LHeC Fast Simulation

This repository provides simulation tools and analysis scripts for studying anomalous quartic gauge couplings (AQGCs) in electron-proton collisions at the Large Hadron-Electron Collider (LHeC). The project focuses on simulating and analyzing key processes that probe the dynamics of $\gamma\gamma \to W^+ W^-$ interactions.

---

Key Features

Anomalous Quartic Gauge Couplings

The study of anomalous quartic gauge couplings (AQGCs) is essential for exploring deviations from Standard Model predictions. AQGCs arise from higher-dimensional operators and provide insights into new physics at high-energy scales.

Simulated Processes

The following processes are central to this project:

1. $\gamma\gamma \to W^+ W^-$
   • Simulated to analyze photon-induced interactions in the LHeC environment.
2. $e^- p \to e^- W^+ W^- p$
   • A key electroweak process in electron-proton collisions, where the scattered electron and proton remain in the final state.

---

Objectives

• Simulate high-energy $\gamma\gamma$ and $e^- p$ collisions at the LHeC.
• Analyze AQGC sensitivity through the processes $\gamma\gamma \to W^+ W^-$ and $e^- p \to e^- W^+ W^- p$.
• Provide tools for differential cross-section analysis, rapidity distributions, and invariant mass studies of the final-state particles.

---

Repository Contents

• Scripts and Tools: Python scripts and analysis tools for parsing LHE files, generating histograms, and visualizing distributions.
• Simulations: Includes configurations for MadGraph5_aMC@NLO to simulate relevant processes.
• Plots and Results: Sample plots for $P_T$, $\eta$, rapidity, and invariant mass distributions.

---

📊 Anomalous Quartic Gauge Couplings (aQGC) Sensitivity at the LHeC

This repository contains a full pipeline for estimating the sensitivity of the LHeC to the anomalous quartic gauge coupling parameter $f_{M2}/\Lambda^4$ via shape-based profile likelihood scans. The analysis is based on semi-leptonic $W^+W^-$ production from $\gamma\gamma$ fusion, including realistic event simulation, machine learning-based signal-background discrimination, and statistical limit setting.

---

📁 Project Structure

├── ExtractData_Train_PlotROC.py           # Extracts and saves histogram-based ML features
├── train_xgboost_and_plot_roc.py          # Trains a BDT and computes ROC + saves scores
├── limit_scan_fm2_profile_likelihood_v5.py  # Full likelihood-based limit calculation
├── output_histograms.root                 # ROOT file with preselection histograms
├── ml_input_from_histograms.csv           # Raw ML inputs from histograms
├── ml_with_bdt_scores.csv                 # ML inputs with predicted BDT scores

---

🧠 ML Pipeline

1. Data Preparation

Run:

python3 ExtractData_Train_PlotROC.py

• Extracts samples from histogram templates in output_histograms.root
• Samples kinematic variables from:
  • Leptons
  • Jets
  • MET
• Saves as ml_input_from_histograms.csv with label, weight, and process.

2. Training the BDT Classifier

Run:

python3 train_xgboost_and_plot_roc.py

• Trains XGBoost BDT
• Computes AUC & ROC curve
• Appends bdt_score and process columns
• Saves as ml_with_bdt_scores.csv

---

📐 Profile Likelihood Limit Setting

Run:

python3 limit_scan_fm2_profile_likelihood_v5.py

Features:

• Calculates optimal BDT cut based on AMS.

• Computes:

  • ML efficiency
  • Preselection efficiency
  • Total efficiency = ML × Preselection

• Constructs Asimov dataset (background only).

• Defines likelihood:

  $$ \mathcal{L}(f_{M2}) = \prod_{i=1}^{N_{\text{bins}}} \text{Poisson}(n_i^{\text{obs}} \mid s_i(f_{M2}) + b_i) $$

• Scans $f_{M2}/\Lambda^4$ and finds 95% CL exclusion:

  $$ q(f_{M2}) = -2 \log \frac{\mathcal{L}(f_{M2})}{\mathcal{L}(0)} > 3.84 $$

---

📊 Output

• profile_likelihood_fm2_plot.pdf showing $q(f_{M2})$
• 95% CL bounds printed in console:

📉 95% CL Excluded FM2 Range: [-0.3733, +0.3901] TeV⁻⁴

---

🧾 Citation

If you use this repository, please cite:

Hamzeh Khanpour, LHeC Semi-leptonic WW Study — Sensitivity to aQGC via BDT + Likelihood Analysis, 2025

---

🧪 Requirements

• Python 3.10+
• ROOT with Python bindings
• xgboost, matplotlib, scipy, pandas, numpy

---

Made with ❤️ by Hamzeh Khanpour

Usage

1. Clone the repository:

   git clone https://github.com/your-username/LHeC_Fast_Simulation.git
   cd LHeC_Fast_Simulation
   

---

Contacts

For additional information or questions, contact us using the email addresses below:

• Hamzeh Khanpour (Hamzeh.Khanpour@cern.ch)
• Laurent Forthomme (Laurent.Forthomme@cern.ch)
• Krzysztof Piotrzkowski (Krzysztof.Piotrzkowski@cern.ch)