PaperlayerView exact revision on arXiv

Source-linked AI summary

MadGraph/MadEvent v4: The New Web Generation

Johan Alwall, Pavel Demin, Simon de Visscher, Rikkert Frederix, Michel Herquet, Fabio Maltoni, Tilman Plehn, David L. Rainwater, Tim Stelzer

arXiv:0706.2334v1hep-ph

TL;DR

MadGraph/MadEvent addresses the need for integrated, flexible simulation of collider signals and backgrounds across physics models and event-generation stages. It provides user-driven code creation, web-based generation, model extensions, and matched multi-jet production. The release demonstrates these capabilities across Higgs, resonance, supersymmetry, single-top, and W+jets applications.

  • Problem

    Collider studies require accurate, integrated simulation of signals and backgrounds across hard interactions, parton showers, hadronization, and detector reconstruction.

  • Method

    The package generates process-specific code from user requests, supports multiple physics models and web-controlled event generation, and matches matrix elements with parton showers for multi-jet samples.

  • Results

    The release provides web-accessible event generation and detector simulation within one framework, with applications spanning Higgs, resonance, supersymmetry, single-top, and W+jets studies.

  • Takeaways & Limitations

    MadGraph/MadEvent enables users to simulate diverse signal and background processes through a common, user-driven workflow from code creation to downloadable events and plots.

  • Takeaways & Limitations

    The new-physics framework requires available Feynman rules and remains limited to interaction structures compatible with HELAS.

Abstract

from arXiv · show

We present the latest developments of the MadGraph/MadEvent Monte Carlo event generator and several applications to hadron collider physics. In the current version events at the parton, hadron and detector level can be generated directly from a web interface, for arbitrary processes in the Standard Model and in several physics scenarios beyond it (HEFT, MSSM, 2HDM). The most important additions are: a new framework for implementing user-defined new physics models; a standalone running mode for creating and testing matrix elements; generation of events corresponding to different processes, such as signal(s) and backgrounds, in the same run; two platforms for data analysis, where events are accessible at the parton, hadron and detector level; and the generation of inclusive multi-jet samples by combining parton-level events with parton showers. To illustrate the new capabilities of the package some applications to hadron collider physics are presented: 1) Higgs search in pp \to H \to W^+W^-: signal and backgrounds. 2) Higgs CP properties: pp \to H jj$in the HEFT. 3) Spin of a new resonance from lepton angular distributions. 4) Single-top and Higgs associated production in a generic 2HDM. 5) Comparison of strong SUSY pair production at the SPS points. 6) Inclusive W+jets matched samples: comparison with the Tevatron data.

1. Introduction

MadGraph/MadEvent is designed to reduce barriers between physics ideas and experimentally testable event simulations. The new release extends this automation to broader models, web-based detector simulation, and unified signal-background studies.

  • Accurate simulation of signals and Standard Model backgrounds is important for interpreting distinctive collider signatures such as mass distributions, kinematic edges, and angular correlations.
  • Recent progress includes general-purpose matrix-element generators, multiparton tools, and Monte Carlo programs incorporating NLO corrections.
  • Hadronic-collision simulation requires integrating hard matrix elements with parton showers and hadronization, using matching algorithms to avoid inconsistent descriptions.
  • MadGraph/MadEvent automates the path from matrix-element calculation to unweighted parton-level events, allowing physicists to focus on physics analysis.
  • The current release adds MSSM, 2HDM, and HEFT models, user-defined models, full hadronic and detector simulation, and web-based generation from concept to reconstructed events.
  • The paper illustrates these capabilities through package overviews, model implementations, matched multi-jet generation, and hadron-collider applications.

2. The path to event generation

MadGraph/MadEvent organizes event generation as a card-controlled pipeline from process-specific code creation through detector-level reconstruction. Users can operate it through the web, locally, or as developers with the full package.

