{"id":"b2f6561d-7ddd-40ff-b95e-51c334c63419","arxiv_id":"2506.15644","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"CDG-2, a clump of four globular clusters in the Perseus cluster, is validated as an almost dark galaxy: faint diffuse emission appears in both HST and Euclid imaging, with globular clusters contributing at least 16.6% of its light.","lead":"Using stacked Hubble Space Telescope images and new Euclid data, astronomers detected extremely faint diffuse starlight around a clump of four globular clusters in the Perseus cluster. This makes the clump, called CDG-2, the first galaxy ever identified purely from its globular clusters, and one of the most dark-matter-dominated galaxies known.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim rests on an unquantified diffuse-emission detection: Section 5 never reports a signal-to-noise ratio or a blank-sky control, so PSF-wing residuals or background-subtraction systematics could masquerade as the ~27 mag arcsec^-2 light.","rationale":"Good-faith reading: the paper does two things—re-detects a tight GC grouping with a fourth GC and quantifies its improbability, and claims detection of very faint diffuse light in HST and Euclid. The second is what elevates CDG-2 from a GC clump to a galaxy. For the central claim, it must be true that the ~27 mag/arcsec^2 emission is actual galaxy light. The paper provides images and model residuals but no S/N, no blank-sky control, no residual chi-square or detection probability. The two independent datasets add real support, but because the same procedural systematics (background subtraction scale, PSF-wing over/under-subtraction, isophote flexibility) can affect both, independence alone does not quantify significance. The absence of code and details of the background mesh make this hard to audit. I agree with the reader's weakest-assumption identification. The proposed control experiment would settle the concern. If it passes, the conditional verdict can become acceptance pending spectroscopy; if it fails, the claim of 'exceptionally strong evidence' is not supported. The halo-mass and GCLF ratios are explicitly acknowledged by the authors as assumption-heavy and are not the central issue; the detection of diffuse light itself is the load-bearing step.","tokens_in":23790,"tokens_out":4781,"duration_ms":62675,"concrete_test":"Re-run the exact Section 5 pipeline on (a) ~100 random blank-sky positions in the same Euclid ERO tile and (b) ~100 realizations with only the four Table 1 GCs injected as PSFs at CDG-2's position into otherwise blank local sky, keeping the same background mesh and isophote settings. Compute the distribution of recovered diffuse flux. If CDG-2's recovered diffuse flux is less than 5 sigma above these control distributions, the diffuse-light detection is not established; if it is >5 sigma and the residual map shows no PSF-correlated pattern, the galaxy interpretation is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's strongest claim—that CDG-2 is a galaxy because faint diffuse emission is seen around its four GCs—depends entirely on the Section 5 isophote analysis. That analysis has no quantitative detection significance. The quoted uncertainty (M_iso = -12.3 ± 0.62 mag) is the dispersion over 100 iterations of the same fitting pipeline; it does not compare the fitted diffuse flux to the noise at that position, nor to a control sample. The pipeline first applies a local background estimator, then PSF-subtracts the GCs, then smooths and fits elliptical isophotes; any residual PSF-wing light from the four GCs (especially in Euclid VIS, which has broad scattered-light wings) or a background-subtraction mismatch on the scale of the 6'' aperture can be absorbed by the isophote model and appear as 'diffuse emission.' The HST/Euclid morphological agreement is encouraging but is not a significance test, since similar model-fitting systematics can operate in both images. The GC spatial clustering itself is quantified (posterior/prior ~2000; chance probability ~1.5e-5), but the diffuse component—the feature that turns a GC clump into a galaxy—has no equivalent chance probability. Section 5 also lacks the size of the local background mesh and the residual noise statistics, so the 16.6% GC light fraction inherits this uncalibrated systematic.