{"id":"79d963a3-d575-4c70-a6f1-606b89865fe7","arxiv_id":"2509.07059","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A new Bayesian framework recovers EVPA rotation in IXPE data and favors a slab coronal geometry for GX 13+1.","lead":"This paper introduces QUEEN-BEE, a new Bayesian statistical method for detecting rotation in X-ray polarization data, and applies it to three IXPE observations of the neutron star binary GX 13+1. It reports clean detections of polarization angle rotation in two epochs and suggests a slab-like corona for the source.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"EVPA linearity assumption is load-bearing for the slab/shell claim; if the true rotation is nonlinear, QUEEN-BEE's recovered PDs are biased and the coronal geometry conclusion may flip.","rationale":"The reader's weakest_assumption identified the linear EVPA model as the key assumption, and I agree that this is the most load-bearing concern for the central claim. The paper's headline contributions are the QUEEN-BEE framework and the coronal geometry constraint. The framework's ability to detect rotation is robust to mild nonlinearity—any smooth rotation will produce evidence for the rotating model—but the recovered PD, and hence the slab/shell discrimination, is sensitive to the exact functional form. The factor-of-two inconsistency in Eqs. 9–10 (q(t)=p cos(2ψ0+ωt) instead of p cos(2ψ0+2ωt)) is a real internal error, but it affects the interpretation of the rotation-rate parameter rather than the PD-based geometry conclusion, provided the code implements the likelihood in Eq. 11 correctly. The post-dip transient detection lacks an explicit trials correction, but that affects only the third-observation transient claim, not the first-observation energy-resolved result. Thus the linearity assumption is the least secure link: it is unquantified, acknowledged as a limitation, and directly feeds the most novel astrophysical conclusion. A simulation-based test with nonlinear injected trajectories would settle whether the bias is significant at the precision claimed. Since the reader already set a CONDITIONAL verdict, my analysis does not require changing it.","tokens_in":921,"tokens_out":1021,"duration_ms":117846,"concrete_test":"Run a simulation campaign matching GX 13+1 Obs 1 count rates and background: generate event lists with known PD and several nonlinear EVPA trajectories (e.g., ψ=ψ0+ωt+βt² with β chosen to produce ~10–20° deviation, and a piecewise linear rotation), then fit with QUEEN-BEE's scout_rot. Check whether the recovered PD in 2–4 and 4–8 keV bands deviates from the input by more than the reported 1σ uncertainties, and whether the 4–8 minus 2–4 keV PD trend reverses or changes sign. If it does, the linearity assumption is unsafe for the slab/shell conclusion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest physical claim—that energy-resolved QUEEN-BEE analysis 'appears to exclude the shell geometry favoring a slab corona'—rests entirely on the PD values recovered under the linear EVPA model ψ(t)=ψ0+ωt (Eq. 8). In the first observation, QUEEN-BEE yields PD=2.3+0.4/−0.3% at 2–4 keV and PD=3.2±0.4% at 4–8 keV, a rising trend that is compared to MONK slab/shell predictions in Fig. 8. PCUBE, by contrast, gives nearly flat PD (~1.4–1.5%) across the same bands. The entire difference comes from modeling the EVPA rotation linearly. If the true EVPA evolution is not strictly linear—e.g., accelerating, piecewise, or energy-dependent in a more complex way—the residual mismatch between the model and data can bias the recovered PD. The paper itself acknowledges (Sec. 5) that QUEEN-BEE is model-dependent and 'best utilized in scenarios with physical motivation for smoothly rotating EVPA', but it does not quantify how large a nonlinearity would be required to change the PD trend. Since the slab-vs-shell conclusion hinges on a modest PD rise (~0.9%) between two energy bins, an unvalidated linearity assumption is the least secure link in the central argument.