{"id":"1cdb7c57-48e8-444f-a793-8ae46459a0a3","arxiv_id":"1908.02496","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Stacked Fermi-LAT observations reveal a 32-sigma cumulative gamma-ray signal from 172 individually undetected extreme blazars, with an average photon index of 2.08 and a predicted contribution of at least 10% to the extragalactic gamma-ray background at 100 GeV.","lead":"Astronomers stacked gamma-ray data from 172 faint extreme blazars that Fermi-LAT could not detect individually and found a clear cumulative signal, equivalent to about 32 standard deviations. The result suggests these objects contribute at least 10% of the extragalactic gamma-ray background at 100 GeV and may guide targeting for the next-generation Cherenkov Telescope Array.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Known-redshift sample selection may bias the stacked average flux and the 10% EGB lower limit; the paper does not test sample representativeness.","rationale":"The paper's central methodological contribution, the likelihood-profile stacking pipeline, is convincingly validated by the empty-sky control (Section 4.1) and the simulation (Section 4.2). The detection of a cumulative signal from the 172 known-redshift undetected extreme blazars is statistically overwhelming (TS=1062) and unlikely to be a background artifact. The reader's weakest assumption correctly identifies the known-redshift selection as the point where the robust detection is extrapolated to a population-wide average flux and an EGB contribution. This concern is load-bearing because the abstract's 10% EGB claim would be invalid if the selected subsample is biased bright; however, it does not undermine the primary stacking detection or the method validation. Since the concern is already acknowledged as a caveat in the reader's verdict and the paper's footnote 7 admits the EGB estimate is approximate, I do not think it changes the accept verdict, but a restacking test on the full 3HSP extreme blazar sample would settle it definitively.","tokens_in":8789,"tokens_out":21693,"duration_ms":209542,"concrete_test":"Restack the 10-1000 GeV Fermi-LAT data using all extreme blazars in the 3HSP catalog with synchrotron peak above 1e17 Hz that are not individually detected, regardless of whether a redshift is known, and compare the best-fit average flux and TS with the 172-source known-redshift result. Also compare the redshift distributions of the stacked subsample and the full catalog. If the average flux shifts by more than the quoted 1-sigma statistical error, or if the known-redshift subsample skews toward z<0.35 while the full sample contains a significant high-z tail, then the selection bias is real and the 10% EGB lower limit should be re-derived with completeness corrections.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is that the 172 stacked sources are selected from the 337 extreme blazars with known redshift (Section 3), a subsample of the 3HSP catalog. Redshift determination requires optical spectroscopy, so the known-redshift subset is likely biased toward optically brighter, lower-redshift, and possibly gamma-ray-brighter objects. The stacked average flux F=6.5e-12 ph/cm2/s and the subsequent 'at least ~10% of the EGB at 100 GeV' claim (Section 5, footnote 7) assume this subsample represents the full undetected extreme blazar population. If the known-redshift subsample is systematically brighter, both the measured average flux and the derived EGB lower limit are inflated. The paper does not quantify the redshift completeness of the 3HSP catalog or test whether the 172 undetected sources are drawn from the same flux/redshift distribution as the full extreme blazar sample. This is the weakest link between the robust stacked detection and the headline astrophysical interpretation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a stacking technique for Fermi-LAT data that combines individual source likelihood profiles to measure the average gamma-ray properties of a population of unresolved sources. The method is applied to 172 gamma-ray-undetected extreme blazars selected from the 3HSP catalog with known redshift, yielding a cumulative detection with TS = 1062 (quoted as ~32σ for 2 dof), an average 10–1000 GeV flux of 6.51e-12 ph/cm2/s, and a photon index of 2.08. The authors validate the method with 172 empty-sky positions that show null stacked TS and with simulated sources whose input flux and index are recovered. They further compare the stacked spectra with CTA sensitivity limits, concluding that the unresolved population lies below CTA's detection threshold whereas the Fermi-LAT-detected extreme blazars are promising CTA targets. Finally, they use the average spectrum to estimate that unresolved extreme blazars contribute at least ~10% of the extragalactic gamma-ray background at 100 GeV.","tokens_in":8905,"tokens_out":4458,"duration_ms":51186,"significance":"If correct, the paper provides a computationally efficient and flexible stacking tool that extends Fermi-LAT reach to populations roughly an order of magnitude fainter than individual-source detection thresholds. The validation strategy is a strength: 172 empty-sky positions yield a stacked TS consistent with the null hypothesis, and simulated sources are recovered with input parameters, giving