Large Hadron Collider Research Articles

End-to-end codesign of Hessian-aware quantized neural networks for FPGAs

We develop an end-to-end workflow for the training and implementation of co-designed neural networks (NNs) for efficient field-programmable gate array (FPGA) hardware. Our approach leverages Hessian-aware quantization (HAWQ) of NNs, the Quantized Open Neural Network Exchange (QONNX) intermediate representation, and the hls4ml tool flow for transpiling NNs into FPGA firmware. This makes efficient NN implementations in hardware accessible to nonexperts, in a single open-sourced workflow that can be deployed for real-time machine-learning applications in a wide range of scientific and industrial settings. We demonstrate the workflow in a particle physics application involving trigger decisions that must operate at the 40 MHz collision rate of the CERN Large Hadron Collider (LHC). Given the high collision rate, all data processing must be implemented on FPGA hardware within the strict area and latency requirements. Based on these constraints, we implement an optimized mixed-precision NN classifier for high-momentum particle jets in simulated LHC proton-proton collisions.

Dip-Bump Structure in Proton’s Single Diffractive Dissociation at the Large Hadron Collider

By extending the dipole Pomeron (DP) model, successful in describing elastic nucleon–nucleon scattering, to proton single diffractive dissociation (SD), we predict a dip-bump structure in the squared four-momentum transfer (t) distribution of proton’s SD. Structures in the t distribution of single diffractive dissociation are predicted around t=−4GeV2 at LHC energies in the range of 3 GeV2≲|t|≲ 7 GeV2. Apart from the dependence on s (total energy squared) and t (squared momentum transfer), we predict also a dependence on missing masses. We include the minimum set of Regge trajectories, namely the Pomeron and the Odderon, indispensable at the LHC. Further generalization, e.g., by the inclusion of non-leading Regge trajectories, is straightforward. The present model contains two types of Regge trajectories: those connected with t-channel exchanges (the Pomeron, the Odderon, and non-leading (secondary) reggeons) appearing at small and moderate −t, where they are real and nearly linear, as well as direct-channel trajectories α(M2) related to missing masses. In this paper, we concentrate on structures in t neglecting (for the time being) resonances in M2.

Production of fully-heavy tetraquark states through the double parton scattering mechanism in pp and pA collisions

The production of fully-heavy tetraquark states in proton–proton (pp) and proton–nucleus (pA) collisions at the center-of-mass energies of the Large Hadron Collider (LHC) and at the Future Circular Collider (FCC) is investigated considering that these states are produced through the double parton scattering mechanism. We estimate the cross sections for the T4c,\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$T_{4c},$$\\end{document}T4b\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$T_{4b}$$\\end{document} and T2b2c\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$T_{2b2c}$$\\end{document} states and present predictions for pp, pCa and pPb collisions considering the rapidity ranges covered by central and forward detectors. We demonstrate that the cross sections for pA collisions are enhanced in comparison to the pp predictions scaled by the atomic number. Moreover, our results indicate that a search of these exotic states is, in principle, feasible in the future runs of the LHC and FCC.

Distilling particle knowledge for fast reconstruction at high-energy physics experiments

Knowledge distillation is a form of model compression that allows artificial neural networks of different sizes to learn from one another. Its main application is the compactification of large deep neural networks to free up computational resources, in particular on edge devices. In this article, we consider proton-proton collisions at the High-Luminosity Large Hadron Collider (HL-LHC) and demonstrate a successful knowledge transfer from an event-level graph neural network (GNN) to a particle-level small deep neural network (DNN). Our algorithm, DistillNet, is a DNN that is trained to learn about the provenance of particles, as provided by the soft labels that are the GNN outputs, to predict whether or not a particle originates from the primary interaction vertex. The results indicate that for this problem, which is one of the main challenges at the HL-LHC, there is minimal loss during the transfer of knowledge to the small student network, while improving significantly the computational resource needs compared to the teacher. This is demonstrated for the distilled student network on a CPU, as well as for a quantized and pruned student network deployed on an field programmable gate array. Our study proves that knowledge transfer between networks of different complexity can be used for fast artificial intelligence (AI) in high-energy physics that improves the expressiveness of observables over non-AI-based reconstruction algorithms. Such an approach can become essential at the HL-LHC experiments, e.g. to comply with the resource budget of their trigger stages.

Dark matter searches at accelerators

The understanding of the nature of dark matter is one of the most pressing questions begging for an answer in fundamental physics today. In this contribution we will review the searches performed recently at accelerators, in particular the ones at the Large Hadron Collider.

