{"id":"a6a46835-814a-4b6e-adb3-b0c08fc33da8","arxiv_id":"2507.18437","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"QPO lags between disc and coronal X-ray bands in MAXI J1820+070 vary with hard X-ray flux, and disc-corona coherence drops at and below the QPO frequency, an effect seen in high-inclination but not low-inclination sources.","lead":"This paper shows that the time delay between X-rays from the disc and the corona in a black hole binary changes on timescales of tens of seconds, tied to the quasi-periodic oscillation (QPO) that the source shows. The behavior appears only in high-inclination systems, which may reveal how the corona changes shape.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Flux-binning selection and two-component mixing are not quantitatively ruled out in §4.1; a simulation is needed before the variable-lag claim can be accepted.","rationale":"The reader's weakest assumption identifies exactly the same soft spot: the flux-binning analysis in §3.1 can in principle create the observed lag-flux relation through selection effects or through a changing mixture of spectral components, and §4.1 does not quantitatively exclude this. I read the paper in good faith and credit its real strengths: the analysis uses public NICER data, the methods are standard spectral-timing tools, the coherence drop at and below the QPO frequency is a striking and less model-dependent result, and the cross-source comparison with MAXI J1803-298 and GX 339-4 is a valuable consistency check. None of those, however, substitute for the missing quantitative test of the flux-binning step. The central claim is not just that lags differ between flux bins; it is that the intrinsic disc-corona QPO lag changes on timescales of a few QPO cycles. That inference requires the null hypothesis of constant intrinsic lags plus mixing/selection to be shown incapable of producing the observed >0.5 rad shift. The spectral fit in Fig. 9 shows only that time-averaged spectra differ by ~10%, which does not bound the phase shift of a two-component cross-spectrum, because the phase shift depends on the relative variability amplitudes of the disc and power-law components at the QPO frequency, not on their time-averaged fluxes alone. The lack of a simulation is therefore not a cosmetic omission; it is the difference between a detection and a possible artifact. I would not reject the paper, because the coherence result and the cross-source phenomenology are likely to stand on their own, and the variable-lag claim may well survive a simulation. But the appropriate verdict is conditional acceptance pending that specific check, with the authors invited to either add the simulation or provide an analytic bound strong enough to exclude the mixing/selection null hypothesis.","tokens_in":24649,"tokens_out":7845,"duration_ms":92170,"concrete_test":"Run a Monte Carlo simulation using the public NICER light curves for ObsIDs 135-137. Generate 18-s segments with the observed mean count rates and rms-flux relation, a constant intrinsic QPO lag equal to the time-averaged value from Fig. 5, and Poisson noise; bin by 3-10 keV flux exactly as in §3.1 and recompute the QPO lag per bin. In the same simulation, give the soft band a two-component mixture with the ~10% flux-bin spectral variations from Fig. 9 and assign the disc component a π/2 phase lag with amplitude g spanning 0.1-5. If the simulated low-to-high flux-bin lag difference exceeds 0.5 rad in a substantial fraction of realizations for any g, the intrinsic-variability interpretation fails; if it stays below ~0.1 rad for all g values allowed by the observed soft-band power spectra, the §3.1 conclusion survives.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The load-bearing step is the flux-binning analysis in §3.1. The conclusion that the QPO lags are intrinsically variable requires that the >0.5 rad lag difference between the lowest and highest hard-flux bins is not produced by (a) the changing mixture of disc and power-law components in the soft bands induced by selecting on 3-10 keV flux, or (b) a statistical selection bias from using the reference band itself as the binning variable. Section 4.1 addresses (a) by spectral fitting and finds only ~10% changes in the diskbb/thcomp ratio, but it never computes the lag shift that such a ratio change would produce. For a two-component mixture, S = αD + βP, with D = g e^{iφ}P at the QPO frequency, the measured soft-hard phase is atan[αg sinφ / (β + αg cosφ)]; with α/β ~ 1 (the reported ~50% disc contribution in the very soft band), φ ~ π/2 and g of order a few, a 