{"id":"2408dddf-5106-4389-a67b-a0c83ef1a86a","arxiv_id":"1908.06924","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A graphene/WS2/MoS2 stack detects light with responsivity up to 4.4×10^6 A/W at 30 fW by using a MoS2 floating quantum well to gate the graphene channel.","lead":"This paper reports a graphene/WS2/MoS2 photodetector that reaches a responsivity of 4.4 million amperes per watt when light power is only 30 femtowatts. The device stores light-generated electrons in a MoS2 layer that acts like a floating gate, so a single photon creates a large current change in graphene.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline responsivity and NEP both scale inversely with the uncalibrated optical power stated as 'incident on the junction'; without calibration details, a factor-of-10 error collapses the >10^6 A/W claim.","rationale":"The reader identified the weakest assumption as the accuracy of the optical power incident on the junction, and my stress-test agrees: this is the single most load-bearing concern because R, NEP, and the benchmark comparison in Table 1 all scale inversely with that number. I considered two other potential objections. First, the trap-mediated gain mechanism with slow de-trapping (tau_MT/tau_TM = 5e-4) might seem to conflict with the claimed fast response and zero reminiscent photocurrent, but the absolute trap time constants are not specified, so a tau_TM below 10 ms could satisfy both the gain requirement and the observed transient behavior; this is a modeling ambiguity, not a decisive flaw. Second, direct sub-bandgap absorption in MoS2/WS2 defect states could in principle contribute at 851 nm, but the wavelength-independent responsivity trend and the fast transient response mitigate this concern, and it does not affect the central quantitative claim as directly as the power calibration does. The reader's CONDITIONAL verdict is appropriate: the paper should be accepted only if the authors supply the missing optical calibration details, beam-size information, and measurement uncertainty, and clarify that the model agreement is a fit rather than an independent prediction.","tokens_in":15152,"tokens_out":16473,"duration_ms":184767,"concrete_test":"Reconstruct the optical path and measure, with a NIST-traceable calibrated power meter and a beam-profiling camera placed at the sample plane, the power and intensity profile of the 851 nm beam for the same attenuator settings used in Fig. 5. Integrate the measured intensity over the 6 um^2 junction area to obtain Pop_true, then recompute R = Iph/Pop_true and NEP = NRMS/R for the 30 fW point. If Pop_true is within 10% of 30 fW and R remains above 1e6 A/W, the concern is resolved; if Pop_true is more than 300 fW, the headline responsivity and NEP claims fail.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim, R = Iph/Pop = 4.4e6 A/W at Pop = 30 fW, depends entirely on the accuracy of the optical power said to fall on the 6 um^2 junction. The paper nowhere describes how Pop was calibrated at the sample plane, how the beam spot size was measured relative to the junction area, how the 30 fW level was established (e.g., ND filter chain, power-meter noise floor, stray-light rejection), or what the estimated uncertainty is. At 30 fW, the power is far below the noise floor of typical calibrated photodiode power meters, so an unstated extrapolation through attenuators is required; such extrapolations can easily be off by an order of magnitude or more. Since R is inversely proportional to the true junction power, a factor-of-10 power underestimate changes R from 4.4e6 to 4.4e5 A/W, invalidating the '>10^6 A/W' headline. The NEP claim (<4 fW/sqrt(Hz)) is likewise affected because NEP = NRMS/R, so it would increase by the same factor. The rate-equation model in Eqs. (1)-(4) cannot rescue this: the model is fitted to the same power axis and its parameters are not independently constrained. This is not a dispute about the photogating physics, which is plausible; it is a claim that the empirical anchor of the entire performance benchmark is undocumented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a vertical FLG/2L-WS2/1L-MoS2 heterojunction photodetector in which photoexcited electrons in few-layer graphene