{"id":"aa6824e0-96dd-4128-8567-a2b909689021","arxiv_id":"2506.03306","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A new large-area MIRI F770W survey and catalog from COSMOS-Web yield 7.7 micron number counts spanning 0.2 to 2300 microJy.","lead":"The COSMOS-Web team releases MIRI F770W imaging and a source catalog covering about 0.2 square degrees, reaching 5-sigma depths near 25.5 AB magnitude. The paper also presents 7.7 micron galaxy number counts across five orders of magnitude in flux, consistent with earlier JWST and Spitzer measurements.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Faint-end number counts likely depend on a flux-dependent effective area, which the paper assumes away.","rationale":"The paper is a careful, well-documented data-release and reduction paper, with several independent checks that support its central claims: the F770W versus IRAC CH4 comparison has only a small offset, the SE++ versus SE AUTO photometry shows good consistency, and the counts broadly agree with CEERS, SMILES, and JADES within the quoted uncertainties. Those checks give real confidence that the catalog and reductions are largely sound. The weakest load-bearing step is the number-count computation at the faint end. The text in Section 5.3 explicitly assumes the effective area is independent of flux, while the depth map in Table 1 shows a nontrivial shallow component. The faintest bin at 0.22 µJy lies below the 5σ depth of the two-exposure areas, so those regions cannot contribute at the same completeness as the rest of the survey. If the counts use the full geometric area without a flux-dependent correction, the faintest point is biased low by an amount comparable to the quoted cosmic variance. This does not overturn the central claim, because the affected point is one of twenty-two and the qualitative agreement with other surveys is dominated by brighter bins, but it does mean the 'robust five orders of magnitude' statement needs either a flux-dependent area correction or an explicit demonstration that the effect is below ~5%. The reader's conditional verdict already captures this, so no change in verdict is needed. Separately, Table 3's units appear mislabeled: values such as 0.03 at 1602 µJy are unphysical if interpreted as 10^-6 sr^-1, and the header likely should read 10^6 sr^-1. This is a presentation issue that should be fixed but is less scientifically load-bearing than the effective-area assumption.","tokens_in":25893,"tokens_out":15966,"duration_ms":208180,"concrete_test":"Recompute N(>S) for the first three bins (0.22, 0.37, 0.58 µJy) using a flux-dependent effective area built from the per-visit completeness functions and the exposure-depth zones in Table 1, e.g., A_eff(S) = A_2exp*C_2exp(S) + A_4exp*C_4exp(S) + A_6/8exp*C_6/8exp(S), instead of the full 683 arcmin^2. If the 0.22 µJy bin shifts by more than ~10% (≈1.5 times the adopted 6.5% cosmic variance), the uniform-area assumption fails and the faint-end counts must be re-derived before comparing with CEERS/SMILES.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.3 states that the effective survey area is treated as uniform across flux bins because ~85% of the MIRI footprint reaches the 25.5 mag depth. Table 1, however, shows that 89 arcmin^2 (13%) of the 683 arcmin^2 footprint has only two exposures and a 5σ depth of 25.15 mag (~0.32 µJy). For the faintest cumulative bin at 0.22 µJy, sources in this shallow region sit at only ~3.5σ, so their completeness is well below the 80% threshold used to define the usable sample. If counts are divided by the full geometric area rather than by A_eff(S) = Σ_i A_i C_i(S), the effective area at 0.22 µJy is overestimated and N(>S) is underestimated by roughly 10%. The manuscript neither constructs a flux-dependent effective area from the exposure/weight map nor demonstrates that the residual bias is small compared with the 6.5% cosmic variance added in quadrature. This affects the faintest decade of the claimed five-order-of-magnitude range and the flux regime where agreement with CEERS and SMILES is assessed.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents the data reduction, catalog construction, and source number counts from the MIRI F770W parallel imaging in the COSMOS-Web survey. The authors describe a custom background subtraction that suppresses large-scale gradients, measure survey depths as a function of exposure count, construct an SE-based catalog, validate photometry against IRAC CH4, and derive completeness-corrected 7.7 μm cumulative number counts over roughly 0.2–2300 μJy. They compare the counts to other JWST surveys and to galaxy formation models.","tokens_in":26165,"tokens_out":8038,"duration_ms":83485,"significance":"If the results hold, this is a valuable dataset and reference: it is the largest MIRI F770W imaging program to date, with a detailed reduction description and a public data release. The number counts span five orders of magnitude at 7.7 