Search for $t\bar tt\bar tW$ Production at $\sqrt{s} = 13$ TeV Using a Modified Graph Neural Network at the LHC

TL;DR

This paper develops a modified graph neural network with physics-informed features to improve detection of rare four-top-quark production events at the LHC, outperforming traditional classifiers.

Contribution

Introduces a hybrid, physics-informed GNN with quantum and cross-attention components for better event classification in high-energy physics.

Findings

GNN achieves higher significance and ROC-AUC than BDT and Xgboost.

Model trained on Monte Carlo simulations with realistic event normalization.

Framework enhances precision in event selection at the LHC.

Abstract

The simultaneous production of four top quarks in association with a ($W$) boson at $(\sqrt{s} = 13)$ TeV is an rare SM process with a next-to-leading-order (NLO) cross-section of $(6.6^{+2.4}_{-2.6} {ab})$\cite{saiel}. Identifying this process in the fully hadronic decay channel is particularly challenging due to overwhelming backgrounds from $t\bar{t}, t\bar{t}W, t\bar{t}Z$, and triple-top production processes. This study introduces a modified physics informed Neural Network, a hybrid graph neural network (GNN) enhancing event classification. The proposed model integrates Graph layers for particle-level features, a custom Multi Layer Perceptron(MLP) based global stream with a quantum circuit and cross-attention fusion to combine local and global representations. Physics-informed Loss function enforce jet multiplicity constraints, derived from event decay dynamics. Benchmarked against conventional methods, the GNN achieves a signal significance $(S/\sqrt{S+B})$ of $0.174$ and ROC-AUC of 0.974, surpassing BDT's significance of $0.148$ and ROC of $0.913$, while Xgboost achieves a significance of $0.149$ and ROC of $0.920$. The classification models are trained on Monte Carlo (MC) simulations, with events normalized using cross-section-based reweighting to reflect their expected contributions in a dataset corresponding to $350\;$fb$^{-1}$ of integrated luminosity. This enhanced approach offers a framework for precision event selection at the LHC, leveraging high dimensional statistical learning and physics informed inference to tackle fundamental HEP challenges, aligning with ML developments.