transfer to a monolayer MoS2 quantum well, producing a photogating effect. The authors report a responsivity of 4.4×10^6 A/W at 30 fW incident power at 851 nm, an NEP < 4 fW/√Hz, response times below 10 ms, and no persistent photocurrent. They support the mechanism with temperature-dependent PL quenching measurements and a rate-equation model with fitted parameters.","tokens_in":15511,"tokens_out":4311,"duration_ms":41804,"significance":"If the performance figures are reproducible, the detector would represent a substantial advance over graphene-absorption photodetectors, combining high gain with millisecond-scale response. The PL experiments provide direct evidence of carrier storage in MoS2, and the fabrication of two devices (D1, D2) with consistent behavior is a strength. The principal issue is that the headline responsivity and NEP rest on an uncalibrated optical power value at the junction; until that is documented, the magnitude of the claimed improvement is not verifiable.","major_comments":[{"comment":"The paper does not describe how the 30 fW optical power falling on the 6 µm² junction was determined. The text states only that this is the smallest power 'incident on the junction' during the experiment, with no details of the calibration chain (e.g., attenuator accuracy, beam spot size relative to the junction, power meter traceability). Because R = Iph/Pop and NEP = NRMS/R, any error in Pop propagates linearly into both headline figures; a factor-of-10 underestimate would reduce R to 4.4×10^5 A/W. The authors must provide a complete optical calibration description and an uncertainty estimate.","section":"§2.4, Fig. 5a and Eq. (6)"},{"comment":"The 'model predicted' responsivity curve is obtained by fitting three free parameters (τMG/τGM=10, τMT/τTM=5×10^-4, NT=5×10^10 cm^-2), and the internal quantum efficiency η is not assigned a value. The agreement over 11 orders of magnitude is therefore a demonstration of the model's flexibility, not an independent prediction. The manuscript should either determine these parameters independently or clearly label the curve as a fit and discuss the sensitivity of the conclusion to the fitted values.","section":"§2.4, Eqs. (1)-(4) and Fig. 5a"},{"comment":"None of the measured responsivity or SNR data points carry error bars, and the NEP of <4 fW/√Hz is derived from a single SNR measurement (SNR≈10 at 30 fW). The authors should provide repeated measurements or a statistical uncertainty analysis to support the stated performance margins.","section":"Fig. 5a and Fig. 5c"}],"minor_comments":[{"comment":"The term 'reminiscent photocurrent' is used; the standard term is 'persistent photocurrent'.","section":"Throughout"},{"comment":"The text says 'Plank constant'; it should be 'Planck constant'.","section":"§2.4, Eq. (4)"},{"comment":"The x-axis of Fig. 5a is 'Power Density (W/m^2)' while Fig. 5c and the text use 'Power (W)'; the conversion between the two, based on the 6 µm² junction area, should be stated explicitly.","section":"Fig. 5a and Fig. 5c"},{"comment":"The exclusion of direct sub-bandgap absorption in MoS2/WS2 is argued from noise and transient response; a direct measurement such as photocurrent excitation spectroscopy below the bandgap would strengthen this conclusion.","section":"§2.3"},{"comment":"The rise and fall times are reported as <10 ms and stated to be limited by the measurement equipment; the bandwidth of the measurement setup should be specified to make this bound meaningful.","section":"§2.3, Fig. 4c-e"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a promising device concept and the PL data support the proposed carrier-storage mechanism. However, the central performance claims depend on an optical power calibration that is not described at all, and the supporting model is a fit rather than an independent prediction. These issues are fixable in revision, but they currently prevent a reliable assessment of the key result. I recommend major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The device concept is genuinely new, and the measured transients at 785 and 851 nm look clean. But the headline 4.4×10^6 A/W rests on a 30 fW optical power that the paper never calibrates, and the model that 'confirms' the mechanism is a three-parameter fit to the same data. The physics is plausible; the proof is incomplete.