on recent developements in madgraph - 大阪大学 · 2011-06-08 · on recent developements in...

On Recent Developementsin MadGraph

2011.06.07

at Osaka University

Kaoru Hagiwara

KEK Theory Center and Sokendai

1. Pre-history (1981-1987)KH, on 2011.06.06 at PPP9 in NCU

2. The vision of MadGraph and FeynrulesJohan Alwall, on 2011.05.03 at MG 2011 Spring in Fermilab

3. FeynRules: status and plansClaude Duhr, on 2011.05.05 at MG 2011 Spring in Fermilab

4. MadLoopValentin Hirschi, on 2011.05.04 at MG 2011 Spring in Fermilab

5. Bridging Theory with ExperimentFabio Maltoni, on 2011.05.09 at PHENO 2011 in Madison

6. How shall we enjoy LHC physics ?KH, on 2011.06.06 at PPP9 in NCU

Pre-history (1982-1987)

1981 Goebel, Halzen, Leveille showed that the cross section zero in the process du → γW−dσ

d cos θ(du → γW−) = 1

2sΣλγ,λW

|M(cos θ)λγ,λW|2 1

16π= (cos θ + 1

3)2 × factor

implies that all the 6 helicity amplitudes have a common factor:

M(cos θ)λW,λγ= (cos θ + 1

3) × fλW ,λγ

1983 I calculate the photon ‘polairzation amplitude’ for the processdu → γW− with Jose Cortes and Franz Herzog.

1985 I calculate the helicity amplitudes for a 2 to 6 processe+e− → L+L− → (l+νN)(l−νN)by using a numerical code with Dieter Zeppenfeld. Phase convention for fermionwave functions is fixed.

1986 I applied the numerical method to SUSY 2 to 6 processes likedu → W−Z → (l−νγ)(l−l+γ)with Howie Baer and Xerxes Tata. Relative phase between u and v spinors(v = CuT) is fixed.

1987 I calculated e+e− → nγ with Nick Brown and Alan D. Martin, but the computer codefor large n was slow. In addition to the n! permutations, 2n terms needed to beevaluated for massive propagator in the old method. An idea of computing‘off-shell’ fermion wave function came into my mind after this work.

The vision of MadGraph and Feynrules

Johan Alwall

on 2011.05.03 at MG 2011 Spring in Fermilab

Johan Alwall - The Vision of MG/FR 1

The VISION of MadGraph and FeynRules

Johan AlwallFermilab

MadGraph Spring 2011, Fermilab, May 3, 2011

Johan Alwall - The Vision of MG/FRJohan Alwall - The Vision of MG/FR 4

The Beginnings


The Beginnings


First “offical” BSM


The Web Generation


Tools and BSM functionality

arXiv:0712.2770arXiv:1010.4255

arXiv:0805.2554

arXiv:0809.2410

arXiv:1007.3300

See talk by Kentarou


Speed

arXiv:0908.4403

arXiv:0909.5257

arXiv:1010.0748

See talks by Junichi and Yoshitaro


Next-to-leading Order

arXiv:1004.2905

arXiv:0908.4272

arXiv:1004.2905

arXiv:1103.0621

See talk by Valentin+ Automatic MC@NLO (to be presented by Paolo)+ Automatic SUSY@NLO (see talk by MadGOLEM team)

mailto:MC@NLO

mailto:SUSY@NLO


MadGraph

● MadGraph has evolved from pure matrix element generator to a multi-purpose platform for a large number of automatized tools

● Very strong trademark – very powerful tool● However, inherent limitations and rigid output

structure still place a limit on user friendliness and development possibilities

● Many developments have been veritable “tour-de-force” efforts with massive post-processing


FeynRules

arXiv:0806.4194


FeynRules


FeynRules

● Easy-to-use but powerful Mathematica package– Generation of Feynman rules from any Lagrangean

– Output of generator specific files using multiple generator translation interfaces

– Continuous new improvements and developments

● However – output limited by the capabilities of the target generator in terms of Lorentz/color structures, multiparticle vertices, effective vertices etc.– e.g., implemented HELAS routines

See talk by FeynRules team


Vision (anno 2008/9)

● Using FeynRules as a cornerstone, allows for unprecedented validation and testing of models, and efficient communication between theorists and experimentalists at any point in the simulation chain

arXiv:0906.2474


Replace this...

