WorksPhysics and Astronomy
Magnetic Field-Controlled Mixed Modulation in Magnetoelectric Sensors
Abstract
Magnetoelectric (ME) magnetic field sensors commonly rely on one of the two modulation principles: the nonlinear dependence of magnetostrictive strain on the applied field or the stress-induced change in magnetization susceptibility. While both effects coexist in any ME device, different readout schemes can be chosen to utilize one or the other effect for magnetic field sensing. This work demonstrates that both principles can be simultaneously implemented in a single electrically modulated ME sensor with inductive readout (a converse ME sensor). This mixed modulation approach significantly enhances low-frequency sensitivity while not affecting the sensitivity at higher frequencies. This leads to a nontrivial dependency of the sensor sensitivity on the frequency of the magnetic field to be measured and can effectively decrease the sensor bandwidth by up to an order of magnitude. We show that the contribution of the modulation from the nonlinearity of the magnetostrictive strain to the sensor sensitivity can be changed by applying a magnetic bias field, offering an additional dimension to the design of ME sensors, especially for potential applications in the unshielded environment.
Introduction
Magnetic field sensors play a crucial role in a wide range of modern applications, including wireless communication systems[1] and biomedical diagnostics[2]. The latter often require detection of extremely weak magnetic fields at low frequencies below 10 kHz with a bandwidth of up to 1 kHz. [3] Recent decades saw substantial progress in the development of magnetic field sensors that target these applications, including sensor concepts utilizing magnetoelectric (ME) composite materials.[4, 5, 6] Such materials comprise mechanically coupled piezoelectric and magnetostrictive components, allowing the conversion of magnetic fields into electrical voltage (direct ME effect) and vice versa (converse ME effect) with a much higher conversion factor in comparison to the single-phase ME materials.[7]
Various magnetic field sensor concepts based on the ME effect have been presented; many of them utilize resonator structures, such as cantilevers.[4] A typical ME cantilever consists of three main components: a piezoelectric layer sandwiched between two electrodes, a silicon substrate, and a magnetostrictive layer on the rear side of the substrate. The sensitivity of the sensors based on the direct ME effect is significantly amplified when the frequency of the magnetic field to be measured matches the resonance frequency of the sensor element. Extremely low limits of detection of a few pT/ have been achieved in this operation mode,[8] however, restricted to narrow-band signals near .
Modulation techniques based on the nonlinearity of magnetostriction have been employed to detect low-frequency, low-amplitude magnetic fields over broader frequency ranges.[9]
Examples include concepts based on the and converse ME effects.[4] These concepts correspondingly employ the dependency of the magnetostrictive strain[10] and the susceptibility of magnetization to stress[11] on the applied magnetic field. Although both modulation techniques have been comprehensively studied in the literature[9, 10, 11, 12, 13], the case of mixed modulation and its effect on the signal and bandwidth of converse ME sensors was not investigated. Yet, it is expected that the effect in the magnetostrictive layer will lead to the mixed modulation of magnetization oscillation in converse ME sensors and, by that, make some contribution to the sensitivity of such sensors. Here, we explain how significant this contribution is and how it changes depending on the applied magnetic bias field. Furthermore, we demonstrate its crucial effect on the sensor sensitivity as a function of the frequency of the measured field, which ultimately determines the bandwidth of the magnetic field sensor and results in a complex dependency of the bandwidth on the bias field.
