{"id":"6072e68b-4f9e-454c-8a1a-28568f084dc7","arxiv_id":"2507.13251","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The infrared echo of AT 2019qiz reveals parsec-scale dust, implying a remnant AGN torus and a peak luminosity much higher than the optical output, which points to missing extreme-ultraviolet energy.","lead":"Infrared echoes from the tidal disruption event AT 2019qiz show that the dust near its black hole sits more than a parsec away, much farther than expected for a quiet black hole. The finding points to a leftover torus from a recently faded active galactic nucleus and suggests the flare emitted far more energy than we see, likely in the extreme ultraviolet.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The quoted R > 1.2 pc lower limit is based on an ad hoc top-1% likelihood cutoff, not a valid confidence interval; a profile-likelihood bound could be much lower.","rationale":"The reader's verdict is CONDITIONAL, and I agree with that overall judgment. My stress-test identifies a specific, load-bearing flaw that the reader mentioned only in passing: the paper's headline lower limit of 1.2 pc is not a statistically valid confidence bound. The paper takes the top 1% of posterior samples by likelihood and takes the minimum R among them, discarding the 90% credible interval because it contains low-likelihood fits. This is equivalent to reporting the best-fit region rather than a confidence interval, and it can substantially overstate the lower limit when parameters are strongly degenerate, which the paper itself acknowledges. The robust geometric floor of 0.68 pc remains, and even if the true 90% lower limit were ~0.7–0.9 pc, the qualitative conclusions (torus remnant, super-Eddington missing energy) would survive in weaker form. Therefore the verdict should remain CONDITIONAL, with the requirement that the authors provide a proper likelihood-ratio-based lower limit and, ideally, fit the full light curve including the early bump. My concern is complementary to, but not identical with, the reader's weakest assumption about the delta-function and single-temperature idealization; hence partial agreement.","tokens_in":18624,"tokens_out":9927,"duration_ms":117804,"concrete_test":"Re-fit the last five epochs using the same model and data, but compute the profile likelihood of R by maximizing over α, β, σ_d for each fixed R, and identify R_90 where Δχ² = 2.71 (or use the 90% HPD of the posterior). Also, refit with all nine epochs including a second dust component for the early bump and check if the best-fit R shifts. If R_90 < 1.0 pc, the abstract's '>1.2 pc' claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim — an inner torus radius >1.2 pc — is derived not from a standard posterior credible interval but from the minimum R among the top 1% of MCMC samples by likelihood (Section 3.2). The paper explicitly discards the 90% credible interval because it contains low-likelihood fits, then redefines the 'lower limit' via this high-likelihood subset. This is not statistically valid: it conditions on good fits and ignores posterior weight, so it cannot provide a confidence bound. The model already has a known degeneracy: with only the rising and plateau phases observed, the four parameters (R, α, β, σ_d) are poorly constrained (the paper itself compares it to solving two equations for three unknowns). A proper profile-likelihood or HPD analysis could easily yield a 90% lower limit closer to the geometric floor R_min = c(t_max - t_peak)/2 = 0.68 pc. If the true bound is ~0.7–0.9 pc rather than 1.2 pc, the torus-remnant interpretation is weakened quantitatively (though a factor ≳5 above the ~0.1 pc sublimation radius would remain), and the derived peak bolometric luminosity lower limits (which scale as R²) would drop by a factor ~0.3–0.6, softening but not eliminating the missing-energy claim. The paper should report a proper likelihood-ratio-based lower limit before the >1.2 pc value is used as a headline.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes the WISE/NEOWISE mid-infrared light curve of the tidal disruption event AT 2019qiz. The IR emission rises steadily for ~4 years and then plateaus, with a roughly constant dust temperature over the last five epochs. The authors construct a thin, inclined dust torus model and fit the last five epochs to infer an inner torus radius R > 1.2 pc. They interpret this large radius as a remnant of a recently faded AGN torus, connecting the source to the broader TDE-QPE unified scenario. They further use the dust echo to derive lower limits on the peak bolometric luminosity of the TDE, finding values (6.6, 9.5, 1.0) × 10^44 erg/s for silicate, SiC, and graphite grains, respectively, all much larger than the observed optical-UV peak, which they attribute to missing EUV energy. The paper also argues that IR-bright TDEs are preferential hosts for QPEs, citing AT2022upj and SDSSJ1335+0728.","tokens_in":18906,"tokens_out":6111,"duration_ms":70600,"significance":"If