confidence that the method is not stacking diffuse background fluctuations. The result that extreme blazars form a genuinely unresolved GeV-emitting population and potentially contribute ~10% of the EGB at 100 GeV is astrophysically interesting. The work is also timely for CTA planning, as it identifies Fermi-LAT-detected extreme blazars as better targets than the currently unresolved population.","major_comments":[{"comment":"The stacked sample is built from 337 extreme blazars with known redshift drawn from the 2011-source 3HSP catalog. Redshift determination requires optical spectroscopy, so the known-redshift subset is plausibly biased toward optically brighter, lower-redshift, and possibly gamma-ray-brighter sources. The paper does not quantify the redshift completeness of 3HSP, nor does it compare the redshift/flux/synchrotron-peak distributions of the known-redshift subset with those of the full 3HSP extreme-blazar population. This is directly relevant to the headline claim that unresolved extreme blazars contribute 'at least ~10%' of the EGB at 100 GeV, because that estimate uses the average flux of the 172 stacked sources as if it represented the whole population. If the known-redshift subset is systematically brighter, the EGB lower limit is inflated. I request that the authors either (a) test representativeness by comparing the known-redshift sample with the full 3HSP sample in observable properties (e.g., radio flux, X-ray flux, synchrotron peak frequency, redshift distribution) and propagate any offset, or (b) explicitly restrict the EGB claim to the known-redshift subset and soften the population-level wording in the abstract and Section 5.","section":"Section 3, sample selection (also Section 5/footnote 7)"},{"comment":"The stacking pipeline assumes that the population is described by a single power law with a common flux and index, and the simulation validation in Section 4.2 draws simulated sources from exactly that model (Gaussian index around 2.1, log-normal flux). This confirms the method recovers input parameters under the assumed model, but it does not test how the technique behaves when the true population has a dispersion in spectral indices or spectral curvature (e.g., log-parabolic or EBL-absorbed spectra). Since the reported average index 2.08 is used to infer the position of the inverse-Compton peak and to compute the EGB contribution, it would be valuable to run a simulation with a realistic range of indices and with EBL attenuation to show that the stacked likelihood-profile method still recovers the flux-averaged index without significant bias. Such a test is within the scope of the paper and would strengthen the applicability of the method beyond the single-template case.","section":"Section 2.2 and Section 4.2"}],"minor_comments":[{"comment":"The conversion from TS = 1062 to 'more than 32σ' is only valid in the Gaussian limit with 2 dof approximately as sqrt(TS); the authors should state this explicitly or quote the chi-square tail probability instead of implying exact 32σ Gaussian significance.","section":"Abstract and Section 3"},{"comment":"The color-bar label 'TS (Extreme blazars)' is redundant and slightly confusing because the panel already shows TS contours; consider labeling it as 'TS' or removing it.","section":"Figure 1, right panel"},{"comment":"The text says the EGB contribution is derived from the average spectrum but does not specify whether the EBL-corrected or observed spectrum is used in that calculation; since the EGB is an observed quantity, the observed (intrinsic minus EBL absorption) spectrum should be used, and this should be clarified.","section":"Section 5 and Figure 2, right panel"},{"comment":"The threshold of TS > 25 for adding unmodeled sources is mentioned, but the paper does not state the resulting number of such excesses in the preprocessing of the 172 undetected sources; a brief note on how many tentative sources were added and then excluded would improve reproducibility.","section":"Section 2.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid methods contribution with convincing internal validation. The main concern is the unquantified selection bias introduced by restricting to known-redshift 3HSP sources, which directly affects the 10% EGB lower-limit claim. This is fixable with a representativeness test or a scaled-back claim, so I recommend major revision rather than rejection. The power-law validation issue is secondary but worth addressing in the same revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The stacking detection is real and the validation is the strongest part of the paper. Empty-sky positions give a null TS distribution and simulated sources recover their input flux and index, so the 32-sigma signal from 172 apparently undetected extreme blazars is credible. The method itself is a modest but legitimate extension of summed-likelihood stacking: generating per-source likelihood profiles independently and then summing them is flexible and parallelizable, and the paper explains the steps clearly enough to reproduce in principle.\n\nThe new astrophysical results are also worth having: an average 10–1000 GeV flux of 6.5e-12 ph/cm2/s with photon index 2.08, an EBL-corrected stacked spectrum that stays hard up to 1 TeV, and a rough lower limit of ~10% of the EGB at 100 GeV from this population. The CTA prospects discussion is sensible and appropriately cautious.