{"id":"23bd03e2-0d65-41e1-9eda-ebb6c9f7018d","arxiv_id":"2412.18310","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"SAW actuation at 375 MHz followed by AFAM mapping shows suspended multilayer graphene supports flexural waves with wavelengths near 2 micrometers, about five times shorter than the incoming surface wave.","lead":"Researchers show that surface acoustic waves can vibrate suspended graphene membranes at 375 MHz, and that atomic force acoustic microscopy can map the resulting wave pattern on the membrane. The suspended graphene shrinks the acoustic wavelength from 10 to about 2 micrometers, a local wave modulation that could be used for on-chip manipulation of small particles or acoustic fields.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Dispersion fit uses zero-tension clamped-plate wavenumbers for low-frequency modes, inconsistent with the pre-tensioned plate model, biasing fitted bending rigidity and tension.","rationale":"The reader's CONDITIONAL verdict is appropriate. My concern sharpens one specific technical flaw in the quantitative analysis: the low-frequency mode wavenumbers are taken from the zero-tension clamped-plate roots, which are not the correct wavenumbers for the pre-tensioned plate model being fitted. For the reported devices, tension dominates at low frequencies, so the discrepancy is significant. This biases the fitted effective bending rigidity and tension, and the R^2 > 0.99 agreement does not validate the model because the low-frequency points are placed using inconsistent wavenumbers. However, the central experimental demonstration — SAW actuation of suspended graphene at 375 MHz and AFAM mapping of the resulting wavelength reduction from ~10 um to ~2 um — is directly observed and does not depend on the dispersion fit. The phase-velocity increase from ~160 m/s to ~700 m/s also survives qualitatively, though the low-frequency value would shift if corrected wavenumbers were used. Thus the core claim is not overturned, but the quantitative consistency claim is weakened. The reader's weakest_assumption correctly identified the clamped-plate wavenumber assumption as part of the problem; my concern goes further by noting that the fitted model itself makes those wavenumbers invalid. A concrete numerical test of the characteristic equation would settle whether the fitted parameters are self-consistent. No change in verdict is needed; the paper should be revised or supplemented to address this internal inconsistency.","tokens_in":9952,"tokens_out":11470,"duration_ms":107660,"concrete_test":"For each device, numerically solve the clamped circular pre-tensioned plate characteristic equation derived from Eq. (1) with clamped boundary conditions, using the reported fitted D/rho_h and T/rho_h. Then compare the predicted first three eigenfrequencies to the measured f01, f11, f21 and the predicted high-frequency fringe spacing to the AFAM data. If the predicted frequencies deviate by more than the experimental resolution, the assumed wavenumbers are inconsistent and the fitted parameters are biased. Re-fit the data self-consistently, allowing the mode wavenumbers to depend on T R^2 / D, and check whether the revised dispersion curve still yields R^2 > 0.99.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative consistency claim rests on a two-parameter fit of Eq. (1) to three laser-interferometry resonance frequencies and one AFAM wavenumber at 375 MHz. The low-frequency wavenumbers are assigned using the clamped circular plate (zero-tension) roots kR = 3.1962, 4.6109, 5.9057. However, Eq. (1) describes a pre-tensioned plate; for a clamped circular plate with tension, the mode wavenumbers are not these roots but solutions of a transcendental equation depending on the dimensionless ratio alpha = T R^2 / D. For the reported resonances (f01 ~ 15-18 MHz, R = 5 um, h = 9.4-37 nm), the implied tension dominates at low frequencies (phase velocities ~150-200 m/s give T/rho_h >> D/(rho_h R^2)), so alpha is large and the true wavenumbers lie close to the membrane Bessel zeros (e.g., k01 R ~ 2.405 rather than 3.196). Using zero-tension plate roots overestimates low-frequency wavenumbers by 25-30%, biasing fitted T/rho_h and D/rho_h and invalidating the claimed R^2 > 0.99 as evidence for the plate model. This is an internal inconsistency: the data are interpreted with wavenumbers that are not solutions of the model being fitted. The directly observed wavelength reduction (10 um to ~2 um) is unaffected, but the fitted parameters and the statement that the data are 'consistent with plate theory' are not reliably established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a combined experimental platform in which 375 MHz surface acoustic