{"id":"58c72a6a-1879-4404-9f87-5f4d74028e56","arxiv_id":"2501.04096","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"JWST NIRCam infrared monitoring of Sgr A* finds fully continuous variability on seconds-to-year timescales, a faint/bright two-population flux distribution, and a 3-40 second lag of 4.8 micron behind 2.1 micron emission.","lead":"JWST watched the black hole at the center of our Milky Way, Sgr A*, for about 48 hours across seven days in 2023 and 2024, catching it flickering constantly in two infrared colors. The data show that the near-infrared light from the black hole never turns off, and that brighter flares have a slightly different color than the faint constant flicker.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the non-stop variability claim is robust to background uncertainties.","rationale":"The reader's conditional verdict is appropriate for the paper's secondary claims (the two-population spectral-index split, the 3-40 s time delay, and the magnetic-field / cutoff-energy estimates), which have real weaknesses: the faint-end spectral indices depend on the uncertain 4.8 micron background, the time delay is not significance-tested, and the physical parameters come from a six-parameter fit to five flares. However, these issues do not bear directly on the strongest claim as framed by the reader, which is the non-stop correlated variability. That claim is supported by stable reference stars, two-color correlation, and high signal-to-noise fluctuations. A constant background error does not remove time-varying signal. Therefore, while the paper should remain CONDITIONAL because of the overreach in derived quantities, the central observational result stands without a load-bearing objection. The agreement_with_reader is set to 'disagree' because the reader's identified weakest assumption (4.8 micron background) is not the load-bearing concern for the non-stop variability claim; it is a concern for a different, secondary claim.","tokens_in":25396,"tokens_out":11349,"duration_ms":112099,"concrete_test":"Independently reduce the 2024 time-series data using PSF-fitting photometry instead of aperture photometry, and repeat the 4.8 micron background subtraction with the background level perturbed by +1 and +2 mJy. If the short-timescale fluctuations and epoch-to-epoch pedestal variations persist in both cases, the non-stop variability claim is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of non-stop infrared variability is supported by multiple independent lines of evidence: the reference star S0-17 and the field star 'Ref' are stable to within their noise, the 2.1 and 4.8 micron fluctuations are correlated in all epochs, and the variability amplitude is roughly a factor of 500 above the photometric noise at 2.1 micron. A constant background offset, even if misestimated by 1-2 mJy at 4.8 micron, cannot generate time-correlated fluctuations; it only shifts the zero point. The paper's own admission of subjectivity in the 4.8 micron background and in the polynomial pedestal fits (Sections 2.4 and Appendix A) is relevant to the spectral-index decomposition and the two-population interpretation, but it does not undermine the existence of continuous variability. No load-bearing concern is identified for the strongest claim as stated.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents JWST/NIRCam photometry of Sgr A* in the F210M and F480M filters over seven epochs from April 2023 to April 2024, totaling roughly 48 h of nearly continuous monitoring. The authors report correlated variability at 2.1 and 4.8 micron in all epochs, continuous short-timescale (seconds-to-minute) fluctuations with no detected truly steady quiescent level, epoch-to-epoch changes in the underlying pedestal, a break in the flux-flux relation near 3 mJy at 2.1 micron, two spectral-index regimes (steep for faint, shallow for bright), 3-40 s lags of 4.8 micron relative to 2.1 micron, counterclockwise loops in spectral-index versus flux diagrams, and synchrotron flare modeling that yields magnetic field strengths of roughly 40-90 G and cutoff energies of 420-720 MeV. The data processing is described in unusual detail, with MAST-archived data, ALMA-based astrometric registration, reference-star noise checks, and explicit saturation-timing corrections.","tokens_in":25508,"tokens_out":9183,"duration_ms":84130,"significance":"The strongest claim, that Sgr A* is essentially always varying in the near-infrared rather than flaring atop a steady baseline, is