{"id":"784d2bac-c60c-497d-b7cc-1f7f1c066aaf","arxiv_id":"2608.05295","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A PTA likelihood expressed in terms of low-order spherical harmonics of the Earth term and pulsar-term variance retains roughly 95% of the information about a stochastic background, and ell_max=3 plus the pulsar-term dipole retains at least 90% for point sources.","lead":"This paper shows that most of the gravitational-wave information in pulsar timing array data can be compressed into about ten numbers per frequency bin: the lowest spherical harmonic coefficients of the Earth-term signal and the pulsar-term variance. This makes searches for the background and individual sources much cheaper after a single full-data analysis.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Point-source compression is validated only for amplitude Fisher information; the ten-coefficient summary is not yet shown to preserve localization or other source parameters, so the 'carry all of the signal' claim outruns the evidence.","rationale":"The paper is carefully argued and does much of what a compression claim requires: an explicit likelihood, analytic harmonic coefficients, public code, and an appendix that directly tests the Gaussian pulsar-term approximation (App. A), finding small Hellinger/KL deviations for NG15 noise realizations. I do not rest the objection on Eq. 6. The weakest load-bearing step for the broad central claim is the identification of 'information about the signal' with the Fisher information of a single amplitude parameter. Eq. 58 defines R via scalar Fisher ratios, and Eqs. 69-73 apply it only to A_s with an orientation-averaged template. The conclusion then states that the summary contains 'all of the signal' and is sufficient for 'any signal or hypothesis testing.' That inference is not yet demonstrated. For point sources, the scientifically important information includes sky location; the paper itself emphasizes resolving individual SMBHBs. High-ell modes contribute to the narrow Earth-term pattern near the source, so a summary that preserves 90% of amplitude SNR may not preserve the Fisher information for position. The proposed check directly computes the localization Fisher matrix in the same NG15 setup; it would either confirm that ell_max=3 plus the pulsar dipole is nearly lossless for source localization or show that the summary is nearly lossless only for amplitude. This supports the reader's CONDITIONAL verdict rather than moving it: the paper should either narrow the claim to amplitude retention or add the parameter-retention demonstration.","tokens_in":23339,"tokens_out":6849,"duration_ms":71227,"concrete_test":"Compute the full 5- to 7-parameter Fisher matrix for a deterministic source (amplitude, sky longitude and latitude, inclination, polarization, phase, and frequency/chirp mass) at the NG15 f2 bin for the full likelihood and for the compressed statistic ell_max=3 plus the pulsar dipole, replacing d_A_s in Eq. 70 with derivatives with respect to each source parameter via Eq. 74. If the compressed Cramer-Rao bounds on sky location are more than roughly 10-20% larger than the full-data bounds, or if the compressed maximum-likelihood position is biased by more than the full-data 1-sigma error, then the 'carry all of the signal' conclusion fails for point sources.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing gap is in the point-source half of the central claim. The quantitative evidence for compression is the scalar Fisher ratio R^2 = tilde-F_AA / F_AA (Eq. 58), which measures information about a single overall amplitude A. For a deterministic source, the compressed derivative is d_A_s <tilde-a_L> = F_L a_L + F_LH a_H (Eq. 69), and the template is polarization/inclination averaged (Eq. 73). This establishes retention of matched-filter SNR for source strength, but not retention of information about sky location, polarization, inclination, phase, or frequency/chirp parameters. The Abstract and Sec. VI nonetheless conclude that the truncated coefficients 'carry all of the signal' and can replace the full dataset for 'any signal or hypothesis testing.' Localization information is not automatically preserved by a high amplitude-retention ratio: near the source the Earth-term pattern is a narrow cross generated by high-ell modes (Fig. 2, left), and the paper itself finds the worst retention for sources near sensitive pulsars. A truncated map with 90% amplitude retention could still degrade or bias position recovery, especially because high-ell leakage through F_LH aliases into the retained low-ell modes, and the Wigner rotation in Eq. 74 makes source-position derivatives different from amplitude derivatives. The stochastic-background case is less affected because an isotropic background has essentially one amplitude parameter; the point-source 'across the entire sky' claim is where the evidence and the conclusion are not yet matched.