{"id":"e2dd8ba2-887d-4e18-b432-6cbc3f05ae43","arxiv_id":"2509.26582","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A piezo-MEMS cantilever coupled to a VO2 memristor oscillator directly converts nanometer-scale mechanical vibrations into rate-coded spike trains in the 100–1000 Hz range, shaped as biphasic pulses for cochlear implant stimulation.","lead":"A research team combined a piezoelectric MEMS cantilever with a VO2 memristor oscillator to turn mechanical vibrations into neural-like spike trains. The proof-of-concept could make future fully implantable cochlear implants smaller and more energy-efficient by replacing conventional signal processing with a neuromorphic front end.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'biologically realistic ~10 nm' and SPL claims rest solely on nominal 6 nm/V actuator calibration; actual cantilever displacement is never measured during rate-encoding experiments.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing concern: unmeasured actuator calibration. I reviewed the manuscript for alternative weak points—lack of error bars in Fig. 7d, biphasic-shaping demonstrated only with uncontrolled audio stimuli, and no electrode-tissue interface testing—but none is as directly connected to the paper's central quantitative claims as the calibration of the mechanical stimulus. The rate-encoding mechanism itself is plausible and the oscillator physics is standard, but the 'biologically realistic ~10 nm' and '82–92 dB SPL' statements rest entirely on the nominal 6 nm/V sensitivity. The vibrometer was available and used for cantilever characterization, but not to verify the actual motion in the rate-encoding setup. This is a concrete, correctable experimental gap rather than a fundamental flaw. The reader's CONDITIONAL verdict already reflects this concern appropriately, and our independent analysis does not change that verdict. Hence, I recommend UNCHANGED.","tokens_in":12150,"tokens_out":3507,"duration_ms":32250,"concrete_test":"Repeat the rate-encoding experiment at 637 Hz with the laser Doppler vibrometer aimed at the cantilever surface (or an adjacent point on the chip) while driving the piezo actuator with the same voltage amplitudes used for the nominal 26–60 nm range. Measure the actual mechanical displacement under the exact sample-holder configuration of Fig. 2b. Compare the measured displacement to the nominal 6 nm/V × V_drive. If the deviation exceeds 20% or is nonlinear, recalibrate Fig. 7d's x-axis and re-evaluate whether the claimed displacement range and 82–92 dB SPL mapping remain valid. Also collect at least three repeated measurements per amplitude to establish error bars and check for drift or hysteresis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II A states that nominal mechanical excitation amplitudes are calculated from the piezoelectric actuator drive voltage using its 6 nm/V sensitivity, and Fig. 7d's caption explicitly notes that displacement values S0 and S are calculated, not measured. The laser Doppler vibrometer used for cantilever characterization (Fig. 2a, 6b) was not used in the rate-encoding experiments; the actual displacement delivered to the cantilever through the custom sample holder, PCB, and flexible ribbon cables at 637 Hz is unverified. This matters because the paper's central quantitative claims—operation at 'biologically realistic excitation amplitudes in the ~10 nm range' and the conversion of 26–60 nm amplitudes to 82–92 dB SPL—depend directly on that calibration. If the mechanical coupling attenuates or distorts the actuator motion, the x-axis of Fig. 7d and the inferred SPL range shift systematically, weakening the biological-relevance claim. The rate-encoding trend versus drive voltage would remain, but the mapping to physiological stimuli would be unproven. This is the most load-bearing concern because it undermines the headline '~10 nm' and '82–92 dB' statements rather than merely the engineering details.