Measurements of isolated photon pair production in association with jets in pp collisions at sqrt s = 8 TeV with the ATLAS detector
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.
Kai: We're looking at the paper: "Measurements of isolated photon pair production in association with jets in pp collisions at sqrt s = 8 TeV with the ATLAS detector".
Mira: Here's the gist of the paper. Measurements of integrated and differential production cross sections for two isolated photons in events with and without associated hadronic jets are presented, based on the 2012 dataset from ATLAS collected at sqrt s = 8 TeV with an integrated luminosity of 20.3 fb-1 in proton–proton collisions.
Kai: The analysis considers photons with specific pseudorapidities (eta gamma 22 GeV).
Mira: Jets are selected based on their pseudorapidity (e.g., p jT > 25 GeV for eta j 50 GeV for 2.4 < eta j < 4.4).
Kai: Events are categorized by the number of jets (N j) into exclusive and inclusive categories, and fiducial cross sections are measured for each jet category, which are further measured differentially across sixteen observables depending on N j.
Paper summary: Kai: So we're looking at this paper today, "Measurements of isolated photon pair production in association with jets in pp collisions at sqrt s = eight TeV with the ATLAS detector <ref:2611.06142#pg1,Measurements of isolated photon pair production in association with jets in>." Basically, they took data from the ATLAS experiment at CERN, specifically from twenty point three femtobarns of proton-proton collisions at an energy of eight teratesla <ref:2611.06142#pg1,of proton-proton collisions at>.
Mira: Right, and what they're doing is measuring how often you can produce two isolated photons together with some associated hadronic jets. The main point they are making is testing the predictions from perturbative quantum chromodynamics, or pQCD.
Lev: For us, the interesting part is that they're looking at this process in association with jets because it's a tough place to calculate using pQCD because of things like back-to-back photon production or when the diphoton system recoils against just one jet <ref:2611.06142#pg2>.
Kai: Exactly, and they are using these measurements as a tool to test those theoretical predictions, specifically fixed-order next-to-leading order calculations for multiple jets <ref:2611.06142#pg3>. The thesis is that these measurements help establish good theoretical knowledge for this process in association with jets.
Mira: And the way they set up their measurements is important because they consider photons with specific pseudorapidities, like those under two point two or one point five six to two point three seven <ref:2611.06142#pg1>. They also require jets to have a certain transverse energy, like more than twenty two giga-electron volts for the photon energy times its transverse momentum <ref:2611.06142#pg1>.
Lev: From an engineering standpoint, it's interesting how they are categorizing these events by the number of jets into exclusive and inclusive groups <ref:2611.06142#pg3>. That way, they can measure fiducial cross sections for each category separately.
Kai: And what they present is that these fiducial cross sections are differential across sixteen observables depending on how many jets there are in the event <ref:2611.06142#pg1>. That means they're looking at a lot of different ways to describe the production process, from the invariant mass of the photons to things involving jet separations.
Mira: The key result they highlight is that fixed-order next-to-leading order calculations from frameworks like GoSam and BlackHat describe the data reasonably well <ref:2611.06142#pg3>. Specifically, GoSam next-to-leading order is compatible within uncertainties in both one jet and two jet categories <ref:2611.06142#pg3>.
Lev: That compatibility level is what we really need to consider for real hardware; it tells us how much theoretical uncertainty we can actually tolerate before the data becomes unusable <ref:2611.06142#pg3>. But they also compare this against the Sherpa two point two two one four next-to-leading order predictions, which they say offer the best overall description across all jet categories and observables <ref:2611.06142#pg3>.
Kai: And what's really striking is that Sherpa one point four point one point one prediction significantly underestimates the measured integrated cross sections across all jet categories except for the three or more jet inclusive category, which they say is coincidentally good <ref:2611.06142#pg3>.