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a summary statistic for pulsar timing array (PTA) data based on the spherical-harmonic coefficients of the Earth-term map and the pulsar-term variance map. It derives a likelihood (Eq. 9) in which unknown pulsar phases are marginalized using a Gaussian pulsar-term approximation (Eq. 6), and it shows that the expected angular power spectrum of any GW signal is the steeply decaying Hellings–Downs spectrum (Eq. 17). Using Fisher information with the NANOGrav 15yr noise properties and sky locations, it computes the fraction of amplitude Fisher information retained when the maps are truncated at low multipoles (Eq. 58). The main quantitative results are that ell_max=2 for the Earth term plus the pulsar-term monopole retain about 95% of the information about an isotropic background, and that for point sources ell_max=3 plus the pulsar-term dipole achieve at least 90% retention across the sky. Appendix A checks the Gaussian pulsar-term approximation via Edgeworth and KL expansions, quoting an upper Hellinger-distance bound near 0.13 for NG15.","tokens_in":23622,"tokens_out":5463,"duration_ms":51920,"significance":"If the full claim were established, the paper would provide an interpretable and computationally cheap compression of PTA data to roughly ten coefficients per frequency bin, with downstream searches performed on the compressed summary. The use of the steep HD multipole decay to motivate low-dimensional summaries is a natural and valuable idea, and the quantitative Fisher forecasts grounded in public NG15 noise products are a useful contribution. The careful Appendix A examination of the Gaussian pulsar-term approximation is a strength, as is the public code release. However, the evidence presented supports retention of information about a single overall amplitude, not the broader statement that the compressed coefficients 'carry all of the signal' for any signal or hypothesis test. The point-source localization gap is the main obstacle to the paper's strongest conclusion.","major_comments":[{"comment":"","section":""}],"minor_comments":[{"comment":"The caption contains a typo: 'correponds' should be 'corresponds'.","section":"Fig. 1 caption"},{"comment":"The power spectrum notation C_0, C_1, C_2 for the pulsar-term variance map is used in Eqs. (24)-(25) without an explicit definition; stating the normalization relative to Eq. (17) would improve readability.","section":"Eqs. (24)-(25)"},{"comment":"The right panel uses both point size and color to indicate the Earth-term SNR per pulsar, but the figure as reproduced provides no quantitative color scale; a colorbar or explicit text values would help the reader interpret the map.","section":"Fig. 3, right panel"},{"comment":"The phrase 'fully characterize' overstates what is demonstrated; the quantitative results concern the Fisher information of an overall amplitude, so the abstract and conclusion should either use more limited wording or add the missing parameter-space information.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the journal's scope and the public code release is a strength. The main issue is the gap between the amplitude-only retention results and the stronger claims about full signal characterization; if the authors compute localization retention or narrow the claims appropriately, I would support publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new thing here is not the harmonic-space expansion itself, which goes back to Roebber and Holder and the harmonic-analysis papers that followed. It is the explicit likelihood (Eq. 9) written in terms of the Earth-term map and the pulsar-term variance, and the quantitative demonstration that compressing to a handful of spherical-harmonic coefficients retains most of the Fisher information about the overall amplitude of a stochastic background or a point source. They use the real NG15 noise properties and sky locations, and they ship code and data products that reproduce the figures. That is concrete and reproducible. The stochastic-background half is in good shape. The HD multipole decay is standard; the Fisher retention ratios are clearly defined; the CURN marginalization is handled; the 95 percent number for ell_max=2 plus the monopole is believable. The Gaussian pulsar-term approximation is the load-bearing step for the point-source half, and they check it in Appendix A with Edgeworth and KL expansions against the true phase-marginalized ring distribution. The median Hellinger bound is small; the 1-sigma upper bound near 0.13 is okay but not nothing. The soft spot is the point-source half of the abstract's 'carry all of the information' claim. What is actually computed for point sources is the Fisher ratio R^2 for a single overall amplitude A_s, using a template averaged over polarization and inclination (Eqs. 69-73). That establishes retention of matched-filter SNR for source strength. It does not establish retention of information about sky location, polarization, inclination, phase, or chirp parameters. Near the source the Earth-term map is a narrow cross built from high-ell modes, and the paper itself shows the worst retention for sources near sensitive pulsars. A map keeping 90 percent of amplitude information