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces QUEEN-BEE, a Bayesian nested-sampling framework for unbinned IXPE event data, and applies it to three IXPE observations of the neutron star low-mass X-ray binary GX 13+1. The framework compares three models—constant EVPA, linearly rotating EVPA, and unpolarized—using Bayesian evidences and Bayes factors. For the first observation the authors report a rotation rate of 42±4 deg/day, consistent with earlier binned analyses, and an energy-resolved PD increase (2.3% at 2–4 keV, 3.2% at 4–8 keV) that they interpret as favoring a slab coronal geometry over a shell geometry. The second observation is consistent with a constant EVPA, and the third shows weak full-band rotation but a strong rotating signal in a post-dip window (170+20−40 deg/day). The paper concludes that QUEEN-BEE can recover rotation signals where binned methods fall below detection thresholds and that GX 13+1's corona is slab-like.","tokens_in":19899,"tokens_out":7011,"duration_ms":79185,"significance":"If the methodological and source-specific claims hold, QUEEN-BEE would be a useful addition to the IXPE polarimetric-timing toolbox: it replaces ad hoc time binning with a coherent likelihood and provides formal model comparison. The code is archived on Zenodo, and the nested-sampling/evidence formalism is standard and reproducible. The apparent detection of rotation in low-significance energy/time bins and the proposed slab-vs-shell discrimination for GX 13+1 would be of interest to the X-ray polarimetry community. However, the current manuscript does not yet establish the central physical conclusion: the slab-vs-shell claim rests on an internal equation inconsistency, a purely visual comparison to fixed model curves, and a post-hoc selected time window. The methodological core is sound and likely fixable, but the source-specific interpretation is over-stated relative to the presented evidence.","major_comments":[{"comment":"The rotating model is internally inconsistent. With ψ(t)=ψ0+ωt (Eq. 8), the instantaneous Stokes parameters should be q(t)=p cos(2ψ0+2ωt) and u(t)=p sin(2ψ0+2ωt). As written, Eqs. (9)–(10) imply a Stokes EVPA that rotates at ω/2, while the likelihood in Eq. (11) uses ψ(t) directly and therefore rotates at ω. The authors must clarify which expression is actually implemented in QUEEN-BEE. If the code follows Eqs. (9)–(10), all reported rotation rates in Tables 2, 4, and 6 are off by a factor of two. This is load-bearing because the rotation rate is a primary output and because the recovered PD depends on the modeled EVPA phase.","section":"§2.3, Eqs. (9)–(10) vs Eq. (11)"},{"comment":"The slab-vs-shell conclusion is not quantitatively supported. The QUEEN-BEE energy-resolved PD points (2.3±0.4% and 3.2±0.4%) are overplotted on fixed MONK slab/shell curves, with no goodness-of-fit statistic, no chi-square or Bayes factor comparing the two geometries, and no systematic uncertainties on the model curves (inclination, spectral state, Comptonization parameters, or wind contamination). The text states that QUEEN-BEE 'appears to exclude the shell geometry', but a visual separation of two points from a curve is not a statistical exclusion. The authors should provide a quantitative model comparison, include model uncertainties, or soften the claim accordingly.","section":"§3.3 and Fig. 8"},{"comment":"The time-resolved 'discovery' of rotation in the third observation uses the window 15.9–27.0 h after the start, which was selected because it corresponds to a light-curve dip identified in earlier analyses of the same data. This is a post-hoc selection. The reported lnB=11.4 for rotating vs constant EVPA in this window is conditional on the window choice; no trials factor or look-elsewhere correction is applied. Without such a correction, the evidence for a transient rotating event in Observation 3 is over-stated. The authors should either pre-specify analysis windows, correct for the number of windows tested, or explicitly frame the result as exploratory.","section":"§3.3, Table 6, and §4"},{"comment":"The linear-EVPA assumption ψ(t)=ψ0+ωt is acknowledged in Sec. 5 as making QUEEN-BEE 'best utilized in scenarios with physical motivation for smoothly rotating EVPA', but the paper does not quantify the sensitivity of the recovered PD to departures from linearity. The slab-vs-shell conclusion depends on a modest PD increase from 2.3% to 3.2% between two energy bands; a small bias from mis-modeled EVPA evolution could alter or reverse the trend. The authors should run injection-recovery simulations with nonlinear EVPA evolution (e.g., a changing rotation rate or a piecewise rotation) and show how the recovered PD and rotation rate are biased. This is necessary to support the physical conclusion.","section":"Sec. 5 and §3.3"}],"minor_comments":[{"comment":"Typo: 'inital' should be 'initial'.","section":"Eq. (8)"},{"comment":"The Fig. 8 top caption states the 2–4 keV PCUBE result has 4.1σ significance, while the text in §3.3 reports 4.2σ for the same quantity. Please harmonize.","section":"Fig. 8 caption vs §3.3"},{"comment":"The suggested substitution of sin^-1 for tan^-1 in the likelihood is not quadrant-safe when q can be negative; using atan2 is preferable. If the code uses the sin^-1 form, the likelihood may be