A low-latency graph computer to identify metastable particles at the Large Hadron Collider for real-time analysis of potential dark matter signatures

Image recognition is a pervasive task in many information-processing environments. We present a solution to a difficult pattern recognition problem that lies at the heart of experimental particle physics. Future experiments with very high-intensity beams will produce a spray of thousands of particles in each beam-target or beam-beam collision. Recognizing the trajectories of these particles as they traverse layers of electronic sensors is a massive image recognition task that has never been accomplished in real time. We present a real-time processing solution that is implemented in a commercial field-programmable gate array using high-level synthesis. It is an unsupervised learning algorithm that uses techniques of graph computing. A prime application is the low-latency analysis of dark-matter signatures involving metastable charged particles that manifest as disappearing tracks.

Analysis of the Variety of the Relativistic Approach Velocity of Objects Based on the Data of the Large Hadron Collider

Analysis of the Variety of the Relativistic Approach Velocity of Objects Based on the Data of the Large Hadron Collider

A search for R-parity-violating supersymmetry in final states containing many jets in pp collisions at s\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\sqrt{s} $$\\end{document} = 13 TeV with the ATLAS detector

A search for R-parity-violating supersymmetry in final states with high jet multiplicity is presented. The search uses 140 fb−1 of proton-proton collision data at s\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\sqrt{s} $$\\end{document} = 13 TeV collected by the ATLAS experiment during Run 2 of the Large Hadron Collider. The results are interpreted in the context of R-parity-violating supersymmetry models that feature prompt gluino-pair production decaying directly to three jets each or decaying to two jets and a neutralino which subsequently decays promptly to three jets. No significant excess over the Standard Model expectation is observed and exclusion limits at the 95% confidence level are extracted. Gluinos with masses up to 1800 GeV are excluded when decaying directly to three jets. In the cascade scenario, gluinos with masses up to 2340 GeV are excluded for a neutralino with mass up to 1250 GeV.

Machine learning classification of sphalerons and black holes at the LHC

In models with large extra dimensions, “miniature” black holes (BHs) might be produced in high-energy proton–proton collisions at the Large Hadron Collider (LHC). In the semi-classical regime, those BHs thermally decay, giving rise to large-multiplicity final states with jets and leptons. On the other hand, similar final states are also expected in the production of electroweak sphaleron/instanton-induced processes. We investigate whether one can discriminate these scenarios when BH or sphaleron-like events are observed in the LHC using machine learning (ML) methods. Classification among several BH scenarios with different numbers of extra dimensions and the minimal BH masses is also examined. In this study we consider three ML models: XGBoost algorithms with (1) high- and (2) low-level inputs, and (3) a Residual Convolutional Neural Network. In the latter case, the low-level detector information is converted into an input format of three-layer binned event images, where the value of each bin corresponds to the energy deposited in various detector subsystems. We demonstrate that only a small number of detected events are sufficient to effectively discriminate between the sphaleron and BH processes. Separation between BH scenarios with different minimal masses is possible with an order of 10 events passing the preselection. A sufficient number of events could be observed in combined Run-2 and -3 data, if the production cross section is not much smaller than the present limit ∼0.1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sim 0.1$$\\end{document} fb. We find, however, that a large number of events is needed to discriminate between BH hypotheses with the same minimal BH mass, but different numbers of extra dimensions.

Reconstruction of unstable heavy particles using deep symmetry-preserving attention networks

Reconstructing unstable heavy particles requires sophisticated techniques to sift through the large number of possible permutations for assignment of detector objects to the underlying partons. An approach based on a generalized attention mechanism, symmetry preserving attention networks (SPA-NET), has been previously applied to top quark pair decays at the Large Hadron Collider which produce only hadronic jets. Here we extend the SPA-NET architecture to consider multiple input object types, such as leptons, as well as global event features, such as the missing transverse momentum. In addition, we provide regression and classification outputs to supplement the parton assignment. We explore the performance of the extended capability of SPA-NET in the context of semi-leptonic decays of top quark pairs as well as top quark pairs produced in association with a Higgs boson. We find significant improvements in the power of three representative studies: a search for tt¯H\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$t\\bar{t}H$$\\end{document}, a measurement of the top quark mass, and a search for a heavy Z′\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${Z}^{{\\prime} }$$\\end{document} decaying to top quark pairs. We present ablation studies to provide insight on what the network has learned in each case.

