{"id":"cb922ede-4ace-41e2-ab66-74f835ab44bf","arxiv_id":"2501.07357","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A 64-pixel SNSPD camera achieves 65% system detection efficiency, 100 ps timing jitter, 20 cps/pixel dark count rate, and 645 Mcps maximum count rate at 1550 nm.","lead":"This paper demonstrates a 64-pixel superconducting nanowire camera that detects single photons at 1550 nm with 65% system efficiency, 100 ps timing jitter, and a 645 Mcps maximum count rate. The self-contained rack-mounted system is aimed at SWIR imaging, quantum communication, and LiDAR, where photon-starved, fast-timing measurements are needed.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The per-pixel SPDE measurement uses the total array count rate for every pixel, so the 77.7% average and ±0.6% uniformity do not actually isolate individual pixels; this weakens the abstract's central per-pixel efficiency claim.","rationale":"The paper reports an impressive, plausibly useful 64-pixel SNSPD system, and much of the characterization—jitter, dark count rate, maximum count rate, crosstalk—is credible. The most prominent headline number, however, is the per-pixel 77.7% efficiency with ±0.6% uniformity, and the reader's weakest assumption identifies exactly where the argument is least secure. The per-pixel measurement is not per-pixel: Sec. 3/Fig. 2b states that R_counts is the total array count rate, so the 'per-pixel' SPDE values for all 64 pixels share the same aggregate numerator. The quoted standard deviation therefore reflects scan-to-scan reproducibility of the array response, not pixel-to-pixel efficiency variation. The authors' justification—that neighboring-pixel absorptions would have been absorbed by the addressed pixel if the spot were contained—requires the spot to match the pixel area exactly and ignores fill-factor gaps and lower-efficiency connection lines. This is a methodological mismatch with the claim, not a disagreement with community consensus. The 65% SDE is less affected because aggregate counting is appropriate for an array-level metric, though the selection of the maximum over spot sizes is worth clarifying. The per-pixel efficiency claim is the one a reader would carry away, and it can be settled by re-analyzing existing per-channel TDC data or by a small-spot per-pixel measurement. Since the reader already conditioned the verdict on this issue, my stress-test does not move the verdict; the manuscript should be revised to either report true per-pixel values or rephrase the claim as array-level detection efficiency at each position.","tokens_in":13020,"tokens_out":6763,"duration_ms":68083,"concrete_test":"Recompute the Fig. 2b efficiencies from the saved 64-channel time-tag records using only the addressed pixel's channel: SPDE_i=(C_i-DCR_i)/R_input. Compare the resulting per-pixel mean and standard deviation with the reported 77.7±0.6%, and report the neighbor fraction (1-C_i/R_counts) for each spot. If the per-pixel mean differs by more than the stated ±5% uncertainty or the spread exceeds ±0.6%, the abstract's per-pixel uniformity claim is not supported by the current data. If per-channel records were not archived, repeat the measurement with a focused spot much smaller than the 27.5 µm pixel pitch (e.g., ≤10 µm diameter) and count only the illuminated pixel.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of a 'uniform average per-pixel photon detection efficiency of 77.7%' rests on the SPDE measurement described in Sec. 3 and Fig. 2b. The text states explicitly: 'R_counts is taken as the total array count rate rather than the count rate measured for each individual pixel.' Thus the numerator in (R_counts - R_dcr)/R_input is the same aggregate TDC count for every pixel; moving the spot changes which pixel is centered, but the number includes counts from neighboring pixels, inter-pixel connection wires, and any other pixels that intercept the ~27 µm Gaussian tails. The authors argue that photons absorbed in neighboring pixels would have been absorbed in the measured pixel if the spot were contained, but this is a compensating assumption, not a measurement. The reported σ=±0.6% therefore quantifies reproducibility of the array-level response as a spot is scanned, not pixel-to-pixel variation in single-pixel efficiency. Edge and corner reductions are consistent with this interpretation. The 77.7% figure also conflates fill factor (~85%), internal nanowire detection probability, and optical alignment. If per-pixel counts from the 64-channel TDC were used instead, the mean and spread could differ materially; dead or inefficient pixels could be masked by neighbors in the total-count method. This is load-bearing because the abstract's headline efficiency and uniformity statement is not supported by the measurement as described.