  • The generation pipeline comprises code creation, parton-level events, hadron-level events, and reconstructed detector objects, with each stage controlled by input cards.
  • A process card specifies processes, physics models, coupling orders, and multiparticle labels, after which MadGraph creates downloadable or cluster-runnable process-specific code.
  • A parameter card supplies model-dependent numerical parameters, while a run card specifies event counts, collision settings, PDFs, scales, and detector acceptance.
  • The run produces unweighted Les Houches events, kinematic plots, and a ROOT file, while Pythia and PGS can extend the simulation to hadron and detector levels.
  • Web users need no installation and can download cards and results, whereas local users gain flexibility and larger samples; developers access all functionality locally.
  • The package contains dedicated components for matrix elements, standalone matrix-element generation, process templates, user models, HELAS, and direct particle decays.

3. The Standard Model implementation

The implementation separates model parameters from event-generation inputs and supports Standard Model variants alongside the HEFT extension. HEFT replaces heavy-quark loops with effective Higgs-gluon couplings under a stated kinematic approximation.

  • Standard Model: The Standard Model implementation uses model calculators to derive secondary parameters and widths consistently at tree level, outputting SUSY Les Houches-compliant parameter cards.
  • Standard Model: The SM Calculator can take GF, mZ, and mW as inputs and derive αem and sin θW, while allowing alternative choices of independent electroweak parameters.
  • Standard Model: Users can distinguish quark pole masses from MS masses and evolve the latter to the Higgs-mass scale to improve the perturbative expansion.
  • Standard Model: Available SM variants include a diagonal-CKM minimal model, an smckm model with first- and second-generation mixing, and a no-Higgs nonlinear sigma-model.
  • Higgs Effective Theory: HEFT directly couples Higgs bosons to gluons and photons, approximating heavy-quark loops by infinite heavy-quark mass when relevant kinematic scales remain below 2mt.
  • Higgs Effective Theory: Because MadGraph supports three- and four-point vertices, HEFT represents four-gluon interactions using three-point vertices with an extra non-propagating tensor particle.
  • Higgs Effective Theory: The implementation also includes pseudo-scalar Higgs-gluon couplings and permits mixed CP states by changing the even and odd coupling components.

4. Going beyond the Standard Model

The release extends MadGraph/MadEvent beyond the Standard Model with tested MSSM and 2HDM implementations, a calculator framework, and user-defined model support.

  • The release adds fully tested MSSM and general 2HDM implementations, alongside a framework for setting up new models.
  • 2HDM implementation: The generic 2HDM permits tree-level flavor-changing neutral currents and CP-violating interactions, while a simplified version generates fewer diagrams.
  • 2HDM implementation: TwoHiggsCalc converts Lagrangian inputs into leading-order scalar masses, mixing matrices, decay widths, and branching ratios for MadEvent processes.
  • 2HDM implementation: The 2HDM uses a Higgs basis with one nonzero-vev doublet, while Gen2HB converts parameters from a generic basis into that basis.
  • 2HDM implementation: Yukawa matrices for the second Higgs doublet are free inputs in the physical fermion basis, whereas mass matrices are fixed by fermion masses and CKM mixing.
  • User Model: The user-model framework enables flexible implementation of new interactions but remains limited by MadGraph and HELAS Lorentz and color structures.

5. Matching of jet production by parton showers and matrix elements

MadGraph/MadEvent combines matrix elements with parton showers to model both hard, separated emissions and softer radiation, supporting CKKW and MLM matching.

  • Multijet backgrounds and signal jet activity motivate combining matrix-element generation with parton showering and hadronization.
  • CKKW matching: CKKW separates emissions with a k⊥ measure, assigning low-k⊥ radiation to showers and high-k⊥ radiation to matrix elements.
  • MLM matching: The alternative MLM scheme reweights multiparton events with αs factors and vetoes events whose showers generate emissions above the phase-space cutoff.
  • The release implements both MLM and CKKW matching, with MLM showering and vetoing in the Pythia interface and CKKW αs and Sudakov reweighting at matrix-element creation.
  • MLM matching: MadGraph/MadEvent lets users define MLM phase-space separation with either cone jets or kT-clustered jets.

