Small-Size Magneto-Mechanical Resonators

• Small-Size MMRs are miniaturized mechanical resonators that use magnetic interactions to generate restoring torque, enabling precise frequency tuning and sensing.
• They incorporate varied architectures—torsional, membrane, cantilever, and optomechanical—to measure pressure, force, temperature, and magnetic fields with high sensitivity.
• Scaling laws, design trade-offs, and specialized excitation/readout methods are central to advancing these devices for wireless sensing, signal transmission, and reservoir computing.

Small-size magneto-mechanical resonators (MMRs) are miniaturized mechanical resonators in which magnetic interactions determine restoring torque, actuation, readout, or frequency tuning. In one major class, a rotor magnet suspended by a compliant filament oscillates torsionally against a stator magnet, and the natural frequency depends on the rotor–stator distance; in other implementations, magnetic force, field gradient, magnetostriction, or spin-dependent force drives membranes, cantilevers, trampolines, or optical microresonators (Faltinath et al., 6 Mar 2025, Merbach et al., 13 Feb 2025, Forstner et al., 2011, Scozzaro et al., 2016). Because the resonance frequency or amplitude can encode pressure, temperature, magnetic field, magnetic gradient, or magnetic resonance force, small-size MMRs have been developed as passive wireless sensors, force detectors, transmitters, and multiphysics computing elements (Fischer et al., 2024, Thanalakshme et al., 2020, Grimaldi et al., 7 Jan 2026).

1. Fundamental magneto-mechanical principles

In passive torsional MMRs, the rotor magnetic dipole moment $\mathbf{m}_r$ experiences the stator field $\mathbf{B}_s$, giving a torque $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$. For small torsional deflections about equilibrium, the restoring torque is approximated by $\tau \approx -\kappa(d)\phi$, where $\kappa(d)=m_r B_0(d)$ and $d$ is the center-to-center rotor–stator distance. The corresponding natural frequency is

$$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$

and, under the ideal dipole assumption $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$, it follows the characteristic scaling $f_n(d)\propto d^{-3/2}$ (Faltinath et al., 6 Mar 2025).

The rotor dynamics are commonly written as

$$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$

with $\mathbf{B}_s$0. In the small-angle limit, $\mathbf{B}_s$1, and for high $\mathbf{B}_s$2 the instantaneous oscillation approaches the natural frequency while becoming deflection-amplitude independent. This same form underlies recent model-based estimators for real-time sensing and pose recovery from coil signals (Reiss et al., 23 Feb 2026).

Magnetic fields and gradients can enter the effective stiffness directly. In the cantilever-based magneto-oscillatory wireless sensor, the total stiffness is written as $\mathbf{B}_s$3, with $\mathbf{B}_s$4 and $\mathbf{B}_s$5, so that for $\mathbf{B}_s$6,

$\mathbf{B}_s$7

This formulation separates first-harmonic field coupling from second-harmonic gradient coupling and makes explicit how rotation can decouple the two contributions (Fischer et al., 2024).

Force-based MMRs use the same harmonic-oscillator structure but with linear displacement instead of torsion. In the silicon nitride membrane magnetic-resonance force detector, the displacement obeys $\mathbf{B}_s$8 with

$\mathbf{B}_s$9

and on resonance the amplitude is $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$0. This formulation connects spin-dependent force, mechanical susceptibility, and optical displacement readout in a form directly analogous to other small-size MMRs (Scozzaro et al., 2016).

2. Principal architectures and material platforms

The term “small-size MMR” covers multiple architectures that share magnetic coupling but differ strongly in mechanical element, transduction path, and operating frequency. The table summarizes representative implementations reported in the literature.