Results
Sensor Concept and Mixed Modulation Principle
Modulation Principle based on the effect
The effect manifests in the dependency of the stiffness tensor of the magnetostrictive material on the magnetostrictive strain and its nonlinear dependency on the magnetic field.[10]
In the case of cantilever-based sensors, the change in the stiffness of the magnetostrictive layer causes a change of , which can be read out optically,[14] electrically,[15, 16] or inductively.[17]
For low amplitudes of an alternating field , oscillation of around can be approximated as
with the time . The shift of with can be calculated via the change of the compliance tensor .[13]
where is the mechanical strain. For the bending modes, is determined by . Hence,
where is the susceptibility of magnetization to stress and is saturation magnetostriction constant. The sensitivity of the magnetic field sensors based on the effect is determined by the change of its resonance frequency, which means that the mechanical bandwidth of the cantilever determines the sensor bandwidth.[15]
Modulation Principle based on the Converse ME Effect
An alternative approach that allows broadband signal detection is based on the converse ME effect and the inductive readout of the magnetization change. An illustration of a converse ME sensor and the coordinate system used in this letter is shown in the inset of Figure 1a. In this operation mode, the cantilever is excited at its mechanical resonance by applying the excitation voltage between the electrodes with a frequency . The induced stress in the magnetostrictive layer causes the oscillation of magnetization . If the amplitude of is small, the oscillation of can be described via the susceptibility . The external magnetic field modulates leading to the modulation of as illustrated in Figure 1b. The resulting oscillation of magnetization can be expressed as[11]

Figure 1. Illustration of the sensor concept. (a) Calculated dependency of and that correspondingly determine the sensitivity of converse ME and -effect sensors on the magnetic bias field . The dependencies are calculated for two orientations of the easy axis (EA) of the uniaxial magnetic anisotropy: along the -axis (solid lines) and tilted by relative to the -axis (dashed lines). Inset: the coordinate system used throughout the paper and a schematic illustration of the sensor element comprising the ME cantilever and the pickup coil. (b) Converse ME effect: applied stress induces oscillation of magnetization proportional to modulated by . The amplitude spectrum of the signal read out with the pickup coil is shown on the right. The main peak in the center is the carrier with an amplitude , the amplitude of the sidebands determines the sensitivity of the sensor to the magnetic field. This case illustrates amplitude modulation (AM) of the carrier. (c) effect: shifts the frequency dependency of the carrier magnitude and phase . This leads a amplitude (AM) and phase (PM) modulation of illustrated with the green arrows. (d) Mixed modulation case. The effect causes amplitude and phase modulation (AM+PM) of , which induces oscillation of modulated by . Amplitude spectrum of the output signal is shown on the right. AM of creates harmonic distortion, i.e., additional signal peaks at , PM of increases the main signal peak by adding additional contribution .
This oscillation of magnetization is then read out inductively with a pickup coil.[18, 19] This sensor concept significantly differs from the other ME sensor concepts because in this case, the readout signal is determined by the magnetization change and not the magnetostrictive strain. This removes the restriction on the sensor bandwidth imposed by resonator bandwidth. Reported values of the sensor bandwidth can reach several kilohertz, which is ten times larger than the corresponding resonator bandwidth.[6, 11] In the following, we show that this is true only within a narrow range of the operation parameters, within which the effect is negligible.
Because of the geometry of the sensor element and the used resonance mode, often referred to as the U-mode,[5, 11, 19, 20] we can consider only the -axis components of and , and the -axis component of and , i.e., in Eq. (4) and in Eq. (3).[11]
Oscillation of magnetization is recorded using a pickup coil connected to a low-noise operational amplifier, which is then connected to the input of a lock-in amplifier. The same lock-in amplifier provides an auxiliary output for sensor excitation. Details about the setup have been presented elsewhere.[5, 11, 19] The resulting output voltage at the input of the lock-in amplifier can be approximated as
where is a proportionality factor determined by the geometry of the resonator, the pickup coil, and readout electronics. For the device investigated here, mVs. The detailed derivation and validation of Eq. (4)-(5) and the coefficient is provided in Ref. [11] as a part of the comprehensive sensor system model.
The first term in Eq. (4) describes the oscillation of magnetization caused by the applied stress and determines the root-mean-square (RMS) amplitude of the carrier signal (denoted on the right in Figure 1b). The second term in Eq. (4) describes the modulation of caused by . It represents the modulation of the carrier and the corresponding amplitude of the sidebands visible in the frequency domain, as illustrated in Figure 1b. Because is linearly proportional to , the dependency of on the bias field can be qualitatively analyzed by calculating the dependency of on the bias field . The dependency calculated using a macrospin model is plotted on the left vertical axis of Figure 1a. Details of the model and material parameters used are provided in the Methods.
