{"id":"62227beb-596c-4bf6-8d79-aae69cf2afe0","arxiv_id":"2506.18231","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Blind tSZ detection with NIKA2 at 18.5 arcseconds yields 16 cluster candidates in COSMOS, 8 with counterparts, demonstrating access to low-mass, high-redshift clusters.","lead":"This paper searches 877 square arcminutes of the COSMOS field in 2 mm NIKA2 data and finds 16 blind Sunyaev-Zel'dovich cluster candidates above signal-to-noise 4, of which 8 have known optical or X-ray counterparts. It is the first blind cluster detection at 18 arcsecond millimeter resolution, and shows NIKA2 can reach low-mass, high-redshift clusters that larger surveys miss.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Simulation-based completeness/purity may be optimistic: the mock sky omits radio sources and uses the same UPP for injection and detection, and the pipeline predicts 26 detections vs 16 observed.","rationale":"The reader identified the UPP self-consistency and simulation fidelity as the weakest assumption; I agree that this is the correct area, but I would sharpen it to a concrete, testable discrepancy: the simulation predicts more detections than the data show, and the simulation sky omits a contaminant (radio sources) that visibly affects the real filtered map. This internal tension makes the quantitative completeness and mass-sensitivity claims less secure. However, the paper's primary existence claim — a blind tSZ search at 18 arcsec resolution yielding 16 candidates, 8 with optical/X-ray counterparts — does not collapse: the null-map random detection rate is only 2.36, and the counterparts are independently cataloged. The mass estimates are explicitly model-dependent and the data products are promised, so the correct disposition remains CONDITIONAL, not rejection. The reader's verdict already captures this balance; my concern reinforces it rather than changing it.","tokens_in":21291,"tokens_out":4148,"duration_ms":45106,"concrete_test":"Re-run the §4 pipeline on 1000 mock skies with two changes: (1) include a radio-source population at 150 GHz and (2) draw cluster pressure profiles from the Arnaud et al. (2010) parameter covariance or from an independent hydrodynamical simulation sample at M500 0.5–3×10^14 Msun and z 0.5–1.5. Compare the predicted detection counts and completeness at M500 = 2×10^14 Msun to the quoted values. If the mean count moves from 26 toward the observed 16, or completeness drops below 80%, the sensitivity claim must be re-stated with these systematic uncertainties.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central sensitivity claim — that NIKA2 reaches ~80% completeness at M500 > 2×10^14 Msun and can detect low-mass, high-z clusters — is validated almost entirely inside a simulation loop that uses the same fixed Arnaud et al. (2010) UPP for both injection and matched-filter detection. Because the filter is matched to the injected profile, completeness and purity are self-consistent by construction, not independently verified against real cluster profile scatter. This is not merely a philosophical worry: §4.1 reports 26 ± 5 predicted detections at S/N > 4 while the real map yields 16, a 2σ deficit. The simulations also omit radio sources and cross-correlations between components (§4.1), yet §3.3/Fig. 3 shows point-source-induced negative S/N tails in the actual filtered map. If the simulated sky is more optimistic than the real COSMOS sky, the completeness contours of Fig. 9 and the median mass M500 = 1.54×10^14 Msun quoted in §6.2 are upper-envelope values. The 8 matched counterparts support the existence of a signal, but the quantitative low-mass/high-z sensitivity claim is only as good as the sky model used to establish it.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a blind search for galaxy clusters via the thermal Sunyaev-Zel'dovich (tSZ) effect in the NIKA2 2 mm map of the COSMOS field, covering 877 arcmin^2 at 18.5 arcsec angular resolution. The authors reduce 195 hours of N2CLS data with a dedicated pipeline, measure the map transfer function, and apply a matched filter based on the Arnaud et al. (2010) universal pressure profile to extract cluster candidates. They report 16 candidates at S/N > 4, of which eight have optical or X-ray counterparts in the literature. Completeness and purity are characterized with 1000 simulated cluster catalogs plus null-map noise and SIDES point sources, yielding completeness above 80% for M500 > 2e14 solar masses and a median candidate mass of about 1.54e14 solar masses. The paper concludes that NIKA2 provides first blind tSZ cluster detection at this angular resolution and demonstrates sensitivity to low-mass, intermediate/high-redshift clusters.","tokens_in":21525,"tokens_out":6834,"duration_ms":69067,"significance":"If the quantitative sensitivity claims survive scrutiny, this is a significant technical milestone: it is a first blind tSZ cluster search at 18 arcsec resolution, with a candidate sample that occupies