{"id":"5cbad26b-8fac-4494-908d-34db43db9db3","arxiv_id":"2501.00980","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Voltage-induced strain from a PMN-PT substrate rotates the magnetic easy axis of Bi-YIG films by 90 degrees and shifts reversal fields in patterned microstructures.","lead":"Voltage applied to a piezoelectric crystal can stretch a thin magnetic garnet film and flip the direction its magnetization prefers to point. The effect, shown in bismuth-substituted yttrium iron garnet films and microstructures, hints at lower-energy ways to write magnetic memory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim rests on unmeasured strain transfer through the SiO2 buffer and on excluding non-strain voltage effects; this premise is plausible but not yet directly supported, so conditional acceptance is appropriate.","rationale":"The reader's conditional verdict is well calibrated. I read the manuscript carefully; the hysteresis loops in Fig. 3 show a clear, monotonic increase in squareness along x and decrease along y with voltage, and the domain images in Figs. 4 and 5 are consistent with an easy-axis reorientation. The second sample provides a reproducibility check. However, the central causal statement—'strain-mediated'—depends on the assumption that the piezoelectric strain of the PMN-PT is transmitted through the 2.4 nm SiO2 buffer and that no other voltage-driven mechanism contributes. This is the load-bearing assumption because if strain is not transmitted, or if the changes are dominated by charge or heating, the quantitative claims (ME coefficient of 1.05×10^-7 s/m, comparison with other oxide systems) are not supported. The paper does not measure the film strain, nor does it include a control on a non-piezoelectric substrate. The concern is not that the data contradict the interpretation; rather, an essential link in the argument is inferred rather than observed. An in-situ XRD strain measurement would settle this directly. If that measurement confirms the expected strain, the paper would be acceptable; if not, the central claim would need revision. I agree with the reader's weakest-assumption identification, and I recommend no change to the conditional verdict.","tokens_in":13794,"tokens_out":8633,"duration_ms":81821,"concrete_test":"Perform in-situ synchrotron X-ray diffraction on the PMN-PT/SiO2/Bi-YIG stack while sweeping the voltage from 0 V to 450 V, measuring the Bi-YIG in-plane lattice spacings along x and y. If the measured film strain does not reproduce the expected PMN-PT strain (compressive along x and tensile along y at 450 V relative to the poled 0 V state) with a magnitude sufficient to produce the observed anisotropy changes, the magnetoelastic interpretation fails. This single measurement directly tests the unmeasured strain-transfer premise.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main claim is that voltage-induced strain from the PMN-PT substrate reorients the easy axis of Bi-YIG by 90°. The paper infers the film strain from the known piezoelectric response of PMN-PT (refs [30,46]) and interprets the MOKE hysteresis loop changes using the magnetoelastic energy expression in Sec. III with a literature λ_s ≈ −4×10^-6. What is never measured is the actual strain in the 45.6 nm Bi-YIG film after transfer through the 2.4 nm amorphous SiO2 buffer. If the buffer partially decouples the film from the substrate, or if the poled remanent strain relaxes differently in the film, the sign and magnitude of the magnetoelastic anisotropy could differ from the model, and the squareness changes could instead reflect voltage-driven charge accumulation at the Bi-YIG/SiO2 interface or Joule heating. The paper contains no control experiment on a non-piezoelectric substrate and no in-situ strain measurement, so the central 'strain-mediated' attribution is not uniquely established. That said, the hysteresis data are internally consistent with a strain-driven easy-axis rotation, so the issue is missing evidence rather than a demonstrated contradiction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports voltage-controlled magnetic anisotropy and magnetization reversal in 45.6 nm (and 55 nm) polycrystalline Bi-substituted YIG films grown on (011)-oriented