You are on page 1of 62

MadGraph/MadEvent

– Introduction and mini-tutorial –


Johan Alwall, SLAC
For the MadGraph team

arXiv:0706.2334
arXiv:0809.2410
and other publications
Tsing Hua University, Taiwan, 9 July 2009
Outline
● Simulating physics processes at colliders
● MadGraph/MadEvent – an automatized matrix
element generator and event simulator
● New Physics simulations
– Implementing new models in MadGraph
– Decay chains
– Decay widths
● Standard Model background simulations
– Matching of Matrix elements and Parton showers

Johan Alwall - MadGraph/MadEvent 2


Simulating physics processes
at hadron colliders

Johan Alwall - MadGraph/MadEvent 3


Elements of a simulation
1. Hard interaction

● Process dependent
● Model dependent

● Needs (in general) Tools:


specialized tools, MG/ME, AlpGen,
process by process Sherpa, Whizard, ...
Johan Alwall - MadGraph/MadEvent 4
Elements of a simulation
1. Hard interaction

2. Parton
showers

● Process independent
● Purturbative

QCD/QED in the Tools:


soft/collinear limit Pythia, Herwig,
approximation Sherpa
Johan Alwall - MadGraph/MadEvent 5
Elements of a simulation
1. Hard interaction 3. Hadronization,
hadron decay

2. Parton
showers

● Process independent
● Not calculable from

first principles Tools:


● Different models used Pythia, Herwig,
in different MCs Sherpa
Johan Alwall - MadGraph/MadEvent 6
Elements of a simulation
1. Hard interaction 3. Hadronization,
hadron decay

2. Parton
showers

● Process independent 4. Underlying


● Not calculable from event
first principles Tools:
● Different models used
Pythia, Herwig,
in different MCs Sherpa, Jimmy, ...
Johan Alwall - MadGraph/MadEvent 7
Elements of a simulation
1. Hard interaction 3. Hadronization,
hadron decay

2. Parton
showers

4. Underlying
event

5. Detector simulation
By experiments, or general (PGS, Delphes)
Johan Alwall - MadGraph/MadEvent 8
● Parton showering
● Hadronization
● Underlying event
● Detector simulation
→ Independent of “hard process”
→ Solve once and for all, tune to experimental data
(competition/alternatives good for cross checks)
● “Difficult”, process and model dependent part:
Hard process generation
Johan Alwall - MadGraph/MadEvent 9
Process list in Pythia

Implementation
of nearly 500
processes by hand:
➔ Painful, error prone,
time consuming
➔ Only small subset of
relevant processes
(2 → 2 processes
from SM, SUSY
and Technicolor)

Johan Alwall - MadGraph/MadEvent 10


Imagine a hard process simulation
program which can simulate any
process in any model at any collider
with good efficiency and minimal risk
for errors?

Too good to be true?

Johan Alwall - MadGraph/MadEvent 11


Imagine a hard process simulation
program which can simulate any
process in any model at any collider
with good efficiency and minimal risk
for errors?

It exists! (or something


very close)
(and is easy to use)

Johan Alwall - MadGraph/MadEvent 12


MadGraph/MadEvent
[Long, Stelzer, 1994; Maltoni, Stelzer, 2003]

Matrix element / event generation of any process:


● User requests a process (ex. pp>tt~jjj) and the
corresponding code is generated on the fly
(up to 10 external particles, max. 10000 diagrams)
● User inputs model/collider parameters/cuts, and code
runs locally or on our computer clusters
● Returns Feynman diagrams, cross section,
parton-level events
● Limitations:
– Tree-level amplitudes, based on Feynman diagrams
– Optimization on single procs limited by generality
Johan Alwall - MadGraph/MadEvent 13
MadGraph/MadEvent v. 4
[JA et al., arXiv:0706.2334]

Features of 2007 release, v. 4.0:


● Complete web simulation:
MadGraph/MadEvent → Pythia → PGS/Delphes
● Personal web databases for all users
● Several complete models: SM, HEFT, MSSM, 2HDM
● USRMOD: Easy new model implementation from SM
● Multiple processes in single generation
● “SUSY Les Houches Accord” for all model parameters
● Analysis platforms: ExRootAnalysis and MadAnalysis
Johan Alwall - MadGraph/MadEvent 14
MG/ME workflow
FeynRules

Johan Alwall - MadGraph/MadEvent 15


Input cards


proc_card.dat: Defines the process to generate

param_card.dat: Defines model parameters,
masses, couplings, decay widths

run_card.dat: Defines which collider to simulate
and cuts to use
● plot_card.dat, pythia_card.dat, 
pgs_card.dat: Defines parameters for analysis,
hadronization, and detector simulation

Johan Alwall - MadGraph/MadEvent 16


So, how do I use MG/ME?

