{"id":"94b53f85-0af9-47d9-baab-0e57d9941c49","arxiv_id":"2505.13637","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A bilayer-graphene/hBN moiré device counts single photons through a bistable negative-differential-conductance state at wavelengths up to 11.3 micrometers and temperatures up to 25 K.","lead":"A stacked graphene-plus-boron-nitride device can register individual light particles at mid-infrared and visible wavelengths while running at 25 kelvin. The detection works through a sharp electrical bistability, and the paper argues this bistability comes from exotic high-field electron motion in the moiré lattice.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-photon claim rests on a power-law and Poissonian-click test that cannot distinguish one-photon triggering from linearized multi-photon or bolometric response; no absolute photon-number calibration is provided.","rationale":"The reader's formal weakest_assumption is the gapped stacking domain behind the negative-differential-velocity mechanism. I regard the absence of a calibrated single-photon measurement as more load-bearing for the headline claim: even if the domain and NDV mechanism are correct, the experiments do not currently prove that a single photon triggers a switching event. The reader's rationale does mention the missing calibration, non-Poissonian mid-IR statistics, and the lack of a detection efficiency, so my agreement is partial rather than full. I keep the CONDITIONAL verdict: the NDC/bistability phenomenology, transport data, spatial maps, and two-region model are substantial and internally consistent, and the authors explicitly flag that the microscopic switching process remains open. The decisive missing experiment is an attenuation sweep that reaches a calibrated mean photon number per reset below one and checks whether the count rate remains linear in photon number there. If that test fails, the paper should be reported as a photosensitive bistable switch without a validated single-photon counting claim; if it passes, the visible-wavelength single-photon claim would be supported, while the mid-IR claim would still require equivalent statistics under reduced background radiation.","tokens_in":29980,"tokens_out":8054,"duration_ms":90546,"concrete_test":"Perform a calibrated attenuation sweep at lambda = 675 nm at the bias point used for Fig. 3C: attenuate the beam in calibrated steps so that the mean number of absorbed photons per reset period spans from the reported ~10^4 down to below 1, recording Gamma_light and click histograms at each step. If Gamma_light per absorbed photon remains constant into the sub-photon-per-reset range and variance=mean persists there, the single-photon triggering interpretation is supported; if Gamma_light/P rolls off, becomes sublinear, or is buried in dark counts before reaching <1 photon per reset, the single-photon claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Load-bearing concern: the 'single-photon counting' conclusion is not established by the reported evidence. The two arguments offered are (i) Gamma_light proportional to P with exponent x=1 near Ic (Figs. 2D, 3B) and (ii) Poissonian click histograms (Fig. 3C,D). Neither is sufficient. For any detector with a smooth response, Gamma(P) is linear at sufficiently low power; the reported powers are not in a single-photon regime. Example: at 675 nm, 0.1 nW corresponds to about 3.4e8 photons/s; with freset = 5.6 kHz this is about 6e4 absorbed photons per reset interval. Linearity in this regime cannot distinguish one-photon triggering from a linearized multi-photon or hot-electron response. Poissonian counts in 235 ms bins are expected for any rare independent Bernoulli switching process, regardless of whether a click requires one, two, or many photons; variance=mean therefore has no discriminating power. The paper itself states that the microscopic switching mechanism is an open question and that gain/carrier multiplication is speculative, while the mid-IR case has demonstrably non-Poissonian statistics (Supplementary Section 5). Without an absolute calibration connecting click probability to photon flux, or a measurement pushed to mean photon number per reset below one, the data support a highly photosensitive bistable switch but not demonstrably single-photon counting.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports high-bias transport in bilayer graphene aligned to hexagonal boron nitride, showing negative differential conductance (NDC) that is spatially localized within the device. The NDC is attributed to Esaki-Tsu negative differential velocity rather than electric-field localization, based on the displacement-field (bandwidth) dependence of the critical voltage. Using a series load resistor to engineer bistability near the critical current, the authors observe stochastic switching events whose rate increases under mid-infrared (11.3 μm) and visible (675 nm) illumination, with linear power dependence near the critical bias and Poissonian click statistics in the visible. The paper claims single-photon counting at these wavelengths up to 25 K. The switching mechanism is