{"id":"55c2d67a-c94e-4b1d-b2b1-b7f7c092901f","arxiv_id":"2506.06453","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Most fast radio bursts occur in the disks of their host galaxies, with about 11% consistent with origins in globular clusters.","lead":"Fast radio bursts are flashes of radio light from distant galaxies, and where they go off inside their home galaxies tells us what kind of star or object produced them. This study maps 37 FRB locations against host galaxy light and globular cluster distributions, finding most bursts in galactic disks while a minority come from old globular clusters.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 11±5% GC fraction rests on an early-type scaling relation applied to a mostly late-type sample; the paper's own exclusion of FRB 20211127I shows the relation can fail, so the headline number is not yet robust.","rationale":"The paper is a careful and transparent population study: the astrometric and Galfit modeling are detailed, the coflash package is open source, and the authors explicitly acknowledge key assumptions (Section 4.2 footnote, Section 5.2 exclusion, Section 7). The disk-majority claim is robust to reasonable threshold variations and persists in the low-redshift subsample. However, the most quantitative headline number, the 11±5% GC fraction, depends on extrapolating a scaling relation calibrated on early-type galaxies to a mostly late-type sample, and the paper's own data provide at least one failure of that relation (FRB 20211127I). This matches the reader's weakest assumption. The proposed test—directly comparing the relation to resolved late-type GC systems and re-running the analysis—would settle whether the GC fraction survives or shifts. Since the reader already recommended a conditional accept and my concern supports that posture, the verdict is unchanged.","tokens_in":49195,"tokens_out":6108,"duration_ms":66684,"concrete_test":"Use M81 (host of FRB 20200120E), the only sample host with directly resolvable GCs, as a calibration check: compare the predicted r_e,GC ≈ 9.09 kpc for log M* ≈ 10.86 (Table 3) with the measured effective radius of M81's globular cluster system. Extend the same comparison to the Milky Way and M33, for which GC system sizes are also known. If the relation is off by a factor ≳2 for these late-type calibrators, re-derive r_e,GC for all late-type hosts using the observed late-type relation (or recalibrated intercept), regenerate the GC models in coflash, and recompute ΔBIC preferences and the inferred GC fraction. A shift of >5 percentage points, or a change in whether FRBs 20210117A and 20220105A reach ΔBIC>3, would show the headline is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central GC fraction (Abstract: 11±5%) is built on synthetic GC radial distributions from the Lim et al. (2024) r_e,GC–M* relation (their Eq. 8; Section 4.2), calibrated on 118 early-type galaxies. The FRB host sample is 35/37 late-type, and the paper itself flags failure modes: in Section 5.2, FRB 20211127I is excluded because its predicted r_e,GC < r_e, 'likely due to a limitation in the applicability of the Lim et al. (2024) relation to low-mass, late-type galaxies.' The two non-confirmed GC candidates, FRBs 20210117A and 20220105A, are late-type hosts whose ΔBIC>3 GC preferences are derived from these extrapolated models. If the true GC systems of late-type galaxies have a different radial scale, the model likelihoods shift and the population-level GC fraction changes. Because the 11±5% value is a headline quantitative claim, this extrapolation is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents deep optical and near-infrared imaging of the host galaxies of 37 well-localized fast radio bursts (34 from CRAFT and three from the literature) and uses a likelihood framework to compare each FRB's position against four models: the smooth Sérsic light profile of the host, residual substructure (primarily spiral arms), and two synthetic globular cluster distributions (ellipsoidal and spherical) built from the Lim et al. (2024) r_e,GC–M* relation. The authors report a median projected offset of 4.2^{+5.7}_{-2.5} kpc (1.0^{+1.5}_{-0.6} r_e), find that the majority of FRBs favor the smooth-disk light model, identify 11±5% of the sample (four FRBs) as favoring a globular cluster origin at ΔBIC>3, and estimate that 20–46% of low-redshift (z<0.15) FRBs favor spiral-arm association. They conclude that FRB progenitors have multiple formation channels, with most associated with massive-star formation and a minority formed through dynamical channels.","tokens_in":49447,"tokens_out":10522,"duration_ms":97337,"significance":"The paper combines a comparatively large, homogeneous FRB host-galaxy sample with a new statistical tool (coflash) and carefully constructed synthetic GC distributions. The seven newly presented CRAFT FRBs and the public code are valuable community resources. If the central claims hold, the paper would provide one of the first quantitative population-level constraints on the fraction of FRBs originating in globular clusters, supporting multiple magnetar formation channels. The authors are transparent about several limitations, and the disk-versus-GC offset analysis is a useful diagnostic. However, the headline GC fraction is