{"id":"cfa861c1-82f9-4d0e-8b56-aceab7353d6e","arxiv_id":"2412.15192","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"AxiECAMB computes CMB and matter power spectra for ultralight axions across 10^-33 to 10^-18 eV by time-averaging fast field oscillations, with metric and fluid matching that reduces errors far below cosmic variance for observationally viable models.","lead":"Working cosmologists built a fast, accurate computer code, AxiECAMB, that predicts how ultralight axions would imprint on the cosmic microwave background and on galaxy clustering over a huge mass range. The code fixes numerical artifacts in older axion codes, so upcoming surveys like CMB-S4, DESI, and Rubin can be used to test whether such axions exist.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Survey-ready accuracy in the 10^-27.5 to 10^-25 eV window rests on switch-phase and metric-partition rules validated only on representative cases; a dense (m, fDM) grid check is needed before the default settings are certified as a uniform accuracy guarantee.","rationale":"The CONDITIONAL verdict is appropriate. The paper has real strengths: public code, detailed derivations, consistency checks of ETA against KG in Figs. 1, 2, 23, 24, and 27, a clear O(H/m)^2 scaling argument, and explicit discussion of limitations. My concern is not that the method is wrong—the internal convergence evidence is strong—but that the accuracy guarantee for the most sensitive mass window is a managed, empirical property rather than a controlled approximation. The empirical constant in W and the phase-tuned switch are the weakest links because they are calibrated on a limited set of modes and models and then assumed to cover the full parameter space. The high-accuracy benchmark being the same code with a later switch is a further reason the absolute error in the problematic window is not independently established, even though the (H*/m)^2 scaling makes residual errors at m/H* = 100 likely small. A dense grid check against a higher-switch reference would either clear the concern or force the authors to weaken the 'approaches cosmic variance' claim. Since the paper itself acknowledges the need for technical refinements and external checks, the conditional verdict stands unchanged.","tokens_in":34839,"tokens_out":8456,"duration_ms":85096,"concrete_test":"Run AxiECAMB with the published default settings over a dense grid in the window 10^-27.5 < m/eV < 10^-25 (at least 10 masses per decade) and fDM = {0.05, 0.1, 0.2, 0.5, 1}, comparing TT/TE/EE against the high-accuracy reference recomputed at both m/H* = 100 and m/H* = 200. If any grid point has Δχ2 (Eq. 36, ℓmax = 2400) above the ΛCDM floor of ≈ 0.47 and cannot be brought below by the θ* shift correction (Eq. 41), then the default settings are not uniformly survey-ready and the switch-selection rules need to be made adaptive. Also record the best switch phase at each point; if the global 2β ≈ 7.08π rule is not near-optimal at most points, the phase tuning is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central accuracy claim is not a uniform bound. In the mass window 10^-27.5 ≲ m/eV ≲ 10^-25 with fDM near unity, the default accuracy is achieved by three empirical interventions: raising m/H* above the baseline 10, moving the switch below z ≈ 800, and setting the switch phase 2β ≈ 7.08π (Sec. II D, App. E, Fig. 27). The phase rule is extracted from a single mode (k = 0.1 Mpc^-1) of a single model (m = 10^-26 eV, fDM = 1), and the metric partition W in Eq. (16) contains the constant 3 chosen to 'roughly optimize' the CMB error. The validation in App. C covers m = 10^-26...10^-29 eV and fDM ≳ 0.1, but the Cℓ accuracy maps in Figs. 13-14 do not prove that the same fixed rules keep Δχ2 < 1 for all combinations of m, fDM, and background parameters. Because the high-accuracy benchmark is the same code with a later switch, it inherits the same ETA and W model; it validates the (H*/m)^2 scaling but cannot independently certify the empirical choices. If the phase rule or the W partition is off for some untested region, the advertised 'approaches cosmic variance' accuracy is not delivered.