could still degrade localization. The conclusion reaching beyond amplitude to 'any signal or hypothesis testing' is not supported by the evidence presented. This is a wording-and-scope problem rather than a flaw in the derivations, but it matters because the point of compression is to enable downstream searches for individual sources. Minor points: noise parameters and the spectral index are fixed in the forecasts; Eq. 27 notes the frequency-evolution caveat; and the point-source compressed statistic is defined with respect to a known template, so the summary is not fully signal-agnostic in that regime, though the formalism itself is. Verdict: worth a serious referee. The central claim for stochastic backgrounds holds; for point sources the authors should either soften 'carry all of the signal' to 'carry most of the SNR for source amplitude' or add a localization Fisher check. Either way, this deserves peer review.","headline":"A genuinely useful compression result for the stochastic background, with the point-source 'carry all the signal' wording running ahead of the amplitude-only Fisher evidence.","tokens_in":812,"tokens_out":796,"would_cite":true,"duration_ms":26481,"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":"The information content of a pulsar timing array's gravitational-wave signal reduces to roughly ten spherical harmonic coefficients per frequency bin.","keywords":["pulsar timing arrays","gravitational-wave background","spherical harmonics","Hellings-Downs curve","data compression","Fisher information","point-source search","pulsar-term variance"],"falsifier":"Simulate timing residuals from the exact single-source likelihood of Eq. A2, sampling the pulsar-term phase uniformly rather than treating it as Gaussian, for a source at the sky location where the paper's retention is lowest, and recompute the Fisher retention ratio $R$ for $\\ell_{\\max}=3$ with a pulsar-term dipole; if $R$ drops below 0.9, the central compression claim fails.","tokens_in":23137,"feed_emoji":"🔭","tokens_out":7543,"duration_ms":63150,"temperature":0.7,"pith_summary":"Gravitational-wave signals leave two imprints in pulsar timing residuals: a coherent Earth term and an incoherent pulsar term whose phases are unknown because pulsar distances are poorly measured. This paper argues that once pulsar distances are marginalized out, all the information a pulsar timing array can extract from those residuals is carried by the spherical harmonic coefficients of the Earth-term map and of the pulsar-term variance map. Because every signal produces the same steeply falling Hellings-Downs angular spectrum on average, the information is concentrated at low multipoles. Using the noise properties and sky positions of the 15-year dataset as a realistic case, the paper finds that keeping only the $\\ell=2$ Earth-term modes plus the monopole of the pulsar-term variance retains about 95% of the information about an isotropic background, and that $\\ell_{\\max}=3$ with the pulsar-term dipole keeps at least 90% of the information about a point source anywhere on the sky. If correct, the full likelihood can be replaced by on the order of ten coefficients per frequency bin, making every downstream search cheap and consistent.","feed_headline":"Ten coefficients per frequency bin hold 95% of PTA signal","feed_subtitle":"Earth-term harmonics plus pulsar-term variance compress the full PTA likelihood with little loss.","key_machinery":"The load-bearing object is the modified PTA likelihood, Eq. 9, in which the Earth term is expanded as $z_e = \\sigma_h \\sum_{\\ell m} a_{\\ell m} Y_{\\ell m}$ and the total per-pulsar variance is $\\sigma_i^2 = \\sigma_h^2 \\sum_{\\ell m} b_{\\ell m} Y_{\\ell m}(\\hat{n}_i) + \\sigma_{n,i}^2$. The pulsar term is absorbed as a Gaussian variance, Eq. 6, rather than a grid over unknown phases. The universal Hellings-Downs spectrum in harmonic space, $C_\\ell \\propto 1/[(\\ell+2)(\\ell+1)\\ell(\\ell-1)]$, is the reason the compression works: it guarantees that a handful of low-$\\ell$ coefficients carries almost all signal-to-noise, and the pulsar-term variance map separately captures low-$\\ell$ auto-correlation information, including a dipole that helps locate point sources.","core_discovery":"The paper's central discovery is that the PTA likelihood can be rewritten, Eq. 9, so that the data enter only through the spherical harmonic coefficients $a_{\\ell m}$ of the Earth-term map $z_e(\\hat{n})$ and $b_{\\ell m}$ of the pulsar-term variance map $\\sigma_p^2(\\hat{n})$, after marginalizing over unknown pulsar phases with a Gaussian prior on the pulsar term. The harmonic-space form of the Hellings-Downs correlation, $C_\\ell \\propto 1/[(\\ell+2)(\\ell+1)\\ell(\\ell-1)]$, is what makes the signal low-dimensional: the quadrupole dominates and higher multipoles are strongly suppressed. The paper quantifies information retention with the Fisher ratio between the compressed and full likelihoods, using realistic noise and sky coverage from the 15-year dataset. The result is that $\\ell_{\\max}=2$ plus the pulsar-term monopole preserves