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a rack-mounted, direct-readout 64-pixel NbTiN superconducting nanowire single-photon detector array operating at 920 mK, with each pixel read out through a 64-channel time-to-digital converter. The authors claim an average per-pixel photon detection efficiency of 77.7% with a standard deviation of ±0.6% at 1550 nm, a system detection efficiency (SDE) of 65%, a dark count rate of ~20 cps per pixel, a per-pixel FWHM jitter of ~100 ps, a 3-dB maximum count rate of 645 Mcps, and no evidence of crosstalk at the 0.1% level. The central claim is that this is the highest-performing time-correlated single-photon counting SWIR imaging array for this pixel count.","tokens_in":13272,"tokens_out":4811,"duration_ms":44924,"significance":"If the per-pixel efficiency claim is properly supported, this system is a notable advance for SWIR single-photon imaging: it combines high system efficiency with high per-pixel count rates, low dark counts, and low crosstalk in a self-contained apparatus. The paper's strengths include direct experimental characterization with stated power-meter uncertainty, a detailed description of the cryogenic and optical setup, and a useful comparison with commercial 36-pixel direct-readout arrays. However, the headline per-pixel efficiency and uniformity numbers are not justified by the measurement as described, which materially weakens the central claim and needs to be addressed before publication.","major_comments":[{"comment":"The per-pixel SPDE values are computed from the total array count rate R_counts, as the text explicitly states: 'R_counts is taken as the total array count rate rather than the count rate measured for each individual pixel.' Therefore the 77.7% average and σ=±0.6% describe the array-level response when the focused spot is centered at each pixel, not the intrinsic detection efficiency of each pixel. Counts from neighboring pixels, inter-pixel wiring, and any other pixels intercepting the ~27 µm Gaussian tails are included in the numerator for every pixel, so the measurement cannot detect a dead or inefficient pixel that is masked by its neighbors. The conclusion that 'all 64 pixels achieving a maximum SPDE in the range 76.2–79.2%' is not supported by the described measurement. Please reanalyze the data using per-pixel counts from the 64-channel TDC, or alternatively revise the abstract and conclusion to state 'array-level efficiency when the spot is centered on each pixel' and remove the per-pixel uniformity claim.","section":"Sec. 3, Fig. 2b"},{"comment":"The paper reports a per-measurement error of approximately ±5% from power-meter uncertainty and spot-alignment error, yet claims a uniformity spread of only ±0.6% standard deviation and all values within about 3%. Under this stated error budget, the observed pixel-to-pixel differences are not statistically significant. The uniformity claim should either be supported by a proper uncertainty propagation that separates common-mode systematic error from pixel-to-pixel variation, or be presented as a reproducibility check of the array-level measurement rather than as evidence of per-pixel uniformity.","section":"Sec. 3, Fig. 2b caption"}],"minor_comments":[{"comment":"The crosstalk discussion refers to Figures 4, 5, and 6 inconsistently (e.g., 'as shown in Fig. 6' followed by 'In Fig. 4' and 'As seen in Fig. 5'); please renumber the figures and ensure all citations point to the correct panels.","section":"Sec. 3, Crosstalk"},{"comment":"The caption contains a typo: 'the highlighted red and blue lines represent thefrom 2 individual pixels' should read 'represent two individual pixels'.","section":"Fig. 3 caption"},{"comment":"The text describes a '50/50 beamsplitter' but Fig. 1c labels a '45:55 Beamsplitter'; please make the split ratio consistent between text and figure.","section":"Sec. 2, Optical setup"},{"comment":"The