argued to involve a structurally determined stacking domain whose gap at the secondary Dirac point suppresses Schwinger/Zener transitions, allowing carriers to reach negative-velocity states.","tokens_in":30209,"tokens_out":9603,"duration_ms":89199,"significance":"If the single-photon claim were established, this would be a substantial advance: a non-superconducting, bistable-switching detector operating at 25 K and 11.3 μm would fill a genuine gap left by SNSPDs and SPADs, and the use of gate-tunable moiré band structure to engineer the bistability is conceptually new. The paper has real strengths independent of that headline: the observation and spatial localization of NDC in BLG/hBN, the in-situ test separating Esaki-Tsu NDV from sequential tunneling via the displacement-field dependence of V_C (Fig. 4A), the two-region model explaining the narrow NDC window, and a thorough supplementary set including self-gating corrections, heating diagnostics, RCWA absorption estimates, and kinetic-equation modeling that captures the filling-factor dependence of I_C.","major_comments":[{"comment":"The evidence presented does not establish single-photon triggering. (i) The linear power dependence Γ ∝ P with x = 1 (Fig. 2D) is expected from any detector with a smooth response at sufficiently low power, and the reported fluxes are far above the single-photon level: at 675 nm and 0.1 nW the incident flux is about 3.4×10^8 photons/s, i.e. about 6×10^4 photons per reset interval at f_reset = 5.6 kHz, and even after the 2% absorption correction quoted in the Fig. 3B caption this leaves roughly 10^3 absorbed photons per reset interval; the mid-IR data at 4.8 nW with ~4% absorption are further still from the single-photon regime. Any detector with a smooth monotonic response will show Γ linear in P at such fluxes, so the x = 1 exponent cannot distinguish one-photon triggering from a linearized multi-photon, hot-electron, or local-bolometric response. (ii) Poissonian histograms with variance = mean (Fig. 3C,D) are the generic statistics of any rare, independent Bernoulli switching process counted in fixed time bins, independent of whether a click requires one, two, or many photons; they therefore have no discriminating power for the single-photon question. (iii) Supplementary Section 5 itself reports non-Poissonian statistics for the mid-IR, so the abstract's mid-IR claim rests only on the linear power dependence. The manuscript's own statements that the microscopic switching mechanism and the required gain or carrier multiplication are open questions (main text, 'Single-photon detection mechanism'; Supplementary Section 16) reinforce that the single-photon level is asserted, not demonstrated. To support the headline claim, the authors should provide an attenuation series extending to mean photon number per reset below unity together with an absolute detection-efficiency calibration, or a direct comparison against an independent single-photon detector, or they should substantially temper the claim in the title, abstract, and discussion.","section":"Main text, Figs. 2D and 3C–D; Supplementary Section 5"},{"comment":"The text states that experiment and calculation agree 'without any fitting parameters' and 'within an order of magnitude,' but the Fig. 4C caption says the experimental data were 'scaled by a factor 3,' and Supplementary Section 7 states that the kinetic-equation model yields IV characteristics only 'up to an unknown constant in the electric field strength.' With an explicit scale factor of 3 and an unknown field constant, the comparison is not parameter-free as claimed. The more robust point is the filling-factor dependence peaking near ν = +2, which does not depend on these constants; please present the comparison with both scalings stated explicitly, or remove the parameter-free wording.","section":"Origin of NDC, Fig. 4C"},{"comment":"The attribution of NDC to Esaki-Tsu negative differential velocity rests on the premise that the photoactive domain has a stacking configuration opening a gap at the secondary Dirac point large enough to suppress Schwinger/Zener transitions. This premise is inferred from spatial maps (Fig. 3A), local 4-probe NDR, and tight-binding calculations showing the gap appears 'for only a few stacking orders' (Supplementary Section 8), but the stacking order and the gap of the actual photoactive region are not directly measured. Because the NDV mechanism conclusion depends on this unverified premise, the main text should state plainly that it is inferred rather than measured, or support it with a direct local probe; the current wording ('Our analysis suggests...') is weaker than the strength of the evidence, and the referee agrees this is a genuine unresolved point rather than a mere stylistic one.","section":"Origin of NDC and Supplementary Section 8"},{"comment":"The quoted absorbed-power fractions are internally inconsistent: the Fig. 2D caption states absorbed power is calculated 'to be 4%,' the Fig. 3B caption states '2% at λ = 675 nm,' Supplementary Section 19 reports α ≈ 2% at 675 nm and ≈3% at 11.3 μm, and the Fig. S10 caption states '~4%.' Since