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the effective time-average (ETA) / effective fluid approximation (EFA) method of Passaglia & Hu (2022) to ultralight axion masses down to 10^-33 eV, where the switch from Klein-Gordon evolution to a fluid description occurs near or after matter-radiation equality. The authors derive an ETA for the metric perturbations with an empirical partitioning weight, match the EFA pressure continuously to the ETA at the switch, add boundary (switch-source) terms to the CMB line-of-sight integrals, and implement the method in a public Boltzmann code, AxiECAMB. They validate the method against a high-accuracy run that solves the KG system to m/H = 100, quantify errors with a cosmic-variance-limited Δχ² statistic at ℓmax = 2400, and compare with axionCAMB across the mass-fraction plane. The main accuracy results are presented for total dark-matter (fDM = 1), fractional dark-matter (fDM = 0.1), and total dark-energy (fDE = 1) cases.","tokens_in":34970,"tokens_out":6709,"duration_ms":59279,"significance":"If the accuracy claim holds, this is a substantial advance: AxiECAMB would be the first ULA Boltzmann code delivering survey-ready CMB and matter power spectra over the full mass range 10^-33–10^-18 eV, with errors below the cosmic-variance limit for observationally viable models and with order-unity corrections to axionCAMB predictions near ΛCDM. The paper includes a public code, careful internal convergence tests showing (H*/m)^2 scaling, a new iteration scheme for the ETA background, and a thorough comparison with previous codes, including identification of axionCAMB failure modes. These strengths make the paper valuable for the community regardless of the caveats below.","major_comments":[{"comment":"The switch-phase rule 2β ≈ 7.08π is calibrated on a single mode (k = 0.1 Mpc^-1) of a single model (m = 10^-26 eV, fDM = 1) and then applied to the entire range 10^-27.5 ≲ m/eV ≲ 10^-25 via Eq. (B16). Since the photon-baryon response to unresolved KG oscillations is explicitly k-dependent (as noted in Sec. II D and Appendix E), the optimal phase may depend on k and on the background parameters in the window. The paper asserts that the tuning suffices but does not show a dense grid of (m, fDM) accuracy tests in this window, particularly for the observationally relevant fDM = 0.1 case. This is load-bearing for the claim that the default settings provide uniform, survey-ready accuracy; either a denser validation or a qualifying statement limiting the guarantee to tested regions is needed.","section":"Sec. II D, Appendix E, Fig. 27"},{"comment":"The metric-partition weight W contains a constant 3 that is 'chosen to roughly optimize' the CMB error. Figure 24 validates the resulting amplitude ratio ΔηKG/ΔσKG against exact KG oscillations for a set of masses and fDM ≳ 0.1, which is encouraging. However, the paper does not demonstrate that the final Cℓ accuracy is insensitive to this constant across the parameter space, nor does it provide a first-principles derivation of the weight. If the optimal constant changes with m or fDM, the default choice could produce larger errors in untested regions. A sensitivity test varying the constant (e.g., 1–5) and showing Δχ² is flat, or a derivation of the partition from the KG equations, would remove this concern.","section":"Eq. (16), Sec. II B, Appendix C"},{"comment":"The paper's central claim in Sec. IV that 'the default accuracy approaches the cosmic variance limit at ℓmax = 2400' is supported by Table I for fDM = 0.1 at m = 10^-28–10^-31 eV and for fDE = 1 cases, but the high-Δχ² region at m = 10^-27 eV, fDM = 1 (Δχ² = 53) is outside the observationally viable regime. The accuracy maps in Figs. 13 and 14 do not resolve, for the phase-tuned window 10^-27.5 ≲ m/eV ≲ 10^-25, whether the default settings keep Δχ² < 1 for all fDM that pass current constraints (roughly fDM ≲ 0.05 at these masses according to the cited axionCAMB limits). The statement 'default accuracy approaches the cosmic variance limit' should be qualified to distinguish the observationally viable region from the full allowed-prior space, or additional grid tests should be provided.","section":"Sec. III, Table I, Figs. 13 and 14"}],"minor_comments":[{"comment":"The first sentence states the motivated mass range as 10^-33 ≲ m/eV ≲ 10^-12, while the fourth sentence says the method covers up to 10^-18 eV; consider clarifying that the paper's range is a subset of the motivated range to avoid confusion.","section":"Abstract"},{"comment":"There is a duplicated word: 'The cancellation of the KG oscillations before the switch also requires requires very dense' should read 'also requires very dense'.","section":"Appendix E (after Eq. E1)"},{"comment":"The caption reads 'including (including) the switch sources'; the second parenthesis should likely be '(solid)' to match the dashed/solid distinction in the text.","section":"Fig. 29 caption"},{"comment":"The phrase 'in the forseeable future' contains a typo ('foreseeable').","section":"Sec. IV"},{"comment":"The linearized θ⋆ correction is presented as a diagnostic. It would be helpful to state explicitly that this correction is not applied in the default pipeline and is only used to identify the source of the error, to avoid a reader misunderstanding that the reported default Δχ² already includes it.","section":"Eq. (41) and Sec. III A"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically strong and the core ETA construction with metric partitioning is well motivated and tested against KG oscillations. My main hesitation is the breadth of the accuracy warranty: the phase-tuning and weight W are empirical, and the high-accuracy benchmark inherits the same ETA model, so it cannot independently certify these choices. I would be satisfied with either (a) a denser (m, fDM) grid in the phase-tuned window, including fDM = 0.1 at m ~ 10^-26–10^-25 eV, or (b) a revised abstract/introduction that claims 'survey-ready accuracy for observationally viable fractions' rather than a uniform default guarantee across the full mass–fraction plane. The paper is not far from acceptable; the issue is scope of the central claim, not the method's validity."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper will get cited. It does something genuinely new: extends the Passaglia-Hu effective time-averaging (ETA) program to ULAs whose switch epoch sits near or after matter-radiation equality, which requires time-averaging the metric perturbations and matching to an effective fluid with continuous pressure so that hydrostatic equilibrium is preserved. The CMB switch-source boundary terms and the phase-tuned switch placement are real technical additions, and the internal validation is convincing: residuals scale as (H*/m)^2, and the errors for fDM ~ 0.1 are below cosmic variance at lmax=2400. The code is public, and the appendices are unusually careful about where the method might break.\n\nThe soft spots are the usual ones for numerical methods papers, and the stress-test note is fair. The accuracy claim is not a uniform bound. In the window 10^-27.5 ≲ m/eV ≲ 10^-25 with fDM = 1, the default settings produce Delta chi^2 up to 53, which only comes down after engineering: raising m/H*, moving the switch away from recombination, and setting the phase 2 beta ≈ 7.08 pi. That phase rule is extracted from one mode (k = 0.1 Mpc^-1) of one model (m = 10^-26 eV, fDM = 1). The metric partition W in Eq. (16) has a constant 3 that is chosen to 'roughly optimize' CMB error. And the high-accuracy benchmark is the same code with a later switch, so it validates the (H*/m)^2 scaling but cannot independently certify the empirical choices. A dense (m, fDM) grid check would close the gap.\n\nNone of this kills the paper. The errors in the observationally viable fDM ~ 0.1 region are small, and the paper is honest about the fDM = 1 caveats (Table I is a model of transparency). But the advertised 'approaches the cosmic variance limit' should be understood as a managed property, not a theorem. For peer review, I would ask for: a versioned release with a commit hash, an independent cross-check of the m/H* = 100 benchmark (e.g., a direct KG integration to much later times or a comparison with a different code), and a sensitivity scan showing how C_l and Delta chi^2 vary with W and the switch phase on a grid of m and fDM.\n\nBottom line: send it out. The right referee will ask for these validations, and the core method looks sound. If the authors can provide the grid check, this becomes the reference tool for translating Stage IV data into ULA constraints.","headline":"Solid methods paper that will become the reference ULA Boltzmann code if the empirical knobs survive a denser validation grid.","tokens_in":35698,"tokens_out":5717,"would_cite":true,"duration_ms":34534,"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 presents AxiECAMB, a Boltzmann code that computes CMB and matter power spectra for ultralight axions from 10^-33 to 10^-18 eV with errors at the cosmic variance limit.","keywords":["ultralight axions","Boltzmann code","effective fluid approximation","effective time averaging","cosmic microwave background","matter power spectrum","cosmic variance","fuzzy dark matter"],"falsifier":"Run the default AxiECAMB settings against the high-accuracy Klein-Gordon solution (switch at $m/H_*=100$) on a dense grid spanning $10^{-27.5} < m/{\\rm eV} < 10^{-25}$ at $f_{\\rm DM}=1$, including masses and switch phases not covered by the paper's tuning rules, and check whether the CMB $\\Delta\\chi^2$ at $\\ell_{\\rm