{"id":"c2df4ebf-7b41-4820-b6f5-dc00b3bca858","arxiv_id":"2608.04108","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Reconstructing PSR J1713+0747's pulse profile from the two temporally stable central Gaussian components yields a phase-connected timing solution across the 2021 profile-change event with 4.454 microseconds residuals.","lead":"This paper reconstructs radio pulse profiles of a key pulsar using only the two stable central emission components, and shows that timing across a known profile-change event remains phase-connected. It offers pulsar timing arrays a way to keep timing precision when pulse shapes change, using low-frequency uGMRT data.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prior-constrained stability and self-referential validation leave the central timing claim untested across the profile-change event.","rationale":"I read the paper as a methods demonstration: use physically motivated Gaussian decomposition to identify stable components and time on them. The method is plausible and has some genuine support: the pre-event reconstruction matches original-profile timing at the few-microsecond level, and the recovered annual solar-wind DM variations indicate that chromatic information survives reconstruction. Those are real, independent checks. The soft spot is that the post-event timing claim cannot be falsified within the paper. Stability is claimed for components whose phase priors already confine them to a narrow interval; reconstruction uses those same components; and the only post-event timing validation is against the reconstructed data themselves. The 4.454 microsecond WRMS and the non-detection of the exponential dip are internal-consistency results, not external evidence that the selected components track absolute rotational phase. The reader's weakest assumption (transfer of EPN high-frequency priors to 300-500 MHz) is a related but slightly different failure mode; the more immediate issue is that even granting the priors, the paper does not provide an independent post-event phase reference. A cross-check against high-frequency PTA ToAs would settle it. Because the method could still be correct and useful, and the missing check is obtainable, I keep the CONDITIONAL verdict rather than rejecting the paper.","tokens_in":17840,"tokens_out":6152,"duration_ms":57537,"concrete_test":"Take the post-event epochs (MJD > 59321) and compare ToAs from the reconstructed profiles with contemporaneous ToAs from an independent high-frequency dataset (e.g., NANOGrav 15/20-yr, EPTA DR2full, or InPTA Band-5), using the same timing model. Form per-epoch ToA differences, weighted by their uncertainties, and test consistency with zero; also split by frequency sub-band to search for a frequency-dependent trend. A nonzero mean offset or a chromatic trend would falsify the claim that G2+G3 preserves unbiased rotational phase, while agreement within the combined ToA uncertainties would support it.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that G2 and G3 are intrinsically stable rotational-phase references whose reconstruction yields unbiased timing across the 2021 profile-change event—rests on a validation loop. Section 5 establishes stability from the same decomposition whose priors (Table 1) restrict phi2 and phi3 to a shared U(0.47,0.52) interval at every epoch. Any event-induced shift larger than this prior support cannot appear as a centroid shift; it is absorbed into the free amplitudes/widths or into G2/G3 interchange. The Section 6 reconstruction uses only these prior-constrained components, and the Section 8 phase-connected 4.454 microsecond solution is evaluated only against that reconstruction. A phase-connected solution across the event is also attainable for any smoothly varying phase curve by fitting F0, so the WRMS measures scatter, not absolute phase accuracy. The external checks—pre-event comparison (Fig. 8) and solar-wind DM recovery (Fig. 7)—do not cover the post-event interval and test chromatic delay, not rotational phase. If the apparent stability is a prior artifact, the reconstructed ToAs carry an epoch- and frequency-dependent bias that the current analysis cannot detect.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a Bayesian five-Gaussian decomposition of frequency-resolved uGMRT Band-3 (300-500 MHz) pulse profiles of PSR J1713+0747, with phase priors informed by EPN polarization profiles at higher frequencies. Tracking the centroids of the five components across the observing baseline, the authors identify the two central components G2 and G3 as being phase-stable across the 2021 profile-change event, reconstruct frequency-resolved profiles using only these components while adding realistic noise, and then estimate epoch-wise dispersion measures and perform conventional tempo2 timing. The central claim is that profiles reconstructed from G2 and G3 preserve the rotational phase information and yield