Pith. sign in

REVIEW 5 major objections 6 minor 2 cited by

Self Biased Integrated Magnonic Device

T0 review · 5 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A silicon-integrated magnonic device operates at zero external magnetic field, using on-chip permanent micromagnets to generate a tunable bias from 20.5 to 11 mT and achieving up to 120 degrees of phase shift at 6 GHz.

desk verdict A genuinely first silicon-integrated, self-biased magnonic phase shifter with direct zero-field spin-wave evidence; the tuning curve is fit-based and needs a caveat but the core demonstration holds up. read the letter →

arxiv 2502.03186 v1 pith:JVBDYF6O submitted 2025-02-05 physics.app-ph

Maria Cocconcelli , Federico Maspero , Andrea Micelli , Alberto Toniato , Andrea Del Giacco , Nicola Pellizzi , Alejandro Enrique Plaza , Andrea Cattoni
show 8 more authors
Marco Madami Raffaele Silvani Abbass Hamadeh Christoph Adelmann Philipp Pirro Silvia Tacchi Florin Ciubotaru Riccardo Bertacco
This is my paper · ORCID
classification physics.app-ph
keywords magnonicsself-biaseddevicepermanentmicromagnetsmagneticfluxconcentratorsDamon-EshbachspinwavesphaseshiftersiliconintegrationRFsignalprocessing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper reports what it calls the first standalone magnonic device on a silicon chip that needs no external magnetic field: a CoFeB waveguide with two gold antennas, flanked by T-shaped magnetic flux concentrators and SmCo permanent micromagnets. The micromagnets create an internal transverse bias field $H_0$ in the waveguide, stabilizing Damon-Eshbach spin waves at zero applied field. Varying the fabrication distance $D$ between the concentrators and the magnets tunes $H_0$ from about 20.5 mT down to 11 mT, which shifts the spin-wave band and changes the phase of the RF transmission. The authors demonstrate phase tuning of up to 120 degrees at 6 GHz and propagation over the 3–8 GHz range in a $100\times150~\mu$m footprint. If established, this would remove a key obstacle to practical magnonic RF devices, since bulky electromagnets are currently the main barrier to on-chip integration.

What carries the argument

The load-bearing element is the symmetric assembly of two T-shaped magnetic flux concentrators (MFCs), made of a Py/Cr multilayer, coupled to rectangular SmCo permanent micromagnets on either side of the CoFeB waveguide. The T-shape is chosen because it concentrates the non-uniform stray field of the permanent magnets more effectively than trapezoidal or bar shapes, giving a measured gain $G \approx 2.4$ over the bare micromagnet field. The mechanism is geometric: moving the micromagnets a distance $D$ from the MFCs changes how much of the stray field is captured, and therefore tunes the internal bias $H_0$ independently of the field-line distribution. The DE spin-wave dispersion in the conduit is then modeled with a dipole-exchange (Kalinikos–Slavin-type) dispersion for metallic stripes, and the measured ${\rm Im}(S_{12})$ is fit to extract $H_0$ as a free parameter.

What would settle it

Measure the static field in the CoFeB conduit directly for each gap $D$, for instance by nitrogen-vacancy magnetometry or by comparing the zero-field BLS spectrum to spectra taken under accurately known external fields; if the field does not fall from about 20.5 mT to about 11 mT as $D$ goes from 0 to 12 $\mu$m, or if the phase shift at 6 GHz does not approach 120 degrees over $D=0$–8 $\mu$m, the central claim is not supported.

Watch

Extended reading notes

Core claim

The central claim is that a magnonic phase shifter can be made to operate without any external magnetic bias by co-integrating permanent magnets and flux concentrators with the spin-wave waveguide on a silicon substrate. In the device, two rectangular SmCo micromagnets are magnetized transversely and their stray field is funneled by T-shaped Permalloy flux concentrators into a 3.6-µm-wide CoFeB conduit, producing a uniform internal field $H_0$ strong enough to set a Damon-Eshbach configuration ($H_0$ perpendicular to the wavevector). By changing the gap $D$ between magnets and concentrators from 0 to 12 $\mu$m, the field is tuned from 20.5 mT to 11 mT, moving the spin-wave band and, at 6 GHz, the relative phase of the transmitted RF signal shifts by about 120 degrees over $D=0$–8 $\mu$m. The authors also report spin-wave decay lengths of 7.6 µm (zero field) versus 5.6 µm (100 mT) from micro-BLS, showing comparable propagation at zero bias. They present this as the first monolithic silicon integration of a self-biased, all-electric-input/output magnonic device.

