{"id":"2ef673ef-3649-43ea-aee8-c74977d131a1","arxiv_id":"1908.01791","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A blind search for extra CO sources inside targeted PHIBSS2 observations yields 67 candidates and new CO luminosity functions that match previous blind surveys.","lead":"This paper searches 110 targeted observations of distant galaxies for extra, unplanned carbon monoxide (CO) signals and finds 67 candidate field galaxies. It uses these serendipitous detections to measure how much molecular gas galaxies contain at different cosmic epochs, showing that this repurposed-data approach agrees with dedicated blind surveys.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Counterpart-based redshift/transition assignment and exclusion of 36% of candidates leave the luminosity functions as unquantified lower limits; the claimed agreement with blind surveys may be coincidental.","rationale":"I read the paper's aim as demonstrating that serendipitous sources in targeted observations can reproduce field CO luminosity functions and gas density evolution. The method is creative, and the reliability/completeness machinery is a real contribution. The load-bearing step is the conversion of 67 candidates into a binned LF via Eq. 9. That conversion is only as good as the redshift/transition assignment from optical counterparts and the treatment of the 36% without counterparts. The paper is transparent about the lower-limit nature, but the abstract's 'very good agreement' treats the resulting points as measurements. The robustness of the agreement to the excluded sources and to photometric-redshift misassignment is not tested. This is not an accusation of error; it is the place where the argument is least secure. The proposed re-analysis is straightforward with the published tables and would settle whether the agreement is robust. I therefore move the recommended verdict from ACCEPT to CONDITIONAL: accept if the re-analysis confirms the published boxes; otherwise require qualification of the central claim.","tokens_in":31023,"tokens_out":17970,"duration_ms":208349,"concrete_test":"Recompute the published CO(2-1), CO(3-2), and CO(4-3) luminosity functions in Fig. 9 under three variants: (1) restrict to candidates with a single optical counterpart and, where available, spectroscopic/grism redshifts; (2) include all 67 candidates, assigning the 24 no-counterpart sources the transition/redshift distribution estimated from the 43 matched candidates (or from the primary-beam-weighted volume), and re-derive the binned Φ; (3) bootstrap the per-source weights by drawing R, C, Pa, and Pz from their uncertainty distributions instead of point estimates. Compare each variant to the published Fig. 9 boxes and to the COLDz/ASPECS LP constraints. If any luminosity bin shifts by more than the published 1σ Poisson range, or if the agreement with the blind-survey points is no longer within 1σ, the central claim requires qualification and acceptance should be conditional.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires that the CO luminosity functions in Fig. 9 are reliable field measurements. Eq. 9 builds them from per-source weights R_j/C_j Pa,j Pz,j, where transition/redshift come solely from the one-beam optical-counterpart match in §4.1.1 against 3D-HST/CANDELS photometric redshifts. Three unquantified systematics sit directly under the headline: (a) 24/67 candidates (36%) have no counterpart and are dropped, and the paper states the results are lower limits (§4.3); (b) for the 43 matched sources the EAZY posterior can permit several CO transitions, and Pa is a heuristic, not a normalized probability; (c) Fig. 6 shows most matched sources have CO-derived molecular masses 1–2 orders of magnitude above SFR-based estimates, an offset the paper attributes to obscured star formation but which is also the signature of a non-negligible fraction of associations or transition assignments being wrong. The error bars in Fig. 9 are Poisson-only (Gehrels 1986) and do not propagate C, R, Pa, or Pz. If the dropped or misassigned sources populate the faint end or specific redshift bins preferentially, the binned LF shape shifts and the 'very good agreement' with Riechers et al. (2019) and Decarli et al. (2019) could be fortuitous. The paper's use of that agreement to vouch for the counterpart IDs (§4.1.1) makes the external comparison doubly important.