{"id":"5368e56c-07b6-406d-a77a-dd6a6ab38dc7","arxiv_id":"2607.14211","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Planck CMB anisotropy data allow up to roughly 3–5% of the CMB energy density to come from high-redshift dust, while excluding a fully non-primordial CMB at very high significance.","lead":"This paper tests whether a small fraction of the cosmic microwave background could have been produced by dust around very early galaxies rather than by the Big Bang itself. Using Planck's maps of tiny temperature fluctuations, it finds that a few percent of the CMB energy density could be non-primordial, consistent with the 1.4% prediction from Gjergo & Kroupa, but the data cannot detect it.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Recombination history not rescaled with ε_CMB: the quoted bounds may shift once recfast is made consistent with the reduced photon density at z>zt.","rationale":"The reader's weakest_assumption pinpoints exactly the place where the model's self-consistency is least secure. The paper implements the ε_CMB rescaling in the background and perturbation equations but leaves recfast using the unmodified photon temperature. Since z_t=17 is well below recombination, the rescaling is active throughout the epoch when the acoustic peaks and damping tail are set. The authors' physical reasoning for the step function implies the recombination-era radiation field should be the primordial component only, so its effective temperature should be ε_CMB^{1/4}T0(1+z); using the full T0(1+z) in recfast applies a different recombination history. The magnitude of the resulting shift in z* and in the ε constraint is not quantified. This is not an external-model criticism; it is an internal consistency check of the very model being fit. I do not see a reason to reject the paper on this basis — the analysis is careful, the baseline recovery is good, and the Run 2/Run 3 checks are sensible — but the central numerical bounds should be regarded as provisional until the recombination treatment is made consistent. A single modified-recfast MCMC run would settle whether the effect is negligible or not. I therefore recommend leaving the conditional verdict unchanged.","tokens_in":37758,"tokens_out":7595,"duration_ms":80455,"concrete_test":"Re-run the Run 1 and Run 2 MCMCs after modifying recfast/HyRec to use the rescaled radiation field at z>z_t — either as a blackbody with T_eff = ε_CMB^{1/4} T0(1+z) or, more accurately, as a grey-body intensity ε_CMB B_ν(T0(1+z)) — keeping all other settings and likelihoods unchanged. If the resulting 95% lower bound ε_CMB ≳ 0.953 shifts by more than the Run 2 1σ width (~0.025), or if Δχ² changes by more than ~1 for the fixed best-fit ε values, the quoted constraints and the consistency claim are not yet established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing issue is the inconsistent treatment of recombination. In the model, Eq. (1) reduces the photon energy density to ε_CMB ρ_γ(z) at all z > z_t = 17, including the entire recombination epoch. In the GK25 scenario the dust is emitted at z≈17, after recombination, so the radiation field during recombination is purely primordial and its temperature is T_γ(z) = ε_CMB^{1/4} T0(1+z). The paper states in §3.1 that recfast is left unmodified, with T_γ(z)=T0(1+z), and §5.2 defends this as a deliberate modelling choice on the grounds that no unique effective temperature can be assigned because the component is \"assumed to remain Planckian in form\". This defence is questionable: for a Planckian spectrum, ρ∝T^4, so rescaling the energy density by ε_CMB at fixed spectral shape is equivalent to a lower temperature (or to a grey-body with emissivity ε_CMB, which also changes the photoionisation rates used in recfast). Since z_t=17 is far below recombination, the inconsistency is not a small late-time correction: it changes the ionisation history, z*, the sound horizon, and the Silk damping tail over the whole range of multipoles used in the likelihood. The magnitude of the induced shift in ε_CMB is not estimated in the paper, so the central claim that percent-level contributions are allowed could be altered by an effect of unknown size.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a one-parameter extension of flat ΛCDM in which the photon energy density is rescaled by a factor ε_CMB above a transition redshift z_t (fiducial value 17), motivated by the GK25 proposal that dust-enshrouded starbursts at z∼15–20 contribute a non-negligible fraction of the observed CMB. The authors implement this modification in CAMB and constrain ε_CMB and the other six ΛCDM parameters with Planck 2018 TT/TE/EE, low-ℓ polarization, lensing, and BAO data. They report that a few-percent non-primordial contribution is fully consistent with the data: a one-sided prior run gives ε_CMB≳0.971 (0.953) at 68% (95%) credibility, a free-parameter run gives ε_CMB=1.0104^{+0.025}_{-0.024}, and the GK25 1.4% estimate lies well within the 1σ region. A