{"id":"0bbc2e0d-4d00-4051-b20d-3cdaa413113c","arxiv_id":"2506.11601","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A systematic search in the STUDIES and COSMOS-Web fields yields 13 confirmed strongly lensed faint dusty star-forming galaxies and a 450um lensed number count about an order of magnitude fainter than previous measurements.","lead":"Using the deepest SCUBA-2 submillimeter maps and JWST imaging, astronomers found 13 faint dusty galaxies that are magnified by foreground galaxies acting as gravitational lenses. This extends measurements of the abundance of such lensed systems to fluxes about ten times fainter than earlier surveys, offering a new test of models of galaxy masses.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"All 13 confirmed lenses are in the COSMOS field, and the paper's own weak-lensing citation suggests STUDIES-COSMOS is overdense; without a field-to-field systematic the headline ~1% lensing fraction may not represent the average sky.","rationale":"The paper's central claim is a statistical one: a ~1% lensing fraction among faint DSFGs and counts extending about seven times fainter than previous limits, compared with model predictions. For that claim to hold as an average-sky constraint, the surveyed fields must be representative. All 13 confirmed lenses are in COSMOS; the STUDIES-COSMOS footprint is cited by the authors as possibly overdense based on weak-lensing E-mode shear (Massey et al. 2007), and zero lenses are found in UDS where roughly one to two are expected. No field-to-field systematic is incorporated into the quoted uncertainties. This is the most load-bearing concern because a biased high lensing optical depth directly inflates the headline fraction and the counts. The alternative concerns, visual-selection completeness and per-source dust confirmation, are real but either bias the counts in the opposite direction (incompleteness makes them lower limits) or are mitigated by the stacking analysis and the cross-check with Nightingale et al. (2025). The proposed test quantifies the foreground overdensity and rescales the counts; if the rescaled counts still match Negrello et al. (2017), the concern is retired. I agree with the reader's weakest assumption and keep the conditional verdict.","tokens_in":15287,"tokens_out":8007,"duration_ms":81801,"concrete_test":"Estimate the foreground overdensity in the STUDIES-COSMOS footprint using the COSMOS2020 catalog: compute the surface density of galaxies at the lens redshifts (z~0.2-1.4) within the footprint relative to the full COSMOS-Web area and to PRIMER-UDS. Then rescale the measured lensed counts by the inverse of the inferred optical-depth enhancement and re-compare with Negrello et al. (2017). If the rescaled counts remain within 1-sigma of the model, the COSMOS overdensity does not undermine the conclusion; if not, the counts should be treated as a COSMOS-specific measurement and the '~1%' fraction needs a field-to-field systematic error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The cumulative 450um counts in Section 3.4 are computed over the combined STUDIES footprint in COSMOS and UDS, yet all 13 confirmed lenses (and all 18 candidates) lie in the COSMOS field; none is found in PRIMER-UDS, where ~1-2 would be expected at the COSMOS density. The paper explicitly notes (end of Section 3.4) that the STUDIES-COSMOS footprint may cover an overdense foreground, citing enhanced E-mode shear (Massey et al. 2007). No field-to-field systematic error is assigned to the counts, and the lensing fraction of ~1% is quoted as an average-sky constraint. If the COSMOS line of sight has an above-average foreground mass density, the strong-lensing optical depth, and hence the measured counts and the ~1% fraction, is biased high relative to the cosmic mean. Because the central claim is a statistical constraint on the number density of strongly lensed faint DSFGs, this representativeness assumption is load-bearing; the agreement with Negrello et al. (2017) could be fortuitous.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a systematic search for strongly galaxy-galaxy lensed faint dusty star-forming galaxies (DSFGs) using the ultra-deep SCUBA-2 450um survey STUDIES in the COSMOS and UDS fields, combined with JWST NIRCam imaging from COSMOS-Web and PRIMER. The search combines visual inspection of JWST cutouts for lensing morphology with a red F444W/F150W color selection, followed by photometric redshift analysis with EAZY and lens modeling with Lenstronomy. The authors identify 18 candidate lensed systems in COSMOS, of which 13 are confirmed as strong lenses with best-fit magnifications >2. Using the 7 confirmed lensed STUDIES sources, they construct cumulative 450um number counts of strongly lensed DSFGs down to about 3.7 mJy, about seven times fainter than previous limits, and compare them with model predictions from Negrello et al. (2017) and Sedgwick et al. (2025). The measured lensing fraction is about 1% relative to unlensed sources, consistent