⚛️ high-energy experiments

Measurements of isolated photon pair production in association with jets in pp collisions at $\sqrt{s}$ = 8 TeV with the ATLAS detector

Using 8 TeV proton-proton collision data collected by the ATLAS detector, this paper presents integrated and differential cross-section measurements for isolated photon pair production with associated jets, comparing the results to various theoretical predictions and finding that NLO-matched Sherpa 2.2.14 provides the best overall description of the data.

Original author: ATLAS Collaboration

Published 2026-11-05

📖 4 min read🧠 Deep dive

CC BY 4.0

Original author: ATLAS Collaboration

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

In the heart of the Standard Model, the theory that describes the fundamental building blocks of our universe, light behaves in a peculiar way. While light usually travels alone as a stream of particles called photons, the intense energy of particle collisions can sometimes force two of them to appear together, born from the same violent event. Physicists call this "diphoton" production. It is a rare occurrence, but when it happens alongside other particles like jets—sprays of debris created when quarks or gluons are knocked loose—it offers a unique window into the forces that hold matter together. Specifically, these events allow scientists to test the rules of the strong nuclear force, known as quantum chromodynamics, under extreme conditions. If the predictions of our current theories match what is seen in nature, the model holds; if they diverge, it hints that our understanding is incomplete and that new physics might be waiting to be discovered.

To test these theories, a researcher using the ATLAS detector at the Large Hadron Collider in Geneva examined a massive collection of data from proton collisions that occurred in 2012. They focused on a specific scenario: two isolated photons appearing in the same event as one or more jets. The researcher sifted through 20.3 units of collision data, a volume so large it represents a significant fraction of the total energy recorded by the machine that year. Their goal was to measure exactly how often these events happened and to map out the details of how the photons and jets were arranged in space and energy. They categorized the events based on how many jets were present, ranging from zero jets up to three or more, and measured the energy and direction of every particle involved. This level of detail allowed them to compare their real-world observations against the most advanced computer simulations available.

The researcher found that the universe is more complex than the simplest computer models predict. When they compared their measurements to older simulations that only accounted for the most basic interactions, those models failed to capture the full picture, often underestimating how many events occurred or misrepresenting the angles at which the particles flew apart. However, when they turned to newer, more sophisticated calculations that included higher levels of complexity, the agreement improved significantly. The best match came from a specific simulation that combined detailed mathematical calculations with a model of how particles shower and evolve after a collision. This advanced model successfully described the data across almost all categories, from events with no jets to those with several.

Despite this success, the study revealed that even the best current theories struggle in specific situations. As the number of jets in an event increased, and particularly when looking at the angles between the particles, the predictions began to drift away from the actual measurements. The discrepancies were most noticeable in the way the particles were oriented relative to one another, suggesting that the calculations are missing some higher-order effects that become important when many particles are produced at once. The researcher also noted that while the overall number of events was predicted reasonably well, the details of how the energy was distributed among the jets showed small but consistent gaps between theory and reality. These gaps are not errors in the experiment but rather signs that the mathematical tools used to describe the strong force need further refinement to handle the most crowded and energetic collisions.

The measurements provided a rigorous check on the current state of particle physics. By confirming that the most advanced simulations work well for simpler events but start to falter as complexity grows, the study highlights exactly where future theoretical work needs to focus. The data serves as a precise benchmark, showing that while our understanding of how light and matter interact is robust, it is not yet perfect. The results confirm that the Standard Model holds up under scrutiny but also point clearly toward the need for more complete calculations to fully explain the behavior of particles in the most extreme environments created by human hands.

Technical Summary: Measurements of isolated photon pair production in association with jets in $pp$ collisions at $\sqrt{s}=8$ TeV with the ATLAS detector

Problem and Motivation
The production of two isolated prompt photons in association with jets ( $\gamma\gamma + \text{jets}$ ) in proton-proton collisions serves as a critical testing ground for perturbative Quantum Chromodynamics (pQCD). While fixed-order calculations are available at next-to-leading order (NLO) for processes involving multiple jets, these calculations can struggle to accurately predict overall cross-section normalizations and the full kinematic dynamics, particularly in regions with high jet multiplicity or specific angular configurations (e.g., back-to-back photons or diphoton systems recoiling against a single jet). Furthermore, resummation techniques are often required for specific phase-space regions. Monte Carlo (MC) generators offering NLO matrix elements matched to parton showers (NLO+PS) are expected to provide a more comprehensive description of both normalization and kinematics. This paper aims to provide a rigorous comparison of experimental data against various theoretical predictions to validate and guide the improvement of these MC simulations.

Methodology
The analysis utilizes the 2012 dataset from the ATLAS experiment at the Large Hadron Collider (LHC), corresponding to an integrated luminosity of $20.3 \text{ fb}^{-1}$ at a center-of-mass energy of $\sqrt{s} = 8$ TeV.