fully non-primordial CMB is excluded at very high significance, while the standard ΛCDM parameters are not significantly shifted. A run with z_t free yields a flat posterior for z_t, indicating insensitivity to the transition epoch.","tokens_in":38252,"tokens_out":9539,"duration_ms":89854,"significance":"If the quantitative results survive scrutiny, the paper provides a genuinely new, achromatic probe of non-primordial CMB contributions, complementary to the FIRAS monopole constraints. In the spectrally degenerate limit (β=0, σ_d=0) the monopole alone allows a fully dust-generated CMB; the anisotropy constraint closes this loophole. The analysis is carefully executed in several respects: the baseline ΛCDM run reproduces Planck 2018 values, the MCMC convergence criterion (R−1<0.005) is stringent, and the quoted Δχ² values are small and consistent across runs. The central quantitative claims, however, rest on a recombination treatment that is not rescaled with ε_CMB, and the one-sided bounds are statistically mislabelled. Both issues are fixable but require additional work.","major_comments":[{"comment":"The recombination history is not rescaled with ε_CMB. Equation (1) reduces the photon energy density at all z>z_t, including the recombination epoch, but recfast is left with T_γ(z)=T0(1+z). If the primordial component is Planckian, its temperature is uniquely T_prim=ε_CMB^{1/4}T0(1+z); the statement in §5.2 that 'no unique effective temperature can be assigned' is incorrect. The ionisation history, z*, the sound horizon, and the Silk damping tail all depend on this temperature. The authors must either implement the rescaled recombination or quantify the resulting shift in the quoted ε_CMB bounds. Without this, the central claim that percent-level contributions are allowed is not yet established.","section":"§3.1 and §5.2"},{"comment":"The '68 (95) per cent lower bound' is not a one-sided bound. The lower edge of the central 68% credible interval (the 16th percentile) corresponds to an 84% one-sided lower limit, and the 2.5th percentile used for the 95% statement gives a 97.5% one-sided limit. To quote a 68% (95%) lower bound, the 32nd (5th) percentile should be used. Because the Run 1 posterior rises monotonically to the prior boundary, these quantiles will shift the quoted numbers, likely to lower values of ε_CMB. The qualitative conclusion may survive, but the headline numbers in the abstract should be re-computed or rephrased as 'lower edge of the X% credible interval'.","section":"§4.2 and abstract"},{"comment":"The abstract states the model is implemented 'self-consistently', while §3.1 states that the ionisation history is unchanged 'by construction'. These statements are in direct tension. Since the recombination history is part of the CMB anisotropy prediction, the self-consistency claim is not supported. The authors should either extend the model to a consistent recombination calculation or explicitly qualify the claim as conditional on the unmodified recfast treatment.","section":"Abstract and §3.1"}],"minor_comments":[{"comment":"The 42σ exclusion of ε_CMB→0 is computed using the lower error bar (1.0104/0.024≈42); using the upper error bar gives ≈40σ. Please quote the significance consistently, e.g., using the mean of the asymmetric errors.","section":"§4.3"},{"comment":"The term 'grey-body' is used for the deficit. In radiative transfer, a grey body with emissivity ε has spectrum ε B_ν(T), not a Planckian at T_eff=ε^{1/4}T. Clarify which spectral model is intended; this ambiguity is directly relevant to the recombination issue in the major comments.","section":"§2.2"},{"comment":"The summary states that standard parameters are recovered without significant shifts; this is broadly true, but ω_b h² shifts by ~1.1σ in Run 1. A qualifier such as 'with the exception of a marginal ~1σ shift in ω_b h²' would be more precise.","section":"§5.1 and Table 5"},{"comment":"The modified CAMB code is 'available on request'. For reproducibility, please provide a persistent public repository (e.g., GitHub or Zenodo) with a version tag or commit hash.","section":"§6"},{"comment":"The legend labels 'Emissivity 1' and 'Emissivity 1.5' are ambiguous. Label the curves as ΛCDM baseline, Run 1 (ε_CMB∈[0.5,1.0]), Run 2 (ε_CMB∈[0.5,1.5]), and Run 3 (z_t free) to match the text.","section":"Figure 9"},{"comment":"The Hitchhiker's Guide footnote attached to the 42σ significance is entertaining but out of place in a journal article; consider removing it.","section":"§4.3 footnote"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a topical scenario with a well-structured analysis, and the numerical work appears careful: baseline ΛCDM recovery, stringent convergence, and small Δχ² across runs. The main scientific obstruction is the inconsistent recombination calculation, which is acknowledged but not quantified. If the authors implement a consistent recombination treatment (or, at minimum, estimate the shift in the bounds) and correct the one-sided quantile statistics, the paper would be suitable for publication. I would also encourage a public code release. The broad citation of non-standard cosmology literature is appropriate given the subject; the tone of the 42σ footnote is unnecessary but not a blocking issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the first anisotropy-spectrum test of the GK25 non-primordial CMB idea, and it's worth reading even though the model has a real inconsistency in how recombination is treated.