well supported by the data presented: the reference stars are stable to their noise, the two bands vary in a correlated manner, and the variability amplitude is far above the photometric noise. If the spectral-index bimodality, short time delays, and magnetic-field estimates survive further scrutiny, this dataset will provide valuable new constraints on particle acceleration and cooling in the inner accretion flow. The openness of the data and the detailed treatment of systematic backgrounds are strengths. The remaining concerns are about the statistical support for several derived claims and about whether those claims are robust to the stated systematic uncertainties.","major_comments":[{"comment":"The two-population spectral-index decomposition is not robust to the adopted 4.8 micron background. The paper states in §3.3 that increasing the 4.8 micron background by 1-2 mJy makes the positive spectral indices of the faint emission disappear. The slopes a± in the flux-flux plane are offset-independent, but the conversion to α− and α+ and the quoted values in Table 6 depend on the assumed background. Please propagate the estimated 1-2 mJy background uncertainty through to α−, α+, and S0, and show explicitly how the claimed bimodality and the loop interpretation change within that allowed range. As written, the title claim of distinct faint and bright populations is not yet established.","section":"§3.3, Table 6, Fig. 7e"},{"comment":"The reported 3-40 s inter-band delays are comparable to or shorter than the 18-46 s sampling cadence. The 4.8 micron data are interpolated onto the 2.1 micron grid, and the cross-correlation is applied to strongly correlated red-noise light curves, which can produce nonzero CCF peaks by chance. No confidence intervals, bootstrap/permutation significance levels, or null-hypothesis tests are reported. The authors should provide a significance estimate for each delay and demonstrate that the delays are not artifacts of the interpolation or of the saturation timing correction described in §2.2.","section":"§3.4, Figs. 7-8"},{"comment":"The low-frequency break f0 reported in Table 5 may be systematically affected by the pedestal-removal procedure. Appendix A describes 4th-degree polynomial fits that are constrained to lie below the light curve; if these fits (or equivalent detrending) are removed before computing the power spectra, low-frequency power is suppressed by construction and the inferred f0 would not be intrinsic. The manuscript should state explicitly whether the PSDs are computed on raw or detrended light curves, and it should validate the f0 and n recovery using synthetic light curves with known power spectra processed through the same pedestal procedure. The Monte Carlo uncertainties in Table 5 do not include this systematic.","section":"§3.2, Appendix A"},{"comment":"The claim of a bimodal flux distribution is not statistically quantified. Table 3 lists χ2 values but no degrees of freedom, reduced χ2, or model-selection criterion. At 4.8 micron the two-log-normal fit (χ2=82.11) is only slightly better than the log-normal-plus-power-law fit (χ2=85.77), despite the latter using a different functional form, and at 2.1 micron the log-normal-plus-power-law is preferred. The identification of two populations should be supported by a likelihood-ratio or information-criterion comparison that accounts for the number of parameters and for the small number of bright points.","section":"§3.1.5, Table 3"},{"comment":"The synchrotron model is described as having six free parameters (B, Q0, Ec, t0, t1, t2), but the fits in Table 7 also vary β1 and β2, giving eight parameters per flare. Fitting two light curves with eight parameters for each of five flares is likely to be strongly degenerate, so the quoted B=38-92 G and Ec=424-736 MeV ranges may not be meaningful as a 'direct estimate'. Please report parameter covariances or corner plots, specify whether the ranges are joint confidence intervals or merely the spread of best-fit values, and discuss how the assumed E^-2 injection spectrum and fixed E1=2 MeV affect the inferred B.","section":"§4.1, Appendix D, Table 7"}],"minor_comments":[{"comment":"The text refers to Table 3 for the mean, median, and sigma fluxes, but these quantities appear in Table 4; Table 3 contains the flux-distribution fit parameters.","section":"§3.1.5"},{"comment":"The text cites 'Dodds-Eden et al. (1999)' in several places, but the bibliography lists Dodds-Eden et al. (2009); the year should be corrected