{"id":"8c33e43c-2e95-4835-b81a-139010818b55","arxiv_id":"2505.11592","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A template-based analysis of mock LiteBIRD CMB data suggests the satellite could constrain the photon-ALP coupling to below 6.5e-12 GeV^-1 for ALP masses below 1e-14 eV, about ten times stronger than the current CAST limit.","lead":"This paper forecasts that the planned LiteBIRD CMB satellite could detect the subtle signature of axion-like particles converting into CMB photons in the Milky Way's magnetic fields, reaching couplings an order of magnitude below current lab limits. It introduces a template-matching technique to capture the non-Gaussian spatial shape of this signal across LiteBIRD's frequency bands.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Null calibration missing: the zero-signal posteriors in Fig. 8 peak away from g=0, so the quoted 95% limits are not yet calibrated and the headline constraint is not established as stated.","rationale":"The central claim is a forecast: LiteBIRD's multi-frequency ILC-cleaned maps, analyzed with the template likelihood, will bound g<6.5e-12. For this to be true, the likelihood must be calibrated: under g=0 the posterior should be centered at zero and the 95% upper limit should have correct frequentist coverage. The paper's own Fig. 8 violates this: zero-signal posteriors peak away from zero. The text attributes the shift to ILC residuals and template misfit but does not fold a systematic into the quoted interval. The patch analysis in Sec. 4.3 is a diagnostic, not a correction. This is the most load-bearing soft spot because it is internal and directly affects the headline number; the Galactic-model dependence identified by the Reader is real but is a standard forecast assumption, and the paper discusses it honestly. The proposed null-calibration run would settle whether the shift is within statistical fluctuation or a real bias. A successful calibration would leave the forecast intact; failure would require adding a systematic term, which could weaken the bound. This is a fixable issue, so I do not recommend rejection; I keep the conditional verdict.","tokens_in":22380,"tokens_out":8900,"duration_ms":100149,"concrete_test":"Null calibration check: generate at least 100 independent LiteBIRD mock skies with gaγ=0 (independent CMB, foreground, and noise realizations), run the exact ILC + template + MCMC pipeline used for Fig. 8, and record the posterior median and 95% upper limit in each. If the medians are not centered on zero within Monte Carlo error, or if more than ~5% of the null realizations exclude g=0 at 95%, add a nuisance parameter for the ILC-residual amplitude (or subtract the empirical mean null residual and inflate σ_R accordingly), marginalize, and recompute. The headline survives only if the calibrated median 95% bound remains at or below 6.5×10^-12 GeV^-1 for m=10^-14 eV.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The forecast rests on the likelihood in Eq. (4.6), logL = -[(Σ_i S_i^2) - (g/g*)^4 Σ_i M_i^2 - ⟨R⟩]^2/(2σ_R^2), where ⟨R⟩ and σ_R are estimated from 500 null realizations. If the ILC-cleaned mock data contain foreground residuals that are not statistically identical to those realizations, the statistic (ΣS^2 - ⟨R⟩) has a non-zero mean and the posterior is shifted away from the injected value. Fig. 8 shows exactly this: for the fiducial case g=0, the posterior peaks are visibly displaced from zero for all three masses. The paper attributes this to non-Gaussian foreground residuals and template misfit, and states the shift 'can be accounted for' by better cleaning or template matching, but no such correction is applied before quoting g<6.5e-12. The patch-based systematic study in Sec. 4.3 diagnoses direction-dependent variation (up to a factor 2-3) but does not add a global systematic term to the headline interval. Since the best-fit templates also have reduced chi-squared ~1.6, residual template mismatch is non-negligible. Therefore the reported 95% confidence intervals are not calibrated under the null; the central claim is conditional on an unapplied correction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a forecast for constraining the photon–ALP coupling gaγ using resonant photon–ALP conversion of CMB photons in the Milky Way, as would be observed by LiteBIRD. The authors simulate ALP polarized-intensity maps for masses 5×10^-14, 10^-14, and 5×10^-15 eV using the Jansson–Farrar Galactic magnetic-field model and the NE2001 electron-density