\n\nThe soft spots are real but not fatal. The known-redshift selection (337 from 3HSP, of which 172 are undetected) could bias the sample toward optically brighter, lower-redshift, and possibly gamma-ray brighter objects. The paper does not quantify redshift completeness or test whether the undetected subset is representative of the full extreme blazar population. The \"at least ~10%\" EGB number is derived under a uniform-sky assumption and would be inflated if the known-redshift subsample is systematically brighter. That is a caveat on the interpretation, not on the detection itself. The single power-law assumption for the stacked spectrum is a simplification, though standard for this kind of analysis. Code is not shipped, which is minor for a Letter but worth noting.\n\nThis paper deserves peer review and publication. The central measurement is well validated, the method is useful, and the caveats can be addressed with a few sentences acknowledging the selection effect and its possible impact on the EGB lower limit.","headline":"A well-validated stacking analysis that delivers a robust cumulative detection of 172 undetected extreme blazars; the EGB lower limit carries a selection-bias caveat, but the central result holds.","tokens_in":9511,"tokens_out":2255,"would_cite":true,"duration_ms":25432,"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":"Stacking the likelihood profiles of 172 Fermi-LAT-undetected extreme blazars produces a cumulative gamma-ray detection at $32\\sigma$, with an average 10--1000 GeV flux of $6.51\\times10^{-12}$ ph cm$^{-2}$ s$^{-1}$ and photon index $2.08$.","keywords":["gamma-ray stacking","extreme blazars","Fermi-LAT","likelihood profile","extragalactic gamma-ray background","Cherenkov Telescope Array","BL Lacertae objects","EBL absorption"],"falsifier":"Run the same likelihood-profile stack on the full extreme-blazar catalog, including sources with photometric redshift limits or no redshift measurement. If the recovered average flux falls significantly below $6.51\\times10^{-12}$ ph cm$^{-2}$ s$^{-1}$ or the 100 GeV EGB share drops well below 10%, the known-redshift subsample is not representative.","tokens_in":8533,"feed_emoji":"🔭","tokens_out":7008,"duration_ms":70078,"temperature":0.7,"pith_summary":"This paper claims that a population of 172 extreme blazars not individually detected by Fermi-LAT nevertheless emits gamma rays as a class, and that the emission can be pulled out by stacking their individual likelihood profiles. The combined signal is detected at about $32\\sigma$, which the authors take as proof that the method works and that these unresolved objects are real GeV emitters. The stacked average spectrum is hard, and after correcting for extragalactic background light absorption it shows no softening up to 1 TeV, implying the inverse-Compton peak lies above that energy. Comparing the stacked flux with CTA sensitivity, the paper concludes that the unresolved population as a whole will probably stay below CTA's threshold, while Fermi-LAT-detected extreme blazars are the better targets. Finally, the paper estimates that at 100 GeV at least about 10% of the extragalactic gamma-ray background comes from these unresolved extreme blazars.","feed_headline":"Fermi-LAT stacking finds 32-sigma glow from 172 faint blazars","feed_subtitle":"The method recovers their average spectrum and links ~10% of the 100-GeV background to them.","key_machinery":"The method is likelihood-profile stacking. For each undetected source, the pipeline builds a two-dimensional grid of gamma-ray photon flux and photon index, fits the Fermi-LAT data at every grid point with all other model parameters fixed except the diffuse backgrounds, and subtracts the log-likelihood at the lowest flux to obtain a test-statistic profile. Because log-likelihoods are additive, the individual TS profiles are summed across the 172 sources and the location of the combined peak, found by spline fitting, gives the average flux and spectral index of the population. The paper validates the machinery against empty-sky positions and injected simulated sources.","core_discovery":"The central discovery is a cumulative gamma-ray signal from 172 Fermi-LAT-undetected extreme blazars: stacking their likelihood profiles yields TS = 1062, corresponding to more than $32\\sigma$ for 2 degrees of freedom. The best-fit average spectral parameters are $F_{10-1000\\,{\\rm GeV}} = 6.51^{+0.36}_{-0.35}\\times10^{-12}$ ph cm$^{-2}$ s$^{-1}$ and $\\Gamma = 2.08^{+0.07}_{-0.06}$. This signal is validated by stacking empty sky positions, which reproduce the null hypothesis, and by simulations of 100 injected sources, whose input flux and index are recovered. The authors use the average spectrum to argue that the unresolved extreme-blazar population contributes at least $\\sim$10% of the extragalactic gamma-ray background at 100 GeV, and that the EBL-corrected spectrum remains hard to 1 TeV, placing the inverse-Compton peak above that energy.","pith_inferences":["If the known-redshift requirement biases the 172-source sample toward brighter or lower-redshift objects, the true population average could be fainter than $6.5\\times10^{-12}$ ph cm$^{-2}$ s$^{-1}$, and the 10% EGB share