The easy axis (EA) of uniaxial magnetic anisotropy is set along the -axis (solid lines) and tilted by relative to the -axis (dashed lines). The tilt of the EA is introduced to mimic the limited manufacturing control and a possible slight deviation from the exact orientation of EA along the -axis. Figure 1a shows that if the EA is parallel to the -axis, and, hence, the sensitivity of the converse ME sensor should stay constant within a certain range of . The tilt of EA causes the tilt of the curve.
With the same macrospin model, the dependency of , which determines the sensitivity of the -effect sensors (Eq. (3)) on the bias field is calculated and plotted in Figure 1a. In contrast to , increases almost quadratically with , which results in the typical dependency of on for such samples [13, 21]. The tilt of the EA causes the shift of the minimum of the curve.
Fig 1a highlights the main difference between the ME sensor concepts whose sensitivity is determined by the magnetostrictive strain and those whose sensitivity is determined by the magnetization change. Under ideal conditions, in the sensor configuration considered here, the sensitivity is nearly quadratic with in the first case and linear in the second case. In the following, we show that the effect contributes significantly to the signal of converse ME sensors, leading to a nontrivial dependency of the sensor signal and bandwidth on the bias field.
Mixed Modulation. Combination of the Converse ME and Effects
To include the contribution of the effect to the sensor signal, we need to consider the modulation of the resonance frequency of the cantilever induced by . The modulation of results in the shift of the frequency dependency of the carrier magnitude and phase as illustrated in Figure 1c. Consequently, if the excitation frequency is kept constant, the modulation of leads to amplitude (AM) and phase modulation (PM) of (highlighted with a green arrow in Figure 1c). If the cantilever is excited exactly at its resonance frequency, the corresponding modulation of (Figure 1d) can be expressed as
Here, is the amplitude of , is a phase lag between and for , is amplitude sensitivity, and is the phase sensitivity expressed as
The amplitude modulation of creates harmonic distortion, visible as additional signal peaks at in fig:1d. For the sake of clarity, we restrict our discussion to the impact of the effect on the main signal peak and assume .
Plugging (7) into (4) and (5) gives the final expression for the output signal that combines both the converse ME and the effects, as illustrated in fig:1d
can be calculated with, e.g., a finite-element model of the magnetoelectrical resonator, as was described in [22]. However, to avoid these computationally expensive calculations, can be estimated from measurements via
where is the measured carrier signal of the sensor. The contribution of the effect to can be then estimated as
where denotes lock-in demodulation of the time-domain signal and selection of its value at .
Influence of the Mixed Modulation on the Signal and Bandwidth of the ME Sensor
In this section, we will demonstrate how the described contribution of the effect affects the output signal of a converse ME sensor.
Signal
The measured sideband amplitude as a function of the magnetic bias field is shown in Figure 2a. A test field with nT and Hz was used for the measurements. Figure 2a shows that reaches its minimum of µV at mT and increases with increasing reaching µV at mT. The contribution of the effect estimated from the measurements with Eq. (10) and the corresponding change in the resonance frequency is also plotted in Figure 2a. follows the same trend as , showing its minimum of µV at mT and reaching a maximum of µV at mT. The resonance frequency shows the opposite trend, reaching the maximum at mT and decreasing with increasing . Such dependency of on is expected from the effect for the current configuration of the uniaxial magnetic anisotropy and the applied anisotropic stress axis.[13] The increasing change of with leads to higher (Eq. (7),(10)), which exceeds 50% of at mT.

Figure 2. Impact of the effect on the signal of the converse ME sensor as a function of the bias field . (a) Left axis: measured sideband amplitude for V and the contribution of the effect to estimated with Eq. (10). Right axis: resonance frequency of the sensor element. (b) Simulated sideband amplitude with and without considering the effect.
Additionally, we performed a simulation of with Eq. (7) using the same sensor signal model as in Ref. [11]. The phase sensitivity is estimated from the measurements using Eq. (9). The sideband amplitude was then calculated as . The resulting dependencies of on for two cases, with and without the effect, are plotted in Figure 2b. While without the effect, barely changes with , with the effect, increases with increasing field magnitude following the increasing , as expected from Fig, 1a. This increase of perfectly resembles the trend in the measured data (Figure 2a). Overall, the data indicate that the contribution of the effect can increase the output signal of the sensor system.