\n\nWhat is new: few-layer graphene serves as both absorber and conduction channel, and a fraction of the photoelectrons transfer into a floating monolayer MoS2 quantum well through a WS2 barrier. That keeps graphene's broadband response while adding photogating gain, which earlier high-gain designs lost by using other absorbers. The PL quenching and low-temperature valley polarization results support charge storage in the MoS2. Two devices show similar responsivity and fast switching, and the absence of persistent photocurrent is a practical plus.\n\nSoft spots, in order. First, the optical power calibration. At 30 fW the paper does not describe how power at the junction was established: no spot-size measurement relative to the 6 µm² junction, no attenuation chain details, no uncertainty estimate or stray-light check. Since R and NEP both scale inversely with true junction power, a factor-of-10 error drops the claim to 4.4×10^5 A/W and invalidates the '>10^6 A/W' headline. This is the load-bearing number. Second, the rate-equation model in §2.4 has three fitted parameters (τMG/τGM=10, τMT/τTM=5×10^-4, NT=5×10^10 cm^-2), and the 'model predicted' curve is a fit to the same responsivity data, so the agreement over 11 orders is not independent validation. Third, there are no error bars anywhere; some indication of device-to-device variation would help.\n\nMinor: the noise heuristic for why 532 nm is noisier is plausible but not proven. The benchmarking table is a bit loose, but the relative claim against prior graphene-absorption detectors looks credible.\n\nWho this is for: experimentalists in 2D photodetectors and photogating. The paper deserves a serious referee but needs major revision before the headline number can be trusted. The authors should provide the calibration method, spot-size measurement, and error analysis, and should present the model as a fit, not a prediction.\n\nRecommendation: send it to peer review, with reviewers instructed to press on the optical calibration. If that holds up, this is a significant result.","headline":"A genuinely new photogating device that could be important, but the headline responsivity depends on an uncalibrated 30 fW power, so the central number is not yet trustworthy.","tokens_in":16007,"tokens_out":3284,"would_cite":false,"duration_ms":29599,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["85.60.Gz","73.50.Pz"],"model":"deepseek-v4-flash","headline":"A graphene photodetector reaches 4.4 million A/W sensitivity","keywords":["graphene photodetector","photogating","floating quantum well","MoS2","WS2","van der Waals heterojunction","responsivity","noise equivalent power"],"falsifier":"Measure the actual optical power at the junction using a calibrated power meter and a beam spot that is fully mapped relative to the 6 $\\mu$m$^2$ active area, then recompute the responsivity; if the true power is ten times higher, the headline $4.4\\times 10^6$ A/W drops to about $4\\times 10^5$ A/W. Independent reproduction of the FLG/WS2/MoS2 stack with the same measurements would also decide whether the device performs as claimed.","tokens_in":14979,"feed_emoji":"🔬","tokens_out":7118,"duration_ms":61257,"temperature":0.7,"pith_summary":"This paper claims that a vertical stack of few-layer graphene on bilayer WS2 on monolayer MoS2 detects light with a sensitivity that outperforms every previously reported graphene-absorption-based photodetector by several orders of magnitude. The design keeps graphene as both the light absorber and the conduction channel, while a fraction of photoexcited electrons transfer rapidly into a floating MoS2 quantum well that electrostatically gates the graphene and amplifies the photocurrent. The authors report a responsivity of $4.4\\times 10^6$ A/W at 30 fW incident power, a noise equivalent power below 4 fW/$\\sqrt{\\mathrm{Hz}}$, and millisecond-scale response with no lingering photocurrent. If the claim holds, graphene-based detectors become viable for low-intensity, broadband, room-temperature photodetection without slow recovery or refresh circuitry.","feed_headline":"Graphene photodetector hits 4.4 million A/W sensitivity","feed_subtitle":"A