6. Tools

The package provides interoperable tools for storing, decaying, simulating, and analyzing events from parton generation through hadronization and detector simulation.

  • Les Houches event files defer plotting, PDF-error estimation, and scale variations, avoiding costly event-generation reruns and enabling generator-independent analysis.
  • ExRootAnalysis: ExRootAnalysis stores events in ROOT trees and provides event loops, selections, object classification, and analysis modules.
  • MadAnalysis: MadAnalysis reads parton- and detector-level event formats, applies selection cuts, and produces plots in ASCII, GnuPlot, or TopDrawer formats.
  • Decay: Decay handles 68 Standard Model decay modes for τ, W, Z, t, and h, efficiently producing multiparticle final states when detailed spin correlations are unnecessary.
  • Hadronization and detector simulation: The Pythia-PGS package links MadEvent files to Pythia hadronization and PGS fast detector simulation, with LHAPDF, STDHEP, and executable interfaces included.
  • Hadronization and detector simulation: Outputs across the simulation chain remain consistently numbered and can be read by ExRootAnalysis or MadAnalysis for stage-specific plots and analysis.

7. Applications to hadron collider physics

The paper demonstrates MadGraph/MadEvent through concise signal, background, model, spin, supersymmetry, and matched-jet studies aimed at both theoretical and experimental analyses.

  • The applications illustrate how the package's new features support signal and background analyses with both theoretical and experimental aims.

7.1 Higgs search in pp →h →W +W −: signal and backgrounds

The example generates Higgs-to-WW signal and relevant backgrounds together, preserving resonant and non-resonant contributions through parton-, hadron-, and detector-level analysis. Angular and jet-veto observables illustrate how simulation choices affect signal-background characterization.

  • Signal and backgrounds: The study targets pp →h →W +W −→e−µ+¯νeνµ for a moderate-mass Higgs, using the lepton azimuthal separation ∆φ as the discriminating variable.The signal’s lepton proximity reflects angular-momentum constraints and left-handed W couplings.
  • Signal and backgrounds: MadGraph can generate signal and multiple backgrounds in one run, mixing unweighted events in proportion to their cross sections.This produces the correct event mixture automatically.
  • Simulation and analysis: Pythia, PGS, and MadAnalysis provide interfaces for decay, hadronization, detector simulation, plotting, and comparisons across simulation stages.The analysis supports parton-level and detector-level event formats.
  • Signal and backgrounds: The background sample includes irreducible WW production and reducible non-resonant, single-top resonant, and tt̄ double-resonant contributions.The processes are included consistently by selecting couplings and vetoing the Higgs as an intermediate particle.
  • Simulation and analysis: Initial-state radiation changes the average lepton pT considerably, while leaving the ∆φ distributions broadly similar between parton-level and fully simulated events.The comparison is made between parton-level and hadron-level distributions.
  • Background composition: Jet vetoes make non-resonant and single-resonant contributions exceed 50% of the final WbWb background, whereas generic phase space is dominated by tt̄.Single-resonant contributions are otherwise completely negligible.

7.2 Higgs CP properties: pp →hjj in the HEFT

The HEFT implementation is used to study Higgs CP properties through jet-angle distributions in pp →hjj. With forward-backward jet cuts, the normalized ∆φjj distributions are compared for scalar and pseudo-scalar Higgs bosons.

  • Method: The study probes Higgs CP properties through the angle ∆φjj between the transverse momenta of the two jets.The process is pp →hjj and is evaluated at parton level.
  • Event selection: The jet selection requires pT (j) > 20 GeV, ∆Rjj > 0.4, |ηj1 −ηj2| > 4, and ηj1 ·ηj2 < 0.These conditions select one very forward and one very backward jet.
  • Results: The selected cuts produce a signal comparable with Higgs production through W-boson fusion, and ∆φjj is plotted for pure scalar and pure pseudo-scalar cases at Mh = 120 GeV.The plotted distributions are normalized.