the large inner radius is robust, this work provides a new piece of evidence for the existence of parsec-scale torus remnants around recently faded AGNs and strengthens the proposed link between IR-bright TDEs and QPEs. The geometric lower-limit argument (R > 0.68 pc, Eq. 4) is simple, transparent, and independent of the detailed dust model, and it already places the echoing dust far outside the ~0.1 pc sublimation radius expected for a 10^6 M_sun black hole. The missing-energy interpretation is also timely and important. The paper is clearly written and makes explicit the degeneracy in its model. However, the headline quantitative claim of R > 1.2 pc rests on a non-standard statistical procedure that, as the authors themselves state in §3.2, is 'not used for uncertainty estimation' but then is used to define a 'lower limitation'. This undermines the precision of the central claim and needs to be fixed with a proper profile-likelihood or posterior-based interval.","major_comments":[{"comment":"The quoted lower limit R > 1.2 pc is not a statistically valid confidence bound. The authors select the top 1% of posterior samples by likelihood and take the minimum R among them, explicitly discarding the 90% credible interval because it contains low-likelihood samples. This conditions on good fits and ignores posterior mass, so it is neither a frequentist confidence interval nor a Bayesian credible interval. Given the acknowledged degeneracy among R, α, β, and σ_d (the text compares the situation to 'two equations ... for three unknowns'), the correct procedure is to compute a profile likelihood in R, maximizing over the other parameters, and then derive a likelihood-ratio-based lower limit. The paper should report this profile and the resulting confidence bound, and should also state what fraction of the posterior mass lies below 1.2 pc. If the profile-likelihood lower limit is closer to the geometric floor of 0.68 pc (Eq. 4), the headline claim '>1.2 pc' must be revised.","section":"§3.2"},{"comment":"The prolonged IR rise is modeled assuming the OUV flare is a delta function, the torus is geometrically thin, and the dust temperature is constant over the fitted epochs. The early bump and the decreasing temperature over the first four epochs are excluded from the fit. These assumptions can bias the inferred radius upward: if the heating source itself brightens over years, or if the emitting dust spans a range of radii and temperatures, the light-travel-time interpretation of the rise is not unique. The robust model-independent result is R_min = c(t_max - t_peak)/2 = 0.68 pc (Eq. 4), which does not rely on these assumptions. To support the specific claim R > 1.2 pc, the authors should demonstrate robustness by convolving the actual OUV light curve (e.g., Hammerstein et al. 2023) with the echo model and by testing a radial dust distribution with a temperature gradient. At minimum, a finite-width pulse test should be shown.","section":"§3.1"},{"comment":"The peak bolometric lower limits are derived from Eq. (15) using R = 1.2 pc and therefore inherit the statistical uncertainty of the R estimate. Since the grain equilibrium condition gives L_bol ∝ R² at fixed dust temperature and composition, reducing the lower limit from 1.2 pc to 0.68 pc would lower the quoted L_peak values by a factor of about 0.32. This would still leave the silicate and SiC limits above the observed OUV peak of 2.7×10^43 erg/s, so the missing-energy conclusion may survive, but the quantitative claim changes materially. The paper should provide a sensitivity analysis of L_peak as a function of R and clearly distinguish the pulse luminosity L_pulse computed from Eq. (15) from the actual peak luminosity; the sentence 'the peak bolometric luminosity Lpeak ~ 1.4 Lpeak' is self-referential and should read L_peak ≈ 1.4 L_pulse.","section":"§4.2"}],"minor_comments":[{"comment":"The abstract states 'resulting in an inner radius >1.2 pc' without mentioning that this is a model-dependent lower limit based on the last five epochs and a non-standard statistical selection; the model-independent geometric limit is 0.68 pc and should be cited as the firm floor.","section":"Abstract"},{"comment":"The sentence 'The posterior samples are well within the prior boundaries and are sufficiently distant from the edges' is misleading because Rmin is the lower boundary of the prior; the authors should specify that the top-1% selected samples are not at this edge.","section":"§3.2"},{"comment":"The shaded bands in Figure 5 show the spread of the top-1% high-likelihood samples, not the posterior predictive distribution; the caption should state this explicitly to avoid misinterpretation.","section":"Figure 5"},{"comment":"The table headers use 'lower limitation' and 'upper limitation'; these should be 'lower limit' and 'upper limit', and column (4) 'logσ_d max' should be clarified as the upper limit on the surface density from the same