ワンクリックで
madgraph
MadGraph5_aMC@NLO — process generation, card writing/editing, event generation, LHE parsing, parameter scans
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
MadGraph5_aMC@NLO — process generation, card writing/editing, event generation, LHE parsing, parameter scans
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Numerically evaluate a FormCalc-reduced one-loop amplitude with LoopTools 2.16 (Passarino-Veltman integrals via the Wolfram/LoopTools MathLink) and emit a scattering/v1 JSON (σ_SI, σ_SD) for direct detection. Unblocks the 2HDM+a loop-only DD path. Self-heals via _shared/installs/looptools preflight.
Reduce a FeynArts amplitude list with FormCalc 9.10. Produces amp_reduced.m and amp_reduced.meta.json sidecar conforming to amp_reduced.meta/v1. Self-heals via _shared/installs/formcalc preflight.
Generate Feynman diagrams and amplitudes using FeynArts 3.11. Supports built-in models (SM, SMQCD, THDM, MSSM) and SARAH-generated models. Outputs FeynAmpList.m, diagrams.pdf, topologies.json, and metadata sidecar.
Compute direct-detection likelihoods and 90%-CL exclusion verdicts using DDCalc 2.2.0. Leaf consumer of scattering/v1 JSON from /micromegas or /formcalc.
MadDM — dark matter relic density, direct detection cross-sections, indirect detection rates, parameter scans with experimental limit comparison
Run HiggsBounds-5 + HiggsSignals-2 constraint checks on a model SLHA file. Computes hb_allowed (per-channel AND), hs_consistent (Δχ² < 6.18), p-values, and per-channel CSV. Supports single-point and scan-directory modes.
| name | madgraph |
| description | MadGraph5_aMC@NLO — process generation, card writing/editing, event generation, LHE parsing, parameter scans |
Full interface for MadGraph5_aMC@NLO (MG5): define processes, write cards, generate events, parse LHE output, and run parameter scans. Covers leading-order and next-to-leading-order generation with any UFO model.
MG5 is the starting point of the HEP Monte Carlo chain. It computes hard-process matrix elements and generates parton-level events in LHE format. Parton showering and hadronization of the LHE output is left to an external Pythia8 driver (a dedicated pythia-config skill is on the roadmap). For dark matter relic density and detection rates using a MG5 plugin, use the maddm skill.
Before any other action, run:
bash plugins/hep-ph-toolkit/_shared/installs/madgraph/detect.sh
plugins/hep-ph-toolkit/_shared/installs/madgraph/INSTALL.md into context and follow it (or
invoke /install madgraph). When the install completes, re-run
detect.sh before proceeding. If it still fails, halt with the
blocker code from the install reference.What are you trying to do?
If the user says use <name>, /madgraph use dark_su3, or --model <name>,
resolve registered paths before writing any MG5 script:
UFO = python3 scripts/resolve_named_model.py <name> --key ufo
SLHA = python3 scripts/resolve_named_model.py <name> --key latest_slha
UFO is non-empty, use import model <UFO> in the generated MG5 script.SLHA is non-empty, supply it as param_card.dat during launch.Example generated MG5 script (passed as a script-file, NOT via -c):
# /tmp/dark_su3_run.mg5
import model /home/user/.local/share/hephaestus/models/dark_su3/ufo/DarkSU3
generate p p > psiD psiD~
output /tmp/dark_su3_out
launch /tmp/dark_su3_out
/home/user/.local/share/hephaestus/models/dark_su3/latest_slha/LesHouches.out
done
Invoke with:
mg5_aMC /tmp/dark_su3_run.mg5
See references/generation.md — "Named-model handoff" for the full callout.
Read: references/setup.md
Covers: source install, conda, Docker options, importing UFO models (FeynRules, HEPMDB, GitHub), CUDACPP GPU backend compilation, Python/Fortran environment requirements, linking Pythia8 as a shower plugin.
Read: references/generation.md
Covers: generate / add process syntax, multiparticle labels, model import conventions, excluding particles with / syntax, proc_card.dat structure, param_card.dat SLHA format, run_card.dat field reference, launch workflow, NLO generation with [QCD]/[QED] and MadLoop.
Read: references/analysis.md
Covers: LHE XML structure (<init>, <event> blocks), banner files, cross-section extraction from stdout and banner, parameter scanning (generating param_card grids, batch submission, merging results).
| Process | MG5 syntax |
|---|---|
| Drell-Yan | generate p p > e+ e- |
| tt̄ (LO) | generate p p > t t~ |
| tt̄ (NLO QCD) | generate p p > t t~ [QCD] |
| W+jets (merged) | generate p p > w+ j then add process p p > w+ j j |
| Single top (t-ch) | generate p p > t j $$ w+ w- |
| Higgs via ggF | generate g g > h [QCD] |
| BSM stop pair | import model mssm then generate p p > t1 t1~ |
| Exclude particle | generate p p > t t~ / a (no photon in s-channel) |
| Z+bb | generate p p > z b b~ |
| 4-lepton | generate p p > e+ e- mu+ mu- |
Multiparticle labels (p, j, l+, vl, …) and the full run_card
field reference (beam, cuts, scale, matching) live in references/generation.md
— consult it when writing cards.
| PDF set | lhaid |
|---|---|
| NNPDF4.0 NLO (central) | 331100 |
| NNPDF4.0 NNLO (central) | 331300 |
| NNPDF3.1 NLO | 303200 |
| NNPDF3.1 NNLO | 303400 |
| CT18 NLO | 14000 |
| CT18 NNLO | 14200 |
| MSHT20 NLO | 27400 |
| MSHT20 NNLO | 27000 |
| Particle | MG5 name | PDG ID |
|---|---|---|
| electron | e- / e+ | 11 |
| muon | mu- / mu+ | 13 |
| tau | ta- / ta+ | 15 |
| e-neutrino | ve / ve~ | 12 |
| μ-neutrino | vm / vm~ | 14 |
| τ-neutrino | vt / vt~ | 16 |
| up | u / u~ | 2 |
| down | d / d~ | 1 |
| strange | s / s~ | 3 |
| charm | c / c~ | 4 |
| bottom | b / b~ | 5 |
| top | t / t~ | 6 |
| gluon | g | 21 |
| photon | a | 22 |
| Z boson | z | 23 |
| W boson | w+ / w- | 24 |
| Higgs | h | 25 |
Non-obvious UFO-import failure modes are catalogued in
references/mg5-model-import-gotchas.md:
import model <bare-name> can silently resolve to a $MG5_HOME/models/
stub with a completely different particle set. Always pass the absolute
path to the SARAH-output UFO (or config.json's models.<name>.ufo), never
the bare name.Chi1 → chi1). Use the
lowercased name in all MG5/MadDM commands — this is why the singlet-doublet
workflow passes dm_candidate = "chi1".| Path | Description |
|---|---|
references/setup.md | Installation, environment, GPU backend, Pythia8 linking |
references/generation.md | Process definition → event output, full card reference (multiparticle labels, run_card fields) |
references/analysis.md | LHE parsing, scanning, cross-section extraction |
references/mg5-model-import-gotchas.md | UFO-import failure modes (bare-name stub resolution, particle-name lowercasing) |
scripts/resolve_named_model.py | Named-model resolver: looks up UFO/SLHA paths from config.json |
scripts/card_io.py | SLHA param_card + run_card read/write/update functions |
scripts/lhe_parser.py | LHE XML parser, cross-section extraction, kinematics |
scripts/mg5_batch.py | proc_card content and bash launch script generation |
assets/example_cards/sm_ttbar/ | Known-good SM pp→tt̄ card set at 13.6 TeV NLO |