PMN-PT substrates through a 2.4 nm amorphous SiO2 buffer. Longitudinal MOKE hysteresis loops measured along the two in-plane axes x̂ = [100] and ŷ = [01-1] show that the loop squareness increases along x̂ and decreases along ŷ as the voltage across the 0.5 mm substrate is increased from 0 to 450 V, while polar MOKE loops show no significant out-of-plane response. Domain imaging at fixed magnetic field demonstrates voltage-driven magnetization reversal, and the Mr/Ms versus V data form a butterfly loop. The authors extract a magnetoelectric coefficient of 1.05 x 10^-7 s/m and compare it with other ferroelectric/ferromagnetic bilayers. Voltage also tunes the switching fields of patterned elliptical dots and 5 um racetracks, and a second, 55 nm thick film is reported to reproduce the trend. The central claim is that piezoelectric strain from the PMN-PT rotates the Bi-YIG in-plane easy axis by 90 degrees, from ŷ at 0 V to x̂ at 450 V, interpreted through a magnetoelastic energy expression with a literature value of the magnetostriction lambda_s of about -4 x 10^-6.","tokens_in":14005,"tokens_out":37206,"duration_ms":339598,"significance":"If correct, this is a valuable extension of voltage-controlled magnetism to bismuth-substituted garnets, a material class of interest for magneto-optics, magnonics, and low-damping spintronics. It builds on the authors' own buffer-layer strategy, previously demonstrated for Y-DyIG, and shows that the effect survives in patterned mesostructures. The evidence base is genuinely multi-modal - two film thicknesses, hysteresis and MOKE domain imaging, and three patterned geometries - and the interpretation does not rely on fitted model parameters; the magnetoelastic analysis uses a literature value of the magnetostriction and the known biaxial piezoelectric response of PMN-PT(011). The reported magnetoelectric coefficient places Bi-YIG competitively among oxide/ferroelectric bilayers.","major_comments":[{"comment":"The central attribution of the observed effects to strain is inferred rather than demonstrated. The film strain is never measured: the manuscript takes the PMN-PT surface strain from the known piezoelectric coefficients (refs [30,46]) and assumes it is transmitted without relaxation through the 2.4 nm amorphous SiO2 buffer into the 45.6 nm Bi-YIG film, even though the film underwent a 600 C, 72 h crystallization anneal that is far above the PMN-PT Curie temperature. There is also no control experiment on a non-piezoelectric substrate (Bi-YIG on Si or on fused silica is already available to the authors) under the same voltage protocol, so voltage-driven charge accumulation or Joule heating are excluded only by plausibility. I want to be clear that the strain mechanism is likely: the easy-axis response develops with opposite signs along the two in-plane axes, which is the signature of the biaxial PMN-PT(011) piezoelectric response, and the stack is largely insulating, making carrier-mediated effects unlikely. However, because the manuscript's headline claim is explicitly 'strain-mediated voltage control', the authors should either add a control or a strain measurement (e.g., in-situ XRD or a strain gauge on the same poling sequence), or explicitly re-frame the claim as consistent with strain mediation and quantify the transfer-efficiency uncertainty.","section":"Secs. II-IV; Fig. 3"},{"comment":"The magnetoelastic interpretation in Sec. III requires the strain along each in-plane axis to change sign between the 0 V remanent state and the 450 V state: after poling and relaxation the text assigns tensile strain along x̂ and compressive along ŷ, while at the applied voltage the same axes are assigned compressive (negative d31) and tensile (positive d32). The easy-axis data are consistent with this assignment, but the sign relationship between the remanent and field-induced strains of the specific PMN-PT poling protocol is never documented, and the voltage polarity convention is not stated. Please provide the strain-vs-voltage butterfly for the substrate (measured, or quoted quantitatively from ref [46] or ref [30]), and state the polarity, so that the statement 'consistent with a negative magnetostriction' at both 0 V and 450 V can be verified rather than taken