sensitive to sample selection and to the extrapolation of a scaling relation calibrated on early-type galaxies, so the quantitative result is not yet robust.","major_comments":[{"comment":"The sample is not representative of the FRB population because three non-CRAFT FRBs were added deliberately on account of being spiral or (candidate) globular-cluster events. Two of these additions, FRB 20200120E and FRB 20240209A, are among the four events that drive the 11±5% GC fraction. In the 34-object CRAFT-only subset, only FRBs 20210117A and 20220105A have ΔBIC>3 GC preferences, giving a fraction of 2/34≈6%, not 11±5%. The headline fractions should be reported separately for the unbiased CRAFT sample and for the full sample, with a discussion of how the intentional inclusion of exceptional events biases the population estimate.","section":"Section 2; Section 6.2; Abstract"},{"comment":"The exclusion of FRB 20211127I is applied inconsistently. The authors exclude this FRB because its inferred r_e,GC < r_e is attributed to a limitation of the Lim et al. (2024) relation for low-mass, late-type galaxies. However, FRB 20210117A, which is retained as a GC candidate with ΔBIC>3 in the NIR, has exactly the same property in that band: from Table 3, its K-band r_e = 1.96 kpc and r_e,GC = 1.30 kpc, so r_e,GC < r_e. This FRB also has the lowest stellar mass in the sample (log M*/M⊙ = 8.59). The paper needs to justify why the limitation that excluded FRB 20211127I does not also disqualify the NIR preference of FRB 20210117A, or consistently remove both; otherwise the 11±5% GC fraction is not robust.","section":"Section 4.2; Section 5.2; Table 3"},{"comment":"The BIC definition is non-standard. The paper defines k as \"the number of parameters in the model (representing the pixel intensity values)\" and then argues that k and n are identical across models, so ΔBIC reduces to Δ(-2lnL). This is a pure likelihood-ratio comparison with no penalty for model complexity. The Kass & Raftery (1995) thresholds (2<ΔBIC<6 as positive evidence, etc.) are calibrated for genuine BIC differences with meaningful parameter counts, and applying them to this unpenalized statistic is not justified. The claim that 11 FRBs have model preferences with ΔBIC>3 should be re-cast as a likelihood-ratio statement, or a proper information criterion with a defined parameter count should be used.","section":"Section 5.2, Eq. (2)"},{"comment":"The abstract mixes two different statistical standards in the same sentence. The statement that \"the majority of FRBs favor locations within the disks\" is based on the raw maximum-likelihood model preference (e.g., 13±3 Profile vs. 12±2 combined GC in the optical), while the \"11±5%\" GC fraction is based on the stricter ΔBIC>3 subset. Only 11 of 37 FRBs have any model preference reaching ΔBIC>3, and only four of those are GC. The abstract should either apply a consistent statistical threshold to both claims or explicitly state that the disk majority is a preferred-model count rather than a statistically significant association.","section":"Abstract; Section 5.1; Section 5.2"}],"minor_comments":[{"comment":"The phrase \"spiral structure is apparent in 86% of our sample of FRB hosts\" would be clearer if it noted that this applies to the z<0.15 subsample and that the spiral-arm fraction is explicitly a lower bound.","section":"Abstract"},{"comment":"Since coflash is described in a companion paper (Deller et al. in prep.), the present paper should include a brief mathematical description of how the likelihood is computed from the convolved probability maps so that the analysis is self-contained; the GitHub link alone is not sufficient for reproducibility.","section":"Section 4.3"},{"comment":"The paper would benefit from a compact table listing, for each GC candidate, the stellar mass, r_e, r_e,GC, and the filter(s) in which the GC preference appears, so that the exclusion criterion applied to FRB 20211127I is transparently visible for all events.","section":"Table 3"},{"comment":"There is a typo in the acknowledgments: \"Reserach Council\" should be \"Research Council\".","section":"Acknowledgements"},{"comment":"The coordinate entry for FRB 20241027B lists R.A. as \"2:24:07.29\" with a missing leading zero; this should be \"02:24:07.29\" for consistency with the other entries.","section":"Table 1"},{"comment":"The sentence \"we find no support for a majority of observed FRBs arising from GCs\" is a reasonable qualitative conclusion, but it would be strengthened by a direct computation of the GC fraction using only the CRAFT-selected sample, as suggested in the major comments.","section":"Section 6.2"}],"recommendation":"major_revision","confidential_remarks":"This is a valuable observational study with a novel methodological framework and new data, and the authors are commendably transparent about many of its limitations. However, the headline GC fraction of 11±5% is not yet robust: it is inflated by the deliberate inclusion of two GC-selected events, and one of the retained non-GC-confirmed candidates (FRB 20210117A) sits in exactly the regime the authors use to exclude another event (FRB 20211127I). The ΔBIC statistic as defined also does not support the invoked information-theoretic thresholds. These issues are fixable in revision, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is the first systematic census of where FRBs sit inside their hosts, and it is worth a serious look. The authors add seven new CRAFT hosts and apply a coherent likelihood framework (coflash) comparing four spatial models — smooth light, residual substructure, and two synthetic GC distributions — across 37 well-localized FRBs. The headline that most FRBs trace the disk, with a minority (11±5%) favoring GC-like distributions and a modest low-z spiral-arm fraction (20–46%, explicitly a lower bound), is a real step forward for the subfield.