a region of the mass-redshift plane not probed by Planck, ACT, or SPT. The manuscript is not circular in the narrow sense: the cluster templates are externally motivated, the detection is applied to real maps, jackknife maps are used to check the noise, and the cross-match rates against external catalogs are explicitly compared with random association rates. Strengths include the transfer-function-corrected matched filtering, the 1000-map simulation suite, the public data products, and a careful external validation of the matched candidates. The central caveat is that the completeness, purity, and mass sensitivity are calibrated with simulations that inject the same universal pressure profile used for detection and that omit radio sources; the quantitative claim of low-mass/high-redshift sensitivity is therefore not yet independently anchored.","major_comments":[{"comment":"The quantitative sensitivity claims — 80% completeness for M500 > 2e14 solar masses, the completeness contours in Fig. 9, and the median mass M500 = 1.54e14 solar masses — rest on a simulation setup in which the injected cluster signal and the matched-filter template both use the Arnaud et al. (2010) universal pressure profile with fixed shape parameters and the same scaling relation (Eqs. 4-6). The filter is therefore matched to the injected population by construction, and the simulations do not test the effect of real cluster profile scatter. This concern is not merely formal: Sec. 4.1 reports 26 ± 5 predicted detections at S/N > 4 while the real map yields 16, a difference that is formally inside 2 sigma but is in the direction expected if the simulated sky is more optimistic than the real one, and the simulations explicitly omit radio sources and component cross-correlations. I ask the authors to quantify the robustness of the completeness and purity to (i) pressure-profile shape variations, for example mass-dependent or simulation-based profiles, (ii) a radio-source population in the mock sky, and (iii) the S/N distribution of detected peaks rather than only the total count; the completeness contours and the median mass should be revised if these tests change the detection rate.","section":"Sec. 4.1, Figs. 4 and 9"},{"comment":"The mass estimates presented in Table 1 and Fig. 9 are derived by fitting the same UPP model used for detection, with the Arnaud et al. (2010) scaling relation enforced through Eq. 6; as the authors acknowledge, these estimates are only valid within that model. The quoted uncertainties are therefore conditional on the UPP and do not include shape-parameter, scaling-relation, or hydrostatic-mass-calibration systematics. In support of the central claim that NIKA2 detects low-mass (M500 around 1-3e14 solar masses) intermediate/high-redshift clusters, at least for the eight matched candidates with redshift priors I ask for an external mass check using weak-lensing or X-ray hydrostatic estimates available in the COSMOS field, or alternatively an explicit statement in the abstract and conclusions that the low-mass sensitivity and median mass are model-dependent predictions rather than externally calibrated measurements.","section":"Sec. 6.2 and Table 1"},{"comment":"The candidate list mixes detections with and without external counterparts, and the paper does not quantify how many of the eight unmatched candidates could be produced by point-source ringing or noise. The random-detection estimate in Sec. 3.4 (2.36 detections per null map above S/N = 4) deliberately excludes point sources, yet Fig. 3 shows that the real filtered map has a pronounced negative S/N tail due to point sources, and Sec. 4.3 notes that point sources degrade purity. For the unmatched candidates, particularly the highest-S/N object NK2-CL J100045.8+020514.3, I ask for the local noise level, the angular distance to the nearest bright 2 mm source used in the masking step, and a point-source-only simulation estimate of the expected number of S/N > 4 peaks; without this, the statement that eight candidates have counterparts should be presented as an upper bound on the confirmed fraction rather than as the main validation of the sample.","section":"Sec. 3.3-3.4 and Sec. 5.1"}],"minor_comments":[{"comment":"The phrase 'we produced 104 simulated maps' appears to have lost a superscript; the intended number of simulated maps per catalog and the total number should be stated explicitly.","section":"Sec. 4.1"},{"comment":"In the last row of Table 1, the matching distance '14,8' uses a comma as a decimal separator; it should be '14.8' for consistency with the other rows.","section":"Table 1"},{"comment":"The caption 'Pixel negative S/N distribution' is confusing because the figure shows the full S/N distribution including both positive and negative tails; the caption should be reworded and the relationship between the 'negative S/N tail' due to point sources and the 'positive tail' attributed to cluster candidates should be clarified.","section":"Fig. 3 caption"},{"comment":"The