top-1% selection.","section":"Table 1"},{"comment":"The notation in Eq. (16) and the surrounding text is confusing: 'Lpeak' is used for both the pulse luminosity from Eq. (15) and the true peak luminosity, leading to the self-referential 'Lpeak ~ 1.4 Lpeak'. Use two distinct symbols, e.g., L_pulse and L_peak.","section":"Equation (16)"}],"recommendation":"major_revision","confidential_remarks":"To the editor: The paper addresses a timely and interesting question, and the geometric lower limit of 0.68 pc is robust and valuable. However, the headline quantitative claim of R > 1.2 pc is based on a statistically invalid selection of posterior samples, and the authors themselves admit that their method is 'not used for uncertainty estimation' in §3.2. This needs to be replaced with a profile-likelihood or proper credible-interval analysis before the paper can be accepted. The missing-energy conclusion is probably robust to the revised radius, but the quantitative values will change. I recommend major revision with a request for a clear statistical treatment of the radius lower limit and a sensitivity study of the luminosity estimates."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Ning,\n\nHere's my read of Wu et al. on the AT2019qiz IR echo.\n\nWhat's actually new: they apply the Dou et al./Jiang et al. torus-echo model to the first normal optical TDE with QPEs, and use it to argue for a parsec-scale dust structure (a faded AGN torus remnant) and for a bolometric peak well above the optical/UV value. That is a timely and interesting package, and they are open about the methodology being inherited from those earlier papers. The geometric light-travel argument is solid: with the last epoch at tau ~ 1350 days and a delta-function flare, R_min = c*tau_max/2 = 0.68 pc is a clean, parameter-free floor. That alone is enough to say the dust is parsec-scale and far beyond the ~0.1 pc sublimation radius for this SMBH. The observation that the temperature is roughly constant in the last five epochs is also a real constraint that favors a thin, well-defined inner edge over a distributed medium.\n\nWhere it gets shaky: the specific claim R > 1.2 pc. The stress-test note is right. They take the minimum R among the top 1% of MCMC samples ranked by likelihood. That is not a confidence interval or any standard bound. A profile-likelihood or HPD interval, accounting for the known degeneracy between R, alpha, beta, would likely push the lower limit back toward the geometric floor. The paper even acknowledges the degeneracy (\"two equations, three unknowns\"), so the 1.2 pc number should not be a headline. If the true bound is ~0.7-0.9 pc, the torus-remnant interpretation is still plausible (a factor >5 above the sublimation radius), but the missing-energy lower limits scale as R^2 and would drop by a factor ~0.3-0.6. The qualitative missing-energy claim survives; the quantitative values do not.\n\nOther soft spots, in minor order: they fit only the last five epochs and ignore the early bump, which they attribute to extra line-of-sight dust; fine as a caveat, but it means the model never explains the full light curve. The single-temperature, single-composition, single-size dust assumption is acknowledged, and Table 1 shows the geometric parameters are stable across compositions, which helps. The W1/W2-derived temperatures and luminosities are standard but only as good as the assumed Q_abs. No data or code release, which would help future work.\n\nWho this is for: TDE and QPE people, and anyone using IR echoes as bolometers. The central scenario — a parsec-scale dusty structure around a recently faded AGN, and a bolometric output significantly above the optical/UV — is probably right and worth saying. The paper's way of quantifying it is not. This deserves a serious referee, but the referee should ask for a proper likelihood-based bound on R and a fit to the full light curve (or an explicit argument for why the early bump cannot be part of the same torus). I'd send it to review, conditionally, and cite it for the geometric floor and the AT2019qiz-QPE connection, not for R > 1.2 pc.\n\nCheers.","headline":"AT 2019qiz's IR echo is a real parsec-scale story, but the R > 1.2 pc headline is an artifact of a shaky top-1% cutoff; the robust geometric floor is 0.68 pc, and the central conclusions survive in weakened form.","tokens_in":19503,"tokens_out":2493,"would_cite":true,"duration_ms":28803,"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":"The infrared echo of tidal disruption event AT 2019qiz is best explained by dust in a thin torus with inner radius greater than 1.2 parsecs, implying a remnant of a recently faded active galactic nucleus and revealing that most of the…","keywords":["tidal disruption events","infrared echoes","AGN torus","quasi-periodic eruptions","missing energy","dust reverberation","AT 2019qiz","extreme ultraviolet"],"falsifier":"If the mid-infrared light curve can be caught in decline, the turnover time