{"id":"3c03cc4c-d05b-44d6-b20c-8ca74a31948b","arxiv_id":"2412.09314","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"First MeerKAT S-band extragalactic catalogue of the DEEP2 field, with 1,199 sources and source counts consistent with prior surveys above the completeness limit.","lead":"Astronomers used the MeerKAT radio telescope's new S-band receivers to image the DEEP2 field for about 70 minutes per frequency band, producing the first widefield catalogue of 1,199 radio sources at 2.5 GHz. The catalogue and source counts match earlier radio surveys, showing that MeerKAT S-band can deliver scientifically useful extragalactic images in short observing sessions.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The completeness correction for the source counts assumes a fixed SKADS point-source fraction of 1/3, but the authors note that most faint star-forming galaxies are likely unresolved, so the true point-source fraction may be higher and bias the faint-end counts and the consistency claim.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the completeness correction's dependence on the assumed point-source fraction. This is indeed the most fragile step in the central claim because the corrected source counts, and hence the consistency with literature, are derived using a fixed 1/3 fraction taken from SKADS, while the authors acknowledge that most faint SFGs are likely unresolved. The internal inconsistency in Section 4.1 (resolved/unresolved modeling sentence) is noted but the completeness curves in Fig. 7 indicate the intended modeling, so it is likely a typographical error and not the primary concern. Varying the point-source fraction is a straightforward sensitivity test that would settle whether the consistency claim is robust. The catalogue itself remains a valuable data product, and the paper's conditional verdict is appropriate.","tokens_in":23107,"tokens_out":6560,"duration_ms":64986,"concrete_test":"Recompute the R=0.3 corrected source counts in Table C1 using the Fig. 7 completeness curves with point-source fractions of 0.5 and 0.7 instead of 1/3, propagating the Poisson and completeness uncertainties, and test whether the corrected counts in the three faintest bins (S < 0.06 mJy) remain within 1σ of the Matthews et al. (2021a) counts. If they shift by more than the quoted uncertainties, the consistency claim depends critically on the assumed size distribution.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 4.4, the completeness-corrected differential source counts assume that point sources constitute one third of the source population, following the intrinsic size distribution in the SKADS simulation. This assumption is load-bearing because the completeness curves for point and extended sources differ substantially (Fig. 7), especially below ~0.1 mJy where the correction factors are large. The authors themselves note in footnote 8 that star-forming galaxies are 'always extended' in the SKADS definition even though most are likely unresolved at the current resolution. Since the faint source population is dominated by star-forming galaxies (Algera et al. 2020), the true fraction of point-like sources is likely larger than 1/3, meaning the blended completeness is underestimated and the corrected counts in the faintest bins are overestimated. The paper's claim of consistency with Matthews et al. (2021a) and Smolčić et al. (2017) therefore rests on an unverified size-distribution assumption. A secondary inconsistency appears in Section 4.1, where the text reads 'Resolved and unresolved sources are respectively modelled as delta functions and Gaussians', which is inverted relative to the expected modeling and contradicts the completeness curves in Fig. 7; this suggests a typo, but it highlights the need for a clear statement of the simulation setup.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents MeerKAT S-band (2.5 GHz) continuum imaging of the DEEP2 field from about 70 minutes on source per sub-band, with detailed calibration, primary-beam correction, source detection, astrometric verification against the Matthews et al. (2021a) L-band catalogue, spectral index measurements, resolved-source classification, completeness and purity simulations, differential source counts, and image-plane stacking. The main scientific claims are that the resulting R=0.3 catalogue (1199 sources) and R=-0.5 catalogue (670 sources) constitute the first S-band extragalactic catalogue of this field, that the differential source counts are consistent with literature measurements and SKADS within the expected variance, and that the spectral index distributions have medians of alpha_L^S1 = -0.52 and alpha_S1^S4 = -0.61.","tokens_in":23411,"tokens_out":5904,"duration_ms":61571,"significance":"The paper is a carefully executed pilot study with substantial technical value: the calibration and imaging steps are described in enough detail to be reproduced, the astrometric offsets are small and well characterised, the noise is tested for Gaussianity, and the completeness and purity simulations follow standard practice. The catalogue and