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a single-channel auditory sensing concept for fully implantable cochlear implants. A piezoelectric ScAlN MEMS cantilever is mechanically excited by a piezo actuator, and its rectified output drives a VO2 nanogap memristor relaxation oscillator. The authors show that changing the nominal drive amplitude changes the oscillator spike rate, claim this rate encoding operates at biologically realistic nanometer-scale displacements and produces auditory-nerve-like rates, and demonstrate that adding a parallel inductor converts the unipolar output into a biphasic waveform suitable for cochlear stimulation. The central claim is an FFT-free, hardware-level conversion from mechanical vibration to rate-coded spikes.","tokens_in":12451,"tokens_out":4361,"duration_ms":36478,"significance":"If the rate-encoding result is robust and the displacement calibration is verified, this is a valuable experimental proof of concept for low-power neuromorphic auditory front ends. The manuscript is careful in describing the VO2 device fabrication and oscillator operation, and the biphasic shaping via a parallel LR circuit is a useful practical contribution. However, the headline quantitative claims—operation at ~10 nm amplitudes, 82–92 dB SPL equivalence, and 100 Hz–1 kHz rate range—are supported by a single device, a nominal actuator calibration, and a limited range of stimulus amplitudes. The concept is interesting, but the evidence as currently presented is not sufficient for the strength of the conclusions.","major_comments":[{"comment":"The mechanical displacement values S0 and S are not measured; they are calculated from the piezo actuator drive voltage using its nominal 6 nm/V sensitivity, and the Fig. 7d caption states this. Because the cantilever is mounted in a custom holder with PCB and ribbon cables, the actual displacement delivered at 637 Hz could differ systematically. The x-axis of Fig. 7d, the '~10 nm' biological-relevance statement, and the inferred 82–92 dB SPL range all depend on this calibration. I request either a direct measurement of the cantilever displacement under the same experimental conditions (e.g., with the vibrometer already used for Fig. 6b) or a clearly stated uncertainty/transfer-function calibration, and correspondingly hedged claims.","section":"II A and Fig. 7d"},{"comment":"The rate-encoding result is presented as a linear dependence f_osc vs S0, but no error bars, repeated measurements, device-to-device variation, or goodness-of-fit metrics are reported. With a single device and a small number of amplitude steps, the linearity and reproducibility of the encoding are not established. At minimum, the authors should report multiple trials, variability, and a fit with confidence bounds, and state whether the trend is linear or sigmoid-like as claimed.","section":"III B / Fig. 7d"},{"comment":"The stated ranges overstate the demonstrated data. The electromechanical rate-encoding experiment uses S0 = 26–60 nm, not '~10 nm', and the measured spike rates are 100–800 Hz, not '100 Hz–1 kHz'. The abstract and conclusion should be bounded by the measured parameter range, or additional experiments covering ~10 nm and rates up to 1 kHz should be provided.","section":"Abstract, III B, Conclusion"}],"minor_comments":[{"comment":"Panel (c) is mislabeled as '(e)' in the caption.","section":"Fig. 6"},{"comment":"In the list of oscillator parameters, 'resistance states R_HRS, R_HRS' should read 'R_HRS, R_LRS'.","section":"III B"},{"comment":"Typos: 'standrad' in the first paragraph and 'Freqency' in Fig. 6b should be corrected.","section":"Introduction"}],"recommendation":"major_revision","confidential_remarks":"The central concept is promising, and the calibration issue is disclosed by the authors rather than hidden. However, the quantitative biological-relevance claims should not be accepted on a nominal 6 nm/V value without verification. I recommend major revision; with direct displacement measurement or substantial hedging, the paper could become acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is a genuine demonstration of the direct chain piezo-MEMS cantilever -> rectifier -> VO2 memristor oscillator -> rate-coded spikes, with an LC addition that makes the output biphasic. The components are known, but the combination and the end-to-end rate-encoding result are new. The authors also cite their own prior cantilever and memristor work appropriately; there is no circularity here, just engineering extension.\n\nWhat the paper does well: it shows clean spiking at stable frequency, a linear dependence of spike rate on the drive amplitude, and a plausible way to get biphasic pulses suitable for cochlear implants. The qualitative story — bigger vibration in, faster spikes out — is convincing. The audio experiments, even if less controlled, back up the electromechanical data.