Architecture	Representative realization	Reported metrics
Torsional rotor–stator passive sensor	4 mm diameter spherical and/or cylindrical permanent magnets, $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$1 mm	$\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$2 Hz depending on geometry; calibrated median deviation generally below $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$3 (Faltinath et al., 6 Mar 2025)
Wireless passive pressure MMR	Cylindrical acrylic housing $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$4 mm, $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$5 mm; 4 mm spherical rotator, 4 mm cylindrical stator, 3D-printed membrane	$\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$6 Hz mbar$\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$7 below 100 mbar; $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$8 Hz readout in dynamic tests (Merbach et al., 13 Feb 2025)
Gd-shielded temperature MMR	Spherical NdFeB stator $\boldsymbol{\tau}=\mathbf{m}_r\times\mathbf{B}_s$9 mm, rotor $\tau \approx -\kappa(d)\phi$0 mm, Gd shell $\tau \approx -\kappa(d)\phi$1m	$\tau \approx -\kappa(d)\phi$2 kHz baseline; peak $\tau \approx -\kappa(d)\phi$3 Hz/K at $\tau \approx -\kappa(d)\phi$4 K for $\tau \approx -\kappa(d)\phi$5m (Faltinath et al., 29 Aug 2025)
SiN$\tau \approx -\kappa(d)\phi$6 membrane force detector	Square low-stress SiN$\tau \approx -\kappa(d)\phi$7 membrane, $\tau \approx -\kappa(d)\phi$8m, $\tau \approx -\kappa(d)\phi$9 nm	$\kappa(d)=m_r B_0(d)$0 at 300 K, $\kappa(d)=m_r B_0(d)$1 at 4 K; $\kappa(d)=m_r B_0(d)$2 fN/$\kappa(d)=m_r B_0(d)$3 at 300 K (Scozzaro et al., 2016)
Cavity optomechanical magnetometer	Toroidal whispering gallery mode resonator, approximately $\kappa(d)=m_r B_0(d)$4m in size, with Terfenol-D actuator	Peak $\kappa(d)=m_r B_0(d)$5 nT/$\kappa(d)=m_r B_0(d)$6 achieved; modeling predicts up to $\kappa(d)=m_r B_0(d)$7 fT/$\kappa(d)=m_r B_0(d)$8 (Forstner et al., 2011)
Cantilever-based magnetic field/gradient sensor	Magneto-oscillatory wireless sensor with footprint $\kappa(d)=m_r B_0(d)$9 mm$d$0	sub-$d$1T field resolution and $d$2T/m gradient resolution (Fischer et al., 2024)
Magnetic trampoline resonator	Single-crystal LSMO thin film, 100 nm thick; $d$3m$d$4 central pad	quality factor up to $d$5k and $d$6 products reaching $d$7 Hz (Manca et al., 21 Jan 2025)

These implementations span torsional, flexural, membrane, trampoline, and optomechanical regimes. The shared theme is not a single geometry but the use of magnetic interaction as stiffness source, actuation path, or measurand transducer.

Material choice is central. Low-stress SiN$d$8 provides high $d$9, low mass, and large accessible surfaces; BeCu crossed-flexure pivots are favored where low structural damping and large transverse stiffness are required; LSMO offers a structural element that is itself magnetic; Terfenol-D provides magnetostrictive actuation; and NdFeB remains the dominant permanent-magnet material in torsional passive MMRs and magneto-mechanical transmitters (Scozzaro et al., 2016, Li et al., 2024, Manca et al., 21 Jan 2025, Forstner et al., 2011, Faltinath et al., 6 Mar 2025).

3. Excitation, readout, and model-based estimation

A defining feature of many small-size MMRs is remote excitation combined with passive readout. In torsional rotor–stator sensors, a weak external magnetic field temporarily deflects the rotor, and after the drive is removed the rotor performs damped oscillations whose time-varying magnetic moment is detected inductively. In the three-coil configuration used for passive wireless sensing, the torsional oscillation appears dominantly in one lateral receive channel, while twice the torsional frequency appears in the orthogonal lateral channel; a vertical coil provides a null channel when the sensor is correctly centered (Faltinath et al., 6 Mar 2025).

The pressure-sensing implementation uses a separable square-shaped Helmholtz-like coil clamped around the column. During a transmit window, a weak oscillatory magnetic field drives the torsional deflection; during a receive window, the same coils detect the induced signal by reciprocity. The reported signal chain comprises a class-D transmitter amplifier, a low-noise receiver amplifier, a RedPitaya Stemlab 125-14 DAC/ADC, real-time frequency/phase-controlled re-excitation, and analog filtering to suppress 50 Hz mains harmonics (Merbach et al., 13 Feb 2025).

Recent model-based sensing work has formalized this workflow as a nonlinear inverse problem. The induced voltage vector is written as

$$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$0

where $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$1 is the receive-coil sensitivity matrix and $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$2 is the orientation. Time-domain and time-frequency estimators fit $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$3, $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$4, $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$5, $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$6, and projection vectors $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$7. The simplified methods reduce estimation time by up to two orders of magnitude at the expense of less than $$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$8 deviation for large maximum deflection angles (Reiss et al., 23 Feb 2026).