evaluated incorrectly for half of the q-u plane.","section":"Footnote 8"},{"comment":"The claim of 'marginal but consistent increases in the overall measured PD for epochs where the EVPA rotation is identified' is not consistent across the three observations: Observation 2 shows no rotation and has the highest PD. Please qualify the statement to refer only to the rotating epochs or energy bins where the increase is actually observed.","section":"Abstract and §4"},{"comment":"In the selected post-dip window, the constant-EVPA model is disfavored relative to the unpolarized model (lnB = -5.2 in Table 7). This is an unusual intermediate state and deserves a brief comment, since it means the only polarized evidence in that window comes from the rotating model.","section":"§3.3, Table 6"}],"recommendation":"major_revision","confidential_remarks":"The paper is a method-plus-application manuscript. The methodological framework is standard nested sampling applied to a known event-level likelihood, so the novelty is incremental but potentially useful. The headline astrophysical result—slab vs shell coronal geometry—is currently the least supported part of the paper; it depends on internal equation consistency, a post-hoc time window, and a visual model comparison. I believe the issues are fixable within the manuscript's scope, but the authors should be asked to either provide quantitative support for the geometry claim or substantially reframe it as speculative. I would not recommend rejection if the major comments are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: QUEEN-BEE is a sensible and potentially valuable addition to the IXPE analysis toolkit, and the GX 13+1 results are worth being aware of. But the headline claim that the energy-resolved analysis excludes a shell corona is not as secure as the abstract suggests.\n\nThe genuinely new piece is the combination of an event-based Stokes likelihood with nested sampling for model comparison. That is a natural extension of Marshall's approach, and it gives the community a way to test constant vs. rotating vs. unpolarized models without arbitrary binning. The method recovers the known 42 deg/day rotation in Obs 1, finds no rotation in Obs 2, and picks up a likely transient rotation after a dip in Obs 3. The code is on Zenodo, and the comparison with PCUBE is clearly presented. For that, the paper deserves credit.\n\nThe soft spots are real but not fatal. First, Eqs. 9 and 10 define q(t) and u(t) with 2ψ0 + ωt, while Eq. 11 is the likelihood for ψ(t) = ψ0 + ωt. That is a factor-of-two inconsistency in the rotating model. The code likely uses Eq. 11, but the paper needs to say so in clear terms. Second, the Obs 3 transient is identified by selecting a specific post-dip time bin without any trials correction. The Bayes factor in that bin is large, but the look-elsewhere effect is not addressed. Third, the slab-vs-shell conclusion rides on ~0.9% difference in PD between two energy bands, under the assumption that the EVPA rotation is exactly linear. If the true rotation is even mildly nonlinear, the recovered PDs shift, and the comparison with fixed MONK curves (with no systematic error) could easily flip. The paper acknowledges the model dependence but does not quantify how much nonlinearity would change the answer.\n\nWho is this for? People working on IXPE polarimetry, especially EVPA variability and coronal geometry in X-ray binaries. It will also be useful as a reference for Bayesian model comparison in low-count polarimetric settings. With the equation inconsistency fixed, the trials issue addressed, and the geometry claim appropriately softened, I would be comfortable citing it or sending it to a journal with confidence. As it stands, it deserves peer review, but the authors should not get a free pass on those three points.\n\nMy recommendation: send it to a serious referee, and the referee should ask for the factor-of-two fix and a more cautious discussion of the geometry conclusion.","headline":"Useful new method and real rotation detections, but the slab-vs-shell claim is shakier than the abstract lets on, and the rotating-model equations have a factor-of-two slip that needs fixing.","tokens_in":20333,"tokens_out":2953,"would_cite":true,"duration_ms":31450,"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 Bayesian event-by-event sampler detects X-ray polarization-angle rotation that binned methods miss, and its energy-resolved results favor a slab-like corona over a shell for GX 13+1.","keywords":["X-ray polarimetry","EVPA rotation","Bayesian nested sampling","GX 13+1","neutron star low-mass X-ray binaries","coronal geometry","IXPE","Stokes