sentence reporting Nharms = 8 and stating that this is equivalent to high-pass filtering above 1/3 Hz deserves a short justification or a reference, because the harmonic cutoff alone does not obviously define a 1/3 Hz high-pass filter.","section":"Sec. 2.2"},{"comment":"The 'semi-public spectroscopic redshift catalog' of Khostovan et al. (2025) is not described; since it is used to identify redshift peaks around candidates, the survey, selection function, and redshift range should be specified.","section":"Sec. 5.2"},{"comment":"The claim of a 'first blind detection at 18 arcsec' would benefit from one sentence stating the scope of the search and the absence of prior blind SZ cluster searches at this angular resolution, in order to make the priority claim precise.","section":"Abstract and Sec. 1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of A&A and the central detection result is credible; my recommendation of major revision is driven entirely by the need to make the simulation-based sensitivity calibration robust to the self-consistency concern described in the report. I do not see a citation or novelty problem, and the reader's skeptical concern about circular completeness estimates is, on reading the paper, a real load-bearing issue rather than a philosophical one."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the take: this is the first blind tSZ cluster detection at 18″ resolution, and that claim holds. The paper is a competent capability demonstration: careful data reduction, transfer function from white-noise simulations, null-map noise estimates, matched filtering with the standard UPP, and 1000 simulated sky realizations to characterize completeness and purity. They find 16 candidates at S/N>4, 8 with optical/X-ray counterparts, median mass around 1.5×10^14 Msun at z~0.74. The cutouts show real negative signal at 2mm. The result genuinely complements Planck/ACT/SPT in the low-mass, high-z regime.\n\nThe soft spots are real but not fatal. The completeness/purity analysis injects the same Arnaud+10 UPP that the matched filter assumes, so the 80% completeness at M500>2e14 is self-consistent by construction, not externally validated. More tellingly, the simulations predict 26±5 detections at S/N>4 while the real map yields 16—about a 2-sigma deficit. The mock sky also omits radio sources and inter-component cross-correlations. So the quantitative sensitivity contours are probably optimistic. The direction is right, but I'd quote them with caution. The 8/16 confirmation rate is not a flaw; at these S/N levels and with the stated purity, some unmatched candidates are expected. Still, the sample is a candidate list, not a catalog of confirmed clusters.\n\nThe paper is honest about the model dependence of the masses and about the contamination from point-source ringing. It also makes maps and catalogs available, though not analysis code. That is fine for this kind of paper.\n\nWho gets value from this: anyone working on tSZ cluster surveys, high-resolution SZ instruments, or the NIKA2 program. It deserves a serious refereeing; the results are reproducible in principle and the core detection claim is supported. I would accept it for peer review and require the authors to temper the completeness language, add the 26-vs-16 tension to the discussion, and release the candidate catalog in machine-readable form if not already. Recommend: engage.","headline":"First blind tSZ cluster search with NIKA2 at 18″ is a real capability milestone; the completeness/purity numbers are simulation-limited and likely optimistic, but the core result stands.","tokens_in":22416,"tokens_out":2163,"would_cite":true,"duration_ms":22744,"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":"The paper reports the first blind detection of galaxy clusters via the Sunyaev-Zel'dovich effect at 18.5-arcsecond resolution, finding 16 candidates in the COSMOS field.","keywords":["galaxy clusters","thermal Sunyaev-Zel'dovich effect","NIKA2","COSMOS field","matched filter","blind detection","millimeter astronomy","large-scale structure"],"falsifier":"Pointed follow-up of the eight unmatched candidates would settle the purity claim: if most show no galaxy overdensity, X-ray emission, or 2 mm decrement at the fitted position, the simulated ~60% purity at S/N > 5 overstates reality. A second decisive check is the paper's own flagged gap (Section 4.1): primary CMB power at $\\ell > 10^4$ is unmeasured and was extrapolated with a power law, so a direct measurement of small-scale CMB fluctuations would confirm that CMB confusion is indeed negligible in these 18.5-arcsecond maps.","tokens_in":21102,"feed_emoji":"🔭","tokens_out":20429,"duration_ms":164013,"temperature":0.7,"pith_summary":"This paper claims the first blind detection of galaxy clusters at millimeter wavelengths through the thermal Sunyaev-Zel'dovich effect at 18.5-arcsecond angular resolution, using 195 hours of NIKA2 2 mm data over $877\\,\\mathrm{arcmin}^2$ of the COSMOS