Load-bearing premise

The reported internal bias field $H_0$ is not measured directly; it is inferred by fitting transmission spectra to a pure Damon-Eshbach dispersion model, and for the largest gap $D=12~\mu$m the authors themselves state that a pure DE configuration is not achieved, so the 11 mT figure rests on an assumed flux-concentrator gain $G=2.2$ applied to a loop shift.

Editorial extensions

If this is right

  • If the claim is right, magnonic RF components such as phase shifters and delay lines can be made on a silicon process with no power-hungry electromagnet, shrinking the footprint to roughly $100\times150~\mu$m.
  • The demonstrated phase tuning range (up to 120° at 6 GHz) is achieved by choosing $D$ during fabrication, i.e. a discretely tunable device; the paper argues that mounting the magnets on MEMS actuators could make the tuning continuous and real-time.
  • The frequency band (3–8 GHz in this proof of concept) is set by the CoFeB material and the bias field magnitude, and the paper argues that thicker or higher-remanence SmCo magnets could push the band into the FR3 range (7–24 GHz) targeted for beyond-6G.
  • The demonstrated zero-field spin-wave decay lengths comparable to the biased case mean that other magnonic functions (filters, interferometers) could in principle also be co-integrated on the same platform.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because $H_0$ is extracted via a dispersion fit, an independent check of the MFC gain (e.g., a magnetometry measurement of the concentrated field under a known applied field) would directly test the tuning curve without relying on the DE model.
  • The phase-shift figure at 6 GHz is read from the relative phase of $S_{12}$ across different physical devices, not a single reconfigurable device; the implied extension is that a single device with movable magnets should reproduce the same phase shift continuously, which is testable with a MEMS implementation.
  • If a lower-damping material such as YIG could be grown on the same platform, the same self-biasing scheme would extend the propagation distance far beyond the 5–10 µm scale set by CoFeB damping, potentially making practical RF filters feasible.
  • The claim 'first monolithic silicon integration' is a strong priority statement; the practical comparison with existing zero-static-power filters (e.g., magnetostatic-wave filters with integrated biasing circuits) would be the benchmark to watch as the technology matures.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper reports a proof-of-concept, fully integrated magnonic device on silicon that is claimed to operate without any external magnetic bias. The device consists of a CoFeB spin-wave waveguide with two RF antennas, flanked by T-shaped magnetic flux concentrators and SmCo permanent micromagnets; changing the distance D between concentrators and magnets is claimed to tune the internal transverse bias field from 20.5 mT (D=0 µm) to 11 mT (D=12 µm), enabling zero-field Damon-Eshbach spin-wave propagation and a phase shift of up to 120° at 6 GHz over D=0–8 µm. The central evidence is VNA transmission spectroscopy showing spin-wave oscillations at zero applied field (Fig. 2g–2i, Fig. 5a) and micro-BLS spectra and decay-length measurements at zero applied field (Fig. 4e–4g). The paper also claims a first monolithic silicon integration of a self-biased magnonic RF device with a 100×150 µm² footprint.

Significance. If the claims hold, this is a meaningful advance toward practical integrated magnonic RF components: it replaces the external electromagnet with on-chip permanent micromagnets and flux concentrators, demonstrates electric input/output on silicon, and provides direct VNA and BLS evidence of spin-wave propagation at zero applied field. The fabrication of the monolithic device and the zero-field propagation data are the paper's clear strengths. However, the quantitative tuning range (20.5→11 mT) and the derived time-delay/phase-shift tuning rest on model-extracted H0 values, and the D=12 µm point in particular is not obtained by the same fit procedure as the other points. These quantitative claims therefore need additional support or tempering before the paper can be accepted at its current strength.

