{"id":"63fc1a2a-2946-41d0-a31c-b48ba02ef7a8","arxiv_id":"2506.23669","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Massive star-forming galaxies at z~2 have dust attenuation A_V of 2-3 (sometimes 4), far higher than pre-JWST surveys found, because the new NIRCam photometry fixes the photometric redshifts of the reddest galaxies.","lead":"Using JWST/NIRCam data, astronomers find that massive star-forming galaxies at z~2 are much redder and more dust-obscured than earlier surveys indicated. The new photometry corrects previously overestimated redshifts for the reddest galaxies, revealing a strong evolution in dust attenuation that was previously hidden.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The red-tail claim hinges on photo-z accuracy for V-J>2.5 galaxies, yet the validation sample is only 25 spectroscopic redshifts; a larger spec-z comparison should settle whether the tail is real.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing point: the evolutionary claim depends on the reliability of photometric redshifts for the reddest, most attenuated galaxies. The paper's own comparison uses only 25 spectroscopic redshifts for V-J>2.5 galaxies, and the text explicitly notes that it is undetermined whether the improvement comes from better photometry or better templates. This is not an internal inconsistency, but it is a genuine correctness risk: a small or biased spec-z sample cannot fully secure the claim that the red tail is real rather than a photo-z artifact. The paper does provide useful independent checks, including the sub-mm cross-match and the bagpipes A_V comparison for a sub-sample, and these support the dust interpretation without independently fixing the redshifts. A larger spec-z sample, combined with a re-derivation of colors and A_V at fixed spectroscopic redshift, would directly settle the concern. Since this is the same concern the reader already raised and the paper is transparent about the limitation, the CONDITIONAL verdict remains appropriate, so no verdict change is needed.","tokens_in":12875,"tokens_out":5314,"duration_ms":65384,"concrete_test":"Expand the Section 3.1 validation by merging all public spectroscopic redshifts in the five CANDELS fields (3D-HST, MOSDEF, VANDELS, JWST/NIRSpec surveys) and matching them to the DJA catalog, targeting more than 100 galaxies with V-J>2.5. For this sample, recompute rest-frame V-J and A_V using the spectroscopic redshift instead of the photometric redshift. If the median V-J and A_V for M*>3e10 M_sun at 1.51.5 develop a V-J>2.5 tail absent at lower z, stands or falls with the photometric redshifts of the reddest galaxies. In Section 3.1 the authors show that earlier 3D-HST photo-zs for 25 matched V-J>2.5 objects were overestimated (median zph=2.12 vs zsp=1.68), while the new DJA photo-zs agree with spec-z (median 1.64 vs 1.68). This is the only direct validation of the crucial population. With 25 objects, however, the comparison cannot rule out a selection bias among spectroscopically targeted red galaxies or a correlated photo-z error that still shifts a subset of the tail across the z=1.5 boundary. The paper itself leaves open whether the improvement is due to improved photometry or improved templates, and a systematic photo-z offset for the faintest, reddest objects would propagate directly into rest-frame V-J, stellar masses, and A_V, potentially manufacturing or amplifying the claimed evolution. Independent support from sub-mm detections confirms dust but does not independently pin down the redshifts of the tail.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses JWST/NIRCam photometry from the DAWN JWST Archive to construct a stellar-mass-complete sample of about 28,000 galaxies with M* > 10^9 M_sun in the redshift range 0.5 < z < 2.5, and studies the evolution of rest-frame U-V and V-J colors and the dust attenuation parameter A_V. The central claim is that the V-J color distribution of star-forming galaxies evolves strongly with redshift: massive galaxies (M* > 3e10 M_sun) at z > 1.5 exhibit a pronounced tail reaching V-J > 2.5, which is absent at z < 1, and this tail is interpreted as arising from dust attenuation with A_V in the range 1.5-3.5 for M* ~ 1e11 M_sun at z ~ 2. The authors attribute the previous lack of detection of this population to overestimated photometric redshifts for the reddest objects in earlier catalogs, and support their interpretation with a comparison to 25 spectroscopic redshifts and with cross-matching to sub-mm selected galaxies. A check with an independent SED-fitting code (bagpipes) is presented for the A_V values of the reddest subsample.","tokens_in":13080,"tokens_out":4456,"duration_ms":53929,"significance":"If the central claim holds, this is an important result: it would mean that massive star-forming galaxies at cosmic noon are significantly more dust-attenuated than previously inferred from rest-frame optical colors, with A_V values extending to 3-4, and that earlier ground- and space-based surveys systematically underestimated this population because of photo-z biases. The paper is concise, uses high-quality public JWST