{"id":"ef04f31a-c551-4cb1-9449-eeda0658176a","arxiv_id":"2502.03186","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"A fully silicon-integrated magnonic phase shifter with on-chip permanent micromagnets operates without external bias and is tunable from 20.5 to 11 mT, giving up to 120 degrees of phase shift at 6 GHz.","lead":"This paper reports a tiny silicon chip that guides magnetic spin waves using built-in permanent micromagnets, so it works without any external electromagnet. The device shifts microwave signals by up to 120 degrees in the 3 to 8 GHz range, a step toward integrating magnonic radio components into phones.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 20.5-to-11 mT tuning curve rests on a DE-only fit for D=0–8 µm and, at D=12 µm, on a fixed MFC gain G=2.2 where the authors exclude pure DE; an independent field measurement is needed.","rationale":"The reader identifies the same weakest assumption, and I agree. The paper's strongest independent evidence is that zero-field VNA oscillations and BLS propagation do occur at D=0, so the basic 'standalone' claim has real support; my concern is not with existence of propagation but with the quantitative H0(D) curve that is a headline result. The microMOKE data corroborate D=0–8, so this is not a blanket rejection of the fitting procedure. However, the D=12 point is outside the model by the authors' own admission and the fixed-gain estimate is unconvincing. The phase-shift result is less vulnerable than the field-tuning curve, because the raw S12 phase curves still shift with D even if the absolute H0 values were wrong; but the paper's interpretation of that shift via the DE dispersion depends on H0(D). Fabrication variability between the discrete D devices is a secondary issue, and the magnet degradation is disclosed by the authors, so I would not raise those as primary. The reader's CONDITIONAL verdict is appropriate; making the condition explicit — require an independent measurement or a full multimode fit at D=12 — would tighten it.","tokens_in":17699,"tokens_out":8789,"duration_ms":83387,"concrete_test":"Decisive check: perform scanning NV-center magnetometry or micro-BLS field-map measurements on the standalone devices with D=0, 8, and 12 µm after the same 5.4 T magnetization protocol, measuring the magnetic field profile under the antennas and comparing it with the H0 values in Fig. 5f. The D=12 device is the critical case: the measured field should fall within the uncertainty band implied by the G=2.2 vs 2.4 discrepancy (about ±1.5 mT) if the tuning curve is correct; if it does not, the headline tuning range and the mechanism invoked for the 120° phase shift need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim — internal bias field tunable from 20.5 mT (D=0) to 11 mT (D=12), and the use of that curve to interpret the 120° phase shift — is not established by direct measurement. For D=0–8 µm, H0 is extracted by fitting Im(S12) to a single-mode Damon–Eshbach dispersion model (Eq. 1 of the Supporting Information, Sec. 3) with H0 as a free parameter. The microMOKE loop-shift data provide an independent estimate for D=0 and D=8 (21.6±3.4 and 15±2.4 mT), but only if the MFC gain G=2.4±0.3 is assumed linear and local. The D=12 point is weaker still: the authors state that at D=12 µm 'a pure DE configuration cannot be achieved,' yet Fig. 5f includes H0≈11 mT for that case, obtained by multiplying a roughly 5 mT map shift by G=2.2. The same text notes that G can fall below unity when the MFC partially saturates, so a single fixed G cannot be carried across the D range. With no error bars on the fitted H0 values and no direct measurement of the on-waveguide field, the tuning range '20.5→11 mT' and the D-to-phase-shift interpretation rest on a model whose validity is explicitly limited to D≤8 µm. That is the load-bearing assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","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.","tokens_in":18067,"tokens_out":7989,"duration_ms":67528,"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":[{"comment":"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.","section":"Section 2.2, Figure 5f"},{"comment":"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).","section":"Section 2.4, Eq. (1)"},{"comment":"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.","section":"Sections 2.1 and 2.2"},{"comment":"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.","section":"Section 2.4, Figure 5c,f"},{"comment":"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.","section":"Section 3 vs. Section 2.4"}],"minor_comments":[{"comment":"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.","section":"Eq. (1) and SI Section 3"},{"comment":"The caption says \"D = 0 mm\"; this should be \"D = 0 µm\".","section":"Figure 5d caption"},{"comment":"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.","section":"Section 3"},{"comment":"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.","section":"Section 2.4 and Discussion"},{"comment":"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.","section":"Figure 5f caption"},{"comment":"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.","section":"Abstract and Discussion"}],"recommendation":"major_revision","confidential_remarks":"The zero-field propagation evidence is direct and credible; the main risk is that the paper overextends its quantitative tuning and phase-shift claims beyond what the model-extracted H0 values and the single D=12 estimate can support. I would advise the editor to request an independent measurement of the on-waveguide field (or a clear reduction of the claimed tuning range), error bars on Fig. 5f, and reconciliation of the time-delay numbers before publication. This is fixable within the scope of a major revision and does not require rejecting the core standalone-operation result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is the first standalone magnonic device with all-electric I/O integrated on silicon that runs without any external magnet. The zero-field evidence is direct — VNA transmission oscillations at 0 mT in Fig. 2g–2i and BLS spectra at 0 mT in Fig. 4e–4g — not just inferred. The paper is also unusually honest about its limits: it says outright that D is fixed during fabrication, that the spin-wave signal is only 1% of the RF transmission, and that the SmCo magnets degraded over a month. That honesty raises my confidence in the parts that matter.