max}=2400$ exceeds unity at any grid point; one such point would falsify the blanket cosmic-variance accuracy claim.","tokens_in":34418,"feed_emoji":"🌌","tokens_out":10493,"duration_ms":75742,"temperature":0.7,"pith_summary":"The paper presents AxiECAMB, the first Boltzmann code that computes cosmic microwave background and matter power spectra for ultralight axions (ULAs) accurately and efficiently across the mass range $10^{-33}$ to $10^{-18}$ eV. Until now, accurate effective treatments existed only for masses much heavier than the Hubble rate at matter-radiation equality, and previous codes made errors that could bias parameter constraints from current and next-generation surveys. The code solves the exact Klein-Gordon system until oscillations become rapid, then matches to a time-averaged effective fluid through a new effective time average that also covers the metric perturbations. If the central claim is right, AxiECAMB's default settings are accurate to the cosmic variance limit at $\\ell_{\\rm max}=2400$, comparable to standard $\\Lambda$CDM calculations, and correct order-unity errors in the ULA-induced change from $\\Lambda$CDM in some parameter regions.","feed_headline":"First ULA Boltzmann code reaches cosmic-variance accuracy","feed_subtitle":"AxiECAMB extends effective time-averaging to the full axion mass range, fixing order-unity errors in prior codes.","key_machinery":"The machinery is a two-stage approximation: solve the Klein-Gordon system until the switch epoch $m/H_* \\sim 10$, then continue with an effective fluid approximation (EFA). The bridge is the effective time average (ETA), which separates the oscillating field into cosine and sine auxiliary components whose cycle averages obey energy-momentum conservation, and which is extended to the metric perturbations $\\eta$ and $\\sigma$ with an empirical weight $W=(k/aH)^2/(3+(k/aH)^2)$ that removes the leading-order oscillations in whichever variable carries them. The ETA sets switch conditions for the EFA, including a pressure-matched equation-of-state parameter $A_w=(p/\\rho)(m/H_*)^2$, and the CMB line-of-sight integral is augmented with boundary switch sources from the integration by parts. This combination removes $O(H/m)$ oscillations, keeps the fluid in hydrostatic equilibrium below the Jeans scale, and makes the default accuracy set by $m/H_*$ a controllable parameter.","core_discovery":"The central discovery is a complete prescription for switching from the Klein-Gordon equations of an ultralight axion to an effective fluid once $m/H \\ge 10$, with errors of order $(H/m)^2$ in all matter and metric variables. The innovation is to construct the effective time average (ETA) of the axion and, crucially, of the metric perturbations $\\eta$ and $\\sigma$, using auxiliary cosine and sine field components and an empirical scale-dependent weighting, so that the leading oscillations are removed and the average is insensitive to the phase at which the switch catches the field. The effective fluid's equation of state is then matched to the ETA pressure, preserving hydrostatic equilibrium below the Jeans scale and avoiding spurious pressure-wave oscillations. Boundary terms in the CMB source function that arise from the switch are included explicitly, and the switch time and phase are chosen to avoid recombination and to minimize residual photon-baryon responses. The paper validates the method against a high-accuracy calculation that solves the Klein-Gordon system to $m/H=100$ and reports default-setting errors at the cosmic variance limit, whereas the previous axionCAMB code can be off by up to five orders of magnitude in $\\Delta\\chi^2$ in extreme regions.","pith_inferences":["A testable implication is that the case-by-case switch-time and phase tuning, needed for $10^{-27.5} \\lesssim m/{\\rm eV} \\lesssim 10^{-25}$ at $f_{\\rm DM}=1$, may not extend to a uniform accuracy guarantee; a grid scan of untested parameter space would show where the advertised cosmic-variance accuracy breaks down.","The same ETA/EFA bridge could be carried to next order in $H/m$, which would likely remove the need for phase tuning and make the accuracy claim robust without case-by-case engineering.","Because the paper fixes $\\Omega_{\\rm ax}h^2$ rather than the initial field value in one part of its convention, part of the residual $\\theta_*$ error is an artifact of that convention; an alternative normalization could reduce the reported $\\Delta\\chi^2$ even further.","Applying the method to isocurvature perturbations, which the paper leaves for future work, would