μm, beyond any other JWST survey, and provide a useful cross-check of model predictions. The photometric comparison with IRAC is an independent validation. The analysis is largely standard and transparent, with empirical completeness simulations and clearly reported statistical uncertainties. However, the issues discussed below (effective-area uniformity and a likely units error in the counts table) affect the interpretation of the faint end and the reported numbers.","major_comments":[{"comment":"The assumption that the effective survey area is uniform across flux bins is not supported by the depth map. Table 1 shows that 89 arcmin^2 (13% of the 683 arcmin^2 footprint) has only two exposures with a 5σ depth of 25.15 AB (≈0.32 μJy), whereas the faintest cumulative bin in Table 3 reaches 0.22 μJy. In this shallow region, sources at 0.22 μJy lie below the 5σ threshold, so their completeness is far below the 80% cutoff applied for the usable sample. If the counts are normalized by the full geometric area instead of a flux-dependent effective area A_eff(S) = Σ_i A_i C_i(S), the faintest bins will be biased low by roughly 10% or more. This exceeds the 6.5% cosmic variance added in quadrature and affects the claimed five-order-of-magnitude range and the agreement with CEERS/SMILES at Sν ≲ 1 μJy. Please recompute the counts using the exposure/weight map or demonstrate that the residual bias is negligible.","section":"Sec. 5.3, Table 1"},{"comment":"The cumulative counts in Table 3 appear to have a units error. With the header '(10^-6 sr^-1)', the first entry N(>0.22 μJy) = 705.76 × 10^-6 sr^-1 = 7.06 × 10^-4 sr^-1, which multiplied by the survey solid angle (~6.1 × 10^-5 sr for 0.2 deg^2) yields about 4 × 10^-8 sources, an impossible value. If the unit in the header were '10^6 sr^-1', the first entry would correspond to roughly 4.3 × 10^4 sources over the survey, which is plausible. Please correct the header and verify all entries and error bars.","section":"Table 3"},{"comment":"The reported uncertainties combine only Poisson and cosmic-variance terms. The completeness correction, photometric zero-point, and the effective-area treatment introduce systematic uncertainties that are not captured in Table 3. In particular, the faintest bins may be affected by the area issue discussed above, and the quoted errors (e.g., ±3.4 on 705.76) appear to be purely statistical. The authors should quantify and include systematic error contributions, or at least discuss their expected magnitude, before the counts can be considered robust over the full stated flux range.","section":"Sec. 5.3.1 and Table 3"}],"minor_comments":[{"comment":"The median astrometric offset in RA is stated as 0.53 mas in the text but 0.35 mas in the Figure 2 caption; please reconcile the two values.","section":"Sec. 3.3 and Figure 2"},{"comment":"The IRAC CH4 magnitude limit is 22.5 in the text and 22.3 in the figure caption; please unify the value and definition.","section":"Sec. 5.1 and Figure 6"},{"comment":"The total MIRI area is given as 722 arcmin^2 in the text, but the areas in Table 1 sum to 683 arcmin^2; please clarify what the additional area corresponds to (e.g., gaps or the coronagraphic field).","section":"Sec. 2 and Table 1"},{"comment":"The abstract quotes the depth for 0.3'' circular apertures, while Section 3.4 uses 0.27'' radius (FWHM) apertures; please ensure the quoted aperture size is consistent throughout.","section":"Sec. 3.4 and Abstract"},{"comment":"The completeness simulations inject only point sources; while the justification is reasonable, the authors could state the expected magnitude of the bias for marginally extended sources at intermediate fluxes.","section":"Sec. 4.3"},{"comment":"The fitted values of the noise-model parameters α and β are not reported; please provide them for reproducibility.","section":"Sec. 4.1, Eq. (1)"},{"comment":"The sentence about repeat visits #154 and #167 is confusing and appears to contain a typo; please rephrase to clarify which visits share reference positions.","section":"Appendix note"}],"recommendation":"major_revision","confidential_remarks":"The paper is an important data release and the analysis is careful overall. The central concern raised in review is the flux-dependent effective area; the authors have the exposure and weight maps to implement the correction. The units typo in Table 3 is easily fixed. With these changes, the paper would be suitable for publication. I have no concerns about the novelty or scope; the data and catalog are valuable for the community. The reader's stress-test about the effective-area assumption is on point and should be taken seriously."