parameters"],"falsifier":"Simulate IXPE-like photon events with a known nonlinear EVPA evolution, such as a sudden 70-degree step or a sinusoidal swing, and run QUEEN-BEE on them: if the framework returns a confidently nonzero linear rotation rate with an inflated polarization degree, the method's central detection claim fails. For the coronal conclusion, a future IXPE observation of GX 13+1 with better statistics, or a third energy bin, that shows polarization degree falling with energy when rotation is modeled would exclude the slab geometry the paper favors.","tokens_in":1940,"feed_emoji":"🐝","tokens_out":2028,"duration_ms":70459,"temperature":0.7,"pith_summary":"The paper introduces QUEEN-BEE, a Bayesian nested-sampling framework that models IXPE X-ray polarimetry event-by-event instead of in time or energy bins, and uses it to ask whether the polarization angle of the neutron-star binary GX 13+1 rotates and what that implies for the corona's shape. Applied to three IXPE observations, it finds decisive evidence for a steadily rotating polarization angle in the first (42±4 degrees/day), none in the second, and a fast transient rotation of about 170 degrees/day in the third right after a light-curve dip. The payoff is physical: in energy-resolved analysis of the first observation, the recovered polarization degrees favor a slab-like corona over a shell-like corona, a distinction the earlier binned analysis could not make. The broader claim is that unbinned Bayesian model comparison can recover rotation signals in regimes where binning leaves polarization below detection thresholds, making QUEEN-BEE a general diagnostic for sources with smoothly rotating polarization angles.","feed_headline":"New sampler finds GX 13+1's corona is a slab, not a shell","feed_subtitle":"An event-by-event Bayesian analysis recovers polarization-angle rotation that binned methods miss, tipping a geometry debate.","key_machinery":"The central object is an event-by-event likelihood over photoelectron azimuthal angles, in which the polarization angle is parametrized three ways: constant, linearly rotating with ψ(t)=ψ0+ωt, or absent (unpolarized). Nested sampling computes the Bayesian evidence for each model from the unbinned event list, and pairwise Bayes factors read against the modified Jeffreys scale select among them. The rotation rate ω, in degrees per day, carries the physical signal, and the use of evidence for model comparison rather than binned point estimates is what lets the method detect rotation in low-signal regimes where binning destroys it.","core_discovery":"QUEEN-BEE models the azimuthal scattering angle of every detected photoelectron with a likelihood that allows the EVPA to rotate linearly in time, ψ(t)=ψ0+ωt, and uses nested sampling to compute Bayesian evidence for three competing hypotheses: constant EVPA, linearly rotating EVPA, and unpolarized emission. On the first IXPE observation of GX 13+1 the rotating model is decisively favored in the full 2–8 keV band and in both the 2–4 and 4–8 keV sub-bands, with rotation rates of 37±6 and 48+6/−7 degrees/day; the same analysis of the third observation isolates a 170+20/−40 degrees/day rotation in a roughly 10-hour post-dip window where binned PCUBE analysis reports only about 1σ polarization.","pith_inferences":["A natural calibration test, not run in the paper, is to feed QUEEN-BEE simulated IXPE events with known nonlinear EVPA evolution (a step change or a sinusoid) and check whether it reports a confidently nonzero linear rotation rate with inflated polarization degree; the paper's own model-dependence caveat suggests this is the missing validation.","If the slab conclusion holds, a coordinated spectroscopy campaign during a rotation epoch should show correlated changes in wind diagnostics such as blueshifted absorption and column density only if wind scattering drives the rotation; the absence of such correlation would favor a purely geometric origin such as spin-axis misalignment.","The marginal polarization-degree increases QUEEN-BEE reports for rotating epochs imply that other IXPE sources with known EVPA rotation, such as Cir X-1 and Sco X-1, may have their intrinsic polarization underestimated by binned analyses; re-running those datasets through the same framework should lift their measured degrees similarly if the effect is real.","The very large Bayes factors in the full-band fits partly reflect enormous event counts, so cross-source comparisons of evidence values should be size-normalized rather than read as absolute model probabilities."],"forward_implications":["If correct, QUEEN-BEE recovers rotation rates consistent with previous binned