data, and makes a falsifiable prediction that a comprehensive spectroscopic survey of high-A_V galaxies at z > 1 should confirm the high attenuation. The independent bagpipes check for the reddest galaxies and the sub-mm cross-match are notable strengths. However, the central claim rests on the reliability of photometric redshifts for a small and extreme subset of the sample, and the current validation sample is too small to fully secure that footing.","major_comments":[{"comment":"The claim that the red V-J tail at z > 1.5 is real and was previously missed because of photo-z biases rests on the comparison with only 25 spectroscopic redshifts for galaxies with V-J > 2.5. This is a very small sample on which to base a strong evolutionary claim, and the paper itself notes that it remains undetermined whether the improvement comes from better photometry or better templates. The authors should quantify the impact of plausible photo-z errors on the inferred color distributions, for example by applying the old 3D-HST photo-z error distribution to the current sample and showing how the V-J tail and the z=1.5 boundary would shift. They should also report how many galaxies in the final sample have V-J > 2.5 in each redshift bin, and how many of those have spectroscopic redshifts, so that the reader can judge the statistical weight of the 25-object validation.","section":"Section 3.1, photo-z validation paragraph"},{"comment":"The statement that the red tail 'does not exist at z < 1' is based on running medians (shown as solid lines) and the visual appearance of the scatter plots, without any confidence intervals or counts of the tail population. Given that the running median is not sensitive to the existence of a tail and the reddest objects are sparse, the authors should provide a quantitative characterization of the tail—for example, the fraction of massive star-forming galaxies with V-J > 2.5 (or > 2.0) as a function of redshift, with bootstrap uncertainties. Without such a statistic, the absence of the tail at low redshift is not demonstrated to be significant.","section":"Section 3.1 and Figure 1"},{"comment":"The A_V values and the V-J colors are both outputs of the same EAZY SED fits, so the correlation between V-J and A_V used to interpret the red tail is partly built into the model. The bagpipes check is welcome, but it is applied only to the V-J > 2.5 subsample, leaving the redshift evolution of A_V for the full star-forming population untested in an independent way. The authors should either run bagpipes on a representative subsample spanning the full mass-redshift range, or explicitly state that the A_V evolution is a model-dependent inference whose robustness has only been verified for the extreme tail.","section":"Section 3.2, A_V evolution"}],"minor_comments":[{"comment":"The sentence 'the photometric redshift estimates for the reddest... has markedly improved' contains a subject-verb agreement error; 'estimates' is plural and should take 'have'. The same issue appears in the abstract.","section":"Abstract and Section 3.1"},{"comment":"The phrase 'These data product derive' should be 'These data products derive'.","section":"Section 2"},{"comment":"The color bar on the right side of the figure is labeled 'Log(M/M_sun)' but it is not immediately clear that this color coding applies to the plotted points; please clarify in the caption.","section":"Figure 2"},{"comment":"The footnote on the origin of the term 'quenching' is interesting but tangential; consider moving it to the main text or removing it, as it interrupts the flow of the introduction.","section":"Section 1, footnote 1"}],"recommendation":"major_revision","confidential_remarks":"This is a well-written short paper that presents an interesting and potentially important empirical result. The main concern is the small spectroscopic validation sample for the reddest population, which is the linchpin of the redshift-evolution claim. The authors are clearly aware of the limitation, but they need to address it more quantitatively before the paper can be accepted. The paper fits the scope of A&A and, with the requested additions, would be a solid contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a solid, honest, short paper that establishes a stronger redshift evolution of rest-frame V-J color and AV for massive star-forming galaxies than previous surveys could see. The new thing is identifying that the reason the red tail was missed is a systematic photometric redshift offset for the very reddest objects, and showing that with the DJA/NIRCam photometry the photo-z agree with spec-z. The cross-checks (bagpipes AV, sub-mm matches, running medians) are the right sort of evidence.