HF-FMR: High-Field, High-Frequency Resonance

• HF-FMR is the extension of ferromagnetic resonance into high-field and high-frequency regimes, resolving magnetic anisotropy, demagnetizing fields, and magnon structures.
• It employs innovative methodologies such as transmission-mode HF-ESR, EUV XFMR, SAW-driven excitation, and voltage-controlled FMR to extract key magnetic parameters.
• HF-FMR enhances element-specific imaging and device integration, offering precise analysis of anisotropy, damping, and domain dynamics in a variety of magnetic materials.

Searching arXiv for the specified paper and closely related HF-FMR work. I’ll proceed using the provided arXiv corpus as the source set for the article and cite the relevant papers directly. High-field/high-frequency ferromagnetic resonance (HF-FMR) is the extension of ferromagnetic resonance into regimes where the static field, the excitation frequency, or both are large enough to resolve magnetic anisotropy, demagnetizing fields, domain-dependent resonances, and low-energy magnon structure with high spectral discrimination. In the cited literature, HF-FMR directly probes the $q=0$ magnon excitations and their anisotropy gaps, and is implemented in transmission-mode HF-ESR/HF-FMR up to 16 T and 850 GHz, in device-centered electrical platforms above 30 GHz, and in element-specific or nanoscale x-ray methods that combine resonance spectroscopy with chemical contrast or imaging (Zeisner et al., 2020, Beier et al., 21 Jul 2025, Tanksalvala et al., 2024, Bonetti et al., 2015, Skowronski et al., 2019).

1. Definition and scope

HF-FMR is used to extract gyromagnetic ratios, effective $g$-factors, saturation magnetization, anisotropy fields, anisotropy gaps, linewidths, and domain-state information from the field and frequency dependence of resonance branches. In the studies summarized here, the accessible experimental space spans from surface-acoustic-wave harmonics at 172 MHz to HF-ESR/HF-FMR at 850 GHz, and from sub-10 mT bias fields in YIG-based magnetometry to 16 T superconducting-magnet measurements in van der Waals magnets (Weiler et al., 2010, Colombano et al., 2019, Beier et al., 21 Jul 2025).

The term also covers several physically distinct situations. In CrCl$_3$, HF-FMR is used in the field-polarized low-temperature phase, where all measured resonance fields at 4 K exceed the saturation fields and therefore probe a single-domain, ferromagnetic-like state. In Fe$_3$GeTe$_2$, the same methodology resolves a finite anisotropy gap and separates multi-domain and single-domain branches. In thin metallic films and patterned devices, HF-FMR is often analyzed through Kittel-type dispersions or macrospin models, whereas in strain-coupled, optomechanical, and x-ray-resolved implementations the resonance is embedded in broader transduction chains (Zeisner et al., 2020, Beier et al., 21 Jul 2025, Alfonsov et al., 2016, Skowronski et al., 2019).

A recurring theme is that HF-FMR is not restricted to conventional microwave magnetic-field driving. The cited work includes magnetoelastic “tickle”-field excitation by surface acoustic waves, voltage-controlled magnetic anisotropy in magnetic tunnel junctions, optomechanical detection through magnetostriction, and element-specific x-ray readout at transition-metal absorption edges. This breadth is methodological rather than merely instrumental: the resonance condition remains the central observable, but the drive and detection channels vary substantially across material classes and geometries (Weiler et al., 2010, Skowronski et al., 2019, Colombano et al., 2019, Tanksalvala et al., 2024).

2. Instrumentation and measurement geometries

The experimental implementations represented in the literature differ mainly in how they deliver microwave excitation, define the sample geometry, and detect the resonant response.