on trust.","section":"Sec. III, Fig. 3"},{"comment":"The quantitative claim that the easy axis reorients by exactly 90 degrees is based on hysteresis loops and domain images taken only along the two orthogonal in-plane directions x̂ and ŷ (Figs. 3 and 5). The data are fully consistent with an easy axis that is closer to ŷ at 0 V and closer to x̂ at 450 V, but for a polycrystalline film without strong texture (Fig. 1a), the actual easy-axis angle is set by the ratio of the two biaxial anisotropy components and cannot be fixed from two axes alone. An angular series of remanence or coercivity at 0 V and 450 V would directly establish the 90 degree rotation; alternatively, the claims in the abstract and in Sec. VI should be tempered to 'reorientation between the two orthogonal in-plane directions'.","section":"Secs. III-IV, Figs. 3-5"}],"minor_comments":[{"comment":"The film composition is given as Bi2.13Y1.40Fe5Ox; the cation sum Bi + Y = 3.53 does not match the garnet formula, and the oxygen content is unspecified. Please give the correct formula and state how the literature value lambda_s = -4 x 10^-6 (refs [47,48]) applies to this specific Bi content.","section":"Sec. II"},{"comment":"In the paragraph describing reversal along x̂, the parenthetical 'which corresponds to an easy x̂' is inconsistent with the surrounding discussion of the 0 V hard-axis case; please clarify which voltage value and which panels (top or bottom) of Fig. 4b are being compared.","section":"Secs. III-IV, Fig. 4"},{"comment":"The statement that white-contrast domains 'increase as the voltage is reduced' should be reconciled with the subsequent explanation about 'weakening of contrast'; the observed increase is in fact consistent with the decrease of Hc along x̂ from 27 mT to 25 mT (Fig. 3a) and could be explained that way.","section":"Sec. IV, Fig. 5"},{"comment":"For the magnetoelectric coefficient alpha_E = mu0 Delta M / Delta E, please state explicitly that t in Delta E = Delta V / t is the PMN-PT substrate thickness (0.5 mm), and specify the voltage-sweep protocol used to obtain the butterfly loop of Fig. 3c (sequence, increment size, and dwell time).","section":"Sec. IV"},{"comment":"The instrument citation '[40]' for the Rigaku diffractometer does not match the reference (Wu et al., Physical Review Applied 2021, on spin-orbit-torque MRAM); the reference list should be checked for this citation and for the Y-DyIG growth citation [28].","section":"Sec. II"},{"comment":"In the conclusion, 'pulsed layer deposition' should read 'pulsed laser deposition'.","section":"Sec. VI"},{"comment":"The reproducibility claim for the 55 nm film (magnetoelectric coefficient 0.9 x 10^-7 s/m) is quoted without supporting data; please show the corresponding loops or place them in the Supplementary Information.","section":"Sec. IV"},{"comment":"The quoted FMR linewidth of about 200 mT is very large relative to the expected resonance field range given 4 pi M_s of about 0.13 T; please provide the field and frequency ranges and the fitting procedure so the reader can judge the reliability of this value, or soften the statement.","section":"Sec. IV, SI S2"},{"comment":"The statement that the as-deposited sample is isotropic in the plane is an important baseline; the corresponding pre-poling hysteresis loops should be shown rather than described.","section":"Sec. III"}],"recommendation":"major_revision","confidential_remarks":"The novelty relative to the authors' own earlier work (refs [28,30], DyIG on PMN-PT with the same SiO2-buffer process) is real but incremental: Bi-YIG is a distinct material with magneto-optical relevance, the patterned mesostructures are new, and the magnetoelectric-coefficient comparison is useful. The main gatekeeping question is whether the strain-mediation claim, which leans on refs [28,30] and [46] for the substrate behavior, needs a direct strain measurement or control experiment; I consider this necessary for the strong form of the claim, while the raw voltage-control observations themselves are well supported. I would also suggest that the editor consider whether the FMR subsection, which is much weaker than the rest of the manuscript, is needed at all."