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a blind search for serendipitous CO line emitters in the 110 targeted PHIBSS2 data cubes of z~1-2 main-sequence galaxies. The authors develop a matched-filter line search with per-cube noise modeling, estimate candidate reliability from the statistics of negative noise peaks, and characterize completeness by injecting 2500 artificial sources per cube. They obtain a catalog of 67 candidates, find tentative 3D-HST/CANDELS counterparts for 43 (64%), assign CO transitions and redshifts via photometric-redshift posteriors, and construct CO(2-1) through CO(6-5) luminosity functions over z~0.5-5 plus a molecular gas mass density evolution measurement. The main astrophysical claim is that these serendipitous-source measurements agree well with previous blind-survey results (COLDz, ASPECS, PdBI HDF-N), demonstrating the utility of targeted observations for field CO luminosity function measurements.","tokens_in":31316,"tokens_out":6156,"duration_ms":58092,"significance":"If the central claim holds, the paper provides a valuable new route to field CO luminosity functions and gas density evolution by repurposing existing targeted observations, and the catalog of 67 candidates is a useful community resource. The methodological care is explicit: false-positive probabilities come from a parametric fit to negative noise peaks, completeness is measured via injection-recovery simulations with beam and primary-beam corrections, and volumes account for primary-beam attenuation. The multi-sightline approach also reduces cosmic variance relative to contiguous deep fields, as quantified with the Driver & Robotham (2010) formalism. These strengths are real and should be credited. However, the luminosity functions depend on photometric-redshift-based transition identification for 43 sources, with 24 sources excluded as unresolved without counterparts, so the quantitative agreement with earlier work is contingent on the reliability of those identifications.","major_comments":[{"comment":"The reported error bars in Figure 9 and Table 5 are Poissonian only (Gehrels 1986) and do not include uncertainties in the weights R, C, Pa, and Pz. Because the luminosity functions are weighted sums over these quantities, the systematic uncertainties—especially in the completeness correction C and the redshift probability Pz—can be comparable to or larger than the Poisson errors, particularly in the faintest bins. The 'very good agreement' with COLDz and ASPECS claimed in the abstract and §5 depends on these error bars. I recommend propagating all uncertainties (e.g., through a Monte Carlo or bootstrap over the noise-model fits, completeness fits, counterpart association, and redshift posteriors) or, at minimum, presenting the luminosity functions with an explicit systematic-error band and adjusting the agreement claim in the abstract accordingly.","section":"§4.3, Eq. (9), and §4.3.1"},{"comment":"The redshift/transition assignment for each serendipitous source relies on matching a single observed line frequency to CO(1-0) through CO(6-5) using the EAZY photometric-redshift posterior. For several sources (e.g., xc55-3, gn010-6, gn018-2, eg012-1) the posterior permits multiple transitions, and the adopted Pz is described qualitatively rather than defined precisely. The derivation should state the exact formula used for Pz and demonstrate that the resulting luminosity functions are insensitive to alternative, equally plausible posterior-weighting schemes, or show how the LFs change when restricted to sources with spectroscopic/grism redshifts. This is directly relevant to the CO(2-1) and CO(3-2) bins compared to COLDz and ASPECS in §5.1.1.","section":"§4.1.1 and Table 4"},{"comment":"The large (1-2 order-of-magnitude) offset between CO-based molecular gas masses and SFR-based masses for most matched sources is attributed to obscured star formation, but the same signature would be produced by a non-negligible fraction of misassociated counterparts or incorrectly assigned transitions. The paper uses the agreement of the derived luminosity functions with blind-survey results as evidence that misidentification is not the dominant effect, which is circular because the luminosity functions themselves are built from these identifications. An independent validation would strengthen the paper: for example, a comparison of the LF after removing sources with low R × Pa, a stacking analysis to confirm candidate lines at counterpart positions, or a demonstration that the subset with secure (spectroscopic/grism) redshifts alone reproduces the LF.","section":"§4.1.1, Figure 6, and §5.1.1"}],"minor_comments":[{"comment":"The coordinate for eg012-1 is given as '2d51m9.70s', presumably '52d51m9.70s'; please correct the typo.","section":"Table 4, EGS rows"},{"comment":"The statement that simulating sources 