Approach	Geometry or contrast	Reported capability
cw HF-ESR/HF-FMR (Zeisner et al., 2020, Beier et al., 21 Jul 2025)	Transmission, Faraday geometry	20–330 GHz in CrCl$_3$; 40–850 GHz in Fe$_3$GeTe$_2$; up to 16 T
EUV XFMR (Tanksalvala et al., 2024)	Reflection-mode T-MOKE on opaque substrates	Verified CW synchronization to 62 GHz; element specificity at Fe, Co, Ni M-edges
Microwave soft x-ray STXM (Bonetti et al., 2015)	Quasi-stroboscopic XMCD imaging	5–10 GHz with 35 nm spatial resolution
Cantilever-detected FMR (Alfonsov et al., 2016)	Mechanical torque detection with rotation	Up to 160 GHz and 15 T
SAW-driven FMR (Weiler et al., 2010)	Magnetoelastic excitation in a delay line	172 MHz to 3.6 GHz
Voltage-induced FMR in MTJs (Skowronski et al., 2019)	VCMA drive and spin-diode detection	Frequencies exceeding 30 GHz
FMR-assisted optomechanical readout (Colombano et al., 2019)	Magnetostriction to WGM cavity transduction	50 MHz to $\approx 1.1$ GHz

Transmission-mode HF-ESR/HF-FMR remains the most direct route to field–frequency maps over broad $f$-$g$0 space. In CrCl$g$1, a homemade continuous-wave spectrometer spanning 20–330 GHz with a Keysight PNA-X, oversized waveguides, and a superconducting solenoid up to 16 T was used in transmission geometry in the Faraday configuration, with complementary X-band measurements at $g$2 GHz. In Fe$g$3GeTe$g$4, a millimetre-wave vector network analyzer with an Oxford magnetocryostat provided 40–850 GHz, 1.7–300 K, and fields up to 16 T, again in transmission and Faraday geometry (Zeisner et al., 2020, Beier et al., 21 Jul 2025).

Device-compatible methods emphasize geometry as strongly as bandwidth. The tabletop HHG-EUV XFMR platform uses a lithographically patterned coplanar waveguide on thermally oxidized Si, a $g$5m EUV focus matched to the center conductor, and reflection-mode T-MOKE at 50° from grazing incidence; its present field range is limited to $g$6 T, but it operates on opaque substrates and active-device geometries. The microwave soft x-ray STXM approach instead relies on synchrotron timing, single-photon counting, and quasi-stroboscopic sampling, with permanent magnets up to 0.8 T perpendicular to the film plane and an electromagnet up to 0.25 T in-plane (Tanksalvala et al., 2024, Bonetti et al., 2015).

Mechanical and hybrid schemes extend HF-FMR into regimes where absorption is inferred indirectly. The cantilever spectrometer detects resonance through the change in magnetic torque on a piezoresistive AFM-type cantilever and uses a double-modulation scheme to subtract nonresonant backgrounds. The SAW experiment launches and detects GHz surface acoustic waves with interdigital transducers across an $g$7m acoustic delay line and reads resonance from changes in $g$8. The optomechanical magnetometer excites FMR in a YIG film with a shorted-end microstrip waveguide and detects the magnetostrictively driven motion of a BTS microsphere through a whispering-gallery-mode cavity (Alfonsov et al., 2016, Weiler et al., 2010, Colombano et al., 2019).

3. Resonance formalisms and parameter extraction

The common theoretical core of HF-FMR is the Landau–Lifshitz–Gilbert equation and the evaluation of resonance from the curvature of the magnetic free energy. In SI units, the cited work uses

$g$9

with $_3$0 the gyromagnetic ratio, $_3$1 the Gilbert damping parameter, and $_3$2 containing external, anisotropy, demagnetizing, exchange, or magnetoelastic terms (Bonetti et al., 2015, Weiler et al., 2010).

For field-polarized CrCl$_3$3, the resonance fields were calculated by minimizing a phenomenological free-energy density

$_3$4

and applying the Smit–Beljers resonance condition

$_3$5

evaluated at the equilibrium orientation. The authors did not use specialized closed-form Kittel formulas; instead they computed the equilibrium orientation and the resonance frequencies numerically (Zeisner et al., 2020).

