{"id":"7003c055-bb4b-44a8-80a4-f8a2f3f33a82","arxiv_id":"1908.10520","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"A 32 by 32 array of superconducting nanowire single-photon detectors was built and operated, achieving a 99.4% baseline yield, up to 8% system detection efficiency at 1550 nm, and 250 to 400 ps timing jitter.","lead":"This paper demonstrates a camera-like detector made of 1,024 superconducting nanowires that can record individual photons. The result is the largest such single-photon detector array reported at the time, a step toward low-light imaging in astronomy, quantum optics, and lidar.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pixel-level claims rely on an unvalidated coincidence-pairing algorithm; a ground-truth multi-spot test would settle whether the reported yield, SDE, and images are accurate.","rationale":"I agree with the reader that the coincidence-pairing analysis is the weakest assumption supporting the most quantitative parts of the central claim. The row-column multiplexing scheme is inherently ambiguous for multi-photon events within the timing resolution, and the paper explicitly acknowledges this in Sections 2 and 5.3. The reported images and yield/SDE figures all pass through the sequential time-tag pairing algorithm, whose false-positive rate is not measured against a known illumination pattern. This is a genuine soft spot. However, I do not think it overturns the verdict of ACCEPT. The paper is framed as a proof of concept, the array is physically fabricated with 1024 pixels, the row and column readout channels show photoresponse, and the persistence text images provide qualitative evidence that pixel localization works at low count rates. The main open question is quantitative accuracy, not the existence of the kilopixel device or the general feasibility of the architecture. A single calibration-target experiment, as described in the concrete test, would settle whether the coincidence algorithm introduces material misattribution at the demonstrated rates. Until that test is run, the strongest defensible position is unchanged acceptance with the noted caveat, not rejection or unverdictability.","tokens_in":8031,"tokens_out":10303,"duration_ms":119240,"concrete_test":"Use a calibrated mask or a pair of independently attenuated 1550 nm beams to illuminate known pixels (e.g., two spots separated by several pixels) at total count rates matching the SDE/jitter measurements (roughly 0.3-1 Mcps) and at the high-rate limit (10-30 Mcps per row). Acquire time tags with the same TDC settings and run the exact sequential-pairing analysis. Compare reconstructed pixel maps to the known mask, reporting per-pixel true-positive rate, false-positive rate, and the fraction of counts assigned to non-illuminated pixels. If the false/streak fraction is below about 1% at the imaging rates and below about 5% at the claimed maximum rate, the central imaging claim is confirmed; otherwise the yield and SDE numbers must be corrected and the 'kilopixel imaging' claim revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Every pixel-level result—the 99.4% yield, the 4%/8% system detection efficiency, the hot/dead pixel inventory, and the images—depends on the Section 2/4 analysis that pairs row and column time tags sequentially within a coincidence window. The paper explicitly concedes (Section 2, Section 5.3) that two events with distinct rows and columns within the timing resolution are unassignable, and that at 320 Mcps a 6 ns window puts roughly 30% of detected events into misattributed pairs. However, no ground-truth test is reported for the actual imaging conditions. The only quantitative bound (at most 0.6% of hot pixels’ counts, Section 5.2) is an estimate of false counts contributed from hot pixels, not a measured true-positive/false-positive rate for the sequential pairing algorithm. Moreover, the maximum-count-rate measurement (Section 5.3) explicitly uses totalized row counts rather than coincidences, so it does not validate pixel-resolved imaging at high rates. If the pairing algorithm misassigns even a few percent of events at the rates used for the SDE and jitter measurements, the reported yield and efficiency figures, and the fidelity of the claimed kilopixel imager, would be weaker than stated. This is the weakest load-bearing step in the argument; the fabrication and row/column electrical evidence for a 32x32 structure is otherwise solid.