field. A matched filter built on the universal pressure profile yields 16 candidates above signal-to-noise 4, eight of which coincide with previously known optical or X-ray clusters and groups. Simulation-based completeness and purity arguments place the sample above 80% complete for $M_{500} > 2\\times10^{14}\\,M_\\odot$ and give a median candidate mass near $1.5\\times10^{14}\\,M_\\odot$ at median redshift $z\\approx0.74$. If the detections are real, high-angular-resolution tSZ cameras can populate a region of the cluster mass-redshift plane that arcminute-resolution surveys cannot reach.","feed_headline":"First blind cluster hunt at 18.5 arcseconds finds 16 candidates","feed_subtitle":"The 18.5-arcsecond view reaches low-mass, high-redshift clusters that wide-field surveys miss.","key_machinery":"The matched filter is the load-bearing object: the optimal linear filter of Haehnelt & Tegmark (1996) applied to the single 2 mm map, with a cluster template built from the universal pressure profile of Arnaud et al. (2010), a generalized Navarro-Frenk-White model of the intracluster electron pressure with all shape parameters fixed and only mass and redshift free. Fifteen templates spanning $\\theta_{500}$ from $18.5''$ to $185''$ are scanned over the map, each convolved with the NIKA2 beam and corrected by the pipeline transfer function so that the large-scale filtering of the data reduction is modeled rather than ignored. The same filter is run on jackknife null maps for the noise statistics and on simulated maps that combine the Tinker et al. (2008) mass function, the SIDES dusty-galaxy sky model, and null-map noise to produce the completeness and purity curves; an MCMC fit to the same template then converts detections into $Y_{500}$, $\\theta_{500}$, and $M_{500}$ estimates.","core_discovery":"The central claim is that blind tSZ cluster detection works at sub-arcminute resolution: in a single 2 mm NIKA2 map of the COSMOS field, the authors find 16 cluster candidates with S/N > 4, of which eight match known clusters, and they argue from end-to-end simulations that such a survey is more than 80% complete for $M_{500} > 2\\times10^{14}\\,M_\\odot$ and reaches masses near $10^{14}\\,M_\\odot$ at $z > 1$. The matched candidates carry derived masses between about 1 and $3\\times10^{14}\\,M_\\odot$ at intermediate and high redshift, with median sample mass $M_{500} = 1.5^{+0.8}_{-0.3}\\times10^{14}\\,M_\\odot$; the highest-S/N candidate, NK2-CL J100045.8+020514.3, has no cataloged counterpart and is proposed as a follow-up target. The paper argues that at $18.5''$ scales primary CMB anisotropies are negligible while dusty galaxies remain point-like, so a single-frequency matched filter suffices, and that a dedicated survey with this camera would extend tSZ cluster catalogs into a low-mass, high-redshift regime where deviations from self-similarity can be tested.","pith_inferences":["If the unmatched candidates are confirmed, the sample's count of low-mass, high-redshift clusters could be compared against the Tinker mass function used in the simulations; a mismatch would indicate either a biased completeness model or new astrophysics, a test the paper does not perform.","Because the fixed universal pressure profile is the main systematic, measuring resolved pressure profiles of the confirmed candidates in X-rays would show whether the mass estimates are biased near $10^{14}\\,M_\\odot$ and would calibrate the profile shape at high redshift.","Applying the identical pipeline to the other N2CLS field, GOODS-N, would provide an independent reproducibility check: the detected counts and masses should match the COSMOS-based completeness predictions if the simulation framework is sound.","The single-band approach leaves radio source contamination as a residual; adding the 1.2 mm channel as a contaminant screen or masking radio catalogs more aggressively could raise purity above the reported values without much extra observing time."],"forward_implications":["A dedicated NIKA2 survey would add a new low-mass, high-redshift region to the mass-redshift plane covered by Planck, ACT, and SPT tSZ catalogs, because the compact beam limits dilution of distant clusters and keeps the completeness contours steep at high redshift.","The eight candidates without known counterparts, especially the brightest one, become concrete targets for optical, X-ray, and interferometric follow-up; confirming even a fraction of them would validate the purity model.","At the angular scales probed, primary CMB fluctuations are negligible, so cluster detection can proceed with a single 2 mm band and a matched filter rather than requiring multifrequency component separation.","The completeness threshold of roughly $M_{500} \\sim 2\\times10^{14}\\,M_\\odot$ at all redshifts is set by survey depth rather than resolution, so deeper integration in the same fields would push the detectable mass floor lower.","Mass estimates for the confirmed candidates place the sample