1. Open your browser


2. Go to one of our sites
3. Create a process
4. Generate events
Sounds easy? It is!
Let me show you!

Johan Alwall - MadGraph/MadEvent 17


Syntax examples

pp > tt~j : Generate all combinations of processes


for particles defined in multiparticle labels p / j
pp > tt~ > bb~W+W- : Only allow diagrams with
t and t~ in s-channel
pp > W+W- / h : Exclude any diagrams with h
pp > (t > bW+)(t~ > b~W-) : Decay chain,
only diagrams compatible with given decay;
only t / t~ close to mass shell

Johan Alwall - MadGraph/MadEvent 18


Recent developments (after v. 4.0)
[JA et al., arXiv:0809.2410]

● FeynRules – models directly from Lagrangian


● USRMOD 2 – user model starting from any model
● Decay chain specifications
● Decay width calculation A>BC...
● Matching for SM&BSM processes
● Grid Version
● Dipole subtraction for NLO real contributions
● MadWeight
● MadOnia
Johan Alwall - MadGraph/MadEvent 19
Recent developments (after v. 4.0)
[JA et al., arXiv:0809.2410]
In th
is le
FeynRules – models directly from Lagrangian ctur

e
● USRMOD 2 – user model starting from any model
● Decay chain specifications
● Decay width calculation A>BC...
● Matching for SM&BSM processes
● Grid Version
● Dipole subtraction for NLO real contributions
● MadWeight
● MadOnia
Johan Alwall - MadGraph/MadEvent 20
Implementing new models

Ways to implement your own model in MG/ME:


1) Modify existing model (e.g. changing couplings)
2) User model framework (new particles/interactions)
– Add new particles
– Add new interactions
– Enter expressions for the new couplings
– Script generates all files needed by MadEvent!
3) FeynRules
– Directly from Lagrangian to implementation
Johan Alwall - MadGraph/MadEvent 21
User model implementation

● Easy and quick implementation, especially when


complexity of added sector is not too large
● Example:
QCD T' pair production with
T' → Aht in Little Higgs
model with T parity

Hubisz, Meade [hep-ph/0411264]

Johan Alwall - MadGraph/MadEvent 22


User model implementation

Specify new particles and interactions


particles.dat

interactions.dat

Johan Alwall - MadGraph/MadEvent 23


User model implementation

Specify new particles and interactions


particles.dat

interactions.dat

Run script. Modify couplings


couplings.f

Johan Alwall - MadGraph/MadEvent 24


User model implementation

Specify new particles and interactions


particles.dat

interactions.dat

Run script. Modify couplings You are ready to


couplings.f
start generating
processes!

Johan Alwall - MadGraph/MadEvent 25


FeynRules
[Christiansen, Duhr, arXiv:0806.4194]

Model-file
Lagrangian
Particles, parameters, ...

FeynRules

TeX Feynman rules

MC interfaces

FeynArts CalcHep MadGraph Sherpa more...

FormCalc Pythia
Johan Alwall - MadGraph/MadEvent 26
FeynRules
[Christiansen, Duhr, arXiv:0806.4194]

● Mathematica package for derivation of


Feynman rules from any Lagrangian
● Specify model content (particles, coupling
relations) and Lagrangian
● Interfaces to MC tools
– Theory tools: FeynArts/CalcHep/MadGraph (loops,
DM constraints, parton level MC/plots)
– Experimental tools: MG/ME, Sherpa (already in exp.
frameworks)
● Implementation example in backup slides
Johan Alwall - MadGraph/MadEvent 27
USRMOD II

● From v. 4.4.26, FeynRules-generated


implementations for all models included in
MadGraph
● New USRMOD can start from any FeynRules-
generated model, made by any author
● Useful e.g. to add content to the MSSM

Johan Alwall - MadGraph/MadEvent 28


Recent developments (after v. 4.0)

● FeynRules – models directly from Lagrangian


● USRMOD 2 – user model starting from any model
● Decay width calculation A>BC...
● Decay chain specifications
● Matching for SM&BSM processes
● Grid Version
● Automatic dipole subtraction
● MadWeight
● MadOnia
Johan Alwall - MadGraph/MadEvent 29
Decay width calculation