consistently.","section":"References"},{"comment":"The Figure 7e caption says the 4.8 micron subtracted background is reduced by 1 and 2 mJy, while §3.3 says it is increased by 1 and 2 mJy; these statements need to be reconciled.","section":"Fig. 7e caption and §3.3"},{"comment":"The caption contains 'log-term quiescent variability'; this should be 'long-term'.","section":"Fig. 6 caption"},{"comment":"The sentence about S24 and S29 spilling 'more flux in 2004 compared to 2003' appears to be a typo for 2024 and 2023.","section":"§2.4"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the non-stop variability result is the most defensible and important claim in the paper. The two-population, time-delay, and magnetic-field claims are presented with insufficient statistical support and with a systematic background uncertainty that the authors themselves acknowledge. If the authors are willing to reframe the paper around the variability result and mark the spectral-index bimodality, delays, and B-field estimates as preliminary, or to add the requested joint analyses, the paper would be suitable for publication. The discrepancy between the Figure 7e caption and §3.3 regarding the background adjustment should be fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the paper's headline claim—that Sgr A* is essentially never quiescent in the NIR and that the 2.1 and 4.8 µm fluctuations track each other continuously—is well supported. The genuinely new bits are the simultaneous two-color JWST light curves, the reported 3–40 s lag of 4.8 µm behind 2.1 µm, and the loop diagrams. Non-stop flickering itself was already known from ground-based work (Witzel et al. 2018; Weldon et al. 2023), so the novelty is the simultaneous cadence and the delay, not the constant variability.\n\nWhat the paper does well is the data reduction. The astrometric registration against ALMA, the noise characterization using S0-17 and the second reference star, the explicit treatment of the 4.8 µm saturation timing offset, and the public MAST data all make the photometry reproducible. The reference stars are flat; the variability amplitude at 2.1 µm is roughly 500 times the noise; a misestimated constant background cannot manufacture time-correlated fluctuations. The continuous variability claim is robust, and I agree with the stress-test note that no load-bearing objection lands on it.\n\nThe soft spots are real, though. First, the two-population spectral-index story leans on the 4.8 µm background estimate. The paper itself shows (Fig. 7e) that the positive faint spectral indices vanish when the background is raised by 1–2 mJy. The slopes in flux-flux space are background-independent, but the spectral-index interpretation is not, and the background estimate is explicitly subjective (Section 2.4). Second, the 3–40 s delay is presented without any significance testing. Cross-correlation peaks from red-noise light curves need a Monte Carlo or surrogate null before they carry weight. Third, the abstract's 'direct estimate' of B=40–90 G and Ec=420–720 MeV is misleading: those numbers come from a six-parameter fit to five hand-picked flares with a linear baseline subtracted. The fit is a reasonable illustrative exercise, but the parameters are not direct measurements.\n\nNone of this kills the main result. The paper is honest about its own subjectivity, and the variability detection carries independent weight. People modeling Sgr A* accretion will find the light-curve data and the model fits useful even if they treat the physical parameters cautiously. Send it to peer review; the referee's main job should be to pin down the lag significance and clearly separate background-dependent spectral-index claims from background-independent ones like the flux-flux slopes. If the delay survives proper null testing, it becomes a genuinely new observable for Sgr A* flare physics.","headline":"Solid JWST light curves convincingly show Sgr A* is never quiet in the NIR; the two-population spectral-index split and the 3–40 s lag need stronger statistical and background treatment before carrying the physical claims.","tokens_in":26179,"tokens_out":4111,"would_cite":true,"duration_ms":35560,"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":"JWST observations of Sgr A* at 2.1 and 4.8 microns show its near-infrared emission flickers non-stop, with faint and bright variable components that have distinct spectral indices and a 3–40 second lag between the two bands.","keywords":["Sgr A*","near-infrared variability","JWST