floating MoS2 quantum well turns graphene's weak absorption into record sensitivity, low noise, and millisecond speed.","key_machinery":"The load-bearing object is the vertical van der Waals heterojunction few-layer graphene / bilayer WS2 / monolayer MoS2, where the monolayer MoS2 acts as a floating quantum well gate for the graphene channel. The mechanism is quantum-confined photogating: ultrafast interlayer electron transfer from graphene into the MoS2 quantum well builds a stored electron population that shifts the graphene Dirac point and changes its conductance, giving high photoconductive gain. The thin WS2 layer controls both the injection rate into the well and the tunnel-back rate to graphene, and the conduction-band offset between MoS2 and graphene provides a built-in field that swiftly empties the well when light is turned off. The argument is carried by a coupled rate-equation model for the photoinduced carrier density in graphene, the electron density in the MoS2 conduction band, and the trapped-carrier density in MoS2, with parameters fitted to reproduce the responsivity-versus-power curve.","core_discovery":"The central discovery is that a floating monolayer MoS2 quantum well, coupled to a few-layer graphene channel through an ultrathin WS2 spacer, produces a quantum-confined photogating effect strong enough to give a measured responsivity of $4.4\\times 10^6$ A/W at 30 fW incident power (851 nm) and a noise equivalent power below 4 fW/$\\sqrt{\\mathrm{Hz}}$, while the built-in band offset discharges the MoS2 rapidly when illumination stops, giving response times under 10 ms with zero reminiscent photocurrent. The paper frames this as the first time a graphene-absorption detector simultaneously achieves ultra-high responsivity, low noise, and fast, self-restoring response at room temperature. A coupled rate-equation model with density-dependent recombination in graphene, interlayer transfer in both directions, and trapping in MoS2 reproduces the measured responsivity over 11 orders of magnitude of incident power density.","pith_inferences":["The fitted model's slow de-trapping time ($\\tau_{TM} \\gg \\tau_{MG}$) implies that MoS2 defect states dominate the gain at low power, so engineering the defect density in MoS2 should tune the responsivity; this is a testable prediction.","At 30 fW with SNR near 10, the device is approaching the few-hundred-photon-per-second regime; whether it can be pushed toward single-photon detection depends on reducing the dark current noise further and on the exact optical power calibration.","The same floating-quantum-well architecture could be transferred to other transition-metal-dichalcogenide pairs with suitable band offsets, and if graphene remains the absorber, the principle should hold at longer wavelengths than the TMD bandgaps.","The absolute responsivity values should be read as upper bounds until the optical power at the junction is independently calibrated; the relative comparison across wavelengths is less sensitive to that uncertainty."],"forward_implications":["Graphene-absorption photodetectors can reach sensitivities that rival detectors with separate absorbing layers, while preserving graphene's broadband absorption from the visible to the near-infrared.","The automatic discharge mechanism removes the need for the gate-voltage refresh pulses used in earlier photogating detectors, simplifying the system.","With NEP below 4 fW/$\\sqrt{\\mathrm{Hz}}$ and response times under 10 ms, the device is practical for room-temperature low-light sensing in the near-infrared.","Because the responsivity follows a single trend line across 532 to 1050 nm, the photogating mechanism should extend across the full graphene absorption band."],"supporting_citations":[{"why":"Establishes the 2.3% monolayer graphene absorption that the paper cites as the starting limitation.","marker":"[17]"},{"why":"Supplies the density-dependent carrier recombination rate in graphene used in the rate-equation model.","marker":"[19]"},{"why":"Benchmark metal/graphene/metal detector with picosecond speed but sub-milliamp responsivity, the speed-sensitivity trade-off the paper aims to beat.","marker":"[20]"},{"why":"Graphene/Ta2O5/graphene detector with high