with Negrello et al. (2017). The authors treat the color-selected counts as lower limits due to incompleteness and note a field-to-field variation, with all candidates in COSMOS and none in UDS.","tokens_in":15512,"tokens_out":7226,"duration_ms":67119,"significance":"If the statistical claims hold, this is the first faint-end constraint on the number density of strongly lensed dusty galaxies, extending previous work that was limited to bright Herschel, SPT, and Planck sources. The paper leverages the deepest SCUBA-2 450um data and JWST imaging to identify a sample of 13 lensed systems, 11 of which are new, with a clearly described methodology based on standard tools (EAZY, Lenstronomy). The authors are commendably transparent about known limitations: they explicitly label the color-selected counts as lower limits, discuss the possibility that the STUDIES-COSMOS field is overdense, and note the need for follow-up of two systems without ALMA data. The agreement with the Negrello et al. (2017) model across two orders of magnitude in flux is a useful test, though the current uncertainties are large. The paper also provides a concrete prediction for future Euclid/Roman searches. The main weaknesses are the unquantified completeness of the visual selection, the lack of a field-to-field systematic error, and the large source photo-z uncertainties that are not propagated into the magnification-based classification.","major_comments":[{"comment":"The cumulative number counts are computed over the combined STUDIES area in COSMOS and UDS, yet all seven strongly lensed STUDIES sources (and all 18 candidates) lie in COSMOS. The paper itself notes at the end of Section 3.4 that the STUDIES-COSMOS footprint may cover an overdense foreground, citing enhanced E-mode shear from Massey et al. (2007). No field-to-field systematic or cosmic-variance term is assigned to the counts in Table 2 or to the ~1% lensing fraction quoted in the Summary. Because the central claim is a statistical constraint on the average-sky number density, the counts and the comparison with Negrello et al. (2017) may be biased high if COSMOS is overdense, and the agreement could be fortuitous. I recommend either restricting the claim to the COSMOS field, or adding a systematic uncertainty that reflects the observed 0/7 (or 0/18) in UDS, for example a Poisson or field-to-field term.","section":"Section 3.4 and Table 2"},{"comment":"The candidate selection relies entirely on visual inspection ('morphology resembling lensing arcs or rings, or configurations that are consistent with red emission located near foreground large bluer sources'), with no quantitative selection function or completeness estimate for this step. The counts in Table 2 inherit the Gao et al. (2024) corrections for spurious fraction, flux boosting, and completeness for 450um source detection, but not for the efficiency of identifying which of those sources are strongly lensed. The small overlap with the Nightingale et al. (2025) sample (9 of 18 candidates) indicates that the visual screen is incomplete. If some strongly lensed STUDIES sources were missed, the cumulative counts are lower limits, not measurements. The color-selected counts are already treated as lower limits for this reason, but the STUDIES counts are not. Please quantify the completeness of the visual selection (e.g., by comparing with the Nightingale et al. sample within the STUDIES footprint) or explicitly label the STUDIES counts as lower limits.","section":"Section 3.1 and Section 3.4"},{"comment":"The source photometric redshifts have very large uncertainties (e.g., COSMOS-001: z=3.1 +3.1/-1.5; COSMOS-005: z=4.8 +0.9/-3.2), and the lensing magnification errors in Table 1 appear to reflect only the model fit, not the source redshift uncertainty. The strong-lensing classification (magnification > 2) is the basis for including sources in the counts. For systems with best-fit magnifications only slightly above 2 (COSMOS-005: mu=2.4; STUDIES-COS-195: mu=2.2; STUDIES-COS-232: mu=2.7), the classification could change if the source redshift is varied across its allowed range. The authors should either propagate the source photo-z posteriors into the magnification posterior or demonstrate that the mu>2 classification is robust to the source redshift uncertainty.","section":"Section 3.2 and Table 1"},{"comment":"Two systems without ALMA data, STUDIES-COSMOS-450-035 and STUDIES-COSMOS-450-218, are included in the counts even though their MIPS or VLA emissions appear closer to the foreground lens and foreground contamination cannot be definitively excluded; the paper states that follow-up observations are needed to rule out 450um emission from the foreground lenses. These two systems contribute to the faint-end cumulative counts in Table 2 (and to the ~1% lensing fraction). If either is a foreground 450um emitter, the counts and the lensing fraction decrease. I recommend presenting the counts with and without these two systems, or providing a quantitative estimate of the