array SDE of 65% is the maximum over the spot sizes shown in Fig. 2c; the abstract's phrase 'the observed system detection efficiency (SDE) across the entire array was 65%' would be clearer if it stated that this is the maximum value obtained with an optimized, expanded spot and that the SDE varies with illumination size.","section":"Sec. 3, SDE measurement"},{"comment":"The supplemental document contains a typo: 'crysostat' should be 'cryostat'.","section":"Supplemental document"}],"recommendation":"major_revision","confidential_remarks":"The per-pixel efficiency claim is the principal obstacle to acceptance. If the authors can extract per-pixel count rates from the 64-channel TDC and re-plot Fig. 2b using those counts, the uniformity claim may be salvageable. If not, the paper should be revised to frame the 77.7% figure as an array-level measurement and to remove per-pixel language from the abstract and conclusion. The comparison with the Single Quantum 36-pixel array should also clarify that the reported SDE is for 24 connected pixels, not a direct 64-pixel comparison."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this for the device, not for the per-pixel efficiency methodology. The 65% system SDE, 100 ps jitter, 20 cps/pixel dark count, and 645 Mcps count rate are a strong combination for a 64-pixel direct-readout array at 1550 nm, and the rack-mounted system looks practical. The abstract's 77.7% 'uniform average per-pixel SPDE' is the problem. As the text admits in Sec. 3, R_counts is the total array count rate, not per-pixel counts. So the ±0.6% standard deviation measures how the array responds as a spot is scanned across it, not pixel-to-pixel variation. The edge/corner drops support that reading. The authors argue photons hitting neighbors would have been absorbed in the centered pixel if the spot were contained, which is a reasonable correction but still an assumption, not a measurement. A dead or inefficient pixel could be masked by its neighbors. This is the load-bearing part of the abstract claim, and it doesn't hold up.\n\nThat said, the rest of the characterization is credible. The array SDE measurement is a direct input-flux vs counted-photons ratio with stated uncertainty, and the jitter/count-rate/dark-count values look consistent with the plotted data. The comparison to Single Quantum's 36-pixel arrays is fair but narrow; those have better jitter, so 'record high' should be qualified. A few metrics (jitter, MCR) lack error bars.\n\nThe paper deserves a serious referee. The device is real and useful, and the issues are fixable with clearer framing: present the uniformity measurement for what it is (array-level spot response), report per-pixel counts if available, and soften the per-pixel efficiency claim. It's a good-enough engineering contribution to publish after revision.","headline":"Solid engineering paper with a usable 64-pixel SNSPD camera, but the per-pixel efficiency uniformity claim is weaker than the abstract suggests.","tokens_in":13945,"tokens_out":2517,"would_cite":true,"duration_ms":24875,"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 64-pixel superconducting nanowire detector array reaches 77.7% per-pixel efficiency at 1550 nm with 645 Mcps count rate.","keywords":["superconducting nanowire single-photon detector","SNSPD array","single-photon imaging","SWIR detection","photon detection efficiency","timing jitter","maximum count rate","direct readout"],"falsifier":"Re-measure the single-photon detection efficiency of each pixel using that pixel's own count rate, with the focused spot contained within the pixel, and compare the per-pixel distribution to the claimed 77.7% average and 0.6% standard deviation; a discrepancy larger than the stated error bars would invalidate the uniform per-pixel efficiency claim.","tokens_in":12799,"feed_emoji":"📷","tokens_out":4216,"duration_ms":33984,"temperature":0.7,"pith_summary":"This paper demonstrates a self-contained 64-pixel superconducting nanowire single-photon detector (SNSPD) array for the short-wave infrared, claiming record-class performance across all key metrics at once: an average per-pixel photon detection efficiency of 77.7% at 1550 nm, a system detection efficiency of 65%, 20 dark counts per second per pixel, 100 ps timing jitter, a 3-dB maximum count rate of 645 Mcps, and no detectable