could be an overestimate rather than a floor; the paper does not correct for this selection.","The same profile-stacking approach could be applied to other position-catalogued but gamma-ray-undetected classes, such as star-forming galaxies or galaxy clusters; the simulation test the paper performs for extreme blazars would need repeating for each class.","A testable extension is to stack the known extreme blazars in two redshift slices; if the average spectrum is redshift-independent after EBL correction, the claim of a hard intrinsic TeV spectrum is strengthened."],"forward_implications":["The unresolved population of extreme blazars is established as a genuine collective GeV source, so it must be counted in models of the gamma-ray sky and of the extragalactic gamma-ray background.","Taking the stacked spectrum at face value, CTA's 50-hour sensitivity is unlikely to detect the average unresolved extreme blazar; any CTA strategy should prioritise the 165 Fermi-LAT-detected extreme blazars.","The absence of EBL-corrected softening up to 1 TeV locates the inverse-Compton peak of extreme blazars above 1 TeV, implying intrinsically hard TeV spectra.","At 100 GeV, unresolved extreme blazars make up at least about 10% of the extragalactic gamma-ray background, so they cannot be ignored in background accounting.","The stacking pipeline can extract signals roughly an order of magnitude fainter than the Fermi-LAT point-source threshold, making it suitable for other unresolved astrophysical populations."],"supporting_citations":[{"why":"Supplies the 4FGL source catalog used to model the gamma-ray sky in each region of interest.","marker":"The Fermi-LAT collaboration 2019a"},{"why":"Provides the total extragalactic gamma-ray background measurements, including the three Galactic foreground models, against which the extreme-blazar contribution is compared.","marker":"Ackermann et al. 2015"},{"why":"Supplies the EBL absorption model used to correct the stacked spectra and infer the intrinsic TeV spectrum.","marker":"Domínguez et al. 2011"},{"why":"Presents an earlier Fermi-LAT stacking method based on co-added count maps, against which the new likelihood-profile approach is positioned.","marker":"Huber et al. 2012"},{"why":"Provides fermiPy, the likelihood-analysis tool used in the pre-processing step to generate source models and TS maps.","marker":"Wood et al. 2017"},{"why":"Defines the test statistic used for detection significance and for the stacked TS profiles.","marker":"Mattox et al. 1996"},{"why":"Supplies the 3HSP catalog from which the 337 extreme blazars with known redshift are selected.","marker":"Chang et al. 2019"},{"why":"Defines the CTA sensitivity curves used to judge the detectability of the stacked extreme-blazar population.","marker":"Actis et al. 2011"}],"fun_headline_variants":["Stacking Fermi data: 32-sigma signal from 172 faint blazars","Unseen extreme blazars glow at 32-sigma in stacked Fermi data","Fermi stacking uncovers 32-sigma emission from extreme blazars","32-sigma stacked signal reveals hard TeV spectra in extreme blazars","Stacking analysis exposes extreme blazars and CTA prospects"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 172 unresolved sources with known redshifts are representative of all extreme blazars; if the redshift requirement systematically excludes faint or distant members of the class, the measured average flux and the derived 10% contribution to the gamma-ray background would be biased upward.","fun_headline_variants_meta":{"raw":{"variants":["Stacking Fermi data: 32-sigma signal from 172 faint blazars","Unseen extreme blazars glow at 32-sigma in stacked Fermi data","Fermi stacking uncovers 32-sigma emission from extreme blazars","32-sigma stacked signal reveals hard TeV spectra in extreme blazars","Stacking analysis exposes extreme blazars and CTA prospects"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000273,"raw_usage":{"total_tokens":1704,"prompt_tokens":1084,"completion_tokens":620,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":700,"completion_tokens_details":{"reasoning_tokens":518}},"tokens_in":700,"tokens_out":620,"duration_ms":6397,"temperature":1.0,"reasoning_tokens":518,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:42:20.906192+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same likelihood-profile stack on the full extreme-blazar catalog, including sources with photometric redshift limits or no redshift measurement. If the recovered average flux falls significantly below $6.51\\times10^{-12}$ ph cm$^{-2}$ s$^{-1}$ or the 100 GeV EGB share drops well below 10%, the known-redshift subsample is not representative.","supporting_citations":[{"cited_title":"2015, ApJ, 799, 86 —","cited_arxiv_id":null,"evidence_quote":"Provides the total extragalactic gamma-ray background measurements, including the three Galactic foreground models, against which the extreme-blazar contribution is compared."},{"cited_title":"2012, A&A, 547, A102","cited_arxiv_id":null,"evidence_quote":"Presents an earlier Fermi-LAT stacking method based on co-added count maps, against which the new likelihood-profile approach is positioned."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the 3HSP catalog from which the 337 extreme blazars with known redshift are selected."}],"review_version":1}