Mechanically driven SMR-based NEMS magnetoelectric antennas

—Mechanically driven magnetoelectric (ME) antennas have been demonstrated to be one of the most effective methods to 14 miniaturise antennas compared to state-of-the-art compact antennas. However, the nanoelectromechanical systems (NEMS) ME antennas 15 are fragile due to their suspended thin-film heterostructure, and have very low power handling capabilities. Here we show that solidly 16 mounted resonator (SMR)-based NEMS ME antennas on a Bragg acoustic resonator, which have a circular resonating disk of 200 𝝁𝒎 17 diameters and operate at 1.75 GHz, show a high antenna gain of -18.8 dBi and 1dB compression point (P1dB) of 30.4 dBm. Compared to 18 same-size thin-film bulk acoustic resonator (FBAR) ME antennas with a free-standing


Introduction 1
In the last few decades, explosive growth in miniaturised, low-profile and cost-effective antennas has emerged due to the proliferation 2 of innovative applications such as Internet-of-Things (IoT) devices, 5th generation (5G) wireless systems, millimetre-wave (mm-3 wave) and Terahertz (THz) applications, etc. 1,2,3,4 There is an insatiable need for small-size and highly efficient antennas. Numerous 4 $ ( − ) + ! , where # is the anisotropy field, is the angle of the applied DC magnetic field, is the tilt of the UMA axis 23 relative to the 90 degrees axis, and H0 is a constant field defined by the chosen fixed resonance frequency. The obtained magnitude 24 of # and the tilt of the effective magnetic anisotropy axis is 1920 A/m and 12.8 degrees for the reference film and 720 A/m and 11.3 25 degrees for the ME antenna. 26 27 5 The source of UMA can be in principle attributed to magnetocrystalline anisotropy, substrate surface topography, magnetic field 1 induced anisotropy and stress induced anisotropy 38 . Due to the amorphous structure of the thick seed layer and smooth substrate 2 surface, we assume only contributions from the induced anisotropies to the UMA. Wang et al. 38 demonstrated the stress-induced 3 UMA in FeCoC films. However, the deposited film stress was not analysed. In the current work, a detailed investigation on the 4 anisotropic stress of FeGaB films is carried out and shown in Supplementary Fig. 1, which suggests an additional stress-induced 5 UMA from the initial layers during growth. 6 7 The general magnetic anisotropy alignment is confirmed by spatially resolved magnetic analysis of an identical device (Fig 2 g-h and  8 Supplementary Fig. 6). The magnetic domains walls tend to align roughly along the ~90 degrees axis. Despite the exhibited roughness 9 of the samples, the magnetic anisotropy alignment is still determining the magnetic microstructure. Deviations of the domain wall 10 alignment are related to the laminated structure of the multilayer. This and the observed wide magnetic domain walls, an indication 11 of low magnetic film coupling, confirm the quality of the film structure. Noticeable is the curving of the magnetisation from the edges 12 are caused by stress relaxation 39 , leading to a stress-induced anisotropy contribution aligned parallel to edges. This indicates tensile 13 stress for the positive magnetostrictive films. This further transfers into a change of the magnetic anisotropy strength in the more 14 central region of the disc shaped device, leading to a spatially varying anisotropy strength alteration. Another factor that plays a role 15 is the change of the magnetic properties in the anchor region, visible from the analysis of the magnetisation process (Supplementary 16 Fig. 6). Overall, the tilt of magnetic anisotropy and the change of integral anisotropy strength of the patterned device structure relative 17 to the full film can be related to spatially varying stress induced anisotropy contributions and the effect of roughness on the coercivity 18 of the sample is not significant in this case. 19 20 Design and simulations of the SMR-based ME antenna 21 Prior to fabricating SMR devices, various simulation methods were implemented to design and optimise the performance of SMR-22 based ME antennas with a focus on the Bragg reflector and ME composites. The first analysis was carried out by using the 1D 23 Mason's model to estimate the reflection coefficient frequency response of a carefully arranged Bragg reflector 40,41,42 . The equivalent 24 circuit model of the SMR-based ME antenna analysed by the Mason's model and based on the transmission line theory is shown in 25 Fig. 3a. The load acoustic impedance decreases while the total number of Bragg reflector layers (n) increases. Therefore, a large value 26 of n is desired for the resonator structure. Typically, three pairs of low/high acoustic impedance layers (n = 6) are chosen for SMRs. 27 28 6 In order to correlate the working frequency bandwidth of the Bragg reflector with the resonant frequency of the piezoelectric resonator, 1 the thickness of each layer is decided by the equation: %&'()*+& = %&'()*+& /4 , where %&'()*+& and %&'()*+& are the thickness and 2 acoustic velocity of the Bragg reflector layers, respectively. In this work, the working frequency of SMR-based ME antenna was not 3 specified. Therefore, the thickness of each layer was not exactly a quarter wavelength of the acoustic wave. The simulated reflection 4 coefficient of the designed Bragg reflector consisting of three pairs of SiO2/W layers was calculated. The results of which are shown 5 in Fig. 3b. The total reflection seen from the resonant structure is formed over a wide frequency range from 1 to 3 GHz. The influence 6 on the reflection coefficient with varying number of periods is also exploited and displayed in Fig. 3b. As more periods are applied, 7 more energy is reflected from Bragg acoustic lattice, and so the frequency band widens. The calculation results for more periods of 8 SiO2/W layers are presented in Supplementary Fig. 3, which illustrates minimal improvement in the reflection coefficient when the 9 number of periods exceeds three. 10

