{"id":"ab6c69ce-f108-4ace-9a9c-521c9cca6717","arxiv_id":"2608.12708","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"ALMA follow-up reveals that most HST/Spitzer-selected z~10 candidates are low-redshift interlopers, with one confirmed as a z=2.54 ULIRG via CO(9-8), and stricter colour thresholds improve high-z sample purity.","lead":"Six pre-JWST z~10 galaxy candidates were observed with ALMA; one turned out to be a dusty ultra-luminous infrared galaxy at z=2.54 with a CO(9-8) line, and the rest appear to be lower-redshift interlopers. The paper also shows that stricter Lyman-break colour cuts are needed to build reliable high-z samples for future wide surveys.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The CO(J=9-8) identification and ULIRG classification hinge on an unpublished NIRSpec redshift; without that spectrum the central line claim is not independently verifiable.","rationale":"The reader and I identify the same most load-bearing weakness. The paper's central new claim is the CO(J=9-8) line from a ULIRG at z=2.54, and every quantity built on that line (L'_CO from Eq. 1, the L_FIR-L'_CO relation in Fig. 3, and the ULIRG classification) requires the line to be real and correctly assigned. The ALMA evidence alone is a 4.5-sigma line with a 4.5% false-positive probability from the clump search; the assignment as CO(9-8) is made because an unpublished NIRSpec redshift is claimed. This is a verifiability gap rather than an internal inconsistency: the NIRCam SED fit gives z=2.56, independently supporting the low-z nature of UDS_18697, and the 30-sigma continuum detection makes the source real. But the photometric redshift cannot identify the line transition, and the line flux itself has only moderate significance. The secondary magnification issue (mu=1.35 left uncorrected, which would place L_FIR near 0.8e12 L_sun, below the ULIRG threshold) reinforces that the ULIRG label is less secure than the abstract implies, though it alone would not change the verdict. The interloper conclusions and the colour-threshold analysis are more robust: they are supported by NIRCam photometry, SED fitting, and the JADES/CANUCS spectroscopic samples, and they would likely stand even if the CO identification were revised. The appropriate verdict thus remains CONDITIONAL, pending access to the NIRSpec spectrum, exactly as the reader concluded.","tokens_in":27201,"tokens_out":9061,"duration_ms":91158,"concrete_test":"Obtain or independently reduce the GO1758 NIRSpec/PRISM spectrum of UDS_18697 and measure the redshift from the detected emission lines; then test whether the 293.27 GHz ALMA line matches the expected CO(J=9-8) frequency at that redshift within one line FWHM. If the independent z_spec is not 2.54, or the frequency match fails, the CO(J=9-8) identification, L'_CO, and the placement in Figure 3 are unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline result—a CO(J=9-8) line from a ULIRG at z=2.54 (Sections 3.2.1 and 3.3; Eq. 1; Fig. 3)—is anchored by the statement that JWST/NIRSpec PRISM from GO1758 (R. Larson et al. in prep.) gives z=2.54. That spectrum is not shown, no line list or signal-to-noise is given, and the only quoted consistency is the ALMA line frequency itself, which corresponds to z_CO=2.5356±0.0004. The ALMA detection is marginal on its own: a 4.5-sigma peak, line-flux S/N of about 2.9, a 0.16 arcsec offset larger than the expected 0.10 arcsec astrometric accuracy, and a 4.5% clump-search false-positive probability (Appendix B). Without an independently accessible NIRSpec redshift, the 293.27 GHz line could be a noise fluctuation or a different transition; the CO(9-8) assignment and L'_CO are therefore unverified. The NIRCam SED fit in Section 4 gives z_NIRCam=2.56+0.19-0.30 and independently supports the low-z interloper conclusion, but it cannot by itself assign the ALMA line to CO(9-8). A secondary issue reinforces the fragility of the ULIRG label: Table 1 lists a magnification mu=1.35 for UDS_18697, and because the luminosities are not corrected, L_FIR=1.1e12 L_sun would become about 0.8e12 L_sun after dividing by mu, below the usual 1e12 L_sun ULIRG threshold.