probability that they are genuine lensed background sources.","section":"Section 3.4 and Table 2"}],"minor_comments":[{"comment":"The abstract says the counts extend 'about an order of magnitude fainter' than previous measurements, while Section 3.4 says 'about seven times fainter'; please harmonize these statements.","section":"Abstract and Section 3.4"},{"comment":"The text refers to the PRIMER-UDS footprint as '10 times smaller' than COSMOS-Web, but the stated areas (255.88 arcmin^2 for PRIMER-UDS and ~0.54 deg^2 for COSMOS-Web) give a factor of about 7.6; please check and reconcile the numbers.","section":"Section 3.1 and Section 3.4"},{"comment":"The sentence 'The cut of 10 sigma in the threshold catalog corresponds to the F444W-to-F150W flux ratio of about 1' is confusing; it would help to explain explicitly how a 10-sigma cut on the F444W-F150W difference map translates to a flux ratio threshold.","section":"Section 3.1"},{"comment":"The visual selection criteria are described only qualitatively; even a brief statement of the number of inspectors and any inter-rater agreement would improve reproducibility.","section":"Section 3.1 and Figure 4"},{"comment":"The 1% lensing fraction is mentioned in the Abstract and Summary but never defined precisely; please state the flux range over which the ratio is computed and the unlensed counts used as the denominator.","section":"Section 3.4 and Figure 4"},{"comment":"The claim of agreement with model predictions within '+/-1-sigma uncertainty' is not quantified; please specify the uncertainty on the model comparison (e.g., the relevant error bar from Table 2).","section":"Section 4 (Summary)"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest about its main caveats, and the discovery sample is valuable. The load-bearing issue is that the central statistical claims (number counts, ~1% lensing fraction, consistency with Negrello et al. 2017) do not yet account for the unquantified completeness of the visual selection or the field-to-field variation that the authors themselves document. None of these problems appears unfixable: the authors could (1) add a field-to-field systematic or restrict claims to COSMOS, (2) use the Nightingale et al. (2025) catalog to estimate selection completeness, and (3) propagate source photo-z uncertainties into the magnification classification. I would be willing to review a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the first systematic attempt to count strongly lensed faint dusty galaxies at 450um, and the authors find 13 confirmed systems and a cumulative count that extends roughly seven times fainter than Herschel/SPT/Planck. The numbers are plausible and the paper is worth engaging with. The main thing to keep in mind: every confirmed lens sits in COSMOS, and the paper itself cites weak-lensing evidence that this line of sight is overdense. So the absolute ~1% lensing fraction is more fragile than the relative shape of the counts.\n\nWhat's actually new: they apply a known red JWST selection (Barger & Cowie, McKay) to the STUDIES and COSMOS-Web fields, then confirm candidates with Lenstronomy. That is not a new method, but the systematic application yields 13 confirmed lensed faint DSFGs, 11 of them new, and the first 450um number counts of lensed sources at these faint fluxes. The lens modeling is standard; they have spectroscopic redshifts for 16 of 18 foreground lenses, and they cross-check magnifications against Nightingale et al., which is a good sanity check. The paper is unusually candid: it calls the color-selected counts lower limits, flags the two STUDIES sources without ALMA data, and devotes a paragraph to the COSMOS overdensity.\n\nSoft spots, in order of importance. First, the all-COSMOS distribution is a real limitation. The counts use the combined COSMOS+UDS area while the lenses are all in COSMOS; if COSMOS is overdense, the lensing fraction is biased high. The authors say this, but they do not assign a field-to-field systematic error, so the quoted ~1% is really 'COSMOS plus zero from UDS.' That should be quantified or softened. Second, the candidate selection is visual and unquantified, which makes completeness hard to assess; the match with Nightingale helps but does not close it. Third, the source photo-z's have large uncertainties (some are essentially unconstrained), though the order z_source > z_lens holds for all. Fourth, two systems need ALMA to rule out foreground 450um emission; the authors know this.