Optical readout is equally prominent in other MMR classes. The SiN$$f_n(d)=\frac{1}{2\pi}\sqrt{\frac{\kappa(d)}{I}}=\frac{1}{2\pi}\sqrt{\frac{m_r B_0(d)}{I}},$$9 membrane magnetic-resonance detector uses fiber-optic interferometry with 1550 nm light, a $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$0m spot, $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$1W incident power, and an interferometer noise floor $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$2 pm/$B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$3 (Scozzaro et al., 2016). The cavity optomechanical magnetometer instead reads the motion of a toroidal whispering gallery mode resonator with 980 nm laser light evanescently coupled by a tapered fiber and detected through direct transmission spectroscopy (Forstner et al., 2011).

Other readout modalities emphasize electrical compactness. The cantilever-based magneto-oscillatory sensor is driven by excitation coils and read out by two wired magnetometers in differential mode, using ring-down fitting to extract frequency with sub-mHz precision (Fischer et al., 2024). Magnetomotive string resonators in water use Lorentz-force drive and motional-EMF detection in a $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$4 T Halbach field, eliminating optical alignment and enabling simultaneous drive and detection of multiple immersed resonators (Venstra et al., 2010). At still smaller scale, the magneto-mechanical reservoir uses a $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$5 lattice of mass-spring resonators, each node carrying an MTJ spin diode; the rectified voltage $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$6 is computed from injection-locked spin dynamics and used as the reservoir state (Grimaldi et al., 7 Jan 2026). Coupled-oscillator metrology adopts a phase-locked drive leading by $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$7, with a frequency counter referenced to a rubidium-standard 10 MHz timebase, so that the oscillation frequency tracks the undamped resonance (Bouche et al., 2024).

4. Sensing modalities and demonstrated performance

Force detection is one of the highest-sensitivity MMR use cases. The SiN$B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$8 membrane detector achieved $B_0(d)\simeq \frac{\mu_0}{4\pi}\frac{m_s}{d^3}$9 fN/$f_n(d)\propto d^{-3/2}$0 at 300 K for the loaded membrane and projected a bare-membrane sensitivity $f_n(d)\propto d^{-3/2}$1 aN/$f_n(d)\propto d^{-3/2}$2 at 4 K. With 100 mW RF, $f_n(d)\propto d^{-3/2}$3 G, and $f_n(d)\propto d^{-3/2}$4 G/$f_n(d)\propto d^{-3/2}$5m, the membrane reached $f_n(d)\propto d^{-3/2}$6 Å, corresponding to a force $f_n(d)\propto d^{-3/2}$7 fN and a polarized moment $f_n(d)\propto d^{-3/2}$8 J/T, or $f_n(d)\propto d^{-3/2}$9 electron spins (Scozzaro et al., 2016).

Pressure sensing in passive wireless MMRs proceeds by converting membrane deflection into a change in magnet spacing. In the reported process-engineering device, sensitivity in the lower range $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$0 mbar was $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$1 Hz mbar$$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$2 for one device and $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$3 Hz mbar$$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$4 for another. Across the full range, average frequency standard deviations were $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$5 Hz and $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$6 Hz; the maximum pressure range reached $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$7 mbar before the travel limit was reached. Dynamic measurements at a 2 Hz frame rate yielded an overall standard deviation of $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$8 mbar without averaging. The same study also found a pronounced thermal cross-sensitivity, with $$\ddot{\varphi} + \frac{\omega_{\text{nat}}}{Q}\,\dot{\varphi} + \omega_{\text{nat}}^2 \sin\varphi = 0,$$9 Hz $\mathbf{B}_s$00C$\mathbf{B}_s$01 and $\mathbf{B}_s$02 (Merbach et al., 13 Feb 2025).

Temperature sensing can also be engineered magnetically rather than mechanically. In the Gd-coated stator design, the ferromagnetic-to-paramagnetic transition of Gd near $\mathbf{B}_s$03 K modulates the stator field at the rotor through temperature-dependent shielding. For $\mathbf{B}_s$04m, the reported peak sensitivity is $\mathbf{B}_s$05 Hz/K at $\mathbf{B}_s$06 K, with a mean sensitivity of $\mathbf{B}_s$07 Hz/K across $\mathbf{B}_s$08 K and a frequency excursion $\mathbf{B}_s$09 Hz (Faltinath et al., 29 Aug 2025).