TH EXPIdea

Lagrangian

Feyn. rules

Amplitudes

Xsecs

Signature

Paper

PHENO

Aut. Feyn. rules

Any amplitude

Any xsec

Events

Paper

Pythia+PGS

More signatures

New MC

Paper

Pythia

Detector simulation

Data

New pythia

2→2 Amps


… by this!

TH EXPIdea

Lagrangian

FeynRulesFeynRules

ME generator

Signatures

Events

Pythia

Detector simulation

Data

PGS

Theorist

Experimentalist


… by this!

TH EXPIdea

Lagrangian

FeynRulesFeynRules

ME generator

Signatures

Events

Pythia

Detector simulation

Data

PGS

Theorist

Experimentalist

Communicationpossible at any level – optimal flexibility for all involved


… by this!

TH EXPIdea

Lagrangian

FeynRulesFeynRules

ME generator

Signatures

Events

Pythia

Detector simulation

Data

PGS

Theorist

Experimentalist

(and of course, papers can be written at any level as well)


Since then, we've raised the bar!

● Development of MadGraph 5 started in November 2009

● Explicit goal – remove all limitations on old MadGraph– Speed

– Number of particles

– Types of interactions

– Output languages and formats

– Flexibility and modularity

● Did we succeed?


Speed

Matrix Elementgeneration:

Matrix Elementevaluation (Fortran):

+ Ongoing workwith recursionfor multipartoncalculations in MG5, see talk by Yoshitaro


Number of particles


Types of interactions

Higgs Effective Theory Multi-fermion vertices


Types of interactions

Diquark cross sections with coupling 0.01

7 TeV LHC

Jet pT:s, fully matched

pp → D + 0,1,2 jets

Color sextet and εijk implementations


Output languages and formats#include "SigmaProcess.h"#include "Parameters_sm.h"using namespace std; namespace Pythia8 {//==========================================// A class for calculating the matrix elements for// Process: u u~ > t t~ QED=0 @1// Process: c c~ > t t~ QED=0 @1// Process: d d~ > t t~ QED=0 @1// Process: s s~ > t t~ QED=0 @1//--------------------------------------------------------------------------

class Sigma_sm_qq_ttx : public Sigma2Process { public:

// Constructor. Sigma_sm_qq_ttx() {} // Initialize process. virtual void initProc(); // Calculate flavour-independent parts of cross section. virtual void sigmaKin(); // Evaluate sigmaHat(sHat). virtual double sigmaHat(); // Select flavour, colour and anticolour. virtual void setIdColAcol(); ...

Pythia 8 C++ output

Compactified and optimized multiprocess output for MadEvent

+ Standalone Matrix Element output in Fortran and C++


But the biggest advances..

… are in a new interface from FeynRules!


But the biggest advances..

… are in a new interface from FeynRules!

Universal FeynRules Output (UFO)

ALOHA


The biggest advances

● Universal FeynRules Output (UFO) – Includes color and Lorentz structure

– Allows for complete specification of effective/non-renormalizable vertices

– Allows for automatic output of model parameter calculations for any model and language

● Automatic Language-independent Output of Helicity Amplitudes (ALOHA) – Automatic generation the necessary helicity

amplitude code for any new model (including effective theories, multi-fermion vertices,...) in Fortran/C++/Python/...

See talk by Olivier


Universal FeynRules Output (UFO)

vertices.py:V_2 = Vertex(name = 'V_2', particles = [ P.G, P.G, P.G ], color = [ 'f(1,2,3)' ], lorentz = [ L.VVV1 ], couplings = {(0,0):C.GC_4})

particles.py:G = Particle(pdg_code = 21, name = 'G', antiname = 'G', spin = 3, color = 8, mass = 'ZERO', width = 'ZERO', texname = 'G', antitexname = 'G', line = 'curly', charge = 0, LeptonNumber = 0, GhostNumber = 0)

lorentz.py:VVV1 = Lorentz(name = 'VVV1', spins = [ 3, 3, 3 ], Structure =

'P(3,1)*Metric(1,2) - P(3,2)*Metric(1,2) - P(2,1)*Metric(1,3) + P(2,3)*Metric(1,3) + P(1,2)*Metric(2,3) - P(1,3)*Metric(2,3)')

couplings.py:GC_4 = Coupling(name = 'GC_4', value = '-G', order = {'QCD':1})