\n\nThe paper does something genuinely new. GK25 proposed that dust-enshrouded starbursts at z~15-20 add a percent-level component to the CMB energy density. Existing tests used the FIRAS monopole, which is chromatic and degenerate when the dust is spectrally identical to a blackbody. The authors instead rescale the photon energy density above z_t=17 by a factor ε_CMB and fit Planck 2018 TT/TE/EE + lensing + BAO in CAMB. This is a fair, achromatic probe of the dynamical effect. Their main results: with ε≤1 they get a 68% lower bound ε≳0.971, with ε free they get ε=1.010±0.025, and z_t is unconstrained. The GK25 1.4% estimate sits comfortably inside the 1σ region, and standard ΛCDM parameters barely shift. The pipeline validation looks solid, and the paper is unusually honest about its limitations.\n\nThe soft spot is the one you flagged: recombination is not rescaled with ε_CMB. Above z_t=17 the model has ρ_γ → ε_CMB ρ_γ, but recfast still uses T0(1+z). For a Planckian component that's not a separate choice: ρ∝T^4, so the temperature is ε^{1/4}T0(1+z). The authors wave this away in §5.2, claiming no unique effective temperature can be assigned, but that's not right. This affects the ionisation history, z*, the sound horizon and the damping tail, and the shift in ε_CMB is not quantified. It may be small — the data don't see a deviation anyway, Δχ²≈+1 — but the quoted bounds are not from a fully self-consistent model. That needs to be fixed before I'd trust the numbers.\n\nThe 42σ exclusion of ε→0 is also not a meaningful number. It's a Gaussian tail extrapolation from a posterior measured near unity; no chain samples ε~0, and the recombination inconsistency makes the prediction for that limit especially unreliable. The qualitative statement that a fully non-primordial CMB is ruled out is fine, but the specific significance is rhetorical.\n\nThe code is \"available on request,\" which isn't adequate for a modified CAMB paper.\n\nBottom line: this deserves a serious referee. The idea is new, the analysis is careful, and the central conclusion — a percent-level dust contribution is compatible with Planck — will almost certainly survive a more consistent treatment. But the paper should go back for a recombination-consistent rerun and a public code release before acceptance.","headline":"First anisotropy-based test of the GK25 non-primordial CMB scenario; the central conclusion is plausible but the quoted bounds rest on an inconsistent recombination treatment and an extrapolated 42σ claim.","tokens_in":38604,"tokens_out":5384,"would_cite":false,"duration_ms":59275,"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":"Using the CMB anisotropy power spectrum, this paper shows that a percent-level non-primordial dust contribution is consistent with current data, while a fully non-primordial CMB is excluded at 42σ.","keywords":["cosmic microwave background","CMB anisotropies","dust emission","grey-body radiation","acoustic peaks","cosmological parameters","sound horizon","photon energy density"],"falsifier":"Re-run the MCMC with the recombination temperature set to T_eff = ε^{1/4} T0(1+z) inside the recombination solver; if the best-fit ε_CMB shifts by more than roughly its quoted uncertainty, or the lower bound moves outside the quoted interval, the central bound is not robust.","tokens_in":37693,"feed_emoji":"🌌","tokens_out":6740,"duration_ms":64272,"temperature":0.7,"pith_summary":"The paper sets out to test whether the cosmic microwave background could be partly non-primordial—specifically, whether dust-enshrouded star formation in the progenitors of massive early-type galaxies at z≈15–20 could contribute a small fraction of the CMB energy we see today. It introduces a single dimensionless parameter ε_CMB that rescales the primordial photon energy density above a transition redshift, and fits it to the full CMB anisotropy, lensing, and baryon-acoustic-oscillation data. The central result is that a 1.4 per cent dust contribution is fully consistent with the data: the 68 per cent lower bound allows up to about 3 per cent of the CMB energy density to be non-primordial, while a contribution approaching 100 per cent is excluded at 42σ. The standard cosmological parameters, including H0 and σ8, are recovered unchanged, so precision cosmology is insensitive to a non-primordial component at this level.","feed_headline":"CMB anisotropy data allow a 1.4% non-primordial dust component","feed_subtitle":"Up to ~3 percent is