{"id":"1c135e9e-7a63-4e0a-9cfb-e51658e142aa","arxiv_id":"2412.09735","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A redesigned HgCdTe LmAPD array reduces ROIC glow by 7x, reaches a dark current below 0.1 e-/pixel/kilosecond, and shows individual photon jumps at high gain.","lead":"Researchers tested a new version of an infrared avalanche photodiode array and found that its readout glow dropped by about a factor of seven, to 0.012 electrons per pixel per frame. They also measured a near zero dark current and showed signs that the device can detect individual photons at high gain.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Photon-counting claim lacks calibrated verification: step-like signals are not shown to scale with photon flux, and the assertion that glow/dark are unamplified is unquantified.","rationale":"The reader's weakest-assumption analysis correctly identifies the photon-jump identification as the principal gap. I agree that the 4 V glow and dark-current measurements are experimentally sound: the two-dataset subtraction and the 20-hour integration are appropriate, and the glow reduction from 0.08 to 0.012 e-/pixel/frame is a clear result. The single-photon demonstration, however, is the part that would justify the 'towards photon counting' framing and the HWO relevance, and it currently lacks a control: no flux calibration, no expected photon rate, no false-positive rate, and no quantitative mask comparison. The proposed test, varying the calibrated source flux and checking that the jump rate follows it, directly tests whether the jumps are photon-induced. Should the test fail, the paper would remain a solid glow/dark-current characterization but the photon-counting claim would need to be softened. Since the reader's CONDITIONAL already reflects this, the verdict is unchanged.","tokens_in":7927,"tokens_out":6291,"duration_ms":66678,"concrete_test":"Run a 6000-frame up-the-ramp sequence at 12 V with the source set to two measured attenuation levels and then fully blocked, using the same changepoint detector on a fixed set of pixels. If the jump rate scales linearly with the calibrated flux and falls to the dark rate when blocked (and if the jump-amplitude histogram exhibits a peak at the gain-calibrated single-photon signal), the photon-counting claim is supported; if the rate does not scale or the amplitudes are not quantized, the claim is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the detector can count individual photons rests on interpreting the step-like signals in Section 4.1 as single photoelectrons. This interpretation is not independently verified: no calibrated photon flux is used, no detection efficiency is given, no false-positive analysis of the changepoint algorithm is presented, and no amplitude histogram shows a resolved single-photon peak. The mask comparison in Section 4.2 is only qualitative ('few jumps', 'smaller amplitudes') and does not report the actual counts or the predicted glow-event rate. The related assertion that 'glow and dark-current do not fully experience the amplification process' is assumed rather than demonstrated. If dark current at 12 V were fully amplified, the masked pixels would exhibit a jump rate comparable to illuminated pixels, and the jumps would not be uniquely attributable to photons. The glow/dark-current measurements at 4 V in Section 3 are plausible and well presented, but they do not constrain the 12 V jump behavior.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the characterization of a redesigned linear-mode avalanche photodiode (LmAPD) array with reduced readout integrated circuit (ROIC) glow. Using two matched Fowler-sampled datasets, one with drop frames and one without, the authors separate the ROIC glow from the intrinsic dark current and report a glow level of 0.012 e-/pixel/frame and an intrinsic dark current of 0.07 e-/pixel/kilosecond, about seven times lower than their previous chip. In a separate high-bias-voltage test at 12 V, the authors claim individual photon detection, identifying step-like jumps in light curves with a changepoint detection algorithm. They further argue, based on a qualitative comparison between masked and illuminated pixels, that glow photons are only partially amplified and can therefore be distinguished from signal photons.","tokens_in":8037,"tokens_out":2168,"duration_ms":23296,"significance":"If the photon-counting claim holds, this would be the first demonstration of single-photon detection in a mature, large-format near-infrared imaging array, with direct relevance to future space missions such as the Habitable World Observatory. The glow and dark-current measurements are well designed and credible: the matched datasets with and without drop frames cleanly separate the two contributions, and the 20-hour dark-current test is consistent with a null intrinsic dark current. The reported glow reduction from 0.08 to 0.012 e-/pixel/frame is itself a useful engineering milestone. However, the photon-counting evidence as presented is incomplete and requires additional calibration and statistical analysis before the central claim can be accepted.","major_comments":[{"comment":"The photon-counting claim rests on the identification of