{"id":"cda35b2a-b86a-44ed-9c96-c276853b6ec0","arxiv_id":"2412.11816","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A Langevin model with a simplified diode-bridge description reproduces measured power-versus-load curves of a piezoelectric harvester under broadband vibrations and reveals time asymmetry from single-variable correlations.","lead":"The authors model a piezoelectric energy harvester driven by random vibrations and show that a stochastic equation with a simplified diode-bridge circuit reproduces measured power curves. The work helps designers pick the load resistance that extracts the most power from vibration harvesters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The diode-bridge agreement in Figs. 9-10 rests on undisclosed effective parameters Ik0 and G; until they are calibrated independently, the model's nonlinear power curve is not shown to be a prediction rather than a fit.","rationale":"The stress-test pass focused on the strongest claim: that the stochastic model reproduces the measured power-versus-load-resistance curves in both configurations. The linear configuration is supported by an analytical expression and by independently checked Gaussian voltage histograms in Fig. 5, so that part is reasonably secure. For Configuration (II), the only quantitative validation of the nonlinear power curve is numerical simulation with two effective diode parameters whose origin is undisclosed. This is the least secure link in the chain: the success of the nonlinear model depends on those two numbers, and without knowing whether they were fixed a priori or tuned to reproduce Fig. 10, the reader cannot distinguish prediction from fit. The paper does not overclaim: it explicitly admits the v_DC distribution is not reproduced and calls the time-asymmetry comparison qualitative. That honesty supports a CONDITIONAL rather than REJECT verdict. The proposed test would settle the issue by re-fitting on a single R value and checking all other points, or by deriving Ik0 and G from direct diode measurements. Since this concern matches the reader's weakest assumption and the proposed remedy is a verification step rather than a change of verdict, UNCHANGED is appropriate.","tokens_in":8314,"tokens_out":4767,"duration_ms":48373,"concrete_test":"Fit Ik0 and G by maximum likelihood (or by matching <v_DC> and <v_DC^2>) using only the R=3300 ohm time series at a=0.8 g, then freeze these values and simulate the full R sweep for both acceleration levels. If the predicted P_harv(R) deviates from the experimental values by more than the experimental uncertainty, or if the predicted optimal resistance leaves the 20-40 kOhm range, the Fig. 10 agreement is a fit rather than a validated prediction. As a complementary check, derive Ik0 and G from a direct measurement of the four-diode bridge I-V characteristic and compare with the reported values.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim for Configuration (II) is that the stochastic model (1)-(6), with fixed parameters Ik0=9e-9 A and G=-133e-11 ohm^-1, reproduces P_harv^(II)=<v_DC^2>/R over the full R sweep in Fig. 10. The paper does not state how Ik0 and G were chosen, does not report error bars on P_harv, and does not provide data/code. Since the paper itself admits that the full v_DC distribution is not reproduced (Fig. 11), matching <v_DC> and <v_DC^2> is weak evidence: with two free parameters one can match two moments at a given R, and the non-monotonic optimum at R* ~ 30 kOhm can be tuned by the negative slope G. The constant values of Ik0 and G also need to be validated against the measured diode-bridge I-V characteristic across the output-voltage range. The agreement in Fig. 10 is therefore load-bearing: if Ik0 and G are actually re-fit to the whole R curve, the central claim reduces to a two-parameter interpolation, not an independent model prediction. The paper's own admission that the v_DC distribution mismatch is 'probably due to the simplifications introduced in the model' reinforces this concern, but does not settle it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a stochastic Langevin model for a piezoelectric energy harvester driven by broadband Gaussian vibrations and compares it with experiments in two load configurations. In Configuration (I), a linear resistive load is treated analytically, yielding the closed-form power expression in Eq. (7); the model parameters are fitted at one acceleration, and the curve is then compared with data at a second acceleration. In Configuration (II), a diode-bridge rectifier with a capacitor and load resistance is modeled by the effective nonlinear equation in Eq. (6), and numerical simulations are