Bandwidth
According to Eq. (4), the bandwidth of the converse ME sensors should be determined by the frequency dependency of the susceptibility i.e., by the magneto-mechanical loss in the magnetostrictive layer.[11] At the same time, the bandwidth of the -effect sensors is determined by the effective loss of the cantilever, i.e., the mechanical resonator as a whole.[15] Consequently, the bandwidth of the -effect sensors is narrower in comparison to the converse ME sensors.[6, 11] However, if the effect dominates the signal of the converse ME sensor, the bandwidth of the mechanical resonator should become the dominating factor determining the sensor bandwidth.
To prove this hypothesis, we measured both sensor and the resonator bandwidth over the bias field . To estimate the mechanical bandwidth of the resonator , the carrier amplitude is measured as a function of the excitation frequency , assuming that is directly proportional to the displacement of the cantilever. The result of the measurement performed at V and mT is plotted in Figure 3a. The mechanical bandwidth of the resonator, estimated as half-width at half-power[23] is Hz.

Figure 3. Impact of the effect on the sensor bandwidth. (a) Carrier heights as a function of the excitation frequency . The resonator bandwidth calculated as half of the half-power bandwidth of is highlighted in green. (b) Dependency of the sensor signal as a function of the magnetic field frequency recorded at different magnitudes of the magnetic bias field. The values are normalized to for each value individually. (c) Dependency of the sensor bandwidth and the contribution of the effect to the sensor signal on the bias field .
The sensor bandwidth is estimated by sweeping the frequency of the applied alternating field with nT and measuring the height of the signal peak . The resulting dependency normalized to recorded at mT is plotted in Figure 3b. At mT, the dependency shows the expected low-pass behavior.[25, 28] At mT and mT, the shape of the curve changes, showing two overlapping resonators with different bandwidths.
The resonator with the smaller bandwidth represents the mechanical resonator, and the resonator with the larger bandwidth represents the mechanically driven magnetic resonator. At mT, the contribution of the effect becomes dominant, and the sensor bandwidth reaches its minimum value equal to the bandwidth of the mechanical resonator.
From the measured dependency a -3 dB bandwidth is calculated and plotted in Figure 3b as a function of the bias field . For comparison, the contribution of the effect to the sensor signal calculated as is plotted on the right vertical axis of Figure 3b. Around mT, the sensor bandwidth reaches its maximum of Hz. At the same , the contribution of the effect is the lowest with %. With the increasing contribution of the effect, the bandwidth decreases, reaching its minimum level of Hz at mT. For mT the contribution of the effect exceeds 50%, resulting in the narrow sensor bandwidth Hz (marked red in Figure 3b). The same correlation between increasing and decreasing is also visible in the negative fields . The curve is asymmetrical around mT because of the tilt of the magnetic anisotropy axis, as was mentioned before. A clear correlation between the effect and the sensor bandwidth is visible, confirming our hypothesis.
Conclusion
This paper demonstrates that the effect plays a crucial role in the performance of magnetoelectric sensors with inductive readout (converse ME sensors). During the operation of such sensors, the effect causes amplitude and phase modulation of the stress within the magnetostrictive layer, leading to a mixed modulation of magnetization oscillation. It results in an increase of the signal at low frequencies and a nontrivial frequency dependency of the output signal, leading to a sensor bandwidth that strongly depends on the magnetic bias field.
Notably, in the case of the investigated sensor, the volume fraction of the magnetostrictive material relative to the substrate and other functional layers is about three times smaller compared to reported -effect sensors.[15, 16, 22, 24] Still, as we have shown, the impact of the effect cannot be neglected. This implies that magnetoelastic nonlinearities and other effects connected with the effect[22, 25] are also relevant for designing converse ME sensors, especially if they are intended to be used in an unshielded environment. Design strategies investigated for -effect sensors[24, 26] could be transferred to converse ME sensors to tune the contribution of the effect and potentially benefit from this second signal pathway. For suppressing the contribution of the effect, higher thickness ratios of the substrate to the magnetostrictive layer might be considered. Likewise, to achieve a higher magnetic field tunability of the bandwidth, a larger contribution of the effect to the signal would be beneficial, which can be achieved by reducing the substrate thickness. Increasing the thickness of the magnetostrictive layer can increase the signal contribution from the converse ME and the effect.