Paper summary: Mira: That underestimation points to some limitations in how those fixed-order calculations handle the full dynamics of the process <ref:2611.06142#pg2>. And they also flag some common issues in all these theoretical predictions <ref:2611.06142#pg3>.
Lev: They mention a progressive undershoot at large cosine theta star gamma gamma values across every category, and there's a slight deficit at the highest jet transverse momenta too <ref:2611.06142#pg3>. Plus, they note an underestimate at large dijet rapidity separations in the two jet category <ref:2611.06142#pg3>.
Kai: And they see an underestimate at small phi gamma gamma separations consistently across multiple jet multiplicities <ref:2611.06142#pg3>. The measurement uncertainty itself is spread from ten percent up to sixteen percent across all the jet categories <ref:2611.06142#pg3>.
Mira: So, while they have these measurements and these comparisons, the paper also lays out where the theoretical models are falling short <ref:2611.06142#pg3>. They are comparing several different theoretical tools to see which one matches what they actually measured from the twenty point three femtobarn dataset <ref:2611.06142#pg3>.
Lev: It’s important to remember that for real hardware, we need to know how much overhead those NLO matched matrix elements add on top of the fixed-order ones <ref:2611.06142#pg3>. That's where the practical constraints really come in when you try to simulate these kinds of events <ref:2611.06142#pg3>.
Kai: So, we have this paper that uses data from eight teratesla collisions at CERN to measure isolated photon pair production with jets <ref:2611.06142#pg1>. It's a detailed look at how well different theoretical calculations match what the ATLAS detector actually saw <ref:2611.06142#pg3>.
Mira: And the conclusion of this paper is that while some tools, like Sherpa two point two two one four, provide a good overall description across all observables, there are clear gaps where they fall short <ref:2611.06142#pg3>.
Lev: The implications for us is that if we were to try and run this kind of process on real hardware, we'd have to account for those systematic uncertainties up to sixteen percent <ref:2611.06142#pg3>. We can't just rely on the best-looking prediction without checking those specific regions where they undershoot <ref:2611.06142#pg3>.
Kai: So, in simple terms, this paper is about measuring how often you get two photons and jets together at high energy <ref:2611.06142#pg3>. It shows that the math works reasonably well for one or two jets with some modern tools, but there are specific kinematic regions where even the best models struggle to keep up with the data <ref:2611.06142#pg3>.
Mira: And this paper really pushes us toward needing better theoretical frameworks that can handle those complex jet environments more accurately <ref:2611.06142#pg3>. It’s about understanding the full dynamics of these interactions, not just the average cross section <ref:2611.06142#pg2>.
Lev: I think the main thing is that as we build better error correction for real systems, we need to know this kind of precision from theory so we can set realistic goals for what our hardware can actually achieve <ref:2611.06142#pg3>. That's where the measurement and the theory have to meet up <ref:2611.06142#pg3>.
Conclusion: Kai: So we're wrapping up this look at those measurements of isolated photon pair production with jets in proton proton collisions at eight teratesla energy.
Mira: This paper, by Kai and Mira, is essentially comparing what the ATLAS detector saw against a bunch of different theoretical calculations.
Lev: It’s about testing those next-to-leading order predictions for how photons and jets are made together.
Kai: The core idea is showing that while some models fit the data well, others have specific problems when you look at certain things, like the angular variables.
Mira: They found that fixed-order calculations from groups like GoSam and BlackHat actually describe a lot of the data reasonably well across one jet and two jet categories.
Lev: But they also pointed out where those models fall short, especially with the angular variables when you have more jets involved.
Kai: The authors flag some consistent issues across all these calculations, like a progressive undershoot at large cosine theta star gamma gamma values.
Mira: They also noted a slight deficit at the highest jet transverse momenta and an underestimation of dijet rapidity separations in the two jet category.
Lev: From my side, that underestimation in those separation measurements is where the real trouble starts when you try to map this onto actual hardware simulations.