\n\nThe soft spots are real but not fatal. The biggest one: the stimulus displacement is never measured during the rate-encoding experiments. The caption to Fig. 7d and Section II A both say S0 and S are calculated from the actuator drive voltage using its nominal 6 nm/V sensitivity. That makes the 'biologically realistic ~10 nm' and '82–92 dB SPL' claims dependent on an unverified coupling through the custom holder and ribbon cables. The laser vibrometer was used for cantilever characterization but not here. The rate-encoding trend versus drive voltage stands, but the mapping to physiological units is unproven. This needs to be fixed with a direct displacement measurement.\n\nSecond: the central result is shown on a single device, with no error bars and no mention of repeatability. For an experimental claim about encoding, that is thin. Third, the stated 100 Hz–1 kHz range is only demonstrated up to 800 Hz. Minor but sloppy. Fourth, the paper talks about low power, small footprint, and reduced latency but measures none of these. Those are speculative until shown.\n\nIs it a serious paper? Yes. It is a proof-of-concept, not a system, and it should be treated as such. The central idea is plausible and the data support the qualitative mechanism. The missing calibration and statistics are exactly what peer review should catch.\n\nMy recommendation: send it to review. Ask for direct displacement measurement during the encoding experiments, repeated measurements with error bars, and a toned-down conclusion that separates what is measured from what is projected.","headline":"A neat single-channel proof-of-concept that shows real rate-encoding from a MEMS cantilever through a VO2 oscillator, but the headline '10 nm / dB SPL' numbers lean on an unverified actuator calibration and there are no repeated measurements.","tokens_in":12981,"tokens_out":1716,"would_cite":true,"duration_ms":17083,"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 that a single sensing channel—a piezo-MEMS cantilever coupled to a VO2 memristor oscillator—converts nanometer-scale mechanical vibrations directly into rate-coded spiking signals in the 100 Hz–1 kHz range, and shape","keywords":["action potential","cochlear implant","frequency-resolved MEMS array","nanogap memristor","piezo-MEMS","rate encoding","tonotopy","VO2"],"falsifier":"Measure the cantilever's real displacement with a laser interferometric vibrometer while sweeping the piezo drive voltage used in the rate-encoding experiment; if the true displacement does not track the nominal 6 nm/V conversion, or the spike-rate versus true-displacement curve is not the reported approximately linear ramp, the central claim is not supported.","tokens_in":12068,"feed_emoji":"🦻","tokens_out":5381,"duration_ms":45999,"temperature":0.7,"pith_summary":"This paper is trying to establish that a single hardware channel can do what a cochlear implant's front end normally does in software: take a tiny mechanical vibration, pick out its frequency, and encode its loudness as a train of neural-style spikes. The proposed chain is a piezo-MEMS cantilever that resonates at a specific frequency, a rectifier that turns its AC output into a DC level, and a VO2 nanogap memristor oscillator that converts that DC level into current spikes. The authors report spikes between roughly 100 Hz and 1 kHz whose frequency grows with the stimulus amplitude in the nanometer range, matching the rate-encoding behavior of auditory nerve fibers. They also show that adding a parallel inductor-resistor network turns the unipolar spikes into a biphasic waveform, the charge-balanced form required for implant electrodes. If true, this is a direct, energy-lean path to a fully implantable cochlear implant that avoids FFT-based signal processing and its associated power, footprint, and latency costs.","feed_headline":"Nanometer vibrations drive 100-to-800-Hz neural-style spike trains","feed_subtitle":"A MEMS cantilever plus a VO2 memristor turns sound into rate-coded spikes without a Fourier transform.","key_machinery":"The load-bearing element is a relaxation oscillator built from a resistor, capacitor, and a vanadium dioxide (VO2) nanogap memristor: a planar device with a 30–60 nm gap between electrodes. The memristor switches between insulating and metallic states through an insulator-metal transition at a threshold voltage, and each switching cycle emits a current spike. The RC values set the baseline spike timescale, while the DC input