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Best to know: this is the first report of voltage-induced 90 degree easy-axis reorientation in Bi-YIG on PMN-PT, and the data are more consistent than the typical \"one loop at two voltages\" paper. You get hysteresis squareness changes along both in-plane axes in the right direction, MOKE domain images at fixed field that track the voltage, a second 55 nm sample with the same trend, and a butterfly Mr/Ms vs V. That is real experimental evidence, and I believe the central claim.\n\nWhat is new: applying the established SiO2-buffer/PMN-PT strain platform to high-Bi-YIG, which has larger magnetostriction than YIG, and showing electrically tunable reversal fields in patterned dots and racetracks. The ME coefficient 1.05e-7 s/m lands in the same range as other oxide/ferroelectric bilayers. The paper is honest that FMR linewidth is broad (200 mT) and attributes it to polycrystalline microstructure and stress.\n\nSoft spots, in proportion. The biggest one is exactly what the stress-test says: the strain in the 45.6 nm film is never measured. The model uses bulk PMN-PT piezoelectric coefficients and a literature lambda_s, and assumes the 2.4 nm amorphous SiO2 buffer transfers the strain faithfully. If the buffer decouples or the poled remanent strain relaxes inhomogeneously, the sign and magnitude of the magnetoelastic anisotropy could differ. Charge accumulation at the Bi-YIG/SiO2 interface or Joule heating are not excluded by any control experiment. That is missing evidence, not a demonstrated contradiction, but it makes the \"strain-mediated\" attribution slightly weaker than the paper's confidence suggests.\n\nMinor issues: the ME coefficient is calculated from the slope of an Mr/Ms vs V butterfly with no error bar and only two points near the maximum; the threshold voltage for easy-axis switching is not precisely defined; and the title says \"magnetization reversal\" when the data show easy-axis rotation and modulation of reversal fields, not full deterministic reversal. The patterned structures are 5-30 um, so claims about \"energy efficient non-volatile memory\" are speculative but that is standard for this field.\n\nThe citation pattern is fine. The self-citations are for the specific growth method and prior Y-DyIG result, which is the relevant prior work.\n\nWho is this for: people working on voltage-controlled garnets, magneto-optics, or multiferroic heterostructures. It is a useful data point, not a breakthrough. It deserves a serious referee: the experiments are careful enough to warrant full review, and the strain-transfer concern is addressable (e.g., in-situ strain measurement via XRD, control on a non-piezoelectric substrate). I would send it out, with the expectation of revisions rather than rejection.","headline":"First voltage-controlled easy-axis rotation in Bi-YIG on PMN-PT, with consistent MOKE and hysteresis evidence; the strain-transfer mechanism is plausible but not directly proven, so conditional.","tokens_in":14574,"tokens_out":1767,"would_cite":true,"duration_ms":14647,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Applying a voltage rotates the magnetic easy axis of a Bi-YIG film by 90 degrees.","keywords":["magnetoelastic coupling","voltage-controlled magnetism","bismuth substituted yttrium iron garnet","PMN-PT piezoelectric substrate","magnetic anisotropy rotation","MOKE microscopy","magnetoelectric coefficient","racetrack memory"],"falsifier":"Measure the in-plane strain of the Bi-YIG film directly (for example, by synchrotron X-ray diffraction or by comparing films on PMN-PT with identical films on a non-piezoelectric substrate) while sweeping the voltage, and check whether the hysteresis-loop changes track the measured strain. If the loop changes persist without corresponding film strain, or if the strain flattens while the magnetic changes continue, the magnetoelastic mechanism is falsified; a simpler test is applying the same voltage to Bi-YIG on a fused silica substrate and seeing whether the MOKE loops change at all.","tokens_in":13553,"feed_emoji":"⚡","tokens_out":6444,"duration_ms":50611,"temperature":0.7,"pith_summary":"This paper reports that applying a voltage to a PMN-PT piezoelectric substrate can rotate the magnetic easy axis of a bismuth-substituted