crosstalk at the 0.1% level. The authors argue that direct readout, with one coaxial cable per pixel, is the practical path to simultaneously high efficiency, count rate, and low jitter at this pixel count. If the numbers hold, the system is a record-class SWIR single-photon camera and a ready platform for high-dimensional quantum photonics, LiDAR, and photon-starved imaging.","feed_headline":"64-pixel single-photon camera hits 77.7% efficiency","feed_subtitle":"A direct-readout superconducting nanowire array also delivers 645 Mcps count rate and 100 ps jitter at 1550 nm.","key_machinery":"The central object is the SNSPD array itself: 64 NbTiN nanowire pixels, each meandered over a roughly 27.5 by 27.8 micron active area on a 30 micron pitch, giving an 80 to 85 percent fill factor, with a gold back mirror and anti-reflection coating forming a low-quality-factor optical cavity at 1550 nm to boost absorption. The array is read out directly, one coaxial cable per pixel, with cryogenic amplifiers at 40 K and a 64-channel time-to-digital converter; the direct-readout architecture is the mechanism that avoids the count-rate and multi-photon tradeoffs of multiplexed arrays. The key identity in the efficiency measurement is $\\text{SPDE} = (R_{\\text{counts}} - R_{\\text{dcr}})/R_{\\text{input}}$, though the authors use the total array count rate for $R_{\\text{counts}}$ rather than per-pixel counts.","core_discovery":"The authors report the fabrication and full characterisation of an 8x8 array of 27.5x27.8 micron NbTiN nanowire pixels on a 30 micron pitch, with a gold back mirror and bilayer anti-reflection coating forming a low-finesse cavity centred at 1550 nm. They measure a mean per-pixel single-photon detection efficiency of 77.7% with a standard deviation of 0.6%, a maximum system detection efficiency of 65%, an average dark count rate of about 20 counts per second per pixel, per-pixel timing jitter around 100 ps FWHM, a 3-dB maximum count rate of about 645 Mcps for the array, and inter-pixel crosstalk constrained below 0.1%. They conclude this is the most optimally performing time-correlated single-photon counting imaging SNSPD array to date in the SWIR, and that the direct-readout architecture is what allows all parameters to remain high simultaneously.","pith_inferences":["Because the per-pixel efficiency is computed from the total array count rate, the reported 0.6% uniformity may partly reflect the constancy of the total count rate rather than true pixel-to-pixel variation; a genuine per-pixel measurement could reveal a larger spread that matters for imaging uniformity.","The same pixel design scaled to smaller pixels would trade fill factor for higher per-pixel count rate and lower geometric jitter, suggesting a design family where pixel size is chosen to match the application's required frame rate and timing precision.","The rack-mounted, self-contained cryostat and direct-readout electronics make this a turnkey detector; a natural next step is to connect it to a multi-plane light converter or complex-medium circuit to perform projective measurements on high-dimensional spatial modes."],"forward_implications":["An SWIR single-photon camera with roughly 78% per-pixel efficiency and about 100 ps timing resolution would enable single-shot multi-spatial-mode measurements for high-dimensional quantum key distribution and entanglement certification.","The 645 Mcps array count rate, about 8 to 11 Mcps per pixel, is sufficient for fast-acquisition single-photon LiDAR and high-data-rate optical communication, and it exceeds multiplexed arrays by orders of magnitude, as the 400,000-pixel camera runs at roughly 10 kcps.","The measured 65% system detection efficiency, given the 80 to 85% fill factor and window and filter losses, implies the per-pixel internal detection efficiency is near the 77.7% level and that residual losses are mostly geometric and optical.","The absence of crosstalk at the 0.1% level over a 10 ns correlation window means the array can be used for coincidence-based quantum measurements without corrections for correlated noise."],"supporting_citations":[{"why":"Supplies the early proof-of-principle imaging arrays that this system is compared against for efficiency and fill factor improvements.","marker":"[39]"},{"why":"Provides the previous 64-pixel NbTiN array that this direct-readout