Magnetic field and gradient sensing have been demonstrated directly in miniature oscillatory devices. The millimeter-scale magneto-oscillatory wireless sensor is capable of detecting magnetic fields with sub-$\mathbf{B}_s$10T resolution to at least $\mathbf{B}_s$11 mT and simultaneously detects magnetic field gradients with a resolution of $\mathbf{B}_s$12T/m to at least $\mathbf{B}_s$13 mT/m. Its average frequency precision was $\mathbf{B}_s$14 mHz, corresponding to $\mathbf{B}_s$15 nT in the reported laboratory environment (Fischer et al., 2024). The cavity optomechanical magnetometer achieved a peak magnetic-field sensitivity of $\mathbf{B}_s$16 nT/$\mathbf{B}_s$17 at resonance at room temperature, while theoretical modeling of an optimized geometry predicted $\mathbf{B}_s$18 fT/$\mathbf{B}_s$19 (Forstner et al., 2011).

Distance-dependent frequency itself can be the sensed quantity. For 4 mm spherical and cylindrical permanent-magnet MMRs, the adapted dipole model

$\mathbf{B}_s$20

described the measured frequency–distance relationship with median deviations generally below $\mathbf{B}_s$21 after calibration, whereas the baseline model overestimated frequency systematically with median deviations $\mathbf{B}_s$22 depending on geometry (Faltinath et al., 6 Mar 2025).

Real-time estimation accuracy is now a performance metric in its own right. In parameter-estimation experiments on small and large passive MMRs, time-frequency methods achieved near-instant fits relative to multi-second time-domain ODE-based fits, and the simplified methods reduced estimation time by up to two orders of magnitude while keeping deviations below $\mathbf{B}_s$23 for large maximum deflection angles (Reiss et al., 23 Feb 2026).

At the most optimistic end of the reported metrology landscape, the coupled-oscillator framework yielded a simulated noiseless force resolution of $\mathbf{B}_s$24 zN measured over 1 s and a magnetic-gradient resolution of $\mathbf{B}_s$25 aT/cm at a single point within $\mathbf{B}_s$26m (Bouche et al., 2024).

5. Arrays, communication, and unconventional functions

Not all small-size MMR research is directed toward scalar sensing. A parallel line of work treats MMRs as sources of ultra-low-frequency magnetic radiation. In resonant magneto-mechanical transmitters, one or more permanent-magnet rotors are driven near torsional resonance so that their rotating dipole moments generate time-varying magnetic fields below 3 kHz. Demonstrated single-rotor devices operated near $\mathbf{B}_s$27 Hz, multi-rotor devices reached $\mathbf{B}_s$28 Hz, and cylindrical micromagnet implementations exceeded 1 kHz, with a reported $\mathbf{B}_s$29 Hz example. The single-rotor prototype achieved $\mathbf{B}_s$30 fT at resonance with $\mathbf{B}_s$31 W, and amplitude modulation via on-off keying was demonstrated at 5 bps and 10 bps (Thanalakshme et al., 2020).

Efficient array operation depends strongly on the bearing system. Crossed-flexure pivot bearings have been proposed as compliant supports that allow large angular rotation, high transverse stiffness, and compact interlocking assembly of closely spaced rotor arrays. In ring-down tests, BeCu pivot-bearing MMRs showed a damping coefficient up to 80 times lower than corresponding ball-bearing MMRs, and $\mathbf{B}_s$32 mm BeCu pivot bearings supported rotational resonances of $\mathbf{B}_s$33 Hz while keeping the lateral resonance near $\mathbf{B}_s$34 Hz (Li et al., 2024).

Coupled-oscillator amplification extends the sensing role of MMRs into a frequency-metrology regime. In the reported MEMS implementation, one small $\mathbf{B}_s$35 mm N52 magnet on a mechanical oscillator is coupled to a large $\mathbf{B}_s$36 mm N52 magnet, and the distance-dependent magnetic force shifts the oscillator’s effective stiffness and undamped resonant frequency. This architecture is intended for zeptonewton and attotesla-per-centimeter metrology rather than conventional passive wireless tracking (Bouche et al., 2024).

A still more unconventional direction is magneto-mechanical reservoir computing. The reported proof-of-concept device is a $\mathbf{B}_s$37 lattice of nonlinear mass-spring resonators with one MTJ spin diode on each mass. Mechanical excitation is injected directly as elastic acceleration, while magnetically coupled MTJs supply the electrical readout through $\mathbf{B}_s$38. On a 243-sample vowel-recognition task, the best trials reached validation accuracy above $\mathbf{B}_s$39, and the study found that modest inhomogeneity in elastic constants could improve performance rather than degrade it (Grimaldi et al., 7 Jan 2026).

These developments broaden the functional definition of an MMR. Small-size MMRs are not limited to being frequency-shifting passive sensors; they also serve as resonant transmitters, signal amplifiers, and nonlinear dynamical substrates for edge computation.