yttrium iron garnet (Bi-YIG) film by 90 degrees, switching the preferred in-plane magnetization direction between two orthogonal axes. The rotation is driven by voltage-induced strain transmitted through a thin amorphous SiO2 buffer into the garnet film, which couples to the film's negative magnetostriction. In-situ magneto-optical Kerr effect (MOKE) microscopy shows hysteresis loops becoming progressively square along one axis and harder along the other as the voltage increases from 0 to 450 V, and domain imaging confirms the easy-axis switch. The resulting magnetoelectric coefficient, 1.05 × $10^{-7}$ s/m, is comparable to other ferrimagnetic oxide films on ferroelectric substrates. This opens a voltage-based route to controlling magnetization in a garnet, a material class of interest for spintronic and magnonic applications.","feed_headline":"Voltage rotates a garnet film's easy axis by 90 degrees","feed_subtitle":"Strain from a piezoelectric substrate tunes magnetization direction in Bi-YIG, a step toward low-power magnetic memory.","key_machinery":"The mechanism is the magnetoelastic energy of a negative-magnetostriction film under anisotropic in-plane strain: $F_{me} = -\\frac{3}{2}\\lambda_s \\frac{Y}{1+\\nu}(\\varepsilon_{xx}\\sin^2\\theta\\cos^2\\varphi + \\varepsilon_{yy}\\sin^2\\theta\\sin^2\\varphi)$. With $\\lambda_s \\approx -4\\times 10^{-6}$, the energy is minimized when magnetization lies along the compressively strained in-plane direction. The PMN-PT substrate, poled along [011], produces opposite-sign strains along [100] (x) and [011] (y) when a voltage is applied, so the compressive direction—and hence the easy axis—switches from y to x. The SiO2 buffer is what allows the garnet to grow without epitaxy while still transmitting the strain.","core_discovery":"The central claim is that a 90-degree reorientation of the magnetic easy axis in a polycrystalline Bi-YIG film can be achieved purely by electric field via strain transfer from a poled (011) PMN-PT substrate. Poling the substrate along its thickness leaves a remanent strain that makes y the easy axis; applying voltage reverses the relative signs of the in-plane strains, so that x becomes easy above a threshold. The paper demonstrates this with in-plane MOKE hysteresis loops and domain imaging at fixed field, and extends the effect to patterned dots and racetracks where the switching field is voltage-tunable. The magnetoelectric coefficient is measured as 1.05 × $10^{-7}$ s/m.","pith_inferences":["If the strain transfer through the 2.4 nm SiO2 is as coherent as assumed, then thicker or more compliant buffers would proportionally weaken the effect; measuring the film strain directly would place quantitative bounds on the achievable anisotropy rotation.","A control experiment on a non-piezoelectric substrate, or with voltage applied but no strain path, would separate any charge-mediated contribution to the MOKE signal from the magnetoelastic one; the paper does not rule out such contributions.","The requirement of a 90-minute poling step before the effect appears suggests that the ferroelectric domain state of PMN-PT, not just its instantaneous piezoelectric response, is part of the memory; engineering that domain state could reduce the switching voltage or add nonvolatility."],"forward_implications":["Voltage-controlled easy-axis switching in Bi-YIG could enable magnetoelectric memory bits with femtojoule-scale write energy instead of the high current densities required for spin-transfer-torque writing.","The demonstrated voltage tuning of switching fields in 5–30 µm dots and racetracks suggests a path toward strain-controlled domain-wall motion in garnet-based racetrack or neuromorphic devices.","Because the SiO2 buffer decouples garnet crystallization from the piezoelectric substrate's lattice, the same growth strategy may transfer to other garnet compositions and to amorphous dielectric layers in integrated circuits.","The measured magnetoelectric coefficient places Bi-YIG on par with other oxide/ferroelectric bilayers, indicating practical coupling strengths despite the polycrystalline film and buffer layer."],"supporting_citations":[{"why":"Establishes the SiO2-buffer growth strategy for garnets on PMN-PT