Magnetostrictive LC Circuit Sensors* 1

Transcription

1 Materials Transactions, Vol. 45, No. 2 (2004) pp. 244 to 248 Special Issue on Materials and Devices for Intelligent/Smart Systems #2004 The Japan Institute of Metals Magnetostrictive LC Circuit Sensors* 1 Eckhard Quandt* 2 and Michael Frommberger Center of Advanced European Studies and Research (caesar), Ludwig-Erhard-Allee 2, Bonn, Germany Magnetostrictive thin films in electric resonant LC circuits are attractive as novel strain sensors. In order to achieve high resonance frequencies exchange coupled nanometer multilayers were used. LC circuits incorporating Fe 50 Co 50 /Co 80 B 20 multilayers exhibited a gauge factor ((f =f Þ=") of the order of This LC circuit sensor enables wireless interrogation, which is demonstrated in a torque measurement. (Received November 6, 2003; Accepted January 14, 2004) Keywords: magnetostriction, high frequency, inductors, permeability, torque sensor 1. Introduction Highly sensitive sensors for strain or stress measurements are of increasing interest for industrial applications in areas like automotive, aerospace, process control, automation and biotechnology. In comparison to classical metallic or piezoresistive strain gauges, magnetostrictive thin film sensors show a higher sensitivity, almost no substrate limitations, and allow the fabrication of small sensor elements. Furthermore, the LC circuit sensor design allows wireless operation using either radar reflectivity or inductive coupling. Inverse magnetostriction (Villari effect) is a powerful transducer mechanism for strain sensors. As a result of the inverse magnetostriction the orientation of the magnetic domains changes in response to an applied strain. An illustration of this mechanism is given in Fig. 1. The figure shows domain rotation of small FeCoBSi discs on a Si substrate. The stress was applied perpendicular to the initial domain orientation and increased from left to right. The sense of the rotation is determined by the sign of both the magnetostriction and the strain and by the initial magnetic domain pattern. In materials with positive magnetostriction the saturation domain orientation will be preferably parallel or antiparallel to the direction of the tensile strain, while compressive strains will induce a perpendicular domain Fig. 1 Kerr images of the domain rotation of 500 nm thick FeCoBSi discs caused by applied stress. The discs are on a Si substrate and have a diameter of 200 mm. The stress was increased from left to right image, applied perpendicular to the initial domain orientation. (by kind permission of J. McCord, IFW Dresden). * 1 This research is funded by BMBF (03N3089) and ONR (N ). * 2 Corresponding author, quandt@caesar.de, +49 (0) Moment (norm.) max. stress decreased stress no stress H ext in mt Fig. 2 Vibrating sample magnetometer measurements of an FeCo/CoB multilayered thin film inserted in a bending test jig. By applying stress, the magnetization turns from hard to easy axis along with an important change in permeability. pattern. This relationship is reversed for materials with negative magnetostriction. In combination with other effects this change of the domain pattern can be transferred into an electrical signal. The rotation of the domains influences e.g. the magnetic permeability of the magnetostrictive material as shown in a VSM measurement in Fig. 2. This change in permeability can also be detected by impedance- 1,2) or inductivity measurements. 3) As the permeability determines the skin depth of a magnetic material, the impedance is changed as a function of the applied strain. Accordingly if the magnetostrictive material serves as core of a coil its change in permeability leads to a change in inductance, or, if the sensor is part of a LC circuit, to a change of its resonance frequency. Stress sensitive sensors may also be built by exploiting magnetoresistive (MR) measurements. The resistance in AMR- (anisotropic MR 4) ) but even more the GMR- (giant MR 5 8) ) or TMR- (tunnel MR 9) ) devices is dependent on the orientation of the magnetic domains. Replacing part of the magnetic material by a magnetostrictive one results in very sensitive strain gauges, especially in the case of TMR junctions. In this paper LC circuit strain sensors based on magnetostrictive thin films will be discussed and compared to alternative strain gauges.