a phase-connected timing solution across the event, with a post-fit WRMS residual of 4.454 microseconds, thereby demonstrating a physically motivated mitigation of pulse-profile variability.","tokens_in":18071,"tokens_out":4532,"duration_ms":44117,"significance":"If the central claim is validated, this is a useful contribution to PTA data-analysis methodology: it addresses the well-known J1713+0747 profile-change event at low radio frequencies, where chromatic effects are severe, and it makes explicit use of physically motivated priors from polarization data. The recovery of the expected annual DM variations near solar conjunction and the pre-event comparison between original and reconstructed ToAs are constructive validation steps. The paper is also clearly written and the decomposition is carefully described. However, the key validation is largely self-referential: the stability of G2 and G3 is assessed with the same priors that constrain them to a narrow phase interval, and the post-event timing solution is evaluated only against profiles reconstructed from those very components. The external checks cover only the pre-event interval or test chromatic delay rather than absolute rotational phase. These issues are addressable, but they are load-bearing for the main claim.","major_comments":[{"comment":"The claim that G2 and G3 are intrinsically phase-stable is partly an artifact of the priors. The centroid priors for phi2 and phi3 are both U(0.47,0.52), a support of only 0.05 in pulse phase, and the constraint phi1 < phi2,phi3 < phi4 < phi5 is imposed. Any event-induced shift larger than this support cannot appear as a centroid shift; it will be absorbed into the amplitudes and widths or into an interchange between G2 and G3. The paper should demonstrate that the posterior widths for phi2 and phi3 are substantially narrower than the prior widths, and should test sensitivity to a wider prior (for example U(0.44,0.55) with the ordering constraint only). Without such a test, the stability shown in Fig. 5 does not establish that these components are an invariant rotational-phase reference.","section":"Section 5 and Table 1"},{"comment":"The timing validation is circular in a way that matters. The post-event phase-connected solution is obtained exclusively from reconstructed profiles built from G2 and G3, the same components selected for stability in Section 5. Fitting only F0 to a smoothly varying phase curve will always produce a phase-connected solution, so the 4.454 microsecond WRMS measures scatter rather than absolute phase accuracy. The pre-event comparison in Fig. 8 does not cover the post-event interval, and the recovered DM variations in Fig. 7 test the chromatic delay, not the rotational-phase reference. The absence of an exponential-dip signature is likewise expected if the dip is carried by the outer components that were removed by construction, so it is not an independent confirmation. To support the central claim, the authors should compare the reconstructed post-event ToAs against an independent phase reference, such as simultaneous Band-5 or L-band observations, or perform an injection-recovery test with a simulated event-induced phase step.","section":"Section 8 and Figure 9"},{"comment":"The priors are transferred from EPN polarization profiles at 728-1369 MHz to the 300-500 MHz Band-3 data. The paper itself notes that radius-to-frequency mapping shifts the apparent locations of components with frequency, and that one leading feature is displaced at low frequency. This raises the risk that the physical emission regions associated with G2 and G3 at 300-500 MHz are not the same regions identified at higher frequencies. The eight sub-bands within Band-3 span 200 MHz, so the authors should quantify how the posterior centroids of G2 and G3 evolve across sub-bands. If the centroids drift systematically with frequency, the reconstructed template carries a chromatic phase offset that the subsequent DM fit cannot absorb, and the timing solution would inherit a frequency-dependent bias.","section":"Section 3 and Table 1"},{"comment":"The pre-event comparison shows a substantial degradation in timing precision: the WRMS increases from 2.505 microseconds for the original profiles to 3.707 microseconds for the reconstructed profiles, a factor of about 1.5. The paper presents this as a successful validation, but for a method intended to improve precision timing it should discuss whether this loss is acceptable and what fraction of the S/N is removed when G1, G4, and G5 are discarded. In addition, the comparison is not strictly like-for-like because the original and reconstructed analyses use independent sub-band flagging and quality filtering, so the WRMS difference may partly reflect different ToA sets rather than the reconstruction alone. A quantitative statement of the median ToA offset and its scatter between the two analyses would be more informative.","section":"Section 