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a workmanlike data-release paper that should be accepted with minor revisions. The genuinely new product is the COSMOS-Web MIRI F770W mosaic and catalog over ~0.2 deg^2, the largest MIRI area from JWST Cycle 1, plus 7.7 um number counts spanning ~0.2-2300 uJy. The reduction is standard (calwebb plus a master background subtraction similar to Yang et al. and Perez-Gonzalez et al.), and the depth measurements, point-source completeness simulations, and cross-checks with IRAC CH4 and prior JWST counts are all carefully done. The counts agree with CEERS, JADES, SMILES, and SDWFS within uncertainties, which is what you want from a survey paper.\n\nThe biggest issue is Table 3. The header says 10^-6 sr^-1, but the values are clearly in 10^6 sr^-1. The Poisson error on the first row (705.76 ± 3.4) matches the expectation for ~43,000 sources in 0.2 deg^2, so the unit is a typo. It should be fixed, but it is not a scientific problem.\n\nThe stress-test worry about a flux-dependent effective area is, on reading the paper, not a real bias. Section 5.3's statement that the coverage is 'largely uniform across visits' is poorly worded, but the completeness simulations were run visit by visit and then merged. That means the completeness curve is effectively the area-weighted average over the different exposure depths, so dividing by the total area with that average completeness is equivalent to using A_eff(S) = Σ A_i C_i(S). The text should explain this instead of hand-waving about uniformity. If the authors want to be extra cautious, they could show the effective area as a function of flux, but the method is sound.\n\nThe bright end of the counts is limited by small-number statistics, as acknowledged. The model comparison is fine but not deep. The photometric redshift distribution is a useful addition.\n\nThis paper is for anyone using the COSMOS-Web MIRI imaging for galaxy evolution, PAH studies, or AGN work. It deserves a serious referee and will likely become a standard reference for the data release. My recommendation is accept after minor revisions: fix the Table 3 units, clarify the completeness/area argument in Section 5.3, and perhaps add a sentence on the bright-end uncertainty.","headline":"Solid COSMOS-Web MIRI data-release paper; fix Table 3 units and clarify the uniformity language, then accept.","tokens_in":26783,"tokens_out":10969,"would_cite":true,"duration_ms":121145,"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":"COSMOS-Web MIRI F770W data release produces 7.7 micron number counts from 0.2 to 2300 microJy that agree with other JWST and Spitzer surveys.","keywords":["JWST MIRI","F770W","number counts","COSMOS-Web","mid-infrared galaxies","data reduction","source catalog","7.7 micron"],"falsifier":"Recompute the faint-end counts using only regions with four or more exposures and compare them with the fiducial counts; if the faintest flux bins shift by more than the quoted Poisson-plus-cosmic-variance errors, the uniform-area assumption is the cause.","tokens_in":25738,"feed_emoji":"🌌","tokens_out":7943,"duration_ms":73220,"temperature":0.7,"pith_summary":"This paper presents the MIRI F770W component of the COSMOS-Web survey, which maps roughly 0.2 square degrees of the COSMOS field in the mid-infrared with a single broad filter. The authors aim to show that their reduction, built on the standard JWST calibration pipeline plus a custom background subtraction, produces a reliable photometric catalog and galaxy number counts at 7.7 microns. They reach a 5-$\\sigma$ point-source depth near $m_{F770W}\\sim25.51$ AB mag, better than pre-flight predictions, and find F770W fluxes agree with Spitzer IRAC CH4 measurements to a median offset below 0.05 mag. Their completeness-corrected cumulative counts span about 0.2 to 2300 microJy and agree with other JWST surveys within 1 $\\sigma$, with the largest area coverage of any JWST mid-infrared survey to date. These counts matter because they connect the faint JWST-selected population to brighter Spitzer-era measurements in a single field, and they provide a new mid-infrared constraint on galaxy formation models.","feed_headline":"COSMOS-Web: 7.7-micron counts span five decades of flux","feed_subtitle":"The largest JWST survey's MIRI catalog matches Spitzer and other JWST counts across a 0.2 deg2 field.","key_machinery":"The load-bearing mechanism is the combination of the custom 'master background subtraction' step and the completeness-corrected number-count construction. The background step builds a time-resolved sky model from contemporaneous, source-masked exposures and subtracts it from each image, which is what allows the survey to reach depths roughly 0.7-0.8 mag better than exposure-time-calculator predictions. The number counts rest on Source Extractor photometry with empirical PSF aperture corrections and on injected-source completeness simulations that define the 80 percent completeness limit at the median 5-sigma depth.","core_discovery":"The central claim is that the COSMOS-Web MIRI F770W reduction and catalog yield robust 7.7 micron number counts spanning five orders of magnitude in flux density, roughly 0.2 to 2300 microJy, and that these counts agree with estimates from other JWST