● MadGraph/MadEvent need info on total decay


width of unstable particles
● For new models: two ways of automatically
calculate decay widths:
– BRIDGE: Dedicated tool for calculation of 2-
and 3-body decays [Meade, Reece, 2007]
– MG/ME: Specify all relevant decays in
proc_card, run as usual (result in GeV)
Example: tp > zp t
Also multiparticle decays work: h > l+l-l+l-
Johan Alwall - MadGraph/MadEvent 30
SUSY vs. UED – spin effects
Long decay chains give information on intermediate
particle masses and spins through edge and endpoint
positions and shapes
Beware of common
simplifications:
0. Production and decay
factorized
1. Spin ignored
2. Chains only through
1 → 2 decays.
3. Narrow width
approximation
employed.
4. Non-resonant
diagrams ignored.
Johan Alwall - MadGraph/MadEvent 31
Decay chains in MG/ME
gg>(go>u~(ul>u n1))(go>b~(b1>(b(n2>mu+(mul->mu- n1)))))
1. Full matrix element with all
correlations between
production and decay
2. 1 → N decays possible
3. BW for all resonances
4. Non-resonant contributions
can be included only where
relevant

Johan Alwall - MadGraph/MadEvent 32


Decay chains in MG/ME
gg>(go>u~(ul>u n1))(go>b~(b1>(b(n2>mu+(mul->mu- n1)))))
1. Full matrix element with all
correlations between
production and decay
2. 1 → N decays possible
3. BW for all resonances
4. Non-resonant contributions
can be included only where
relevant
Example safe simplification: factorize process at scalar decay
gg>(go>u~ ul)(go>b~ b1)
Decay scalars at event level with MG or BRIDGE
ul > u n1
b1 > b(n2>mu+(mul->mu- n1))
Johan Alwall - MadGraph/MadEvent 33
Recent developments (after v. 4.0)

● FeynRules – models directly from Lagrangian


● USRMOD 2 – user model starting from any model
● Decay width calculation A>BC...
● Decay chain specifications
● Matching for SM&BSM processes
● Grid Version
● Automatic dipole subtraction
● MadWeight
● MadOnia
Johan Alwall - MadGraph/MadEvent 34
Simulating QCD radiation

● For any New Physics signature involving hard


jets, it is crucial to correctly simulate hard jet
production in Standard Model backgrounds
● Besides top quark decay, the only source of
hard jets in the SM is QCD radiation. Examples:
– W/Z production plus jets
– Hard photon plus jets
– Top quark pairs plus extra jets
(esp. > 4 jet or dilepton + > 2 jet signals)
– QCD multijet production
Johan Alwall - MadGraph/MadEvent 35
Simulating QCD radiation

● For any New Physics signature involving hard


jets, it is crucial to correctly simulate hard jet
production in Standard Model backgrounds
● Besides top quark decay, the only source of
hard jets in the SM is QCD radiation. Examples:
g Z
g g

_
(beam) q q (beam)
Johan Alwall - MadGraph/MadEvent 36
The Parton shower approach

● For soft jets, and jets at large rapidity (small angle


with beam), use Parton Shower approach:
– Based on 1→2 splittings in collinear approximation
of QCD (cf. photon emissions in QED)
– Step-by-step subsequent QCD emissions
– Fast, computationally cheap
– No limit on particle multiplicity
● However, only formally valid in the soft and
collinear regions of phase space

Johan Alwall - MadGraph/MadEvent 37


The Matrix Element approach

● For high-pT, central and widely separated jets, full


matrix element calculations necessary
– Includes subleading, non-logarithmic terms
– Includes interference between diagrams
– Describes jet production away from the soft and
colllinear region
– Fixed parton multiplicity
– Slow, large computer resources needed
● Diverges in the soft and collinear region (due to
non-resummation of logarithms)
Johan Alwall - MadGraph/MadEvent 38
Simulating QCD radiation

Cross section (pb)

Jet def. cutoff ʃ pT(N-th jet) > x (GeV)


Parton showers can get multiple hard jet production
from QCD radiation wrong by orders of magnitude
Johan Alwall - MadGraph/MadEvent 39
Simulating QCD radiation
We need to simultaneously simulate jets
throughout both the hard and soft/collinear regions,
for several jet multiplicities (e.g. Z+0,1,2,3,4 jets).
Problems:
– Double counting between samples