compared with measurements of the mean rectified voltage and harvested power as functions of load resistance, with a non-monotonic power curve and an optimum near R* ~ 30 kOhm. The paper also examines three- and four-point connected correlation functions of vDC, claiming that their time asymmetry reveals the nonequilibrium nature of the dynamics, with qualitative agreement between experiment and simulation.","tokens_in":8618,"tokens_out":3971,"duration_ms":41153,"significance":"If the nonlinear configuration were shown to be genuinely predictive, this would be a valuable contribution to stochastic modeling of piezoelectric harvesters with rectified loads, especially the compact effective description of the diode bridge in Eq. (6). The linear configuration is handled carefully: Eq. (7) is analytical, the parameters are fitted at one acceleration and the model then captures data at a second acceleration, and the Gaussian voltage distributions in Fig. 5 are consistent with the linear model. The proposed time-asymmetry analysis for a single measured variable is also interesting and well motivated by earlier work on linear systems. However, the nonlinear part of the paper, which is the main new experimental claim, is currently under-supported: the two effective diode parameters are introduced without disclosing how they were determined, no error bars are given for the central quantities, and the paper itself admits that the full vDC distribution is not captured by the model. These issues make it impossible to judge whether the agreement in Figs. 9 and 10 is an independent prediction or a two-parameter fit.","major_comments":[{"comment":"The effective diode parameters Ik0 = 9e-9 A and G = -133e-11 Ohm^-1 are introduced without stating how they were obtained. If they were chosen by matching the experimental power curve in Fig. 10, then the agreement shown there is a two-parameter interpolation rather than an independent model prediction. Please disclose the calibration procedure, and ideally validate Ik0 and G independently, for example by comparing the effective I-V characteristic in Eq. (6) with a measured diode-bridge characteristic, or by fitting at a single resistance and predicting the full R sweep.","section":"Section III B, Eq. (6)"},{"comment":"No error bars or statistical uncertainties are reported for the mean rectified voltage and the harvested power. Without such uncertainties, the claim of good agreement for the non-monotonic power curve, and in particular for the optimal resistance near R* ~ 30 kOhm, cannot be assessed quantitatively. Please report error bars from repeated measurements or from bootstrap resampling, and give a quantitative measure of agreement such as relative error or chi-square per degree of freedom.","section":"Figs. 9 and 10, Section III B"},{"comment":"The paper states that while average and variance are well described, the full vDC distribution is not reproduced, and attributes this to simplifications in the diode model. Since the harvested power in Eq. (8) is proportional to the second moment of vDC, the distribution tails directly affect the central quantity of the paper. Please quantify the contribution of the tail mismatch to <vDC^2> and discuss whether the model remains reliable for the power curve outside the tested range of load resistances.","section":"Fig. 11 and Section III B"},{"comment":"The claim that time asymmetry is revealed by the high-order correlation functions is based on a visual comparison of C(3) and C(4) with their time-reversed counterparts. No statistical test, confidence intervals, or quantitative asymmetry measure is provided, and the simulation curves show only qualitative agreement with experiment. Please quantify the asymmetry, include error estimates, and state explicitly whether the simulation reproduces the experimental asymmetry within statistical uncertainty.","section":"Section III C, Figs. 12 and 13"}],"minor_comments":[{"comment":"The experimental section reports the sampling rate fs = 5 kHz but not the duration of each time series or the number of independent realizations; these details are needed to assess stationarity and statistical precision.","section":"Section II"},{"comment":"The definitions of C(3) and C(4) are not obviously normalized cumulants; please clarify the normalization and state explicitly that the time-reversed functions are computed from the same data with the time argument reversed.","section":"Eqs. (9) and (10)"},{"comment":"Using a consistent plotting style across panels, such as filled markers for experimental data and lines for simulations, would improve readability; where error bars are absent this should be stated in the