{"id":"f32feae0-b00a-418b-abb6-c17259f4573d","arxiv_id":"2506.20382","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"PBjam 2.0 adds three Bayesian models for automatic identification of l=1 mixed-mode frequencies in main-sequence, subgiant, and red-giant stars.","lead":"This paper describes PBjam 2.0, an open-source tool that automatically measures dipole mode oscillation frequencies in stars, from main sequence to red giants. If it works reliably, it would let astronomers characterize thousands of stars' masses, radii, ages, and internal rotation much faster than manual analysis.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim lacks quantitative validation: only three hand-picked stars, no independent frequency comparison, and demonstration stars may sit inside the prior sample.","rationale":"The paper's central claim is about automatic and reliable dipole-mode identification. The underlying physics and code structure are plausible: the asymptotic p-mode relation (Eq. 7), the coupled-cavity eigenvalue problem (Eq. 9), and the uniform-coupling root-finding (Eq. 12) follow established mixed-mode theory (Ong & Basu 2020), and the code is open source and reproducible. However, the only evidence that the pipeline actually identifies l=1 modes correctly is visual agreement on three stars, one per regime, with no independent frequency comparison. Because the prior is constructed from a large Kepler/TESS sample that may contain those same stars, the demonstrations cannot exclude the possibility that the results are partially memorized rather than predictive. The reader's weakest assumption about the subgiant coupling-matrix compression (Section 2.2.2) is a legitimate secondary concern: the p_L and p_D scalings are demonstrated on one red-giant model (Fig. 5), and residual errors are absorbed by per-mode Gaussian jitter of width 3% Delta-nu, which could mask systematic misplacement in real subgiants. But even if that compression were perfect, the central claim would remain unvalidated without a quantitative test on a broader sample. Therefore the paper should be accepted only conditionally: acceptance should require the excluded-prior rerun and a simulated-recovery study. This recommendation matches the reader's CONDITIONAL verdict, though it identifies the validation gap as the more load-bearing concern than the coupling compression alone.","tokens_in":21942,"tokens_out":4731,"duration_ms":55416,"concrete_test":"Exclude KIC 5184732, KIC 5723165, and KIC 4448777 from the prior sample and rerun the full mode-identification and peakbagging pipeline on those same power spectra, comparing the recovered l=1 frequencies with published independent measurements (e.g., Li et al. 2020a; Appourchaux 2020). In parallel, generate simulated power spectra from GYRE eigenfrequencies for 20-50 stars spanning the MS, SG, and low-luminosity RG regimes, inject them into realistic noise, and run PBjam 2.0 blind; report the fraction of l=1 modes recovered within the 3% Delta-nu prior width and the rate of radial-order misassignment. If recovery is accurate and unbiased with the demonstration stars removed from the prior, the central claim is supported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The strongest claim is that PBjam 2.0 automatically identifies l=1 modes in main-sequence, subgiant, and low-luminosity red-giant stars. The evidence presented for this claim consists of three demonstration stars, one per model (KIC 5184732, KIC 5723165, KIC 4448777; Figs. 3, 4, 6), with no quantitative success metric. No comparison is made between the recovered dipole frequencies and independent measurements, no simulated spectra with known injected mode frequencies are analyzed, no failure rate or radial-order misassignment rate is reported, and the sensitivity of the results to the manually chosen N_p or to the user-selected model (MS/SG/RG) is not assessed. The demonstration is therefore consistent with the method working on favorable examples without establishing that it works automatically across the intended population. A related concern is that the nonparametric prior (Section 2.4.1) is built from 13,413 Kepler targets plus 30,812 stellar-model grid samples; the paper does not state that the three demonstration stars are excluded from this sample. If any of these stars contributed their previously measured l=1 parameters to the prior, the demonstrations would be partially retrospective, not predictive. The abstract's 'automatically identify' is thus not yet supported by the presented tests.