11
To achieve a frequency domain analysis of the admittance spectrum, a 2D model of the ME antenna was developed, and the coupling 12 between electrical potential and mechanical displacement in the antenna was simulated by the finite element method (FEM) in 13 COMSOL Multiphysics v5.1 42 . The magnitude of the total displacement profile of the longitudinal waves and the standing wave 14 amplitude as a function of depth of the designed ME antenna at the electromechanical resonant frequency of 1.7 GHz are presented 15 in Fig. 3c and d. The 2D strain assumption is used in this model and the thickness of each layer is defined with the value as shown in 16 Supplementary Table 1. The developed Bragg reflector is shown with optimised performance as the acoustic wave energy is well 17 confined within the ME composites with little energy dissipating into the substrate. The return loss curve of the ME antenna is plotted 18 in Fig. 3e, which is achieved by the frequency domain analysis with the 2D FEM COMSOL model. The thicknesses of each individual 19 layer in the Bragg reflector and ME composites determines the working frequency of SMR antenna. As explained previously, the 20 performance of SMRs can be improved by adjusting the thickness of the Bragg reflector layer to be a quarter wavelength of the 21 acoustic wave. 22

23
Mechanical ME antenna performance 24 The antenna gain was characterised in an anechoic chamber by utilising a calibrated linear polarisation standard horn antenna. As 25 shown in Fig. 4a, the return loss curve (S11), receiving (S12) and transmitting behaviour (S21) of the SMR antenna with a resonant 26 frequency of 1.75 GHz and antenna gain of -18.8 dBi are presented. The S12 and S21 curves overlap with each other. The 7 electromechanical resonance frequency, which is defined by the thickness of the ME disk and is validated by the 2D COMSOL model, 1 is expressed as: 2 where T is the thickness of the ME disk, E and ρ are the equivalent Young's modulus and equivalent density of the resonator, 4 respectively. In Fig. 4b, the schematics and the fitting parameters for the modified Butterworth Van Dyke (MBVD) model with 5 electrical and equivalent mechanical components of SMR antenna are presented. The electromechanical coupling coefficient ( * $ ) 6 and quality factor (Q) are calculated as 1.0% and 95, respectively. The MBVD fitting curve for return loss S11 is plotted in Fig. 4c  7 and matches the measurements well. Compared to the released FBAR ME antenna, the SMR antenna has a 10 dBi higher gain. The 8 gain enhancement of the SMR antenna is attributed to the Bragg reflector helping to confine more acoustic energy in the ME films 9 allowing for greater amplitude in EM waves. 10 11 A high linearity is always wanted for the components in the RF systems such as filters, amplifiers and antennas. This is desired 12 because there are numerous different bands, and they have to be protected from any undesired signals. Furthermore, nonlinearity of 13 the devices can undesirably degrade the performance of the system. It has been reported that the acoustic resonators exhibit a nonlinear 14 behaviour at high power levels 43,44,45 . Since the power density and temperature play significant roles in controlling the nonlinearity, 15 various methods such as the device structure, area and materials etc. are used to improve the linearity of acoustic resonators 46 . 16