Thin-film analyses often reduce to Kittel-type expressions. In the HHG-EUV XFMR study, the in-plane and perpendicular thin-film relations are written as

$_3$6

and

$_3$7

with $_3$8. The same study uses a dynamic-susceptibility fit to phase and amplitude, and quotes the standard linewidth relation

$_3$9

For short-wavelength modes, the exchange-augmented in-plane dispersion is

$_3$0

These forms are also the basis for several thin-film and imaging studies in the data set (Tanksalvala et al., 2024, Bonetti et al., 2015).

Strain-coupled and epitaxial systems require an enlarged free-energy description. Theoretical work on epitaxial films includes Zeeman, cubic anisotropy, demagnetization, and magnetoelastic terms, and predicts nonmonotonic $_3$1 across strain-induced spin reorientation transitions. SAW-driven FMR uses the same LLG framework but replaces an external microwave magnetic field with an internal magnetoelastic drive field $_3$2, obtained from the strain-dependent magnetoelastic energy. This distinction is operationally important because it produces a symmetry-selective angular dependence absent in standard microwave-driven absorption (Pertsev et al., 2011, Weiler et al., 2010).

4. Anisotropy, demagnetization, and domain structure

HF-FMR is especially effective at separating demagnetizing, magnetocrystalline, and domain-state contributions. CrCl$_3$3 is a particularly clean example. Single-crystal platelets were approximated as extended flat plates with $_3$4 and $_3$5, the 4 K HF-FMR data were taken entirely above the low-temperature saturation fields, and the measured frequency–field dispersions for $_3$6 and $_3$7 were reproduced with $_3$8. The extracted parameters were $_3$9, $_2$0, and $_2$1, used in the simulations as $_2$2. The apparent anisotropy is therefore dominated by demagnetizing fields set by platelet geometry, while the magnetocrystalline anisotropy is negligible within experimental resolution (Zeisner et al., 2020).

The temperature-dependent ESR in the same material demonstrates that the resonance field deviates from the paramagnetic position well above the 2D ordering temperature $_2$3 K. The shift $_2$4 is negative for $_2$5 and positive for $_2$6 below $_2$7 K, with larger magnitude for the hard-axis orientation. This establishes ferromagnetic short-range correlations above the ordered state and corroborates the quasi-2D easy-plane character inferred from the field-polarized FMR response (Zeisner et al., 2020).

Fe$_2$8GeTe$_2$9 represents the opposite limit, where magnetocrystalline anisotropy is large and decisive. At 2 K, the field dependence of the three observed magnon branches is described by a semiclassical domain-based model that yields an anisotropy field $_3$0 T, an anisotropy gap $_3$1 GHz, $_3$2, a domain-wall angle $_3$3, an in-plane saturation field $_3$4 T, and a uniaxial anisotropy constant $_3$5 erg/cm$_3$6. The branch $_3$7 softens and bends toward $_3$8, whereas $_3$9 and $_3$0 track the high-field single-domain response. A single-domain model fails to reproduce the bending of $_3$1, which makes the multi-domain interpretation structurally necessary rather than optional (Beier et al., 21 Jul 2025).

The temperature evolution in Fe$_3$2GeTe$_3$3 further shows that $_3$4 remains finite at $_3$5, with $_3$6 GHz, and vanishes only near $_3$7 K. Orientation-dependent resonance-field shifts persist to the same temperature scale. These data show quasi-static anisotropic short-range order above the static long-range ordering temperature and underscore the role of anisotropy in stabilizing ferromagnetism in a quasi-2D metallic system (Beier et al., 21 Jul 2025).