Additionally, the results show the possibility of tuning the bandwidth and signal output of the sensor system by carefully adjusting operational parameters, which offers an additional dimension to the design of ME sensors. Hence, this work contributes to the fundamental understanding of magnetoelectric magnetic field sensors and marks a significant step towards tailoring magnetoelectric devices to specific applications in the future.
Methods
Signal measurement. The signal measurements were performed with a lock-in amplifier HF2LI from Zurich Instruments and a low-noise operational amplifier LT1363 (from Analog Devices) in non-inverting op-amp configuration. The alternating magnetic field and static magnetic field were created with two coaxial solenoids and two low-noise current sources (6221, Keithley Instruments, Solon, USA). The measurements were performed in a magnetically, electrically, and acoustically shielded environment, including an additional multilayer mu-metal cylinder (Model ZG1, Aaronia).[27]
Macrospin model. The macrospin model used for calculating and includes uniaxial magnetic anisotropy, stress-induced anisotropy, and Zeeman energy. Details of the model are provided elsewhere.[11] The saturation magnetostriction constant , effective magnetic anisotropy constant , and magnetic anisotropy angle relative to are set to ppm,[28] J/m, and .
Acknowledgements
The research was supported by the German Research Foundation (DFG) through the Collaborative Research Centre CRC 1261 “Magnetoelectric Sensors – From Composite Materials to Biomagnetic Diagnostics” and by the Carl Zeiss foundation via the project “Memristive Materials for Neuromorphic Electronics” (MemWerk).
We would like to thank our colleagues Jingxiang Su and Florian Niekiel from Fraunhofer ISITS in Itzehoe for deposition of the piezoelectric AlN and providing the silicon wafers as part of the CRC 1261.
References
- [1]a) T. Nan, H. Lin, Y. Gao, A. Matyushov, G. Yu, H. Chen, N. Sun, S. Wei, Z. Wang, M. Li et al., Nat. Commun. 2017, 8, 296; b) C. Dong, C. Sun, L. Chen, Y. He, Y. Liu, B. Luo, N. X. Sun, Materials Today Electronics 2025, 100135.
- [2]a) S. Kopyl, R. Surmenev, M. Surmeneva, Y. Fetisov, A. Kholkin, Materials Today Bio 2021, 12, 100149; b) H. Heidari, Magnetic Sensors for Biomedical Applications, John Wiley & Sons Incorporated, Newark, 2020; c) K. Zhu, A. Kiourti, IEEE Open J. Antennas Propag. 2022, 3, 732.
- [3]a) E. Elzenheimer, C. Bald, E. Engelhardt, J. Hoffmann, P. Hayes, J. Arbustini, A. Bahr, E. Quandt, M. Höft, G. Schmidt, Sensors 2022, 22, 1018; b) K. Sternickel, A. I. Braginski, Supercond. Sci. Technol. 2006, 19, S160-S171.
- [4]X. Liang, A. Matyushov, P. Hayes, V. Schell, C. Dong, H. Chen, Y. He, A. Will-Cole, E. Quandt, P. Martins et al., IEEE Transactions on Magnetics 2021, 57, 1.
- [5]L. Thormählen, P. Hayes, E. Elzenheimer, E. Spetzler, G. Schmidt, M. Höft, J. McCord, D. Meyners, E. Quandt, Appl. Phys. Lett. 2024, 124.
- [6]E. Elzenheimer, P. Hayes, L. Thormählen, E. Engelhardt, A. Zaman, E. Quandt, N. Frey, M. Höft, G. Schmidt, IEEE Sensors Journal 2023, 23, 5660.
- [7]J. van den Boomgaard, A. M. J. G. van Run, J. V. Suchtelen, Ferroelectrics 1976, 10, 295.
- [8]P. Durdaut, S. Salzer, J. Reermann, V. Röbisch, P. Hayes, A. Piorra, D. Meyners, E. Quandt, G. Schmidt, R. Knöchel et al., IEEE Sensors Journal 2017, 17, 2338.
- [9]Y. K. Fetisov, G. Srinivasan, Journal of Applied Physics 2024, 135.
- [10]E. W. Lee, Rep. Prog. Phys. 1955, 18, 184.
