-off process. All samples were finally capped with a 3 nm Ta layer to prevent surface oxidation. By reactive ion etching, a trench (200–240 µm × 7.5 mm) was etched into the backside of the silicon cantilever to 528

1.5

10¬7

10¬8

b

Without EB at optimum bias field Without EB no bias field With EB no bias field

lower its mechanical resonance frequency. All samples were annealed at a base pressure of 10−5 mbar at 250 ◦ C for 15 min for Fe70.2 Co7.8 Si12 B10 and 350 ◦ C for 1 h for Fe50 Co50 , respectively. During annealing and subsequent cooling a magnetic field µ0 Hcool = 0.3 T was present. For the structural investigation TEM was performed using a Tecnai F30 STwin microscope (300 kV, field-emission gun, spherical aberration constant Cs = 1.2 mm). For energy-dispersive X-ray analysis, a Si/Li detector (EDAX System) was applied. STEM Z -contrast images were recorded by using a HAADF detector. Magnetic hysteresis curves were recorded by vibrating sample magnetometry using a LakeShore model 7300 system. The magnetostrictive properties were measured in a cantilever deflection set-up after ref. 25. This measurement determines the magnetoelastic coefficient b as a function of the applied field. According to ref. 26, the magnetostrictive strain λ can be calculated using λ = −b

1 + νf Ef

with Ef and νf being the Young modulus and the Poisson ratio of the magnetostrictive layer, respectively. λ is calculated using Ef = 230 GPa for Fe50 Co50 and Ef = 150 GPa for Fe70.2 Co7.8 Si12 B10 with a Poisson ratio νf of 0.3 (refs 27,28). The piezomagnetic coefficient dm was calculated by numerical derivation of the magnetostriction curves. For the measurement of the magnetoelectric properties, an electromagnet was used to generate a d.c. bias field or to compensate the static magnetic field to zero, respectively. A pair of Helmholtz coils generated an a.c. driving field of 1 Oe for Fe70.2 Co7.8 Si12 B10 and 3 Oe for Fe50 Co50 with a frequency matching that of the mechanical resonance of the cantilever samples. The samples were clamped at one end to the sample holder and were then placed with the longitudinal axis parallel to the driving-field direction. A lock-in amplifier triggered by the driving-field frequency was used to obtain the magnetoelectric response. For the measurement of the noise voltage density spectrum, the output of the magnetoelectric sensor was connected to a charge amplifier with a charge-to-voltage coefficient of 2 × 1010 V C−1 (voltage gain of 30.16). The overall noise from the sensor and the amplifier was measured with Stanford Research Systems dynamic signal analyser SR785 under well-shielded conditions without any additional magnetic field. The sensitivity curve was measured at the output of the charge amplifier by using Stanford Research Systems lock-in amplifier SR830. The additional a.c. magnetic field was provided by Keithley’s a.c. current source 6221.

Received 28 November 2011; accepted 19 March 2012; published online 22 April 2012

References 1. Ma, J., Hu, J., Li, Z. & Nan, C-W. Recent progress in multiferroic magnetoelectric composites: From bulk to thin films. Adv. Mater. 23, 1062–1087 (2011). 2. Dong, S., Cheng, J., Li, J. F. & Viehland, D. Enhanced magnetoelectric effects in laminate composites of Terfenol-D/Pb(Zr,Ti)O3 under resonant drive. Appl. Phys. Lett. 83, 4812–4814 (2003). NATURE MATERIALS | VOL 11 | JUNE 2012 | www.nature.com/naturematerials