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports the discovery and validation of Candidate Dark Galaxy-2 (CDG-2), a compact grouping of four globular cluster (GC) candidates in the Perseus cluster, originally found in Li et al. (2025a) using a statistical spatial-clustering analysis of GCs. The authors stack two HST/ACS images and use Euclid Early Release Observations to claim the detection of extremely faint diffuse emission (~27 mag/arcsec^2) surrounding the four GCs. They present a combined PSF-photometry and elliptical-isophote model to estimate the diffuse light, apply a mock-galaxy injection correction for flux outside the fitted isophotes, and cross-check the GC light fraction with Subaru and CFHT data. Under the assumption that the four GCs are the complete GC population, they derive a total V-band luminosity of (6.2±3.0)×10^6 L_sun and a GC light fraction of at least 16.6%; assuming a canonical GCLF raises the fraction to ~33%. They conclude that CDG-2 is the first galaxy discovered purely through its GC population and one of the most dark-matter-dominated galaxies known.","tokens_in":24037,"tokens_out":7875,"duration_ms":84930,"significance":"If the diffuse-emission detection holds up, this is an important result: CDG-2 would be the first galaxy confirmed through its GC population alone, and its extreme GC-to-total light ratio (16.6% or higher) would provide a strong constraint on galaxy formation scenarios in which most star formation occurs in massive, dense clusters. The statistical detection of the GC clump is well quantified (posterior/prior ~2000; chance probability ~1.5×10^-5), and the independent confirmation in Euclid data is a notable strength. The paper also benefits from a transparent description of its assumptions and from the mock-injection test in Appendix A. However, the central claim—the detection of diffuse emission—currently lacks a quantitative significance assessment, and the isophote-fitting procedure is not specified in enough detail to rule out systematic artifacts. These gaps undermine the 'exceptionally strong evidence' for a galaxy until addressed.","major_comments":[{"comment":"The diffuse-emission detection is asserted as 'significant' but no signal-to-noise ratio or detection significance is reported. The quoted uncertainty M_iso = -12.3 ± 0.62 mag is the dispersion of the isophote model over 100 iterations of the same fitting pipeline, not a comparison with the local noise or with a blank-sky control. Please provide: (i) the S/N of the diffuse component in both the stacked HST and Euclid images; (ii) the residual noise statistics after background subtraction; and (iii) a control experiment (e.g., running the same PSF+isophote fit at random 'empty' positions, or injecting only the four GCs without diffuse light) to show that PSF-wing residuals and background-subtraction systematics cannot produce the detected flux.","section":"Section 5, Figure 3(b)"},{"comment":"The background subtraction and isophote-fitting parameters are under-specified. The text mentions a 'SExtractor-like local background estimator' and a 6'' aperture for masking CDG-2 but does not give the background mesh size, the box size, or the criteria used to 'reject unrealistic results' when isophote centers are allowed to float. These choices directly affect the measured diffuse flux at ~27 mag/arcsec^2, which is close to the sky-noise limit. Please specify all parameters (including Gaussian smoothing scale in physical units) and demonstrate stability of the diffuse flux and GC light fraction to reasonable variations in the background scale and isophote constraints.","section":"Section 5"},{"comment":"The Subaru and CFHT cross-checks are described as supporting the 'quite robust' GC light-fraction estimate, but the manuscript provides no uncertainties or procedural details for these data, reporting only '~18%' and '~22%'. Without error bars or a description of the fitting procedure (or a reference to a companion paper), these numbers cannot be used to validate the Euclid-based result. Please either present the full analysis or temper the robustness claim.","section":"Section 5"},{"comment":"The argument that d_CDG-2 ≳35 Mpc is based on the claim that a bottom-heavy GCLF is 'rather unlikely' and 'has never been found before.' This uses the observed GC magnitude distribution to constrain the distance under an assumed GCLF, while the paper's own Section 6.3 acknowledges that dwarf-regime galaxies and UDGs may have non-canonical GCLFs. Please clarify that this distance constraint is conditional on the canonical-GCLF assumption and does not independently support the