at $M_{500}$ roughly $1$ to $3\\times10^{14}\\,M_\\odot$ and redshift $z\\approx0.24$ to $1.42$, a regime of interest for testing deviations from cluster self-similarity at low mass."],"supporting_citations":[{"why":"Supplies the universal pressure profile whose fixed shape parameters define the cluster template for both the matched filter and the mass estimates.","marker":"Arnaud et al. (2010)"},{"why":"Derives the optimal linear filter used to extract the cluster signal from the noise-dominated map.","marker":"Haehnelt & Tegmark (1996)"},{"why":"Provides the PyMF implementation of the matched filter that is run on the data and on the simulations.","marker":"Erler et al. (2019)"},{"why":"Establishes the matched-filter methodology for blind SZ cluster extraction that this work applies in single-band form.","marker":"Melin et al. (2006)"},{"why":"Mass function used to generate the mock cluster catalogs underlying the completeness and purity calculations.","marker":"Tinker et al. (2008)"},{"why":"The SIDES simulated infrared dusty extragalactic sky whose galaxies are injected into the simulated maps as point-source contamination.","marker":"Béthermin et al. 2017, 2022"},{"why":"Provides the 2 mm tSZ spectral factor that converts the Compton-parameter template into surface brightness.","marker":"Ruppin et al. (2018)"},{"why":"Documents the NIKA2 calibration baseline and instrument performance on which the data reduction builds.","marker":"Perotto et al. (2020)"},{"why":"Describes the data reduction pipeline that produces the NIKA2 maps and defines the transfer function used in the templates.","marker":"Ponthieu et al. (2025)"},{"why":"The PSZ2 arcminute-resolution tSZ sample against which the NIKA2 mass-redshift coverage is compared.","marker":"Planck Collaboration et al. (2016)"}],"fun_headline_variants":["First blind SZ cluster hunt at 18'' finds 16 candidates","NIKA2's sharp eye spots 16 galaxy clusters blindly","Sub-arcminute SZ survey reveals 16 cluster candidates","Blindly hunting clusters at 18 arcsec: 16 found","Sharpest blind cluster search yet nets 16 candidates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The pipeline assumes that the universal pressure profile of Arnaud et al. (2010), with shape parameters fixed from massive nearby clusters, describes the gas in the low-mass, high-redshift clusters being sought, and that simulated maps built from that model, null-map noise, and the SIDES galaxy sky faithfully reproduce the real COSMOS field.","fun_headline_variants_meta":{"raw":{"variants":["First blind SZ cluster hunt at 18'' finds 16 candidates","NIKA2's sharp eye spots 16 galaxy clusters blindly","Sub-arcminute SZ survey reveals 16 cluster candidates","Blindly hunting clusters at 18 arcsec: 16 found","Sharpest blind cluster search yet nets 16 candidates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000829,"raw_usage":{"total_tokens":3759,"prompt_tokens":1221,"completion_tokens":2538,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":837,"completion_tokens_details":{"reasoning_tokens":2448}},"tokens_in":837,"tokens_out":2538,"duration_ms":17110,"temperature":1.0,"reasoning_tokens":2448,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:51:40.796033+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Pointed follow-up of the eight unmatched candidates would settle the purity claim: if most show no galaxy overdensity, X-ray emission, or 2 mm decrement at the fitted position, the simulated ~60% purity at S/N > 5 overstates reality. A second decisive check is the paper's own flagged gap (Section 4.1): primary CMB power at $\\ell > 10^4$ is unmeasured and was extrapolated with a power law, so a direct measurement of small-scale CMB fluctuations would confirm that CMB confusion is indeed negligible in these 18.5-arcsecond maps.","supporting_citations":[{"cited_title":"W., Piffaretti , R., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the universal pressure profile whose fixed shape parameters define the cluster template for both the matched filter and the mass estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Derives the optimal linear filter used to extract the cluster signal from the noise-dominated map."},{"cited_title":"E., Basu , K., & Bertoldi , F","cited_arxiv_id":null,"evidence_quote":"Provides the PyMF implementation of the matched filter that is run on the data and on the simulations."},{"cited_title":"B., Bartlett , J","cited_arxiv_id":null,"evidence_quote":"Establishes the matched-filter methodology for blind SZ cluster extraction that this work applies in single-band form."},{"cited_title":"W., et al","cited_arxiv_id":null,"evidence_quote":"Provides the 2 mm tSZ spectral factor that converts the Compton-parameter template into surface brightness."},{"cited_title":"X., Beelen , A., et al","cited_arxiv_id":null,"evidence_quote":"Describes the data reduction pipeline that produces the NIKA2 maps and defines the transfer function used in the templates."}],"review_version":2}