2 Magnetostrictive LC Circuit Sensors Experimental The devices presented here were fabricted by magnetron sputtering using an Ardenne CS cluster tool and subsequent photo lithographic steps using a Süss MA6 mask aligner and, in case of dry etching, an Oxford ion beam etcher. The films were deposited on intentionally unheated substrates using dcor rf-magnetron sputtering. In most cases a static in-plane magnetic bias field of approximately 8 ka/m was applied during the deposition. Furthermore, some of the films were annealed in a dc magnetic field in order to induce magnetic anisotropy while other samples were annealed without the presence of a magnetic field in order to release initial stresses. The magnetic properties were determined using a Lake Shore 735 vibration sample magnetometer (VSM) as well as cantilever based devices for measuring magnetostriction 10) The high frequency permeability was determined using a strip-line permeameter. 11) The sensor itself was characterized by means of specially designed test jigs. Magnetic domain patterns as a function of the applied strain were investigated in cooperation with R. Shulls group at NIST, Gaithersburg, USA, using the MOIF method 12) and with R. Schäfers group at IFW, Dresden, Germany, using Kerr microscopy. 13) A small in situ 4-point-bending device was used to apply well defined strains. The microstructure of the films as well as of the devices was determined using a focused ion beam system (FIB, Leo 1540X cross beam). The composition of the films was determined by energy dispersive x-ray measurements (EDX) or in the case of multilayers by Auger electron spectroscopy (AES) using a Physical Electronics 690 Auger-Nanoprobe. 3. Results and Discussion The main requirements for magnetostrictive materials to be used as cores in microinductors are (1) a cut-off frequency well above the operating frequency of the device, (2) low losses at the operating frequency in order to achieve high quality factors, and (3) a significant inverse magnetostrictive effect at the operating frequency. In order to achieve high cut-off frequencies the magnetic easy axis has to be aligned such that the magnetic field is parallel to the hard axis of the material. In this configuration magnetic reversal is obtained by domain rotation only that, in contrast to domain wall motion, can respond to fields even in the GHz regime. If the rf magnetic field is applied in-plane but along the hard axis of the thin film the maximum possible operating frequency is limited by the ferromagnetic resonance frequency while the bandwidth of thicker materials is controlled by eddy current losses. The ferromagnetic resonance frequency is approximately given by f res ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðh k þ H Bias ÞM S 2 where is the gyromagnetic constant, H K the anisotropy field, H Bias a superimposed dc magnetic field, and M S the saturation magnetization. This frequency limitation is not independent of the permeability, which, along the hard axis, is approximately given by ha ¼ M S : H K Therefore, the main practical engineering parameters for high frequency magnetic materials for tayloring the ferromagnetic resonance frequency and permeability are the anisotropy field (H K ) and the saturation magnetization (M S ). In previous investigations it was found that exchange coupled CoB/FeCo multilayers with individual layer thicknesses in the range of 1 15 nm are a very attractive material system for tailoring of high frequency properties in a wide range. 11,14) As an example, the increase of the cut-off frequency of such a multilayer in comparison to a CoB single layer having both the same permeability is shown in Fig. 3. It can be seen that at constant permeability, the ferromagnetic resonance frequency could be shifted from 1.5 GHz for the single layer to over 3 GHz for the multilayer. At the same time, the damping in the multilayered film increased slightly. The change of the real part of ac permeability determined at 1 GHz as a function of the applied strain is shown in Fig. 4. Due to the positive magnetostriction in this material tensile strain results in a gradual rotation of the easy axis into the rf ac permeability µ' FeCo/CoB µ' CoB film thickness: 250nm frequency in GHz Fig. 3 AC permeability of a 250 nm thick FeCo/CoB multilayer film in comparison to a CoB single layer film. While the permeability remains, the ferromagnetic resonance frequency could be shifted from 1.5 GHz to about 3.1 GHz by layering CoB with FeCo. µ ac 1 GHz compressive tensile ε in % Fig. 4 Dependence of the ac permeability of an FeCo/CoB multilayer on applied strain ( 0 was determined at 1 GHz).