images are made publicly available. The main caveat is the completeness correction, which assumes a fixed point-source fraction derived from SKADS; because the faint-end consistency claim rests on corrected counts in a regime where the correction is large, this assumption needs to be tested explicitly. If the counts survive such a robustness test, the paper will be a useful reference for planning future MeerKAT S-band surveys.","major_comments":[{"comment":"The completeness-corrected source counts assume that point sources constitute one third of the population. Fig. 7 shows that the point-source and extended-source completeness curves differ substantially below roughly 0.1 mJy, where the correction factors are large. Footnote 8 acknowledges that star-forming galaxies are classified as extended in the SKADS catalogue even though most are likely unresolved at the current resolution. Since the faint population is dominated by star-forming galaxies, the true point-source fraction is plausibly larger than 1/3; in that case the blended completeness used here is underestimated and the corrected counts in the faintest bins of Table C1 are overestimated. The claimed consistency with Matthews et al. (2021a) and Smolčić et al. (2017) at S < 100 uJy is therefore contingent on an unverified size-distribution assumption. I request a robustness test, e.g. recomputing the corrected counts for point-source fractions of 0.5, 0.7, and 1.0, or, if such a test is not feasible, an explicit statement that the faint-end consistency is not robust to this assumption.","section":"Section 4.4 and footnote 8"},{"comment":"The stacked S-band flux density is 26% lower than expected from the L-band catalogue assuming alpha = -0.7. The paper attributes this to a systematic effect but does not quantify its possible impact on the catalogue flux scale or on the corrected source counts. A global offset in the faint-end S-band flux scale would shift sources between flux density bins and could bias the counts, so this is not only a stacking issue. I ask for a short assessment of the maximum plausible effect of this 26% deficit on the source counts in the affected flux range, or an explicit statement that the deficit is confined to the stacking analysis and cannot affect the catalogue-based counts.","section":"Section 4.3"}],"minor_comments":[{"comment":"The sentence 'Resolved and unresolved sources are respectively modelled as delta functions and Gaussians' is inverted relative to the SKADS description given in the preceding sentence; unresolved sources are the delta functions and resolved sources are the Gaussians. Please correct this and confirm that the simulation code uses the intended mapping.","section":"Section 4.1"},{"comment":"There is a duplicated word in 'we do not expect expect extreme outliers'; it should read 'we do not expect extreme outliers'.","section":"Section 4.2"},{"comment":"There is a duplicated article in 'the the majority of these sources'; it should read 'the majority of these sources'.","section":"Section 3.2"},{"comment":"There is a duplicated article in 'subtract the the fitted uncertainty on S_peak'; it should read 'subtract the fitted uncertainty on S_peak'.","section":"Section 3.3.2"},{"comment":"There is a duplicated 'where' in 'where where simulations and different observations begin to disagree'; it should read 'where simulations and different observations begin to disagree'.","section":"Section 4.4"}],"recommendation":"major_revision","confidential_remarks":"This is a solid pilot study with public data products and careful technical validation. The only load-bearing issue is the SKADS-based point-source fraction in the completeness correction; if the authors add a robustness test or explicitly qualify the faint-end consistency claim, I would expect the paper to be acceptable. The unexplained 26% stacking deficit should also be addressed as a systematic uncertainty for the source counts. No concerns about novelty, referencing, or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good pilot data paper, worth a proper refereeing: the catalogue is new and publicly available, the analysis is standard and careful, and the main scientific claim (counts consistent with the literature) is broadly supported, with one caveat about the completeness correction that should be quantified before publication.\n\nThe genuinely new thing is the data product: the first widefield extragalactic continuum catalogue with MeerKAT S-band in DEEP2, 1199 sources down to ~17 uJy at R=0.3, plus a higher-resolution catalogue, and the images are on the SARAO archive. The calibration and imaging are described in detail, including the primary-beam correction from holographic models. The source counts are derived with the usual machinery: PyBDSF, completeness from SKADS injections, purity from negative detections, and cosmic variance from SKADS realisations. The spectral indices between L-band and S-band and within the band are a useful extra. The paper is honest that this is a shallow pilot; it does not claim more than it delivers.