8 and Figure 8"}],"minor_comments":[{"comment":"The sentence 'As an example of our successful bayesian gaussian decomposition of a post-event epoch (MJD=59551) together with its associated AD statistic (see Fig.3) - the corresponding p-value is 0.528' is grammatically incomplete; the main clause is missing.","section":"Section 4.4"},{"comment":"The DM time series is shown without error bars, so the significance of the annual solar-wind variations cannot be assessed. Please include uncertainties on the DM points or state that they are smaller than the plotted markers.","section":"Section 7 and Figure 7"},{"comment":"The data set mixes profiles with N_bin=128, 256, and 512 phase bins. The paper does not discuss whether the different phase resolutions affect the Gaussian decomposition or the reconstructed-profile timing, particularly for the narrow G3 component whose width prior extends down to sigma=0.002.","section":"Section 2"},{"comment":"The claim that 'no discernible systematic offsets' exist between original and reconstructed ToAs is only qualitative. A quantitative offset estimate with an uncertainty would strengthen the pre-event validation.","section":"Section 8"},{"comment":"The related work by Nichols et al. (2026) on Gaussian-component modeling of the same profile-change event is cited only once. Given the direct overlap in methodology and target, a more explicit comparison of the low-frequency approach with that high-frequency analysis would help the reader understand the novel contribution.","section":"Section 1 and Discussion"}],"recommendation":"major_revision","confidential_remarks":"The central concern is the self-referential validation chain: the stability of G2/G3, the construction of the reconstructed profiles, and the timing validation all derive from the same five-Gaussian model with narrow phase priors. I would encourage the editor to request an independent post-event phase comparison (e.g., simultaneous Band-5/L-band timing or a synthetic injection test) before accepting. The paper is otherwise well organized and the data set is valuable, so the issues appear fixable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper does something new and mostly useful — frequency-resolved Gaussian decomposition of uGMRT Band-3 profiles, reconstruction from the two stable central components, and a phase-connected timing solution across the 2021 profile-change event — but its central validation is partly circular. The method deserves a serious referee, with the expectation that the authors add an independent post-event check and release code/data.\n\nWhat's genuinely new: the 300–500 MHz demonstration that profiles reconstructed from G2 and G3 preserve phase coherence across the event. The frequency-resolved treatment is what lets them recover the expected annual DM variations from the solar wind, which is a good sanity check that the chromatic information survives reconstruction. The pre-event comparison (original 2.505 µs vs reconstructed 3.707 µs WRMS) shows a degradation but no obvious systematic offset, which is reassuring.\n\nThe soft spots are real but not fatal. The stability of G2/G3 is measured from the same five-Gaussian decomposition that is later used for reconstruction, and the phase priors in Table 1 restrict φ2 and φ3 to the same U(0.47,0.52) range at every epoch. That range is actually broad relative to the component widths — 0.05 in phase is a lot for a narrow central component — so the circularity is not as severe as a first pass might suggest. Still, stability is only demonstrated within the prior support, and the exponential-dip check is expected to vanish once you remove the varying outer components; it is not an independent confirmation.\n\nThe bigger gap is that there is no post-event comparison against an external reference. A phase-connected solution across the event can be obtained by fitting F0 even if the reconstructed ToAs carry a smooth epoch-dependent bias; the 4.454 µs WRMS measures scatter, not absolute phase accuracy. The physical assumption that the high-frequency EPN polarization features (728–1369 MHz) persist at 300–500 MHz is also the weakest link, and the paper's morphological alignment is suggestive rather than conclusive.