NIRCam","spectral index","black hole accretion flow","synchrotron cooling","Galactic center","flares"],"falsifier":"Measure the actual 4.8 µm stellar contamination independently, for example from high-resolution spectroscopy or a stellar-PSF model of S0-2, S24, and S29 at the observed epochs; if the true contamination exceeds the adopted background by more than about 2 mJy, the faint positive spectral indices disappear and the claimed two-population split in spectral index collapses.","tokens_in":25080,"feed_emoji":"🕳️","tokens_out":5834,"duration_ms":51594,"temperature":0.7,"pith_summary":"The paper argues that Sgr A*'s near-infrared emission never turns off: across seven JWST epochs totaling about 48 hours, the flux at 2.1 and 4.8 microns fluctuates at all times, with no measurable steady quiescent level. It claims the variability splits into two populations—faint continuous flickering with steep spectral index $\\alpha \\approx -1.6$ and bright flares with shallower index $\\alpha \\approx -0.85$—meeting at a break near 3 mJy at 2.1 microns. It also claims that 4.8 micron emission lags 2.1 micron emission by 3–40 seconds, producing counterclockwise loops in spectral-index plots, and that synchrotron-cooling fits give magnetic field strengths of 40–90 G. If correct, the inner accretion flow is always active, with faint flickering tied to turbulence and bright flares tied to episodic reconnection.","feed_headline":"Sgr A* never stops flickering, JWST shows","feed_subtitle":"Seven epochs at 2.1 and 4.8 microns reveal two flare populations and a few-second infrared lag.","key_machinery":"The central machinery is simultaneous NIRCam F210M and F480M photometry of Sgr A* with astrometric registration to ALMA 230 GHz images and per-epoch background subtraction, combined with (1) broken power-law fits to temporal power spectra, (2) broken linear fits to the 4.8-versus-2.1 µm flux-flux relation whose slope change marks the boundary between faint and bright populations, and (3) an age-stratified synchrotron model in which electrons are injected with an $E^{-2}$ spectrum and upper cutoff energy $E_c$, cool in a constant magnetic field $B$, and produce optically thin emission at the two NIR frequencies. The model's free parameters allow the observed loops and lags to be reproduced, yielding estimates of $B$ and $E_c$.","core_discovery":"On its own terms, the paper establishes that Sgr A*'s near-infrared light curves, observed simultaneously at 2.1 and 4.8 µm with JWST NIRCam over seven epochs in 2023–2024, show continuous variability on timescales from seconds to hours in every epoch, with the pedestal level itself changing from day to day and year to year. The flux-flux correlation has a break near 3 mJy at 2.1 µm: fainter emission has spectral index $\\alpha \\approx -1.58$, brighter emission $\\alpha \\approx -0.85$, and the flux histogram is best described by two log-normal components. Cross-correlation reveals a 3–40 s lag of 4.8 µm behind 2.1 µm, and each flare traces a counterclockwise loop in spectral-index versus flux space. The paper interprets the loops as synchrotron cooling in an evolving, age-stratified electron population and derives magnetic field strengths of 40–90 G and cutoff energies of 420–720 MeV.","pith_inferences":["If the two-population picture holds, future simultaneous two-band JWST observations could test whether the pedestal level predicts the amplitude or rate of bright flares; the paper notes a qualitative correlation between strong flaring and elevated pedestal but does not quantify it.","Because the positive spectral indices of faint emission depend on a 4.8 µm background that is uncertain by 1–2 mJy, deeper imaging that resolves S0-2, S24, and S29 could either confirm the faint-component spectral index or show it is an artifact.","The 3–40 s lag may be a general signature of synchrotron cooling in accreting black holes; if the lag scales with black hole mass, similar simultaneous two-band monitoring could constrain magnetic field strengths in other galactic nuclei.","A direct test of the non-stop variability claim would be a longer, gap-free observation at 2.1 µm alone: if a true quiescent floor ever appears, the 'always fluctuating' interpretation would need revision."],"forward_implications":["If the non-stop variability claim is correct, there is no steady quiescent near-infrared component from Sgr A*, so models must explain an always-on flickering process rather than flares superimposed on a stable baseline.","The ~3 