ALOHA output SUBROUTINE VVV1_0(V1,V2,V3,C,VERTEX) IMPLICIT NONE DOUBLE COMPLEX V1(6) DOUBLE COMPLEX V2(6) DOUBLE COMPLEX V3(6) DOUBLE COMPLEX C DOUBLE COMPLEX VERTEX DOUBLE PRECISION P2(0:3),P3(0:3),P1(0:3)

P2(0) = DBLE(V2(5)) P2(1) = DBLE(V2(6)) P2(2) = DIMAG(V2(6)) P2(3) = DIMAG(V2(5)) P3(0) = DBLE(V3(5)) P3(1) = DBLE(V3(6)) P3(2) = DIMAG(V3(6)) P3(3) = DIMAG(V3(5)) P1(0) = DBLE(V1(5)) P1(1) = DBLE(V1(6)) P1(2) = DIMAG(V1(6)) P1(3) = DIMAG(V1(5))

VERTEX = C*( (V3(1)*( (V1(1)*( (V2(2)*( (0, -1)*P1(1)+(0, 1) $ *P3(1)))+( (V2(3)*( (0, -1)*P1(2)+(0, 1)*P3(2)))+(V2(4)*( (0, $ -1)*P1(3)+(0, 1)*P3(3))))))+( (V2(1)*( (V1(2)*( (0, 1)*P2(1) $ +(0, -1)*P3(1)))+( (V1(3)*( (0, 1)*P2(2)+(0, -1)*P3(2))) $ +(V1(4)*( (0, 1)*P2(3)+(0, -1)*P3(3))))))+( (P1(0)*( (0, 1) $ *(V2(2)*V1(2))+(0, 1)*(V2(3)*V1(3))+(0, 1)*(V2(4)*V1(4)))) ...

Fortran


ALOHA outputvoid VVV1_0(complex<double> V1[], complex<double> V2[], complex<double> V3[], complex<double> C, complex<double> & vertex){ double P2[4], P3[4], P1[4]; P2[0] = V2[4].real(); P2[1] = V2[5].real(); P2[2] = V2[5].imag(); P2[3] = V2[4].imag(); P3[0] = V3[4].real(); P3[1] = V3[5].real(); P3[2] = V3[5].imag(); P3[3] = V3[4].imag(); P1[0] = V1[4].real(); P1[1] = V1[5].real(); P1[2] = V1[5].imag(); P1[3] = V1[4].imag(); vertex = C * ((V3[0] * ((V1[0] * ((V2[1] * (complex<double> (0., -1.) * P1[1] + complex<double> (0., 1.) * P3[1])) + ((V2[2] * (complex<double> (0., -1.) * P1[2] + complex<double> (0., 1.) * P3[2])) + (V2[3] * (complex<double> (0., -1.) * P1[3] + complex<double> (0., 1.) * P3[3]))))) + ((V2[0] * ((V1[1] * (complex<double> (0., 1.) * P2[1] + complex<double> (0., -1.) * P3[1])) + ((V1[2] * (complex<double> (0., 1.) * P2[2] + complex<double> (0., -1.) * P3[2])) + (V1[3] * (complex<double> (0., 1.) * P2[3] + complex<double> (0., -1.) * P3[3]))))) + ((P1[0] * (complex<double> (0., 1.) * (V2[1] * V1[1]) + complex<double> (0., 1.) * (V2[2] * V1[2]) + complex<double> (0., 1.) * (V2[3] * V1[3]))) + (P2[0] * (complex<double> (0., -1.) * (V2[1] * V1[1]) + complex<double> (0., -1.) * (V2[2] * V1[2]) + complex<double> (0., -1.) * (V2[3] * V1[3]))))))) + ...

C++


So the foundation has been laid

… for a new, amazing set of tools

… for astonishing physics applications

… for making the LHC era a success!