\n\nThis is a worthwhile paper for PTA timing people and anyone working on profile variability. It is clearly written, honestly motivated, and the data analysis is reproducible in principle — but no code or data are currently provided, which is a checkpoint-level issue. I would send it to peer review and ask for: (1) an external validation of the reconstructed ToAs across the event, e.g., comparison with high-frequency EPTA/PPTA/NANOGrav ToAs or a simulated profile-change injection; (2) a statement of how many epochs/sub-bands were excluded; (3) public code and reconstructed archives.","headline":"Low-frequency profile-domain timing for J1713+0747 is promising but its validation loop needs an external post-event check before I'd trust it in PTA pipelines.","tokens_in":18740,"tokens_out":4199,"would_cite":true,"duration_ms":37350,"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":"The two central Gaussian components of PSR J1713+0747's low-frequency pulse profile stay phase-stable across the 2021 profile-change event, and timing on profiles rebuilt from only those components stays phase-connected at 4.454…","keywords":["millisecond pulsars","pulse profile variability","Bayesian inference","Gaussian decomposition","profile-domain timing","PSR J1713+0747","dispersion measure","pulsar timing arrays"],"falsifier":"Take a single high-signal-to-noise post-event epoch, fit the five-component model with the central-component phase priors removed or widened far beyond the adopted ranges, and check whether the recovered G2/G3 centroids remain inside the original prior range at all eight sub-bands; if they drift out of that range, the reported stability is an artifact of the priors. A stronger test is to compare the reconstructed low-frequency timing solution epoch-by-epoch against simultaneous high-frequency (around 1.4 GHz) timing across the profile-change event: any frequency-dependent jump or drift in the residuals would show that the stable components are not a rotational-phase invariant.","tokens_in":17667,"feed_emoji":"⏱️","tokens_out":9406,"duration_ms":67886,"temperature":0.7,"pith_summary":"The paper argues that even when a pulsar's overall pulse shape changes dramatically, the rotational phase information needed for precision timing survives in a subset of the emission components. For the millisecond pulsar PSR J1713+0747, which underwent a sudden, persistent profile change in April 2021, the authors decompose low-frequency (300–500 MHz) pulse profiles into five Gaussian components, each tied to a physical emission region via priors from high-frequency polarization data. Two central components remain localized in phase across the whole observing baseline, so the authors rebuild each profile using only those two components. Timing on the reconstructed profiles stays phase-connected across the event, with a post-fit weighted RMS residual of 4.454 microseconds, and recovers the expected annual dispersion-measure variations. If this holds, pulsar timing arrays can keep using data from pulsars whose profiles change, instead of discarding post-event observations.","feed_headline":"Two stable pulse components preserve timing across profile change","feed_subtitle":"Timing from only the two stable central components stays phase-connected across the 2021 event at 4.454 microseconds.","key_machinery":"Five Gaussian components fitted to each frequency-resolved (eight sub-bands) pulse profile via Bayesian inference, with informative but flexible phase priors derived from high-frequency polarization morphology. The load-bearing assumption is the component identity across frequency and time: the two central components G2 and G3 are allowed overlapping phase priors because they occupy nearly the same phase range at low frequency, yet their amplitudes and widths are tracked independently. The reconstruction step then discards the unstable outer components and writes only G2+G3 back into the data archives, preserving the metadata and noise characteristics, so that standard cross-correlation timing can proceed.","core_discovery":"The central claim is that the centroid phases of the two central Gaussian components, G2 and G3, are stable to within their posterior uncertainties across the October 2019–September 2025 baseline, including across the 2021 profile-change event, while the outer components G1, G4, and G5 wander more widely. Because the pulse profile is modeled as a sum of components anchored to persistent emission regions, the phase of the stable central pair can serve as a rotational-phase reference. Reconstructing each frequency sub-band profile as just G2+G3, with realistic noise restored, and carrying that reconstruction through standard dispersion-measure estimation and timing, produces a phase-connected timing solution with a post-fit WRMS residual of 4.454 microseconds. The paper presents this as a physically motivated route to precision timing for pulsars with evolving profiles.","pith_inferences":["If the central components trace the core emission close to the stellar surface, this method effectively turns pulsar timing into core-component timing, which may be inherently more robust to magnetospheric reconfigurations that mostly alter conal emission.","A natural extension is to fit the Gaussian decomposition and the dispersion measure simultaneously rather than sequentially, which could remove any residual covariance between chromatic delays and component centroid shifts.","Applying the same reconstruction to simultaneous high-frequency (around 1.4 GHz) data and checking cross-band consistency would test whether the selected components are