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports the fabrication and characterization of a 32 x 32 (1024-element) superconducting nanowire single-photon detector (SNSPD) array using row-column multiplexing. The array covers a 1.6 x 1.6 mm area with a 50 um pixel pitch. Using a 64-channel time-tagging readout, the authors measure a baseline yield of 99.4%, a system detection efficiency of up to 8% at 1550 nm, and average pixel jitter of 250-400 ps. They demonstrate imaging by sweeping a focused laser spot to produce persistence images, including spelling text, and they probe the maximum count rate up to 10 Mcps per row. The paper concludes that row-column multiplexing can be extended to kilopixel-scale SNSPD arrays.","tokens_in":8292,"tokens_out":8480,"duration_ms":78440,"significance":"If the results hold, this is an important milestone for SNSPD arrays, demonstrating a viable route to large-format detectors for imaging, spectroscopy, and photon-starved applications. The work combines careful microfabrication with a practical readout scheme, and the persistence-imaging demonstration is a compelling proof of concept. The authors are also transparent about the fundamental ambiguity of row-column multiplexing in multi-photon events and about the limitations of the current readout, including hot pixels and current redistribution. This paper would be a valuable reference for the superconducting-detector community.","major_comments":[{"comment":"The statement that misattributions 'consist of at most 0.6% of the corresponding hot pixels' counts' is not supported by a derivation or a citation. Please provide the calculation or a direct measurement, since this bound is used to argue that the pixel-level efficiency and yield figures are reliable despite the row-column ambiguity.","section":"Section 5.2"},{"comment":"The pixel efficiency values of 12% at 3 uA and 23% at 4 uA are reported without any uncertainty. Please include the 2-D Gaussian fit parameters, the estimated laser spot size and power calibration, and an error analysis that accounts for counting statistics and possible misattribution, so that the headline system efficiency of 8% can be assessed.","section":"Section 5.2"}],"minor_comments":[{"comment":"The baseline yield is stated as 99.4%, but four dead pixels out of 1024 correspond to 99.6%. Please clarify whether the two low-response pixels are counted as non-yielding, and define 'baseline yield' precisely.","section":"Section 5.1"},{"comment":"The jitter is computed from the average of the row and column time tags; please clarify whether this averaging reduces the jitter relative to a single-channel readout, and if so, specify the relationship.","section":"Section 5.3"},{"comment":"The sentence beginning 'at an average count rate of 320 Mcps across the whole array...' introduces a hypothetical scenario; please label it explicitly as an estimate and note that the array was not operated at this rate in the reported measurements.","section":"Section 5.3"},{"comment":"Consider adding a scale bar or amplitude annotation to Fig. 2 so the reader can judge the pulse height relative to the comparator threshold.","section":"Figure 2"}],"recommendation":"minor_revision","confidential_remarks":"The stress-test concern about the coincidence-pairing algorithm is partially addressed by the persistence-image demonstration, which serves as a qualitative ground-truth test at low count rates. The authors should be encouraged to add a quantitative validation or an explicit error analysis in a revised version, but the central feasibility claim is sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nYou should know this paper reports the first 1024-pixel SNSPD array, a 32x32 row-column multiplexed device with 99.4% baseline yield, SDE up to 8% at 1550 nm, and 250–400 ps jitter. It also shows real imaging: a focused laser spot and a swept beam spelling text. That is a genuine capability step, roughly 16x more pixels than prior multiplexed arrays.\n\nWhat's new is the scaling, not the architecture—the same groups published row-column multiplexing for 2x2 and 8x8. The engineering is the contribution: running 1024 pixels with 64 readout channels, calibrating per-channel delays, and getting a working imager. The fabrication and electrical evidence are solid.\n\nThe paper is also honest about the scheme's blind spot. In row-column multiplexing, two photons arriving within the coincidence window are ambiguous; at high count rates this creates false pixels and streaks. They show this explicitly in Fig. 4b and estimate misattributions at ≤0.6% of hot-pixel counts.\n\nThe soft spot is that the pixel-level claims—yield, SDE, hot/dead inventory—rest on the coincidence-pairing algorithm, which never gets a ground-truth test. A multi-spot illumination experiment, where you know which pixels are lit, would give a true positive/false positive rate. Without it, the reported 99.4% yield and 4%/8% SDE carry some unmeasured systematic uncertainty. The max-count-rate test deliberately uses totalized row counts, so it doesn't validate high-rate pixel-resolved imaging, but the paper doesn't claim it does.\n\nAlso, no error bars on efficiency or jitter, and no data/code. Minor: the per-pixel resistance is estimated from a row I-V slope, not measured per pixel.\n\nNone of this undermines the core claim. At the low bias used for imaging, no misattribution is evident, and the ambiguity is a known, fundamental property of the architecture, openly discussed. This is an honest proof-of-concept, not an overreach.