0-jet ME event

+
1-jet ME event

Johan Alwall - MadGraph/MadEvent 40


Simulating QCD radiation
We need to simultaneously simulate jets
throughout both the hard and soft/collinear regions,
for several jet multiplicities (e.g. Z+0,1,2,3,4 jets).
Problems:
– Double counting between samples

0-jet ME event
+ PS

+ Double counting!
1-jet ME event

Johan Alwall - MadGraph/MadEvent 41


Simulating QCD radiation
We need to simultaneously simulate jets
throughout both the hard and soft/collinear regions,
for several jet multiplicities (e.g. Z+0,1,2,3,4 jets).
Problems:
– Double counting between samples
– Discontinuities in distributions

PS ME Matched
ME+
PS

Johan Alwall - MadGraph/MadEvent 42


Simulating QCD radiation

Solution: “Jet matching” between ME and PS


– Separate “hard jet” and “soft/collinear jet” regions
using phase-space cutoff
– Allow ME jets to populate only “hard” region and PS
emissions only “soft” region
– Modify ME description to mimick the parton shower
near the cutoff
– Schemes: Catani, Krauss, Kuhn, Webber [2001],
M.L. Mangano [2002, 2006]
Multiple variants (J.A. et al, 2007)

Johan Alwall - MadGraph/MadEvent 43


Matching in MadGraph/MadEvent
Example: Simulation of e+e- with 0, 1, 2, 3 jets
proc_card.dat: run_card.dat:
pp > e+e­     @0 …
…   1 = ickkw Matching on
pp > e+e­j    @1 …
…  15 = xqcut
pp > e+e­jj   @2
… Matching scale
pp > e+e­jjj  @3 (cut between ME and PS)

Matching automatically done when run


through MadEvent and Pythia interface
Johan Alwall - MadGraph/MadEvent 44
Sanity checks for matching

● pT(W+/-) at the Tevatron (run 1)

● Comparison of W+jets production between


AlpGen, MadGraph, Sherpa et al. [arXiv:0706.2569]
Johan Alwall - MadGraph/MadEvent 45
Conclusions

The MadGraph philosophy:


– Provide you with a complete set of tools for
simulation and analysis of Standard Model and
New Physics processes
– Minimize time you spend on simulation difficulties
– Allow you to spend your time on physics problems
– Our approach:
● Complete web interface (process and event generation)
● Completely open and easily customizable source code
● Support to spin-offs, periferal tools, custom needs
● Fast response to user requests
Johan Alwall - MadGraph/MadEvent 46
Visit us,
generate processes,
generate events!
http://madgraph.hep.uiuc.edu/
http://madgraph.physics.ucl.ac.be/
http://madgraph.roma2.infn.it/

謝謝您
Johan Alwall - MadGraph/MadEvent 47
Backup slides

Johan Alwall - MadGraph/MadEvent 48


Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 49
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 50
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 51
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 52
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 53
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 54
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 55
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 56
Slides contributed by C. Duhr
Johan Alwall - MadGraph/MadEvent 57
MLM matching

Keep Discard unless highest multiplicity


Johan Alwall - MadGraph/MadEvent 58
Shower kT scheme

● Keep/reject event based on kT of hardest shower emission


(as reported by Pythia)
● Highest multiplicity treatment as in CKKW, use min dparton
as cutoff
● No jet clustering
● No need of “fiducial region”, can use kTmatch = dcutME

● Need similar kT definitions in ME and PS (only “new”, pT-


ordered showers at present)

Johan Alwall - MadGraph/MadEvent 59


MadWeight
[P. Artoisenet, V. Lemaitre, F. Maltoni, O. Mattelaer]

Tool to find matrix element weight of exp. events


for (almost) any process in any model

Phase space integration Find likelihood for model


using automatic change of parameters (here top mass)
variables aligned with peaks
Johan Alwall - MadGraph/MadEvent 60
MG/ME on the Grid
[Mad team]
● Optimized/specialized code for given process
● MG code creation as usual
● Selection of parameters (cards) as usual
● Train grids + get relative subprocess cross
sections once and for all in a “gridpack”
● Quick and efficient generation of few events on
single machine – only run relevant channels
● Only input: random seed, number of events

Johan Alwall - MadGraph/MadEvent 61


Automatic dipole subtraction
[R. Frederix et al, arXiv:0808.2128]

● Automatic divergence subtraction for the real


corrections in any QCD NLO calculation
– Catani-Seymour subraction scheme
– Standalone implementation
– Both for SM and BSM
– Massless and massive external particles

Johan Alwall - MadGraph/MadEvent 62

You might also like