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports ALMA Band 7 observations of six pre-JWST z~10 galaxy candidates, targeting the [OIII] 88 micron line. A 4.5-sigma line and a 29.9-sigma dust continuum are detected in UDS_18697; the line is identified as CO(J=9-8) at z=2.5356 because unpublished JWST/NIRSpec PRISM observations reportedly give z=2.54. The authors classify UDS_18697 as a ULIRG with L_FIR=1.1e12 L_sun, using a dust temperature from Sommovigo et al. (2022) and a fixed beta=2.0, and place it on the L_FIR-L'_CO relation. SED fitting with new NIRCam photometry gives z~2-2.5 for three objects, and NIRSpec results (in prep.) indicate low redshifts for two more; J2140+0241 remains unidentified. Finally, using JADES and CANUCS spectroscopic catalogues, the authors propose that raising the Lyman-break colour threshold F115W-F150W from 0.5 to 1.0 increases the purity of z>8.5 samples from ~45% to ~75%, with implications for Euclid, Roman, and GREX-PLUS.","tokens_in":27517,"tokens_out":6819,"duration_ms":63423,"significance":"If the results hold, the paper provides a quantitative, empirically grounded caution about contamination in wide-field high-redshift galaxy surveys and a concrete selection guideline. The use of independent spectroscopic catalogues for the colour-purity analysis and the explicit clump-search false-positive estimate are strengths. However, the headline CO(J=9-8) identification and the consequent ULIRG classification rest entirely on an unpublished NIRSpec spectrum; the ALMA detection alone is marginal. The magnification-uncorrected luminosities further weaken the ULIRG classification. These issues affect the central claims and must be addressed before the paper can be considered archival.","major_comments":[{"comment":"The identification of the 293.27 GHz line as CO(J=9-8) rests entirely on the unpublished JWST/NIRSpec redshift of z=2.54 (R. Larson et al., in prep.). The ALMA data alone are marginal: peak S/N of 4.46, line-flux S/N of ~2.9, a 0.16 arcsec offset versus the expected 0.10 arcsec astrometric accuracy, and a 4.5% false-positive probability from the clump search in Appendix B. Without the NIRSpec data, the line could be a noise fluctuation or a different transition. The manuscript should include the NIRSpec redshift information (line list, signal-to-noise, or spectrum) or, failing that, explicitly state that the CO(9-8) identification and all derived quantities (L'_CO, L_FIR-based ULIRG classification, position on the L_FIR-L'_CO diagram) are tentative pending publication of the NIRSpec results.","section":"Section 3.2.1 and 3.3; Eq. (1)"},{"comment":"The ULIRG classification is not robust against the uncorrected magnification factor mu=1.35 listed in Table 1. The paper quotes L_FIR=1.1e12 L_sun with mu uncorrected; dividing by mu=1.35 gives about 0.8e12 L_sun, below the conventional 1e12 L_sun ULIRG threshold. Moreover, the reported 1-sigma range (L_FIR=1.1+1.0-0.3 x 10^12 L_sun) already extends below 1e12 L_sun. The authors should either correct the luminosities for magnification (quantifying the uncertainty in mu) or restate the ULIRG classification as conditional on mu<1.35, and should adjust the abstract and conclusion accordingly.","section":"Table 1 and Section 3.3"},{"comment":"The quoted CO line luminosity, L'_CO(J=9-8) = 1.53 ± 0.54 K km s^-1 pc^2, is missing a factor of 10^9. Using Eq. (1) with the listed line flux (0.397 Jy km/s), D_L at z=2.536, and nu_obs=293.27 GHz gives L'_CO ~ 1.5e9 K km s^-1 pc^2, consistent with the axis range in Figure 3. The text should read (1.53 ± 0.54) x 10^9 K km s^-1 pc^2. This is a numerical error in the