{"id":"55326ddc-7d05-48a3-91ef-5ac309797bdc","arxiv_id":"2505.09779","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"Using ZTF light curves of 915 quasars, the authors find a universal anti-correlation between variability amplitude and Eddington ratio, expressed as log λEdd = -0.71 log Fvar - 1.52.","lead":"Quasars that accrete closer to the Eddington limit show smaller optical brightness fluctuations, a pattern that holds across most of cosmic time. The paper quantifies this as a single equation connecting variability amplitude to Eddington ratio, potentially letting astronomers estimate black hole feeding rates from light curves alone.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The universal λEdd–Fvar relation is not supported by the paper's own bin fits: slopes range from -0.99 to -0.16 and the 2<z<3 bin has R²=0.01, so Eq. 5 may be a between-redshift artifact rather than a universal relation.","rationale":"The reader correctly identifies the comparability of Fvar across redshift as a weak point, especially the rest-frame baseline effect acknowledged in Section 4.4. My concern is more direct: the paper's own bin-by-bin fits are internally inconsistent with the claimed universal relation. Slopes vary by roughly a factor of six, and the highest-redshift bin, which is essential for a 0<z<3 claim, has R²=0.01. This does not require an external assumption about quasar variability; it is visible in Table 4. The concrete interaction test is well-defined and would settle whether the slope differences are statistically real. If they are, the correct conclusion would be a redshift-dependent relation, not a universal one; if they are not, a restricted version might survive. I therefore keep the CONDITIONAL verdict rather than upgrading to REJECT, because the needed test is clear and the underlying anti-correlation is plausible and consistent with prior work. Credit is due for the public filtering code and for the explicit caveats about selection effects and rest-frame baselines, but those caveats do not by themselves establish the universality of Eq. 5.","tokens_in":16222,"tokens_out":6400,"duration_ms":68054,"concrete_test":"Run a WLS regression of log λEdd on log Fvar, a five-level redshift-bin factor, and the log Fvar × redshift-bin interaction, using the same weights as Section 4.1, and test whether all interaction coefficients are zero using an F-test or ΔBIC against the no-interaction model. If the interaction is significant, no single universal slope exists and Eq. 5 cannot serve as a redshift-independent estimator; as a follow-up, truncate all light curves to a common 500-day rest-frame baseline and refit to test whether the baseline mismatch drives the apparent high-redshift trend.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. 5 encodes the central claim: one slope and intercept predict λEdd from Fvar across 0<z<3. The paper's own Table 4 contradicts the required invariance. The WLS slopes per bin are -0.99, -0.32, -0.48, -0.43, and -0.16 from z<0.5 through 2<z<3, with R² values 0.18, 0.03, 0.10, 0.12, and 0.01. Section 4.4 describes the low- and intermediate-redshift fits as all very similar, but -0.99 versus -0.32 is a factor of three, and the highest-redshift bin has essentially no within-bin correlation. The global r=-0.31 and slope -0.71 therefore plausibly arise from between-bin covariance: higher-redshift quasars have higher λEdd and lower Fvar, the latter partly from the shortened rest-frame baseline (~500 days at z~3 versus ~2000 days at z~0, noted in Sections 4.1 and 4.4). That is a population-mixing signature, not evidence for a universal physical relation. In addition, even if the slope were stable, R²=0.15 gives little support for the claim that Eq. 5 is particularly useful for estimating individual Eddington ratios.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes ZTF g-band photometric variability for 915 quasars from the AQMES-MED sample and correlates the fractional variability Fvar with bolometric luminosity, black hole mass, and Eddington ratio across five redshift bins from 0 to 3. The authors report an anti-correlation between Fvar and luminosity, a redshift-dependent Fvar-MBH relation, and a global anti-correlation between log lambda_Edd and log Fvar. This global relation is expressed as Eq. (5) and is claimed to be nearly redshift independent and useful for estimating Eddington ratios of quasars at 0<z<3.","tokens_in":16531,"tokens_out":3496,"duration_ms":36518,"significance":"If the claimed universal lambda_Edd-Fvar relation were established, it would provide a simple photometric estimator of Eddington ratio across a wide redshift range and a useful constraint for accretion-disk variability models. The paper contributes a carefully constructed sample, explicit checks on emission-line contamination in the photometric bands, publicly available filtering code, and a machine-readable catalog of variability and spectral properties. However, the central claim of a universal, redshift-invariant relation is not supported by the paper's own