the absolute absorbed photon flux is the central quantity for any single-photon calibration (and for the per-reset photon numbers invoked in the main text), these numbers must be reconciled and a single consistent set of values, with uncertainties, should be quoted.","section":"Figs. 2–3 captions and Supplementary Sections 19, 11"}],"minor_comments":[{"comment":"The click threshold is defined as 0.02 V below the median of the differentiated waveform; please provide a robustness check showing that Γ is insensitive to this threshold choice over a reasonable range, since the threshold is a free parameter of the counting procedure.","section":"Methods, click-counting scheme"},{"comment":"The caption states that the doping level in (E) and (F) was 'slightly different' from that in (C); given the extremely steep dependence of Γ on bias, please clarify how comparisons across these panels are normed.","section":"Fig. 2E–F caption"},{"comment":"The abstract claims operation 'up to 25 K,' while the Fig. 5B text says detector performance is unchanged 'up to 20 K' and the Fig. 5A caption places NDC survival 'around 30 K'; please reconcile these numbers.","section":"Abstract and Fig. 5"},{"comment":"There is a typo: 'All measured were performed at 4 K except E' should read 'All measurements were performed at 4 K except E.'","section":"Fig. 2 caption"},{"comment":"In the two-region model, the fraction of the NDC region l2/(l1+l2) is a free parameter; please state the value used for the black curve in Fig. 4B and how it was chosen.","section":"Supplementary Section 13, Fig. 4B"},{"comment":"The threshold condition for electric-field localization is quoted as V_SL/a > W, but the proportionality constant between the measured critical voltage V_C and the bandwidth W used for the red dashed line in Fig. 4A is not stated; please give the assumed field distribution and constant so that the comparison is reproducible.","section":"Fig. 4A and Supplementary Section 10"},{"comment":"References 3 and 46 contain malformed titles and dates ('Nature Reviews Physics 2021 4:3 4, 194–208' and 'Sci Adv 5, eaay8897 (2024)'); these should be cleaned up before publication.","section":"References 3 and 46"}],"recommendation":"major_revision","confidential_remarks":"The abstract and title assert a single-photon demonstration that the paper's own evidence and admissions do not support. The missing element—an absolute calibration or an attenuation series reaching mean photon number per reset below unity—is precisely the hard part of any single-photon claim, so the authors should be asked either to supply it or to reframe the claim as 'highly sensitive bistable photodetection with linear power dependence consistent with, but not proving, single-photon sensitivity.' The NDC/NDV transport story, including the bandwidth-dependence discrimination test, is independently solid and would survive such a reframing. I also note that the authors cite resonant-tunneling-diode single-photon counting (ref. 48) without explaining how their evidence standard differs from that prior work; the reader may reasonably ask what distinguishes the present Poissonian-statistics argument from claims already made for RTDs."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one. The interesting part is not the headline claim, it's the high-field transport: they show NDC in BLG/hBN moiré devices, tie it to Esaki-Tsu negative differential velocity rather than electric-field localization via the bandwidth dependence, and support it with two devices, local four-probe measurements, spatial maps, and a Boltzmann calculation that captures the doping dependence of the critical current within an order of magnitude (after a factor-3 scaling). The bistable switching is real, and the temperature/power discrimination (Fig. 2E/F, S16) is a good-faith attempt to rule out bolometric heating. For a proof-of-concept, that is a solid physical result.\n\nThe soft spot is exactly where the stress-test points: 'single-photon counting' is not established. Linear Γ(P) is expected for any smooth detector at sufficiently low power; 0.1 nW at 675 nm is about 3e8 photons/s, orders of magnitude above one photon per reset interval. Poissonian histograms in 235 ms bins mean variance=mean, but that holds for any rare independent switching process, regardless of whether a click needs one, two, or many photons. The mid-IR statistics are admittedly non-Poissonian because of background. So the data support a highly photosensitive bistable switch, not demonstrably single-photon detection. That is a large gap relative to the abstract, and it is fixable with attenuation calibration, a regime with mean photon number per reset well below one, or an absolute efficiency measurement.\n\nOther soft spots: the NDC mechanism leans on the authors' own tight-binding model for the bandwidth W(D), and the factor-3 scaling in the critical-current comparison is not fully explained. The gapped structural domain that suppresses Schwinger tunneling is inferred from spatial maps and stacking statistics, not directly measured. The paper is candid about this and explicitly leaves the microscopic switching mechanism open, which I respect. The citation pattern is fair; Esaki-Tsu, Berdyugin, and SNSPD literature are properly credited.