major comments (5)
  1. [Section 2.2, Figure 5f] The D=12 µm point in the tuning curve is not derived from the DE fit used for D=0–8 µm. The text states that at D=12 µm "a pure DE configuration cannot be achieved" and that at zero applied field "a reliable fit cannot be performed"; the H0≈11 mT value is instead obtained as a ~5 mT loop shift multiplied by an assumed MFC gain G=2.2. This is load-bearing for the abstract claim that the internal bias field is tunable from 20.5 to 11 mT. Please provide a direct, model-independent measurement of the on-waveguide field at each D (at least at D=12), or restrict the reported tuning range to D≤8 µm and remove the D=12 point from Fig. 5f.
  2. [Section 2.4, Eq. (1)] The H0 values used to compute the theoretical dispersions and propagation times in Fig. 5d,e are outputs of the same DE-dispersion fit used to fit Im(S12) in Fig. 5a. Consequently, the agreement between the measured impulse-response delays (Fig. 5b) and the theoretical delays (Fig. 5e) is a consistency check of the model rather than an independent corroboration of the extracted H0 values. Please state this limitation explicitly and, if the time-delay tuning claim is to be quantitative, include an independent field estimate (e.g., BLS at zero field for each D).
  3. [Sections 2.1 and 2.2] The conversion H0 = G × ΔHa relies on the MFC gain G, which is measured as 2.4±0.3 at D=0 but assumed to be 2.2 at D=12 without an explained provenance or an uncertainty. The manuscript itself notes that G can drop below unity when the MFC partially saturates, so a single fixed G cannot be assumed across the D range. Report how G=2.2 was determined and propagate its uncertainty into Fig. 5f, or replace the extracted H0 curve with directly measured on-waveguide fields.
  4. [Section 2.4, Figure 5c,f] The phase-shift tuning claim of up to 120° at 6 GHz is presented without error bars, and the phase comparison is made across different fabricated devices. Because device-to-device variations in antenna placement, probe position, and reference-plane phase can contribute to S12, the uncertainty of the extracted relative phase should be quantified. Add error bars (or a table of repeated measurements) to Fig. 5f and state how many devices per D value were measured.
  5. [Section 3 vs. Section 2.4] The Discussion states that tunable delays "up to 150 ps" are achievable, but the experimental propagation times reported in Section 2.4 vary from 370 to 450 ps over D=0–8 µm, an 80 ps range. Please reconcile these numbers; if the 150 ps refers to a different definition (e.g., theoretical group-delay variation over the operating band), state that explicitly.
minor comments (6)
  1. [Eq. (1) and SI Section 3] The fitting function in the main text Eq. (1) uses sin(kr+φ), while SI Section 3 writes the same fit as cos(kr+φ); the phase convention should be made consistent.
  2. [Figure 5d caption] The caption says "D = 0 mm"; this should be "D = 0 µm".
  3. [Section 3] The estimated current density for generating 20 mT with a current line is given as "4·10-6 A/cm2"; an order-of-magnitude estimate for a 3.6 µm-wide, 1 µm-thick wire at 1 µm distance gives roughly 10^6 A/cm2, so the exponent appears to be a typo.
  4. [Section 2.4 and Discussion] The text says the phase shift at fixed frequency can be tuned by "about 100 degrees" in Section 2.4 but "about 120 degrees" in the Discussion and Abstract; please clarify whether these refer to different frequencies or different figures, and make the numbers consistent.
  5. [Figure 5f caption] The caption says the right scale reports H0 estimated from the fit of curves in panel 5a, but the D=12 value is not obtained by that fit; the caption should state the actual estimation method and its uncertainty for that point.
  6. [Abstract and Discussion] The "first demonstration" claim should be qualified against prior integrated zero-field magnonic work (Refs. [11] and [12]) by specifying the precise novelty (e.g., monolithic silicon integration with permanent-magnet bias and electrical I/O), to avoid an overbroad priority claim if those references contain similar integrated elements.

Circularity Check

1 steps flagged · score 4.0 of 10

The standalone zero-field operation and phase-shift data are directly measured; only the 'theoretical time delay' consistency check is partly built from the field values fitted to the same S12 traces.

  1. fitted input called prediction [Section 2.4 (Fig. 5d–e); Eq. (1) and Supporting Information Section 3]
    "In panel 5d we show the band dispersion of DE modes in our conduits according to the analytical model for SWs in ferromagnetic metallic stripes, for values of applied fields corresponding to the H0 from our analysis of S12(f) in standalone devices at various D (see panel 5f), together with the simulated antenna efficiency. ... This is qualitatively consistent with the trend of the experimental propagation time reported in Figure 5b."