7.3 Spin of a new resonance from lepton angular distributions

The study uses lepton angular distributions in X →µ+µ− to distinguish spin-0, spin-1, and spin-2 resonances. It examines production-mode effects and tests the stability of the distributions after radiation, hadronization, and detector simulation.

  • Related implementation: The HEFT diagram selection uses an auxiliary tensor particle T to represent four-gluon interactions, while the right-hand diagram is absent for a pseudo-scalar Higgs.The auxiliary-particle construction is part of the effective-vertex implementation.
  • Setup: The process p¯p →X →µ+µ− is evaluated for spin-0, spin-1, and spin-2 s-channel resonances.The angular distributions are presented as a method for studying the intermediate particle’s spin.
  • Angular observable: The Collins–Soper angle θ minimizes effects from non-zero transverse momentum of the intermediate resonance.At leading-order parton level, the muon pair has zero transverse momentum and θ coincides with θ⋆.
  • Spin discrimination: The normalized cos θ distribution is compared across the three resonance spins in q¯q →X →µ+µ−.No cuts are applied in the corresponding figure.
  • Production dependence: For spin-2 production, changing the initial state from gluon fusion to the combined gluon-and-quark sample dramatically changes the muon angular distribution.The comparison uses a 1 TeV spin-2 resonance.
  • Simulation effects: Initial-state radiation, showering, and hadronization do not significantly modify the lepton distributions, while PGS provides relatively reliable reconstructed results for CMS studies.The reconstructed and parton-level distributions are compared with a muon rapidity cut |ηµ+,µ−| < 2.4.

7.4 Single-top associated Higgs production in a generic 2HDM

Single-top associated Higgs production is smaller than naive rate estimates because dominant amplitudes undergo strong destructive interference. The generic 2HDM study shows that model-dependent enhancements remain constrained by cancellations and suppressed emission configurations.

  • SM and MSSM: The SM single-top associated Higgs cross section is about 100 fb for mh ≃100 GeV, rather than the naive estimate of about 300 fb.The reduction arises from strong destructive interference between the two dominant amplitudes.
  • SM and MSSM: Each dominant diagram contains a term proportional to mt that violates unitarity at high energies and cancels in their sum.This cancellation explains the suppressed physical cross section.
  • MSSM: In the MSSM, an additional charged-Higgs diagram contributes for heavy scalar H0 because of the large W ±H∓H0 coupling.Its amplitude is roughly the same order and sign as the W-mediated contribution.
  • 2HDM: For a light pseudoscalar A0 in a type I 2HDM, increasing mH± enhances the cross section, but it remains much smaller than the SM value because of the reduced top Yukawa coupling.In type II models, varying mH± has only a small effect.
  • 2HDM: For initial-state b-quark emission, the overall pseudoscalar cross section is slightly suppressed, by roughly a factor of 2, relative to the SM across mH± values.The conclusion attributes the small rates to unitarity cancellations and suppressions from different emission configurations.

7.5 Comparison of strong SUSY pair production at the SPS points

The SPS comparison demonstrates MadGraph/MadEvent’s ability to generate mixed strong-SUSY samples, process them through detector simulation, and compare subprocess contributions and event-scale distributions across benchmark points.