10% change in α/β can produce changes of order 0.1-0.3 rad, and the exact value depends on the disc variability amplitude at the QPO frequency, which is not reported. The spectral fit also uses a simplified model with a 5% systematic and does not propagate parameter uncertainties into a predicted lag. Effect (b) is entirely unaddressed: because the 3-10 keV band is both the flux selector and the reference band, and because the rms-flux relation makes high-flux segments have higher QPO power, the averaging within each flux bin has flux-dependent SNR; any residual bias in the phase estimator can masquerade as a lag-flux relation. A quantitative simulation or a closed-form bound is needed before the intrinsic-variability claim can be accepted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a spectral-timing analysis of NICER observations of the black hole X-ray binary MAXI J1820+070 in its 2018 hard state, focusing on type-C quasi-periodic oscillations (QPOs). The authors measure cross-spectral phase lags and coherence between soft energy bands (with significant disc emission) and a harder 3-10 keV band, and further study how the QPO lags depend on the instantaneous 3-10 keV flux by binning 18-s light-curve segments into four flux bins. They report that the QPO hard lag between disc-dominated and power-law-dominated bands anticorrelates with the hard flux, varying on timescales of a few QPO cycles, with the effect strongest below about 0.3 Hz. They also find that the intrinsic coherence between disc and power-law bands is strongly reduced at and below the QPO frequency, recovering above it, and that this pattern tracks the QPO frequency across the outburst. The same coherence and lag features are observed in the high-inclination source MAXI J1803-298 but not in the lower-inclination source GX 339-4. The authors interpret these results as evidence for variations in the vertical extent of the corona on timescales slightly longer than the QPO cycle.","tokens_in":24949,"tokens_out":12766,"duration_ms":143227,"significance":"If the results hold, this is a significant observational contribution to the study of QPOs and disc-corona coupling. The paper provides a direct, high-significance measurement that QPO lags are not stationary but vary on timescales of tens of seconds, and it reports a previously unrecognized link between the QPO frequency and a coherence drop at and below that frequency. It also offers a promising inclination dependence, with the effect present in two high-inclination sources and absent in one lower-inclination source, which is a falsifiable prediction for QPO models. The analysis uses standard cross-spectral methods, reports significance values above 5 sigma for the softest bands in multiple observation groups, and includes checks such as spectral fitting of flux bins and a comparison to a pre-QPO observation. The paper promises a reproduction package on Zenodo, which aids reproducibility. The main caveat, discussed below, is that the flux-binning analysis used for the variable-lag claim needs an explicit quantitative demonstration that selection effects and spectral mixing cannot produce the observed lag-flux relation.","major_comments":[{"comment":"The conclusion that the QPO lags are intrinsically variable rests on ruling out the possibility that the observed lag-flux relation is produced by flux-dependent changes in the spectral composition of the soft band. The authors show that the diskbb/thcomp ratio changes by only about 10% between flux bins, but they do not translate this change into a predicted phase-lag difference. For a two-component mixture in the soft band (S = αD + βP) measured against a power-law-dominated reference (P), the measured cross-spectral phase is arg(αg e^{iφ} + β), where g is the disc-to-power-law variability amplitude ratio at the QPO frequency and φ is the intrinsic disc-power-law phase lag. If φ ≈ π/2 and g ≈ 1, a 10% change in α/β changes this phase by only about 0.05 rad, far less than the observed >0.5 rad difference. The authors' qualitative conclusion is therefore likely correct, but the paper should include this calculation or a small simulation, and should propagate the spectral-fit uncertainties into the predicted lag range, rather than leaving the inference implicit.","section":"Section 4.1 and Fig. 9"},{"comment":"The flux-binning analysis uses the 3-10 keV count rate both as the variable that defines the flux bins and as the reference band for the cross-spectral phase measurements. Because of the rms-flux relation, the flux bins differ systematically in QPO amplitude and