central measurement of the paper and must be corrected.","section":"Section 3.3, after Eq. (1)"},{"comment":"The L_FIR estimate is derived from a single continuum measurement with a fixed emissivity index beta=2.0 and a dust temperature assigned from the Sommovigo et al. (2022) relation with a uniform metallicity prior of Z=0-2 Z_sun. The resulting L_FIR has an asymmetry (1.1+1.0-0.3 x 10^12 L_sun) that spans below the ULIRG threshold even before magnification correction, and the central value is largely set by the assumed T_d. The authors should discuss the sensitivity of the ULIRG classification to the choice of beta and to the T_d relation, and ideally provide a table of L_FIR for a grid of T_d and beta values.","section":"Section 3.3, dust temperature and beta"}],"minor_comments":[{"comment":"The line properties for COSMOS-z10-2 are inconsistent between the text and Table 4: the text gives line flux 0.475±0.169 Jy km/s, central frequency 334.66±0.07 GHz, and FWHM 463.06±73.77 km/s, while Table 4 lists 0.439±0.157 Jy km/s, 334.67±0.06 GHz, and 415.42±63.16 km/s. Please reconcile these values.","section":"Section 3.2.2 vs Table 4"},{"comment":"The purity estimates of ~45% and ~75% are quoted without the number of galaxies in the respective redshift bins. Since the underlying JADES/CANUCS sample is incomplete and may be biased toward strong breaks, it would be helpful to state the sample sizes (N_z>8.5 and N_interloper) and the exact cuts used (including S/N_F090W<2 and S/N_F150W>5) so that readers can assess the statistical weight of the proposed F115W-F150W>1.0 threshold.","section":"Section 5.2.1"},{"comment":"The false-positive probability for UDS_18697 uses the number of clumps on the inverted map (N_neg) as the background estimate, but N_pos and N_neg differ by a factor of about 2.5 in the 1-arcsec aperture. The authors note this may be statistical fluctuation, but a bootstrap or Monte Carlo estimate of the false-positive rate would be more robust; currently the 4.5% probability is tied to a single noisy measurement of N_neg.","section":"Appendix B"},{"comment":"The statement that the line is 'co-spatial with the position of UDS_18697' is in tension with the reported 0.16 arcsec offset, which is larger than the expected 0.10 arcsec astrometric accuracy. Please clarify the positional tolerance used in this assessment.","section":"Section 3.2.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's central scientific claim depends on an unpublished JWST/NIRSpec spectrum, and the ALMA detection alone is marginal. The paper should either include the NIRSpec data or visibly soften the CO identification and ULIRG classification. The missing 10^9 factor in L'_CO is a serious typo that must be corrected. The colour-purity analysis is potentially useful, but the purity numbers should be presented with sample sizes and caveats. The paper is not suitable for acceptance in its current form, but the issues are addressable within the scope of a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know up front.\n\nFirst, the part that will survive is the interloper story. Three of the six pre-JWST z~10 candidates get NIRCam photometry, and the SED fits place all three at z~2-2.5. That is convincing on its own, and it lines up with the NIRSpec redshifts quoted for the others. The reconstruction of why the old HST+Spitzer selection failed – deblending in IRAC [3.6], sparse wavelength coverage, and faintness near the detection limit – is clear and matches what people have been finding since JWST. The colour-threshold analysis using JADES/CANUCS is the genuinely new piece: going from F115W-F150W > 0.5 to > 1.0 raises estimated purity for z>8.5 galaxies from ~45% to ~75%. That is a concrete, useful number for Euclid and Roman planning.