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper describes the PBjam 2.0 software extension for automatically identifying dipole (l=1) mode frequencies in solar-like oscillators. Three new models are introduced: an asymptotic p-mode relation for main-sequence (MS) stars (Eq. 7), a frequency-dependent coupling-matrix formalism for subgiants (SG, Eq. 9) in which the full coupling matrices are compressed into two scalars p_L and p_D, and a uniform coupling model for red giants (RG, Eq. 12). The mode identification is embedded in the Bayesian framework of the original PBjam, using a nonparametric prior constructed from 13,413 Kepler targets, 288 TESS targets, and 30,812 model-grid samples (Section 2.4.1), followed by a detailed peakbagging stage (Section 3). The paper demonstrates the method on three stars, one per model: KIC 5184732 (MS), KIC 5723165 (SG), and KIC 4448777 (RG), showing visually plausible fits in Figures 3, 4, and 6.","tokens_in":22224,"tokens_out":2116,"duration_ms":24042,"significance":"If the claimed capability is genuine, this is a valuable contribution to the asteroseismology software ecosystem: it extends an open-source, widely used package toward automated l=1 identification across evolutionary stages, which would enable large-scale analyses of mixed-mode frequencies and downstream stellar characterization. The mathematical ingredients (asymptotic relations, Hermitian eigenvalue coupling, JWKB characteristic equations) are standard, and the code is released openly. However, the paper's central claim is currently supported only by three hand-picked demonstrations with no quantitative validation, and one prior-construction detail raises a possible circularity. These issues are fixable within the manuscript's scope, but they must be addressed before the claim 'automatically identify l=1 modes' is established.","major_comments":[{"comment":"The central claim that PBjam 2.0 'automatically identify[ies] l=1 modes' is not quantitatively validated. The evidence consists of three demonstration stars with visually plausible fits; there is no comparison between the recovered dipole frequencies and independent measurements, no injection-recovery test on simulated spectra with known mode frequencies, no reported failure rate or radial-order misassignment rate, and no sensitivity analysis to the manually chosen N_p or to the user-selected model (MS/SG/RG). As written, the demonstrations establish consistency with favorable examples but not automatic performance across the intended population. I recommend adding a validation section that (a) analyzes simulated spectra with injected l=1 modes across MS, SG, and RG regimes and reports recovery accuracy and misassignment rates, and (b) compares PBjam 2.0 results for a modest sample of stars against independent peakbagging catalogs.","section":"Section 2.2, Figs. 3, 4, 6, and Abstract"},{"comment":"The nonparametric prior is built from 13,413 Kepler targets and 30,812 model-grid samples, and the paper does not state whether the three demonstration stars (KIC 5184732, KIC 5723165, KIC 4448777) are excluded from this sample. If any of these stars contributed their previously measured l=1 parameters to the prior, the demonstrations would be partially retrospective rather than predictive. The paper should explicitly state the exclusion protocol for validation targets, or alternatively re-run the demonstrations with a prior constructed after removing the demonstration stars from the sample, and report whether the results change materially.","section":"Section 2.4.1 (priors)"},{"comment":"The compression of the frequency-dependent coupling matrices L and D into two scalars p_L and p_D, with the index-dependence scalings L_ij ≈ ω_g,j^2 p_L and D_ij ≈ ω_g,j/ω_p,i p_D, is validated only on one red-giant structural model (Fig. 5). The SG model is intended for subgiants, where the coupling regime differs (few g-modes coupling to many p-modes), and the residual frequency dependence, acknowledged in the text as '~1 µHz' discrepancies, is then absorbed by a per-mode Gaussian jitter σ_nl ~ N(0, 3% Δν). The paper should demonstrate on subgiant stellar models that (a) the same scalings suppress index dependence, and (b) the 3%-Δν jitter does not systematically bias the recovered mixed-mode frequencies. Without this, the eigenvalue problem in Eq. 9 may place mixed modes incorrectly in real subgiants, and the jitter would mask rather than reveal the error.","section":"Section 2.2.2, Eq. (9), Fig. 5"}],"minor_comments":[{"comment":"The paper says 'the user to decide when using each method is appropriate' but the abstract claims PBjam 'automatically identify[ies] l=1 modes'. The automaticity is therefore