{"id":"4cb40f8c-edf7-4ae5-8c9e-093e4e12c6ca","arxiv_id":"1908.02746","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The inclusive isolated-photon production cross section at 13 TeV has been measured up to 2.5 TeV and matches NLO and NNLO QCD predictions within uncertainties.","lead":"The ATLAS experiment measured the production rate of high-energy isolated photons in proton-proton collisions at 13 TeV using a tenfold larger data sample than before. The results reach up to 2.5 TeV photon energy and test next-to-next-to-leading-order quantum chromodynamics predictions for the first time in ATLAS.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"NNLO 'excellent description' rests on an unquantified 3.6% αem-scheme shift relative to the NLO predictions.","rationale":"The reader's identified weak assumption (Rbg=1) is indeed an assumption, but the paper directly tests it in validation regions and assigns an explicit systematic uncertainty (<2.5%, typically <1%; Section 7). Because the signal purity is above 90% in all bins, even a 40% deviation in Rbg translates into a small shift in the signal yield, and the assigned uncertainty covers it. The stress-test pass found a more load-bearing issue on the theory side. The NNLO prediction, which the paper highlights as the precision benchmark, is computed with αem in the Gμ scheme (1/132.232), whereas the NLO predictions use αem(0)=1/137.036. Since the cross section is linear in αem, this is a flat 3.6% normalization shift. The NNLO scale uncertainties are 0.6%–5%, so this convention choice is of the same size as the dominant quoted theoretical uncertainty. The paper does not mention or quantify this difference when comparing Nnlojet with Jetphox/Sherpa or when claiming excellent agreement. A simple rerun with αem(0) would establish whether the central claim survives; if the NNLO prediction moves to a systematic underestimate, the conclusion should be softened. The suggested check is cheap and decisive. The paper is otherwise careful and the measurement itself appears sound, so the recommendation is conditional acceptance pending this quantitative check.","tokens_in":51196,"tokens_out":16663,"duration_ms":197572,"concrete_test":"Recompute the Nnlojet NNLO prediction with αem = 1/137.036 instead of 1/132.232, keeping all other inputs (scales, NNPDF3.1, hybrid isolation, NP corrections) fixed, and regenerate the ratio plots in Figures 11 and 12. If the NNLO/data ratio drops by ~3.6% and the points move systematically outside the quoted total uncertainty, the 'excellent description' claim is not robust to the αem convention. As a cross-check, compute the Nnlojet NLO prediction with both αem values and compare with Jetphox at the same αem to separate the genuine NNLO QCD correction from the coupling scheme.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The most load-bearing unquantified assumption is not the Rbg=1 background correlation (tested in validation regions and assigned a <2.5% systematic in Section 7), but the electromagnetic-coupling scheme used for the headline NNLO prediction. In Section 8, Nnlojet uses αem = 1/132.232 (Gμ scheme), while Jetphox and Sherpa use αem = 1/137.036. The isolated-photon cross section is proportional to αem at every perturbative order, so this single choice changes the Nnlojet prediction by a flat 3.6% relative to the other predictions. The NNLO scale uncertainty is only 0.6%–5% (Section 8.2), so the αem convention is comparable to, or larger than, the dominant quoted NNLO theory uncertainty. No uncertainty or cross-check for this scheme choice is reported. The paper's conclusion that the NNLO prediction gives an 'excellent description' (Section 10) therefore rests on an unquantified 3.6% normalization convention; switching to αem = 1/137.036 would lower the prediction by about 3.6%, likely degrading the central-region agreement and worsening the already noted underestimate in 1.56<|η|<1.81. The measurement itself is not affected, but the theory-data consistency claim is not robust to this choice.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a measurement of the inclusive isolated-prompt-photon differential cross section in proton-proton collisions at sqrt(s)=13 TeV using 36.1 fb^-1 of ATLAS data. Photon candidates with E_T>125 GeV and |eta|<2.37 (excluding the calorimeter transition region 1.37<|eta|<1.56) are selected with tight identification and an isolation requirement. The dominant background from jets misidentified as photons is subtracted with a two-dimensional sideband technique in the isolation-tightness plane; the residual electron background is estimated from simulation. Background-subtracted yields are unfolded to particle level with bin-by-bin correction factors computed with Pythia, with Sherpa used for a model-dependence systematic and an iterative Bayesian unfolding as a cross-check. The cross sections are presented as functions of E_T in four |eta| regions and as a function of |eta| in E_T bins. The results are compared with NLO QCD predictions from JETPHOX and SHERPA 2.2.2 and with an NNLO QCD prediction from NNLOJET using several PDF sets. The authors conclude that the predictions give a good description of the data within uncertainties and that the NNLO prediction provides an excellent description except for a tendency to underestimate the data in 1.56<|eta|<1.81.","tokens_in":51466,"tokens_out":8822,"duration_ms":101028,"significance":"If