\n\nThe critical soft spot, as the authors themselves say, is that the validation of the reddest population rests on 25 spectroscopic redshifts. That is a small number, and the stress-test note is right that it cannot rule out a bias among spec-z-targeted red galaxies. The paper is also openly undecided whether the improvement comes from better photometry or better templates; that matters for how much we trust the photo-z at the extremes. Still, the 25 objects show a large offset in 3D-HST (median 2.12 vs 1.68) and a corrected DJA photo-z (1.64 vs 1.68), and the full matched sample of 7000+ objects shows no offset, which gives some confidence that the issue is localized and fixed. The independent bagpipes check on the V-J>2.5 sample gives AV values within 0.2 mag, and the sub-mm detections land exactly in the red tail. I don't think the central argument is in doubt: quiescent galaxies cannot produce V-J>2.5, so dust attenuation is required.\n\nThere is a mild circularity issue: colors and AV come from the same SED fits, so the correlation between V-J and AV is partly model-driven. The paper addresses this by showing that the interpretation is consistent with independent indicators, and it is cautious about the physical explanation (patchy dust, birth clouds, gas fractions).\n\nWho is this for: anyone working on galaxy evolution, dust attenuation, or photo-z. It is a reference result that will be cited heavily. The next step, a larger spectroscopic sample of high-AV galaxies, is explicitly flagged by the authors. My recommendation: send it to a good referee, accept after the photo-z caveat is made more prominent if needed. This deserves to be published.","headline":"A solid, transparent demonstration that massive star-forming galaxies at z~2 are much redder and more dust-attenuated than previously thought, with the main caveat being the small spectroscopic validation sample for the reddest objects.","tokens_in":13694,"tokens_out":2178,"would_cite":true,"duration_ms":23873,"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":"JWST/NIRCam photometry shows that massive star-forming galaxies at z≈2 are far dustier than earlier surveys indicated, with V-J colors past 2.5 and visual extinction up to 3.5 magnitudes.","keywords":["galaxy evolution","dust attenuation","rest-frame colors","V-J color","JWST NIRCam photometry","photometric redshifts","star-forming galaxies","massive galaxies"],"falsifier":"A spectroscopic redshift survey of galaxies with $V-J>2.5$ at $z\\approx1.5$–$2.5$ would settle the claim: if the new photometric redshifts show the same roughly 0.4 overestimate seen in earlier catalogs, the red tail and the inferred rise in $A_{\\mathrm{V}}$ would largely disappear.","tokens_in":12691,"feed_emoji":"🔴","tokens_out":13392,"duration_ms":121584,"temperature":0.7,"pith_summary":"The paper shows that the rest-frame $V-J$ color — the difference in brightness between the optical $V$ band and the near-infrared $J$ band — of star-forming galaxies evolves strongly with cosmic time. Massive galaxies ($M_\\star > 3\\times10^{10}\\,M_\\odot$) at redshift $z>1.5$ develop a tail of very red objects reaching $V-J>2.5$ that has no counterpart at $z<1$, and the paper interprets these colors as dust attenuation, with visual extinction $A_{\\mathrm{V}}$ between 1.5 and 3.5 magnitudes for $M_\\star\\approx10^{11}\\,M_\\odot$ galaxies at $z\\approx2$. Earlier surveys missed this evolution, the paper argues, because their photometric redshifts for exactly these reddest, most attenuated galaxies were biased; JWST/NIRCam photometry removes the bias. If the claim is right, the dust content of massive high-redshift galaxies has been systematically underestimated, and their stellar masses and star-formation rates need to be rethought.","feed_headline":"JWST finds galaxies at z≈2 are dustier than we knew","feed_subtitle":"NIRCam colors reveal a red V-J tail beyond 2.5, with visual extinction up to 3.5 mag, that older catalogs missed.","key_machinery":"The central machinery is the rest-frame $V-J$ color, the difference between rest-frame $V$-band and $J$-band magnitudes, measured from JWST/NIRCam photometry combined with HST imaging. In the $U-V$ versus $V-J$ diagram, dusty star-forming galaxies are red in $V-J$ while quiescent galaxies are blue in $V-J$, so this color separates attenuation by dust from the reddening caused by old stellar populations. The measurement works because NIRCam provides rest-frame near-infrared coverage up to $z\\approx2.5$, and because the precise photometry removes the systematic redshift overestimates that previously pushed the reddest galaxies to the wrong distances and washed out the tail.","core_discovery":"The central discovery is that the rest-frame $V-J$ versus stellar mass relation for star-forming galaxies becomes very steep at $z>1.5$, with high-mass galaxies ($M_\\star > 3\\times10^{10}\\,M_\\odot$) reaching $V-J\\approx2$–$3$, while the same galaxies at $z<1$ have $V-J\\approx1$–$1.5$. The reddest colors cannot be produced by old stellar populations, so they must come from dust; spectral-energy-distribution fitting gives $A_{\\mathrm{V}}\\approx1.5$–$3.5$ for $M_\\star\\approx10^{11}\\,M_\\odot$ at $z\\approx2$, with about 90% of such galaxies having $A_{\\mathrm{V}}>2$ at $z>1.5$. The paper attributes the previously unseen evolution to improved photometric redshifts for the reddest galaxies: for 25 galaxies with spectroscopic redshifts and $V-J>2.5$, the new median photometric redshift matches the spectroscopic median, whereas the older catalog overestimated