GCLF-corrected ratios.","section":"Section 6.1, Figure 4"}],"minor_comments":[{"comment":"The text 'the the final measurements' contains a duplicated article; please fix.","section":"Section 2"},{"comment":"The chance probability of 1.5×10^-5 is quoted without explaining how it is computed from the posterior background counts; a more explicit description or reference would help.","section":"Section 4"},{"comment":"The filter transformation M_IE = M_V - 0.5 mag is given without a citation or SED assumption; please provide the reference and justify the offset.","section":"Section 5"},{"comment":"Please define what the 'Model' and 'Data' curves represent and specify the aperture used for the radial profiles.","section":"Figure 3(b), right panel"},{"comment":"The half-light radii for CDG-2-GCC3 and GCC4 are listed as ≲2 pc; please provide upper limits with proper uncertainty notation or a note clarifying the measurement.","section":"Table 1"},{"comment":"The specific frequency S_N = 345^{+323.1}_{-112.5} is reported without explanation of how the asymmetric uncertainties are derived; please clarify.","section":"Section 6.2"},{"comment":"The list of software includes 'photoutils', which should be 'photutils'.","section":"Software section"},{"comment":"The informal sentence thanking J. Li for 'making sure that his eyes were not playing ticks on him' is out of place in a journal article; consider rewording.","section":"Acknowledgments"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses an interesting question and the independent Euclid confirmation is a strength. The main gap is the lack of a quantitative detection significance for the diffuse emission; this is essential before the galaxy claim can be accepted. The paper is likely suitable for publication after a major revision that includes the requested control tests and parameter specifications. I would not recommend rejection at this stage, as the issue appears addressable with additional analysis of existing data."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nQuick take: this is a genuinely interesting candidate—possibly the first galaxy discovered purely through its globular cluster population—but the core new claim, the detection of diffuse emission, is not quantitatively justified. The GC clustering is on very solid footing; the diffuse light is not.\n\nWhat's new: Li et al. had already reported CDG-2 as a GC clump. Here they add a fourth GC from a deeper catalog, boosting the clustering signal roughly tenfold, and then report faint diffuse emission in stacked HST and Euclid images. The Euclid detection is independent and the morphology matches HST, which is encouraging. They also cross-check the GC light fraction with Subaru and CFHT, getting ~18% and ~22% versus 16.6% from Euclid. That consistency is a real point in their favor. The paper is transparent about its assumptions—distance, GCLF, mass-to-light ratios—and explicitly frames the 16.6% as a lower bound.\n\nThe soft spot is exactly where the stress-test lands: Section 5 never reports a signal-to-noise ratio for the diffuse emission, nor any control test on blank sky. The quoted uncertainty on the isophote magnitude is the scatter across 100 iterations of the same pipeline; it does not measure whether the fitted flux could arise from PSF-wing residuals or background-subtraction systematics. The HST/Euclid agreement helps rule out a PSF artifact from a single instrument, but it does not rule out correlated systematics from the local background estimator, especially at the ~27 mag/arcsec^2 level. Without a quantitative detection significance or an injected-blank-sky test, the 'significant' diffuse emission is an assertion, not a demonstrated result.\n\nThat said, the GC clustering alone is impressive (chance probability ~1.5e-5), so even without the diffuse light, the object is a strong dark-galaxy candidate. The derived physical quantities—L_V, dark matter fraction, S_N—are all conditional on the diffuse detection and on extrapolated scaling relations, so they should be treated with caution.\n\nBottom line: this deserves serious peer review. A referee should demand an explicit detection significance, a blank-sky or PSF-wing control, and ideally keep the diffuse-light claim separate from the GC-clustering claim. The authors have the data and the skills to provide that. If they do, this could be a significant result.