captions.","section":"Figs. 4-13"},{"comment":"The parameter values M, theta, and gamma are quoted with uncertainties, but the fitting procedure is not described; a sentence on the fitting method (e.g., least squares on log P versus R) would be helpful.","section":"Section III A"}],"recommendation":"major_revision","confidential_remarks":"The main issue is the undisclosed calibration of Ik0 and G in the nonlinear configuration. This is a fixable problem: the authors should provide the calibration data, error bars, and ideally an independent validation of the effective diode parameters. The linear part is sound and the time-asymmetry analysis is interesting but needs statistical support. I do not see grounds for rejection, but the nonlinear central claim must be substantiated before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid, modest engineering-model paper. The genuinely new bit is the effective diode-bridge description (Eq. 6) and the single-variable high-order correlation test for time asymmetry. The linear configuration (I) is handled well analytically; Eq. (7) fits the resistor data, and parameters fitted at 0.8 g carry over to 1.0 g. That part is credible.\n\nThe soft spot is the nonlinear configuration. Ik0 and G in Eq. (6) are introduced as 'effective' values with one line saying what they are, and no procedure for obtaining them, no comparison with the diode's actual I-V, no error bars on the measured power, and no data/code. With two free parameters you can match a mean and a second moment at one R; the negative G does a lot of work in producing the optimum near 30 kOhm. The authors admit in Fig. 11 that the v_DC distributions are not reproduced. So the claim that the model 'reproduces the experimental behavior' is fair at the level of two moments, but it is not an independent prediction. The time-asymmetry part is likewise qualitative: both experiments and simulations show a peak that flips under time reversal, but the shapes differ and the quantitative disagreement is hand-waved as due to diode simplifications.\n\nThese flaws are not fatal. The paper is honest about the distribution mismatch and does not overclaim in the conclusions; the effective model is a reasonable simplification for an engineering audience. But the calibration gap is load-bearing enough that the paper should not be accepted without the authors showing how Ik0 and G are chosen, ideally from a measured diode I-V, and giving some statistical error bars. If they turn out to be re-fit to the whole R curve, the 'model' becomes a two-parameter interpolation, and the practical guidance (optimal R ~ 30 kOhm) is a curve-fitting outcome rather than a model prediction.\n\nWho is this for? Researchers in vibration energy harvesting, especially those who want a quick stochastic model for rectified loads. It is not a major theoretical contribution, and the self-citation cluster is fine—the new work is genuinely incremental over [15-17]. I would send it to review rather than desk-reject, with a request for the diode parameter calibration and error analysis. If those are not supplied, I would not rely on the nonlinear power curve quantitatively.","headline":"A competent incremental extension of the same group's Langevin model to rectified loads; the power-vs-R agreement is real but rests on two never-calibrated diode parameters, so the validation claim needs a caveat.","tokens_in":9139,"tokens_out":2341,"would_cite":false,"duration_ms":23319,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.40.-a","84.60.-h"],"model":"deepseek-v4-flash","headline":"A Langevin model with a two-parameter diode bridge reproduces power-versus-load curves of a piezoelectric harvester; high-order correlations of one voltage reveal temporal asymmetry in the nonlinear setup.","keywords":["piezoelectric energy harvester","stochastic Langevin model","broadband vibrations","white noise","diode bridge rectifier","load resistance optimization","time-reversal asymmetry","high-order correlation functions"],"falsifier":"Repeat the experiment at an acceleration not used in calibration (for instance 0.9 g) or with a different storage capacitor, and test whether the numerical model with the same $I_{k0}$ and $G$ still reproduces the measured power-versus-resistance curve; if the predicted optimal resistance or peak power moves away from the data, the effective parameters are specific to the fitted conditions rather than predictive.","tokens_in":8104,"feed_emoji":"⚡","tokens_out":10949,"duration_ms":87185,"temperature":0.7,"pith_summary":"The paper aims to show that a minimal stochastic description—an underdamped Langevin equation driven by white noise, coupled to the load circuit—can