true rotational-phase invariants, and would generalize the method to broadband pulsar-timing-array use.","The achieved 4.454 microsecond residual, compared with 2.505 microseconds pre-event from the full profile, suggests a modest precision cost for using only two components; the trade-off between stability and lost signal-to-noise could be quantified for other pulsars."],"forward_implications":["Post-event observations of PSR J1713+0747 can be kept in pulsar-timing-array datasets rather than flagged, because the reconstructed profiles yield a phase-connected solution across the 2021 event.","Epoch-wise dispersion measures, including the annual solar-wind signature, are recoverable from the reconstructed profiles, so chromatic propagation information survives the reconstruction.","The timing signature of the profile-change event, an exponential dip seen in the original time-of-arrival residuals, disappears from the reconstructed residuals, indicating that the reconstruction removes the profile-change timing bias.","The methodology transfers to other pulsars with abrupt or gradual profile variability, as long as polarization data exist to define the component priors."],"supporting_citations":[{"why":"Provides the high-frequency polarization profiles whose stable features seed the phase priors for the five Gaussian components.","marker":"Dai et al., 2015"},{"why":"Reports the 2021 profile-change event and defines the exponential-dip signature that the reconstructed time-of-arrivals are checked against.","marker":"Singha et al., 2021"},{"why":"Establishes that the profile change is broadband and intrinsic, motivating reconstruction from intrinsic emission components rather than propagation effects.","marker":"Jennings et al., 2024"},{"why":"Demonstrates Gaussian-component timing across the same event at higher frequencies; the present low-frequency reconstruction extends that approach.","marker":"Nichols et al., 2026"},{"why":"Supplies the cross-correlation-based dispersion-measure estimation procedure used to turn reconstructed sub-band profiles into epoch-wise dispersion measures.","marker":"Krishnakumar et al., 2021"},{"why":"Provides the timing model and fit machinery used to obtain the phase-connected solution and the 4.454 microsecond residual.","marker":"Edwards et al., 2006"},{"why":"Supplies the nested-sampling algorithm used to sample the Gaussian component posteriors.","marker":"Feroz et al., 2009; Buchner et al., 2014"},{"why":"Provides the fiducial timing solution used to refold and de-disperse the data before decomposition.","marker":"EPTA Collaboration et al., 2023"}],"fun_headline_variants":["Stable pulse components preserve timing across profile changes","Two stable components anchor timing amid profile variability","Reconstructing profiles from stable components gives robust timing","Stable emission components keep timing phase-connected through change","Profile-stable components beat variability for pulsar timing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The five Gaussian components fitted at 300–500 MHz correspond to the same persistent emission regions identified in the high-frequency polarization profiles, so the phase priors carry that identity into the low-frequency data; if radius-to-frequency mapping changes which physical region dominates the central components at low frequency, the 'stable' reference is not a true rotational-phase invariant.","fun_headline_variants_meta":{"raw":{"variants":["Stable pulse components preserve timing across profile changes","Two stable components anchor timing amid profile variability","Reconstructing profiles from stable components gives robust timing","Stable emission components keep timing phase-connected through change","Profile-stable components beat variability for pulsar timing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000288,"raw_usage":{"total_tokens":1666,"prompt_tokens":902,"completion_tokens":764,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":691}},"tokens_in":518,"tokens_out":764,"duration_ms":7142,"temperature":1.0,"reasoning_tokens":691,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:44:04.365611+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a single high-signal-to-noise post-event epoch, fit the five-component model with the central-component phase priors removed or widened far beyond the adopted ranges, and check whether the recovered G2/G3 centroids remain inside the original prior range at all eight sub-bands; if they drift out of that range, the reported stability is an artifact of the priors. A stronger test is to compare the reconstructed low-frequency timing solution epoch-by-epoch against simultaneous high-frequency (around 1.4 GHz) timing across the profile-change event: any frequency-dependent jump or drift in the residuals would show that the stable components are not a rotational-phase invariant.","supporting_citations":[],"review_version":2}