provide a sharper test of whether the ETA partitioning of $\\eta$ and $\\sigma$ remains valid when the axion field carries a non-adiabatic component."],"forward_implications":["AxiECAMB attains sub-percent accuracy in CMB and matter power spectra across the $10^{-33}$ to $10^{-18}$ eV mass range, with runtimes comparable to CAMB in $\\Lambda$CDM.","Default accuracy approaches the cosmic variance limit at $\\ell_{\\rm max}=2400$; the dominant error is a tiny shift in the angular sound horizon, at most $\\Delta\\theta_*/\\theta_* \\sim 9\\times 10^{-4}$ even in observationally ruled-out extreme models.","Existing ULA constraints based on axionCAMB should be re-examined, because axionCAMB's errors can be a significant fraction of the ULA-induced change from $\\Lambda$CDM and can reach $\\Delta\\chi^2 \\sim 10^7$ in some regions.","Including the switch boundary terms is necessary: omitting them produces phase-dependent errors above cosmic variance at low multipoles.","The same effective method covers dark-energy-like ULAs down to $10^{-33}$ eV and fuzzy dark matter up to $10^{-18}$ eV, with fitting functions remaining the efficient choice at still higher masses."],"supporting_citations":[{"why":"Supplies the original switch-and-effective-time-average method that this paper extends to low masses, metric perturbations, and dark-energy-like ULAs.","marker":"[45]"},{"why":"axionCAMB is the predecessor code whose numerical errors and parameter-space failures this paper quantifies and fixes.","marker":"[42]"},{"why":"Provides the generalized dark matter fluid approximation underlying axionCAMB, which the EFA replaces after the switch.","marker":"[30]"},{"why":"Documents numerical artefacts in the effective fluid approximation that motivate the need for a more accurate ULA Boltzmann treatment.","marker":"[44]"},{"why":"Supplies the Planck fiducial cosmological parameters used for all accuracy tests and benchmarks.","marker":"[52]"},{"why":"Provides the DES S8 measurement used as the large-scale-structure accuracy benchmark for Delta S8.","marker":"[7]"},{"why":"Establishes that fDM=1 is ruled out for light ULAs, framing the maximal-stress test cases for code accuracy.","marker":"[61]"}],"fun_headline_variants":["AxiECAMB: cosmic-variance accuracy for ultralight axions","Axion power spectra fixed: new code hits cosmic variance limit","New axion code beats old by 5 orders in extreme cases","Effective fluid treatment nails axion CMB spectra","Ultralight axion spectra: from Klein-Gordon to fluid with precision"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The accuracy guarantee rests on the assumption that truncating the photon-baryon response to unresolved axion oscillations at leading order after the switch, a residual of order $(8\\pi G\\rho_{\\rm ax}/H^2)(H_*/m)^2$, stays below the cosmic-variance target for every parameter combination, which in the window $10^{-27.5} \\lesssim m/{\\rm eV} \\lesssim 10^{-25}$ with $f_{\\rm DM}=1$ is achieved only through case-by-case switch-time, switch-phase, and recombination-avoidance rules.","fun_headline_variants_meta":{"raw":{"variants":["AxiECAMB: cosmic-variance accuracy for ultralight axions","Axion power spectra fixed: new code hits cosmic variance limit","New axion code beats old by 5 orders in extreme cases","Effective fluid treatment nails axion CMB spectra","Ultralight axion spectra: from Klein-Gordon to fluid with precision"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0008,"raw_usage":{"total_tokens":3578,"prompt_tokens":1066,"completion_tokens":2512,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":682,"completion_tokens_details":{"reasoning_tokens":2422}},"tokens_in":682,"tokens_out":2512,"duration_ms":14657,"temperature":1.0,"reasoning_tokens":2422,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:33:19.476432+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the default AxiECAMB settings against the high-accuracy Klein-Gordon solution (switch at $m/H_*=100$) on a dense grid spanning $10^{-27.5} < m/{\\rm eV} < 10^{-25}$ at $f_{\\rm DM}=1$, including masses and switch phases not covered by the paper's tuning rules, and check whether the CMB $\\Delta\\chi^2$ at $\\ell_{\\rm max}=2400$ exceeds unity at any grid point; one such point would falsify the blanket cosmic-variance accuracy claim.","supporting_citations":[{"cited_title":"Ratra, Phys","cited_arxiv_id":null,"evidence_quote":"Supplies the original switch-and-effective-time-average method that this paper extends to low masses, metric perturbations, and dark-energy-like ULAs."}],"review_version":1}