NATURE MATERIALS DOI: 10.1038/NMAT3306 3. Laletin, V. M. et al. Frequency and field dependence of magnetoelectric interactions in layered ferromagnetic transition metal-piezoelectric lead zirconate titanate. Appl. Phys. Lett. 87, 222507 (2005). 4. Srinivasan, G. et al. Magnetoelectric bilayer and multilayer structures of magnetostrictive and piezoelectric oxides. Phys. Rev. B 64, 214408 (2001). 5. Greve, H., Woltermann, E., Quenzer, H-J., Wagner, B. & Quandt, E. Giant magnetoelectric coefficients in (Fe90 Co10 )78 Si12 B10 -AlN thin film composites. Appl. Phys. Lett. 96, 182501 (2010). 6. Tadahiko, K. & Isao, S. Self bias magnetostrictive material. Japanese patent 09083037 A (1997). 7. Mandal, S. K., Sreenivasulu, G., Petrov, V. M. & Srinivasan, G. Flexural deformation in a compositionally stepped ferrite and magnetoelectric effects in a composite with piezoelectrics. Appl. Phys. Lett. 96, 192502 (2010). 8. Onuta, T-D. et al. Energy harvesting properties of all-thin-film multiferroic cantilevers. Appl. Phys. Lett. 99, 203506 (2011). 9. Vopsaroiu, M., Blackburn, J. & Cain, M. G. A new magnetic recording read head technology based on the magneto-electric effect. J. Phys. D 40, 5027–5033 (2007). 10. Meiklejohn, W. & Bean, C. New magnetic anisotropy. Phys. Rev. 105, 904–913 (1957). 11. Nogués, J. & Schuller, I. K. Exchange bias. J. Magn. Magn. Mater. 192, 203–230 (1999). 12. Van Driel, J., de Boer, F. R., Lenssen, K-M. H. & Coehoorn, R. Exchange biasing by Ir19 Mn81 : Dependence on temperature, microstructure and antiferromagnetic layer thickness. J. Appl. Phys. 88, 975–982 (2000). 13. Li, H. et al. Exchange enhancement and thermal anneal in Mn76 Ir24 bottom-pinned spin valves. J. Appl. Phys. 89, 6904–6906 (2001). 14. Kantola, M. & Tokola, E. X-ray studies on the thermal expansion of copper–nickel alloys. Ann. Acad. Sci. Fennicae Ser. A6: Physica 223, 1–10 (1967). 15. Ellis, W. C. & Greiner, E. S. Equilibrium relations in the solid state of the iron–cobalt system. Trans. Am. Soc. Metals 29, 415–432 (1941). 16. Hu, J-g., Jin, G-J. & Ma, Y-Q. Thickness and angular dependencies of exchange bias in ferromagnetic/antiferromagnetic bilayers. J. Appl. Phys. 92, 1009–1013 (2002). 17. Camarero, J. et al. Origin of the asymmetric magnetization reversal behavior in exchange-biased systems: Competing anisotropies. Phys. Rev. Lett. 95, 057204 (2005). 18. Greve, H. et al. Low damping resonant magnetoelectric sensors. Appl. Phys. Lett. 97, 152503 (2010). 19. Williamson, S. J. & Kauffmann, L. Biomagnetism. J. Magn. Magn. Mater. 22, 129–201 (1981).

ARTICLES 20. Dho, J. & Blamire, M. G. Competing functionality in multiferroic YMnO3 . Appl. Phys. Lett. 87, 252504 (2005). 21. Borisov, P., Hochstrat, A., Chen, X., Kleemann, W. & Binek, C. Magnetoelectric switching of exchange bias. Phys. Rev. Lett. 94, 117203 (2005). 22. Béa, H. et al. Tunnel magnetoresistance and robust room temperature exchange bias with multiferroic BiFeO3 epitaxial thin films. Appl. Phys. Lett. 89, 242114 (2006). 23. Wu, S. M. et al. Reversible electric control of exchange bias in a multiferroic field-effect device. Nature Mater. 9, 756–761 (2010). 24. Hauguel, T. et al. Experimental evidence for exchange bias in polycrystalline BiFeO3 /Ni81 Fe19 thin films. J Appl. Phys. 110, 073906 (2011). 25. Klokholm, E. The measurement of magnetostriction in ferromagnetic thin films. IEEE Trans. Magn. 12, 819–821 (1976). 26. Du Tremolet de Lacheisserie, E. & Peuzin, J. Magnetostriction and internal stresses in thin films: The cantilever method revisited. J. Magn. Magn. Mater. 136, 189–196 (1994). 27. Wetherhold, R. Magnetoelastic interaction in magnetostrictive spring-magnet multilayers. J. Magn. Magn. Mater. 269, 61–69 (2004). 28. Ludwig, A. & Quandt, E. Optimization of the 1E effect in thin films and multilayers by magnetic field annealing. IEEE Trans. Magn. 38, 2829–2831 (2002).

Acknowledgements The authors would like to thank J. McCord for fruitful discussions and the German Science Foundation DFG for financial support through the Collaborative Research Centre SFB 855 ‘ME Composite Materials—Biomagnetic Interfaces of the Future’. Many thanks go to C. Zamponi for accurate TEM sample preparation by focused ion beam technology.

Author contributions E.L., D.M. and E.Q. designed the experiment. C.K. and E.L. were responsible for the preparation and characterization of the Fe70.2 Co7.8 Si12 B10 and Fe50 Co50 samples, respectively. V.H. performed TEM measurements and data analysis. L.K. supervised the TEM measurements and data analysis. R.J. did the noise measurements under the supervision of R.K. E.L. and D.M. developed the angle concept of effective exchange bias and the multilayer stack. D.M. and E.Q. supervised the research. All authors contributed to the manuscript and the interpretation of the data.

Additional information The authors declare no competing financial interests. Supplementary information accompanies this paper on www.nature.com/naturematerials. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to D.M.



529