model, inject them into mock LiteBIRD observations with CMB, PySM foregrounds, and instrument noise, clean the maps with ILC, and fit a family of latitude-based spatial templates. A Gaussian likelihood on the sum of squared ILC-cleaned polarized intensities is then used to infer gaγ. The headline result is a projected 95% upper limit gaγ < 6.5×10^-12 GeV^-1 for ma below 10^-14 eV, about an order of magnitude stronger than the CAST bound, with the strongest individual-mass bound gaγ < 4.5×10^-12 GeV^-1 for ma=5×10^-15 eV.","tokens_in":22610,"tokens_out":7607,"duration_ms":79882,"significance":"The paper addresses an interesting and timely question: whether a full-sky CMB polarization mission can probe ALP parameter space beyond current laboratory bounds using a non-Gaussian Galactic signal. The conversion-probability formalism is standard, the mock-observation pipeline is end-to-end, and the recovery of a non-zero injected coupling in Fig. 9 is a useful sanity check; the scaling of the template by (gaγ/g*)^4 follows from the conversion physics rather than being a fitted calibration. The patch-based systematics diagnostic in Sec. 4.3 is also a constructive idea. However, the central forecast is not yet calibrated: the null-hypothesis posteriors in Fig. 8 peak away from zero, the likelihood in Eq. (4.6) omits signal–noise cross terms, and the template fits have reduced chi-squared values between 1.55 and 1.66. Until these effects are corrected or incorporated as systematic terms, the quoted upper limits are conditional on the template and foreground models and should not be read as a validated LiteBIRD sensitivity projection.","major_comments":[{"comment":"The fiducial (gaγ=0) posterior peaks are visibly displaced from zero for all three masses, and the figure caption itself attributes this to ILC residual foregrounds and template mismatch. The headline upper limits are read from these same uncalibrated posteriors, and the statement that the shifts can be accounted for by template matching or better cleaning is not accompanied by any such correction in the quoted intervals. The patch analysis in Sec. 4.3 diagnoses direction-dependent variation but does not add a global systematic term to the full-sky posterior. The reported 95% confidence intervals are therefore not calibrated under the null hypothesis, and the central claim gaγ < 6.5×10^-12 GeV^-1 is not established as stated.","section":"Sec. 4.2, Eq. (4.6), Fig. 8"},{"comment":"The likelihood implicitly assumes E[Σ_i S_i^2] = ⟨R⟩ + (gaγ/g*)^4 Σ_i M_i^2. If D_i denotes the ILC-cleaned null map, a data map with an injected signal satisfies S_i ≈ D_i + (gaγ/g*)^2 M_i up to template mismatch, so the statistic has expectation 2(gaγ/g*)^2 Σ_i ⟨D_i M_i⟩ + (gaγ/g*)^4 Σ_i M_i^2 + [Σ_i D_i^2 − ⟨R⟩], not the assumed form. The cross term 2(gaγ/g*)^2 Σ_i ⟨D_i M_i⟩ is never estimated or demonstrated to be negligible. Since the residual foregrounds are non-Gaussian and spatially correlated with the Galactic ALP template, this term provides a concrete mechanism for the posterior shift seen in Fig. 8 and means the likelihood calibration is incomplete.","section":"Sec. 4.2, Eqs. (4.6)-(4.7)"},{"comment":"The best-fit reduced chi-squared values 1.55, 1.60, and 1.66 show that the template family does not fully describe the simulated ALP maps, and Sec. 4.3 reports patch-to-patch variation of a factor of 2–3 in the inferred coupling. Neither effect is propagated into the quoted 95% upper limits: the likelihood in Eq. (4.6) has no term for template mismatch, and the patch analysis is used only as a diagnostic. A defensible forecast should either incorporate a template-mismatch or systematic term into the full-sky posterior or explicitly frame the bounds as conditional on the template family and foreground model rather than as projected LiteBIRD constraints.","section":"Sec. 4.1 and Sec. 4.3"}],"minor_comments":[{"comment":"The final sentence has a punctuation and capitalization error: 'CAST at gaγ < 6.6×10^-11 GeV^-1, This shows' should be corrected.","section":"Abstract"},{"comment":"'Interior Linear Combination' should be 'Internal Linear Combination' to match standard usage.","section":"Sec. 4"},{"comment":"The superscript α in ΔI^α_T, ΔI^α_Q, and ΔI^α_U is not defined; please clarify what α labels.","section":"Sec. 2.3, Eqs. (2.15)-(2.17)"},{"comment":"The non-zero coupling case is described only qualitatively; a numerical posterior interval or a table of the inferred gaγ values for each mass would greatly improve reproducibility.","section":"Sec. 