bin-dependent fits, and the global relation's predictive power is weak. The lasting value of the work is likely the descriptive correlations and the data products rather than Eq. (5) as currently stated.","major_comments":[{"comment":"The claim that the WLS fits are \"all very similar\" for 0<z<2 is contradicted by Table 4: the slopes are -0.99, -0.32, -0.48, -0.43, and -0.16 across the five redshift bins. The lowest-redshift slope differs from the adjacent bin by roughly a factor of three, and the 2<z<3 bin has R^2=0.01, which is effectively no within-bin correlation. Therefore Eq. (5), presented as a universal 0<z<3 relation, is not supported by the paper's own bin-resolved fits.","section":"Section 4.4 and Table 4"},{"comment":"The global regression may reflect between-bin covariance rather than an intrinsic physical relation. Higher-redshift quasars in this sample have higher lambda_Edd and lower Fvar, with the latter partly caused by the rest-frame baseline shrinking from about 2000 days at z~0 to about 500 days at z~3 (Section 4.1). The authors should report partial correlations controlling for redshift and quantify the global fit after formally excluding or down-weighting the highest-redshift bin; the statement that the correlation coefficient remained unchanged after excluding that bin is not accompanied by any quantitative result.","section":"Section 4.4, Eq. (5)"},{"comment":"The predictive utility claimed for Eq. (5) is not established. The relation is an in-sample fit to the same 915 objects with R^2=0.15, and no independent validation set or cross-validation is provided. With R^2=0.15, the scatter in inferred lambda_Edd is large relative to the dynamic range of the relation, so the phrase \"particularly useful to estimate the Eddington ratio\" needs quantitative support, such as a scatter plot of predicted versus measured lambda_Edd or a cross-validated prediction interval.","section":"Section 4.4, Eq. (5)"},{"comment":"The systematic redshift dependence of the rest-frame monitoring baseline and rest-frame wavelength is load-bearing for the universality claim. The paper concedes that the rest-frame interval at z>2 may be insufficient to fully capture variability. If high-z Fvar values are systematically underestimated, the observed flattening of the lambda_Edd-Fvar relation at high redshift is an expected artifact, not evidence of a universal relation. The authors should either quantify this effect by recomputing Fvar on matched rest-frame baselines or explicitly restrict the claimed generality of Eq. (5) to z<2.","section":"Section 4.1 and Section 4.4"}],"minor_comments":[{"comment":"The sentence \"an anti-correlation the highest redshifts\" is missing the word \"at\" before \"the highest redshifts.\"","section":"Abstract"},{"comment":"Lu et al. (2019) and Vanden Berk et al. (2004) each appear twice in the reference list with slightly different citation details; these duplicates should be merged.","section":"References"},{"comment":"The second term in Eq. (4) uses sigma_err without a clear definition; if it denotes the standard deviation of the measurement errors, that should be stated explicitly, and the notation should be made consistent with sigma^2_err used in Eqs. (2) and (3).","section":"Eq. (4)"},{"comment":"The caption of Figure 11 describes a green shaded region as a 95% prediction interval, while the text in Section 4.4 refers to the green-shaded region as the 95% confidence interval; these statements should be aligned.","section":"Figure 11 and Section 4.4"},{"comment":"The phrase \"approximately 2000 MJDs in the rest frame per source\" is ambiguous: it likely means roughly 2000 epochs or a baseline in MJD units, but the wording should be clarified to distinguish the number of epochs from the temporal baseline.","section":"Section 2.2"}],"recommendation":"major_revision","confidential_remarks":"The paper's abstract and conclusions assert a universal, redshift-independent Eddington-ratio relation that is not supported by the bin-resolved fits in Table 4. I recommend requiring the authors to either substantially revise the central claim to a redshift-dependent or low-redshift relation, or provide matched-baseline corrections and out-of-sample validation that would justify the current wording. The data products and correlation analysis are useful and could form a solid foundation for a revised paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a useful dataset and a careful photometric variability analysis, but the headline 'universal relation' is not supported by the numbers in the paper itself. The global fit (Eq. 5, slope -0.71, R²=0.15) looks like a between-redshift artifact: within the five redshift bins the WLS slopes are -0.99, -0.32, -0.48, -0.43, and -0.16, and the 2<z<3 bin has R²=0.01. Those are not 'all very similar,' as Section 4.4 claims. The high-z bin barely correlates, and its shorter rest-frame baseline (~500 days vs ~2000 days at z≈0) plus bluer rest-frame g-band means the Fvar values are not directly comparable across bins. The global anti-correlation (r=-0.31) is statistically significant, but it probably reflects the fact that higher-redshift quasars are both more luminous/less variable and observed over shorter rest-frame windows.