Kai: So, what does this mean for us? It means we need theoretical tools that can handle these complex jet environments more accurately than the current best fixed-order calculations.
Mira: They are pushing us toward needing better frameworks that capture the full dynamics of these interactions, not just an average cross section number.
Lev: That’s exactly right, because if we want to build reliable error correction for real hardware, we need to know those specific regions where the theory undershoots.
Kai: So, it boils down to making sure our theoretical models match what the experimental setup actually measured across all those sixteen observables.
Mira: And that comparison is crucial for understanding the underlying physics of how these particles interact at high energy scales.
Lev: Next time, we’ll look closer at how those systematic uncertainties translate into real overhead when you try to run these kinds of processes on actual computing hardware.
The ATLAS Collaboration
EUROPEAN ORGANISATION FOR NUCLEAR RESEARCH (CERN)
hep-ex
Submitted: 2026-11-04
Updated: 2026-11-04
Comments: 56 pages in total, author list starting page 39, 17 figures, 4 tables, submitted to Phys. Rev. D. All figures including auxiliary figures are available at https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/STDM-2016-12/
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 77/100
The gist: Measurements of integrated and differential production cross sections for two isolated photons in events with and without associated hadronic jets are presented…
Key concepts
- Isolated Photons
- These are photons detected in collisions that do not have nearby hadronic activity or jets associated with them. They are used as clean probes to study specific hard scattering processes involving only the photon pair.
- Jets ($N_j$ categories)
- The analysis categorizes events based on the number of associated jets, ranging from exclusive (few) to inclusive (many). This allows researchers to see how the production rate changes depending on whether one or many hadronic sprays are present in the collision.
- Cross Sections and Predictions
- Cross sections measure the probability of a specific particle interaction occurring. The paper compares these measured cross sections with predictions from different theoretical models (like NLO calculations) to test their accuracy against real-world experimental data.
Terminology
Summary
Measurements of integrated and differential production cross sections for two isolated photons in events with and without associated hadronic jets are presented, based on the 2012 dataset from ATLAS collected at sqrt s = 8 TeV with an integrated luminosity of 20. p. 0 D CERN-EP-2026-280
November 5, 2026
Measurements of isolated photon pair production in
association with jets in 𝒑 𝒑 collisions at √
𝒔 = 8 TeV
with the ATLAS detector
The ATLAS Collaboration
Measurements of integrated and differential production cross section of two isolated photons
in events with and without associated hadronic jets based on the 2012 dataset of 20. 3 fb−1
integrated luminosity of proton–proton collisions at √
𝑠 = 8 TeV collected by the ATLAS
experiment are presented. D CERN-EP-2026-280
November 5, 2026
Measurements of isolated photon pair production in
association with jets in 𝒑 𝒑 collisions at √
𝒔 = 8 TeV
with the ATLAS detector
The ATLAS Collaboration
Measurements of integrated and differential production cross section of two isolated photons
in events with and without associated hadronic jets based on the 2012 dataset of 20. 3 fb−1
integrated luminosity of proton–proton collisions at √
𝑠 = 8 TeV collected by the ATLAS
experiment are presented. 3 fb−1
integrated luminosity of proton–proton collisions at √
𝑠 = 8 TeV collected by the ATLAS
experiment are presented. The measurements consider photons with pseudorapidities of
|𝜂
𝛾
| < 1. At small radii, the inner detector (ID) is made up of fine-granularity pixel and microstrip
detectors. These silicon-based detectors cover the pseudorapidity range |𝜂| < 2. 56 < |𝜂
𝛾
| < 2. 37 and transverse energy of 𝐸
𝛾
T
> 22 GeV. Diphoton
events are categorized according to the number of jets (𝑁
j
) into exclusive and inclusive 𝑁
j
categories
and fiducial cross sections are measured for each jet category. Fiducial cross sections are also measured