\n\nSecond, the headline CO(9-8) detection in UDS_18697 is much weaker than it looks. The ALMA line is 4.5-sigma on the moment map, with a 4.5% false-positive probability from the clump search and a 0.16 arcsec offset. The identification as CO(9-8) rests entirely on the unpublished NIRSpec PRISM redshift from Larson et al. (in prep.). No spectrum is shown, no line list, no S/N. The NIRCam photo-z of 2.56 independently says the object is not at z~10, which is fine, but it cannot tell you which molecular line sits at 293.27 GHz. So treat L'_CO and the placement on the L_FIR-L'_CO plot as provisional until the spectroscopy appears.\n\nThe L_FIR estimate has its own caveats. The dust temperature comes from the Sommovigo+2022 scaling, beta is fixed to 2.0, and the magnification 1.35 is not corrected. Dividing by mu puts L_FIR at ~0.8e12 L_sun, below the usual ULIRG threshold. The authors acknowledge this in passing but still use the ULIRG label. That is a bit generous.\n\nMinor: the acknowledgements list ALMA program 2021.1.01396.S while Table 1 says 2021.1.00389.S. Data availability is 'on request'. No code or catalogues shipped, though they use public data.\n\nBottom line: the paper is a solid, honest empirical contribution that deserves a serious referee. The interloper conclusion and the purity recommendation will stand. I would like the referee to ask for (a) the NIRSpec redshift details or a clear caveat on the CO identification, (b) magnification-corrected luminosities and a labelled ULIRG threshold, and (c) fixing the program ID. I would not desk-reject this.","headline":"A useful cautionary tale with a robust interloper conclusion, but the CO(9-8)/ULIRG headline needs the unpublished NIRSpec data to stand.","tokens_in":28252,"tokens_out":3501,"would_cite":true,"duration_ms":33580,"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":"ALMA and JWST show that UDS_18697, once thought to be a galaxy at $z\\sim10$, is a dusty ultra-luminous infrared galaxy at $z=2.54$ with a CO($J=9-8$) line, and that most of the paper's pre-JWST $z\\sim10$ candidates are low-redshift…","keywords":["high-redshift galaxies","Lyman break selection","photometric redshift","ALMA","CO(9-8) line","ultraluminous infrared galaxy","JWST/NIRCam","galaxy contaminants"],"falsifier":"Inspect the NIRSpec PRISM spectrum of UDS_18697: if it does not show multiple optical lines (e.g., H$\\alpha$, [OIII], H$\\beta$) at a common redshift $z=2.54$, or if the ALMA line frequency does not match CO($J=9-8$) at that redshift, the identification fails. Alternatively, re-observing the target with ALMA at higher sensitivity and resolution would settle the marginal $4.46\\sigma$ detection: a genuine CO($J=9-8$) line should reproduce at $>5\\sigma$, remain co-spatial with the $29.9\\sigma$ dust continuum, and show a velocity width matching the NIRSpec lines.","tokens_in":26944,"feed_emoji":"🔭","tokens_out":12398,"duration_ms":89463,"temperature":0.7,"pith_summary":"This paper aims to establish that UDS_18697, a galaxy previously selected from HST and Spitzer photometry as a $z\\sim10$ candidate, is actually an ultra-luminous infrared galaxy at $z=2.54$ that emits a CO($J=9-8$) line, and that five of the six $z\\sim10$ candidates in the ALMA sample are low-redshift interlopers once JWST/NIRCam photometry and NIRSpec spectroscopy are folded in. The finding matters because pre-JWST Lyman-break selections of the earliest galaxies can be severely contaminated by dusty star-forming and Balmer-break galaxies at $z\\sim2$–3, and the paper offers a quantitative remedy: raising the Lyman-break colour threshold from $F115W-F150W>0.5$ to $F115W-F150W>1.0$ raises the purity of a $z>8.5$ sample from roughly 45% to roughly 75%, at some cost in completeness. If the result holds, current and future wide-area surveys such as the Euclid Deep Fields and the Roman High-Latitude Wide-Area Survey should adopt stringent colour cuts and deep short-wavelength coverage to avoid being dominated by interloping bright galaxies.","feed_headline":"ALMA