waves (SAWs) on LiNbO3 actuate suspended multilayer graphene membranes, and atomic force acoustic microscopy (AFAM) maps the resulting standing wavefields. The central observation is that the acoustic wavelength on the suspended graphene is reduced from the 10 µm SAW wavelength to 1.83–2.28 µm, with phase velocities increasing from ~160 m/s at low frequencies to ~700 m/s at 375 MHz. The authors interpret the data with a pre-tensioned plate dispersion model, fit effective bending rigidity and tension parameters per device, and compare the measured wavefields with COMSOL simulations. They also show wavefield modification for cavities of different size and shape.","tokens_in":10235,"tokens_out":3084,"duration_ms":28770,"significance":"If the quantitative interpretation is sound, this work would demonstrate a useful on-chip, material-agnostic route to high-frequency actuation of suspended 2D membranes, with accessible nanoscale readout via AFAM. The direct mapping of the wavelength reduction on suspended graphene is an advance over optical techniques with micron-scale resolution, and the demonstration of geometry-controlled wavefield patterning is a strength. However, the load-bearing dispersion analysis has internal inconsistencies that currently undermine the fitted parameters and the claimed consistency with pre-tensioned plate theory. The experimental phenomenology, especially the wavelength reduction and the circular fringe patterns, remains credible and valuable; the quantitative modeling needs substantial revision.","major_comments":[{"comment":"The wavenumbers k01 = 3.1962/R, k11 = 4.6109/R, and k21 = 5.9057/R are the roots of the clamped circular plate with zero tension. The model in Eq. (1) includes a tension term T k^2, so for a pre-tensioned plate these constants are not the mode wavenumbers. In the reported tension-dominated regime (phase velocities of ~160 m/s at low frequencies imply T/rho_h >> D/(rho_h R^2)), the true wavenumbers approach the membrane Bessel zeros (j01 = 2.405, j11 = 3.832, j21 = 5.136), i.e., they are 25–30% smaller than the zero-tension plate roots used in the fit. This internal inconsistency biases the fitted (D/rho_h)_eff and (T/rho_h)_eff and invalidates the reported R^2 > 0.99 as evidence for the pre-tensioned plate model.","section":"Pre-tensioned plate model"},{"comment":"The dispersion curve is fitted with two free parameters, (D/rho_h)_eff and (T/rho_h)_eff, to only three or four data points per device (three low-frequency resonances and one AFAM point at 375 MHz). With this number of points, a two-parameter fit can achieve R^2 > 0.99 almost by construction; this does not provide independent confirmation of the plate model. The paper should report confidence intervals on the fitted parameters and, ideally, cross-validate using more than one high-frequency wavenumber per device.","section":"Figure 3 and Table 1"},{"comment":"The COMSOL model for device D4 uses the fitted (D/rho_h)_eff and (T/rho_h)_eff as inputs, so the good agreement between the simulated and experimental wavefields in Figure 4 is partly by construction. This agreement is therefore not an independent validation of the pre-tensioned plate model. The qualitative pattern geometry and the scaling with cavity size/shape remain informative, but the quantitative match should be presented as a consistency check, not as predictive confirmation.","section":"Simulation of SAW-induced wavefields on suspended graphene membranes"},{"comment":"The quantitative analysis assumes that the spacing between adjacent bright fringes in the AFAM maps equals half the local flexural wavelength on the suspended graphene. Tip-sample nonlinearities, topography coupling, or a misinterpretation of the demodulated signal could systematically affect this spacing. The paper should provide an explicit calibration or a control experiment (e.g., comparing the fringe spacing on supported graphene to the known SAW wavelength) to justify that the extracted wavenumber k_exp is quantitatively accurate to the level needed for the dispersion fit.","section":"AFAM wavelength extraction"}],"minor_comments":[{"comment":"The caption states 'D1, D3 and D4' but the text refers to devices with 9.4 nm, 14 nm, and 34 nm thickness, which are D1, D2, and D4. The caption should be corrected.","section":"Figure 3 caption"},{"comment":"The statement 'for thin flakes (<20 layers), tension dominates over bending rigidity' is not directly tied to the measured thickness range of 9.4–37 nm (about 28–110 layers); the transition also depends on frequency and wavenumber through Eq. (1). Please rephrase to avoid an apparent inconsistency.","section":"Pre-tensioned plate model text"},{"comment":"The abstract says '~2 µm' and later the text gives the range 1.83–2.28 µm; please keep the values consistent throughout, including the abstract.