{"id":"3075b190-8e83-46fd-8f7e-2abcc65a3c15","arxiv_id":"2509.11055","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Non-linear structure formation changes the global 21-cm signal by up to ~15 mK, a clumping effect that could probe dark matter fluctuations on sub-galactic scales (about 20 million solar masses), under optimistic foreground assumptions.","lead":"This paper simulates how small clumps of dark matter and gas in the early universe alter the sky-averaged 21-cm radio signal, and finds the effect could be detected with future radio arrays. If real, this opens a new way to probe dark matter's behavior on sub-galactic scales, far smaller than current measurements reach.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Foreground degeneracy, acknowledged in Supp. Note 3, undercuts the headline detectability: 3-parameter foreground cuts Dark Ages clumping S/N to 3.4σ and Saturated-Coupling z=30–20 to 0.42σ.","rationale":"The simulation physics appears carefully done: uniform-simulation subtraction, resolution convergence to ~4–7% at N=1024, explicit redshift-space distortion treatment, and inclusion of streaming velocities are real strengths. None of those raise a fatal concern. The load-bearing weak point is not the physical signal calculation, but the translation from predicted ΔTb to detection significance. The reader's weakest_assumption correctly identifies this; my independent reading agrees. The paper is honest in disclosing the foreground degeneracy, but the abstract's 'unambiguous' and 'single antenna suffices' language is stronger than the disclosed numbers support. Verdict stays CONDITIONAL: the physical result is credible, the observational claim is not fully established.","tokens_in":28205,"tokens_out":4180,"duration_ms":50237,"concrete_test":"Recompute all Table 1 clumping S/N values using the 2- and 3-parameter foreground models already defined in Supp. Note 3, but with a realistic, redshift-dependent xα(z) (e.g., the vanilla model in Supp. Fig. 11) replacing the fixed Saturated/Moderate coupling cases. If, with 3 foreground parameters, the single-antenna Cosmic Dawn clumping S/N remains below ~3σ or the Dark Ages array S/N below ~5σ, the headline detectability/discriminating claims should be downgraded from 'unambiguous' to 'conditional on foreground spectral smoothness and instrument calibration.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central forward-looking claim is that the clumping effect is detectable 'unambiguously' with a Dark Ages array (7.47σ) and with a single antenna during Cosmic Dawn (15.4σ, Saturated Coupling). This rests on the observability model in the main text and Supp. Note 3, which fits the foreground with a single free parameter A·ν^{-2.6} and neglects calibration/beam uncertainties. The authors themselves show that adding just one or two extra polynomial foreground terms reduces the Dark Ages array detection from 7.5σ to 3.4σ, and the Saturated Coupling z=30–20 single-antenna clumping detection from 15σ to 0.42σ. A 3.4σ or 0.42σ result is not 'unambiguous.' The supplementary defense—that a realistic, rapidly rising Lyα coupling would break the degeneracy—is plausible but not quantified; no forecast is shown for that case with a 2–3 parameter foreground. Since detectability is the raison d'être of the paper, the abstract and discussion overstate what the presented analysis supports.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses a combination of 3 Mpc GADGET SPH simulations (512^3 particles, 17 combinations of large-scale overdensity and baryon--dark matter streaming velocity, plus WDM-like cutoffs) and a 384 Mpc semi-numerical 21cmSPACE grid to compute the global 21-cm brightness temperature during the Dark Ages and early Cosmic Dawn. The authors find that non-linear clumping changes the sky-averaged signal relative to a homogeneous universe by up to ~0.5 mK in the Dark Ages and up to ~15 mK in the saturated Ly-alpha coupling limit at z~20. They report that the clumping effect is separable from a one-parameter foreground at 7.47 sigma with a Dark Ages antenna array and at 15.4 sigma with a single antenna in the Saturated Coupling z=30--20 case, and that a kcut=100 h/Mpc cutoff could be distinguished at several sigma. They interpret this as a probe of small-scale power down to ~50 kpc / 2x10^7 Msun.","tokens_in":28561,"tokens_out":9680,"duration_ms":123324,"significance":"The physical calculation is a clear step beyond previous small-box studies: it is the first to include streaming