the measurement and comparisons are taken at face value, the paper provides a substantially improved 13 TeV isolated-photon dataset: the ten-fold luminosity increase extends the systematics-dominated region to about 1 TeV and pushes the E_T reach to 2.5 TeV in the central region. The measurement itself is largely model-independent: signal yields are extracted from data and unfolded with generic MC generators, while the tested predictions come from independent programs with externally fitted PDFs. The experimental analysis is careful: the data-driven background estimate is validated in sideband regions, the Rbg=1 assumption is tested and assigned a systematic uncertainty, 76 photon energy-scale components are propagated individually, and alternative unfolding and generator choices are used as cross-checks. The main caveat concerns the normalization convention of the NNLO prediction: the G_mu scheme for alpha_em introduces an unquantified 3.6% shift relative to the on-shell value used by the other predictions. Since the NNLO scale uncertainty is 0.6%-5%, this convention materially affects the strength of the 'excellent description' conclusion.","major_comments":[{"comment":"Section 8 and Table 2 state that the NNLOJET prediction uses alpha_em = 1/132.232 (G_mu scheme) while JETPHOX and SHERPA 2.2.2 use 1/137.036. The isolated-photon cross section in the direct-dominated phase space of this measurement is proportional to alpha_em at each perturbative order, so this convention changes the NNLOJET prediction by a flat +3.6% relative to the other predictions. The quoted NNLO scale uncertainty is only 0.6%-5% (Section 8.2), so the alpha_em scheme choice is comparable to or larger than the dominant quoted NNLO theory uncertainty, yet no cross-check or uncertainty is assigned to it. The conclusion in Section 10 that the NNLO prediction 'gives an excellent description of the data' therefore depends on this convention; a prediction with alpha_em = 1/137.036 would lie lower by about 3.6% and could shift ratio points out of the quoted theory band and worsen the already-noted underestimate in 1.56<|eta|<1.81. I request that the authors either repeat the NNLO calculation with the on-shell alpha_em value, or add an explicit alpha_em scheme uncertainty and state whether the agreement conclusions in Sections 9 and 10 remain unchanged.","section":"Section 8, Table 2; Section 10"},{"comment":"The PDF and alpha_s uncertainties for the NNLOJET prediction are not computed directly; they are approximated by those estimated at NLO with JETPHOX, which uses a different PDF set (MMHT2014) and a different perturbative order from the NNPDF3.1 NNLO set used in NNLOJET. This approximation is understandable, but because the NNLO scale uncertainty is much smaller than at NLO, the approximate PDF/alpha_s component may not be conservative, and the text does not say how large this component is relative to the quoted total NNLO uncertainty. Please add a sentence clarifying the magnitude of the approximate PDF/alpha_s contribution, or provide a direct estimate with the NNPDF3.1 set, so that readers can judge which uncertainty dominates the NNLO band in Figures 11 and 12.","section":"Section 8.2; Figures 11-12"}],"minor_comments":[{"comment":"The statement that the hadronisation and underlying-event uncertainty of the SHERPA 2.2.2 prediction 'can be neglected' is based on an expectation; since this prediction is compared directly at particle level, a short numerical justification or a reference to a tune variation would strengthen the claim.","section":"Section 8.1"},{"comment":"The table caption could state explicitly that the 125<E_T<150 GeV event yields use only 2015 data, while the E_T>150 GeV yields use 2015+2016 data; the text says this, but a footnote would make the table self-contained.","section":"Section 4, Table 1"},{"comment":"The definitions of regions A, B, C and D appear only in the surrounding prose; adding them to the caption of Eq. (1) or to a small table would improve readability.","section":"Section 5, Eq. (1)"},{"comment":"The wording 'excellent description' is stronger than the abstract's 'good description'; given the alpha_em scheme issue and the known underestimate in 1.56<|eta|<1.81, I suggest tempering this wording unless the alpha_em cross-check requested in the major comments is added.","section":"Section 10"}],"recommendation":"major_revision","confidential_remarks":"The paper is generally strong and the measurement is likely publishable after revision. My main concern is the alpha_em scheme used for the NNLOJET prediction; it can be addressed without new data-taking, but it is important for the paper's central theory-comparison claim. I would not recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nWhat you should know: this is a careful, well-executed measurement from ATLAS, and the data are solid. The one thing to flag is that the headline “good description by NNLO QCD” rests on a choice of electromagnetic coupling that is not cross-checked.\n\nThe genuinely new content: 36 fb^-1 of 13 TeV data, extending the photon ET reach to 2.5 TeV, with the first ATLAS comparison to NNLOJET predictions. The systematic treatment is detailed: 76 energy-scale components, data-driven background with validation regions for R_bg, two unfolding methods, and a full breakdown of the experimental uncertainties. The measurement itself is model-independent in the right way: signal yields are extracted from data, and the MC generators are only used for efficiencies and unfolding, with variations taken as systematics.