and JWST expose a 'z~10' galaxy as a ULIRG at z=2.54","feed_subtitle":"A CO(9-8) line plus NIRSpec redshifts demote five of six candidates; a stricter color cut cleans the sample.","key_machinery":"The argument runs through a short chain of identifications and measurements. (1) The ALMA line at 293.27 GHz, co-spatial with UDS_18697, is identified as CO($J=9-8$) because the NIRSpec redshift $z=2.54$ places that transition at the observed frequency; the clump search gives a 4.5% false-positive probability for the line. (2) The $29.9\\sigma$ dust continuum at $368\\pm19\\,\\mu$Jy is modelled as a single-temperature modified blackbody with emissivity index $\\beta=2.0$ and a dust temperature $T_{\\rm d}$ from the Sommovigo et al. (2022) relation, giving $L_{\\rm FIR}\\approx1.1\\times10^{12}\\,L_\\odot$. (3) The source is placed on the $L_{\\rm FIR}$–$L'_{\\rm CO(9-8)}$ diagram against local and $z>2$ samples. (4) The colour–colour criterion $F115W-F150W>0.5$ with $F150W-F356W<1.4$, applied to spectroscopically confirmed galaxies from the JADES and CANUCS catalogues, yields the purity estimate, and the stricter $F115W-F150W>1.0$ threshold raises purity from $\\sim45\\%$ to $\\sim75\\%$ for $z>8.5$.","core_discovery":"The central discovery is that the ALMA line at 293.27 GHz with significance $4.46\\sigma$ toward UDS_18697 is not the targeted [OIII] 88$\\mu$m line at $z\\sim10$ but rather CO($J=9-8$) at $z=2.5356$, making the source a ULIRG with $L_{\\rm FIR}\\approx1.1\\times10^{12}\\,L_\\odot$ and a dust temperature $T_{\\rm d}\\approx42.8$ K, and placing it slightly below, yet consistent with, the local $L_{\\rm FIR}$–$L'_{\\rm CO}$ relation while being among the faintest CO($J=9-8$) emitters at $z>2$. This identification is anchored by JWST/NIRSpec PRISM spectroscopy that fixes the redshift at $z=2.54$ (R. Larson et al., in prep.), and it is corroborated by SED fitting with JWST/NIRCam photometry that revises all three NIRCam-observed targets to $z\\sim2$–2.5. The paper further claims, on the basis of spectroscopically confirmed galaxies from the JADES and CANUCS catalogues, that a stricter Lyman-break colour criterion substantially reduces low-$z$ contamination in high-$z$ samples.","pith_inferences":["If the unpublished NIRSpec redshift is independently confirmed, the CO($J=9-8$) identification makes UDS_18697 a benchmark for CO excitation in a moderately luminous ULIRG at cosmic noon; a direct test would be ALMA observations of lower-$J$ CO lines (e.g., CO(4-3) or CO(7-6)) at the same redshift, which should yield a consistent dynamical mass and line width.","The reported $\\sim4.5\\%$ and $\\sim10\\%$ false-positive probabilities for the two marginal ALMA lines suggest that comparable line searches should routinely report both positive and negative clump counts in their fields; without such statistics, marginal detections are hard to evaluate.","The purity jump from $\\sim45\\%$ to $\\sim75\\%$ is measured on the JADES/CANUCS spectroscopic sample, which may be biased toward blue continua or strong emission-line galaxies; applying the same colour cut to a completeness-corrected spectroscopic survey would sharpen the quantitative claim.","Because the magnification factor $\\mu=1.35$ for UDS_18697 is not corrected, the intrinsic $L_{\\rm FIR}$ and $L'_{\\rm CO}$ could be $\\sim35\\%$ lower, moving the source closer to the local relation; correcting it tests whether the slight CO-bright offset is real."],"forward_implications":["UDS_18697 becomes one of the faintest $z>2$ sources detected in CO($J=9-8$), extending the $L_{\\rm FIR}$–$L'_{\\rm CO}$ relation an order of magnitude downward in luminosity; more such serendipitous ALMA detections can populate the gap between local ULIRGs and high-$z$ DSFGs.","Pre-JWST $z\\sim10$ candidates selected from HST+Spitzer photometry should be re-fitted with NIRCam photometry before being used to