\n\nWho is this for? Anyone working on high-field transport in moiré systems or exploring NDC-based photodetectors. It deserves a serious referee, but the single-photon claim needs to be either calibrated properly or downgraded to 'highly sensitive bistable photodetection with single-photon potential.' I'd send it out, and I'd push for the calibration experiment in revision.","headline":"A genuinely new NDC-based bistable photodetector with rich transport data, but the single-photon headline is not yet supported by the reported calibration.","tokens_in":30927,"tokens_out":2266,"would_cite":true,"duration_ms":23320,"reading_group":"yes","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 claims that a bilayer-graphene/hBN moiré device counts individual photons at 11.3 μm and 675 nm up to 25 K by exploiting negative differential conductance, which it attributes to negative differential velocity in a folded…","keywords":["single-photon detection","moiré superlattice","negative differential conductance","negative differential velocity","bilayer graphene on hBN","bistable switching","mid-infrared photodetection","non-equilibrium electron phases"],"falsifier":"Perform scanning tunnelling spectroscopy directly on the narrow photoactive strip identified in spatial maps: if no band gap appears at the secondary Dirac point, or if the strip moves with gate voltage or bias, the negative-velocity mechanism and the structural-domain explanation would be ruled out.","tokens_in":29743,"feed_emoji":"📡","tokens_out":5432,"duration_ms":53498,"temperature":0.7,"pith_summary":"The paper reports a single-photon detector built from a bilayer-graphene/hexagonal-boron-nitride moiré superlattice. At high bias the device enters a negative differential conductance regime, and with a series resistor this becomes a bistable switch that a photon can trigger. The authors demonstrate counting at 11.3 μm and 675 nm up to 25 K, with Poissonian click statistics for visible light. They argue the negative differential conductance comes from carriers accelerated into negative-velocity parts of the Brillouin zone, not from electric-field localization, because the critical voltage stays nearly constant as the miniband width is tuned with displacement field. A sympathetic reading is that this establishes a non-superconducting mechanism for single-photon detection in a wavelength and temperature range where existing detectors struggle.","feed_headline":"Moiré device counts single photons at 25 K and 11.3 μm","feed_subtitle":"Negative differential conductance turns each absorbed mid-infrared or visible photon into a click, without nanowires or millikelvin cooling.","key_machinery":"The central object is negative differential velocity (NDV), the Esaki-Tsu mechanism in a strongly coupled superlattice: a DC field accelerates carriers past a band inflection where the effective mass changes sign, so the drift velocity decreases with increasing field. Two ingredients make NDV observable here: a displacement-field-opened gap at the secondary Dirac point that blocks Schwinger/Zener breakdown, and a structurally determined stacking domain, inferred from spatial maps and contact-dependent transport, that isolates the lowest miniband. A 200 kOhm series resistor converts the resulting negative differential conductance into a bistable current-voltage loop; photons tip the bistable switch. A generalized two-dimensional kinetic Boltzmann equation on the tight-binding band structure supplies the calculated current-field curve and critical-current-versus-doping dependence.","core_discovery":"The central claim is that a single-photon click can be produced by a non-equilibrium electron phase in a moiré superlattice without superconductivity, and that the detection wavelength is set by the broad infrared absorption of bilayer graphene combined with a field-induced bistability. Concretely, the paper shows count rates that rise linearly with absorbed power (exponent x = 1), Poissonian switching statistics at 675 nm, and a power dependence qualitatively different from temperature-induced dark counts, which rules out laser heating as the cause of the light-driven clicks. It attributes the negative differential conductance to negative differential velocity in an isolated miniband: a rare stacking domain with a gap at the secondary Dirac point suppresses Schwinger pair creation, letting carriers accelerate into negative-velocity states. Band-structure calculations and a two-region circuit model reproduce the measured critical currents within an order of magnitude, and the mechanism is reported to persist to roughly 30 K.","pith_inferences":["If the negative-velocity interpretation is right, Bloch oscillations in a two-dimensional solid become a plausible next step: the same isolated-miniband condition that produces NDV is the condition for coherent acceleration across the Brillouin zone.","Because only a small, rare stacking domain appears to work, deterministic control of local twist or stacking