\n\nThe soft spot is in the theory comparison, not the measurement. In Section 8, NNLOJET uses alpha_em = 1/132.232 (G_mu scheme), while JETPHOX and SHERPA use 1/137.036. The isolated-photon cross section is linear in alpha_em, so the NNLO prediction is a flat 3.6% higher than it would be with the other scheme. The quoted NNLO scale uncertainty is 0.6%–5%, so this convention is about the same size as the largest theory error. The paper says the NNLO prediction gives an “excellent description” of the data, but that conclusion is sensitive to this choice. Lowering the prediction by 3.6% would push the central value below the data in several regions, especially at forward rapidity. The authors do not mention or quantify this dependence.\n\nI don’t think this is a deal-breaker. The measurement stands; the NNLO comparison is still useful, and the discrepancy is within the combined uncertainties. But it is a genuine oversight that a referee would want fixed: either use the same alpha_em as the other predictions or add an alpha_em-scheme uncertainty to the theory band. It would strengthen the paper to state this choice explicitly in the summary.\n\nThe R_bg=1 assumption is tested and assigned a <2.5% systematic, so that’s fine. The only other minor omission is not having tables in the preprint, but that is standard practice with HEPData. The citation pattern looks appropriate, and the comparison to previous ATLAS results is properly referenced.\n\nWho is this for: anyone working on QCD precision measurements, PDF fits, or the phenomenology of direct photon production. It earns a serious referee slot, and a good referee would catch the alpha_em point. My recommendation: engage with it, and if you write a related paper, cite the data but treat the NNLO normalization with care.","headline":"Solid ATLAS isolated-photon measurement with a real but unquantified alpha_em-scheme shift in the NNLO theory comparison; the data are trustworthy, the theory caveat deserves a note.","tokens_in":51976,"tokens_out":3762,"would_cite":true,"duration_ms":38476,"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 13 TeV isolated-photon cross section matches QCD predictions.","keywords":["isolated photon production","inclusive cross section","proton-proton collisions","LHC","ATLAS","perturbative QCD","next-to-next-to-leading-order","parton distribution functions"],"falsifier":"Measure $R_{\\text{bg}}$ directly in a sample of candidate events from which prompt photons have been removed by an independent method, such as a template fit to the electromagnetic shower-shape distribution, and check whether it deviates from 1 by more than the assigned systematic uncertainties; a larger deviation would shift the extracted cross sections and break the claimed agreement with QCD predictions.","tokens_in":50994,"feed_emoji":"⚛️","tokens_out":5968,"duration_ms":60664,"temperature":0.7,"pith_summary":"This paper reports the ATLAS measurement of the inclusive isolated-photon cross section in proton-proton collisions at $\\sqrt{s}=13$ TeV using 36.1 fb$^{-1}$ of data, covering photon transverse energies from 125 GeV to 2.5 TeV in several pseudorapidity regions. The authors compare their measurement with next-to-leading-order QCD predictions from JETPHOX and SHERPA and with a next-to-next-to-leading-order prediction from NNLOJET, using several parton distribution functions. The central claim is that all predictions describe the data within experimental and theoretical uncertainties, with the NNLO prediction reproducing the data in most kinematic regions. If true, this confirms that perturbative QCD accurately describes hard photon production at the highest accessible energies and strengthens the case for using these data to constrain the proton's gluon density.","feed_headline":"Isolated-photon rates at 13 TeV match QCD predictions","feed_subtitle":"ATLAS data from 36.1 fb^-1 agree with next-to-next-to-leading-order QCD from 125 GeV to 2.5 TeV.","key_machinery":"The central object is the two-dimensional sideband, or ABCD, method used to extract the prompt-photon signal from the dominant background of jets misidentified as photons. Photon candidates are divided into four regions defined by tight versus non-tight identification and isolated versus non-isolated transverse energy; the signal yield in the signal region is obtained from Eq. (1) with the background correlation $R_{\\text{bg}}$ set to 1, meaning isolation and identification are assumed uncorrelated for background events. Bin-by-bin unfolding correction factors from Monte Carlo simulation convert the background-subtracted yields to particle-level cross sections, while the theoretical comparisons rely on fixed-cone and hybrid-cone photon-isolation prescriptions and on scale-variation uncertainties for the predictions.","core_discovery":"The paper establishes that the differential cross section for isolated prompt-photon production in $pp$ collisions at $\\sqrt{s}=13$ TeV, measured as a function of $E_\\gamma^T$ in four $|\\eta^\\gamma|$ regions and as a function of $|\\eta^\\gamma|$ in several $E_\\gamma^T$ ranges, is well described by NLO QCD calculations (JETPHOX, SHERPA) and by an NNLO QCD calculation (NNLOJET). The NNLO