step-like jumps as individual photoelectrons, but the manuscript provides no calibrated photon flux, no detection efficiency estimate, no amplitude histogram, and no false-positive analysis for the changepoint algorithm. Without such verification, the jumps could be due to random telegraph noise, single-event upsets, or other non-photon artifacts. Please provide a quantitative validation, for example by varying the incident flux and showing that the jump rate scales linearly, or by demonstrating a resolved single-photon amplitude peak.","section":"Section 4.1"},{"comment":"The mask comparison used to argue that glow photons are only partially amplified is purely qualitative: the text reports 'very few jump detections' and 'smaller amplitudes' without giving the actual counts in masked versus illuminated pixels, the expected number of glow events based on the measured glow rate, or the statistical uncertainties. This is load-bearing for the photon-counting interpretation because it is the only evidence that glow events do not appear as full-amplitude jumps. Please report the jump rates and amplitude distributions in both regions and compare them with a quantitative prediction.","section":"Section 4.2"},{"comment":"The conversion gain reported in Table 1 (1.54 ± 0.01 e-/ADU) is measured at 4 V, but the photon-counting tests are performed at 12 V and no conversion gain for this operating condition is given. Without a high-bias conversion gain, the jump amplitudes cannot be expressed in electrons, and the claim that photoelectron signals are 'notably above the read noise level' cannot be checked quantitatively. Please provide the conversion gain at 12 V or otherwise specify how the jump amplitudes in electrons were obtained.","section":"Table 1 and Section 4.1"}],"minor_comments":[{"comment":"Typographical error: 'essentally' should be 'essentially'.","section":"Abstract"},{"comment":"Typographical error: 'Righ' should be 'Right'.","section":"Figure 5 caption"},{"comment":"The phrase 'Detectors with this ability have been know for decades' contains a spelling error: 'know' should be 'known'.","section":"Section 4.1"},{"comment":"The sentence 'We believe this approach can be a new way to measure many parameters...' is speculative and does not directly support the central claim; consider moving it to the future-work section.","section":"Section 4.1"}],"recommendation":"major_revision","confidential_remarks":"The glow and dark-current results appear solid and timely, and the paper clearly improves on the prior C22 characterization. The main risk is the photon-counting claim, which is central to the abstract and conclusions but currently lacks the calibration and statistical support expected for a claim of first demonstration. If the authors can supply the requested quantitative validation without a major reworking of the experiment, the paper would be suitable for acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the glow and dark-current measurements are credible and well-designed. The two Fowler-sampled datasets, one with drop frames and one without, cleanly separate ROIC glow from intrinsic dark current. The measured glow of 0.012 e-/pixel/frame, about 7x lower than their C22 result, and the 20-hour dark-current measurement consistent with null (<0.1 e-/pixel/kilosecond) are solid, reproducible results. That part of the paper holds up and is a genuine step toward the detector requirements for HWO-class instruments.\n\nSecond, the single-photon detection claim in Section 4 is suggestive but not yet quantitatively nailed. The changepoint algorithm is a reasonable tool, but there is no calibrated photon flux, no detection efficiency, no false-positive analysis, and no amplitude histogram showing a resolved single-photon peak. The mask comparison in Section 4.2 is qualitative--\"few jumps\" and \"smaller amplitudes\"--without actual counts or a predicted glow-event rate. The stress-test concern lands here. The assertion that glow and dark-current do not fully experience the avalanche amplification is assumed rather than demonstrated. This does not sink the paper, because the glow/dark measurement stands on its own, but it means the photon-counting result should be read as preliminary.\n\nWhat is actually new: the redesigned pixel with an extra metal layer to block the JFET leakage path, and the measured 7x glow reduction. The 20-hour dark-current bound is also a useful data point. The photon-counting demonstration, if confirmed, would be a first for a mature near-infrared array, but the evidence is not yet there to claim it as established.\n\nMinor soft spots: the conversion gain is measured at 4V but not stated at 12V, where the jumps are observed. The statement that \"most (~80%) pixels visually display photon detection capabilities\" is vague and unquantified. The citations to their own C22 work are legitimate prior characterization, not circular.