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The experimental core of the paper—SAW actuation of suspended graphene and AFAM mapping—is interesting and likely to be of value to the nanomechanics community. The main obstacle is the quantitative dispersion analysis, which uses zero-tension plate wavenumbers in a tensioned-plate fit. This is a fixable issue in principle, but it requires redoing the fit with the correct mode wavenumbers (or a proper tensioned-plate eigenvalue solver) and re-evaluating the COMSOL comparison. The paper would also be stronger if the authors were more explicit about the limited number of independent data points and provided uncertainty estimates for the fitted parameters."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper shows something genuinely useful: surface acoustic waves at 375 MHz can drive suspended multilayer graphene membranes, and AFAM can map the resulting standing wavefield with sub-micron resolution. The key observation—a reduction in wavelength from 10 µm on the substrate to roughly 2 µm on the suspended part, with phase velocity rising from ~160 to ~700 m/s—is directly visible in the AFAM images and is probably correct. The combination of SAW actuation and AFAM mapping on suspended 2D materials is new and worth having. The extra demonstrations with square and smaller circular cavities add value.\n\nThe soft spot is the quantitative dispersion analysis. The authors fit Eq. (1), a pre-tensioned plate model, to the first three laser-interferometry resonances plus the 375 MHz AFAM point. For the low-frequency modes they assign wavenumbers using the clamped circular plate roots kR = 3.1962, 4.6109, 5.9057. But those are the zero-tension plate roots. For a tensioned plate, the mode wavenumbers are not those roots; if tension dominates, as it does for these devices at low frequency, the wavenumbers move toward the membrane Bessel zeros (kR ≈ 2.405, etc.). Using the zero-tension roots biases the fitted bending rigidity and tension, so the R² > 0.99 agreement with Eq. (1) is not strong evidence for the model. The stress-test note has this right.\n\nThere's also a related issue: the fit uses only three or four points per device, and the COMSOL simulation uses the same fitted parameters, so the agreement with COMSOL is in-sample rather than predictive. The claim of predictive power is thus overplayed.\n\nNone of this kills the central demonstration. The wavelength reduction and the change in wave velocity are observable without the model. But the quantitative extraction of D and T, and the statement that the data are consistent with plate theory, need correction. I'd suggest the authors redo the low-frequency mode assignment for a tensioned plate, report parameters with uncertainties, and ideally validate on a held-out device or with an independent prediction.\n\nThis paper deserves peer review. A good referee will push on the mode assignment and ask for out-of-sample validation, but the experimental platform and the main observation are solid. I would bring it to a reading group working on 2D nanomechanics, though I would not cite the fitted parameters until revised.","headline":"Useful new method for driving and imaging high-frequency wavefields on suspended graphene, but the dispersion fit has an internal inconsistency that needs fixing.","tokens_in":10833,"tokens_out":3313,"would_cite":false,"duration_ms":29887,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["74K20","74H45"],"pacs":["43.35.+d","68.37.Ps","81.05.ue"],"model":"deepseek-v4-flash","headline":"Suspended graphene membranes can be actuated at 375 MHz by surface acoustic waves, and atomic-force acoustic microscopy maps the resulting standing wavefields, showing a wavelength reduction from 10 µm to about 1.8–2.3 µm.","keywords":["surface acoustic waves","suspended graphene","atomic force acoustic microscopy","flexural wave dispersion","pre-tensioned plate model","standing wavefield mapping","two-dimensional material actuation"],"falsifier":"Repeat the AFAM measurement on the same drum at several tip setpoint forces and with different cantilevers, and compare the extracted wavelengths to an independent sub-micron displacement probe (for example a stroboscopic electron-microscopy or short-wavelength interferometric measurement) at the same 375 MHz drive. If the fringe spacing changes with contact conditions or disagrees with the independent wavelength, the half-wavelength interpretation and the fitted $(D/\\rho h)$ and $(T/\\rho h)$ values do not hold.","tokens_in":9735,"feed_emoji":"🌊","tokens_out":10699,"duration_ms":85566,"temperature":0.7,"pith_summary":"This