{"id":"8c9ea165-c500-4134-8d6b-02bf9643e19a","arxiv_id":"2411.14327","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"In the MUSE Extremely Deep Field, 32-51% of high-redshift Ly-alpha emitters show double-peaked line profiles, with the fraction depending on luminosity rather than decreasing steeply with redshift once selection effects are considered.","lead":"A team used the deepest MUSE survey to sort the shapes of Lyman-alpha emission lines from 477 distant galaxies into single, double, triple, or undetected peaks, adding a spatial check to tell real from fake doubles. Between 32% and 51% of the bright sample show double-peaked lines, and the share rises with galaxy luminosity, offering a new benchmark for how gas around early galaxies scatters this signature light.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The luminosity and redshift trends in the double-peak fraction are not separated from each other: after the three cuts, high-redshift bins are still the most luminous, so both the claimed luminosity evolution and the high-z plateau may be selection artifacts; a bivariate or matched test is needed.","rationale":"The reader's conditional verdict is appropriate. The paper is unusually careful: it tests the detection threshold in App. A, runs background spectra in Sect. 3.6, and explicitly acknowledges that the high-z plateau may be caused by the high mean luminosity in the highest-redshift bins (Sect. 4.3.3). Those admissions do not, however, resolve the degeneracy. The concern here is sharper than the reader's weakest_assumption in one respect: the same luminosity–redshift coupling that can create a spurious plateau can also create a spurious luminosity trend, because the low-luminosity bins are preferentially at low redshift. The one-dimensional binning in Figs. 10 and 12 cannot distinguish an intrinsic luminosity dependence from a redshift dependence or from the combined selection effect. A matched or bivariate analysis is therefore a necessary condition for the abstract-level claim that the fraction evolves with luminosity and for the corresponding physical discussion of IGM attenuation. This does not weaken confidence in the 32–51% bracket itself; it weakens the separate trends built on top of it. The test proposed is straightforward with the published sample, and it should be reported as a conditional requirement rather than grounds for rejection.","tokens_in":34338,"tokens_out":7230,"duration_ms":75928,"concrete_test":"Recompute XDP_I and XDP_R in luminosity quartiles within a single narrow redshift window where the sample still spans a wide luminosity range (e.g., 3.0 < z < 4.0), and recompute the redshift trend using only galaxies in a fixed luminosity interval (e.g., 41.0 < log10(L_Lyα/[erg/s]) < 42.0) or by resampling each redshift bin to match the luminosity distribution of the lowest-z bin. If the luminosity trend persists in the fixed-redshift window and the redshift plateau persists in the fixed-luminosity subsample, the concern is resolved; if either trend weakens or flips, that trend should be reported as a selection effect or upper limit rather than an intrinsic property.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline range XDP_I = 51±4% and XDP_R = 32±3% (Eqs. 3–4) is probably robust as a bracket, but the two differential trends are the load-bearing part of the central claim. The 'unbiased' sample is defined by cuts on L_Lyα, integrated S/N, and vsep (Sect. 4.1.2–4.1.3). These cuts do not break the luminosity–redshift degeneracy: Fig. 12 itself shows that the highest-redshift bins have the highest mean Lyα luminosity, and Sect. 4.3.3 concedes that the high-z plateau may be an artifact of this. The same degeneracy can also manufacture the luminosity trend claimed in Sect. 4.3.2. Because faint bins are preferentially at low redshift, where XDP is already low (77% at z~3 versus 42% at z>5.5), the observed increase of XDP with luminosity could be a redshift effect rather than an intrinsic luminosity dependence. The paper computes each trend with one-dimensional bins and never controls for the other variable, so the abstract's statement that the fraction 'seems to evolve dependently with the Lyα luminosity' and the claim of no severe redshift decline are not yet established as independent population properties. This is not an accusation of error; the authors flag part of the issue, but the physical interpretation requires a bivariate test.