allowed; a fully non-primordial CMB is excluded at 42 sigma.","key_machinery":"The engine of the argument is a single dimensionless emissivity parameter ε_CMB that multiplies the standard photon energy density above the transition redshift zt (fiducially z = 17), a step-function rescaling implemented directly in the Boltzmann solver's background-expansion, perturbation-evolution, and sound-horizon routines. Because the photon energy density sets the sound horizon, the photon-baryon ratio, and matter-radiation equality, the acoustic-peak structure becomes a spectral-shape-independent dynamical probe of how much radiation is non-primordial. For comparison with the broader radiation-density literature, the model maps onto an effective neutrino-number shift ΔN_eff ≈ 4.4(ε_","core_discovery":"The paper's central claim is that the CMB anisotropy power spectrum, which probes the radiation content through the sound horizon and acoustic-peak structure, is achromatically sensitive to any non-primordial energy component and therefore provides the first anisotropy-based test of the dust scenario. With ε_CMB free, the posterior is ε_CMB = 1.0104^{+0.025}_{-0.024}; with ε_CMB restricted below unity, the data give ε_CMB ≳ 0.971 at 68 per cent and ≳ 0.953 at 95 per cent, permitting up to ~3 and ~5 per cent non-primordial energy density. The paper therefore concludes that current data neither detect nor exclude the conservative 1.4 per cent estimate, while excluding a fully non-primordial CM","pith_inferences":["The authors leave implicit that the same two-parameter test can be applied to any proposal in which part of the CMB is non-primordial; the framework would give a quantitative anisotropy-based verdict for those models too.","If the paper is right, next-generation small-scale CMB measurements should tighten ε_CMB through the Silk damping tail and could eventually constrain the transition epoch that current data cannot.","The main open modeling question is the use of the unscaled recombination temperature; a fully self-consistent treatment with T_eff = ε^{1/4} T0(1+z) might shift the quoted bounds, so the published numbers should be treated as provisional until that check is done.","The assumed isotropy of the dust foreground could hide a hemispherical component; testing a spatially modulated ε_CMB would connect this scenario to the observed large-scale CMB power asymmetry."],"forward_implications":["A dust contribution at the 1.4 per cent level lies inside the 1σ region, so the conservative estimate survives the anisotropy data.","A non-primordial contribution approaching the full CMB energy density is excluded at 42σ.","The standard six-parameter cosmological model, including H0 and σ8, is recovered without significant shifts, so an unseen few-per-cent foreground of this kind would not bias parameter inference.","The constraint applies equally for any transition epoch between z = 5 and z = 50, meaning current data bound the amplitude of a non-primordial component but not when it was produced.","Combined with the CMB monopole spectral bound, the scenario is bracketed: the monopole is tighter for chromatic dust, while the anisotropy spectrum closes the spectrally degenerate loophole."],"fun_headline_variants":["CMB anisotropy permits up to 3% non-primordial dust","Dust could comprise 3% of CMB, anisotropy data show","CMB anisotropy data allow 3% dust contribution, exclude full non-primordial","Sound horizon constrains dust: up to 3% allowed in CMB","Non-primordial CMB dust capped at 3% by anisotropy peaks"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The analysis rescaled the photon energy density in the expansion and perturbation equations but left the recombination calculation on the unmodified photon temperature, so the acoustic-peak and damping-tail predictions are not fully self-consistent for ε_CMB ≠ 1.","fun_headline_variants_meta":{"raw":{"variants":["CMB anisotropy permits up to 3% non-primordial dust","Dust could comprise 3% of CMB, anisotropy data show","CMB anisotropy data allow 3% dust contribution, exclude full non-primordial","Sound horizon constrains dust: up to 3% allowed in CMB","Non-primordial CMB dust capped at 3% by anisotropy peaks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000799,"raw_usage":{"total_tokens":3458,"prompt_tokens":958,"completion_tokens":2500,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":702,"completion_tokens_details":{"reasoning_tokens":2399}},"tokens_in":702,"tokens_out":2500,"duration_ms":16183,"temperature":1.0,"reasoning_tokens":2399,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T02:45:59.229078+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the MCMC with the recombination temperature set to T_eff = ε^{1/4} T0(1+z) inside the recombination solver; if the best-fit ε_CMB shifts by more than roughly its quoted uncertainty, or the lower bound moves outside the quoted interval, the central bound is not robust.","supporting_citations":[],"review_version":1}