differentially in up to sixteen observables, depending on the number of jets present in an event.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The NLO predictions provide a reasonable description of the data overall, although significant discrepancies emerge as jet multiplicity increases, particularly concerning angular variables. p. 0 14 in all categories. Overall, the NLO predictions
provide a reasonable description of the data, with significant discrepancies observed for
increasing jet multiplicity, particularly for angular variables, attributable to the absence of
higher-order terms in the calculations. Events are classified
into categories based on the number of jets and integrated fiducial cross sections of diphoton
production in association with inclusive and exclusive number of jets are presented. Fiducial
cross sections, differential in sixteen diphoton, jet, dijet and photon–jet observables, are
presented and compared with leading-order Sherpa 1. The Sherpa 2. 14 NLO-matched predictions provide
the best overall description of the data across all jet categories and observables. This
allows for a larger phase space, providing a larger data sample for the measurement that is split into jet
categories. The measured cross sections are compared with fixed-order NLO predictions from the GoSam
and BlackHat frameworks, to LO predictions from Sherpa 1. 3 fb−1
is presented. Diphoton
events are categorized according to the number of jets (𝑁
j
) into exclusive and inclusive 𝑁
j
categories
and fiducial cross sections are measured for each jet category.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Common limitations observed in the theoretical predictions include a progressive undershoot at large theta* gamma gamma values across all categories, a slight deficit at the highest jet transverse momenta, an underestimate at large dijet rapidity separations (delta y jj) in the two-jet category, and an underestimate at small delta phi gamma gamma separations consistently observed across multiple jet multiplicities. The total measurement uncertainty spans 10% to 16% across jet categories. p. 0 The Sherpa 2. 14 NLO-matched predictions provide
the best overall description of the data across all jet categories and observables. 4 Measurement strategy
Fiducial cross sections are measured integrated per jet category and also differentially in up to sixteen
observables. They are measured as a function of diphoton kinematic and angular properties: the diphoton
invariant mass 𝑚𝛾𝛾, the transverse momentum of the diphoton system 𝑝
𝛾𝛾
T
, the opening angle between the
photons in the azimuthal plane Δ𝜙𝛾𝛾 and | cos 𝜃
∗
𝛾𝛾 | defined in the Collins–Soper frame [40]. Dijet kinematic variables, namely the
dijet invariant mass 𝑚jj, the transverse momentum of the dijet system 𝑝
jj
T
, the opening angle between the two
highest-𝑝T jets in the azimuthal plane Δ𝜙jj and their difference in rapidity Δ𝑦jj are measured for events with
at least two jets. The azimuthal angle between the diphoton and dijet systems, Δ𝑆𝛾𝛾,jj ≡ |Δ𝜙𝛾𝛾,jj| ∈ [0, 𝜋],
is also measured.Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI can perform NLO-matched predictions from Sherpa 2.2.14 for diphoton production in association with up to three jets across all jet categories and observables, providing a
best overall description of the data across all jet categories and observables.
-
The improved system can utilize Singular Value Decomposition (SVD) unfolding, generalized to double differential cross sections with regularization based on adjacent bins in the measured observable, to account for detector effects.
-
The AI can employ a figure of merit combining statistical uncertainties and expected biases using sub-linear scaling to choose the regularization strength k for SVD unfolding.
-
The system can quantify systematic uncertainties by varying control regions in the fake-photon background subtraction method, such as changing the number of criteria removed from the tight selection definition.
-
The AI can estimate non-perturbative corrections (NPC) to fixed-order predictions using dedicated Sherpa 1.4.1 samples with and without hadronization and underlying event mechanisms via a Rivet routine.
-
The system can perform differential cross section comparisons across various theoretical frameworks, including NLO fixed-order pQCD from GoSam and BlackHat, as well as NLO-matched predictions from Sherpa 2.2.14 for exclusive jet categories.
-
The AI can analyze angular observables like | cos theta star gamma gamma | and Delta phi gamma gamma for the inclusive jet categories to identify significant discrepancies attributable to the absence of higher-order terms in calculations.