'at the edges of pixels and at the centers of pixels' has negligible impact would be clearer if the text distinguished between pixel edges and primary-beam edges, since the primary-beam sensitivity drop-off is treated separately in the volume calculation.","section":"§3.1.2, last paragraph"},{"comment":"Because the bins are 0.5 dex wide and stepped at 0.1 dex, adjacent points are strongly correlated; the text notes that every 5th bin is independent, but the caption should also state this so that readers do not treat adjacent points as independent.","section":"Figure 9 and Table 5"},{"comment":"'45 out of the 110 data cubes showing sources' followed by a catalog of 67 candidates is clear, but the sentence would benefit from explicitly stating that some cubes contain multiple candidates.","section":"Abstract and §1"}],"recommendation":"major_revision","confidential_remarks":"The paper is carefully executed and the catalog is a useful resource. My primary concern is the understated systematic uncertainties in the luminosity functions and the circularity of using the LF agreement to validate the counterpart identifications; both are addressable with additional analysis or more qualified claims. I would support publication after major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real contribution here is the 67-source serendipitous CO catalog and the demonstration that you can extract field-like CO luminosity functions and gas densities from targeted observations. The false-positive analysis using negative peaks is solid, the completeness injections are done cube by cube, and the luminosity functions are built with explicit per-source weights for reliability, completeness, association, and redshift probability. That is careful work, and the catalog alone is worth having.\n\nWhere I part ways with the reader's ACCEPT-with-confidence is on the strength of the headline claim. The luminosity functions are lower limits: 24 of 67 candidates (36%) have no optical counterpart and are dropped, and the paper says so. That is fine as a stated limitation, but the the error bars are Poisson-only and the weights Pa and Pz are heuristics, not calibrated probabilities. The transition assignment sits on photometric-redshift posteriors from EAZY, and the association probability is essentially a geometric distance weighting. None of that is propagated into the quoted uncertainties. So the \"very good agreement\" with COLDz and ASPECS is reassuring but not quantitatively decisive. The agreement could be fortuitous, especially if the dropped sources live at the faint end or in particular redshift bins. I also find it slightly circular to use that same agreement to vouch for the counterpart identifications in Section 4.1.1.\n\nThat said, the stress-test concern does not sink the paper. The authors are transparent about the lower-limit nature, they test for clustering bias around the primary targets, and the K-S tests on source properties do not reveal obvious pathologies. The Fig. 6 offset between CO-based and SFR-based masses is plausibly explained by obscured star formation, though it could also signal some misassociation. The paper would be stronger if it quantified how much the LFs shift under alternative transition assignments or if all no-counterpart sources were assigned within a plausible range, but that is a revision request, not grounds for rejection.\n\nThis is a paper for observers working on CO luminosity functions and molecular gas evolution, and for anyone thinking about mining archival interferometric data for serendipitous science. It deserves a serious referee; the method is sound and the catalog is a useful community resource. I would send it to review with a request that the authors either propagate the main systematics or explicitly quantify the range of plausible LF shifts from the 36% without counterparts.","headline":"A careful, genuinely useful mining of targeted PHIBSS2 cubes for serendipitous CO sources, with a catalog and luminosity functions that support the method but do not yet nail the quantitative agreement with blind surveys.","tokens_in":31862,"tokens_out":1269,"would_cite":true,"duration_ms":17625,"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 blind search inside targeted galaxy data cubes reconstructs field carbon-monoxide luminosity functions.","keywords":["galaxies: high-redshift","galaxies: evolution","galaxies: ISM","galaxies: luminosity function","CO luminosity function","molecular gas mass density","serendipitous sources","blind line search"],"falsifier":"Spectroscopically determine the redshifts of all 67 CO candidates, especially