constrain luminosity functions or reionisation; this paper shows three targets that previously had $z_{\\rm phot}\\sim10$ collapse to $z\\sim2$–2.5.","Euclid Deep Fields and Roman HLWAS should adopt a stringent Lyman-break colour threshold (equivalent to $F115W-F150W>1.0$) plus deep observations blueward of the break; otherwise bright low-$z$ interlopers will outnumber true $z>8.5$ galaxies.","Balmer-break galaxies at $z\\sim2$ can mimic Lyman-break dropouts; requiring a non-detection in F090W ($S/N<2$) alongside the colour cut removes this class of contaminant."],"supporting_citations":[{"why":"Supplies the JWST/NIRSpec PRISM redshift $z=2.54$ that anchors the CO($J=9-8$) identification.","marker":"R. Larson et al. in prep."},{"why":"Provided the original $z\\sim10$ candidate sample (including UDS_18697) from HST+Spitzer photometry that the paper re-derives with JWST data.","marker":"Finkelstein et al. 2022a"},{"why":"The physically-motivated dust temperature relation used to estimate $T_{\\rm d}\\approx42.8$ K and hence $L_{\\rm FIR}$.","marker":"Sommovigo et al. 2022"},{"why":"The local $L_{\\rm FIR}$–$L'_{\\rm CO}$ relation against which UDS_18697 is compared.","marker":"Liu et al. 2015"},{"why":"The colour-selection criteria (e.g., $F115W-F150W>0.5$) that the paper ports to NIRCam filters and tightens.","marker":"Bouwens et al. 2016"},{"why":"The ASTRODEEP-JWST NIRCam+HST photometric catalogue used for the revised SED fitting.","marker":"Merlin et al. 2024"},{"why":"JADES NIRSpec catalogue supplying the spectroscopically confirmed high-$z$ and low-$z$ galaxies for the purity test.","marker":"Bunker et al. 2024"},{"why":"Provides $z>2$ DSFG CO($J=9-8$) luminosities for the $L_{\\rm FIR}$–$L'_{\\rm CO}$ comparison.","marker":"Riechers et al. 2021"}],"fun_headline_variants":["ALMA line in 'z~10' candidate is CO(9-8), not [OIII] – from a ULIRG at z=2.54","JWST NIRSpec and ALMA reveal: 'z~10' galaxy is actually z=2.54 ULIRG","Stricter Lyman-break color cut needed after misidentified z~10 candidates","ALMA+JWST: 'z~10' candidates are low-z interlopers; stricter cut fixes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the unpublished JWST/NIRSpec PRISM redshift from R. Larson et al. (in prep.) is correct at $z=2.54$; every identification of the ALMA line as CO($J=9-8$), and therefore the ULIRG classification, the luminosity estimates, and the placement on the $L_{\\rm FIR}$–$L'_{\\rm CO}$ relation, collapses if that redshift is wrong.","fun_headline_variants_meta":{"raw":{"variants":["ALMA line in 'z~10' candidate is CO(9-8), not [OIII] – from a ULIRG at z=2.54","JWST NIRSpec and ALMA reveal: 'z~10' galaxy is actually z=2.54 ULIRG","Stricter Lyman-break color cut needed after misidentified z~10 candidates","ALMA+JWST: 'z~10' candidates are low-z interlopers; stricter cut fixes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000761,"raw_usage":{"total_tokens":3519,"prompt_tokens":1229,"completion_tokens":2290,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":845,"completion_tokens_details":{"reasoning_tokens":2168}},"tokens_in":845,"tokens_out":2290,"duration_ms":17410,"temperature":1.0,"reasoning_tokens":2168,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:05:42.604634+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inspect the NIRSpec PRISM spectrum of UDS_18697: if it does not show multiple optical lines (e.g., H$\\alpha$, [OIII], H$\\beta$) at a common redshift $z=2.54$, or if the ALMA line frequency does not match CO($J=9-8$) at that redshift, the identification fails. Alternatively, re-observing the target with ALMA at higher sensitivity and resolution would settle the marginal $4.46\\sigma$ detection: a genuine CO($J=9-8$) line should reproduce at $>5\\sigma$, remain co-spatial with the $29.9\\sigma$ dust continuum, and show a velocity width matching the NIRSpec lines.","supporting_citations":[],"review_version":1}