17
The power handling capability of the FBAR and SMR antennas are characterised by the power sweep results shown in Fig. 4f. The 18 P1dB is acquired by measuring the S21 curves as a function of input power. The FBAR antenna has a P1dB of 7.1 dBm while the 19 P1dB of SMR antenna is 30.4 dBm, which shows a better power handling capability of the SMR antenna. The S11 curves of the FBAR 20 and SMR antennas at different input power levels are measured and shown in Fig. 4d and e. The resonance peak of the FBAR antenna 21 shifts to lower frequency at high input power while the SMR antenna remains unchanged, which means the FBAR antenna starts 22 going into nonlinear region. This can be explained by the self-heating effects at high power levels. As the input power extremely 23 increases, the high power density leads to self-heating of the resonator and higher temperature. Compared to the free-standing FBAR 24 membrane structure, the SMR antenna with acoustic Bragg reflector stacks has a larger thermal conductivity. This compensation 25 effect results in a better power handling capability. The comparison of performance metrics between the FBAR and SMR antennas 26 are listed in Table 1. 27   8   1 The radiation characteristics of the SMR-based ME antennas were tested in a far-field configuration with a distance of 0.76 m between 2 the SMR antenna and the horn antenna. The active radiative element of the SMR antenna is defined by the ZnO/FeGaB ME 3 heterostructure disk with a diameter of 200 µm. Owing to the small size of the SMR antenna and the limitations of the probe station, 4 we were only able to measure the in-plane radiation pattern 180 degrees around the SMR device. The schematics and measured 5 radiation pattern are shown in Fig. 5a. A dipole-like radiation pattern is assumed according to the symmetry of the SMR antenna. The 6 radiation pattern shows an approximate 7-degree shift from the centre of the horn antenna due to the in-plane tilt of the UMA, as 7 shown in Fig. 2. The maximum gain is located along the tilted direction of the anchor length (the 7-degree point as shown in Fig. 5e), 8 which is the hard axis (H.A.) of the magnetostrictive FeGaB film. At 7 degrees, the RF magnetic field component of EM wave is 9 parallel to the magnetisation of the magnetostrictive layer, which results in maximum coupling efficiency between the SMR antenna 10 and EM wave to achieve maximum gain. When in-plane radiation is measured at 97 degrees from the centre, the antenna gain 11 approaches its null value because the RF magnetic field is perpendicular to the magnetisation. 12

13
The polarisation behaviours of SMR ME antennas were measured by rotating the standard horn antenna along three major axes, as 14 shown in Fig. 5b-d. In the schematic representations of Fig. 5, the horn antenna was rotated along the out-of-plane axis (Fig. 5b), in-15 plane axis perpendicular to the anchor direction (Fig. 5c) and in-plane axis along the anchor direction (Fig. 5d). The sinusoidal wave 16 along 0° and 180° in all the schematics indicates the RF H-field component of the EM waves from the horn antenna. The normalised 17 gain plot in Fig. 5f-h show a similar shape of bifold symmetry, which results from the in-plane uniaxial magnetic anisotropy of the 18 FeGaB/ZnO multilayer in the resonating disk of the SMR antenna. 19 20 As presented in Fig. 5f, the highest gain of the SMR antenna is achieved along the easy axis (E.A.) when the Hrf is parallel to the E.A. 21 direction. The lowest gain is measured tilted away from the 0-degree position when the Hrf is parallel to the H.A. direction. The other 22 two polarisation charts in Fig. 5g and h show a similar behaviour where the maximum gain is obtained along the E.A. direction of 23 magnetic anisotropy. This is because the strongest coupling between Hrf and SMR ME antenna is achieved when Hrf is parallel to the 24 magnetic easy axis in the ME disk. The measured results in Fig. 5  The latest mechanically driven antennas have pushed the boundaries of antenna miniaturisation to micrometre dimensions starting 3 with the free-standing membrane FBAR design. The presented SMR structures for ME antennas not only improve upon the antenna 4 performance metrics but also simplifies the process with better device structural integrity and removing the extra packaging steps 5 required for device protection. We have demonstrated a working micro-sized ME antenna that takes advantage of a solidly mounted 6 resonator to confine energy in the magnetostrictive/piezoelectric heterostructure showing improvements in antenna radiation. The 7 SMR ME antenna, with an overall dimension of 700 × 700 (L × W), was designed and optimised with the 1D Mason model 8 and a 2D COMSOL FEM simulation to operate at a resonant frequency of 1.75 GHz and experimentally demonstrated a gain of -18.8 9 dBi. Further improvements can be made by increasing the SMR quality factor and optimising the impedance matching with the signal 10 feedline to the resonator. 11