\n\nNone of this sinks the paper. The central result—a population of lensed faint DSFGs exists and their counts are roughly consistent with model predictions—holds up. But the absolute normalization should be treated as provisional. This is a good refereeing candidate: I'd send it out, ask for a field-to-field systematic or a clear statement that the counts are COSMOS-only, quantification of the visual selection, and ALMA follow-up on the two ambiguous sources. The paper is written for the submm-lensing community, and it will be useful as a reference point for Euclid/Roman forecasts.","headline":"A solid first measurement of faint lensed DSFG counts that deserves refereeing, with a field-to-field caveat the authors themselves flag.","tokens_in":16128,"tokens_out":4291,"would_cite":true,"duration_ms":38260,"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":"A deep submillimeter and JWST search discovers 13 strongly lensed faint dusty star-forming galaxies, yielding the first number counts of such systems at fluxes roughly seven times fainter than previously probed.","keywords":["strong gravitational lensing","dusty star-forming galaxies","submillimeter surveys","SCUBA-2 STUDIES","JWST NIRCam","number counts","galaxy evolution","COSMOS field"],"falsifier":"Measure the cumulative 450-micron counts of strongly lensed sources in an independent, comparably deep field, for example the full UDS or a future Euclid deep field. If, at the faintest flux bins near 3.7 mJy, the counts fall below the 1-sigma lower bounds reported here, the COSMOS overdensity is driving the signal. A second check is spectroscopic redshifts of the 13 background sources: if several turn out to lie in front of their putative lenses, the lensing interpretation collapses.","tokens_in":15108,"feed_emoji":"🔭","tokens_out":6241,"duration_ms":55167,"temperature":0.7,"pith_summary":"This paper aims to establish that strongly lensed dusty star-forming galaxies are not confined to the brightest sources found by Herschel and the South Pole Telescope; a systematic search finds a population of 13 faint ones. Using the deepest 450-micron SCUBA-2 survey (STUDIES) plus a JWST red-color selection, the authors confirm the systems with photometric redshifts and lens modeling, then construct cumulative number counts that reach about seven times fainter than earlier measurements. The measured lensing fraction of roughly 1% relative to unlensed sources at the same fluxes matches the predictions of a model that already reproduces the bright-end counts. If correct, this provides the first statistical handle on the faint end of strongly lensed dusty galaxies, an interval where models disagree and data were previously scarce.","feed_headline":"13 faint lensed dusty galaxies discovered","feed_subtitle":"The ~1% lensing fraction at faint fluxes matches predictions and opens a new statistical probe of dark matter.","key_machinery":"The selection pipeline combines three ingredients. The ultra-deep SCUBA-2 450-micron survey STUDIES provides the submillimeter flux base and a catalog of more than 400 faint dusty galaxies. A JWST red-color cut (F444W flux above 1 microjansky and F444W/F150W ratio above 3.5) selects candidate dusty sources across the wider COSMOS-Web footprint, with isophotal fluxes measured in a threshold catalog to preserve the colors of extended lensed arcs. Candidate systems are then confirmed by EAZY photometric redshifts, which place the sources behind their foreground lenses, and by Lenstronomy lens modeling using singular isothermal ellipse or elliptical power-law mass profiles, which yields magnifications above 2. Number counts are built with STUDIES-based simulation corrections for spurious fraction, flux boosting, and completeness.","core_discovery":"The central claim is that a population of 13 strongly galaxy-galaxy lensed faint dusty star-forming galaxies exists in the COSMOS field, with lensing magnifications above 2 confirmed by lens modeling and by source redshifts that exceed the foreground lens redshifts. From this sample the paper derives the 450-micron cumulative number counts of strongly lensed sources down to about 3.7 mJy, a flux limit roughly seven times fainter than the previous limit, and finds that the lensed fraction is about 1% of the unlensed population at the same fluxes. The counts agree within 1-sigma uncertainty with the model of Negrello et al. (2017), which also reproduces the bright lensed population, while a competing model prediction lies lower. No such systems are found in the comparable UDS field, which the authors attribute to field-to-field variation and possibly an overdense foreground in COSMOS.","pith_inferences":["If the COSMOS field really is a foreground overdensity, as the E-mode shear signal suggests, the sky-averaged lensing fraction may be below 1%; Euclid's wide-area search will provide the calibration.","The same red-color plus magnification selection can be replayed on existing and future JWST surveys to build a sample large enough for spectroscopic follow-up of the background sources' interstellar medium at cosmic noon.","The tension between the two model predictions highlighted in the paper implies that a modest expansion of the sample (tens of systems) will suffice to discriminate which model reproduces the faint lensed population."],"forward_implications":["The faint-end lensed number counts give a new, independent constraint on the abundance and mass distribution of foreground dark-matter halos probed at submillimeter wavelengths.","With a roughly 1% lensing