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Using the MUSE eXtremely Deep Field DR2 catalogue, the paper selects 477 Lyα-emitting galaxies at z = 2.8–6.6 and develops an automated spectral-plus-spatial classification pipeline that assigns each line to no-peak, single-peak, double-peak, or triple-peak categories, with narrow-band imaging used to reject spatially spurious peaks. After applying luminosity, signal-to-noise, and peak-separation cuts, the paper defines an unbiased sample of 206 galaxies and reports a double-peak fraction between XDP_R = 32 ± 3% and XDP_I = 51 ± 4% (Eqs. 3–4). It further reports that this fraction increases with Lyα luminosity, that much of the sample is blue-dominated in B/T, and that the double-peak fraction shows a plateau rather than a strong decline at high redshift, with a possible observational cause for the plateau. The paper also identifies fake double-peaks, interacting pairs, and a small subsample with secure systemic redshifts.","tokens_in":34636,"tokens_out":6137,"duration_ms":60614,"significance":"If the differential trends survive closer scrutiny, this would be one of the first blind, spatially verified censuses of Lyα spectral shapes at z ≈ 3–6. The inclusive/restrictive bracket is a useful way to bound classification uncertainties, and the background-spectrum test (Sect. 3.6) is a clear strength, as is the narrow-band spatial confirmation that eliminates a substantial number of false peaks. The catalogue of peak separations and B/T ratios will be useful for studies of LyC leakers, IGM transmission, and gas kinematics. The headline double-peak fraction range is probably robust as a bracket, but the paper's two differential claims about luminosity and redshift evolution are not yet established because the analysis does not separate those two variables.","major_comments":[{"comment":"The two differential claims in the abstract—that the double-peak fraction increases with Lyα luminosity and does not decline steeply with redshift—are not separated from each other. The unbiased-sample cuts in Sect. 4.1.3 remove some observational biases but do not break the luminosity–redshift degeneracy: Fig. 12 shows that the highest-redshift bins also have the highest mean Lyα luminosity, and Sect. 4.3.3 concedes that this may create the observed plateau. The same covariance can also produce the luminosity trend in Fig. 10, because the faint luminosity bins are preferentially at low redshift, where the double-peak fraction is already low. A bivariate analysis (for example, XDP in two-dimensional luminosity–redshift bins, a luminosity-matched comparison, or an explicit forward-model of the selection) is needed before the abstract's claims of luminosity-dependent evolution and the absence of a severe redshift decline can be taken as intrinsic population properties. As written, the one-dimensional binning leaves these as plausible but non-unique interpretations.","section":"4.3.2–4.3.3, Figs. 10 and 12"},{"comment":"The thresholds defining the unbiased sample (LLyα > 3 × 10^40 erg/s, integrated S/N > 7, vsep > 150 km/s, and the N = 40 detection threshold) are chosen empirically from the same dataset, and the S/N = 7 cut in particular is justified by the plateau in the cumulative double-peak fraction in Fig. 6. The background test in Sect. 3.6 quantifies false positives, but it does not measure completeness: it does not tell what fraction of true double-peaks at a given S/N, vsep, B/T, and redshift are missed by the pipeline. Since the quoted XDP range and the trends in Figs. 10 and 12 depend directly on these choices, the systematic uncertainty from threshold selection is currently not propagated into the quoted errors. An end-to-end injection/recovery test, or at least an explicit demonstration that the conclusions are stable as the thresholds are moved within plausible ranges, is needed to support the 'universal fraction' framing.","section":"4.1.2–4.1.3 and App. A"},{"comment":"The paper presents a large blue-dominated fraction (48/105 in the inclusive unbiased sample) as suggesting unique gas dynamics and inflows, but for most of the unbiased sample the systemic redshift is unknown, so 'blue peak' means only 'bluer than the red peak', not necessarily blueward of the systemic velocity. The only secure check, in Sect. 5.3, gives eight double-peaks with systemic redshifts, of which only one has B/T > 0.5, and that object has both peaks on the red side of the systemic redshift. The manuscript does acknowledge this limitation in Sect. 5.3, but the abstract and Sect. 4.2.3 still convey the physical inflow interpretation without this caveat. Please either quantify how many of the blue-dominated objects can plausibly be true blue-dominated given the systemic-redshift subsample, or state in the abstract that the inflow interpretation is preliminary.","section":"4.2.3 and 5.3"}],"minor_comments":[{"comment":"The sentence 'The fraction of the sample which is