6. Scaling laws, design trade-offs, and limitations

Scaling laws recur across otherwise different implementations. In rotor–stator passive MMRs, uniform miniaturization by a factor $\mathbf{B}_s$40 gives $\mathbf{B}_s$41, $\mathbf{B}_s$42, and, when the gap scales as $\mathbf{B}_s$43, $\mathbf{B}_s$44. The same paper states that miniaturization with proportionally smaller gaps raises $\mathbf{B}_s$45 approximately as $\mathbf{B}_s$46, which is favorable for bandwidth and responsiveness (Faltinath et al., 6 Mar 2025). Crossed-flexure pivot MMRs show an analogous trend, with $\mathbf{B}_s$47, $\mathbf{B}_s$48, and hence $\mathbf{B}_s$49 (Li et al., 2024).

That frequency advantage is not equivalent to monotonic improvement in overall sensing or transmission performance. In magneto-mechanical transmitters, shrinking the rotor reduces $\mathbf{B}_s$50 and raises $\mathbf{B}_s$51, but it also reduces magnetic moment, so the near field at range drops unless oscillation amplitude or rotor count is increased. The reported design rule is therefore to distribute total magnet volume across multiple low-inertia rotors rather than merely shrinking a single rotor (Thanalakshme et al., 2020). The pressure-sensing literature expresses the same kind of compromise differently: thinner membranes increase compliance and $\mathbf{B}_s$52, but reduce maximum pressure range and mechanical robustness; smaller initial magnet spacing $\mathbf{B}_s$53 raises sensitivity but shortens the travel before contact (Merbach et al., 13 Feb 2025).

For membrane force detectors, the key sensitivity expression

$\mathbf{B}_s$54

makes the design targets explicit: lower $\mathbf{B}_s$55, lower $\mathbf{B}_s$56, higher $\mathbf{B}_s$57, and careful choice of $\mathbf{B}_s$58. In tensioned SiN$\mathbf{B}_s$59 membranes, the literature therefore recommends making $\mathbf{B}_s$60 as small as feasible, using low-stress SiN$\mathbf{B}_s$61, maintaining large $\mathbf{B}_s$62 within compact packaging, and operating at cryogenic temperatures where possible (Scozzaro et al., 2016).

Mechanical support and damping control can dominate performance. Pivot-bearing studies emphasize avoiding the hybrid regime $\mathbf{B}_s$63, because coupling between rotational and lateral modes increases dissipation (Li et al., 2024). Pressure MMRs identify air damping, support losses, and alignment scatter as recurring issues (Merbach et al., 13 Feb 2025). Magnetomotive resonators in water show that electrical loading by the complex dielectric response of water can depress the apparent $\mathbf{B}_s$64 below the optically measured value, while resistive heating of the liquid shifts resonance frequency by approximately $\mathbf{B}_s$65 Hz/$\mathbf{B}_s$66C in the reported 200 $\mathbf{B}_s$67m strings (Venstra et al., 2010).

Magnetic modeling also imposes limits. The distance law $\mathbf{B}_s$68 remains valid across the tested mm-scale permanent-magnet geometries, but near-field deviations, finite geometry, glue mass, and misalignment make a single fitted prefactor $\mathbf{B}_s$69 necessary for accurate calibration (Faltinath et al., 6 Mar 2025). Likewise, simplified parameter-estimation models retain excellent speed but become less accurate when higher harmonics or large-angle motion matter; TUSA$\mathbf{B}_s$70 remains within $\mathbf{B}_s$71 over $\mathbf{B}_s$72, whereas TUSAE$\mathbf{B}_s$73 shows larger errors in $\mathbf{B}_s$74 (Reiss et al., 23 Feb 2026).

Materials integration creates a distinct set of trade-offs. LSMO trampoline resonators avoid the interface problems of hybrid structural-plus-magnetic stacks by using a single magnetic oxide as both resonator and functional layer, but their reported $\mathbf{B}_s$75 values remain far below the room-temperature $\mathbf{B}_s$76 cited for state-of-the-art Si-based resonators (Manca et al., 21 Jan 2025). This suggests that fabrication simplicity, magnetic interaction volume, and ultrahigh $\mathbf{B}_s$77 are still being balanced rather than simultaneously maximized.

Across the literature, small-size MMRs are best understood as a family of devices governed by a common theme: magnetic interaction reshapes the stiffness landscape of a miniature mechanical resonator. The specific implementation then determines whether the primary figure of merit is force sensitivity, frequency calibration error, field or gradient resolution, wireless range, transmit efficiency, or computational richness.