velocities and large-scale density variations in this context, it validates the homogeneous baseline against CAMB to ~1%, demonstrates resolution convergence to 4--6.7% at z=30, and uses a uniform-simulation subtraction to control numerical offsets. The predicted clumping correction is a concrete, falsifiable prediction of the CDM model, and the WDM-like cutoff comparisons usefully quantify sensitivity. If the signal can be measured, it would indeed open a new window on small-scale dark matter clustering. The main weakness is that the observability forecasts, which are central to the paper's impact, rest on an optimistic foreground model whose failure is demonstrated in the authors' own Supplementary Note 3.","major_comments":[{"comment":"The headline detection claims (7.47 sigma for the Dark Ages array; 15.4 sigma for Saturated Coupling z=30-20 with a single antenna) are computed with a foreground model containing a single free parameter A*nu^{-2.6}. The authors' own Supplementary Note 3 shows that adding one or two polynomial foreground terms reduces these significance values to 3.4 sigma and 0.42 sigma, respectively. A 3.4 sigma or 0.42 sigma signal is not 'unambiguous'. The abstract and Discussion therefore overstate what the present analysis supports. The qualitative argument in Supp. Note 3 that a realistic, rapidly rising Ly-alpha coupling would break the degeneracy is plausible, but no forecast is shown for that case with a 2--3 parameter foreground; such a forecast is needed to support the Cosmic Dawn detectability claim.","section":"Observability; Table 1; Supplementary Note 3"},{"comment":"The strongest Cosmic Dawn claims rely on the Saturated Coupling prediction at z~20. The simulations deactivate atomic/molecular cooling 'to prevent the collapse of gas in minihalos', and the paper states the gas statistics are valid down to z~20. No test of this assumption is presented. At z~20, minihalos have gone non-linear, and cooling/collapse could plausibly alter the dense-gas temperature distribution that generates the clumping signal. The authors should quantify the sensitivity of Delta_hTb to this modeling choice, e.g., by comparing with runs that include cooling but suppress star formation, or by estimating the effect of unresolved minihalos.","section":"Methods, 'Numerical simulations'; Table 1"}],"minor_comments":[{"comment":"Typo: 'devided' should be 'divided' in the definition of h.","section":"Introduction"},{"comment":"The noise curves in Figure 2 are described as using bins with Delta(ln nu)=1, while the significance calculation in Supp. Note 3 uses Delta nu=1 MHz. Please clarify which binning is used for Table 1 and why the displayed noise curves differ.","section":"Figure 2 caption; Supplementary Note 3"},{"comment":"The approximate mapping between kcut and WDM particle mass (e.g., kcut=100 h/Mpc ~ 7 keV) is given without a formula. Adding the assumed relation would help readers connect the illustrative models to current constraints.","section":"Figure 1 and Supplementary Figure 7"},{"comment":"The text calls Moderate Coupling the 'most pessimistic early Cosmic Dawn case'; this refers only to astrophysical assumptions. The foreground treatment is at the same time explicitly optimistic. Please distinguish these two axes of uncertainty to avoid confusion.","section":"Observability"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this. The clumping calculation is worth taking seriously. What's new: first to combine streaming velocity and large-scale density/velocity fields with resolved sub-Mpc hydrodynamics in a global 21-cm prediction, and to show that small-scale redshift-space distortions change the result at the tens-of-percent level. The validation is genuinely good: CAMB agreement to ~1%, resolution convergence at the 4-6.7% level at z=30, and the uniform-simulation subtraction is the right way to isolate a per-cent-level clumping signal from numerical noise. The physical conclusion—that sub-Mpc clumping shifts the Dark Ages global signal by ~0.5 mK and saturated-coupling signal by ~15 mK at z=20, and that a WDM-like cutoff suppresses it strongly—is well supported. This is a clear step past Xu et al. (2021).