12
The SMR ME antennas have demonstrated more robust features than its FBAR freestanding membrane counterparts. Due to its small 13 size, high operating frequency, high sensitivity, structural stability, semiconductor processing integration, and good power handling 14 capability, it is the ideal device for small-size microwave antennas and remote wireless sensing applications used for compact UAVs 15 (unmanned aerial vehicles), bio-implantable antennas, wearable antennas, IoT (internet of things), NFC (near field communication), 16 RFID (radio frequency identification), satellites, and many more existing and to be envisioned applications. 17 18

Methods 19
Methods and any associated references are available in the online version of the paper. Methods 12 Materials decision. The acoustic impedance of various materials is the key point for creating effective Bragg reflectors used for 13 ME antennas. Silicon (Si), silicon dioxide (SiO2) and aluminium (Al) are common low impedance materials for Bragg reflectors 14 while platinum (Pt), tungsten (W) and iridium (Ir) are introduced as high impedance materials. In this work, we determined that 15 employing sputtered SiO2/W as the low/high acoustic impedance materials to construct the Bragg reflector will provide the best 16 acoustic impedance ratio along with a cost-effective fabrication process. In order to achieve strong ME coupling in the 17 magnetostrictive/piezoelectric ME composites, ferromagnetic/ferroelectric materials with large piezomagnetic/piezoelectric 18 coefficients are desired. Excellent magnetic softness and magnetostrictive behaviours have been realised in FeGa-based thin films 47, 19 48 , which have already been applied in different RF/microwave tuneable devices 22, 25 . Aluminium nitride (AlN) and zinc oxide (ZnO) 20 are two of the most popular piezoelectric materials for FBAR devices, among which, AlN has been chosen as the material for 21 commercial BAW filters that operates at 1-2 GHz because of its quality factor (Q). However, ZnO has larger * $ and wider 22 bandwidth than AlN due to its higher piezoelectric coefficient (d33). Moreover, the deposition, control of texture, and stoichiometry 23 of ZnO is much easier compared with that for AlN 49, 50 . ZnO films with highly c-axis-preferred orientation are essential for realising 24 high-quality SMR-type ME antennas. To fabricate SMR-based ME antennas, SiO2/W as the low/high impedance Bragg reflector 25 and FeGaB/ZnO as the magnetostrictive/piezoelectric ME composite were carefully designed and deposited by RF magnetron 26 sputtering. Platinum (Pt) and gold (Au) were chosen as the bottom and top electrodes, respectively, due to the benefits for growing 27 highly c-axis-textured ZnO thin films and decent conductivity. 28 29 Thin films deposition. All thin-film materials were prepared in a magnetron sputtering system at the Argon (Ar) flux density of 15 30 sccm with a base pressure of ~ 1 × 10 12 Torr. The oxides including SiO2 and ZnO were deposited by reactive RF sputtering and 31 other metallic thin films involving W, FeGaB, Pt, and Au were DC sputtered. All materials were deposited at room temperature 32 except ZnO layer, which was deposited at 450 ℃ to achieve a highly c-axis-orientated structure. For the Bragg reflector with 33 structure of [W (531 nm)/SiO2 (543 nm)]3, W layers were sputtered at a plasma power of 50 W and an Ar atmosphere of 4.5 mTorr; 34