3 246 E. Quandt and M. Frommberger field direction. For low strains this rotation leads to a certain permeability. A strain of approximately 0.04% results in a spin flip into the rf field direction and the permeability suddenly vanishes, as now the easy axis is parallel to the exciting rf-field. The strain that determines the permeability maximum depends on material and design parameters, i.e. shape anisotropy and internal stresses. In response to a compressive strain the easy axis becomes more stable in the direction perpendicular to resulting in a gradual decrease of the permeability. The upper frequency limit due to eddy current losses is given by: f ec ¼ 4 0 r t 2 where t is the layer thickness and the resistivity of the magnetic material. For typical values of the resistivity of amorphous magnetostrictive material this frequency limit is responsible for the cut-off frequency if the t 1:5 mm ( ¼ 300, ¼ 100 mcm). Therefore, an increase in Q can be obtained by lamination of the magnetostrictive material (e.g. CoB/FeCo multilayers) with additional insulating layers (e.g. SiO 2 ) in a kind of transformer core design. 15,16) The benefit of this approach is demonstrated by a comparison of two 2 mm CoB/FeCo multilayers, one as a solid metallic film without SiO 2 interlayers, the other divided into 25 layers separated by 100 nm SiO 2 layers (Fig. 5). For frequencies higher than 50 MHz the SiO 2 layering results in a significant increase of the real part of the permeability ( 0 ). At the same time the imaginary part ( 00 ) decreases up to MHz leading to an important increase of the material quality factor Q ¼ 0 = 00. This factor limits the overall quality factor that can be obtained when using magnetostrictive LC circuits. The magnetostrictive multilayer thin films discussed above have been integrated as sensing elements in a LC circuit fabricted in thin film technique. These LC circuits consist of a strip inductor (Au, 100 mm wide, mm long) surrounded by magnetostrictive layers, which are isolated against the conductor by SiO 2 layers. The values of the inductance and the capacitance were adjusted to yield a LC quality factor µ'/µ'' µm FCCB 25x 80nm FCCB/100nm SiO frequency in GHz Fig. 5 Frequency dependence of the quality factor of a 2 mm thick FeCo/ CoB multilayer and a (FeCo/CoB)/SiO 2 laminated film with equal magnetic volume. Fig. 6 Scanning electron microscopy image of a cross section of the strip inductor made by focused ion beam milling. The image shows the inner conductor, isolating and magnetic layers. Fig. 7 Scanning electron microscopy image of the segmented strip inductor (left) and the corresponding mask layout (right). resonance frequency of approximately 500 MHz. The overall device quality factor at this frequency equaled approximately 9, about the same value as for the material Q itself at this frequency. The magnetic easy axis was oriented parallel to the strip. For the present sensor the magnetostrictive layers consist of 25 CoB/FeCo multilayers of 80 nm thickness each laminated by 50 nm SiO 2 layers (Fig. 6). Further reduction of eddy currents can be obtained by lateral patterning of the magnetostrictive material as shown in Fig. 7. Here, the magnetic material surrounding the inductor strip was segmented to eliminate current flow parallel to the strip. Applying a tensile stress perpendicular to the strip inductor results in a gradual rotation of the magnetic domains into the stress direction. To measure this influence of strain on the LC circuits the devices were mounted into a 4-point bending test jig (Fig. 8). A pusher block controlled by a linear stepping motor, allowed a defined bending of the device. The sensor is remotely interrogated by an excitation and a pick-up coil mounted at a distance of approximately 2 mm. The resonance frequency is determined from the transmitted signal (S 21 ) using a network analyzer (Rohde&Schwarz Vector Network Analyzer 20 khz 8 GHz ZVCE). Figure 9 shows the dependence of the resonance frequency of the LC circuit on the applied stress. Corresponding to the behavior of the permeability shown in Fig. 4, the permeability increased first, leading to an increase of the inductance and consequently a decrease of the resonance frequency. At a certain strain (" ¼ 0:03%, u no external magnetic bias) the permeability suddenly drops to 1 resulting in a corresponding