and the poling/relaxation strain behavior the present study relies on.","marker":"[30]"},{"why":"Supplies the piezoelectric coefficients d31 and d32 of PMN-PT that determine the opposite-sign strains along x and y.","marker":"[46]"},{"why":"One source for the negative saturation magnetostriction λ_s ≈ -4×10^-6 used in the magnetoelastic energy model.","marker":"[47]"},{"why":"The other source for magnetostriction and anisotropy values of bismuth-substituted YIG films.","marker":"[48]"},{"why":"Defines the high-Bi composition Bi2.13Y1.40Fe5Ox and its magneto-optical and magnetostrictive properties used for film growth.","marker":"[43]"},{"why":"Provides the comparison magnetoelectric coefficient for YIG/PMN-PT that situates the present result.","marker":"[50]"},{"why":"Provides the comparison magnetoelectric coefficient for YIG on PMN-PZT ceramic substrates.","marker":"[51]"},{"why":"Provides the comparison magnetoelectric coefficient for La0.67Sr0.33MnO3/PMN-PT oxide bilayers.","marker":"[52]"}],"fun_headline_variants":["Voltage twists magnetic easy axis by 90 degrees","Voltage flips garnet's magnetization easy axis","Electric field rotates magnetic axis in garnet film","90-degree axis switch via voltage in Bi-YIG","Voltage tunes easy axis by 90 degrees in garnet"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The observed easy-axis reorientation is attributed entirely to strain transferred through the thin SiO2 buffer from the PMN-PT substrate, so if the film strain does not follow the substrate's piezoelectric response—or if voltage also changes the magnetic state through charge accumulation, ionic motion, or heating—the central claim would be compromised.","fun_headline_variants_meta":{"raw":{"variants":["Voltage twists magnetic easy axis by 90 degrees","Voltage flips garnet's magnetization easy axis","Electric field rotates magnetic axis in garnet film","90-degree axis switch via voltage in Bi-YIG","Voltage tunes easy axis by 90 degrees in garnet"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000887,"raw_usage":{"total_tokens":3837,"prompt_tokens":960,"completion_tokens":2877,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":576,"completion_tokens_details":{"reasoning_tokens":2800}},"tokens_in":576,"tokens_out":2877,"duration_ms":19321,"temperature":1.0,"reasoning_tokens":2800,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:38:02.902782+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the in-plane strain of the Bi-YIG film directly (for example, by synchrotron X-ray diffraction or by comparing films on PMN-PT with identical films on a non-piezoelectric substrate) while sweeping the voltage, and check whether the hysteresis-loop changes track the measured strain. If the loop changes persist without corresponding film strain, or if the strain flattens while the magnetic changes continue, the magnetoelastic mechanism is falsified; a simpler test is applying the same voltage to Bi-YIG on a fused silica substrate and seeing whether the MOKE loops change at all.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the SiO2-buffer growth strategy for garnets on PMN-PT and the poling/relaxation strain behavior the present study relies on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"One source for the negative saturation magnetostriction λ_s ≈ -4×10^-6 used in the magnetoelastic energy model."},{"cited_title":"Hansen, K","cited_arxiv_id":null,"evidence_quote":"The other source for magnetostriction and anisotropy values of bismuth-substituted YIG films."},{"cited_title":"Hayashi, K","cited_arxiv_id":null,"evidence_quote":"Defines the high-Bi composition Bi2.13Y1.40Fe5Ox and its magneto-optical and magnetostrictive properties used for film growth."},{"cited_title":"Srinivasan, M","cited_arxiv_id":null,"evidence_quote":"Provides the comparison magnetoelectric coefficient for YIG/PMN-PT that situates the present result."},{"cited_title":"Liuyang, P","cited_arxiv_id":null,"evidence_quote":"Provides the comparison magnetoelectric coefficient for YIG on PMN-PZT ceramic substrates."},{"cited_title":"Pesquera, E","cited_arxiv_id":null,"evidence_quote":"Provides the comparison magnetoelectric coefficient for La0.67Sr0.33MnO3/PMN-PT oxide bilayers."}],"review_version":1}