mJy break in the flux-flux slope and the dual log-normal flux histogram imply two distinct particle populations: a faint steep-spectrum component and a bright shallow-spectrum flare component.","The 3–40 s lag of 4.8 µm behind 2.1 µm, together with the counterclockwise spectral-index loops, implies synchrotron cooling in fields of 40–90 G with upper cutoff energies of 420–720 MeV.","Sub-minute flux changes—up to a factor of two in about 1.4 minutes—place the emitting region at horizon scales of a few gravitational radii, linking the infrared variability to the inner accretion flow.","The epoch-to-epoch pedestal changes by roughly a factor of two, showing long-term variability on daily, monthly, and yearly timescales that any complete model of Sgr A* must reproduce."],"supporting_citations":[{"why":"Earlier NICMOS light curves showing persistent low-level flickering that the JWST 'non-stop variability' claim directly extends to continuous, higher-cadence data.","marker":"Yusef-Zadeh et al. 2009"},{"why":"Prior flux-distribution and power-spectrum analysis that supplies the log-normal/power-law baselines and white-noise timescale this paper compares its broken power spectra against.","marker":"Witzel et al. 2018"},{"why":"Source of the two-log-normal fitting method used to separate faint and bright populations in the flux histograms.","marker":"Dodds-Eden et al. 1999"},{"why":"Orbital parameters used to evolve S-cluster stars to the observation epochs and estimate their contaminating flux in the Sgr A* aperture.","marker":"Gillessen et al. 2017"},{"why":"Extinction values $A_{K_s}=2.46$ and $A_M=1.0$ applied to convert measured fluxes to the de-reddened 2.1 and 4.8 µm light curves.","marker":"Schödel et al. 2011"},{"why":"Provides the ~5 $r_g$ hot-spot orbit and ~0.3c velocities used to interpret sub-minute infrared variability as horizon-scale emission.","marker":"GRAVITY Collaboration et al. 2018"},{"why":"GRMHD simulations of the inner accretion flow whose predicted magnetic field strengths bracket the 40–90 G values inferred from the flare fits.","marker":"Ressler et al. 2020a,b"},{"why":"Submillimeter power-spectrum break timescale (~8 hr) that the shorter infrared break frequencies are compared against to argue the low-frequency turnover changes between epochs.","marker":"Dexter et al. 2014"}],"fun_headline_variants":["JWST: Sgr A* flickers constantly in two distinct modes","Sgr A*'s infrared lag reveals two flare populations","JWST sees Sgr A* vary sub-minute, two emission processes","Sgr A*'s flickering never stops, JWST shows dual nature"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 4.8 micron background subtraction—including contamination from S0-2, S24, S29, and diffuse emission—is correct to within about 1–2 mJy; the paper shows that raising that background by 1–2 mJy erases the positive spectral indices of the faint emission, so the two-population spectral split depends on it.","fun_headline_variants_meta":{"raw":{"variants":["JWST: Sgr A* flickers constantly in two distinct modes","Sgr A*'s infrared lag reveals two flare populations","JWST sees Sgr A* vary sub-minute, two emission processes","Sgr A*'s flickering never stops, JWST shows dual nature"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000194,"raw_usage":{"total_tokens":1435,"prompt_tokens":1110,"completion_tokens":325,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":726,"completion_tokens_details":{"reasoning_tokens":246}},"tokens_in":726,"tokens_out":325,"duration_ms":3850,"temperature":1.0,"reasoning_tokens":246,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:41:15.869168+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the actual 4.8 µm stellar contamination independently, for example from high-resolution spectroscopy or a stellar-PSF model of S0-2, S24, and S29 at the observed epochs; if the true contamination exceeds the adopted background by more than about 2 mJy, the faint positive spectral indices disappear and the claimed two-population split in spectral index collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier NICMOS light curves showing persistent low-level flickering that the JWST 'non-stop variability' claim directly extends to continuous, higher-cadence data."},{"cited_title":"P., Morris M","cited_arxiv_id":null,"evidence_quote":"Prior flux-distribution and power-spectrum analysis that supplies the log-normal/power-law baselines and white-noise timescale this paper compares its broken power spectra against."}],"review_version":1}