conditional on a user-provided model choice; this nuance should be stated in the abstract or conclusions to avoid overclaiming.","section":"Section 2.2 (introductory paragraph)"},{"comment":"In the Appendix table, the row for the coupling matrix D lists 'pL' twice; the second entry should be 'pD'.","section":"Table 3"},{"comment":"The statement 'We therefore only apply this method to SG stars' is slightly confusing given the preceding sentence says the construction is also usable for MS and RG stars. Clarify that the limitation is purely computational, not physical.","section":"Section 2.2.2, last paragraph"},{"comment":"The stopping criterion for Dynesty is described as 'a change in log-evidence falls below the percent level'; this is vague. Specify the exact threshold or the Dynesty default used.","section":"Section 2.4.2"},{"comment":"The paper would benefit from a short description of the Kepler/TESS data reduction used for the demonstration stars (e.g., which light curves, preprocessing, and whether the spectra are taken from a public catalog).","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is a software/technical contribution with a broad user community, and the prior-overlap concern is easily fixable but should not be left implicit. If the authors add a real validation section, the paper could become acceptable; if they argue that three demonstrations suffice, the abstract's automatic-identification claim should be softened."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi — the short version: this is a legitimate and useful software paper, but the headline claim ('automatically identify l=1 modes') runs ahead of the evidence. The three models are not new physics — Tassoul, Deheuvels & Michel, Ong & Basu, and Shibahashi are all cited — but the integration into one Bayesian pipeline, the two-scalar compression of the coupling matrices, and the data-driven prior construction are real extensions. The paper reads well, the methods section is specific, and the code is open source. That counts.\n\nThe demonstrations on KIC 5184732, KIC 5723165, and KIC 4448777 are visually plausible, and the echelle diagrams suggest the identifications are reasonable. I agree with the reader's concern, though: there is no quantitative validation. No injection-recovery, no comparison with independently measured mode frequencies, no failure or misassignment rate, no sensitivity to N_p or to the choice of MS/SG/RG model. With only three hand-picked stars, the automatic claim is not established across the intended population. This is the load-bearing soft spot, and it is real.\n\nThe prior-construction issue is also worth raising. Section 2.4.1 builds the prior from 13,413 Kepler targets plus 30,812 model-grid samples, and the paper does not state that the three demonstration stars are excluded. If they are in the sample, the demonstrations are partially retrospective. That should be easy to check, but the omission matters.\n\nThe smaller worry — the SG model's compression of L and D into two scalars, shown on one RG structure model — is a legitimate concern but not damning. The paper itself acknowledges residual index dependence and adds a 3% Delta-nu jitter to absorb it. That is a reasonable stopgap, but it does mean the validation gap is wider for the SG model specifically.\n\nNet: the integration is genuinely useful, the methods are reproducible in principle, and the physics is not exotic. The paper deserves a serious referee. I would ask for quantitative validation before publication: at minimum an injection-recovery test on simulated spectra and an independent comparison for the three stars, plus a statement on prior overlap. If those come back clean, PBjam 2.0 is exactly the kind of tool the community will use.","headline":"PBjam 2.0 is a useful, clearly explained software extension for automated l=1 identification, but the central claim runs ahead of the quantitative evidence.","tokens_in":22825,"tokens_out":1730,"would_cite":true,"duration_ms":19354,"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":"PBjam 2.0 automatically identifies dipole-mode frequencies in solar-like oscillators across three coupling regimes, from main-sequence p-modes to red-giant mixed modes.","keywords":["asteroseismology","dipole modes","mixed modes","mode identification","subgiant stars","red giant stars","Bayesian inference","PBjam"],"falsifier":"Compute the full coupling-matrix elements from Equation (11) for a grid of subgiant stellar models spanning a range of masses and evolutionary states, solve the full eigenvalue problem, and compare those exact mixed-mode frequencies with