capture how a real piezoelectric energy harvester behaves when shaken by random broadband vibrations. The practical point is that the model gives a quantitative route to the load resistance that maximizes extracted power, which is exactly the design question for powering small sensors from ambient vibration. Experiments on a commercial harvester are compared with the model for two load circuits: a plain resistor, and a diode-bridge rectifier with a storage capacitor (the configuration needed to supply DC electronics). In both cases the model reproduces the measured non-monotonic power-versus-resistance curve, with optimal resistances around $3$–$4$ k$\\Omega$ and about $30$ k$\\Omega$, respectively. The paper further claims that, in the nonlinear setup, time-reversal asymmetry—a hallmark of non-equilibrium dynamics—can be detected from high-order correlation functions of a single recorded voltage, which was not true in the linear setup.","feed_headline":"Stochastic model predicts optimal load for random-vibration harvesters","feed_subtitle":"Simulations match both load types, locating the resistance that maximizes harvested power.","key_machinery":"The load-bearing object is the coupled system of stochastic differential equations (1)–(6): an underdamped Langevin equation for the tip-mass displacement and velocity, an electromechanical coupling term $\\theta v_p$ feeding the mechanical dynamics back into the electrical side, and a load equation whose current-voltage characteristic $f(v_p)$ is the only part that changes between configurations. For the linear load $f(v_p)=v_p/R$, the model is Gaussian and exactly solvable, giving the analytic power formula (7) used for parameter calibration. For the diode-bridge rectifier, $f(v_p)$ takes the effective form in Eq. (6) with two constants $I_{k0}$ and $G$, which the paper treats as the entire description of the nonlinear load. The second main tool is the set of connected three- and four-point correlation functions $C^{(3)}_{v_{\\rm DC}}(t)$ and $C^{(4)}_{v_{\\rm DC}}(t)$ defined in Eqs. (9)–(10); their asymmetry under time reversal is the diagnostic that reveals non-equilibrium dynamics from a single measured time series.","core_discovery":"The central claim is that the stochastic model of Eqs. (1)–(6) is a valid minimal description for this harvester in both load configurations. In the linear case the model is exactly solvable and yields a closed formula for the extracted power, $P_{\\rm harv}^{(I)} = \\frac{D_0 M^2 R\\theta^2}{M(\\gamma+R\\theta^2)+C_p R\\gamma(C_p K_s R+\\gamma+R\\theta^2)}$, whose parameters are fitted to one acceleration value and then successfully describe a second one. For the nonlinear diode-bridge configuration, the paper proposes a simplified two-parameter effective model of the rectifier, $I_{k0} = 9 \\times 10^{-9}$ A and $G = -133 \\times 10^{-11}\\,\\Omega^{-1}$, and numerical simulations of the same Langevin equations reproduce the experimentally measured average voltage and extracted power over the whole tested range of load resistances. The power curve is non-monotonic with a maximum near $30$ k$\\Omega$, which is the optimal load for this rectifier circuit. The same data analysis shows that connected three- and four-point correlation functions of the single measured voltage $v_{\\rm DC}$ are not symmetric under time reversal, revealing the non-equilibrium nature of the dynamics in the nonlinear configuration.","pith_inferences":["Since the paper does not explain how $I_{k0}$ and $G$ are obtained, the nonlinear agreement may be a calibration rather than an independent prediction; a sharper test would fix these two constants on one load and then predict the whole power curve.","The admitted failure to reproduce the full $v_{\\rm DC}$ distribution suggests that any quantity sensitive to higher cumulants of the rectified voltage—such as peak voltage stress on the storage capacitor or efficiency under intermittent loads—may require a more detailed diode model.","The high-order-correlation diagnostic demonstrated here could be applied to other nonlinear energy-conversion or biological systems where only one observable is accessible; the paper does not establish the data length needed for reliable detection.","If the effective diode parameters turn out to be universal for a given diode type or only depend on easily measured quantities, the model becomes a design tool; otherwise each circuit would need its own calibration."],"forward_implications":["For the linear resistive load, the optimal resistance is independent of the forcing acceleration, so one load value serves a range of vibration levels.","The same fitted mechanical parameters describe