Event Selection: Events are selected requiring two isolated prompt photons with transverse energy $E_T^\gamma > 22$ GeV and pseudorapidity $|\eta_\gamma| < 1.37$ or $1.56 < |\eta_\gamma| < 2.37$ . Jets are reconstructed using the anti- $k_t$ algorithm ( $R=0.4$ ) with $p_T^j > 25$ GeV for $|\eta_j| < 2.4$ and $p_T^j > 50$ GeV for $2.4 < |\eta_j| < 4.4$ .
Background Decomposition: The signal yield is extracted using a data-driven 4 $\times$ 4 matrix method to separate prompt diphotons ( $\gamma\gamma$ ) from backgrounds involving fake photons from jets ( $\gamma j, j\gamma, jj$ ). A separate data-driven method estimates the reducible background from electrons misidentified as photons.
Unfolding: Measured yields are unfolded to the particle level using Singular Value Decomposition (SVD) to correct for detector effects, including trigger efficiency, reconstruction efficiency, and pileup. The unfolding relies on Sherpa 1.4.1 MC samples.
Theoretical Comparisons: The measured fiducial cross sections (integrated and differential) are compared against:
- Sherpa 1.4.1: Leading-order (LO) predictions matched to parton showers (used for background estimation and normalization).
- GoSam and BlackHat: Fixed-order NLO calculations.
- Sherpa 2.2.14: NLO-matched predictions (NLO matrix elements for $\gamma\gamma+0,1$ jets merged with LO for $\gamma\gamma+2,3$ jets).

Key Contributions
The paper presents a comprehensive differential analysis of isolated diphoton production in association with jets at $\sqrt{s}=8$ TeV. Key contributions include:

Integrated and Differential Cross Sections: Measurements are provided for exclusive and inclusive jet categories ( $N_j = 0, 1, 2, \geq 3$ ) across sixteen observables. These include diphoton kinematics ( $m_{\gamma\gamma}, p_T^{\gamma\gamma}, |\cos\theta^*_{\gamma\gamma}|, \Delta\phi_{\gamma\gamma}$ ), jet kinematics, dijet variables ( $m_{jj}, p_T^{jj}, \Delta y_{jj}, \Delta\phi_{jj}$ ), and photon-jet separations ( $\Delta R_{\gamma j}$ ).
Systematic Uncertainty Evaluation: A detailed breakdown of uncertainties is provided, identifying the definition of non-tight control regions for fake-photon background subtraction and jet energy scale as dominant systematic sources in most regions.
Comprehensive Theory Benchmarking: The results offer a stringent test of NLO fixed-order calculations (GoSam, BlackHat) and NLO+PS matched generators (Sherpa 2.2.14) across a wide phase space.

Results

Integrated Cross Sections: The measured fiducial cross sections range from $28.5$ pb (0-jet) to $0.82$ pb ( $\geq 3$ -jet). Fixed-order NLO predictions (GoSam, BlackHat) are generally compatible with data within uncertainties, though they tend to underestimate the central values (e.g., GoSam is 11% and 9% below data for 1-jet and 2-jet categories, respectively). Sherpa 2.2.14 shows good agreement within scale uncertainties, with minor underestimations in the 0-jet (9%) and 2-jet (15%) exclusive categories.
Differential Distributions:
- Sherpa 2.2.14 provides the best overall description of the data across all jet categories and observables, generally remaining within 15% of the data over the bulk of the distributions.
- Fixed-order NLO (GoSam/BlackHat) provide reasonable descriptions in many regions and are compatible with data within uncertainties, but exhibit significant discrepancies as jet multiplicity increases, particularly for angular variables. For instance, while compatible within uncertainties, they show trends of underestimation in the $p_T^{\gamma\gamma}$ distribution at high values and deficits in dijet invariant mass ( $m_{jj}$ ) at high masses.
- Sherpa 1.4.1 (LO) generally underestimates the integrated cross sections (except in the $\geq 3$ -jet category where it coincidentally matches) but often describes the shapes of distributions reasonably well when normalized to data.
Specific Discrepancies: All theoretical models show a progressive undershoot at large $|\cos\theta^*_{\gamma\gamma}|$ values and a slight deficit at the highest jet transverse momenta. Sherpa 2.2.14 underestimates data at small $\Delta\phi_{\gamma\gamma}$ separations, a trend also seen in lower jet multiplicities.

Significance
The paper claims that the NLO predictions provide a reasonable description of the data in most regions, but significant discrepancies observed for increasing jet multiplicity and specific angular variables highlight the limitations of fixed-order calculations due to the absence of higher-order terms. The results demonstrate that NLO-matched predictions (specifically Sherpa 2.2.14) offer the most accurate description of the process kinematics and normalization simultaneously. Consequently, these measurements constitute a benchmark for future improvements to NLO-matched event generators and provide a stringent test of perturbative QCD predictions for diphoton production in association with jets.

More like this