order, or deliberate supermoiré patterning, could convert this proof-of-principle into a detector with a much larger active area.","The bias-dependent count-rate exponent suggests a tunable threshold detector: biased close to the critical current it registers single photons, while biased farther away it registers multi-photon events, which could be useful for photon-number-resolving applications.","The two-regime Arrhenius analysis of dark counts implies two distinct escape barriers; if those barriers also control photon-triggered switching, engineering the potential landscape of the bistable state may reduce noise more effectively than simply cooling further."],"forward_implications":["Mid-infrared single-photon counting at 11.3 μm becomes possible without superconducting nanowires or millikelvin cooling; the paper reports detection up to about 25 K and negative differential conductance up to about 30 K.","The same mechanism, tuned through miniband engineering and displacement fields, should transfer to other moiré and van der Waals superlattice systems; the paper reports bistabilities in magic-angle twisted bilayer graphene as supporting evidence.","Reset rates above 38 kHz are limited by the measurement circuit, not the device, so dedicated electronics and fiber-coupling should push speed and efficiency beyond the demonstrated values.","Because the count rate grows exponentially with bias and linearly with power near threshold, detector efficiency and dark count rate can be traded against each other by choosing the operating point.","The near-constant critical voltage under displacement-field tuning separates the negative-velocity mechanism from electric-field localization, giving a practical design rule for future NDC-based detectors."],"supporting_citations":[{"why":"Establishes the out-of-equilibrium criticalities (Schwinger pair production) in graphene moiré superlattices that the present device must suppress, and supplies the ring-like dI/dV structure.","marker":"[26]"},{"why":"Defines the competing mechanism of electric-field localization in superlattices that the paper argues against.","marker":"[38]"},{"why":"Introduces the Esaki-Tsu negative differential velocity mechanism that the paper invokes to explain the NDC.","marker":"[39]"},{"why":"Supplies the discriminating test: critical field independent of bandwidth favours NDV over localization.","marker":"[40]"},{"why":"Supplies the displacement-field-dependent bandwidth and gap calculations used to tune and interpret the experiment.","marker":"[41]"},{"why":"Underlies the broadband long-wavelength absorption of bilayer graphene that gives the detector its mid-IR reach.","marker":"[47]"},{"why":"Provides the Geiger-mode resonant-tunnelling diode precedent for photon-triggered switching in NDC devices.","marker":"[48]"},{"why":"Provides the statistical benchmark (Poissonian count distribution) used to establish single-photon detection at visible wavelengths.","marker":"[19]"},{"why":"Supplies the metrology context for correlating detector statistics with source statistics as proof of single-photon sensitivity.","marker":"[35]"}],"fun_headline_variants":["Moiré detector counts single photons at 25 K","Negative differential conductance enables single-photon clicks","Infrared single-photon detection without superconducting nanowires","Bistable moiré phase detects single mid-IR photons","Single-photon counting in moiré superlattices to 25 K"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole interpretation rests on a localized, structurally determined domain whose rare stacking order opens a gap at the secondary Dirac point, allowing carriers to reach negative-velocity states; the stacking and the gap are inferred from transport maps and calculations, not directly measured.","fun_headline_variants_meta":{"raw":{"variants":["Moiré detector counts single photons at 25 K","Negative differential conductance enables single-photon clicks","Infrared single-photon detection without superconducting nanowires","Bistable moiré phase detects single mid-IR photons","Single-photon counting in moiré superlattices to 25 K"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000828,"raw_usage":{"total_tokens":3596,"prompt_tokens":898,"completion_tokens":2698,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":514,"completion_tokens_details":{"reasoning_tokens":2616}},"tokens_in":514,"tokens_out":2698,"duration_ms":20114,"temperature":1.0,"reasoning_tokens":2616,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:13:04.570285+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform scanning tunnelling spectroscopy directly on the narrow photoactive strip identified in spatial maps: if no band gap appears at the secondary Dirac point, or if the strip moves with gate voltage or bias, the negative-velocity mechanism and the structural-domain explanation would be ruled out.","supporting_citations":[{"cited_title":"& Peeters, F","cited_arxiv_id":null,"evidence_quote":"Provides the statistical benchmark (Poissonian count distribution) used to establish single-photon detection at visible wavelengths."}],"review_version":1}