\n\nSoft spots. The abstract's 'detected unambiguously' is not supported by the paper's own analysis. Supp. Note 3 shows that adding a second or third foreground polynomial term drops the Dark Ages clumping significance for an array from 7.5σ to 3.4σ, and the Saturated Coupling z=30–20 single-antenna clumping significance from 15σ to 0.42σ. The suggested remedy—that a realistic, rapidly rising Lyα coupling will break the degeneracy—is plausible, but no forecast with realistic coupling and a multi-parameter foreground is shown. Calibration and beam errors are also set aside. So the forecast S/Ns are conditional on an optimistic foreground model. That is not a fatal flaw, but the abstract and discussion overstate what is demonstrated. Also, the modified GADGET code and 21cmSPACE grid are not public; only one figure's source data is shared. For a paper making per-cent-level claims, independent checking is harder than it should be.\n\nOverall: the simulation result is solid and the foreground caveat is out in the open, even if the headline language ignores it. The value is in the clumping prediction and the RSD effect, not the '10σ' detectability numbers. I would cite it for the former. I want a serious referee to see it; with a more honest observational-claims section it becomes a good paper. The stress-test note holds up.","headline":"Strong simulation result; the 'unambiguous detection' language is not backed by the paper's own foreground-degeneracy analysis.","tokens_in":28983,"tokens_out":2924,"would_cite":true,"duration_ms":32260,"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":"This paper establishes that non-linear structure formation (clumping) alters the sky-averaged 21-cm brightness of hydrogen by up to ~15 mK during Cosmic Dawn and ~0.5 mK during the Dark Ages, making the small-scale power spectrum of matter","keywords":["global 21-cm signal","dark ages","cosmic dawn","dark matter clumping","small-scale power spectrum","warm dark matter","fuzzy dark matter","Lyman-alpha coupling"],"falsifier":"A global 21-cm spectrum measurement covering roughly z=25–70 (ν≈20–55 MHz) with sensitivity to ~0.1 mK residuals after foreground subtraction would settle the central claim: if the observed spectrum matches the homogeneous-universe prediction to better than ~0.1 mK (with no wiggle of ~0.5 mK amplitude), the clumping effect would be ruled out at that amplitude, implying either a suppressed small-scale power spectrum or a foreground model that absorbs the signal.","tokens_in":28139,"feed_emoji":"📡","tokens_out":9851,"duration_ms":85767,"temperature":0.7,"pith_summary":"This paper calculates how the growth of small-scale structure — the clumping of gas and dark matter that later forms the first stars — modifies the sky-averaged (global) 21-cm brightness of neutral hydrogen during the Dark Ages and early Cosmic Dawn. It finds that non-linear density fluctuations shift the signal by up to ~0.5 mK during the Dark Ages and up to ~15 mK once Lyman-α coupling from the first stars is saturated. The authors argue that this clumping signature is not a foreground to be subtracted, but a physical signal that encodes the amplitude of density fluctuations on comoving scales of ~50 kpc, corresponding to masses of ~2×10^7 solar masses. With a single foreground parameter, the effect would be detectable at 7.5σ by an array of global-signal antennas during the Dark Ages, and at 15–20σ by a single antenna during Cosmic Dawn. Because popular dark-matter models such as warm and fuzzy dark matter suppress these small-scale fluctuations, the global 21-cm signal becomes a test of dark matter in an otherwise inaccessible regime.","feed_headline":"Clumping of dark matter shifts the cosmic 21-cm signal by up to 15 mK","feed_subtitle":"A sky-averaged hydrogen radio measurement could probe sub-galactic scales beyond current dark-matter limits.","key_machinery":"The key machinery is a multi-scale simulation pipeline: a 3 Mpc smoothed-particle-hydrodynamics simulation of baryons and dark matter, with chemistry and Compton heating, run with 17 combinations of local overdensity and baryon–dark matter streaming velocity (plus 51 runs with Gaussian-cutoff initial conditions to mimic warm/fuzzy dark matter); and a 384 Mpc semi-numerical grid that assigns the simulated 21-cm signal to each 3 Mpc pixel based on its local density and streaming velocity. The observable of interest is the difference ΔhTb = Tb - Tb,h between the clumped and homogeneous predictions, computed by comparing against uniform simulations that share the same time steps, isolating the s","core_discovery":"The central discovery is that the standard homogeneous-universe calculation of the global 21-cm signal fails at the sub-mK level during the Dark Ages and at the level of ~10% during Cosmic Dawn: in standard CDM, clumping makes the signal less negative at