Abstract
Measurements of integrated and differential production cross section of two isolated photons in events with and without associated hadronic jets based on the 2012 dataset of 20.3 fb-1 integrated luminosity of proton-proton collisions at sqrt s =8 TeV collected by the ATLAS experiment are presented. The measurements consider photons with pseudorapidities of η γ<1.37 or 1.56<η γ<2.37 and transverse energy of E T γ>22 GeV. Jets are required to have p T j>25 GeV for η j<2.4 and p T j>50 GeV for 2.4<η j<4.4. Events are classified into categories based on the number of jets and integrated fiducial cross sections of diphoton production in association with inclusive and exclusive number of jets are presented. Fiducial cross sections, differential in sixteen diphoton, jet, dijet and photon-jet observables, are presented and compared with leading-order Sherpa 1.4.1 predictions in all jet categories, fixed-order next-to-leading-order GoSam and BlackHat calculations in selected categories, and NLO-matched predictions from Sherpa 2.2.14 in all categories. Overall, the NLO predictions provide a reasonable description of the data, with significant discrepancies observed for increasing jet multiplicity, particularly for angular variables, attributable to the absence of higher-order terms in the calculations. The Sherpa 2.2.14 NLO-matched predictions provide the best overall description of the data across all jet categories and observables.
Sources
- All-orders resummation for diphoton production at hadron colliders
- Automated One-Loop Calculations with GoSam
- Next-to-Leading Order Gamma Gamma + 2-Jet Production at the LHC
- Event generation with SHERPA 1.1
- Diphoton production at the LHC: a QCD study up to NNLO
- NNLO QCD corrections to diphoton production with an additional jet at the LHC
- Measurement of direct photon pair production cross sections in ppbar collisions at sqrt(s)=1.96 TeV
- Measurement of the differential cross sections for isolated direct photon pair production in $p \bar p$ collisions at $\sqrt{s} = 1.96$ TeV
- Measurement of the cross section for prompt isolated diphoton production using the full CDF Run II data sample
- Measurement of the isolated diphoton cross-section in pp collisions at sqrt(s) = 7 TeV with the ATLAS detector
- Measurement of isolated-photon pair production in pp collisions at sqrt(s) = 7 TeV with the ATLAS detector
- Measurement of the production cross section of pairs of isolated photons in $pp$ collisions at 13 TeV with the ATLAS detector
- Measurement of the Production Cross Section for Pairs of Isolated Photons in pp collisions at sqrt(s) = 7 TeV
- Measurement of differential cross sections for the production of a pair of isolated photons in pp collisions at sqrt(s) = 7 TeV
- Observation of the electroweak production of $\gamma\gamma jj$ at $\sqrt{s}=13$ TeV in $140\ \text{fb}^{-1}$ of $pp$ collision data with the ATLAS detector
- Measurements of integrated and differential cross sections for isolated photon pair production in $pp$ collisions at $\sqrt{s}=8$ TeV with the ATLAS detector
- Performance of the ATLAS Trigger System in 2010
- Luminosity determination in pp collisions at $\sqrt{s}$ = 8 TeV using the ATLAS detector at the LHC
- New parton distributions for collider physics
- QCD Matrix Elements + Parton Showers
Related papers
- DrSai: an agentic AI system toward autonomous physics analysis workflows at BESIII
- CANTON- mu Proposal: A Next-Generation Muon g-2 Measurement at Sub-0.1 ppm Precision
- Search for dark photon dark matter near 1,meV using a superconducting mixer
- ANUBIS detector and its sensitivity to neutral long-lived particles
- How to scale your HEP ML models: A recipe for robust architecture comparisons at scale
- Search for the 16 O(ppp) to 13 C pi+ pi+ e+ Decay Mode in Super-Kamiokande Using Machine Learning Techniques