signal-to-noise, and the paper does not assess whether this selection can induce a spurious lag-flux correlation when the true QPO phase is constant. This is a load-bearing point for the claim that the QPO lags change intrinsically. The authors should test this explicitly, for example by simulating a stationary process with a constant QPO phase and an rms-flux relation, applying the same 18-s segment binning and phase estimation, and showing that the recovered lag-flux slope is consistent with zero at the level of the observed slopes. Without such a test, part of the reported lag-flux signal could in principle be an artifact of the binning procedure.","section":"Section 3.1 and Fig. 1"}],"minor_comments":[{"comment":"In the first paragraph of Section 2, the text states that the five energy bands are 'listed in Table 1', but the energy bands are actually listed in Table 2; this cross-reference should be corrected.","section":"Section 2, Table 2"},{"comment":"The caption states that the y-axis is scaled logarithmically, but the same panel shows phase lags that can be near zero or negative; please clarify how negative or zero values are displayed, or use a linear or symlog axis.","section":"Section 3.1, Fig. 1 caption"},{"comment":"The text says that the reported σ values represent the probability of obtaining the data if there were no relation; these are significance levels in Gaussian sigmas, not probabilities, and the wording should be adjusted to avoid confusion.","section":"Section 3.1, after Eq. (1)"},{"comment":"The term 'imaginary QPO' is introduced through a reference to Bellavita et al. (2025) but is not defined in the present paper; a brief definition or explanation of the cross-spectral feature would help the reader.","section":"Section 4.4, ObsID 187"}],"recommendation":"major_revision","confidential_remarks":"This is a strong observational paper with careful standard methodology and a potentially important new result. The coherence drop tracking the QPO frequency is robust and well supported by the data, including the pre-QPO control observation. The variable-lag claim is the most novel part, but it currently lacks a quantitative treatment of the two main selection/mixing effects. Adding a short analytical estimate or a light-curve simulation would substantially strengthen the paper. If the authors provide that, I would be happy to see the paper published in MNRAS."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline result is new and worth taking seriously: the QPO lag between disc-dominated and coronal bands in MAXI J1820+070 varies with hard flux on timescales of a few QPO cycles, and the disc–corona coherence drops at and below the QPO frequency, tracking the QPO as it evolves. The cross-source comparison with MAXI J1803-298 and GX 339-4 is a nice touch, and the contrast between high- and low-inclination sources gives the paper a lot of its weight.\n\nWhat the paper does well: the measurements are carefully made with standard cross-spectral methods, the softest bands show >5σ relations in multiple independent observation groups, and the coherence drop is shown to follow the QPO frequency across the outburst and to be absent in a pre-QPO observation. The paper is also honest about the limits of its interpretation, and the discussion of coronal geometry is appropriately speculative rather than overclaimed.\n\nThe soft spot is exactly what the stress-test note identifies. Section 4.1 fits the spectra of the flux bins and finds only ~10% changes in the disc-to-corona ratio, but it never computes what lag shift such a mixing change would produce. Given the two-component mixture and the unknown disc variability amplitude at the QPO frequency, a 10% ratio change could plausibly produce a non-negligible fraction of the observed >0.5 rad effect. More worrying, the paper never addresses the selection bias from using the 3-10 keV band both as the flux binning variable and as the reference band in the lag calculation. The rms–flux relation means high-flux segments have different QPO power and signal-to-noise, and any residual phase-estimator bias could masquerade as a lag–flux relation. This is the load-bearing step for the claim of intrinsic lag variability, and it needs a quantitative simulation or a closed-form bound.\n\nThat said, the coherence drop is a separate, robust result that does not depend on flux binning, and it stands regardless of the lag–flux issue. The overall picture is probably right, but the central claim needs one more piece of analysis before I would call it bulletproof.