\n\nWho it's for: anyone working on SNSPD arrays, single-photon imaging, or dark-matter/optical-communication systems that need large photon-counting focal planes. It deserves a serious referee; I'd send it out with minor revision, asking for error bars and ideally a ground-truth multi-spot test.","headline":"First kilopixel SNSPD array is a real scaling milestone; the pixel-level counting is a bit soft, but the feasibility claim holds.","tokens_in":8896,"tokens_out":3647,"would_cite":true,"duration_ms":38942,"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 reports the first kilopixel-scale superconducting nanowire single-photon detector array, a 32×32 row-column multiplexed imaging array with 99.4% baseline yield, up to 8% system detection efficiency at 1550 nm, and 250–400 ps…","keywords":["superconducting nanowire single-photon detectors","SNSPD array","row-column multiplexing","kilopixel array","single-photon imaging","time-tagging readout","1550 nm detection","timing jitter"],"falsifier":"Illuminate two known pixels simultaneously with synchronized single-photon pulses and compare the reported pixel pairs with the true ones: if the array systematically reports the alternative diagonal pair, or if false pixels appear at the quoted count rates, the coincidence readout does not identify pixels as claimed.","tokens_in":7833,"feed_emoji":"📸","tokens_out":8262,"duration_ms":77325,"temperature":0.7,"pith_summary":"This paper reports the first kilopixel-scale superconducting nanowire single-photon detector array: 1024 pixels on a 1.6 by 1.6 millimeter chip read out through only 64 electrical lines. The authors show that row-column multiplexing, previously demonstrated at 64 pixels, extends to a 32 by 32 grid with a baseline yield of 99.4% working pixels. Under 1550 nm illumination the array reaches about 8% system detection efficiency at 4 microamps of bias per pixel, with per-pixel timing jitter of 250 to 400 ps. The result matters because low-flux astronomy, dark-matter searches, and time-resolved single-photon imaging depend on detector arrays larger than any previously built from superconducting nanowires.","feed_headline":"First 1024-pixel superconducting single-photon imaging array built","feed_subtitle":"Row-column readout yields 99.4% working pixels and 8% system efficiency at 1550 nm.","key_machinery":"The load-bearing mechanism is row-column multiplexing: each pixel is a superconducting nanowire in series with a roughly 50–200 Ω resistor; pixels in a row share one bias line, and pixels in a column share an inductor and ground path. When a photon strikes a pixel, the nanowire becomes resistive and sends opposite-polarity voltage pulses down its row and column lines. Coincidence matching of row and column time tags, using a raw 6 ns window or a 0.6–1.2 ns window after delay calibration, assigns the event to one of the 1024 pixels using only 32 row plus 32 column amplifier channels. The series resistors equalize current sharing, while the column inductors caused a bias instability that the authors circumvented by biasing alternating rows with opposite voltages.","core_discovery":"The central claim is that a 32×32 array of superconducting nanowire pixels, each in series with a resistor and sharing row and column readout lines, can be fabricated over a millimeter-scale area and operated as a single-photon imaging array. A photon is registered by pairing row and column time tags that fall within a coincidence window; after per-channel delay calibration, windows of 0.6–1.2 ns collect more than 99.5% of the counts. The paper reports 1024 pixels with 99.4% baseline yield, up to 8% system detection efficiency at 1550 nm, average jitter of 400 ps at 3 μA and 250 ps at 4 μA, and persistence images of a laser spot swept across the array to spell text. It also states the known cost of coincidence readout: two photons arriving within the timing resolution produce ambiguous row-column pairs, so misattribution grows with count rate.","pith_inferences":["A natural next step is to test whether the same 32×32 geometry with an optical cavity and higher fill factor can push system detection efficiency well above 8%, making the array competitive with single-pixel detectors for photon-starved astronomy.","Because coincidence ambiguity is intrinsic to row-column readout, applications that need multi-photon spatial resolution will require a different encoding such as delay-line, SFQ, or frequency multiplexing, or faster pixels with narrower coincidence windows.","The authors' alternated-bias workaround implies the column inductors are not needed with DC-coupled amplifiers; removing them could simplify biasing and raise the maximum stable bias current.","Pairing row and column time tags in real time on an FPGA, which the paper notes the PCIe output is compatible with, would turn the post-processed demonstration into a live single-photon camera."],"forward_implications":["Because $N\\times N$ pixels are read out with $2N$ lines, an array of