    H0 is the free parameter used to fit Im(S12(f)) via Eq. (1) as described in SI Section 3, and the same fitted H0 values are then inserted into the analytical dispersion to compute the 'theoretical' propagation time Dtth = vg×r in Fig. 5e. The experimental arrival times in Fig. 5b are extracted from the very same gated S12(f) data by Fourier transform, so the frequency-domain fit and the time-domain wavepacket are representations of one dataset. The claimed consistency between Figs. 5b and 5e therefore largely follows from the fit itself rather than from an independent prediction. The directly measured phase-shift tuning (Fig. 5c/f) and zero-field spin-wave observation do not inherit this circularity; microMOKE loop shifts provide a partially independent estimate of H0.

full rationale

The central claim of the paper — a self-biased, all-electric magnonic device on silicon operating at zero applied field — rests on direct VNA time-domain and BLS observations of spin-wave propagation at Ha = 0, not on a fitted parameter. The phase-shift tunability (up to 120° at 6 GHz) is read directly from the relative phase of S12(f) between devices with different D. The only step that reduces toward its own input is the 'theoretical' time-delay comparison of Fig. 5d–e, where the H0 values used to compute vg are exactly the values fitted to the same S12 traces that yield the experimental delays; this is a self-consistency check, not an independent validation. No load-bearing uniqueness theorem or ansatz is imported from the authors' prior work; the on-chip magnet design is supported by COMSOL, microMOKE, and the VNA fits. The D=12 µm bias-field value (11 mT) is an extrapolation using a fixed MFC gain G=2.2 where the DE model is admittedly invalid, which is a correctness/robustness concern rather than a circular derivation. Overall, the central findings retain independent experimental content, so the circularity score is moderate rather than high.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a fitted DE dispersion model with several material and geometric parameters, on micromagnetic simulations for the demagnetizing field and single-domain threshold, and on the assumption that MFC gain and SmCo remanence remain stable. No new physical entities are introduced. The heaviest burden is that the published bias-field tuning curve is inferred, not measured, and uses fitting parameters that shift over time.

free parameters (10)
  • Saturation magnetization Ms = 1.3e6 A/m
    Table 1: used in Kalinikos-Slavin dispersion model and adjusted in fit to reference S12 traces.
  • Exchange stiffness A = 18e-12 J/m
    Table 1: fit parameter in dispersion model.
  • Gilbert damping alpha = 0.0043
    Table 1: fit parameter affecting attenuation length.
  • CoFeB thickness tCoFeB = 26.3 nm
    Table 1: fit to reference devices, nominal 25 nm.
  • Antenna thickness tantenna = 129 nm
    Table 1: fit, nominal 125 nm.
  • Antenna width wantenna = 1 um
    Table 1: fit, nominal 1 um.
  • Silica spacer thickness tspacer = 70 nm
    Table 1: fit, nominal 70 nm.
  • Per-trace amplitude S, phase j, attenuation Latt = varied per curve
    Eq. 1: free parameters in each Im(S12) fit.
  • Internal bias field H0 per device = 20.5 to 11 mT
    Fit parameter in the DE dispersion model used to produce the tuning curve in Fig. 5f.
  • MFC gain G = 2.4 +/- 0.3 (MOKE), 2.2 (used for D=12)
    Used to convert microMOKE loop shift into H0; value derived from slope ratio, not direct field measurement.
assumptions (6)
  • domain assumption Kalinikos-Slavin dipole-exchange DE dispersion model for a transversely magnetized metallic stripe is valid for the CoFeB waveguide
    Section 3 SI: the fit of Eq. 1 relies on this dispersion; authors note it fails below 20 mT for reference conduits and for D=12 standalone.
  • domain assumption Demagnetizing field of the saturated CoFeB conduit is -6.3 mT, as computed by mumax3
    Section 2.2: used to convert fitted effective field He into H0.
  • domain assumption The MFC gain G is constant across the central region and equal to the microMOKE slope ratio of about 2.4
    Section 2.1 and Fig. 1: H0 = G x DHa, and the D=12 estimate uses G=2.2.
  • domain assumption Reference subtraction at a compensating external field removes all spin-wave contributions, isolating the electromagnetic background
    Section 3 SI: residual drift and direct coupling effects are acknowledged; time gating is then applied.
  • domain assumption SmCo micromagnets remain saturated along x with remanent magnetization 0.6 T during and after magnetization
    Simulation section and methods; the authors separately report degradation of SmCo over one month, so this assumption has limited validity over time.
  • standard math Antenna excitation efficiency h(k) equals the spatial Fourier transform of the Oersted field of a rectangular stripline
    Section 3 SI: used in the fitting model to weight spin-wave modes by frequency.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Self Biased Integrated Magnonic Device." pith.science (2026). https://pith.science/paper/JVBDYF6O