the scalar-compressed solution using $p_L$ and $p_D$. If the residuals exceed the 3% of $\\Delta\\nu$ jitter for a substantial fraction of modes, the subgiant model's frequency predictions would be shown to be inaccurate; a test on synthetic spectra with known injected mixed modes would show whether the posterior mode frequencies recover the inputs.","tokens_in":21749,"feed_emoji":"🔭","tokens_out":8662,"duration_ms":84944,"temperature":0.7,"pith_summary":"This paper extends the PBjam software to identify $\\ell=1$ (dipole) oscillation modes automatically. Previously PBjam only handled $\\ell=0$ and $\\ell=2$ modes, which stay regularly spaced through stellar evolution; dipole modes become mixed pressure-gravity modes whose frequencies do not follow a simple pattern once stars leave the main sequence. The update provides three frequency models — an asymptotic p-mode relation for main-sequence stars, a frequency-dependent coupling-matrix model for subgiants, and a uniform coupling model for red giants — and feeds the resulting mode locations into a detailed peakbagging stage. If the models work as claimed, non-experts and large surveys can extract dipole-mode frequencies, and with them core-rotation and age information, from thousands of stars with minimal interaction.","feed_headline":"Three models find dipole modes from MS to red giants","feed_subtitle":"New Bayesian models automatically locate l=1 mixed-mode frequencies in main-sequence, subgiant, and red-giant spectra.","key_machinery":"The carrying device is a two-stage spectral model with a Bayesian prior built from a nonparametric sample. Stage one fits a background plus $\\ell=0,2$ Lorentzian pairs using the asymptotic p-mode relation $\\nu_{n_p,0} = \\Delta\\nu\\left(n_p + \\varepsilon_p + \\frac{\\alpha_p}{2}(n_p - n_{\\max})^2\\right)$; stage two divides the spectrum by that fit and fits the residual with $\\ell=1$ Lorentzians whose frequency centers come from one of three prescriptions. The subgiant prescription is the generalized Hermitian eigenvalue problem of Equation (9) with coupling matrices recovered from the scalars via $L_{ij} \\approx \\omega_{g,j}^2 p_L$ and $D_{ij} \\approx (\\omega_{g,j}/\\omega_{p,i}) p_D$; the red-giant prescription is the characteristic equation $\\tan\\theta_p(\\nu)\\tan\\theta_g(\\nu) - q(\\nu) = 0$ with $q$ sampled from a uniform prior. Priors are kernel-density estimates over a sample of 13,413 Kepler targets, 288 TESS targets, and 30,812 model-grid subgiants, projected into a lower-dimensional latent space for efficient sampling.","core_discovery":"The paper's central claim is that one automated Bayesian pipeline, PBjam 2.0, can identify $\\ell=1$ mode frequencies across all three coupling regimes by choosing among three model prescriptions. For main-sequence stars the dipole modes are treated as pure p-modes offset from the $\\ell=0$ ridges by a constant $d_{01}$. For subgiants, the pipeline builds the mixed-mode frequencies by solving a generalized Hermitian eigenvalue problem whose p-g coupling matrices are compressed into two scalars, $p_L$ and $p_D$, using frequency-based scalings, with a per-mode Gaussian jitter of 3% of $\\Delta\\nu$ to absorb residual model error. For red giants, the dipole modes are the roots of a characteristic equation with a constant coupling strength $q$, found by damped Halley's method. The paper demonstrates the identification on one main-sequence star, one subgiant, and one red giant, and shows the subgiant case through detailed peakbagging, arguing that the same prior methodology used for $\\ell=0,2$ naturally extends to the $\\ell=1$ parameters.","pith_inferences":["A natural extension would apply the same scalar-compression treatment to $\\ell=2$ mixed modes, which the paper notes is out of scope, since the eigenvalue formalism already supports it.","The 3% of $\\Delta\\nu$ per-mode jitter could double as a diagnostic: on stars with independently measured mode frequencies, the width of that jitter posterior would reveal whether model error or intrinsic scatter dominates.","Combining the staged identification with joint multi-dataset likelihoods, for example TESS plus K2 or PLATO plus Kepler data, might rescue faint or short-cadence subgiants that do not constrain the modes alone.","The choice among the main-sequence, subgiant, and red-giant models is left to the user; an automatic evolutionary-stage classifier based on $\\nu_{\\max}$ and $\\Delta\\Pi_1$ would make the pipeline fully hands-off."],"forward_implications":["PBjam 2.0 can automatically deliver $\\ell=1$ mode frequencies for main-sequence, subgiant, and low-luminosity