voltage—which tracks the rectified cantilever signal—tunes the spike rate. The piezo-MEMS cantilever, a spiral-shaped ScAlN resonator with quality factors above 200, supplies the frequency-selective front end by converting a narrow band of mechanical vibration into an AC","core_discovery":"The paper claims, for the first time, a direct conversion of analog microelectromechanical signals into biomimetic spiking signals. In the key experiment, a piezo-MEMS cantilever with a resonance near 637 Hz is excited mechanically at amplitudes of 26–60 nm, calculated from the actuator's nominal 6 nm/V sensitivity. The cantilever's AC output is amplified, rectified, and smoothed to a DC level that drives a VO2 nanogap memristor relaxation oscillator; the oscillator emits current spikes whose frequency increases approximately linearly with stimulus amplitude, falling in the 100–800 Hz range. Adding a parallel LR circuit converts the unipolar spikes into a biphasic waveform whose numerical in","pith_inferences":["If the linear rate-encoding curve holds across cantilever channels tuned to different resonance frequencies, the full array would produce a sound spectrogram directly as a spike raster, with no digital filter bank; the paper does not demonstrate multi-channel behavior.","The paper routes only the rectified DC signal into the oscillator; coupling the unfiltered AC cantilever signal into the oscillator could add phase-locking, a temporal coding strategy the authors mention as possible but do not test.","The nominal 6 nm/V displacement calibration is a single-point assumption; an independent interferometric measurement of the true cantilever displacement during spiking would pin the amplitude scale and directly test the biological-relevance claim."],"forward_implications":["A single analog channel can perform frequency-selective sensing and rate-encoding of amplitude without any Fourier transform or digital signal processing.","The output spike frequencies land in the biologically relevant 100 Hz–1 kHz auditory nerve range, so the circuit can interface with neural coding schemes directly.","The biphasic waveform produced by the added LR network satisfies the charge-balance safety requirement for cochlear implant electrodes.","Because the VO2 switching events are orders of magnitude faster than the millisecond-scale target spikes, the spike rate is set by RC parameters and can be tuned to the neural domain.","The direct analog path could reduce the roughly 10 ms latency experienced with conventional cochlear implants, which would improve sound localization."],"fun_headline_variants":["MEMS cantilever and VO2 memristor turn vibrations into neural spikes","Cochlear implant chip uses memristor to encode sound as spikes","Implantable sensor: piezo-MEMS and memristor produce spike trains","Frequency-selective spike encoding without FFT for cochlear implants"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claimed nanometer-scale stimulus amplitudes and the rate-encoding curve are computed from the piezoelectric actuator's drive voltage using its nominal 6 nm/V sensitivity; if the actual motion delivered to the cantilever through the mounting differs from that value, the biological-relevance and amplitude-to-rate mapping shift.","fun_headline_variants_meta":{"raw":{"variants":["MEMS cantilever and VO2 memristor turn vibrations into neural spikes","Cochlear implant chip uses memristor to encode sound as spikes","Implantable sensor: piezo-MEMS and memristor produce spike trains","Frequency-selective spike encoding without FFT for cochlear implants"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000265,"raw_usage":{"total_tokens":1473,"prompt_tokens":801,"completion_tokens":672,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":545,"completion_tokens_details":{"reasoning_tokens":601}},"tokens_in":545,"tokens_out":672,"duration_ms":6101,"temperature":1.0,"reasoning_tokens":601,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T13:29:56.155747+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the cantilever's real displacement with a laser interferometric vibrometer while sweeping the piezo drive voltage used in the rate-encoding experiment; if the true displacement does not track the nominal 6 nm/V conversion, or the spike-rate versus true-displacement curve is not the reported approximately linear ramp, the central claim is not supported.","supporting_citations":[],"review_version":1}