\n\nThe paper is also unusually honest about its own limitations. Section 5.2 states that FRB 20211127I is excluded because the Lim et al. (2024) r_e,GC–M* relation, calibrated on 118 early-type galaxies, likely fails for low-mass, late-type hosts. That admission is exactly the right thing to do, but it points at a load-bearing assumption: the sample is 35/37 late-type, and the headline 11±5% GC fraction rests on extrapolating an early-type scaling relation to a mostly late-type population. The two unconfirmed GC candidates (20210117A, 20220105A) are late-type hosts whose preferences come from those models. If the true GC radial scales differ, the population fraction moves. The paper's own exclusion proves the mechanism can fail.\n\nOther soft spots are real but smaller. Only about one third of model preferences reach ΔBIC>3; most are weak, which the authors acknowledge. The Anderson-Darling test does not separate the disk and GC offset distributions (p<0.05 only 9–16% of draws). The spiral fraction is honestly labeled a lower bound. None of this breaks the central disk-majority claim, but the quantitative GC fraction should be read as provisional until the GC scaling relation is validated on late types or replaced with direct GC constraints.\n\nWhat is new and solid: the offset catalog, the public coflash extension, the careful treatment of astrometric and localization uncertainties, and a reproducible methodology for classifying FRB environments. Self-citation of Lim et al. is not a problem here because the relation is external and the FRB positions are not used to fit it.\n\nWho is this for? FRB phenomenologists and host-galaxy folks will get immediate value; transient demographics people can use the framework. It deserves a serious referee, with the expectation of heavy revision or at least a robustness test using alternate GC scaling relations and reporting the fraction without the excluded outlier.\n\nRecommendation: send it to review.","headline":"First systematic census of FRB locations within host galaxies; the disk-majority claim is solid, but the headline GC fraction depends on an extrapolated scaling relation and should be treated as provisional.","tokens_in":50069,"tokens_out":2298,"would_cite":true,"duration_ms":24140,"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":"Most fast radio bursts trace the disks of their host galaxies, not globular clusters.","keywords":["fast radio bursts","magnetars","globular clusters","host galaxy environments","spiral arms","galactocentric offsets","ASKAP CRAFT survey","stellar populations"],"falsifier":"Resolve the globular-cluster populations of a handful of nearby (within a few tens of Mpc) late-type FRB host galaxies and measure their radial distribution; if it does not match the early-type-calibrated r_e,GC–M* relation, the synthetic GC models lose their foundation. Alternatively, VLBI localizations of FRBs 20210117A and 20220105A would show whether either coincides with a resolved globular cluster.","tokens_in":49009,"feed_emoji":"📡","tokens_out":4250,"duration_ms":39510,"temperature":0.7,"pith_summary":"This paper maps where 37 well-localized fast radio bursts sit inside their host galaxies and asks which stellar population they trace. It finds that most bursts are consistent with the smooth stellar light of galactic disks, while only 11±5% statistically favor the spatial distribution of globular clusters, almost all at large projected offsets. This matters because it suggests that most FRB progenitors form alongside massive stars, while a smaller dynamical channel produces magnetars in old star clusters. The result implies multiple magnetar formation pathways and would guide searches for FRB counterparts.","feed_headline":"Most fast radio bursts sit in galaxy disks","feed_subtitle":"A 37-burst census finds ~11% favor globular clusters, pointing to multiple magnetar birth channels.","key_machinery":"The analysis rests on a likelihood framework that places each FRB's localization on four normalized spatial models: the smooth Sérsic surface-brightness profile of the host, its residual substructure (mostly spiral arms), and two synthetic globular-cluster distributions (ellipsoidal and spherical) built from a scaling relation between host stellar mass and globular-cluster effective radius. The framework convolves each model with the FRB position uncertainty and compares likelihoods through ΔBIC. This machinery converts a single FRB position into a statement about which stellar population it most plausibly traces.","core_discovery":"The paper's central claim is that the majority of FRBs favor locations within the disks of their host galaxies, while a minority of about 