red-giant stars, removing the need for manual dipole-mode identification.","Mixed-mode frequencies from the subgiant and red-giant models provide access to core rotation and the g-mode period spacing, quantities tied to internal structure and stellar age.","The detailed peakbagging stage inherits the mode identification as priors, so rotation, acoustic glitches, and correlations between neighboring modes can be measured without re-labeling modes.","The nonparametric prior construction can be updated as more subgiant observations accumulate, directly improving the model in the region where the current observational sample is sparse.","The red-giant rotation option that couples rotating p- and g-modes separately by azimuthal order can produce asymmetric multiplet splittings, which are needed to interpret core rotation in advanced red giants."],"supporting_citations":[{"why":"Initial PBjam release; supplies the Bayesian mode-identification methodology, likelihood, and prior construction that this update extends to $\\ell=1$ modes.","marker":"Nielsen et al. 2021"},{"why":"Provides the coupling regimes and the non-asymptotic generalized Hermitian eigenvalue formulation, including the coupling-matrix integrals used by the subgiant model.","marker":"Ong & Basu 2020"},{"why":"Source of the non-asymptotic description of mixed modes that the subgiant eigenvalue model builds on.","marker":"Deheuvels & Michel 2010"},{"why":"Gives the characteristic equation whose roots define mixed-mode frequencies in the red-giant uniform-coupling model.","marker":"Shibahashi 1979"},{"why":"Supplies the asymptotic relations for p-mode and g-mode frequencies that underlie all three models.","marker":"Tassoul 1980"},{"why":"Provides the red-giant structure model used to validate the scalar compression of the coupling matrices and the rotating-frame coupling approach for asymmetric multiplets.","marker":"Ong et al. 2022"},{"why":"Supplies the 30,812 model-grid subgiant samples that fill the sparse observational prior for the $\\ell=1$ parameters.","marker":"Lindsay et al. 2024"},{"why":"Defines the three-term Harvey-like background model used in the first stage of mode identification.","marker":"Kallinger et al. 2014"}],"fun_headline_variants":["PBjam 2.0 adds dipole modes for every stellar phase","Three Bayesian models spot l=1 frequencies in any star","Automated Bayesian pipeline finds mixed modes in stars","From main sequence to red giants: PBjam finds dipole modes","PBjam 2.0:covers l=1 modes in all coupling regimes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that in subgiants, the complicated frequency-by-frequency interaction between the outer pressure modes and the inner gravity modes can be captured by just two constants, $p_L$ and $p_D$, after rescaling by the mode frequencies — a simplification demonstrated on one stellar model. If real subgiants violate this simplification, the predicted frequencies of the mixed modes will be wrong, and the added 3% of $\\Delta\\nu$ scatter could hide the error instead of exposing it.","fun_headline_variants_meta":{"raw":{"variants":["PBjam 2.0 adds dipole modes for every stellar phase","Three Bayesian models spot l=1 frequencies in any star","Automated Bayesian pipeline finds mixed modes in stars","From main sequence to red giants: PBjam finds dipole modes","PBjam 2.0:covers l=1 modes in all coupling regimes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000887,"raw_usage":{"total_tokens":3866,"prompt_tokens":1018,"completion_tokens":2848,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":2761}},"tokens_in":634,"tokens_out":2848,"duration_ms":25246,"temperature":1.0,"reasoning_tokens":2761,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:50:09.796887+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the full coupling-matrix elements from Equation (11) for a grid of subgiant stellar models spanning a range of masses and evolutionary states, solve the full eigenvalue problem, and compare those exact mixed-mode frequencies with the scalar-compressed solution using $p_L$ and $p_D$. If the residuals exceed the 3% of $\\Delta\\nu$ jitter for a substantial fraction of modes, the subgiant model's frequency predictions would be shown to be inaccurate; a test on synthetic spectra with known injected mixed modes would show whether the posterior mode frequencies recover the inputs.","supporting_citations":[{"cited_title":"2010, Ap&SS, 328, 259, doi: 10.1007/s10509-009-0216-2","cited_arxiv_id":null,"evidence_quote":"Source of the non-asymptotic description of mixed modes that the subgiant eigenvalue model builds on."}],"review_version":1}