\n\nThe soft spot is the completeness correction. It assumes a fixed point-source fraction of one third. The authors themselves note in footnote 8 that most star-forming galaxies, which dominate the faint population, are likely unresolved at this resolution, so the true point fraction could be higher. That would bias the corrected faint-end counts downward (i.e., the reported counts may be overestimated where the corrections are large). The corrected counts in the lowest bins are factors of tens above the raw counts, so this is not a negligible effect. The paper should show how the counts change for a plausible range of point-source fractions. It doesn't. Also, Section 4.1 has an inverted sentence: 'Resolved and unresolved sources are respectively modelled as delta functions and Gaussians' — that's backwards. Easy fix.\n\nThe abstract's source-count consistency claim is a little stronger than the data support; below ~50 uJy the completeness corrections become inadequate, and the paper says so in the text, but the abstract doesn't carry that qualification. That should be tidied.\n\nOverall, the paper is a solid reference for MeerKAT S-band survey planning. The catalogue will be cited. With the completeness caveat quantified and the typo fixed, I'd be happy to see it in MNRAS. Send it to a referee; it's not a desk reject.","headline":"A useful pilot data paper with a new public catalogue; the source-count consistency claim needs a quantified completeness caveat.","tokens_in":23993,"tokens_out":2854,"would_cite":true,"duration_ms":29739,"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":"With 70 minutes on source, MeerKAT S-band delivers a 1,199-source catalogue of the DEEP2 field, with source counts consistent with established surveys.","keywords":["MeerKAT","S-band","radio continuum","source counts","DEEP2 field","extragalactic catalogue","spectral index","radio interferometry"],"falsifier":"Observe DEEP2 again at S-band to roughly 1 microjansky per beam rms with resolution comparable to the R=-0.5 image, and measure the point-source fraction and counts below 0.1 mJy. If the point-source fraction deviates strongly from one third, or the Euclidean-normalized counts fall outside the uncertainties reported here, the completeness-corrected counts and the claimed consistency with the literature are falsified.","tokens_in":22969,"feed_emoji":"📡","tokens_out":6099,"duration_ms":55569,"temperature":0.7,"pith_summary":"The paper presents the first widefield extragalactic continuum catalogue made with the MeerKAT S-band receivers, targeting the radio-selected DEEP2 field. By combining the S1 and S4 sub-bands and about 70 minutes on source, the authors produce an image with 4.7 microjansky per beam sensitivity and detect 1,199 sources down to 16.9 microjansky. They show that the completeness-corrected differential source counts agree with established L-band and 3-GHz surveys, and that stacking faint L-band sources pushes a count estimate down to roughly 10 microjansky. The wider point is that even a short single-pointing S-band observation yields scientifically usable extragalactic results, which supports the case for larger MeerKAT S-band surveys.","feed_headline":"1199 radio sources catalogued in first MeerKAT S-band survey","feed_subtitle":"Seventy minutes on source reaches 16.9 microjansky and matches known source counts.","key_machinery":"The load-bearing element is the MeerKAT S-band receiver system, whose band is split into five sub-bands (S0–S4); the paper combines S1 and S4 into a contiguous 1.97–3.50 GHz coverage, excludes spectral windows with high noise, and images with WSClean at two Briggs robust weightings. Source detection uses PyBDSF with a 3-sigma island threshold and 5-sigma pixel threshold. Completeness is measured by injecting SKADS-simulated point and extended sources into residual images and re-running detection; the resulting completeness curves enter the source-count correction under the assumption that one third of sources are point-like and two thirds extended. Primary-beam correction uses per-spectral-window holographic beam models combined with the same weights as the multi-frequency synthesis image, and cosmic variance on the counts is estimated by resampling SKADS skies over the same area.","core_discovery":"The central claim is that MeerKAT's new S-band system is a working widefield extragalactic imaging instrument. Combining sub-bands S1 (1.97–2.84 GHz) and S4 (2.62–3.50 GHz) in a multi-frequency-synthesis image, with 70 minutes on source per sub-band and 52–55 antennas, the authors reach a robust-weighted rms of 4.7 microjansky per beam and extract 1,199 sources at R=0.3 (670 at R=-0.5). They report completeness-corrected Euclidean-normalized differential source counts at 2.5 GHz down to about 20 microjansky that are consistent, within the SKADS-estimated cosmic variance, with the L-band DEEP2 counts and VLA-COSMOS 3-GHz counts. They also derive median spectral indices of about -0.5 to -0.6 across an effective 1.8 GHz frequency baseline, identify 18–22 percent of sources as resolved, and use image-plane stacking to estimate lower-limit counts down to 10 