\n\nBottom line: this deserves a serious referee. The glow/dark-current results are publishable now; the photon-counting claim needs additional verification--a flux scaling test, a false-positive rate, and a quantitative mask comparison--before it can be accepted. I would send it to review, and the review should push on those points.","headline":"Solid glow and dark-current measurements on a redesigned LmAPD; the photon-counting claim is plausible but under-verified.","tokens_in":8629,"tokens_out":1864,"would_cite":true,"duration_ms":20651,"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":"A redesigned infrared avalanche photodiode array cuts its self-generated glow sevenfold and detects individual photons at high gain.","keywords":["infrared detectors","avalanche photodiodes","photon counting","dark current","ROIC glow","HgCdTe","high-contrast imaging","LmAPD arrays"],"falsifier":"Expose the array to a calibrated, steadily attenuated light source and record light curves; if the jumps are photons, the detected jump rate must rise linearly with the input flux and the per-jump amplitude must remain constant. If the jump rate stays flat while flux changes, or if jump amplitudes scale with something other than the photoelectron signal, the photon interpretation fails.","tokens_in":7695,"feed_emoji":"🔭","tokens_out":4940,"duration_ms":46087,"temperature":0.7,"pith_summary":"This paper reports that a redesigned linear-mode avalanche photodiode (LmAPD) array, intended for ultra-low-background infrared astronomy, has its readout glow reduced to 0.012 e-/pixel/frame, about seven times lower than the previous chip. With that glow removed, the intrinsic dark current is below 0.1 e-/pixel/kilosecond at 50 K, effectively indistinguishable from zero over normal exposure times. At a higher bias voltage, the array produces light curves with discrete upward jumps that the authors identify as individual photoelectrons, which would make it the first mature near-infrared imaging array capable of photon counting in a low-noise regime. The result matters because future missions to image Earth-like exoplanets require detectors with dark current below 1 e-/pixel/kilosecond and read noise below 1 e-/pixel/frame at infrared wavelengths, where no flight-qualified technology currently meets both.","feed_headline":"Infrared chip cuts glow 7x, counts single photons","feed_subtitle":"Redesigned LmAPD array reaches dark current below 0.1 e-/pixel/kilosecond, a step toward infrared exoplanet imaging.","key_machinery":"The device is a HgCdTe linear-mode avalanche photodiode array grown by MOVPE, in which photoelectrons created in a P-type absorber diffuse through a p-n junction and are multiplied in a high-field region before being collected by a readout integrated circuit (ROIC). The specific fix at the pixel level is an added metal layer that blocks the suspected glow path: light emitted by a source-follower JFET leaking through a gap in the metal architecture onto the photosensitive region. Measurements separate glow from dark current by comparing two datasets that read the array in a Fowler pattern, one clocked with drop frames and one left unclocked, so that glow appears only in the first. For the photon-counting demonstration, the load-bearing mechanism is the unusually low excess noise factor of HgCdTe avalanche multiplication, which lets a single photoelectron's amplified signal exceed the read noise; a changepoint algorithm then finds the jumps in each pixel's light curve and fits their amplitudes.","core_discovery":"The central claim is that the glow produced by the readout integrated circuit when the array is clocked can be blocked at the pixel level, and that doing so exposes an essentially zero intrinsic dark current and enables single-photon detection. The paper measures a per-frame glow of $0.012 \\pm 0.001$ e-/pixel/frame, a factor of about seven below the $0.08$ e-/pixel/frame reported for the previous chip, and a dark current of $0.07 \\pm 0.03$ e-/pixel/kilosecond over a 20-hour test. At 12 V bias, where the effective read noise is below 1 e-/pixel/frame, the authors use a changepoint detection routine on up-the-ramp light curves to identify discrete jumps that they attribute to individual photoelectrons; roughly 80% of pixels visually show such photon-detection behavior, and the few jumps detected under a dark mask are smaller and about an order of magnitude rarer than expected for fully amplified glow. The implication, stated by the authors, is that glow photons may be only partially amplified and therefore distinguishable from signal photons.","pith_inferences":["A direct test of the photon interpretation would be to measure the jump rate while varying a calibrated input flux: if the jumps are single photoelectrons, the rate should scale linearly with flux and the amplitude distribution should be independent of flux.","The same light-curve jump analysis could be turned into a general-purpose detector characterization tool that extracts gain and excess noise factor per pixel, something the paper suggests but does not implement.","If the partially amplified glow hypothesis is correct, glow suppression may be achievable by engineering the electric field profile so that parasitic photons are absorbed after the multiplication region, not just by blocking their emission."],"forward_implications":["If the measured glow and dark current hold at array scale, LmAPD arrays meet the dark-current and read-noise budget that has been identified for coronagraphic exoplanet imaging in the near-infrared.","Because