11±5% favor the spatial distribution of globular clusters, primarily FRBs with galactocentric offsets ≳3 r_e. Based on 34 ASKAP/CRAFT FRBs plus three literature events, the authors build surface-brightness and residual substructure models of each host and synthetic globular-cluster distributions, then compare likelihoods. They conclude that, assuming FRBs come from magnetars, most progenitors are associated with massive-star formation and a minority form through dynamical channels in old stellar systems. They also find that at z<0.15, where spiral structure is visible in 86% of hosts, roughly 20–46% of FRBs favor association with spiral arms, a fraction well below the ≳88% seen for core-collapse supernovae, arguing that high star-formation efficiency is not the dominant driver of FRB progenitor production.","pith_inferences":["The 11±5% globular-cluster fraction is likely a lower bound because the framework cannot identify GC origins for bursts that project near the bright galaxy center, where smooth light dominates.","If future samples at low redshift with VLBI-class localizations reproduce the ~20–46% spiral-arm fraction, the comparison with core-collapse supernovae would point to a genuine age or environment difference between FRB and CCSN progenitors.","Applying the same likelihood machinery to FRB samples from other telescopes, or to Type Ia supernova positions, would test whether FRB progenitors more closely trace star formation or stellar mass."],"forward_implications":["If the central claim holds, most FRB progenitors are young magnetars born from core-collapse supernovae in galactic disks.","A minority (~11±5%) of FRBs would arise from dynamical channels in globular clusters, implying at least two distinct magnetar formation paths.","The low spiral-arm association fraction relative to core-collapse supernovae implies that the most intense star-forming environments are not required to make FRB progenitors.","Host-normalized offsets beyond about 3 r_e become a useful flag for globular-cluster origins, and only a couple more such events would make the disk and GC offset distributions statistically distinct.","The seven newly presented hosts extend the redshift and imaging baseline for FRB environment studies."],"supporting_citations":[{"why":"Supplies the r_e,GC–M* scaling relation used to build all synthetic globular-cluster distributions.","marker":"Lim et al. (2024)"},{"why":"Miliarcsecond localization that confirmed FRB 20200120E's globular-cluster origin, the anchor for the GC channel.","marker":"Kirsten et al. (2022)"},{"why":"Associated FRB 20200120E with M81 and provided the host properties used in modeling.","marker":"Bhardwaj et al. (2021)"},{"why":"Established FRB 20240209A as a strong globular-cluster candidate in a massive elliptical galaxy.","marker":"Shah et al. (2025)"},{"why":"Presented the host of FRB 20240209A and its quiescent, massive elliptical properties.","marker":"Eftekhari et al. (2025)"},{"why":"Provides the CRAFT sample, host galaxy catalog, and the bulk of the deep optical and near-infrared imaging.","marker":"Shannon et al. (2025)"},{"why":"Supplies HST imaging and offset measurements for earlier FRB hosts that anchor the sample.","marker":"Mannings et al. (2021)"},{"why":"Demographic comparison suggesting FRB progenitors may trace star formation plus stellar mass, the context for multiple channels.","marker":"Horowicz & Margalit (2025)"}],"fun_headline_variants":["Most FRBs reside in galaxy disks, not globulars","FRB mapping: disks host most bursts, globulars ~11%","Fast radio bursts favor disk locations in hosts","Galaxy disks birth most FRBs, globulars a minority"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The inferred globular-cluster fraction and the identification of FRBs 20210117A and 20220105A as globular-cluster candidates rest on a scaling relation for globular-cluster sizes calibrated on early-type galaxies and extrapolated to a sample that is mostly late-type disks; if that relation does not hold for these galaxies, the globular-cluster model probabilities—and therefore the 11±5% fraction—are unreliable.","fun_headline_variants_meta":{"raw":{"variants":["Most FRBs reside in galaxy disks, not globulars","FRB mapping: disks host most bursts, globulars ~11%","Fast radio bursts favor disk locations in hosts","Galaxy disks birth most FRBs, globulars a minority"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000536,"raw_usage":{"total_tokens":2638,"prompt_tokens":1070,"completion_tokens":1568,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":686,"completion_tokens_details":{"reasoning_tokens":1497}},"tokens_in":686,"tokens_out":1568,"duration_ms":12249,"temperature":1.0,"reasoning_tokens":1497,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:55:58.178752+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Resolve the globular-cluster populations of a handful of nearby (within a few tens of Mpc) late-type FRB host galaxies and measure their radial distribution; if it does not match the early-type-calibrated r_e,GC–M* relation, the synthetic GC models lose their foundation. Alternatively, VLBI localizations of FRBs 20210117A and 20220105A would show whether either coincides with a resolved globular cluster.","supporting_citations":[],"review_version":1}