\n\nWhat's new is the monolithic co-integration of CoFeB waveguide, inductive antennas, T-shaped MFCs, and SmCo permanent micromagnets on Si, giving a 100×150 µm footprint. Ref. [11] did zero-field spin waves in YIG nanowaveguides without on-chip bias; Ref. [12] used bias circuitry, not permanent magnets. The authors' own Ref. [15] had the ingredients but not the full integration. So the first-demo claim is credible.\n\nThe soft spot is the tuning curve. H0 is a fitted free parameter in a pure Damon–Eshbach dispersion model for D=0–8 µm, and for D=12 µm the authors themselves say a pure DE configuration can't be achieved, yet they still report 11 mT from a loop shift times a fixed gain G=2.2. They also note G can drop below unity if the MFC partially saturates. So the exact 20.5→11 mT range and the 120° phase-shift-versus-D interpretation rest on a model that is explicitly invalid at one endpoint. That's worth flagging. But it's not fatal: the microMOKE loop shifts give 21.6±3.4 mT (D=0) and 15±2.4 mT (D=8), independent of the DE fit, and the VNA/BLS estimates at D=0 are consistent (18 mT vs 15–16 mT, the spread explained by magnet history). So the standalone operation and rough tunability are solid; only the precise values and the D=12 point need qualification.\n\nWho is this for? Applied magnonics and RF-component people. It's a proof-of-concept, not a deliverable — the 1% signal fraction alone precludes practical use. But as an integration milestone it deserves a serious referee, and the limitations are stated rather than buried. My recommendation: send it out. Ask for error bars on the fitted H0 values and a clear statement that D=12 is an estimate outside the DE model, and it's publishable.","headline":"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.","tokens_in":18698,"tokens_out":1354,"would_cite":true,"duration_ms":14096,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"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.","keywords":["magnonics","self-biased device","permanent micromagnets","magnetic flux concentrators","Damon-Eshbach spin waves","phase shifter","silicon integration","RF signal processing"],"falsifier":"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.","tokens_in":17474,"feed_emoji":"🧲","tokens_out":5392,"duration_ms":41508,"temperature":0.7,"pith_summary":"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.","feed_headline":"First fully integrated magnonic device works with no external magnet","feed_subtitle":"SmCo micromagnets and flux concentrators bias a CoFeB waveguide, shifting RF phase up to 120 degrees at 6 GHz.","key_machinery":"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.","core_discovery":"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.","pith_inferences":["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."],"forward_implications":["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."],"supporting_citations":[{"why":"Establishes the prior method of tuning magnonic devices with on-chip permanent micromagnets that this work extends to full silicon integration.","marker":"[15]"},{"why":"Demonstrates zero-field spin waves in YIG nanowaveguides, serving as the comparative zero-bias approach for this device.","marker":"[11]"},{"why":"Reports frequency-tunable magnetostatic wave filters with zero static power biasing circuitry, another no-bias approach this work contrasts with.","marker":"[12]"},{"why":"Provides the dipole-exchange dispersion model for spin-wave modes in metallic stripes used to fit Im(S12) and extract H0.","marker":"[21]"},{"why":"Supplies the magnonic-waveguide micro-BLS methodology and mode analysis used for the zero-field propagation checks.","marker":"[19]"},{"why":"Supplies the time-of-flight VNA spectroscopy method used to measure spin-wave dispersion.","marker":"[17]"}],"fun_headline_variants":["No external magnet: first self-biased magnonic phase shifter","Integrated micromagnets eliminate external field for spin waves","Self-biased magnonics: built-in magnets oust external bias","First magnet-free magnonic device hits 120° phase shift on silicon","Magnonics goes standalone: integrated permanent magnets replace bulky electromagnets"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"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.","fun_headline_variants_meta":{"raw":{"variants":["No external magnet: first self-biased magnonic phase shifter","Integrated micromagnets eliminate external field for spin waves","Self-biased magnonics: built-in magnets oust external bias","First magnet-free magnonic device hits 120° phase shift on silicon","Magnonics goes standalone: integrated permanent magnets replace bulky electromagnets"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00089,"raw_usage":{"total_tokens":3910,"prompt_tokens":1084,"completion_tokens":2826,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":700,"completion_tokens_details":{"reasoning_tokens":2737}},"tokens_in":700,"tokens_out":2826,"duration_ms":20362,"temperature":1.0,"reasoning_tokens":2737,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T05:35:27.579609+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"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.","supporting_citations":[],"review_version":1}