\n\nRecommendation: send to review, with a request for the missing significance tests.","headline":"Likely a real galaxy, but the paper's central claim of diffuse light lacks a quantitative significance, so the physical numbers are provisional until that is nailed down.","tokens_in":24660,"tokens_out":2161,"would_cite":true,"duration_ms":25699,"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":"First galaxy found purely through its globular clusters.","keywords":["dark galaxies","globular clusters","ultra-diffuse galaxies","low surface brightness galaxies","Perseus cluster","Hubble Space Telescope","Euclid survey","globular cluster luminosity function"],"falsifier":"A targeted observation that would settle the claim: acquire deep, higher-resolution imaging or spectroscopy that can separate the diffuse component from the globular clusters' point-spread-function wings; for example, JWST/NIRCam imaging where the PSF is much narrower, or a spectrum of the diffuse light showing stellar absorption lines. If the diffuse signal disappears once accurate PSF wings are modeled, or if the detected light is shown to be an artifact of background subtraction, the central claim fails.","tokens_in":23554,"feed_emoji":"🌌","tokens_out":5663,"duration_ms":57757,"temperature":0.7,"pith_summary":"Candidate Dark Galaxy-2 (CDG-2) is a tight grouping of four globular clusters in the Perseus cluster with no previously visible galaxy light around them. This paper reports extremely faint diffuse emission around the four clusters in both stacked Hubble Space Telescope images and Euclid survey images, with matching morphology in the two independent data sets. If that diffuse light is real, CDG-2 is a genuine galaxy and the first one ever discovered purely through its globular cluster population. The authors estimate the galaxy's total V-band luminosity at about $L_{V,\\mathrm{gal}} = 6.2 \\pm 3.0 \\times 10^6\\,L_\\odot$, with at least 16.6% of its light coming from the globular clusters themselves, and, under standard scalings between globular cluster counts and halo mass, a dark matter fraction between 99.94% and 99.99%.","feed_headline":"First galaxy found purely through its globular clusters","feed_subtitle":"Faint diffuse starlight detected around four clusters suggests a galaxy that is 99.94% or more dark matter.","key_machinery":"The detection pipeline is a trans-dimensional Markov chain Monte Carlo point-process model that treats globular cluster positions as three overlapping point processes (intergalactic medium, bright galaxies, and UDG/dark galaxies) and outputs a posterior probability map for the location of UDG/dark galaxy centers; the diffuse-light validation proceeds by iterative PSF photometry of the four clusters, Gaussian smoothing of the cluster-subtracted image, and elliptical isophote fitting on both stacked HST and Euclid data. The globular-cluster-to-halo mass scalings, specifically $M_{\\mathrm{GC}}/M_h \\approx 2.9 \\times 10^{-5}$ from Harris et al. (2017) and $M_h \\approx 5 \\times 10^9 N_{\\mathrm{GC}} \\, M_\\odot$ from Burkert & Forbes (2020), convert the four detected clusters into a dark matter halo mass estimate, yielding the 99.94% to 99.99% dark matter fraction.","core_discovery":"The paper claims that CDG-2 is a galaxy. The four globular clusters, spanning roughly 1.2 kpc, are not a chance clumping of intergalactic globular clusters: a Bayesian point-process analysis puts the chance probability at about $1.5 \\times 10^{-5}$, and the stacked HST and Euclid images both show diffuse emission with the same morphology around the clusters. The measured mean surface brightness is about 27 mag arcsec$^{-2}$ in the Euclid band, and after subtracting the cluster point-spread functions and fitting elliptical isophotes, the residual diffuse light yields a total luminosity of $L_{V,\\mathrm{gal}} = 6.2 \\pm 3.0 \\times 10^6\\,L_\\odot$. The authors present this as the first detection of a galaxy that was found through its globular cluster population alone, and they argue that the concordance of two independent data sets, together with the extremely low chance probability of the clustering, justifies calling CDG-2 an almost dark galaxy.","pith_inferences":["The close coupling between the inferred GC light fraction and the assumed distance (Figure 4) suggests that a single radial-velocity measurement of one of the four globular clusters would sharpen the halo mass estimate dramatically; this is a concrete next step, though the paper does not spell it out as a prioritized observation.","The same