early times (positive ΔTb up to 0.37 mK at z~64), then more negative after z~46 (down to -0.51 mK at z~27), and then, when Lyman-α coupling is strong, always weaker in absorption, reaching +14.7 mK at z=20. About half of this effect comes from fluctuations on scales smaller than roughly 50 kpc (comoving), i.e., from halos of mass ~2×10^7 solar masses. A WDM-like cutoff at that scale suppresses the clumping signal by ~50%, so the global signa","pith_inferences":["If the clumping correction is as large as predicted, it should also appear as a systematic contribution to 21-cm power-spectrum measurements from the same epochs; the authors note this but leave it for future work.","The same pipeline could be applied to models with enhanced small-scale clumping (e.g., primordial magnetic fields or primordial black holes), turning the global signal into a discriminating probe of those models as well.","A testable extension: the degeneracy between Lyman-α coupling strength and clumping amplitude can likely be broken by fitting the full redshift evolution, because realistic coupling rises sharply with time while the clumping cutoff is constant; a joint fit would tighten dark-matter constraints.","The predicted dark-ages wiggle in ΔhTb (a rise to ~+0.4 mK, then a fall to ~-0.5 mK, peaking near z~27) is a distinctive shape; a future lunar global-signal measurement that observes the 21-cm spectrum and finds no such feature would indicate either suppressed small-scale power or a foreground that mimics the signal."],"forward_implications":["In standard CDM, the maximum possible 21-cm absorption during Cosmic Dawn is ~13% weaker at z=15 than in a homogeneous universe; forecasts of the strongest absorption must be revised downward.","With a global array (100,000-hr equivalent) and a single foreground parameter, the Dark-Ages clumping effect is detectable at 7.5σ for z=200–20; with a single antenna and saturated Lyman-α coupling, it is detectable at 15–20σ for z=20–40.","Suppressing the smallest-scale fluctuations (as in a roughly 7 keV warm dark matter model) is distinguishable from CDM at 2.9σ in the Dark-Ages case and at 8.7–11.4σ during Cosmic Dawn.","The signal probes density fluctuations on comoving scales around 50 kpc — a mass scale of ~2×10^7 solar masses — roughly three orders of magnitude below current direct constraints.","Ignoring the baryon–dark matter streaming velocity overestimates the clumping signal by ~10–13% at z=30, so precise modeling of the global signal must include it."],"fun_headline_variants":["Dark matter clumping shifts cosmic 21-cm signal by up to 15 mK","Sub-galactic clumps leave 15 mK signature in global 21-cm signal","21-cm signal probes dark matter clumping on 20-million-solar-mass scales","Cosmic 21-cm signal could test dark matter clumping on sub-galactic scales"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The detection significance numbers assume that the radio foreground can be described by a single free parameter of the form A·ν^(-2.6) with negligible calibration and beam errors; if the foreground requires more parameters, the paper's own Supplementary Note shows the Dark-Ages clumping detection drops from 7.5σ to 3.4σ and the Saturated-Coupling single-antenna clumping detection falls to 0.42σ.","fun_headline_variants_meta":{"raw":{"variants":["Dark matter clumping shifts cosmic 21-cm signal by up to 15 mK","Sub-galactic clumps leave 15 mK signature in global 21-cm signal","21-cm signal probes dark matter clumping on 20-million-solar-mass scales","Cosmic 21-cm signal could test dark matter clumping on sub-galactic scales"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000297,"raw_usage":{"total_tokens":1582,"prompt_tokens":794,"completion_tokens":788,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":694}},"tokens_in":538,"tokens_out":788,"duration_ms":8875,"temperature":1.0,"reasoning_tokens":694,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T17:09:45.725231+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A global 21-cm spectrum measurement covering roughly z=25–70 (ν≈20–55 MHz) with sensitivity to ~0.1 mK residuals after foreground subtraction would settle the central claim: if the observed spectrum matches the homogeneous-universe prediction to better than ~0.1 mK (with no wiggle of ~0.5 mK amplitude), the clumping effect would be ruled out at that amplitude, implying either a suppressed small-scale power spectrum or a foreground model that absorbs the signal.","supporting_citations":[],"review_version":1}