@misc{pith2026250203186,
  author       = {Pith},
  title        = {Pith review of: Self Biased Integrated Magnonic Device},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JVBDYF6O}},
  note         = {Machine review of arXiv:2502.03186}
}
abstract

In the race towards "beyond 6G" telecommunication platforms, magnonics emerges as a promising solution due to its wide tunability within the FR3 band (7-24 GHz). So far, however, the need for an external magnetic bias field to allow the coherent excitation of spin waves has been a major bottleneck. Conventional bulky electromagnets are power-intensive and challenging to integrate on-chip, restricting magnonic applications largely to academic research. Here, we present the first demonstration of a standalone, tunable magnonic device featuring all-electric input and output, fully integrated on a silicon substrate with a compact footprint of 100 x 150 $\mu$m. The device consists of a CoFeB waveguide equipped with two radio frequency antennas, flanked by a symmetric configuration of T-shaped magnetic flux concentrators and rectangular SmCo permanent micromagnets. By varying the distance D between the flux concentrators and the permanent magnets from 0 to 12 $\mu$m, the transverse bias field can be tuned from 20.5 mT to 11 mT, respectively. This variation directly modulates the dispersion relation of Damon-Eshbach spin wave modes in the CoFeB waveguide. In these proof-of-concept devices, the spin wave frequency band ranges from 3 to 8 GHz, with precise phase shift tuning of up to 120 degrees at 6 GHz achieved by varying D within the 0-8 $\mu$m range. The operational frequency band could even be pushed to higher frequencies through optimized micromagnet engineering.

Figures

Figures reproduced from arXiv: 2502.03186 by the authors.

Figure 1
Figure 1. a) False-color SEM image and schematic cross section of a device. b) COMSOL Multiphysics simulation of the stray field H produced by the assembly of SmCo permanent magnets and MFC made of Permalloy (field lines and blue color scale for the Hx component), combined with the simulation of the micromagnetic configuration of the CoFeB conduit (arrows and Brown color scale for the Mx component). c) Mx vs Hx loops measured… view at source ↗
Figure 2
Figure 2. Panels a,b,c report VNA measurements on reference CoFeB conduits without MFC and permanent magnets. a) 2D color map of the oscillations related to the imaginary part of S12 vs frequency for various applied fields Ha. b) Experimental S12(f) curves (continuous lines) for selected values of Ha (numbers in mT outside the brackets close to each curve), together with the result of the fit using equation 1 (dashed lines) f… view at source ↗
Figure 4
Figure 4. a) False-color SEM image of the investigated device. b) and e) Spin wave spectra recorded by micro-BLS at μ0Ha =100 mT and μ0Ha = 0 mT, respectively. c)Two-dimensional map of the BLS intensity recorded at a frequency of 12.6 GHz excited by the inductive antenna. An external field μ0Ha =100 mT was applied along the short axis of the conduit. d) and g) Spin-wave intensity (linear scale) as a function of the propagatio… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Resonance-enhanced Floquet cavity electromagnonics

    cond-mat.mes-hall 2025-05 conditional novelty 6.0 of 10

    A series capacitor turns the magnon drive coil into an LC resonator, boosting the Floquet drive by about six times and enabling GHz-to-MHz control of cavity magnon modes.

  2. Review on spin-wave RF applications

    physics.app-ph 2024-11 conditional novelty 3.0 of 10

    A comprehensive review of spin-wave RF components that concludes YIG-based magnonic devices show strong promise for 5G and 6G communication systems.

Reference graph

Works this paper leans on

2 extracted references · 1 canonical work pages · cited by 2 Pith papers

  1. [1]

    Forward V olume - FW

    INTRODUCTION So far magnonics, i.e. the branch of magnetism studying spin waves (SW) in magnetic media, essentially remained an exciting academic research field dealing with this peculiar form of electromagnetic waves in ferro/ferri/antiferro-magnetic materials displaying intriguing phenomena [1][2][3][4]. At variance with other forms of waves, the band-d...

    work page

  2. [5]

    MandMEMS

    EXPERIMENTAL SECTION Device fabrication. The whole fabrication process is made of four main steps. 1) 1 µm thick SmCo permanent magnets (100x40 µm footprint) embedded in a Si(001) coupon (20x20 mm) are realized by sputtering deposition in a trench previously defined by reactive ion etching, with a mesoporous silica layer used for subsequent lift off. Upon...

    arXiv 2021

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.