it by about 0.4. The $U-V$ colors of these galaxies stay relatively blue, so the attenuation must be gray in the ultraviolet-optical, consistent with attenuation of most stars younger than about one gigayear rather than only the youngest star-forming regions.","pith_inferences":["If the redshift improvement turns out to come from better templates rather than better photometry, the same correction could be applied to existing HST-only catalogs, shifting their color distributions without new observations.","If dust attenuation rises this steeply with redshift, the peak may lie beyond $z=2.5$, making the $z>1.5$ tail described here the low-redshift edge of the optically dark, dust-obscured population found at $z\\gtrsim3$.","A direct test of the gray-attenuation interpretation would compare NIRCam-based $A_{\\mathrm{V}}$ with independently measured far-infrared or millimeter dust luminosities for the same individual galaxies.","The steep mass dependence of $A_{\\mathrm{V}}$ suggests that gas fraction and dust geometry, rather than metallicity, set the dust column; galaxy formation models that reproduce this $A_{\\mathrm{V}}$–mass–redshift relation would be needed to explain the heavy obscuration of massive galaxies at $z\\approx2$."],"forward_implications":["At $0.510^{11}\\,M_\\odot$ at $z>1.5$ have $A_{\\mathrm{V}}>2$, so these systems should be treated as heavily dust-obscured rather than moderately reddened.","Older photometric-redshift catalogs systematically misplaced the reddest galaxies toward higher redshift, so samples built from them need to be re-derived or re-calibrated for these objects.","The red $V-J$ tail of the mass-selected sample overlaps with the ALMA sub-mm selected population, indicating that the NIRCam colors and the sub-mm detections trace the same obscured star-forming galaxies.","Because $U-V$ remains relatively blue at high attenuation, using $U-V$ alone to infer dust content or to classify galaxies as quiescent will misclassify massive dusty galaxies."],"supporting_citations":[{"why":"Supplies the HST+NIRCam photometry and the spectral-energy-distribution-derived catalog of redshifts, masses, rest-frame colors, and $A_{\\mathrm{V}}$ on which the whole sample rests.","marker":"Valentino et al. 2023"},{"why":"Provides the template-fitting method used to estimate the photometric redshifts and rest-frame colors.","marker":"Brammer et al. 2008"},{"why":"Defines the $U-V$ versus $V-J$ boundary used to separate star-forming from quiescent galaxies.","marker":"Muzzin et al. 2013"},{"why":"Provides the earlier photometric-redshift catalog shown to be systematically too high for the reddest galaxies.","marker":"Skelton et al. 2014"},{"why":"Supplies the spectroscopic redshifts used to validate the improved photometric redshifts for the $V-J>2.5$ galaxies.","marker":"Merlin et al. 2024"},{"why":"Provides the ALMA sub-mm catalog used to show that the red $V-J$ tail corresponds to dust-obscured star-forming galaxies.","marker":"Adscheid et al. 2024"},{"why":"Supplies the dust-geometry model that explains high $A_{\\mathrm{V}}$ with relatively blue $U-V$ colors.","marker":"Gebek et al. 2025"}],"fun_headline_variants":["JWST reveals massive galaxies at z~2 are far dustier than thought","Redder than expected: JWST uncovers heavy dust in high-z galaxies","JWST finds dusty behemoths at z>1.5 that old surveys missed","V-J colors show dust attenuation up to Av=3.5 in massive z~2 galaxies","JWST's sharp eyes spot extra dust in massive high-redshift galaxies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result depends on the assumption that the photometric redshifts for the reddest, most dust-obscured galaxies are now correct, and the paper's own check of this assumption uses only 25 spectroscopic redshifts.","fun_headline_variants_meta":{"raw":{"variants":["JWST reveals massive galaxies at z~2 are far dustier than thought","Redder than expected: JWST uncovers heavy dust in high-z galaxies","JWST finds dusty behemoths at z>1.5 that old surveys missed","V-J colors show dust attenuation up to Av=3.5 in massive z~2 galaxies","JWST's sharp eyes spot extra dust in massive high-redshift galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000362,"raw_usage":{"total_tokens":2120,"prompt_tokens":1279,"completion_tokens":841,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":895,"completion_tokens_details":{"reasoning_tokens":735}},"tokens_in":895,"tokens_out":841,"duration_ms":8594,"temperature":1.0,"reasoning_tokens":735,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:34:50.575858+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A spectroscopic redshift survey of galaxies with $V-J>2.5$ at $z\\approx1.5$–$2.5$ would settle the claim: if the new photometric redshifts show the same roughly 0.4 overestimate seen in earlier catalogs, the red tail and the inferred rise in $A_{\\mathrm{V}}$ would largely disappear.","supporting_citations":[{"cited_title":"E., Whitaker , K","cited_arxiv_id":null,"evidence_quote":"Provides the earlier photometric-redshift catalog shown to be systematically too high for the reddest galaxies."}],"review_version":1}