Strain engineering introduces a third anisotropy regime. In epitaxial cubic films, the FMR frequency can vary nonmonotonically with epitaxial strain and reach a minimum at a strain-induced spin reorientation transition. The theoretical examples in the cited work predict $_3$8 at a second-order transition in Fe$_3$9Co$_2$0 at $_2$1 under $_2$2 kOe, a first-order transition in CoFe$_2$3O$_2$4 at $_2$5 under $_2$6 kOe, and HF-FMR exceeding 200 GHz for tensile strains $_2$7 in CoFe$_2$8O$_2$9. Voltage-induced strain from ferroelectric substrates adds a further tuning axis, with field-dependent and, in some geometries, strongly enhanced tunability near a critical electric field that triggers spin reorientation (Pertsev et al., 2011).

5. Specialized excitation and imaging modalities

A major development in HF-FMR is the decoupling of resonance spectroscopy from conventional inductive detection. The tabletop HHG-EUV XFMR platform detects FMR in reflection-mode with linearly polarized EUV light at the M-edges of Fe, Co, and Ni, using a coherent high-harmonic source synchronized to an RF frequency comb. The present system demonstrates a continuous-wave bandwidth of 62 GHz, with synchronization verified at 40, 57, and 62 GHz, and reports timing jitter of 1.0–1.1 ps after removing the oscilloscope contribution. Element-resolved dynamics were demonstrated in Ni$\approx 1.1$0Fe$\approx 1.1$1, Co$\approx 1.1$2Fe$\approx 1.1$3, and a Ni/TaO$\approx 1.1$4/Fe multilayer, with phase-resolved fits to the dynamic susceptibility and comparison to in situ inductive FMR on the same coplanar waveguide (Tanksalvala et al., 2024).

Microwave soft x-ray STXM adds real-space imaging to phase-resolved FMR. In the reported implementation, a low-jitter microwave source phase-locked to the synchrotron clock and single-photon counting electronics enable quasi-stroboscopic detection from 5 to 10 GHz with 35 nm resolution. The method directly imaged a 6.27 GHz spin wave from a spin-torque oscillator and uniform FMR at 9.129 GHz in a Co microstrip, where the dynamic absorption cross-section change $\approx 1.1$5 corresponds to an out-of-plane precession amplitude of $\approx 1.1$6 after accounting for pulse-length convolution. This is not merely spectroscopic selectivity but phase-resolved mapping of the mode profile (Bonetti et al., 2015).

Elastic and voltage-based drives change the excitation symmetry. In SAW-driven FMR, the internal RF field is generated by magnetoelastic coupling to a propagating strain wave rather than by an external microwave magnetic field. The attenuation shows an approximately four-fold symmetry versus the in-plane angle $\approx 1.1$7, with no attenuation at $\approx 1.1$8 and $\approx 1.1$9, and maxima near $f$0. The data up to 2.24 GHz were reproduced with $f$1 mT, $f$2 mT, $f$3, and $f$4, and the work attributes the unusually large effective damping to inhomogeneous excitation by the finite SAW wavelength (Weiler et al., 2010).

Voltage-induced FMR in W-buffered CoFeB/MgO MTJs uses voltage-controlled magnetic anisotropy and spin-diode detection rather than an RF magnetic field. In a $f$5m$f$6 MTJ with $f$7 nm, post-annealing at $f$8C produced $f$9 MJ/m$g$00, $g$01 MJ/m$g$02, and $g$03 T, enabling V-FMR up to 31 GHz within a 1 T field limit. Reported damping values include $g$04 for the reference layer, $g$05 for the free layer in one geometry, and $g$06 for the free layer in a nearly perpendicular geometry after stronger annealing (Skowronski et al., 2019).

Two additional modalities show how far HF-FMR can be embedded into hybrid measurement chains. In the optomechanical magnetometer, the response depends on spectral overlap between the YIG FMR and mechanical breathing modes of a BTS microsphere, leading to a peak sensitivity better than 900 pT Hz$g$07 at 206 MHz and sensitivity around a few nT Hz$g$08 up to the GHz range. In the cantilever spectrometer, resonance is detected mechanically through changes in torque; the setup reaches 160 GHz and 15 T and reveals, across several films, a second resonance line with opposite angular dependence and linewidths of roughly 1–2 T at 120 GHz. These studies show that HF-FMR can be a transduction primitive as much as a spectroscopic one (Colombano et al., 2019, Alfonsov et al., 2016).