microjansky, finding them consistent with simulations though lower than some published counts.","pith_inferences":["If the assumed one-third point-source fraction is off, the completeness-corrected counts below about 0.1 mJy are biased; a deeper, higher-resolution S-band image of DEEP2 would settle this directly.","The 26 percent shortfall in stacked S-band flux relative to L-band expectations, if not a systematic of stacking, implies a median spectral index near -1.2 for the faint population, which would steepen all S-band count conversions at the faint end.","The demonstrated short-integration sensitivity suggests the upcoming MeerKAT+ array could push into confusion-limited sub-microjansky S-band imaging over wide fields, not just statistical stacking.","The S1 high-frequency excess noise and beam-elongation peaks mean that survey designers should prefer the S2 sub-band, as the authors note, for maximum usable bandwidth."],"forward_implications":["A single 70-minute MeerKAT S-band pointing reaches source densities that previously required much longer VLA campaigns, making large-area S-band surveys cheap.","The S-band catalogue, combined with the L-band DEEP2 data, yields spectral indices over an effective 1.8 GHz baseline that can separate flat-spectrum core-dominated sources from steeper star-forming and synchrotron populations.","Stacking L-band source positions produces a 62-sigma detection of the faint S-band population, demonstrating that source counts can be extended below the detection threshold.","Because a single MeerKAT S-band pointing covers roughly 16 times the sky of the deepest published S-band survey at comparable sensitivity, future surveys will cut cosmic-variance and Poisson scatter.","The authors recommend a robust weighting of R=0.3 as the optimal choice for broadband extragalactic S-band imaging."],"supporting_citations":[{"why":"Defines the DEEP2 field and supplies the deep 1.28 GHz image used as the L-band benchmark for comparison and cross-matching.","marker":"Mauch et al. (2020)"},{"why":"Provides the L-band catalogue and source counts used for cross-matching, spectral-index fitting, stacking positions, and the source-count comparison.","marker":"Matthews et al. (2021a)"},{"why":"Supplies the SKADS simulated sky used to measure completeness and to estimate cosmic variance on the source counts.","marker":"Wilman et al. (2008)"},{"why":"Offers the VLA-COSMOS 3 GHz source counts to which the S-band counts are converted and compared.","marker":"Smolčić et al. (2017)"},{"why":"Defines the deepest published S-band survey, COSMOS-XS, used as the sensitivity and sky-coverage benchmark for the capability claim.","marker":"van der Vlugt et al. (2021)"},{"why":"Provides the variance-estimation method applied to SKADS resampling to quantify source clustering and Poisson scatter on the counts.","marker":"Heywood et al. (2013)"},{"why":"Supplies the holographic MeerKAT S-band primary-beam measurements used for accurate flux correction.","marker":"de Villiers (2023)"},{"why":"Provides the PyBDSF source finder used for detection and measurement of the catalogue sources.","marker":"Mohan & Rafferty (2015)"}],"fun_headline_variants":["MeerKAT S-band first look: 1199 radio sources found","MeerKAT S-band: 1199 sources, 4.7 uJy rms in 70 min","First MeerKAT S-band survey: 1199 radio sources catalogued","MeerKAT S-band reveals 1199 sources in short 70-min exposure","S-band MeerKAT catalog: 1199 sources, matches known counts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The completeness correction that converts raw detections to true source counts assumes, following the SKADS simulation, that exactly one third of sources are point-like and the rest extended; if the real sky has a different size distribution, the corrected counts in the faint regime where corrections are large will be wrong.","fun_headline_variants_meta":{"raw":{"variants":["MeerKAT S-band first look: 1199 radio sources found","MeerKAT S-band: 1199 sources, 4.7 uJy rms in 70 min","First MeerKAT S-band survey: 1199 radio sources catalogued","MeerKAT S-band reveals 1199 sources in short 70-min exposure","S-band MeerKAT catalog: 1199 sources, matches known counts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000588,"raw_usage":{"total_tokens":2804,"prompt_tokens":1031,"completion_tokens":1773,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":647,"completion_tokens_details":{"reasoning_tokens":1661}},"tokens_in":647,"tokens_out":1773,"duration_ms":13140,"temperature":1.0,"reasoning_tokens":1661,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T17:05:25.934747+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe DEEP2 again at S-band to roughly 1 microjansky per beam rms with resolution comparable to the R=-0.5 image, and measure the point-source fraction and counts below 0.1 mJy. If the point-source fraction deviates strongly from one third, or the Euclidean-normalized counts fall outside the uncertainties reported here, the completeness-corrected counts and the claimed consistency with the literature are falsified.","supporting_citations":[],"review_version":1}