analyses when rotation is present and matches PCUBE results when the EVPA is constant, so it can serve as a direct cross-check for IXPE polarization-timing studies.","The energy-resolved result for the first GX 13+1 observation would shift the coronal interpretation from 'shell or slab unresolved' to 'slab favored,' lending support to the western spectral model of the source.","The framework's evidence-based comparison provides a way to decide statistically between constant, rotating, and unpolarized models without committing to bin sizes in advance.","Rotation rates that change across epochs (about 42 degrees/day versus about 170 degrees/day) indicate the rotation mechanism is transient, tied to changing geometry or episodic wind obscuration rather than a fixed, steady precession.","Because the method works directly on unbinned events, it extends to other IXPE targets and to future instruments with smoothly rotating EVPA signals, including proposed soft X-ray and enhanced timing-polarimetry missions."],"supporting_citations":[{"why":"Supplies the prior binned discovery of roughly 70-degree EVPA rotation in the first GX 13+1 observation, which QUEEN-BEE re-derives as 42±4 degrees/day.","marker":"A. Bobrikova et al. 2024b"},{"why":"Provides the PCUBE binned Stokes-parameter algorithm that serves as the baseline comparison method throughout the paper.","marker":"F. Kislat et al. 2015"},{"why":"Demonstrates the event-based maximum-likelihood approach for Mrk 421 that QUEEN-BEE extends with Bayesian model comparison.","marker":"L. Di Gesu et al. 2023"},{"why":"Supplies the weighted log-likelihood formalism for IXPE data on which QUEEN-BEE's simplified direct likelihood is based.","marker":"H. L. Marshall 2021a,b, 2024"},{"why":"Provides dynesty, the nested sampler used to compute posterior samples and Bayesian evidences.","marker":"J. S. Speagle 2020"},{"why":"Provides bilby, the Bayesian inference environment that QUEEN-BEE's implementation is structured around.","marker":"G. Ashton et al. 2019"},{"why":"Supplies the MONK-based Monte Carlo predictions of polarization degree versus energy for slab and shell coronal geometries used for the geometry conclusion.","marker":"A. Gnarini et al. 2022"},{"why":"Provides the MONK radiative-transfer code whose slab and shell simulations underlie the geometry comparison.","marker":"W. Zhang et al. 2019"},{"why":"Analyzes the third GX 13+1 observation's light-curve dips and EVPA swings, which QUEEN-BEE's time-resolved post-dip result extends.","marker":"A. Di Marco et al. 2025"}],"fun_headline_variants":["GX 13+1's corona is a slab, says Bayesian polarization sampler","Bayesian event-by-event analysis settles GX 13+1 corona shape","QUEEN-BEE finds slab-like corona around neutron star GX 13+1","Polarization rotation in GX 13+1 points to slab corona","New Bayesian method detects hidden polarization swings in GX 13+1"],"cache_read_input_tokens":22016,"weakest_assumption_plain":"The analysis assumes that within each time or energy segment the polarization angle rotates at a single constant rate; if the true angle evolution is not linear, the recovered polarization degree and rotation rate are biased, and the slab-versus-shell conclusion built on those values would not hold.","fun_headline_variants_meta":{"raw":{"variants":["GX 13+1's corona is a slab, says Bayesian polarization sampler","Bayesian event-by-event analysis settles GX 13+1 corona shape","QUEEN-BEE finds slab-like corona around neutron star GX 13+1","Polarization rotation in GX 13+1 points to slab corona","New Bayesian method detects hidden polarization swings in GX 13+1"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000641,"raw_usage":{"total_tokens":2873,"prompt_tokens":913,"completion_tokens":1960,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":657,"completion_tokens_details":{"reasoning_tokens":1871}},"tokens_in":657,"tokens_out":1960,"duration_ms":12272,"temperature":1.0,"reasoning_tokens":1871,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T22:51:01.203260+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate IXPE-like photon events with a known nonlinear EVPA evolution, such as a sudden 70-degree step or a sinusoidal swing, and run QUEEN-BEE on them: if the framework returns a confidently nonzero linear rotation rate with an inflated polarization degree, the method's central detection claim fails. For the coronal conclusion, a future IXPE observation of GX 13+1 with better statistics, or a third energy bin, that shows polarization degree falling with energy when rotation is modeled would exclude the slab geometry the paper favors.","supporting_citations":[],"review_version":1}