- [11]E. Spetzler, B. Spetzler, D. Seidler, J. Arbustini, L. Thormählen, E. Elzenheimer, M. Höft, A. Bahr, D. Meyners, J. McCord, Advanced Sensor Research 2024, 2400109.
- [12]a) N. Tiercelin, V. Preobrazhensky, P. Pernod, H. Le Gall, J. Ben Youssef, J. Magn. Magn. Mater. 2000, 210, 302; b) X. Zhuang, M. L. C. Sing, C. Cordier, S. Saez, C. Dolabdjian, L. Shen, J. F. Li, M. Li, D. Viegland, IEEE Sensors Journal 2011, 11, 2266.
- [13]B. Spetzler, E. V. Golubeva, R.-M. Friedrich, S. Zabel, C. Kirchhof, D. Meyners, J. McCord, F. Faupel, Sensors 2021, 21, 2022.
- [14]Z. Zhang, H. Wu, L. Sang, Y. Takahashi, J. Huang, L. Wang, M. Toda, I. M. Akita, Y. Koide, S. Koizumi et al., ACS applied materials & interfaces 2020, 12, 23155.
- [15]S. Zabel, C. Kirchhof, E. Yarar, D. Meyners, E. Quandt, F. Faupel, Appl. Phys. Lett. 2015, 107, 152402.
- [16]R. Jahns, S. Zabel, S. Marauska, B. Gojdka, B. Wagner, R. Knöchel, R. Adelung, F. Faupel, Appl. Phys. Lett. 2014, 105.
- [17]Z. Chu, C. Dong, C. Tu, X. Liang, H. Chen, C. Sun, Z. Yu, S. Dong, N.-X. Sun, Appl. Phys. Lett. 2019, 115.
- [18]M. D. Mermelstein, Electronics Letters 1986, 22, 525.
- [19]P. Hayes, M. Jovičević Klug, S. Toxvaerd, P. Durdaut, V. Schell, A. Teplyuk, D. Burdin, A. Winkler, R. Weser, Y. Fetisov et al., Sci Rep 2019, 9, 16355.
- [20]J. Schmalz, E. Spetzler, J. McCord, M. Gerken, Sensors 2023, 23, 5012.
- [21]B. Spetzler, E. V. Golubeva, C. Müller, J. McCord, F. Faupel, Sensors 2019, 19, 4769.
- [22]E. Spetzler, B. Spetzler, J. McCord, Advanced Functional Materials 2023, 34, 2309867.
- [23]W. O. Davis, Journal of Microelectromechanical Systems 2011, 20, 968.
- [24]A. D. Matyushov, B. Spetzler, M. Zaeimbashi, J. Zhou, Z. Qian, E. V. Golubeva, C. Tu, Y. Guo, B. F. Chen, D. Wang et al., Adv. Mater. Technol. 2021, 6, 21000294.
- [25]a) Y. K. Fetisov, D. A. Burdin, N. A. Ekonomov, L. Y. Fetisov, A. A. Berzin, P. Hayes, E. Quandt, Appl. Phys. Lett. 2018, 113; b) Z. Chu, C. Dong, C. Tu, Y. He, X. Liang, J. Wang, Y. Wei, H. Chen, X. Gao, C. Lu et al., Phys. Rev. Appl. 2019, 12, 44001.
- [26]a) B. Spetzler, C. Kirchhof, J. Reermann, P. Durdaut, M. Höft, G. Schmidt, E. Quandt, F. F upangel, Appl. Phys. Lett. 2019, 114; b) B. Spetzler, C. Kirchhof, E. Quandt, J. McCord, F. Faupel, Phys. Rev. Appl. 2019, 12, 64036.
- [27]R. Jahns, R. Knöchel, H. Greve, E. Woltermann, E. Lage, E. Quandt in 2011 IEEE International Workshop on Medical Measurements and Applications proceedings. (MeMeA 2011) ; Bari, Italy, 30 - 31 May 2011 (Ed.: M. Savino), IEEE, Piscataway, NJ, 2011, pp. 107–110.
- [28]A. Ludwig, E. Quandt, IEEE Trans. Magn. 2002, 38, 2829.