stacking-and-isophote validation strategy used here could be applied to the other globular cluster overdensities already found in Perseus, notably CDG-1, to determine whether almost dark galaxies are a population rather than a single anomaly.","If the diffuse emission in CDG-2 is confirmed spectroscopically, the GC-overdensity search method becomes a general tool for finding galaxies that are invisible in conventional surface-brightness surveys, potentially enlarging the census of extreme dark-matter-dominated dwarfs in Euclid data."],"forward_implications":["CDG-2 is one of the faintest galaxies known to host globular clusters, with one of the highest globular-cluster light fractions ever measured (at least 16.6%, and likely ~33% if a canonical globular cluster luminosity function is assumed).","If CDG-2 has a canonical GC luminosity function, its dark matter halo mass fraction rises to ≳99.99%, making it potentially the most dark-matter-dominated galaxy known.","The existence of CDG-2 supports formation scenarios in which most star formation occurred in dense, massive clusters that later became globular clusters, with little diffuse star formation in the field.","The confirmation of CDG-2 makes CDG-1, a similar globular cluster clump with no detected diffuse emission, a plausible even-more-extreme twin; if CDG-1 is a real galaxy, it could be the first object with essentially no field stars around its globular clusters.","These objects provide testbeds for dark matter models, such as fuzzy or axionic dark matter, through the presence of compact dark matter halos that can host such extreme systems."],"supporting_citations":[{"why":"Supplies the statistical point-process method that discovered CDG-2 and frames the detection framework.","marker":"Li et al. (2025a)"},{"why":"Proposes the globular-cluster-overdensity search and reports the first candidate dark galaxy, CDG-1, establishing the background approach.","marker":"Li et al. (2022)"},{"why":"Provides the PIPER HST survey data and initial GC catalog from which CDG-2 was drawn.","marker":"Harris et al. (2020)"},{"why":"The deeper DOLPHOT-based GC catalog that added the fourth globular cluster and boosted the detection signal of CDG-2.","marker":"Li et al. (2025b)"},{"why":"The Euclid Early Release Observations of the Perseus cluster that supply the independent diffuse-light detection.","marker":"Marleau et al. (2024)"},{"why":"The globular-cluster-to-halo mass relation used to estimate CDG-2's dark matter halo fraction.","marker":"Harris et al. (2017)"},{"why":"The alternative globular-cluster-number-to-halo-mass scaling used to cross-check the halo mass estimate.","marker":"Burkert & Forbes (2020)"}],"fun_headline_variants":["First galaxy found via its globular clusters","Almost dark galaxy confirmed: 99.94% dark matter","Dark galaxy candidate validated by faint starlight","Perseus cluster galaxy is 99.99% dark matter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The diffuse emission at about 27 mag arcsec$^{-2}$ around the four globular clusters is real galaxy light rather than residual flux from the clusters' point-spread functions or a systematic error in the local background subtraction used before isophote fitting.","fun_headline_variants_meta":{"raw":{"variants":["First galaxy found via its globular clusters","Almost dark galaxy confirmed: 99.94% dark matter","Dark galaxy candidate validated by faint starlight","Perseus cluster galaxy is 99.99% dark matter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000217,"raw_usage":{"total_tokens":1541,"prompt_tokens":1154,"completion_tokens":387,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":770,"completion_tokens_details":{"reasoning_tokens":322}},"tokens_in":770,"tokens_out":387,"duration_ms":4935,"temperature":1.0,"reasoning_tokens":322,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T23:51:57.560142+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A targeted observation that would settle the claim: acquire deep, higher-resolution imaging or spectroscopy that can separate the diffuse component from the globular clusters' point-spread-function wings; for example, JWST/NIRCam imaging where the PSF is much narrower, or a spectrum of the diffuse light showing stellar absorption lines. If the diffuse signal disappears once accurate PSF wings are modeled, or if the detected light is shown to be an artifact of background subtraction, the central claim fails.","supporting_citations":[],"review_version":1}