red peak dominated is 23%' appears to be a typo: the numbers quoted just before it (N = 51 below B/T = 0.5 and N = 15 above) give a blue-dominated fraction of 23% for UDP_R, not a red-dominated fraction.","section":"4.2.3"},{"comment":"The sentence 'around 10% of the peaks detected are spurious (∼76 over 760 detected peaks)' is not directly derivable from Table 2, which reports 7% and 12% of background spectra with a noise peak; please clarify the denominator and the conversion from the background test to the estimated number of spurious peaks.","section":"3.6"},{"comment":"The term 'unbiased sample' is used for a sample that is still restricted by luminosity, S/N, and peak-separation cuts; consider adding a qualifier such as 'unbiased within the selected parameter range' to avoid overstating the degree to which observational selection has been removed.","section":"4.1.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid observational census and the headline double-peak fraction bracket is likely robust. The main weakness is the unbroken luminosity–redshift degeneracy in the differential trends; this is fixable with a bivariate test or with more cautious wording of the abstract claims. The threshold-tuning concern is also real and would benefit from completeness simulations, but it does not by itself invalidate the bracketing approach. I would be comfortable with acceptance after these issues are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is worth your time. It is the first systematic, automated spectral classification of Lyα line shapes in the deepest MUSE field, and it adds a spatial confirmation step that is genuinely new. Prior double-peak fractions came from visual inspection or spectral-only algorithms; this one uses narrow-band images to check that both peaks come from the same spatial location. That step is well motivated and tested on 200 background spectra, yielding a 7-12% false peak rate. The resulting GOLD/SILVER/BRONZE taxonomy is a real contribution, and the identification of 4% 'fake' double-peaks and ~19% interacting systems is useful.\n\nThe central double-peak fraction, 32-51% on a cleaned sample of 206 galaxies, is probably robust. The authors bracket it with inclusive and restrictive samples, they test their detection threshold (N=30/40/50) and show the spatial verification homogenizes the results, and they are honest about the remaining biases. I would trust that range.\n\nThe soft spot is the differential trends. The luminosity and redshift dependence of XDP are presented as separate results, but the selection cuts do not break the luminosity-redshift degeneracy. The highest-redshift bins are also the most luminous, as their own Fig. 12 shows. The authors concede the high-z plateau could be an artifact of high luminosity, but the same degeneracy can manufacture the claimed luminosity trend, since faint bins sit preferentially at low redshift where XDP is low. A bivariate or matched test is needed before either trend is quoted as an intrinsic population property. This is not a fatal flaw; it is an addressable one, and the paper explicitly flags part of it. Also, the full data tables are not in the preprint, which makes verification harder, and the blue-dominated fraction is sensitive to the inclusive/restrictive choice, though the authors note the need for systemic redshifts.\n\nBottom line: this is a careful, honest observational paper with a methods contribution that will be used. It deserves a serious referee, and the revision should ask for the bivariate analysis and the data tables.","headline":"A careful blind census of Lyα double-peak fractions with a genuinely new spatial confirmation step; the 32-51% bracket is probably robust, but the luminosity and redshift trends need a bivariate test before they are quoted.","tokens_in":35268,"tokens_out":2322,"would_cite":true,"duration_ms":22448,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Between 32% and 51% of distant Lyman-alpha emitting galaxies show double-peaked emission, with the fraction growing as luminosity rises and staying flat with redshift.","keywords":["Lyman-alpha emitters","double-peaked Lyα profiles","spectral classification","MUSE extremely deep field","IGM attenuation","blue-to-total flux ratio","peak separation","high-redshift galaxies"],"falsifier":"Measure the double-peak fraction in redshift bins with matched Lyα luminosity distributions, for example with a survey reaching the same faint luminosities at z≈6 as at z≈3; if the high-redshift plateau disappears once luminosity is matched, the claimed IGM attenuation signal is largely a selection artifact, while if a steep decline remains, the plateau is