6. Materials, performance regimes, and research directions

The materials studied by HF-FMR cover isotropic van der Waals magnets, strongly anisotropic metallic ferromagnets, sputtered soft films, Heuslers, oxides, and hybrid device stacks. FeN films fabricated by reactive sputtering illustrate how dynamic parameters track microstructure. In the soft-magnetic window $g$09–$g$10 sccm nitrogen flow rate, clear FMR was observed from 2 to 18 GHz under fields up to 2000 Oe, with $g$11–$g$12, $g$13–$g$14 Oe, and high-field damping $g$15–$g$16 that was nearly insensitive to growth conditions. By contrast, the low-field damping increased strongly with nitrogen flow rate and correlated with the hard-axis coercivity, implicating extrinsic disorder, inhomogeneous broadening, and defect-sensitive relaxation channels (Hwang et al., 2016).

High-frequency access also exposes the limits of uniform-mode modeling. For Fe$g$17Ni$g$18, simultaneous fitting of 120–160 GHz data and low-frequency angular dependences gave $g$19 GHz/T, $g$20 erg/cm$g$21, and $g$22 erg/cm$g$23 at low temperature. For Sr$g$24FeMoO$g$25, however, a plausible fit of the low-field line using $g$26 per formula unit, $g$27 GHz/T, $g$28 erg/cm$g$29, and $g$30 erg/cm$g$31 did not reproduce the angular dependences with the same parameter set. The emergence of a second line above $g$32 GHz and the drastic linewidth increase with frequency suggest that multimode dynamics, including a possible perpendicular standing spin-wave contribution, become important in the millimeter-wave regime (Alfonsov et al., 2016).

Several application directions follow directly from the reported results. CrCl$g$33 is identified as a useful calibration or reference system because its HF-FMR is governed by shape anisotropy rather than intrinsic magnetocrystalline terms. Reflection-mode EUV XFMR is explicitly positioned toward measurements and imaging of active devices on arbitrary and opaque substrates, with a stated path to coherent diffractive imaging, ptychography, and holography. The YIG–microsphere platform uses FMR as the frequency-selective front end of a room-temperature high-speed magnetometer. VCMA-driven MTJs translate HF-FMR into an on-chip electrical format for resonators and detectors above 30 GHz (Zeisner et al., 2020, Tanksalvala et al., 2024, Colombano et al., 2019, Skowronski et al., 2019).

The same literature also defines current limitations. The EUV platform is presently limited by photon flux, $g$34 frame-to-frame EUV intensity fluctuations, incidence-angle optimization, a field ceiling of $g$35 T, and reduced high-frequency signal at small precession cones. The STXM implementation is ultimately constrained by the $g$36 ps synchrotron pulse width in the operating mode. SAW-driven FMR in extended films exhibits broadening attributed to finite-$g$37 excitation. Cantilever-detected HF-FMR requires aggressive background subtraction and encounters pronounced linewidth growth and mode multiplicity. In V-FMR, the linewidth can deviate from linearity at small $g$38 when the PMA is large and the field and magnetization are strongly noncollinear (Tanksalvala et al., 2024, Bonetti et al., 2015, Weiler et al., 2010, Alfonsov et al., 2016, Skowronski et al., 2019).

Taken together, the cited work presents HF-FMR as a family of techniques unified by resonance physics but diversified by geometry, transduction, and target observable. It can isolate pure shape anisotropy in a thin van der Waals platelet, quantify a 170 GHz anisotropy gap in a strongly anisotropic ferromagnet, resolve microstructure-dependent damping in sputtered films, and serve as the operative resonance in x-ray imaging, optomechanical sensing, and voltage-controlled spintronic devices. This suggests that the central evolution of HF-FMR is not a single expansion in frequency or field, but a convergence of high spectral range with element specificity, real-space imaging, hybrid transduction, and device compatibility.