4.2, Fig. 9"},{"comment":"The 95% C.I. notation is used for posterior-based intervals; these should be labeled as credible intervals or the construction should be clarified.","section":"Sec. 4.2"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on Mehta and Mukherjee (arXiv:2505.11592). The genuinely new piece is a template-based, power-spectrum-free search for the Galactic ALP distortion signal in simulated LiteBIRD data, plus a new forecast number: g<6.5e-12 GeV^-1 for ma<1e-14 eV, about an order of magnitude stronger than CAST. The method is described carefully: mock multi-frequency maps with PySM foregrounds, ILC cleaning, spatial template fits, and a Gaussian likelihood on summed squared intensity. The patch-based systematic study is a good addition. This is honest, careful work, and it clearly engages with the literature.\n\nThe main soft spot is exactly the one your stress-test note flags. The zero-coupling posteriors in Fig. 8 peak away from zero for all three masses. The paper says this shift can be accounted for by better foreground cleaning or template matching, but no correction is applied before quoting the 95% limits. So the headline constraint is not calibrated under the null; it is a forecast conditional on an identified but unimplemented systematic correction. The reduced chi-squared values around 1.6 for the best-fit templates reinforce that the template is incomplete. A reader should not treat g<6.5e-12 as an established bound.\n\nThe other soft spots are the ones the paper itself acknowledges: the signal model is tied to the Jansson-Farrar magnetic field and NE2001 electron density, and if those are biased the bound shifts. The template method deliberately discards small-scale information, which is a fair tradeoff but lowers sensitivity. There is no code or data release, which makes reproducibility harder.\n\nWho is this for? People forecasting ALP constraints from CMB experiments and people working on non-Gaussian Galactic signal searches. It deserves a serious referee. A revised version that calibrates the null, by subtracting the shift or adding a systematic term to the likelihood, and quotes limits with that systematic included would be much stronger. If the authors also release the pipeline, this becomes a useful community tool.","headline":"A genuinely new template-based forecast for Galactic ALP signals with LiteBIRD, but the quoted 95% limits are not calibrated under the null because the zero-signal posteriors are visibly shifted from zero.","tokens_in":23198,"tokens_out":3352,"would_cite":true,"duration_ms":34851,"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":"A template-based analysis of Milky Way CMB distortions could constrain photon–axion couplings ten times better than CAST for masses below $10^{-14}$ eV.","keywords":["axion-like particles","photon-ALP conversion","cosmic microwave background","LiteBIRD","Milky Way magnetic field","template matching","CMB spectral distortions","beyond Standard Model"],"falsifier":"Regenerate the mock maps and run the same pipeline with an independent, data-driven reconstruction of the Galactic magnetic field and electron density, for example from pulsar Faraday-rotation and dispersion measures or from a different published field model. If the 95% upper limit on $g_{a\\gamma}$ changes by more than the reported statistical error, the constraint is dominated by model choice rather than by data, and the forecast should not be read as a reliable LiteBIRD reach.","tokens_in":22130,"feed_emoji":"🛰️","tokens_out":9317,"duration_ms":88815,"temperature":0.7,"pith_summary":"The paper argues that the upcoming CMB satellite LiteBIRD can detect or constrain axion-like particles by watching CMB photons convert into ALPs inside the Milky Way's magnetized gas. Because conversion happens only where the ALP mass matches the local plasma frequency, the distortion has a specific spatial and spectral signature that is strongly non-Gaussian. The authors replace the usual power-spectrum analysis with a fast template search on multi-frequency maps, and on mock LiteBIRD observations they recover both a null signal and a nonzero injected coupling. For the lightest masses considered, the forecast reaches $g_{a\\gamma}\\lesssim 4.5\\times10^{-12}$ GeV$^{-1}$ at 95% confidence, roughly an order of magnitude beyond the current CAST laboratory bound. If realized, that would open a new region of ALP parameter space to a dedicated CMB mission.","feed_headline":"Template search lets LiteBIRD beat CAST axion bound