\n\nWho this is for: X-ray timing people and QPO modelers. It deserves a serious referee. I would recommend accept with major revision, asking for the flux-binning simulation and the public code (the Zenodo link is promised but not yet live). That is a feasible request, and the paper is strong enough to warrant it.","headline":"A solid and genuinely new spectral-timing result on MAXI J1820+070, with the flux-binning caveat being real but addressable; deserves peer review and a requested simulation.","tokens_in":25534,"tokens_out":2148,"would_cite":true,"duration_ms":23503,"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":"This paper establishes that QPO lags between the disc and corona in MAXI J1820+070 vary with coronal flux on timescales of tens of seconds, and that the QPO suppresses disc-corona coherence at and below its frequency.","keywords":["quasi-periodic oscillations","X-ray timing","coherence","disc-corona lags","MAXI J1820+070","black hole X-ray binaries","coronal geometry","NICER"],"falsifier":"Simulate light curves with constant intrinsic QPO lags and the measured ~10% spectral component changes, apply the same 18-second flux binning and third-bin lag estimate, and check whether a >0.5 rad lag difference between flux bins emerges; if it does, the intrinsic-lag conclusion fails.","tokens_in":24438,"feed_emoji":"🕳️","tokens_out":8165,"duration_ms":76456,"temperature":0.7,"pith_summary":"This paper uses NICER observations of the black hole X-ray binary MAXI J1820+070 to test whether the time delay between soft disc photons and harder coronal photons at the quasi-periodic oscillation (QPO) frequency is a fixed property. It finds that the QPO lag between a soft band with significant disc emission and a hard coronal power-law band varies strongly with the instantaneous 3-10 keV flux, changing by more than half a radian on timescales of tens of seconds, and that the variation is largest at QPO frequencies below about 0.3 Hz. The same observations show that coherence between disc and coronal bands drops sharply at and below the QPO frequency, even though it stays high above it. The paper argues this points to a corona whose geometry changes on timescales a few times longer than the QPO cycle, rather than a static reprocessing geometry.","feed_headline":"Black hole QPO lags swing with coronal flux within seconds","feed_subtitle":"Disc-corona lags change by up to a radian as the X-ray flux varies over tens of seconds","key_machinery":"The central object is the frequency-dependent cross-spectrum between five NICER energy bands. The authors define five bands from 0.3-10 keV, with the softest two dominated by disc emission and the hardest by coronal power-law emission. Using 18-second segments, about three QPO cycles long, they sort the data into four hard-flux bins and then measure phase lags and intrinsic coherence in the third Fourier bin, which sits on the QPO fundamental. The coherence function, which measures whether variability in two bands is linearly related, is the quantity that reveals the QPO-linked drop, and the lag-versus-flux slope in radians per normalised count rate quantifies how much the lag itself changes.","core_discovery":"The central claim is that the QPO hard lag between disc-dominated soft X-rays and coronal power-law X-rays in MAXI J1820+070 is intrinsically variable, anticorrelated with the hard (3-10 keV) flux, and that the QPO mechanism suppresses linear coherence between disc and coronal bands at and below the QPO frequency. Specifically, the lag at the QPO fundamental between 0.3-0.6 keV and 3-10 keV is almost 1 rad in the lowest hard-flux bin and consistent with zero in the highest bin, with a difference exceeding 0.5 rad across bins, while lags between power-law-dominated bands show little or no flux dependence. The paper shows the coherence drop tracks the QPO frequency across the outburst, appears in the high-inclination source MAXI J1803-298 but not in low-inclination GX 339-4, and is absent in an early observation without a QPO. The authors interpret these patterns as evidence that the vertical extent or geometry of the corona changes on timescales slightly longer than the QPO cycle, which would explain both the variable lags and the reduced coherence.","pith_inferences":["A quantitative simulation of the flux-binning selection effect, with constant intrinsic lags and the measured ~10% spectral changes, would settle whether the intrinsic-lag conclusion holds; the paper argues against the mixing explanation but does not run such a simulation.","If changing coronal height drives the effect, then phase-resolved spectroscopy within each QPO cycle