the 24 without optical counterparts; if a substantial fraction turn out to be real sources at transitions or redshifts outside the assumed CO(1-0) through CO(6-5) ladder, the luminosity functions and gas density points would move off the blind-survey curves.","tokens_in":30844,"feed_emoji":"🔭","tokens_out":19686,"duration_ms":170015,"temperature":0.7,"pith_summary":"This paper asks whether observations taken to study specific high-redshift galaxies can be mined for unrelated galaxies that happen to lie in the same field, turning a targeted program into a blind survey. Searching 110 targeted data cubes, it finds 67 candidate serendipitous carbon-monoxide (CO) sources, identifies tentative optical counterparts for about 64% of them, and uses the sample to build CO luminosity functions for five rotational transitions at median redshifts from about 0.7 to 3.6. Those luminosity functions and the derived cosmic molecular gas density evolution agree with results from dedicated blind surveys of deep fields. If this holds, targeted observations can be recycled as field surveys, greatly expanding the volume over which the molecular gas content of galaxies is measured at fixed observing cost.","feed_headline":"Stray CO detections in targeted cubes trace galaxy gas supply","feed_subtitle":"Mining 110 galaxy-targeted cubes for stray carbon-monoxide sources reproduces the universe's gas history.","key_machinery":"The central mechanism is a blind matched-filter line search over each spectral cube. Every cube is Hanning-smoothed and decimated to five velocity resolutions; the peak of each pixel spectrum is divided by a polynomial model of the local noise to form signal-to-noise maps. The noise side is calibrated empirically from the distribution of negative peaks, fit by an exponentially modified Gaussian, and each candidate is assigned a reliability $R = 1 - N_{\\rm expected}/N_{\\rm beams}$. Completeness comes from injecting 2500 artificial beam-shaped Gaussian sources per cube. The luminosity function estimator $$\\Phi(\\log L_i) = \\frac{1}{V}\\sum_{j=1}^{N_i} \\frac{R_j}{C_j} P_{a,j} P_{z,j}$$ then converts the weighted candidate counts into comoving space densities, dividing by the volume $V$ sampled by each transition, upweighting by completeness $C_j$, and downweighting by reliability, optical-counterpart association, and transition/redshift probability.","core_discovery":"The paper's claim is that serendipitous CO sources found inside targeted observations are field galaxies rather than satellites of the primary targets, and that their counts can be turned into measurements of the CO luminosity function and the cosmic molecular gas density. From 67 candidate sources in 110 data cubes, with per-source reliability, completeness, counterpart-association, and redshift-probability weighting, the paper constructs CO(2-1), CO(3-2), CO(4-3), CO(5-4), and CO(6-5) luminosity functions at median redshifts from about 0.7 to 3.6. It reports that these agree with earlier blind-search constraints, and interprets the agreement as evidence that the serendipitous sample is not biased toward the primary targets and that combining many independent lines of sight reduces cosmic variance. Because sources without optical counterparts are excluded, the luminosity functions are lower limits.","pith_inferences":["Inference: If the same pipeline were applied to other large targeted interferometric surveys, the serendipitous sample could grow to hundreds of sources, allowing finer luminosity-binning and a fainter reach than this 110-cube dataset.","Inference: Because the 36 percent of sources without optical counterparts are dropped, the true molecular gas density may be higher than the reported lower limits, especially if those sources are dust-obscured galaxies missed by rest-optical catalogs.","Inference: The paper's mass-SFR comparison implies that optical-to-8-micron SED fits underestimate star formation in CO-selected galaxies; a testable consequence is that adding far-infrared photometry should bring SED-inferred and CO-inferred gas masses into agreement for most counterparts.","Inference: A redshift-space cross-correlation of the serendipitous sources with the primary targets would give a quantitative check on the no-clustering assumption; a signal at $\\Delta z \\lesssim 0.1$ would require adding a clustering bias term to the luminosity function estimator."],"forward_implications":["Existing targeted CO surveys can be re-analyzed with the same blind-search pipeline to produce field CO luminosity functions, adding deep-field science without new observing time.","The 67-source catalog presents