gives an independent measure of the inner radius: for $R\\sim1.2\\,\\mathrm{pc}$ the model predicts the plateau will end roughly four years after the flare, so a much earlier or much later turnover would falsify the fitted torus geometry.","tokens_in":18386,"feed_emoji":"🌌","tokens_out":9342,"duration_ms":94561,"temperature":0.7,"pith_summary":"AT 2019qiz is a tidal disruption event—a star torn apart by a supermassive black hole—and the first ordinary optical example with detected X-ray quasi-periodic eruptions. Its mid-infrared light curve, from nine epochs of WISE data, rises slowly for years and then plateaus while the dust temperature stays nearly constant. The paper argues this pattern is a geometric light-travel-time echo from a thin dust torus, and fitting the last five epochs yields an inner radius greater than 1.2 parsecs—far beyond the roughly 0.1 parsec sublimation radius expected for a $\\sim 10^6\\,M_\\odot$ black hole. That places the dust in a remnant of a recently faded active galactic nucleus, supporting a unified picture in which faded AGNs produce both TDEs and QPEs. The same echo, treated as a bolometer, requires a peak flare luminosity above $6.6\\times10^{44}\\,\\mathrm{erg\\,s^{-1}}$ for silicate dust, far above the observed optical-UV peak, implying the missing energy was radiated in the extreme ultraviolet.","feed_headline":"Infrared echo places TDE dust 1.2 parsecs out","feed_subtitle":"AT 2019qiz's slow IR rise implies a remnant AGN torus and missing EUV energy","key_machinery":"The engine of the argument is a light-travel-time echo model of a thin, inclined dust torus. The flare is treated as a delta function in time, and the delay $\\tau = (R/c)(1-\\cos\\theta')$ controls which parts of the torus the observer sees illuminated at each epoch; as time passes, the iso-delay surface sweeps out an ever larger area of the torus, producing a slow rise and then a plateau. The observed flux is computed from the intersection area of the torus with the iso-delay surface, the Planck function at the measured dust temperature, and the grain absorption efficiency. With the temperature fixed by the W1/W2 color, the free parameters are the surface density, the inclination angle $\\alpha$, the half-opening angle $\\beta$, and the inner radius $R$, and the last five epochs are fitted with Markov Chain Monte Carlo sampling to give the $R > 1.2\\,\\mathrm{pc}$ lower limit and the required bolometric luminosity.","core_discovery":"The central discovery is that the unusually long infrared rise of AT 2019qiz is the echo of the optical/ultraviolet flare from a well-defined dust structure rather than a slowly brightening source. Modeling the last five epochs, during which the dust temperature is constant, with a thin, inclined torus gives a lower limit on the torus inner radius of $R \\gtrsim 1.2\\,\\mathrm{pc}$, consistent across silicate, silicon carbide, and graphite grains, with a hard light-travel-time floor of $0.68\\,\\mathrm{pc}$. For a $\\sim10^6\\,M_\\odot$ black hole the usual dust sublimation radius is at most $\\sim0.1\\,\\mathrm{pc}$, so the dust must be a leftover, or remnant, of a torus whose inner part has disappeared after the AGN faded. The same model demands a peak bolometric luminosity of at least $6.6\\times10^{44}$, $9.5\\times10^{44}$, and $1.0\\times10^{44}\\,\\mathrm{erg\\,s^{-1}}$ for silicate, silicon carbide, and graphite grains respectively, all well above the $2.7\\times10^{43}\\,\\mathrm{erg\\,s^{-1}}$ optical-UV blackbody peak—so most of the flare's energy was emitted in the extreme-ultraviolet band that the echo reprocesses.","pith_inferences":["The same light-travel-time reasoning could be applied to other slowly rising IR-bright TDEs to infer parsec-scale dust radii; if several show radii well beyond the sublimation radius, the recently faded AGN population would be established as common rather than exceptional.","If the missing extreme-ultraviolet energy is real, then optical-UV blackbody fits systematically underestimate TDE bolometric luminosities, which would shift TDE rates, Eddington ratios, and the luminosity function, implying that many apparently sub-Eddington TDEs are actually super-Eddington.","A testable extension is to search for extended emission-line regions and other faded-AGN signatures in the hosts of IR-bright TDEs; finding them in a larger sample would tie the torus remnant directly to the TDE/QPE sequence.","If the infrared echo peaks long after the optical flare, archival surveys such as WISE can be used retrospectively to identify TDEs that occurred before optical surveys caught them, effectively extending the TDE census."],"forward_implications":["If the torus remnant interpretation is right, AT 2019qiz shows that dusty tori survive for roughly $10^4$ years after an AGN switches off, with the inner dust falling inward and disappearing faster than the outer dust.","The echo-based peak luminosity lower limit ($\\gtrsim6.6\\times10^{44}\\,\\mathrm{erg\\,s^{-1}}$ for silicate dust) makes the peak of AT 2019qiz super-Eddington, with most energy hidden in the extreme ultraviolet, supporting the missing-energy solution for TDEs.","Because the only two optical TDEs with detected QPEs—AT 2019qiz and AT2022upj—are both infrared-bright with parsec-scale dust, the paper predicts that QPEs should preferentially follow IR-bright TDEs, and that monitoring such TDEs at X-ray wavelengths is an efficient way to find new QPEs.","The early infrared bump and the declining dust temperature in the first four epochs point to an additional dust component inside the torus, possibly along the line of sight, that the thin-ring model does not include.","Future mid-infrared observations of the decline phase, with NEO Surveyor or Roman, will turn the current lower limit on the inner radius into a precise measurement and constrain the opening angle and inclination."],"supporting_citations":[{"why":"Supplies the grain absorption efficiencies used to derive dust temperatures from the W1/W2 flux ratio.","marker":"Laor & Draine (1993)"},{"why":"Provides the convex dust ring echo model that the paper adapts for AT 2019qiz.","marker":"Dou et al. (2017)"},{"why":"Provides the optical-UV light curve, the peak blackbody luminosity of $2.7\\times10^{43}\\,\\mathrm{erg\\,s^{-1}}$, and the bolometric flare evolution used to convert pulse luminosity to peak luminosity.","marker":"Hammerstein et al. (2023)"},{"why":"Reports the detection of quasi-periodic eruptions in AT 2019qiz and the observed QPE luminosity used to argue against a QPE-powered infrared echo.","marker":"Nicholl et al. (2024)"},{"why":"Demonstrates the same torus echo methodology on the long-lived infrared echo of PS16dtm, showing such light curves can be produced without QPEs.","marker":"Jiang et al. (2025)"},{"why":"Detects the extended emission-line region in the host galaxy, providing evidence for a recently faded AGN that supports the torus remnant interpretation.","marker":"Xiong et al. (2025)"},{"why":"Proposes the unified scenario connecting faded AGNs, TDEs, and QPEs that this paper's conclusions support and extend.","marker":"Jiang & Pan (2025)"},{"why":"Gives the quasar dust time-lag versus bolometric luminosity relation used to argue that a normal AGN torus around a $10^6\\,M_\\odot$ black hole would have an inner radius near $0.1\\,\\mathrm{pc}$, making $1.2\\,\\mathrm{pc}$ anomalously large.","marker":"Lyu et al. (2019)"},{"why":"Establishes the infrared echo as a bolometer for the extreme-ultraviolet luminosity of tidal disruption events, motivating the missing-energy inference.","marker":"Lu & Kumar (2018)"}],"fun_headline_variants":["TDE's slow IR rise reveals a 1.2-pc torus remnant","TDE's infrared echo points to a faded AGN's torus remnant","Missing TDE energy may lurk in extreme UV, IR echo suggests","AT 2019qiz's IR echo hints at QPE link and missing energy","Remnant torus at 1.2 pc explains TDE's IR echo and missing flux"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the slow infrared rise is purely a geometric light-travel-time echo from a thin dust ring at a single constant temperature, with the optical/ultraviolet flare treated as an instantaneous pulse; if the heating source itself brightened gradually over years, or the emitting dust spans a range of radii and temperatures, the inferred 1.2 parsec inner radius would be too large, although the 0.68 parsec light-travel floor is more robust.","fun_headline_variants_meta":{"raw":{"variants":["TDE's slow IR rise reveals a 1.2-pc torus remnant","TDE's infrared echo points to a faded AGN's torus remnant","Missing TDE energy may lurk in extreme UV, IR echo suggests","AT 2019qiz's IR echo hints at QPE link and missing energy","Remnant torus at 1.2 pc explains TDE's IR echo and missing flux"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000558,"raw_usage":{"total_tokens":2787,"prompt_tokens":1212,"completion_tokens":1575,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":828,"completion_tokens_details":{"reasoning_tokens":1467}},"tokens_in":828,"tokens_out":1575,"duration_ms":11271,"temperature":1.0,"reasoning_tokens":1467,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:27:55.789288+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If the mid-infrared light curve can be caught in decline, the turnover time gives an independent measure of the inner radius: for $R\\sim1.2\\,\\mathrm{pc}$ the model predicts the plateau will end roughly four years after the flare, so a much earlier or much later turnover would falsify the fitted torus geometry.","supporting_citations":[{"cited_title":"Extended Emission-line Region in a Poststarburst Galaxy Hosting Tidal Disruption Event AT2019qiz and Quasiperiodic Eruptions","cited_arxiv_id":"2503.19722","evidence_quote":"Detects the extended emission-line region in the host galaxy, providing evidence for a recently faded AGN that supports the torus remnant interpretation."}],"review_version":1}