should show the disc illumination pattern varying with QPO phase, a testable prediction that follows from the geometric interpretation.","The steepening of the lag-flux slope below roughly 0.3 Hz predicts that other bright high-inclination hard-state sources with low QPO frequencies should show even larger lag swings, providing a direct observational check.","Searching for coherence dips alone, rather than requiring a power-spectral QPO peak, could reveal QPO mechanisms in fainter sources or in states where the QPO is too weak to be detected in the power spectrum."],"forward_implications":["Time-averaged QPO lag measurements will systematically underestimate the disc-corona lag whenever the hard flux varies within the averaging window.","Any QPO model that assumes fixed geometry and fixed lags on timescales of tens of seconds is inconsistent with the observed lag swings of more than half a radian.","The coherence boundary at the QPO frequency implies the QPO mechanism adds or removes variability that is not shared linearly between disc and corona, so unity-coherence assumptions at these frequencies are invalid.","The presence of the effect in two high-inclination sources and its absence in a low-inclination source points to a viewing-angle-dependent geometric origin, strongest when the disc is seen more edge-on.","During the bright decline, the lag and coherence features persist at the expected QPO frequency even when no power-spectral peak is detectable, so the timing signature can outlive the power-spectrum signature."],"supporting_citations":[{"why":"Established the flux-dependent short-term lags and the method of binning by instantaneous coronal flux that this work extends to the QPO frequency.","marker":"Bollemeijer et al. (2024)"},{"why":"Provided the cross-spectral formalism used to compute phase lags and coherence.","marker":"Uttley et al. (2014)"},{"why":"Defined the intrinsic coherence function and its interpretation as a measure of linear correlation.","marker":"Vaughan & Nowak (1997)"},{"why":"Supplied the error estimates for the phase lags used in the flux-bin measurements.","marker":"Ingram (2019)"},{"why":"Proposed the Lense-Thirring precession model that is the main geometric interpretation for the variable lags.","marker":"Ingram et al. (2009)"},{"why":"Provided the vKompth Comptonization model whose QPO lags depend on coronal size and feedback, discussed as an alternative interpretation.","marker":"Bellavita et al. (2022)"},{"why":"Measured disc and coronal spectral parameters varying out of phase in a QPO, supporting the geometric interpretation.","marker":"Stevens & Uttley (2016)"},{"why":"Showed that the disc illumination pattern depends strongly on coronal height, used to argue that changing geometry can produce the lag swings.","marker":"Dauser et al. (2014)"},{"why":"Previously saw a low-frequency coherence drop in another source, providing the comparison that makes the new QPO-connected coherence drop notable.","marker":"Cassatella et al. (2012)"}],"fun_headline_variants":["QPO disc-corona lags vary with coronal flux in seconds","Reduced coherence ties QPO to corona geometry changes","Black hole QPO lags swing with X-ray brightness in seconds","MAXI J1820+070 reveals flux-dependent QPO lags"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that grouping 18-second segments by their 3-10 keV count rate and measuring the lag in the QPO bin does not itself create the lag-flux relation through selection effects or through a changing mixture of disc and coronal emission.","fun_headline_variants_meta":{"raw":{"variants":["QPO disc-corona lags vary with coronal flux in seconds","Reduced coherence ties QPO to corona geometry changes","Black hole QPO lags swing with X-ray brightness in seconds","MAXI J1820+070 reveals flux-dependent QPO lags"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000319,"raw_usage":{"total_tokens":1879,"prompt_tokens":1103,"completion_tokens":776,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":719,"completion_tokens_details":{"reasoning_tokens":702}},"tokens_in":719,"tokens_out":776,"duration_ms":7717,"temperature":1.0,"reasoning_tokens":702,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:13:12.165419+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate light curves with constant intrinsic QPO lags and the measured ~10% spectral component changes, apply the same 18-second flux binning and third-bin lag estimate, and check whether a >0.5 rad lag difference between flux bins emerges; if it does, the intrinsic-lag conclusion fails.","supporting_citations":[],"review_version":2}