\n\nWhat the paper does well: 915 quasars from AQMES-MED with ZTF g-band light curves over ~6 years, a documented filtering pipeline (code on GitHub with DOI), a check of emission-line contamination in the filters, and a machine-readable table. That is genuinely useful, reproducible groundwork. The luminosity–Fvar anti-correlation is clean and consistent with prior work. The redshift-dependent MBH–Fvar trend (positive at z<0.5, negative at z>2) is interesting, even if it may be partly selection-driven.\n\nThe soft spots are in the interpretation, not the data handling. Eq. 5 is fitted to the same 915 objects used to assert its predictive utility, with no independent validation set; with R²=0.15 the scatter is large enough that 'estimating the Eddington ratio of a general quasar from Fvar' is not practically meaningful. The paper acknowledges the rest-frame baseline problem and the selection biases in Sections 4.1 and 4.4, but does not correct for them or test whether the bin slopes are consistent within errors. That last point matters: the claim of redshift independence requires showing the bin slopes are statistically indistinguishable, which the paper does not do.\n\nWho should read it: AGN variability people, especially those working toward LSST and Rubin. It consolidates known correlations and provides a reusable sample. It deserves peer review—the analysis is worth refereeing—but the authors need to either soften the universality claim or demonstrate it with a proper redshift-independent test (e.g., matching rest-frame baselines, fitting a hierarchical model, or validating on an external sample). I would not cite Eq. 5 as a predictive relation, but I might cite the sample/code.\n\nRecommendation: send to a competent referee with a request for major revision, focusing on the bin-slope consistency and validation.","headline":"The underlying Fvar–λEdd anti-correlation is real and the sample is solid, but Eq. 5 as a universal predictor does not survive the paper's own Table 4.","tokens_in":17130,"tokens_out":3117,"would_cite":false,"duration_ms":29694,"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":"Quasar flicker rate predicts black hole feeding pace across z=0 to z=3.","keywords":["quasar variability","Eddington ratio","fractional variability","supermassive black holes","accretion","Zwicky Transient Facility","active galactic nuclei","photometric monitoring"],"falsifier":"Extend the $z\\approx2$-$3$ light curves to a rest-frame baseline comparable to low redshift, roughly 2000 days as LSST monitoring will provide, and recompute $F_{\\rm var}$. If high-redshift $F_{\\rm var}$ values rise substantially, the universal relation steepens or becomes redshift-dependent; if they stay put, the relation is confirmed.","tokens_in":16029,"feed_emoji":"🔭","tokens_out":6083,"duration_ms":56508,"temperature":0.7,"pith_summary":"This paper uses six years of Zwicky Transient Facility g-band photometry for 915 quasars spanning redshifts 0 to 3 to ask whether the amplitude of quasar brightness variations tracks the rate at which the black hole accretes matter. It argues that the fractional variability $F_{\\rm var}$ anticorrelates with Eddington ratio in every redshift bin, while correlations with luminosity and black hole mass change sign or strength with redshift. The combined sample yields a single relation, $\\log \\lambda_{\\rm Edd} = (-0.71\\pm 0.06)\\log F_{\\rm var} - (1.52\\pm 0.06)$, which the paper presents as nearly redshift-independent. If the relation holds, photometric monitoring alone can estimate the Eddington ratio of a quasar at $0<z<3$, turning variability into a probe of accretion state.","feed_headline":"How fast a quasar flickers reveals its black hole's feeding pace","feed_subtitle":"One equation turns six years of ZTF g-band photometry into Eddington ratios for 915 quasars from z=0 to z=3.","key_machinery":"The load-bearing quantity is fractional variability $F_{\\rm var}$, the excess variance of a light curve normalized by its mean flux, computed after subtracting the average measurement-error variance. It is measured from ZTF g-band light curves binned to one-day means, after removing outliers and requiring at least 100 points. Because the Eddington ratio normalizes bolometric luminosity by the Eddington luminosity, it removes most of the mass and redshift dependence carried by luminosity alone; the weighted least-squares regressions of $\\log\\lambda_{\\rm Edd}$ against $\\log F_{\\rm var}$ are the machinery that produces Equation 5.","core_discovery":"The central claim is that the Eddington ratio, not luminosity or black hole mass alone, is the physical parameter that organizes optical quasar variability across cosmic time. Over the full $0<z<3$ sample, the measured fractional variability in the g-band obeys $\\log\\lambda_{\\rm Edd}=(-0.71\\pm0.06)\\log F_{\\rm var}-(1.52\\pm0.06)$ with $R^2=0.15$ and global correlation $r=-0.31$; the authors state that this relation is particularly useful to estimate the Eddington ratio of a general quasar at $0<z<3$ from its measured $F_{\\rm var}$. They interpret the anti-correlation physically: at high accretion rates the disk radiates more efficiently and the stochastic amplitude of brightness changes is suppressed relative to the mean. The paper does not claim the relation is exact; it presents it as a robust empirical framework with redshift-dependent scatter, weakest at $2<z<3$.","pith_inferences":["If the relation survives longer baselines, variability could serve as a cheap proxy for black-hole growth rate in surveys too faint for spectroscopy, complementing X-ray and radio accretion indicators.","A natural test is to apply Equation 5 to r-band or i-band light curves; if the same slope holds across bands, the relation is truly wavelength-independent and not a g-band artifact.","One could check individual high-Eddington quasars known from single-epoch spectra: those with low $F_{\\rm var}$ should systematically be the most efficiently accreting, and outliers may flag unreliable virial mass estimates."],"forward_implications":["Eddington ratios for quasars at $0<z<3$ can be estimated from single-band photometric variability alone, without spectra.","Upcoming high-cadence surveys can map accretion states over large quasar samples and cosmic time using one photometric relation.","The redshift dependence seen in luminosity and black-hole-mass correlations is largely a selection effect; Eddington ratio is the underlying variable.","The weaker high-redshift correlation points to a timescale requirement: variability metrics need rest-frame baselines long enough to capture the full fluctuation power."],"supporting_citations":[{"why":"Supplies the SDSS DR16 quasar catalog values for black hole mass, bolometric luminosity, and Eddington ratio used in every correlation.","marker":"Wu & Shen 2022"},{"why":"Describes the ZTF Science Data System and forced photometry service that produced the light curves.","marker":"Masci et al. 2018"},{"why":"Documents the ZTF forced photometry service used for the time-series data.","marker":"Masci et al. 2023"},{"why":"Establishes the fractional variability and excess variance formalism that defines $F_{\\rm var}$.","marker":"Peterson et al. 1998"},{"why":"Introduces SDSS-V and the Black Hole Mapper program from which the AQMES sample is drawn.","marker":"Kollmeier et al. 2019"},{"why":"Presents the ZTF survey design and observing cadence underlying the six-year light curves.","marker":"Bellm et al. 2018"},{"why":"Provides an earlier framework linking quasar variability to black hole mass and accretion, serving as the comparative baseline for the redshift-dependent mass trend.","marker":"Kelly et al. 2009"},{"why":"Provides ensemble variability-luminosity-mass relations that this paper tests against at higher redshift.","marker":"MacLeod et al. 2010"}],"fun_headline_variants":["Quasar flicker rate reveals black hole feeding speed","One curve links quasar flickering to accretion rate","Flicker amplitude predicts black hole feeding rate","Quasar variability: a universal probe of black hole accretion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison treats $F_{\\rm var}$ measured at different redshifts as equivalent, even though the rest-frame baseline shrinks from about 2000 days at $z\\approx0$ to about 500 days at $z\\approx3$ and the g-band moves to progressively bluer rest wavelengths.","fun_headline_variants_meta":{"raw":{"variants":["Quasar flicker rate reveals black hole feeding speed","One curve links quasar flickering to accretion rate","Flicker amplitude predicts black hole feeding rate","Quasar variability: a universal probe of black hole accretion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000644,"raw_usage":{"total_tokens":3026,"prompt_tokens":1074,"completion_tokens":1952,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":690,"completion_tokens_details":{"reasoning_tokens":1889}},"tokens_in":690,"tokens_out":1952,"duration_ms":14688,"temperature":1.0,"reasoning_tokens":1889,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:24:44.623025+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extend the $z\\approx2$-$3$ light curves to a rest-frame baseline comparable to low redshift, roughly 2000 days as LSST monitoring will provide, and recompute $F_{\\rm var}$. If high-redshift $F_{\\rm var}$ values rise substantially, the universal relation steepens or becomes redshift-dependent; if they stay put, the relation is confirmed.","supporting_citations":[{"cited_title":"2019, Bulletin of the American Astronomical Society, doi: 10.3847/2041-8213/ab2c1f Koz lowski, S","cited_arxiv_id":null,"evidence_quote":"Introduces SDSS-V and the Black Hole Mapper program from which the AQMES sample is drawn."},{"cited_title":"C., Kulkarni, S","cited_arxiv_id":null,"evidence_quote":"Presents the ZTF survey design and observing cadence underlying the six-year light curves."}],"review_version":1}