  • Sample construction: 497 strong-SUSY subprocesses are generated automatically in proportions set by their relative cross sections.The sample distinguishes gluino, squark, and associated-production groups, including different squark flavors.
  • Sample construction: Events for all ten SPS points are passed through Pythia for decays, showering, and hadronization, then through PGS for detector simulation.The workflow uses parameter cards generated with SoftSusy and Sdecay.
  • Generated results: The generated outputs include total and subprocess cross sections, unweighted events at parton, hadronized, and detector levels, and ROOT analysis files.These outputs support analysis at all three event levels.
  • Cross-section comparison: Valence-partner, gluino, and sometimes top-partner production contributes 68% to 92% of the total cross section while comprising about 25 of nearly 500 processes.The dominance is attributed to valence u and d parton distributions, especially for heavier squarks.
  • Kinematic distributions: HT peak positions are about 65–85% of the sum of produced-particle masses, with an increasing ratio at higher masses.SPS5 is an exception: its low-mass stop-pair peak lies below 40% of the summed stop masses.
  • Kinematic distributions: Adding twice the LSP mass to the peak positions yields values between 85% and 100% of the produced-particle mass.The ratio shows a slightly rising trend toward higher masses.
  • Scope and limitation: The study is preparatory because gluino decays were handled with Pythia rather than MadEvent, limiting treatment of spin correlations in angular-distribution studies.The authors note that efficiency becomes important when studying such refinements.

7.6 Inclusive W+jets matched samples: comparison with the Tevatron data

The W+jets study combines matrix elements and parton showers to describe multiple jet multiplicities, and its default simulation reproduces the shape of Tevatron data after a 20% normalization fit.

  • Motivation: W+jets is an important background because hadron-collider signals often contain isolated leptons, missing energy, and hard jets.Parton showers describe soft or nearby jets, whereas hard well-separated jets require matrix elements and matching.
  • Matching method: The matching method uses a variable matching scale to separate matrix-element and parton-shower descriptions, with scale variations used to assess robustness.The study follows a modified MLM approach using kT-clustered jets.
  • W-boson spectrum: At low pT(W), the pure parton shower dominates; above the matching scale, higher jet multiplicities become increasingly important, and near 150 GeV all four included multiplicities are comparable.The matched sum combines the different multiplicity contributions.
  • Comparison with data: The default simulation reproduces the shape of Tevatron data, with the normalization fitted by 20%.The displayed uncertainty band varies the matrix-element αs arguments by factors of 2 upward and downward.
  • Jet-rate diagnostic: The 1→2 differential jet rate illustrates the transition between parton-shower and matrix-element samples across the matching scale.The jet rate is the variable used for matching, with jets restricted to |η| < 2.5 in the stated definition.

8. Conclusions and Outook

The release presents MadGraph/MadEvent as one user-driven framework for simulating signals and backgrounds, from model implementation and event generation through detector simulation. It also outlines ongoing developments in theorist–experimentalist interaction, model implementation, decay-chain generation, and matrix-element-based analysis.

  • MG/ME creates process-specific code from user requests rather than relying on a fixed process library, while supporting extensible Standard Model, HEFT, MSSM, and general 2HDM models.This user-driven design is intended to keep model and process selection flexible.
  • All package functionalities can be accessed through a web interface, including code creation, event generation, detector simulation, and downloading event files and plots.The interface uses input cards and does not require local installation or compilation.
  • The framework is designed to connect theorists and experimentalists by allowing models and collider phenomenology to be tested up to simulated detector level, with both signal and background generation.Experimental groups can subsequently apply detector-specific tools for an exp-grade analysis.
  • A single framework automates matrix-element creation, cross-section integration, and event generation while supporting flexible use without specific programming skills.The stated benefit is to reduce tedious and error-prone tasks for theorists and experimentalists.
  • Planned or tested extensions include theorist–experimentalist interaction tools, decay-chain generation for rich multiparticle final states, and matrix-element-based use of spin correlations in analysis.The decay approach retains correlations from intermediate resonances for later analysis.
  • The current semi-automatic new-physics-model framework requires available Feynman rules and remains limited by the interaction forms supported by HELAS.Planned work targets generic-Lagrangian diagram generation and a more flexible HELAS implementation.
Loading 0706.2334v1…