prediction, which has substantially reduced scale uncertainties, provides a good description of the data except for a tendency to underestimate in the region $1.56<|\\eta^\\gamma|<1.81$. The measurement extends the region where experimental uncertainties dominate to about 1 TeV in $E_\\gamma^T$ and significantly improves both statistical and systematic precision over the previous ATLAS 13 TeV measurement.","pith_inferences":["If the $R_{\\text{bg}}=1$ assumption were replaced by an $R_{\\text{bg}}$ value measured directly from a dedicated background-enriched sample, the background-subtraction systematic could shrink and sharpen the PDF constraining power.","The mild NNLO underestimation in the $1.56<|\\eta^\\gamma|<1.81$ region could point to missing electroweak corrections or to the treatment of isolation in the forward calorimeter transition region; a dedicated calculation including full electroweak effects would test this.","Applying the same analysis strategy to the full Run 2 dataset (about 140 fb$^{-1}$) should extend the reach in $E_\\gamma^T$ and make the photon energy scale the dominant systematic, providing a sharper test of QCD at the TeV scale."],"forward_implications":["The NNLO QCD prediction for isolated-photon production at high transverse energy is validated, so future comparisons in this channel can rely on the smaller NNLO scale uncertainties.","The measured cross sections can be included in global parton distribution function fits, with the potential to tighten the gluon density in the proton.","The extension of the systematic-limited region to about 1 TeV means that additional LHC data can probe the TeV scale where electroweak corrections are expected to become relevant.","The agreement between the data and the Sherpa ME+PS@NLO prediction supports the use of matched matrix-element-plus-parton-shower calculations for photon production.","The consistency between the nominal ABCD extraction and the validation regions constrains the size of any background-correlation bias in the signal yields."],"supporting_citations":[{"why":"Supplies the NNLO QCD prediction from NNLOJET that is the central next-to-next-to-leading-order comparison.","marker":"[16]"},{"why":"Provide the JETPHOX NLO QCD calculation for the direct and fragmentation photon production cross sections.","marker":"[19, 20]"},{"why":"Supplies the SHERPA event generator used both for Monte Carlo signal simulation and for the ME+PS@NLO prediction.","marker":"[21]"},{"why":"The previous ATLAS 13 TeV isolated-photon measurement that this analysis extends with a ten-fold larger data sample.","marker":"[5]"},{"why":"The earlier ATLAS 8 TeV measurement that established the analysis methodology and the isolation requirement used here.","marker":"[4]"},{"why":"Defines Frixione's isolation criterion used in the theoretical predictions to avoid collinear divergences.","marker":"[37]"},{"why":"Provides the measured photon identification efficiencies used to derive the scale factors applied to simulated events.","marker":"[46]"},{"why":"Determines the integrated luminosity and its 2.1% uncertainty that normalizes the measured cross section.","marker":"[26]"}],"fun_headline_variants":["ATLAS photon cross section agrees with NNLO QCD at 13 TeV","Photon production at 13 TeV: NNLO QCD predictions pass ATLAS test","ATLAS measures isolated photons, QCD predictions hold to NNLO","Precise photon rates from ATLAS confirm QCD at 13 TeV","ATLAS extends photon cross-section measurement, QCD fits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The signal yields assume that, for the jet background, whether a photon candidate passes the tight identification is unrelated to whether it is isolated, so the background correlation $R_{\\text{bg}}$ equals 1; if that is false, the background subtraction and every cross-section bin inherit a bias.","fun_headline_variants_meta":{"raw":{"variants":["ATLAS photon cross section agrees with NNLO QCD at 13 TeV","Photon production at 13 TeV: NNLO QCD predictions pass ATLAS test","ATLAS measures isolated photons, QCD predictions hold to NNLO","Precise photon rates from ATLAS confirm QCD at 13 TeV","ATLAS extends photon cross-section measurement, QCD fits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000385,"raw_usage":{"total_tokens":2018,"prompt_tokens":907,"completion_tokens":1111,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":523,"completion_tokens_details":{"reasoning_tokens":1011}},"tokens_in":523,"tokens_out":1111,"duration_ms":11272,"temperature":1.0,"reasoning_tokens":1011,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:35:18.981602+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure $R_{\\text{bg}}$ directly in a sample of candidate events from which prompt photons have been removed by an independent method, such as a template fit to the electromagnetic shower-shape distribution, and check whether it deviates from 1 by more than the assigned systematic uncertainties; a larger deviation would shift the extracted cross sections and break the claimed agreement with QCD predictions.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the measured photon identification efficiencies used to derive the scale factors applied to simulated events."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Determines the integrated luminosity and its 2.1% uncertainty that normalizes the measured cross section."}],"review_version":1}