Idea

Lagrangian

FeynRulesFeynRules

UFO/ALOHA/MG

Precision (NLO)Events Cosmology (DM) Multi-jet … (countless other possibilities!)

To be presented by Matt

FeynRules: status and plans

Claude Duhr


FeynRules

Claude Duhrin collaboration with N. D. Christensen and B. Fuks

MadGraph Meeting 2011Fermilab, 05 May 2011

Status and plans

Mittwoch, 4. Mai 2011

What is FeynRules..?

• FeynRules is a Mathematica package that allows to derive Feynman rules from a Lagrangian.

• The only requirements on the Lagrangian are:

➡ All indices need to be contracted (Lorentz and gauge invariance)

➡ Locality

➡ Supported field types: spin 0, 1/2, 1, 2 & ghosts



cp3

IntroductionFrom FeynRules to FeynArts so far

The new FeynRules interface to FeynArtsConclusion

Welcome in the FeynRules era

C. Degrande The new FeynRules interface

• FeynRules comes with a set of interfaces, that allow to export the Feynman rules to various matrix element generators.

• Interfaces coming with current public version

➡ CalcHep / CompHep

➡ FeynArts / FormCalc

➡ MadGraph 4 & 5

➡ Sherpa

➡ Whizard / Omega© C. Degrande



cp3

IntroductionFrom FeynRules to FeynArts so far

The new FeynRules interface to FeynArtsConclusion

For each tool, the right input

C. Degrande The new FeynRules interface

• FeynRules comes with a set of interfaces, that allow to export the Feynman rules to various matrix element generators.

• Interfaces coming with current public version:

➡ CalcHep / CompHep

➡ FeynArts / FormCalc

➡ MadGraph 4 & 5

➡ Sherpa

➡ Whizard / Omega© C. Degrande


• The input requested form the user is twofold.

How to use FeynRules

F[1] == {ClassName -> q, SelfConjugate -> False, Indices -> {Index[Colour]}, Mass -> {MQ, 200}, Width -> {WQ, 5} }

L = -1/4 FS[G,mu,nu,a] FS[G,mu,nu,a] + I qbar.Ga[mu].del[q,mu] - MQ qbar.q

• The Model File:Definitions of particles and parameters (e.g., a quark)

• The Lagrangian:

L = −14Ga

µν Gµνa + iq γµ Dµq −Mq q q


• Once this information has been provided, FeynRules can be used to compute the Feynman rules for the model:

How to use FeynRules

FeynmanRules[ L ]

• Equivalently, we can export the Feynman rules to a matrix element generator, e.g., for MadGraph 4,

WriteMGOutput[ L ]

• This produces a set of files that can be directly used in the matrix element generator (“plug ‘n’ play”).


Lagrangian

FeynArts

Translation Interfaces

TeX Feynman Rules

Model-fileParticles, parameters, ...

FeynRules

MadGraph CalcHep Sherpa

FeynRules

Whizard Golem Herwig


Weyl fermions and superfields

• Full support for two-component fermions.

• Full support for superfields, automatic solving of eq. of motions for auxiliary fields, etc.

Superfield action

Expansion into Grassmann variables

SUSY Lagrangian in component fields

Solve EoM for auxiliary fields

Physical 4D Lagrangian Old version

SF module

➡ See Benj’s talk


Spin 3/2

• The super-space module opens a completely new way to implement any SUSY, with gravitino.

• Spin 3/2 is fully supported (privately), and all vertices have been checked against the literature.

FA[mu,nu] fNbar[r].(gfNrG1.Ga[mu,r,t]+gfNrG2.Ga[mu,r,s].Ga[5,s,t]).rG[t,nu]

R[1] == { ClassName -> rG, SelfConjugate -> True, Mass -> {MrG, 1000}, Width -> {WrG, 20}, FullName -> "Spin 3/2 G particle" },


Sextet color structures

• So far, FeynRules was limited to triplet and octet color structures (mostly because there was no ME generator for higher representations).

• The new FeynRules version allows to include also particles in the sextet representation.

• Together with MG5, this allows to generate for the first time sextet production without having to ‘hack’ the ME generator.