roughly 95% of the information about an isotropic stochastic background, while point sources need $\\ell_{\\max}=3$ plus the pulsar-term dipole to reach at least 90% retention for every sky location.","pith_inferences":["The paper's Fisher-ratio test could be promoted to a posterior comparison: running the full and compressed likelihoods on the same real 15-year data and checking that parameter posteriors overlap would test the summary in the regime the Fisher calculation only approximates.","Because the summary lives in a fixed spherical-harmonic basis, different pulsar timing arrays could publish their compressed coefficients and combine them without sharing raw arrival-time chains, a possibility the paper leaves implicit.","As arrays add pulsars and the sky sampling becomes more uniform, the number of modes needed should fall rather than rise, which could be checked by repeating the Fisher-ratio calculation on future datasets."],"forward_implications":["A single analysis over the full timing residuals can extract on the order of ten coefficients per frequency bin, and isotropic, anisotropy, and continuous-wave searches can all be run afterward on that summary.","For an isotropic background, truncating the Earth term at $\\ell=2$ and keeping only the pulsar-term monopole loses only about 5% of the Fisher information about the signal amplitude.","For point sources, the pulsar-term variance dipole carries significant information: including it together with $\\ell_{\\max}=3$ Earth-term modes keeps at least 90% of the source amplitude information for every sky location.","The degeneracy between the auto-correlation (CURN) and the Hellings-Downs angular pattern that appears under aggressive truncation is largely closed by measuring the pulsar-term monopole.","Because the retained low-$\\ell$ harmonics alias and recapture the high-$\\ell$ signal in a non-uniform array, the compression loses less information than a naive counting of modes would suggest."],"supporting_citations":[{"why":"Establishes the detected low-frequency gravitational-wave background whose characterization motivates the compression.","marker":"[1]"},{"why":"Supplies the Hellings-Downs correlation that the paper recasts in harmonic space.","marker":"[6]"},{"why":"Introduces CMB-style spherical-harmonic mapping of gravitational-wave backgrounds that the paper extends to a nearly lossless statistic.","marker":"[22]"},{"why":"Gives the harmonic-space form of the Hellings-Downs spectrum used in Eq. 17.","marker":"[23]"},{"why":"Supplies the harmonic-space Earth-term coefficients and the notation for $C_\\ell$ adopted here.","marker":"[26]"},{"why":"Provides the 15-year noise properties and sky locations used in the realistic Fisher forecasts.","marker":"[30]"},{"why":"Supplies the coupling-matrix formalism for the map Fisher matrices in Eqs. 28-29.","marker":"[32]"}],"fun_headline_variants":["Ten coefficients per bin keep 95% of PTA information","PTA likelihood shrinks to a handful of spherical harmonics","Low-dimensional harmonics preserve 95% of PTA data","95% of PTA info survives compression to few harmonics","Simplify PTAs: 10 harmonic coefficients capture 95%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Gaussian approximation for the pulsar term, Eq. 6, is the load-bearing simplification; it is exact for many sources but only approximate for a single dominant source, where the true phase-marginalized distribution is a ring rather than a Gaussian.","fun_headline_variants_meta":{"raw":{"variants":["Ten coefficients per bin keep 95% of PTA information","PTA likelihood shrinks to a handful of spherical harmonics","Low-dimensional harmonics preserve 95% of PTA data","95% of PTA info survives compression to few harmonics","Simplify PTAs: 10 harmonic coefficients capture 95%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000602,"raw_usage":{"total_tokens":2869,"prompt_tokens":1065,"completion_tokens":1804,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":681,"completion_tokens_details":{"reasoning_tokens":1721}},"tokens_in":681,"tokens_out":1804,"duration_ms":14821,"temperature":1.0,"reasoning_tokens":1721,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T15:52:29.390281+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate timing residuals from the exact single-source likelihood of Eq. A2, sampling the pulsar-term phase uniformly rather than treating it as Gaussian, for a source at the sky location where the paper's retention is lowest, and recompute the Fisher retention ratio $R$ for $\\ell_{\\max}=3$ with a pulsar-term dipole; if $R$ drops below 0.9, the central compression claim fails.","supporting_citations":[{"cited_title":"Upper limits on the isotropic gravitational radiation background from pulsar timing analysis.,","cited_arxiv_id":null,"evidence_quote":"Supplies the Hellings-Downs correlation that the paper recasts in harmonic space."},{"cited_title":"The NANOGrav 15-Year Data Set,","cited_arxiv_id":null,"evidence_quote":"Provides the 15-year noise properties and sky locations used in the realistic Fisher forecasts."}],"review_version":1}