responsivity but slow response, a key comparison for the trade-off.","marker":"[30]"},{"why":"Hybrid graphene-quantum-dot phototransistor showing ultrahigh gain via photogating, the approach the paper modifies with a floating quantum well.","marker":"[31]"},{"why":"Graphene/MoS2 Schottky photodetector used as a baseline for graphene-absorption performance.","marker":"[40]"},{"why":"Pure monolayer graphene detector with quantum-dot arrays, showing high photoresponse but slow recovery, another trade-off reference.","marker":"[45]"},{"why":"Demonstrates ultrafast interlayer charge transfer in MoS2/WS2 heterostructures, the physical basis for the fast transfer into the quantum well.","marker":"[48]"}],"fun_headline_variants":["Floating quantum well boosts graphene photodetector to 4.4M A/W","Graphene photodetector reaches 4.4M A/W via floating MoS2 quantum well","Floating quantum well gate unlocks 4.4M A/W graphene photodetector"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim depends on the accuracy of the optical power stated as falling on the 6 $\\mu$m$^2$ junction; the paper does not report power-meter calibration, beam spot size relative to the junction, or measurement uncertainty, so if the true power at the junction is larger than stated, the responsivity is proportionally lower.","fun_headline_variants_meta":{"raw":{"variants":["Floating quantum well boosts graphene photodetector to 4.4M A/W","Graphene photodetector reaches 4.4M A/W via floating MoS2 quantum well","Floating quantum well gate unlocks 4.4M A/W graphene photodetector"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001119,"raw_usage":{"total_tokens":4687,"prompt_tokens":1006,"completion_tokens":3681,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":622,"completion_tokens_details":{"reasoning_tokens":3604}},"tokens_in":622,"tokens_out":3681,"duration_ms":23983,"temperature":1.0,"reasoning_tokens":3604,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:30:47.329028+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the actual optical power at the junction using a calibrated power meter and a beam spot that is fully mapped relative to the 6 $\\mu$m$^2$ active area, then recompute the responsivity; if the true power is ten times higher, the headline $4.4\\times 10^6$ A/W drops to about $4\\times 10^5$ A/W. Independent reproduction of the FLG/WS2/MoS2 stack with the same measurements would also decide whether the device performs as claimed.","supporting_citations":[{"cited_title":"R.; Blake, P.; Grigorenko, A","cited_arxiv_id":null,"evidence_quote":"Establishes the 2.3% monolayer graphene absorption that the paper cites as the starting limitation."},{"cited_title":"A.; Strait, J","cited_arxiv_id":null,"evidence_quote":"Supplies the density-dependent carrier recombination rate in graphene used in the rate-equation model."},{"cited_title":"Graphene Photodetectors for High-Speed Optical Communications","cited_arxiv_id":null,"evidence_quote":"Benchmark metal/graphene/metal detector with picosecond speed but sub-milliamp responsivity, the speed-sensitivity trade-off the paper aims to beat."},{"cited_title":"H.; Chang, Y","cited_arxiv_id":null,"evidence_quote":"Graphene/Ta2O5/graphene detector with high responsivity but slow response, a key comparison for the trade-off."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Hybrid graphene-quantum-dot phototransistor showing ultrahigh gain via photogating, the approach the paper modifies with a floating quantum well."},{"cited_title":"A.; Sinha, R.; Karabiyik, M.; Das, S.; Choi, W.; Pala, N","cited_arxiv_id":null,"evidence_quote":"Graphene/MoS2 Schottky photodetector used as a baseline for graphene-absorption performance."},{"cited_title":"Y.; Liu, T.; Meng, B.; Li, X.; Liang, G.; Hu, X.; Wang, Q","cited_arxiv_id":null,"evidence_quote":"Pure monolayer graphene detector with quantum-dot arrays, showing high photoresponse but slow recovery, another trade-off reference."},{"cited_title":"Ultrafast Charge Transfer in Atomically Thin MoS _2 /WS _2 Heterostructures","cited_arxiv_id":null,"evidence_quote":"Demonstrates ultrafast interlayer charge transfer in MoS2/WS2 heterostructures, the physical basis for the fast transfer into the quantum well."}],"review_version":1}