the glow is per-frame rather than per-second, high-frequency up-the-ramp readout suppresses its accumulated contribution, making fast sampling the natural operating mode for ultra-low-background observations.","The observation that glow-induced jumps under the mask are rarer and smaller than signal jumps implies that residual glow may not fully enter the multiplication region, so it may contribute less effective noise than its raw count suggests.","If single-photon jumps can be counted reliably, the same changepoint measurements provide a new way to measure gain, excess noise factor, glow, and dark current from the quiet portions and jump amplitudes of individual light curves."],"supporting_citations":[{"why":"Supplies the previous chip's glow and dark-current baseline that the new measurements are compared against.","marker":"[10]"},{"why":"Establishes that apparent zero dark current in infrared arrays is often multiplexer glow, motivating the separation performed here.","marker":"[5]"},{"why":"Documents prior performance of SAPHIRA HgCdTe APD arrays, the detector lineage this work extends.","marker":"[7]"},{"why":"Provides an overview of SAPHIRA detector capabilities and operating modes relevant to the low-flux regime.","marker":"[8]"},{"why":"Describes the pixel-level architecture and MOVPE growth of the LmAPD arrays used in the measurements.","marker":"[11]"},{"why":"Describes the SIDECAR ASIC used to operate the detector and digitize its output in the test setup.","marker":"[13]"},{"why":"Reports ROIC glow reduction in other short-wave infrared focal plane arrays, providing the context for the glow-blocking fix.","marker":"[14]"}],"fun_headline_variants":["LmAPD glow cut 7x, enabling photon counting in infrared","Readout glow slashed by 7x, LmAPDs count single photons","Infrared array tames readout glow, detects single photons","Sub-electron dark current in LmAPDs after glow reduction","Glow reduction in LmAPDs paves way for photon counting infrared"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's photon-counting claim rests on the assumption that the discrete jumps it detects in the light curves are single photoelectrons, not random telegraph noise, single-event upsets, or partially amplified glow photons.","fun_headline_variants_meta":{"raw":{"variants":["LmAPD glow cut 7x, enabling photon counting in infrared","Readout glow slashed by 7x, LmAPDs count single photons","Infrared array tames readout glow, detects single photons","Sub-electron dark current in LmAPDs after glow reduction","Glow reduction in LmAPDs paves way for photon counting infrared"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000367,"raw_usage":{"total_tokens":1990,"prompt_tokens":979,"completion_tokens":1011,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":595,"completion_tokens_details":{"reasoning_tokens":914}},"tokens_in":595,"tokens_out":1011,"duration_ms":9752,"temperature":1.0,"reasoning_tokens":914,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T16:46:43.279764+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Expose the array to a calibrated, steadily attenuated light source and record light curves; if the jumps are photons, the detected jump rate must rise linearly with the input flux and the per-jump amplitude must remain constant. If the jump rate stays flat while flux changes, or if jump amplitudes scale with something other than the photoelectron signal, the photon interpretation fails.","supporting_citations":[{"cited_title":"First tests of a 1 megapixel near-infrared avalanche photodiode array for ultra-low background space astronomy,","cited_arxiv_id":null,"evidence_quote":"Supplies the previous chip's glow and dark-current baseline that the new measurements are compared against."},{"cited_title":"Zero dark current in H2RG detectors: it is all multiplexer glow,","cited_arxiv_id":null,"evidence_quote":"Establishes that apparent zero dark current in infrared arrays is often multiplexer glow, motivating the separation performed here."},{"cited_title":"Next-generation performance of SAPHIRA HgCdTe APDs,","cited_arxiv_id":null,"evidence_quote":"Documents prior performance of SAPHIRA HgCdTe APD arrays, the detector lineage this work extends."},{"cited_title":"Overview of the SAPHIRA detector for adaptive optics applications,","cited_arxiv_id":null,"evidence_quote":"Provides an overview of SAPHIRA detector capabilities and operating modes relevant to the low-flux regime."},{"cited_title":"Linear-mode avalanche photodiode arrays in HgCdTe at Leonardo UK,","cited_arxiv_id":null,"evidence_quote":"Describes the pixel-level architecture and MOVPE growth of the LmAPD arrays used in the measurements."},{"cited_title":"SIDECAR low-power control ASIC for focal plane arrays including A/D conversion and bias generation,","cited_arxiv_id":null,"evidence_quote":"Describes the SIDECAR ASIC used to operate the detector and digitize its output in the test setup."},{"cited_title":"ROIC glow reduction in very low flux short wave infra-red focal plane arrays for astronomy,","cited_arxiv_id":null,"evidence_quote":"Reports ROIC glow reduction in other short-wave infrared focal plane arrays, providing the context for the glow-blocking fix."}],"review_version":1}