tenfold","feed_subtitle":"CMB photons converting in the Galaxy's magnetic field could reveal axion-like particles below 10^-14 eV.","key_machinery":"The central object is the photon-ALP resonant conversion probability: at each resonance the conversion probability is $P_i \\approx \\pi \\gamma_{ad,i}/2$, with adiabaticity $\\gamma_{ad} = 2g_{a\\gamma}^2 B_t^2 \\omega(1+z)/|\\nabla\\omega_p^2|$, where $B_t$ is the transverse magnetic field and $\\omega_p$ is the plasma frequency. The method then compresses the 3-D signal into 2-D template maps of the form $\\Delta I(\\theta^*,\\phi^*) = A(\\theta)+B(\\theta)\\cos\\theta\\sin\\phi$, built from latitude-based radial functions $r=f(d,\\theta)$ and coordinate-transformed to the solar system frame. These templates are cleaned with an internal linear combination of LiteBIRD's frequency channels, the best template is selected by reduced $\\chi^2$, and the coupling is inferred by an MCMC Gaussian likelihood built from the distribution of no-ALP map realizations.","core_discovery":"The paper's central claim is that a spatial template, built from latitude-symmetric analytic functions plus an azimuthal asymmetry term, captures the large-scale sky shape of the Galactic photon-ALP conversion signal well enough that a joint temperature-polarization likelihood on multi-frequency LiteBIRD maps can recover the coupling. In mock skies with no ALPs, the 95% upper limits are $g_{a\\gamma}<4.5\\times10^{-12}$ GeV$^{-1}$ at $m_a=5\\times10^{-15}$ eV, $g_{a\\gamma}<6.5\\times10^{-12}$ GeV$^{-1}$ at $m_a=10^{-14}$ eV, and $g_{a\\gamma}<1.9\\times10^{-11}$ GeV$^{-1}$ at $m_a=5\\times10^{-14}$ eV, all stronger than the CAST limit of $g_{a\\gamma}<6.6\\times10^{-11}$ GeV$^{-1}$. When a nonzero coupling $g_{a\\gamma}=10^{-11}$ GeV$^{-1}$ is injected, the lighter-mass cases constrain it both from above and below, and a patch-by-patch analysis finds the all-sky inference consistent with the injected value while exposing which sky regions suffer the largest systematics. The authors flag in Sec. 4.3 that these inferences are limited by the accuracy of modeled galactic foregrounds, electron density, and magnetic field.","pith_inferences":["Editorial inference: the quoted bounds are conditional on the adopted Milky Way magnetic-field and electron-density models; a model error common to all sky patches would survive the patch-by-patch consistency check and bias the coupling silently.","Editorial inference: because the templates fit polarized intensity only, the linear-polarization angle (Q/U phase) is unused; a template that also matches the polarization orientation could separate true conversion from foreground leakage and sharpen the limits.","Editorial inference: the same pipeline is a generic search tool for non-Gaussian, frequency-correlated distortions; adapting it to dark-photon conversion or other exotic spectral distortions would require only changing the spectral dependence and template shape."],"forward_implications":["If the template forecast holds, LiteBIRD will probe $g_{a\\gamma}\\lesssim10^{-12}$ GeV$^{-1}$ for ALP masses around $10^{-15}$ eV, entering parameter space inaccessible to laboratory helioscopes.","A nonzero detection would require 3-D modeling of the Milky Way's magnetic field and electron density to fit small-scale structure and firm up the coupling estimate.","The template-plus-ILC pipeline can be applied to other non-Gaussian microwave signals for which the power spectrum is not a sufficient statistic, such as diffuse synchrotron and dust emission.","The patch-by-patch inference can be used as a diagnostic: outlier sky patches flag poorly modeled foregrounds or template misfit, and comparing patches estimates systematic error alongside the statistical bound.","The method is computationally cheap relative to full 3-D radiative-transfer searches, making full-sky Bayesian searches feasible."],"supporting_citations":[{"why":"Supplies the resonant photon-ALP conversion formalism and the Milky Way signal equations that generate the mock maps.","marker":"[55]"},{"why":"Supplies the fiducial large-scale Galactic magnetic-field model used in the signal simulation.","marker":"[64]"},{"why":"Sets the magnetic-field scale height adopted in the simulation within the fiducial field model.","marker":"[85]"},{"why":"Supplies the NE2001 Galactic electron-density model that determines where resonant conversion occurs along each line of sight.","marker":"[86]"},{"why":"Establishes that the conversion signal is strongly non-Gaussian, motivating