real.","tokens_in":34135,"feed_emoji":"🔭","tokens_out":4980,"duration_ms":42986,"temperature":0.7,"pith_summary":"This paper aims to give the first systematic, algorithm-driven inventory of Lyman-alpha line shapes among a blind, unbiased sample of high-redshift galaxies. Using the deepest MUSE field, it classifies 477 galaxies at z=2.8-6.6 by combining spectral detections with narrow-band images, and defines a cleaned sample of 206 galaxies. The central claim is that between 32% and 51% of these galaxies show double-peaked Lyα profiles, with the fraction increasing toward brighter Lyα luminosity and not declining steeply with redshift, although an artificial rise at the highest redshifts may explain the plateau. If true, the double-peak fraction becomes a usable tracer of how intergalactic gas attenuates Lyα emission across cosmic time, and the surprising number of blue-dominated double peaks would point to a population of galaxies with unusual gas kinematics.","feed_headline":"Up to 51% of distant galaxies show double-peaked Lyman-alpha lines","feed_subtitle":"New spectral-spatial census of 206 galaxies ties the double-peak fraction to luminosity, not to redshift.","key_machinery":"The load-bearing mechanism is a two-stage classification pipeline. First, for each of 477 galaxies, one hundred bootstrap realizations of the extracted spectrum are generated; a pixel counts as signal if the S/N exceeds 1 for two adjacent pixels, and the detection spectrum is thresholded at N=40 to define areas of signal, with a flux-variation analysis catching secondary peaks inside a single area. Second, narrow-band images for each area of signal are run through a source-detection and deblending routine so that each spectral peak must have a spatial counterpart inside the reference segmentation map; this yields the GOLD, SILVER, and BRONZE spatial categories and removes fake multi-peaks. Applying three selection cuts — Lyα luminosity above $3\\times10^{40}$ erg/s, integrated S/N above 7, and peak separation above 150 km/s — produces the unbiased sample of 206 galaxies, and the inclusive versus restrictive definitions of double peaks bound the true fraction.","core_discovery":"The paper's central discovery is a measured universal fraction of double-peaked Lyman-$\\alpha$ emitters: $X_{\\mathrm{DP,I}} = 51\\pm4\\%$ for the inclusive unbiased sample and $X_{\\mathrm{DP,R}} = 32\\pm3\\%$ for the restrictive one. The authors argue that this fraction grows with Lyα luminosity, from roughly 34% in the faintest bin to near 60% in the brightest, and that the expected decrease with redshift is not seen as a steep decline, but rather as a plateau above z≈4 that may be inflated by the higher mean luminosities of the highest-redshift bins. They also report that roughly half of the inclusive double-peaked sample is blue-dominated ($B/T>0.5$), a larger share than previous studies and simulations suggest, and that about 4% of double peaks are spatially fake, with the two peaks arising from different locations.","pith_inferences":["If systemic redshifts become available for a larger fraction of the sample, some of the blue-dominated double peaks may turn out to have both peaks on the red side of systemic velocity, meaning they trace geometry rather than infall; the paper already sees this for one of its fifteen objects with secure redshifts.","The same method applied to shallower, wider surveys could test whether the luminosity trend persists when selection effects differ, and whether the low-redshift rise toward cosmic noon is real.","With sufficiently deep data at z>6, matching luminosity across redshift would separate an intrinsic IGM-opacity signal from a purely selection-driven plateau.","The 4% fake-double-peak fraction is a lower limit; higher-resolution imaging could reveal more spatially distinct components, blurring the boundary between GOLD and BRONZE classifications."],"forward_implications":["If the fraction is truly luminosity-dependent, any survey comparing double-peak fractions across redshift must match Lyα luminosity or the apparent redshift evolution will be contaminated.","A plateau instead of a steep decline at z>4 weakens the simplest expectation that IGM attenuation destroys blue peaks at high redshift.","The finding that roughly 20% of the parent sample lies in interacting or complex environments implies that purely spectral classifications overcount multi-peaked Lyα lines.","A sizeable blue-dominated population, if confirmed with systemic redshifts, would challenge the standard outflow-dominated picture for high-redshift LAEs.","Peak separations out to nearly 1600 km/s, with a mean around 447 km/s, are consistent with earlier samples, providing a cross-check that