surveys within their uncertainties while slightly underpredicting IRAC-based counts on the bright end. The claim is established by constructing a MIRI-selected catalog from F770W mosaics, applying empirical PSF aperture corrections, measuring completeness with injected-source simulations, and adding a cosmic-variance term of about 6.5 percent in quadrature to the Poisson errors. The authors explicitly note that COSMOS-Web is the only JWST survey to date that efficiently samples such a wide flux range over a large contiguous area.","pith_inferences":["We infer that the faintest bins of the number counts could be tested by recomputing them using only the deeper four-or-more-exposure area; if the two-exposure region contributes a disproportionate share of faint sources, the stated uniform-area assumption would not hold.","The same reduction and completeness machinery could be applied to the overlapping PRIMER MIRI coverage, effectively doubling the surveyed area and providing an independent check on the 6.5 percent cosmic-variance estimate.","The MIRI-selected catalog may reveal a population of mid-infrared-only sources whose optical and near-infrared counterparts are too faint for reliable redshift fitting; the authors note this as a direction for future work."],"forward_implications":["The 7.7 micron counts provide a single contiguous-field census from about 0.2 to 2300 microJy, bridging faint JWST surveys and bright Spitzer-era counts.","The released mosaics and MIRI-selected catalog let other teams study source populations in the COSMOS field that are faint or absent in NIRCam imaging.","The agreement with the SHARK and SPRITZ model predictions, and the mild overprediction of faint counts by some other models, gives galaxy-formation models a new mid-infrared constraint.","The small median offset with IRAC CH4 photometry supports the independent calibration of the two instruments and offers a cross-check for filter-dependent SED effects."],"supporting_citations":[{"why":"Defines the COSMOS-Web survey design, area, and observing strategy that the MIRI component extends.","marker":"Casey et al. 2023"},{"why":"Provides the CEERS F770W number counts used for comparison and a precedent for super-background subtraction in MIRI data.","marker":"Yang et al. 2023b"},{"why":"Supplies the SMILES and JADES F770W counts compared against the COSMOS-Web counts, including the faint-end cosmic-variance discussion.","marker":"Stone et al. 2024"},{"why":"Provides the COSMOS2020 catalog whose IRAC CH4 photometry is used for the F770W photometric comparison.","marker":"Weaver et al. 2022"},{"why":"Supplies the Spitzer SDWFS IRAC CH4 number counts used for the bright-end comparison.","marker":"Ashby et al. 2009"},{"why":"Provides the empirical MIRI PSF models used to compute aperture corrections for the F770W photometry.","marker":"Libralato et al. 2024"},{"why":"Supplies the ComEst package used to run the injected-source completeness simulations.","marker":"Chiu et al. 2016"},{"why":"Supplies the Source Extractor code used for source detection and photometry in the MIRI catalog.","marker":"Bertin & Arnouts 1996"},{"why":"Provides the empirical relation used to estimate the cosmic-variance contribution to the number-count errors.","marker":"Driver & Robotham 2010"}],"fun_headline_variants":["COSMOS-Web MIRI: 5 decades of 7.7-µm counts","JWST MIRI 7.7-µm counts: five decades from COSMOS-Web","7.7-µm counts from COSMOS-Web match IRAC and JWST","Five decades of 7.7-µm flux from COSMOS-Web MIRI"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The number counts assume the survey's effective area is the same for every flux bin, even though about 15 percent of the area has only two exposures and is shallower.","fun_headline_variants_meta":{"raw":{"variants":["COSMOS-Web MIRI: 5 decades of 7.7-µm counts","JWST MIRI 7.7-µm counts: five decades from COSMOS-Web","7.7-µm counts from COSMOS-Web match IRAC and JWST","Five decades of 7.7-µm flux from COSMOS-Web MIRI"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001436,"raw_usage":{"total_tokens":5854,"prompt_tokens":1075,"completion_tokens":4779,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":691,"completion_tokens_details":{"reasoning_tokens":4679}},"tokens_in":691,"tokens_out":4779,"duration_ms":31719,"temperature":1.0,"reasoning_tokens":4679,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:06:57.860063+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the faint-end counts using only regions with four or more exposures and compare them with the fiducial counts; if the faintest flux bins shift by more than the quoted Poisson-plus-cosmic-variance errors, the uniform-area assumption is the cause.","supporting_citations":[{"cited_title":"2016, Astronomy and Computing, 16, 79, doi: 10.1016/j.ascom.2016.04.005","cited_arxiv_id":null,"evidence_quote":"Supplies the ComEst package used to run the injected-source completeness simulations."}],"review_version":1}