{"id":"838a297d-d74a-4456-a4d1-359cb010a3d9","arxiv_id":"2607.28888","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"InGaAs-on-LNOI bonded heterostructures demonstrate bias-tunable, mode-dependent shear-horizontal acoustoelectric non-reciprocity up to 174 dB/mm (unpassivated upper bound) and 32 dB/mm stable in passivated devices.","lead":"This paper builds a shear-horizontal surface acoustic wave amplifier using a thin InGaAs semiconductor bonded onto lithium niobate on insulator, and reports stable non-reciprocal transmission of 32 dB/mm in passivated devices at 1.11 GHz. It also finds that the bare InGaAs surface oxidizes within weeks and shows that passivation preserves the response, which matters for compact non-reciprocal RF components.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing reverse-bias control: the paper never shows that reversing VDC reverses the sign of S21−S12, so the attribution of the measured non-reciprocity to acoustoelectric gain is not yet secured.","rationale":"The reader's weakest assumption targets the analytical model and FEM-simulated K2 values, which is a valid concern about quantitative interpretation and the conclusion that excess doping is the dominant limitation. However, that concern is not the single most load-bearing issue for the paper's central claim. Even if the model and K2 values were perfect, the measured S21−S12 contrast would still need to be shown to be a genuine acoustoelectric effect rather than a bias-dependent artifact. The most direct and decisive control is reversing the bias polarity: acoustoelectric gain is inherently drift-direction dependent, so the non-reciprocity must reverse sign. The paper does not report such a measurement, nor does it report a zero-bias reciprocity baseline. This does not warrant rejection—the device may well be performing as claimed—but it does mean the central attribution is conditioned on a simple control that is absent. Since the reader already assigned CONDITIONAL, my assessment does not change the verdict; it sharpens the condition that should be met before accepting the claim. I marked agreement as partial because I share the reader's concern about attribution of S21−S12 to AE, but I locate the weakest point in the missing sign-reversal control rather than in the simulated K2 values.","tokens_in":925,"tokens_out":1288,"duration_ms":72303,"concrete_test":"On the passivated 3-µm-wavelength device, sweep VDC from −45 V to +45 V in 5 V steps and measure time-gated S21 and S12 at the mode-I peak (1.11 GHz) and mode-III peak (2.79 GHz). Also record S21 and S12 at VDC=0 as a null control. If the AE attribution is correct, S21(+V)−S12(+V) should be approximately equal and opposite to S21(−V)−S12(−V) at each bias magnitude, within measurement repeatability (suggested ±3 dB/mm), and the zero-bias contrast should be near zero. If the sign does not track the drift direction, or if a comparable non-reciprocity persists at zero bias, the central attribution to acoustoelectric gain must be qualified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central empirical claim is that the measured S21−S12 contrast is acoustoelectric (AE) gain. Yet Section IV-A and Fig. 3 report only forward-bias sweeps; they do not show that reversing the drift field inverts the non-reciprocity. For a drift-direction mechanism, reversing VDC must swap the roles of S21 and S12: the same carriers that amplify forward waves must attenuate reverse waves, so S21(+V)−S12(+V) should be approximately equal and opposite to S21(−V)−S12(−V) at each mode peak. Without this control, the observed bias-dependent asymmetry could partly arise from current-induced heating, bias-dependent contact or transducer impedance, or passive fabrication asymmetry amplified by gating, none of which constitutes acoustoelectric amplification. The reader's concern about simulated K2 values is legitimate, but it affects the magnitude and mode-ranking interpretation, not the existence of the AE interaction; the missing sign-reversal/null-bias test is more load-bearing because it questions the basic attribution. A zero-bias measurement is also needed to establish that the passive device is reciprocal before bias is applied.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports fabrication and characterization of a shear-horizontal surface-acoustic-wave (SH-SAW) acoustoelectric amplifier built from an epitaxial InGaAs film bonded to X-cut LNOI via an Al2O3-mediated process. The central experimental claim is bias-dependent non-reciprocal transmission: for a 3-μm-wavelength device, the paper reports S21−S12 contrasts of about 160 dB/mm, 19.1 dB/mm, and 174 dB/mm for three acoustic modes in fresh unpassivated devices, and a stable 32 dB/mm fundamental-mode contrast in a passivated device at 30 V bias. The authors attribute the non-reciprocity to acoustoelectric gain, support the mode-dependence with FEM-simulated K2 values inserted into an analytical gain model, use Hall measurements to diagnose an unintentional background doping of ~1e17 cm−3, and identify ambient surface oxidation as a degradation mechanism mitigated by InP or ALD Al2O3 passivation.","tokens_in":11765,"tokens_out":4315,"duration_ms":51225,"significance":"If the central attribution is correct, this is a valuable engineering advance: it demonstrates a wafer-bonded InGaAs/LNOI platform for bias-tunable, non-reciprocal acoustic devices with multi-GHz bandwidth, and it goes beyond a bare demonstration by identifying the dominant material limit (elevated carrier density) with an independent Hall measurement. The mode ordering, the passivation-stability study, and the explicit use of Hall data without free fitting to the AE model are strengths. However, the missing reverse-bias control means the load-bearing claim that the observed S21−S12 asymmetry is acoustoelectric gain is not yet secured; this can be addressed experimentally and is therefore a basis for major revision rather than rejection.","major_comments":[{"comment":"The paper never shows that reversing the DC drift bias reverses the sign of S21−S12. For an acoustoelectric mechanism, reversing VDC must swap the roles of forward and reverse waves, so S21(+V)−S12(+V) should be approximately opposite to S21(−V)−S12(−V) at each mode frequency. The current data show only forward-bias sweeps and a monotonic increase in non-reciprocity. Without this control, the asymmetry could in part be caused by bias-dependent transducer/contact impedance, current-induced heating, or gating of a passive fabrication asymmetry. Please provide zero-bias reciprocal baseline and negative-bias measurements for at least the fundamental mode.","section":"§IV-A, Fig. 3"},{"comment":"There is an inconsistency between the abstract, which describes the 174 dB/mm value as an 'upper bound,' and the body text, which states that peak contrasts of approximately 160, 19.1, and 174 dB/mm 'were measured' for modes I, II, and III. The abstract's wording suggests these values are not direct measurements or are corrected upper limits, while Table II labels the same numbers as 'Mea. NR.' Please clarify what was actually measured, what correction or upper-bound interpretation is being applied, and provide uncertainty estimates. This matters because the quantitative headline claims (32 dB/mm and 174 dB/mm) are central to the paper's significance.","section":"Table II, §IV-A, Abstract"},{"comment":"The calculated non-reciprocity values used to validate the mode-dependence rely on FEM-simulated K2 values and on the analytical gain expression of Ref. [31], which is by a co-author. The measured values are roughly a factor of 3 below the calculated values for all modes, and the model parameters are not independently measured. This does not by itself invalidate the ranking of modes, but it weakens the claim that the K2–frequency trend is quantitatively validated. Please provide an independent estimate of K2 (e.g., from IDT admittance or resonator measurements) or clearly state that the comparison is only a self-consistent trend check, not a quantitative validation. ","section":"§II Eq. (1), Table II, Fig. 4"},{"comment":"The Hall-based consistency check is described qualitatively as reproducing the measured passivated non-reciprocity 'within an order of magnitude' (32 dB/mm), but the manuscript does not show the calculated value obtained by inserting the Hall n and μ into Eq. (1), nor does it compare this against the unpassivated-mode data. Since the claim that excess doping is the dominant limitation rests on this calculation, please present the explicit numbers, the assumed dielectric thicknesses, and the sensitivity of the result to the uncertainty in Hall n and μ. ","section":"§IV-C"}],"minor_comments":[{"comment":"The expression for ω_a appears dimensionally incomplete as written: ε_g v_a/ε_ph has units of velocity, not frequency. The dielectric gap thickness h is introduced in the text but does not appear in the formula. Please correct the equation and define all symbols consistently.","section":"§II Eq. (2)"},{"comment":"The frequency- and time-gating procedure is mentioned but no parameters are given. Please specify the gating window, filter type, and how feedthrough and triple-transit contributions were separated, since the reported S21−S12 values are extracted after gating.","section":"Fig. 3"},{"comment":"The Hall measurement assumes a unity Hall factor. Since the extracted n is central to the doping diagnostic, please state the resulting systematic uncertainty or justify the unity assumption for this InGaAs carrier density.","section":"Table I / §IV-C"},{"comment":"The manuscript uses 'X-cut LNOI' but does not specify the SAW propagation direction relative to the crystallographic axes. Shear-horizontal coupling and K2 are strongly orientation-dependent on LN; please report the in-plane orientation of the IDTs.","section":"§I and §IV-C"},{"comment":"No uncertainty bars or statistics are given for the measured non-reciprocity values. A single-device demonstration is reasonable for a first report, but the precision of the 32/160/174 dB/mm numbers should be stated, particularly because they are quoted in the abstract.","section":"Table II"},{"comment":"The phrase 'reported as upper bounds' in the abstract is ambiguous and inconsistent with Table II, as noted in the major comments. Please rephrase so the reader knows whether the numbers are measured or corrected upper-limit estimates.","section":"Abstract and Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious and useful experimental report, and the authors are appropriately cautious in several places (e.g., calling the demonstration 'initial,' and describing the Hall comparison as a consistency check rather than a validation). The main gap is the absence of a reverse-bias control, which is directly testable and should be decisive for the acoustoelectric attribution. Adding such data, together with clarifying the 'upper bound' wording and showing the explicit Hall-model calculation, would make the central claim persuasive. I do not see the current evidence as sufficient for acceptance, but I also do not see a fundamental error that would force rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is a genuine experimental effort with two things worth taking away: the InGaAs-on-X-cut-LNOI bonded platform works for shear-horizontal acoustoelectric interaction, and the passivation/aging study is a practical contribution that should survive contact with the literature. The Hall diagnostic is a good move — inserting measured n and μ into the model and getting the right order of magnitude gives the doping explanation some teeth.\n\nWhat's new is the specific combination: Al2O3-mediated bonding of epitaxial InGaAs to LNOI, mode-resolved non-reciprocity, and the InP/Al2O3 passivation comparison. The mode-dependence follows the expected K2-frequency trend, which is credible even if the K2 values are simulated rather than measured.\n\nThe soft spots are proportionate but real. The most load-bearing is the missing sign-reversal control. For a drift-direction mechanism, reversing VDC should swap S21 and S12 — the same carriers that amplify forward waves attenuate reverse waves. The paper only sweeps forward bias, so the observed S21−S12 asymmetry could partly come from bias-dependent transducer impedance, heating, or fabrication asymmetry. A zero-bias reciprocity check and a negative-bias sweep would close that gap cheaply. That is a bigger issue than the simulated-K2 circularity, which affects magnitude and mode ranking but not the existence of the interaction.\n\nAlso, measurements are on a single unpassivated device with no error bars or device statistics; the passivated Hall data come from co-located test structures, not the same dies as the unpassivated gain numbers, so the doping explanation for the 174 dB/mm upper bound is an extrapolation. The title says 'amplifier' but the data show differential transmission, not net gain. That overclaim should be fixed.\n\nThe passivation result — bare InGaAs loses its AE response in weeks and an InP or Al2O3 cap stabilizes it — is solid and independently useful. The paper deserves a serious referee; a good one can push the authors to add the reverse-bias control, tone down the language, and report statistics. I'd bring it to the reading group if you follow III-V/ferroelectric integration, and I'd cite it for the passivation and platform data.","headline":"Real fabrication and a useful passivation result, but the 'amplifier' title overclaims: no net gain and no reverse-bias control, so the AE attribution is not yet locked down.","tokens_in":12267,"tokens_out":2097,"would_cite":true,"duration_ms":25424,"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":"This paper demonstrates a compact acoustoelectric amplifier in which a bonded InGaAs film on lithium niobate amplifies one direction of a shear-horizontal surface acoustic wave while attenuating the reverse, achieving a stable 32 dB/mm non-","keywords":["acoustoelectric effect","surface acoustic wave","shear-horizontal SAW","lithium niobate on insulator","InGaAs","non-reciprocal RF devices","wafer bonding","in-band full-duplex"],"falsifier":"Measure the electromechanical coupling of modes I–III directly on an IDT resonator on the same LNOI stack (for example, by admittance fitting or by comparison with a bare-LNOI reference) and compare with Table II; a significant mismatch would change the predicted gain ranking and the doping-limitation conclusion. Alternatively, grow a channel with carrier density near the 3×10^15 cm^-3 design target and check whether non-reciprocity rises toward the model's ~487 dB/mm bound.","tokens_in":11367,"feed_emoji":"🔊","tokens_out":5363,"duration_ms":56316,"temperature":0.7,"pith_summary":"This paper tries to establish that a shear-horizontal surface acoustic wave can be amplified—rather than merely delayed or filtered—by drifting electrons in a bonded InGaAs film on a lithium-niobate-on-insulator (LNOI) substrate. The central demonstration is non-reciprocal transmission: with a DC bias on the semiconductor channel, forward-traveling acoustic waves gain energy from the carriers while reverse waves are damped. The authors report a stable passivated-device non-reciprocity of 32 dB/mm at 1.11 GHz (30 V bias, 64 mW consumed), with unpassivated devices reaching 174 dB/mm as upper bounds across 1.1–2.8 GHz. They also show that the mode-dependent electromechanical coupling sets the achievable gain, that unintentional epitaxial doping (~1e17 cm^-3, two orders above target) is the dominant limiter, and that bare InGaAs surface oxidation destroys the response within weeks unless passivated by InP or ALD Al2O3. If correct, this is a compact, low-power path toward non-reciprocal RF components and loss-compensated acoustic delay lines for in-band full-duplex wireless.","feed_headline":"One-way acoustic gain hits 32 dB/mm on