the new method reproduces established measurements."],"supporting_citations":[{"why":"Supplies the MXDF DR2 catalogue and the reference spectra that define the parent sample of 477 galaxies.","marker":"Bacon et al. 2023"},{"why":"Provides the main MUSE-based comparison for double-peak fraction and peak-separation distributions.","marker":"Kerutt et al. 2022"},{"why":"Earlier UV-selected sample reporting 30% double-peaks, used as a literature baseline for the measured fraction.","marker":"Kulas et al. 2012"},{"why":"Earlier equivalent-width-selected sample reporting 50% double-peaks at z=3.1, used for comparison.","marker":"Yamada et al. 2012"},{"why":"Establishes the relation between Lyα peak separation and neutral hydrogen column density, motivating the physical interpretation of double peaks.","marker":"Verhamme et al. 2015"},{"why":"Simulations predicting B/T distributions and the rarity of blue-dominated spectra; used to interpret the surprisingly large blue-dominated fraction.","marker":"Blaizot et al. 2023"},{"why":"Stacked spectra showing that the blue part of Lyα is increasingly suppressed with redshift, the trend the paper tests against its double-peak fraction.","marker":"Hayes et al. 2021"},{"why":"IGM transmission modeling predicting that intergalactic gas preferentially attenuates the blue side of Lyα, motivating the expected redshift decline in double peaks.","marker":"Laursen et al. 2011"}],"fun_headline_variants":["Up to 51% of distant galaxies have double-peaked Lyman-alpha lines","Double-peaked Lyman-alpha emitters: 51% at most, tied to luminosity","Lyman-alpha double peaks: up to half of galaxies, but luminosity-driven","Distant galaxies' Lyman-alpha lines: double peaks common, luminosity matters","Blue-dominated Lyman-alpha double peaks hint at gas infall in early galaxies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The three selection cuts that define the unbiased sample are assumed to remove the dominant observational biases, so the measured trends with luminosity and redshift reflect intrinsic galaxy properties rather than selection effects.","fun_headline_variants_meta":{"raw":{"variants":["Up to 51% of distant galaxies have double-peaked Lyman-alpha lines","Double-peaked Lyman-alpha emitters: 51% at most, tied to luminosity","Lyman-alpha double peaks: up to half of galaxies, but luminosity-driven","Distant galaxies' Lyman-alpha lines: double peaks common, luminosity matters","Blue-dominated Lyman-alpha double peaks hint at gas infall in early galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00031,"raw_usage":{"total_tokens":1840,"prompt_tokens":1091,"completion_tokens":749,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":707,"completion_tokens_details":{"reasoning_tokens":644}},"tokens_in":707,"tokens_out":749,"duration_ms":5937,"temperature":1.0,"reasoning_tokens":644,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:17:52.418314+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the double-peak fraction in redshift bins with matched Lyα luminosity distributions, for example with a survey reaching the same faint luminosities at z≈6 as at z≈3; if the high-redshift plateau disappears once luminosity is matched, the claimed IGM attenuation signal is largely a selection artifact, while if a steep decline remains, the plateau is real.","supporting_citations":[{"cited_title":"2023, , 670, A4","cited_arxiv_id":null,"evidence_quote":"Supplies the MXDF DR2 catalogue and the reference spectra that define the parent sample of 477 galaxies."},{"cited_title":"Equivalent widths of Lyman $\\alpha$ emitters in MUSE-Wide and MUSE-Deep","cited_arxiv_id":"2202.06642","evidence_quote":"Provides the main MUSE-based comparison for double-peak fraction and peak-separation distributions."},{"cited_title":"R., Shapley , A","cited_arxiv_id":null,"evidence_quote":"Earlier UV-selected sample reporting 30% double-peaks, used as a literature baseline for the measured fraction."},{"cited_title":"2012, The Astrophysical Journal, 751, 29, publisher: The American Astronomical Society","cited_arxiv_id":null,"evidence_quote":"Earlier equivalent-width-selected sample reporting 50% double-peaks at z=3.1, used for comparison."},{"cited_title":"2015, , 578, A7","cited_arxiv_id":null,"evidence_quote":"Establishes the relation between Lyα peak separation and neutral hydrogen column density, motivating the physical interpretation of double peaks."},{"cited_title":"J., Runnholm , A., Gronke , M., & Scarlata , C","cited_arxiv_id":null,"evidence_quote":"Stacked spectra showing that the blue part of Lyα is increasingly suppressed with redshift, the trend the paper tests against its double-peak fraction."}],"review_version":1}