4 Magnetostrictive LC Circuit Sensors 247 resonance frequency in MHz and axis 0 axis turned by 180 Fig. 8 Sketch of the strain measurement set-up consisting of a block pushing against a printed circuit bord. The LC circuit is placed on the 4- point bending jig and excited and read out by two small coils. A magnetic bias field can be applied in different directions by a pair of Helmholtz coils. circuit resonance frequency in MHz / ε = 543 / ε = 991 ε / ε = G H ext perpend. to EA no H ext 19.91G H ext parallel to EA ε in % Fig. 9 Dependence of the resonance frequency of the LC circuit on applied stress for different external magnetic fields. A figure of merit of about 1000 could be obtained with devices like this one. increase of resonance frequency. By applying a magnetic bias field or bias stress (not shown) the characteristic of the resonance frequency on applied stress (Fig. 9) can be shifted resulting in e.g. a gradual increase of resonance frequency with the applied tensile strain (in case of a magnetic bias field of 20 G perpendicular to the initial easy axis). A figure of merit (FoM) of the LC circuit given by the relative change in frequency (f =f 0 ) over the change in strain (") can be defined as FoM ¼ f =": f 0 Values exceeding 1000 have been obtained depending on the bias field and the range of strain. These values are extraordinarily high compared to the resistive type of strain gauges with an equivalent gauge factor FoM ¼ R =" R 0 where R=R 0 denotes the relative change in resistance. f 0 f Maximum gauge factors are 4 for metallic, 180 for piezoresistive and 600 for magnetostrictive TMR strain gauges. 9) Another advantage of LC circuit strain gauges is that the measurement of frequencies instead of analogous signals is significantly less sensitive to disturbances. A possible application for this type of sensor is the measurement of torque. Figure 10 shows preliminary data for a magnetostrictive LC circuit attached to a non-magnetic steel shaft. The data were obtained using the inductive coupling method similar to the read-out scheme shown in Fig Conclusions A LC circuit type strain sensor featuring wireless interrogation has been developed and fabricated. The prototype sensor uses exchange coupled CoB/FeCo multilayers with additional isolating layers as sensing element. The devices are displaying quality factors of about 9 at resonance frequencies of approximately 500 MHz. In strain measurements a gauge factor of about 1000 could be measured. The feasibility of this sensor principle has been demonstrated for torque measurements. Future work will concentrate on design optimization of the sensor element, further improvement of the magnetostrictive material and on the development of a suitable read-out electronic. Acknowledgements The authors acknowledge fruitful discussions with Alfred Ludwig, and Michael Tewes (caesar), and Manfred Wuttig (University of Maryland). The authors would like to thank Jeff McCord (IFW Dresden, Germany) for the domain observations. REFERENCES torque in Nm Fig. 10 Results of a torque measurement using stress sensitive LC circuits. The dependence of the resonance frequency of the LC circuit on applied torque up to 200 Nm is shown. 1) L. P. Shen, T. Uchiyama, K. Mohri, E. Kita and K. Bushida: IEEE Trans. Magn. 33 (1997) ) K. Mohri, T. Uchiyama, L. P. Shen, C. M. Cai and L. V. Pawin: Sensors and Actuators A91 (2001) 85. 3) C. H. Tyrén, D. G. Lord, United States Patent No (1994).

5 248 E. Quandt and M. Frommberger 4) B. B. Pant and D. R. Krahn: J. Appl. Phys. 69 (1991) ) R. C. O Handley and A. R. Childress: IEEE Trans. Magn. 31 (1995) ) M. J. Mawin, G. A. Gurney, D. R. Wilhout and V. S. Speriosa: Appl. Phys. Lett. 72 (1998) ) L. Basil, B. Gurney, D. Wilhout and V. S. Speriosa: J. Appl. Phys. 85 (1999) ) T. Duenas, A. Sehrbrock, M. Löhndorf, A. Ludwig, J. Wecker, P. Grünberg and E. Quandt: J. Magn. Magn. Mater (2002) ) M. Löhndorf, T. Duenas, M. Tewes, E. Quandt, M. Rührig and J. Wecker: Appl. Phys. Lett. 81 (2002) ) E. Quandt and A.E. Clark: Proc. Actuator 98, ) A. Ludwig, M. Tewes, S. Glasmachers, M. Löhndorf and E. Quandt: J. Magn. Magn. Mater (2002) ) R. D. Shull, E. Quandt, A. J. Shapiro and M. Wuttig: submitted to IEEE Trans. Magn. 13) J. McCord, R. Schäfer, M. Frommberger, S. Glasmachers and E. Quandt: submitted to the joint Intermag/MMM conference (2004). 14) A. Ludwig: Appl. Surf. Sci. (2003) in print. 15) M. Senda and O. Ishii: IEEE Trans. Magn. 30 (1994) ) M. Frommberger, A. Ludwig, A. Sehrbrock and E. Quandt: proceedings of Intermag 2003, IEEE Trans. Magn. (2003) in print.