fraction, wide-area surveys such as Euclid and Roman should discover tens of thousands of strongly lensed faint dusty galaxies, sharpening the statistical test.","Because the red-color selection is about 80% complete, the true lensed fraction may be somewhat higher than the measured value, so the reported counts serve as a lower limit for the color-selected portion.","The absence of lensed systems in the UDS field indicates that cosmic variance must be included in any single-field lensing statistic, and multi-field surveys will be needed to average over it."],"supporting_citations":[{"why":"Defines the STUDIES survey, the ultra-deep SCUBA-2 450-micron dataset that provides the submillimeter detections.","marker":"Wang et al. 2017"},{"why":"Supplies the final STUDIES images and the simulation-based method for spurious-fraction, flux-boosting, and completeness corrections used for the number counts.","marker":"Gao et al. 2024"},{"why":"Provides the JWST red-color selection criteria (F444W > 1 microjansky and F444W/F150W > 3.5) used to find the candidate dusty sources.","marker":"Barger & Cowie 2023"},{"why":"Verifies the completeness of the color-selection method, the basis for treating the color-selected counts as a lower limit.","marker":"McKay et al. 2025"},{"why":"Model prediction for the strongly lensed dusty-galaxy number counts that the faint-end measurements are compared against.","marker":"Negrello et al. 2017"},{"why":"Lenstronomy, the lens modeling software used to fit mass profiles and derive magnifications.","marker":"Birrer & Amara 2018"},{"why":"Pilot search in the STUDIES footprint that discovered one of the confirmed systems and established the feasibility of the approach.","marker":"Pearson et al. 2024"},{"why":"Previous discovery of the COSMOS-001 Einstein ring, one of the 13 systems, providing independent confirmation.","marker":"van Dokkum et al. 2024"},{"why":"Independent catalog of strong lenses in COSMOS-Web used to cross-check nine of the systems and validate the magnification estimates.","marker":"Nightingale et al. 2025"},{"why":"Weak-lensing analysis showing enhanced E-mode shear in the STUDIES-COSMOS footprint, the evidence behind the field-overdensity caveat.","marker":"Massey et al. 2007"}],"fun_headline_variants":["13 faint lensed dusty galaxies found in deep survey","JWST+SCUBA-2 uncover 13 faint lensed galaxies","New population of 13 faint lensed dusty galaxies","Lensing fraction ~1% seen in faint dusty galaxies","Counts pushed 7x fainter for lensed dusty galaxies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sample comes from a single field, COSMOS, and the paper assumes its lensing statistics represent the average sky; if COSMOS lies in an overdense foreground, the measured lensing fraction and number counts are biased high.","fun_headline_variants_meta":{"raw":{"variants":["13 faint lensed dusty galaxies found in deep survey","JWST+SCUBA-2 uncover 13 faint lensed galaxies","New population of 13 faint lensed dusty galaxies","Lensing fraction ~1% seen in faint dusty galaxies","Counts pushed 7x fainter for lensed dusty galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00028,"raw_usage":{"total_tokens":1686,"prompt_tokens":993,"completion_tokens":693,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":609,"completion_tokens_details":{"reasoning_tokens":607}},"tokens_in":609,"tokens_out":693,"duration_ms":7549,"temperature":1.0,"reasoning_tokens":607,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:04:11.662611+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the cumulative 450-micron counts of strongly lensed sources in an independent, comparably deep field, for example the full UDS or a future Euclid deep field. If, at the faintest flux bins near 3.7 mJy, the counts fall below the 1-sigma lower bounds reported here, the COSMOS overdensity is driving the signal. A second check is spectroscopic redshifts of the 13 background sources: if several turn out to lie in front of their putative lenses, the lensing interpretation collapses.","supporting_citations":[{"cited_title":"2017, , 850, 37, 10.3847/1538-4357/aa911b","cited_arxiv_id":null,"evidence_quote":"Defines the STUDIES survey, the ultra-deep SCUBA-2 450-micron dataset that provides the submillimeter detections."},{"cited_title":"J., & Cowie , L","cited_arxiv_id":null,"evidence_quote":"Provides the JWST red-color selection criteria (F444W > 1 microjansky and F444W/F150W > 3.5) used to find the candidate dusty sources."},{"cited_title":"The Physical Properties and Morphologies of Faint Dusty Star-forming Galaxies Identified with JWST","cited_arxiv_id":"2503.00102","evidence_quote":"Verifies the completeness of the color-selection method, the basis for treating the color-selected counts as a lower limit."},{"cited_title":"2024, , 527, 12044, 10.1093/mnras/stad3916","cited_arxiv_id":null,"evidence_quote":"Pilot search in the STUDIES footprint that discovered one of the confirmed systems and established the feasibility of the approach."}],"review_version":1}