paper demonstrates that surface acoustic waves (SAWs) can drive suspended multilayer graphene membranes at 375 MHz, a frequency beyond the reach of most existing mechanical actuation schemes, and that the resulting standing vibration field can be imaged with atomic-force acoustic microscopy (AFAM). On the suspended graphene, the acoustic wavelength shrinks from 10 µm on the substrate to about 1.83–2.28 µm, a reduction of more than 76%, and the measured phase velocity rises from roughly 160 m/s at the lowest drum resonances to about 700 m/s at 375 MHz. The authors interpret these observations with the dispersion relation of a pre-tensioned plate: low-frequency modes are tension-dominated, while the high-frequency SAW response is bending-dominated. The combined actuation-plus-imaging scheme matters because it offers a universal, on-chip route to high-frequency actuation of two-dimensional membranes and a way to locally compress acoustic wavelengths for nanoscale manipulation.","feed_headline":"375 MHz sound waves shrink to ~2 µm on suspended graphene","feed_subtitle":"Atomic-force acoustic maps show flexural waves and a velocity jump from 160 to 700 m/s.","key_machinery":"The load-bearing object is the pre-tensioned plate dispersion relation $\\omega_k = \\sqrt{(D/\\rho h)\\,k^4 + (T/\\rho h)\\,k^2}$, with $D$ the bending rigidity, $T$ the tension, $\\rho h$ the areal mass density, and $k$ the wavenumber. It connects the low-frequency clamped-circular modes (wavenumbers $k_{01}=3.1962/R$, $k_{11}=4.6109/R$, $k_{21}=5.9057/R$) measured by laser interferometry to the high-frequency wavenumber extracted from AFAM fringe spacing, so that effective $(D/\\rho h)$ and $(T/\\rho h)$ can be fitted. The enabling readout is AFAM: the cantilever's nonlinear tip-sample interaction converts the 375 MHz, amplitude-modulated vibration into a 10 kHz demodulated signal, producing a spatial map of the standing wavefield with sub-micron resolution.","core_discovery":"The central claim is that a Rayleigh surface wave arriving at a suspended graphene drum is converted into a flexural wave with a much shorter wavelength, and that this conversion can be quantified. Using 375 MHz SAWs generated by interdigital transducers on lithium niobate, the authors excite five multilayer graphene drums of 9.4–37 nm thickness and map the standing wavefields with AFAM, whose nonlinear tip-sample interaction demodulates the amplitude-modulated drive into time-averaged spatial fringes. Fringe spacings give wavelengths of 1.83–2.28 µm on suspended graphene versus 10 µm on supported graphene. Combining these points with the first three clamped-circular-drum resonance frequencies measured by laser interferometry, the authors fit the dispersion relation $\\omega_k = \\sqrt{(D/\\rho h)\\,k^4 + (T/\\rho h)\\,k^2}$ and find phase velocities from roughly 160 m/s to 700 m/s, with fits at $R^2>0.99$. Finite-element simulations using the fitted stiffness parameters reproduce the measured wavefield geometry and dimensions.","pith_inferences":["If the half-wavelength fringe interpretation carries over to other materials, the same platform should map frequency-dependent flexural fields on suspended hexagonal boron nitride or MoS2 membranes, and the compression factor should grow as thickness decreases because $D$ scales as $h^3$.","Because AFAM contrast depends on nonlinear tip-sample contact, a controlled setpoint-force sweep would reveal how much of the fitted stiffness is intrinsic to the membrane rather than induced by the measurement; the paper does not perform that control.","Detuning or removing the reflector IDT would turn the standing-wave measurement into a propagating-wave measurement, separating intrinsic flexural dispersion from resonator boundary effects.","The sub-2 µm acoustic standing patterns are natural templates for trapping or moving nanoparticles on a chip, and could be extended to reconfigurable sorting, though the paper stops short of demonstrating transport."],"forward_implications":["SAW actuation should work on any suspended two-dimensional membrane, conducting or not, because the coupling is purely mechanical, and interdigital transducer design can push the drive into the gigahertz range.","Suspended membranes act as local acoustic lenses: a fixed 10 µm SAW wavelength is compressed to about 2 µm over the drum, so cavity size, shape, and flake thickness can pattern nodal lines and localized wavefields.","The dispersion fit means the effective bending rigidity and tension of a multilayer flake can be extracted from two measurement bands—megahertz drum resonances and the 375 MHz AFAM point—without needing optical resolution to see the short wavelength.","Because the high-frequency branch is