both tested accelerations, indicating that calibrating the model on one condition may transfer to other driving intensities.","The two-parameter diode model is enough to locate the optimal load of the rectifier circuit, which makes the model usable for designing the AC/DC stage without simulating the full diode bridge.","In the nonlinear configuration, a single measured voltage time series is sufficient to expose time-reversal asymmetry through high-order correlations, unlike the linear configuration where cross-correlations between two variables were needed.","Because the model captures the full probability distribution of $v_p$ in both setups, the approach is not limited to mean power but also describes voltage fluctuations, which matter for the electronics being powered."],"supporting_citations":[{"why":"supplies the linear-configuration model, its analytical solution, and the power formula (7) used to fix the mechanical parameters.","marker":"[15]"},{"why":"extends the same stochastic-thermodynamics framework to an electromagnetic harvester, supporting the generality of the modeling approach.","marker":"[16]"},{"why":"establishes the linear-configuration result that single-variable correlations cannot reveal time-reversal asymmetry, the baseline the nonlinear result is contrasted with.","marker":"[17]"},{"why":"documents the pitfall of time-irreversibility inference in linear systems and motivates the use of higher-order correlation functions.","marker":"[20]"},{"why":"provides the standard modeling context for energy harvesters driven by broadband random vibrations that this work builds on.","marker":"[11]"}],"fun_headline_variants":["Stochastic model predicts best load for vibration harvesters","Optimal load for random-vibration harvesters predicted","Model matches power from both linear and nonlinear harvester loads","Non-monotonic power curve exposes optimal load for harvester","Langevin model replicates harvester power across load types"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The nonlinear power comparison depends on treating the diode bridge as two constant effective parameters, $I_{k0}=9\\times10^{-9}$ A and $G=-133\\times10^{-11}\\,\\Omega^{-1}$, valid for every load resistance and both acceleration levels, even though the paper does not disclose how those values were obtained.","fun_headline_variants_meta":{"raw":{"variants":["Stochastic model predicts best load for vibration harvesters","Optimal load for random-vibration harvesters predicted","Model matches power from both linear and nonlinear harvester loads","Non-monotonic power curve exposes optimal load for harvester","Langevin model replicates harvester power across load types"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000276,"raw_usage":{"total_tokens":1699,"prompt_tokens":1051,"completion_tokens":648,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":667,"completion_tokens_details":{"reasoning_tokens":568}},"tokens_in":667,"tokens_out":648,"duration_ms":17223,"temperature":1.0,"reasoning_tokens":568,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:32:29.809152+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the experiment at an acceleration not used in calibration (for instance 0.9 g) or with a different storage capacitor, and test whether the numerical model with the same $I_{k0}$ and $G$ still reproduces the measured power-versus-resistance curve; if the predicted optimal resistance or peak power moves away from the data, the effective parameters are specific to the fitted conditions rather than predictive.","supporting_citations":[{"cited_title":"Stochastic thermodynamics of a piezoelectric energy harvester model","cited_arxiv_id":null,"evidence_quote":"supplies the linear-configuration model, its analytical solution, and the power formula (7) used to fix the mechanical parameters."},{"cited_title":"Stochastic thermodynamics of an electromagnetic energy harvester","cited_arxiv_id":null,"evidence_quote":"extends the same stochastic-thermodynamics framework to an electromagnetic harvester, supporting the generality of the modeling approach."},{"cited_title":"Inference of Time-Reversal Asymmetry from Time Series in a Piezoelectric Energy Harvester","cited_arxiv_id":null,"evidence_quote":"establishes the linear-configuration result that single-variable correlations cannot reveal time-reversal asymmetry, the baseline the nonlinear result is contrasted with."},{"cited_title":"Energy harvesters driven by broadband random vibrations","cited_arxiv_id":null,"evidence_quote":"provides the standard modeling context for energy harvesters driven by broadband random vibrations that this work builds on."}],"review_version":1}