Nuclear Parton Distribution Functions After the First Decade of LHC Data

We present a review of the conceptual basis, current knowledge, and recent progress regarding global analysis of nuclear parton distribution functions (PDFs). After introducing the theoretical foundations and methodological approaches for the extraction of nuclear PDFs from experimental data, we discuss how different measurements in fixed-target and collider experiments provide increasingly precise constraints on various aspects of nuclear PDFs, including shadowing, antishadowing, the EMC effect, Fermi motion, flavor separation, deuteron binding, and target-mass and other higher-twist effects. Particular emphasis is given to measurements carried out in proton–lead collisions at the Large Hadron Collider, which have revolutionized the global analysis during the past decade. These measurements include electroweak boson, jet, light hadron, and heavy flavor observables. Finally, we outline the expected impact of the future Electron Ion Collider and discuss the role and interplay of nuclear PDFs with other branches of nuclear, particle, and astroparticle physics.

Searching for lepton flavor violation with the CMS experiment

Searches for Lepton Flavor Violation (LFV) stand at the forefront of experimental particle physics research, offering a sensitive probe to many scenarios of physics beyond the Standard Model. The high proton–proton collision energy and luminosity provided by the CERN Large Hadron Collider (LHC) and the excellent CMS detector performance allow for an extensive program of LFV searches. This paper reviews a broad range of LFV searches conducted at the CMS experiment using data collected in LHC Run 2, including [Formula: see text]  decays, Higgs boson decays, and top quark production and decays. In each analysis, the online and offline event selections, signal modeling, background suppression and estimation, and statistical interpretation are elucidated. These searches involve various final state particles in a large transverse momentum range, showcasing the capability of the CMS experiment in exploring fundamental questions in particle physics.

Searching for a heavy neutral CP-even Higgs boson in the BLSSM at the LHC Run 3 and HL-LHC

The detection of a heavy neutral CP-even Higgs boson of the B-L\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$B-L$$\\end{document} Supersymmetric Standard Model (BLSSM), h′\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$h'$$\\end{document}, with mh′≃400GeV\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$m_{h'}\\simeq 400~\ ext {GeV}$$\\end{document}, at the Large Hadron Collider (LHC) for a center-of-mass energy of s=14TeV\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sqrt{s}=14~\ ext {TeV}$$\\end{document}, is investigated. The following production and decay channels are considered: gg→h′→ZZ→4ℓ\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$gg\\rightarrow h'\\rightarrow {ZZ}\\rightarrow 4\\ell $$\\end{document} and (with being the Missing Transverse Energy (MET)), where ℓ=e,μ\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\ell =e,\\mu $$\\end{document}, with integrated luminosity Lint=300fb-1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$L_{\ ext {int}}=300~{\ ext {fb}}^{-1}$$\\end{document} (Run 3). Furthermore, we also look into the di-Higgs channel gg→h′→hh→bb¯γγ\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$gg\\rightarrow h'\\rightarrow {hh}\\rightarrow {b{\\bar{b}}\\gamma \\gamma }$$\\end{document} at the High-Luminosity LHC (HL-LHC) with an integrated luminosity of Lint=3000fb-1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$L_{\ ext {int}}=3000~{\ ext {fb}}^{-1}$$\\end{document}. We demonstrate that promising signals with high signal-to-background statistical significance (S/B\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$S/\\sqrt{B}$$\\end{document}) can be obtained through the three aforementioned channels.

Nanosecond anomaly detection with decision trees and real-time application to exotic Higgs decays

We present an interpretable implementation of the autoencoding algorithm, used as an anomaly detector, built with a forest of deep decision trees on FPGA, field programmable gate arrays. Scenarios at the Large Hadron Collider at CERN are considered, for which the autoencoder is trained using known physical processes of the Standard Model. The design is then deployed in real-time trigger systems for anomaly detection of unknown physical processes, such as the detection of rare exotic decays of the Higgs boson. The inference is made with a latency value of 30 ns at percent-level resource usage using the Xilinx Virtex UltraScale+ VU9P FPGA. Our method offers anomaly detection at low latency values for edge AI users with resource constraints.

Non-relativistic quantum chromodynamics in parton showers

Measurements of quarkonia isolation in jets at the Large Hadron Collider (LHC) have been shown to disagree with fixed-order non-relativistic quantum chromodynamics (NRQCD) calculations, even at higher orders. Calculations using the fragmenting jet function formalism are able to better describe data but cannot provide full event-level predictions. In this work we provide an alternative model via NRQCD production of quarkonia in a timelike parton shower. We include this model in the Pythia 8 event generator and validate our parton-shower implementation against analytic forms of the relevant fragmentation functions. Finally, we make inclusive predictions of quarkonia production for the decay of the standard-model Higgs boson.