bonded InGaAs-lithium niobate","feed_subtitle":"A 64 mW biased semiconductor channel amplifies forward surface waves and damps reverse, enabling compact non-reciprocal RF delay lines.","key_machinery":"The load-bearing object is the acoustoelectric gain relation (Eq. 1), which extends the classical separated-medium amplifier to a thin semiconductor film separated from a high-K2 piezoelectric by a thin dielectric bonding layer. Gain arises when the carrier drift velocity vd = μE exceeds the acoustic phase velocity va; the numerator scales with frequency ω, conductivity, and electromechanical coupling K2, while the denominator contains a dielectric-relaxation term (1+ω/ωa)^2 that damps gain at high frequency. The same coefficient set is used to convert Hall-measured carrier density and mobility into predicted non-reciprocity, and FEM-simulated K2 values for the three modes are used to compar","core_discovery":"On its own terms, the paper claims that the acoustoelectric effect in an epitaxial InGaAs channel wafer-bonded to X-cut lithium niobate on insulator produces measurable, bias-dependent, mode-dependent non-reciprocity in SH-SAW transmission. The measured S21−S12 contrast follows the K2-plus-frequency trend of the analytical AE gain model: modes with larger simulated electromechanical coupling (fundamental SH-SAW K2≈21.8%, second-order SH-SAW K2≈13.9%) yield far larger non-reciprocity than the longitudinal-leaky mode (K2≈3.94%). Hall-effect characterization of the transferred film gives n≈1e17 cm^-3, μ≈8000 cm^2/V·s; inserting these into the gain formula reproduces the passivated device's 32 d","pith_inferences":["If epitaxial doping can be suppressed, the same platform should cross from non-reciprocal attenuation contrast into net amplification (forward S21 above 0 dB), which would enable loss-compensated delay lines; the paper stops at non-reciprocity values rather than net gain.","The identified oxidation-aging mechanism is not platform-specific: any bare III-V channel relying on surface carrier response is vulnerable, so the InP or Al2O3 passivation recipe likely transfers to other acoustoelectric hybrids.","The field-confinement role of the buried oxide suggests the dielectric stack itself can be tuned (for example, SiO2 thickness) to trade off K2, confinement, and passivation loss—an extension the paper leaves implicit.","Hall-based doping diagnosis could be turned into an in-line process monitor: measuring carrier density on co-located test structures predicts AE gain before full device measurement, accelerating epitaxial process development."],"forward_implications":["Correcting the channel doping toward the 3×10^15 cm^-3 design target should raise fundamental-mode non-reciprocity toward the model's roughly 487 dB/mm ideal bound and lower zero-bias insertion loss.","Passivated devices sustain 32 dB/mm at 1.11 GHz at 64 mW, making them viable as compact non-reciprocal elements in in-band full-duplex self-interference-cancellation front ends.","Mode engineering—choosing wavelength and mode order to sit in high-K2 regimes—can spread acoustoelectric gain across 1–3 GHz, since the second-order SH-SAW mode shows the largest normalized amplification.","Any practical passivation must be co-optimized with gain, because the protecting layer redistributes the piezoelectric field away from the channel even as it prevents oxidation-driven aging."],"fun_headline_variants":["InGaAs-on-LNOI amplifier yields 32 dB/mm non-reciprocity","Acoustic diode effect on InGaAs-LNOI with 32 dB contrast","Non-reciprocal acoustic gain: 32 dB/mm at 1.11 GHz","32 dB/mm one-way acoustic amplification on InGaAs-LNOI","Bonded InGaAs-LNOI amplifies surface waves one-way"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the FEM-simulated electromechanical coupling coefficients and the analytical gain model accurately describe this bonded geometry; if either is wrong, the mode ranking and the conclusion that excess doping is the dominant limitation would not follow.","fun_headline_variants_meta":{"raw":{"variants":["InGaAs-on-LNOI amplifier yields 32 dB/mm non-reciprocity","Acoustic diode effect on InGaAs-LNOI with 32 dB contrast","Non-reciprocal acoustic gain: 32 dB/mm at 1.11 GHz","32 dB/mm one-way acoustic amplification on InGaAs-LNOI","Bonded InGaAs-LNOI amplifies surface waves one-way"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000976,"raw_usage":{"total_tokens":4042,"prompt_tokens":860,"completion_tokens":3182,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":604,"completion_tokens_details":{"reasoning_tokens":3076}},"tokens_in":604,"tokens_out":3182,"duration_ms":22435,"temperature":1.0,"reasoning_tokens":3076,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T01:20:05.648564+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the electromechanical coupling of modes I–III directly on an IDT resonator on the same LNOI stack (for example, by admittance fitting or by comparison with a bare-LNOI reference) and compare with Table II; a significant mismatch would change the predicted gain ranking and the doping-limitation conclusion. Alternatively, grow a channel with carrier density near the 3×10^15 cm^-3 design target and check whether non-reciprocity rises toward the model's ~487 dB/mm bound.","supporting_citations":[],"review_version":1}