bending-dominated ($\\omega \\propto k^2$), thicker and stiffer membranes will compress the SAW wavelength more, making the compression tunable through flake thickness.","Probe-based mapping is required for these short wavelengths; non-contact optical Doppler vibrometry with micrometer-scale resolution cannot resolve the 1.83–2.28 µm features."],"supporting_citations":[{"why":"Reviews actuation and readout methods for 2D material membranes and frames the gap that SAW actuation fills.","marker":"[5]"},{"why":"Shows SAW devices reaching gigahertz frequencies with on-chip control, justifying the frequency and transducer design choices.","marker":"[12]"},{"why":"Surveys SAW-induced phenomena in 2D materials and motivates SAWs as a promising but little-explored actuation route for suspended membranes.","marker":"[13]"},{"why":"The closest prior demonstration, a graphene resonator detecting shear-horizontal surface waves, which this work extends to imaging and dispersion analysis.","marker":"[17]"},{"why":"Uses AFM-based ultrasonic methods to map dynamic properties of multilayer graphene, supporting the AFAM readout approach.","marker":"[25]"},{"why":"Provides the laser interferometry setup used to measure the low-frequency drum resonances that anchor the dispersion fit.","marker":"[26]"},{"why":"Supplies the deterministic viscoelastic stamping method used to transfer graphene flakes over the cavities.","marker":"[28]"}],"fun_headline_variants":["SAW actuation maps wavefields on suspended graphene","Graphene membranes bend SAWs: wavelength shrinks to 2 µm","375 MHz waves on graphene: velocity jumps to 700 m/s","Flexural waves on graphene mapped via SAW and AFAM"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative interpretation rests on assuming that every bright AFAM fringe on the suspended graphene marks half a local flexural wavelength, and that the flake behaves as a smooth, uniformly stretched, continuously clamped circular plate with the standard mode wavenumbers; if tip-sample contact averaging, topography, or a wrong mode assignment changes the fringe-to-wavelength relation, the extracted wavelengths and the fitted bending rigidity and tension are systematically wrong.","fun_headline_variants_meta":{"raw":{"variants":["SAW actuation maps wavefields on suspended graphene","Graphene membranes bend SAWs: wavelength shrinks to 2 µm","375 MHz waves on graphene: velocity jumps to 700 m/s","Flexural waves on graphene mapped via SAW and AFAM"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000192,"raw_usage":{"total_tokens":1400,"prompt_tokens":1049,"completion_tokens":351,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":665,"completion_tokens_details":{"reasoning_tokens":278}},"tokens_in":665,"tokens_out":351,"duration_ms":3832,"temperature":1.0,"reasoning_tokens":278,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T04:48:23.773269+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the AFAM measurement on the same drum at several tip setpoint forces and with different cantilevers, and compare the extracted wavelengths to an independent sub-micron displacement probe (for example a stroboscopic electron-microscopy or short-wavelength interferometric measurement) at the same 375 MHz drive. If the fringe spacing changes with contact conditions or disagrees with the independent wavelength, the half-wavelength interpretation and the fitted $(D/\\rho h)$ and $(T/\\rho h)$ values do not hold.","supporting_citations":[{"cited_title":"G., Dolleman, R","cited_arxiv_id":null,"evidence_quote":"Reviews actuation and readout methods for 2D material membranes and frames the gap that SAW actuation fills."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows SAW devices reaching gigahertz frequencies with on-chip control, justifying the frequency and transducer design choices."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Surveys SAW-induced phenomena in 2D materials and motivates SAWs as a promising but little-explored actuation route for suspended membranes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The closest prior demonstration, a graphene resonator detecting shear-horizontal surface waves, which this work extends to imaging and dispersion analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Uses AFM-based ultrasonic methods to map dynamic properties of multilayer graphene, supporting the AFAM readout approach."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the laser interferometry setup used to measure the low-frequency drum resonances that anchor the dispersion fit."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the deterministic viscoelastic stamping method used to transfer graphene flakes over the cavities."}],"review_version":1}