Quantum collider probes of the fermionic Higgs portal

We explore the sensitivity of future hadron colliders to constrain the fermionic Higgs portal, with a focus on scenarios where the new fermions cannot be directly observed in exotic Higgs decays. This portal emerges in various models including twin-Higgs scenarios and dark matter models, posing significant challenges for collider tests. Working in an effective field theory (EFT), we determine the reach of the high-luminosity option of the Large Hadron Collider (HL-LHC), the high-energy upgrade of the LHC (HE-LHC) and a proposed Future Circular Collider (FCC) in probing the fermionic Higgs portal through off-shell and double-Higgs production. Notably, we find that quantum-enhanced indirect probes offer a better sensitivity than other direct Higgs measurements. We argue that this finding is valid in a wide class of ultraviolet realisations of the EFT. Our study presents a roadmap of a multifaceted search strategy for exploring the fermionic Higgs portal at forthcoming hadron machines.

Combination of searches for heavy spin-1 resonances using 139 fb−1 of proton-proton collision data at s\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\sqrt{s} $$\\end{document} = 13 TeV with the ATLAS detector

A combination of searches for new heavy spin-1 resonances decaying into different pairings of W, Z, or Higgs bosons, as well as directly into leptons or quarks, is presented. The data sample used corresponds to 139 fb−1 of proton-proton collisions at s\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\sqrt{s} $$\\end{document} = 13 TeV collected during 2015–2018 with the ATLAS detector at the CERN Large Hadron Collider. Analyses selecting quark pairs (qq, bb, tt¯\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ t\\overline{t} $$\\end{document}, and tb) or third-generation leptons (τν and ττ) are included in this kind of combination for the first time. A simplified model predicting a spin-1 heavy vector-boson triplet is used. Cross-section limits are set at the 95% confidence level and are compared with predictions for the benchmark model. These limits are also expressed in terms of constraints on couplings of the heavy vector-boson triplet to quarks, leptons, and the Higgs boson. The complementarity of the various analyses increases the sensitivity to new physics, and the resulting constraints are stronger than those from any individual analysis considered. The data exclude a heavy vector-boson triplet with mass below 5.8 TeV in a weakly coupled scenario, below 4.4 TeV in a strongly coupled scenario, and up to 1.5 TeV in the case of production via vector-boson fusion.

Machine learning techniques for intermediate mass gap lepton partner searches at the large hadron collider

Machine learning techniques for intermediate mass gap lepton partner searches at the large hadron collider

Revisiting a Realistic Intersecting D6-Brane with Modified Soft SUSY Terms

Because there are a few typos in the supersymmetry-breaking sfermion masses and trilinear soft term, regarding the current Large Hadron Collider (LHC) and dark matter searches, we revisit a three-family Pati–Salam model based on intersecting D6-branes in Type IIA string theory on a T6/(Z2×Z2) orientifold with a realistic phenomenology. We study the viable parameter space and discuss the spectrum consistent with the current LHC Supersymmetry searches and the dark matter relic density bounds from the Planck 2018 data. For the gluinos and first two generations of sfermions, we observe that the gluino mass is in the range [2, 14] TeV, the squarks mass range is [2, 13] TeV and the sleptons mass is in the range [1, 5] TeV. We achieve the cold dark matter relic density consistent with 5σ Planck 2018 bounds via A-funnel and coannihilation channels such as stop–neutralino, stau–neutralino, and chargino–neutralino. Except for a few chargino–neutralino coannihilation solutions, these solutions satisfy current nucleon-neutralino spin-independent and spin-dependent scattering cross-sections and may be probed by future dark matter searches.

Improved particle-flow event reconstruction with scalable neural networks for current and future particle detectors

Efficient and accurate algorithms are necessary to reconstruct particles in the highly granular detectors anticipated at the High-Luminosity Large Hadron Collider and the Future Circular Collider. We study scalable machine learning models for event reconstruction in electron-positron collisions based on a full detector simulation. Particle-flow reconstruction can be formulated as a supervised learning task using tracks and calorimeter clusters. We compare a graph neural network and kernel-based transformer and demonstrate that we can avoid quadratic operations while achieving realistic reconstruction. We show that hyperparameter tuning significantly improves the performance of the models. The best graph neural network model shows improvement in the jet transverse momentum resolution by up to 50% compared to the rule-based algorithm. The resulting model is portable across Nvidia, AMD and Habana hardware. Accurate and fast machine-learning based reconstruction can significantly improve future measurements at colliders.