FeynArts interface

• A new interface to FeynArts is being developedthat allows to implement arbitrary Lorentz structures.

(C. Degrande, CD)

FeynRules

FeynArts

‘.mod’-file ‘.gen’-fileModel-dependent couplings

(Model-independent) Lorentz structures

• This development goes along with a new version of FormCalc able to deal with multi-fermion interactions.


The UFO

UFO = Universal FeynRules Output

• Idea: Create Python modules that can be linked to other codes and contain all the information on a given model.

• The UFO is a self-contained Python code, and not tied to a specific matrix element generator.

• The content of the FR model files, together with the vertices, is translated into a library of Python objects, that can be linked to other codes.

• Golem, MadGraph 5 and Herwig++ will use the UFO.

• In particular, the UFO is the default model format for MadGraph 5.

(C. Degrande, CD, B. Fuks, D. Grellscheid, O. Mattelaer, T. Reiter)

➡ See Olivier’s talk


The UFO & ALOHA

• The development of the UFO goes hand in hand with the development of ALOHA.

• Idea: ALOHA uses the information contained in the UFO to create the required HELAS routines on the fly!

particles.py

vertices.py

couplings.py

lorentz.py

UFO

}ALOHA

MG 5


Model database


Validation of new models

• FeynRules does not only provide the power to develop and validate new models, but also to validate them to an unprecedented level!

• A given model can be output to more than one matrix element generator, and their results can be compared

➡ Different conventions➡ Different gauges➡ Different ways of handling large cancellations.

• This procedure can easily be automatized!


Conclusion

• The new version of FeynRules comes with lots of new features.➡ Superfields➡ Spin 3/2 (still private)➡ Sextets➡ New FeynArts interface➡ Interface to UFO, and hence to MadGraph 5

• The chain

is about to open a completely new era of HE phenomenology!

FR UFO ALOHA MG5


MadLoop

Valentin Hirschi


MadLoop

V a l e n t i n H i r s c h iE P F L

2 1 a p r i l 2 0 1 1

M a d G r a p h S p r i n g 2 0 1 1@ F e r m i L a b

Wednesday, May 4, 2011

Valentin Hirschi, 4 May 2011 MadGraph Spring 2011 @ FermiLab

Motivations - NLO Basics - Automation - Results - MG5, Future plans

NLO Basics

8

NLO contributions have two parts

�

md(4)σB

Virtual partCurrent bottleneck of NLO computations

Algorithms for automation known in principle but not yet efficiently implemented

This work brings automation using MadGraph and CutTools interfaced through MadLoop.

σNLO =

�

md(d)σV +

Real emission partAutomated for different methods

Challenge is the systematic extraction of singularities

MadFKS using the FKS subtraction method successfully implemented on MGv4

� ��

m+1d(d)σR+




Subtraction terms

9

IR divergences are dealt with using subtraction terms

σNLO=

�

md(d)σV

+

�

m+1d(d)σR

+

�

md(4)σB

➧

σNLO=

�

m

�d(4)σB

+

�

ld(d)σV

+

�

1d(d)σA

�+

�

m+1

�d(4)σR − d(4)σA

�

Each of the two integrals are separately finite.

The only missing input required from MadFKS is the finite part of the virtual amplitude.

This is the part MadLoop provides!




Towards automationMadGraph is a tree-level matrix element generator

MadFKS is built on MG and computes everything but the finite part of the amplitude.

CutTools is a code which computes very general loop integrals using the OPP method

10

↵Idea: To create a third-party program, MadLoop which interfaces these three tools to automatically compute some loop x-sections.