a CO-selected sample of galaxies with measured fluxes, line widths, reliabilities, and counterpart associations, available for follow-up study of gas-rich galaxies outside the original target selection.","Combining many independent sightlines keeps cosmic variance at roughly 13-18 percent, so serendipitous surveys can be competitive with contiguous deep fields for a small fraction of the observing time.","The CO luminosity functions at redshifts of about 0.7 to 3.6 tighten constraints on the evolution of the cosmic molecular gas density, including an apparent excess of high-J CO-bright galaxies over current theoretical predictions."],"supporting_citations":[{"why":"This earlier blind survey supplies a direct comparison point for the CO(2-1) and CO(3-2) luminosity functions.","marker":"Walter et al. 2014"},{"why":"This early blind-search program supplies pilot CO luminosity-function and gas-density constraints against which the serendipitous results are checked.","marker":"Decarli et al. 2016"},{"why":"This blind CO(1-0) survey at redshift 2.4 provides the anchor for comparing the CO(3-2) measurements after a line-ratio conversion.","marker":"Riechers et al. 2019"},{"why":"This large blind program provides the CO luminosity functions and molecular gas density evolution used for the main comparison.","marker":"Decarli et al. 2019"},{"why":"This paper defines the luminosity conversion used to turn measured CO fluxes, distances, and observed frequencies into CO luminosities.","marker":"Solomon et al. 1997"},{"why":"This paper supplies the brightness-temperature ratios used to convert higher-J CO luminosities to CO(1-0) luminosities.","marker":"Daddi et al. 2015"},{"why":"This work provides the depletion-time scaling relation and the survey context used to compare CO-derived gas masses with SFR-inferred gas masses.","marker":"Tacconi et al. 2018"},{"why":"This work supplies the empirical cosmic-variance estimator used to show that many independent targeted fields reduce cosmic variance.","marker":"Driver & Robotham 2010"}],"fun_headline_variants":["Serendipitous CO sources reveal gas history in targeted fields","Stray CO lines in 110 cubes map cosmic gas density","67 stray CO sources recover galaxy gas history","Secondary CO detections trace cosmic gas density evolution"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Each serendipitous source gets its redshift by being matched to an optical galaxy within one telescope beam and by assuming its detected line is one of six carbon monoxide rotational transitions; the 36 percent of sources with no optical match are excluded, so the measured luminosity functions are lower limits.","fun_headline_variants_meta":{"raw":{"variants":["Serendipitous CO sources reveal gas history in targeted fields","Stray CO lines in 110 cubes map cosmic gas density","67 stray CO sources recover galaxy gas history","Secondary CO detections trace cosmic gas density evolution"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000425,"raw_usage":{"total_tokens":2237,"prompt_tokens":1062,"completion_tokens":1175,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":678,"completion_tokens_details":{"reasoning_tokens":1110}},"tokens_in":678,"tokens_out":1175,"duration_ms":8854,"temperature":1.0,"reasoning_tokens":1110,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:03:08.463615+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Spectroscopically determine the redshifts of all 67 CO candidates, especially the 24 without optical counterparts; if a substantial fraction turn out to be real sources at transitions or redshifts outside the assumed CO(1-0) through CO(6-5) ladder, the luminosity functions and gas density points would move off the blind-survey curves.","supporting_citations":[{"cited_title":"2014, ApJ, 782, 79","cited_arxiv_id":null,"evidence_quote":"This earlier blind survey supplies a direct comparison point for the CO(2-1) and CO(3-2) luminosity functions."},{"cited_title":"2019, ApJ, 882, 138","cited_arxiv_id":null,"evidence_quote":"This large blind program provides the CO luminosity functions and molecular gas density evolution used for the main comparison."},{"cited_title":"J., Genzel, R., Saintonge, A., et al","cited_arxiv_id":null,"evidence_quote":"This work provides the depletion-time scaling relation and the survey context used to compare CO-derived gas masses with SFR-inferred gas masses."},{"cited_title":"P., & Robotham, A","cited_arxiv_id":null,"evidence_quote":"This work supplies the empirical cosmic-variance estimator used to show that many independent targeted fields reduce cosmic variance."}],"review_version":1}