this type can be scaled to larger square grids without multiplying cryogenic cables and amplifiers.","A baseline yield of 99.4% on a 1.6 by 1.6 mm chip shows that the nanowire fabrication process is uniform enough to populate kilopixel arrays, not just small test chips.","The demonstrated 4–8% system detection efficiency at 1550 nm is a starting point; embedding pixels in an optical cavity and raising the fill factor would directly multiply total efficiency.","The measured 250–400 ps jitter is worse than single-pixel SNSPDs, but it is already sufficient for time-resolved imaging and lidar; reducing current redistribution or using a constant-fraction discriminator should tighten it.","Count rate is bounded by coincidence ambiguity: with a 6 ns window, roughly 30% of detected events are misattributed at 320 Mcps across the array, while a 1.2 ns window reduces that to about 3%."],"supporting_citations":[{"why":"Demonstrates the 8x8 row-column multiplexing array that this work extends to 32x32.","marker":"[14]"},{"why":"Introduces the row-column pixel readout scheme for a 2x2 WSi superconducting nanowire array.","marker":"[19]"},{"why":"Provides the 64-channel time-tagging readout electronics used to record row and column pulses.","marker":"[3]"},{"why":"Supplies the 64-channel time-to-digital converter that generates the time tags.","marker":"[20]"},{"why":"Shows a roughly 600-effective-pixel delay-line superconducting nanowire imager, the scale this array surpasses.","marker":"[18]"},{"why":"Represents the direct-readout 64-pixel benchmark that motivates multiplexed readout.","marker":"[12]"},{"why":"Surveys cryogenic readout architectures for superconducting nanowire arrays and frames the multiplexing choice.","marker":"[13]"}],"fun_headline_variants":["Kilopixel SNSPD array achieves single-photon imaging","1024-pixel SNSPD imager with row-column readout","Superconducting nanowires image at 1024 pixels","Single-photon imaging scaled to kilopixel array"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The imaging and pixel-identification claims rest on the assumption that row and column time tags paired within the coincidence window actually identify which pixel fired, because two photons arriving close in time create row-column pairs that the readout cannot distinguish.","fun_headline_variants_meta":{"raw":{"variants":["Kilopixel SNSPD array achieves single-photon imaging","1024-pixel SNSPD imager with row-column readout","Superconducting nanowires image at 1024 pixels","Single-photon imaging scaled to kilopixel array"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000577,"raw_usage":{"total_tokens":2629,"prompt_tokens":760,"completion_tokens":1869,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":376,"completion_tokens_details":{"reasoning_tokens":1799}},"tokens_in":376,"tokens_out":1869,"duration_ms":15130,"temperature":1.0,"reasoning_tokens":1799,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:41:20.078770+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Illuminate two known pixels simultaneously with synchronized single-photon pulses and compare the reported pixel pairs with the true ones: if the array systematically reports the alternative diagonal pair, or if false pixels appear at the quoted count rates, the coincidence readout does not identify pixels as claimed.","supporting_citations":[{"cited_title":"A near-infrared 64-pixel superconducting nanowire single photon detector array with integrated multiplexed readout,","cited_arxiv_id":null,"evidence_quote":"Demonstrates the 8x8 row-column multiplexing array that this work extends to 32x32."},{"cited_title":"A four-pixel single-photon pulse-position array fabricated from wsi superconducting nanowire single-photon detectors,","cited_arxiv_id":null,"evidence_quote":"Introduces the row-column pixel readout scheme for a 2x2 WSi superconducting nanowire array."},{"cited_title":"Status of NASA’s deep space optical communication technology demonstration,","cited_arxiv_id":null,"evidence_quote":"Provides the 64-channel time-tagging readout electronics used to record row and column pulses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the 64-channel time-to-digital converter that generates the time tags."},{"cited_title":"Single-photon imager based on a superconducting nanowire delay line,","cited_arxiv_id":null,"evidence_quote":"Shows a roughly 600-effective-pixel delay-line superconducting nanowire imager, the scale this array surpasses."},{"cited_title":"A 64-pixel NbTiN superconducting nanowire single-photon detector array for spatially resolved photon detection,","cited_arxiv_id":null,"evidence_quote":"Represents the direct-readout 64-pixel benchmark that motivates multiplexed readout."},{"cited_title":"Readout architectures for superconducting nanowire single photon detectors,","cited_arxiv_id":null,"evidence_quote":"Surveys cryogenic readout architectures for superconducting nanowire arrays and frames the multiplexing choice."}],"review_version":1}