the template search in place of power-spectrum statistics.","marker":"[67]"},{"why":"Provides the current laboratory upper limit that the LiteBIRD forecast is compared against.","marker":"[68]"},{"why":"Provides the simulated synchrotron and thermal-dust foreground skies used to build the mock observations.","marker":"[99]"},{"why":"Provides the internal linear combination cleaning method used to separate the ALP signal from foregrounds across frequency channels.","marker":"[105]"},{"why":"Supplies the MCMC sampler used for the Bayesian posterior inference of the coupling.","marker":"[112]"}],"fun_headline_variants":["LiteBIRD template search could tighten axion limits 10x beyond CAST","Template-based CMB analysis gives LiteBIRD order-of-magnitude ALP edge","Galactic axion imprint: LiteBIRD's template search beats CAST tenfold","LiteBIRD's multi-band CMB template offers 10x better axion bounds than CAST"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The forecast stands or falls on the accuracy of the adopted large-scale models of the Milky Way's magnetic field and free-electron density, together with the simulated foregrounds standing in for the real sky; biased profiles would shift every quoted bound.","fun_headline_variants_meta":{"raw":{"variants":["LiteBIRD template search could tighten axion limits 10x beyond CAST","Template-based CMB analysis gives LiteBIRD order-of-magnitude ALP edge","Galactic axion imprint: LiteBIRD's template search beats CAST tenfold","LiteBIRD's multi-band CMB template offers 10x better axion bounds than CAST"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001305,"raw_usage":{"total_tokens":5417,"prompt_tokens":1137,"completion_tokens":4280,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":753,"completion_tokens_details":{"reasoning_tokens":4188}},"tokens_in":753,"tokens_out":4280,"duration_ms":29405,"temperature":1.0,"reasoning_tokens":4188,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:51:13.853965+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Regenerate the mock maps and run the same pipeline with an independent, data-driven reconstruction of the Galactic magnetic field and electron density, for example from pulsar Faraday-rotation and dispersion measures or from a different published field model. If the 95% upper limit on $g_{a\\gamma}$ changes by more than the reported statistical error, the constraint is dominated by model choice rather than by data, and the forecast should not be read as a reliable LiteBIRD reach.","supporting_citations":[{"cited_title":"Mukherjee, R","cited_arxiv_id":null,"evidence_quote":"Supplies the resonant photon-ALP conversion formalism and the Milky Way signal equations that generate the mock maps."},{"cited_title":"Jansson and G.R","cited_arxiv_id":null,"evidence_quote":"Supplies the fiducial large-scale Galactic magnetic-field model used in the signal simulation."},{"cited_title":"Gaensler, G","cited_arxiv_id":null,"evidence_quote":"Sets the magnetic-field scale height adopted in the simulation within the fiducial field model."},{"cited_title":"Cordes,Ne2001: a new model for the galactic electron density and its fluctuations, inMilky Way Surveys: The Structure and Evolution of our Galaxy, vol","cited_arxiv_id":null,"evidence_quote":"Supplies the NE2001 Galactic electron-density model that determines where resonant conversion occurs along each line of sight."},{"cited_title":"Turbulence Induced Non-Gaussian Spectral Distortion in the Microwave Sky from Photon-Axion Conversion in Galaxy Clusters","cited_arxiv_id":"2503.05302","evidence_quote":"Establishes that the conversion signal is strongly non-Gaussian, motivating the template search in place of power-spectrum statistics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the current laboratory upper limit that the LiteBIRD forecast is compared against."},{"cited_title":"Thorne, J","cited_arxiv_id":null,"evidence_quote":"Provides the simulated synchrotron and thermal-dust foreground skies used to build the mock observations."},{"cited_title":"\"Internal Linear Combination\" method for the separation of CMB from Galactic foregrounds in the harmonic domain","cited_arxiv_id":"0811.4277","evidence_quote":"Provides the internal linear combination cleaning method used to separate the ALP signal from foregrounds across frequency channels."},{"cited_title":"Foreman-Mackey, D.W","cited_arxiv_id":null,"evidence_quote":"Supplies the MCMC sampler used for the Bayesian posterior inference of the coupling."}],"review_version":1}