design extends and outperforms.","marker":"[46]"},{"why":"Gives the 400,000-pixel camera benchmark whose roughly 10 kcps count rate highlights the count-rate advantage of direct readout.","marker":"[54]"},{"why":"Supplies the differential-readout concept that motivates the double-ended pixel design used to reduce geometric jitter.","marker":"[55]"},{"why":"Supports the free-space filtering arrangement with short-pass and bandpass filters that keeps dark counts low while maintaining high coupling efficiency.","marker":"[56]"},{"why":"Provides the polarization dependence of SNSPD absorption that explains the parallel-polarization configuration used for maximum efficiency.","marker":"[57]"},{"why":"Describes a commercial 36-pixel SNSPD array at 1064 nm whose SDE and jitter serve as a comparison point for the present system.","marker":"[68]"},{"why":"Reports a commercial 36-pixel SNSPD array at 1550 nm, providing the closest existing product comparison for SDE, jitter, and dark count rate.","marker":"[69]"}],"fun_headline_variants":["64-pixel SNSPD array: 77.7% efficiency, 645 Mcps, 100 ps jitter","8x8 SNSPD array hits 77.7% per-pixel efficiency and 645 Mcps","77.7% efficient 64-pixel SNSPD camera with 645 Mcps","SNSPD array: 77.7% efficiency, 645 Mcps, 100 ps jitter","Direct-readout SNSPD array: 77.7% efficiency, 645 Mcps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported per-pixel efficiency claims assume that using the total array count rate in place of each pixel's own count rate faithfully represents individual pixel performance; if per-pixel counting were used, the 77.7% average and 0.6% uniformity might differ.","fun_headline_variants_meta":{"raw":{"variants":["64-pixel SNSPD array: 77.7% efficiency, 645 Mcps, 100 ps jitter","8x8 SNSPD array hits 77.7% per-pixel efficiency and 645 Mcps","77.7% efficient 64-pixel SNSPD camera with 645 Mcps","SNSPD array: 77.7% efficiency, 645 Mcps, 100 ps jitter","Direct-readout SNSPD array: 77.7% efficiency, 645 Mcps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001114,"raw_usage":{"total_tokens":4721,"prompt_tokens":1105,"completion_tokens":3616,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":721,"completion_tokens_details":{"reasoning_tokens":3484}},"tokens_in":721,"tokens_out":3616,"duration_ms":25466,"temperature":1.0,"reasoning_tokens":3484,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:43:59.998300+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-measure the single-photon detection efficiency of each pixel using that pixel's own count rate, with the focused spot contained within the pixel, and compare the per-pixel distribution to the claimed 77.7% average and 0.6% standard deviation; a discrepancy larger than the stated error bars would invalidate the uniform per-pixel efficiency claim.","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":"Supplies the early proof-of-principle imaging arrays that this system is compared against for efficiency and fill factor improvements."},{"cited_title":"A 64-pixel nbtin superconducting nanowire single-photon detector array for spatially resolved photon detection,","cited_arxiv_id":null,"evidence_quote":"Provides the previous 64-pixel NbTiN array that this direct-readout design extends and outperforms."},{"cited_title":"Jitter characterization of a dual-readout snspd,","cited_arxiv_id":null,"evidence_quote":"Supplies the differential-readout concept that motivates the double-ended pixel design used to reduce geometric jitter."},{"cited_title":"Free-space coupled superconducting nanowire single-photon detector with low dark counts,","cited_arxiv_id":null,"evidence_quote":"Supports the free-space filtering arrangement with short-pass and bandpass filters that keeps dark counts low while maintaining high coupling efficiency."},{"cited_title":"Opticalpropertiesofsuperconductingnanowiresingle-photondetectors,","cited_arxiv_id":null,"evidence_quote":"Provides the polarization dependence of SNSPD absorption that explains the parallel-polarization configuration used for maximum efficiency."},{"cited_title":"Superconducting nanowire single-photon detectors for laser communication,","cited_arxiv_id":null,"evidence_quote":"Reports a commercial 36-pixel SNSPD array at 1550 nm, providing the closest existing product comparison for SDE, jitter, and dark count rate."}],"review_version":1}