MadFKS

diverges as with

11

Phase-space: Divide and Conquer

|Mn+1|2 1

χ2i

1

1− yij

χi =Ei√s

yij = cos θij

Real emission part : dσR = |Mn+1|2dφn+1

Divide phase-space so that each partition has at most one soft and one collinear singularity

dσR =�

ij

Sij |Mn+1|2dφn+1

�

ij

Sij = 1

Use plus distribution to regulate the singularities

�dχ

�1

χ

�

+

f(χ) =

�dχ

f(χ)− f(0)

χ

dσR =�

ij

�1

χi

�

+

�1

1− yij

�

+

χ2i (1− yij)Sij |Mn+1|2dφn+1




CutTools

CutTools uses the OPP method for loop reduction at the integrand level

13

Or how to compute loops without doing so

N(q) =m−1�

i0<i1<i2<i3

�d(i0i1i2i3) + d(q; i0i1i2i3)

� m−1�

i �=i0,i1,i2,i3

Di

+m−1�

i0<i1<i2

[c(i0i1i2) + c(q; i0i1i2)]m−1�

i �=i0,i1,i2

Di

+m−1�

i0<i1

�b(i0i1) + b(q; i0i1)

� m−1�

i �=i0,i1

Di

+m−1�

i0

[a(i0) + a(q; i0)]m−1�

i �=i0

Di

+ P (q)m−1�

i

Di

A(q) =N(q)

D0D1 · · · Dm−1

Di = (q + pi)2 −m2

i , p0 �= 0 .

q2 = q2 + q2 (q · q) = 0

�d(d)σV =

�d(4+�)

�A(q) + A(q)

�

R2 can be obtained with a tree-level-like computation with special Feynman-Rules.

Evaluation of N(q) for different specific q’s allows to algebraically obtain the coefficients a, b, c and d

Reconstruction of the dependance of the numerator gives the cut-constructible part R1 of the finite part of the virtual amplitude

q

�A(q) → R2

�

Finite part = R1 + R2




MadLoop

14

Helping MadGraph and CutTools to be friends

Tree-like operations : Generate and obtain Born and R2 amplitudes

Loop diagrams : Generate cut-loop diagramsSelect a non-redundant basisCompute the color factorsProvide numerator of the loop integrantHandle exceptional PS points

Compute R2 contribution

Square born against virtual amplitude

Carry UV renormalization.

Perform Sanity checks (Double pole, Ward Identity, ...)




Cut-Loop diagrams

15

With a specific example

Consider : e+e− → γ → uu

Loop particles are denoted with a star. When MG is asked for it gives back eight diagrams. Two of them are: e+e− → u∗u∗uu

Selection is performed to keep only one cut-diagram per loop contributing in the process ≡

≡

Diag 1 = [u∗(6)g∗(5)u∗(A)]

Diag 3 = [u∗(A)u∗(6)g∗(5)]

Tags are associated to each cut-diagram. Those whose tags are mirror and/or cyc l ic permutations of tags of diagram already in the loop-basis are taken out.

Additional custom filter to eliminate tadpoles and bubbles attached to external legs.




MadLoop

16

New Features implemented

Recognition of the loop topologies in order to filter L-cut diagrams.

Structure to deal with two MG process simultaneously (the L-cut and the born-like).

Treat color to obtain the squaring of the loop color structure against the born one.

Right form of integrant to CutTools: no denominators, complex momenta and reconstruction of missing propagator for the sewed particles.

Implementation of QCD ghosts.

Implementation of the special R2 vertices and automatic UV renormalization.

�

i=−1,1

�µi (p)�ν∗i (p) → −gµν




MadLoop

19

How to make it faster

Recycling of the tree-structures attached to the loop.

Identify identical contributions (i.e. massless fermion loops of diff. flavors)

Call CT not per diagram, but per set of diagrams with the same loop kinematics.

Use of recursion relations: Big gain in reals, not so much in the virtual.




Local checks

20

Ref. [33] : A. van Hameren et al.

You don’t want the exhaustive list...

The numerics are pin-point on analytical data, even with several mass scales.

We believe the code is very robust - e.g., MadLoop helped spot mistakes in published loop computations ( , )W+W+jjZjj

Analytic computations from an independent implementation of the helicity amplitudes by J.J van der Bij et al.




Integrated Results

21

Running time: Two weeks on a 150+ node cluster

Proof of efficient EPS handling withZtt

Successful cross-check against known results

Large K-factors sometimes

No cuts on b, robust numerics with small PT




Distributions

22

Full machinery at work

Case study of with starring actors:[H/A]tt

MGv4, CT, MadFKS, MadLoop and aMC@NLO interfaced to Herwig6 !




Future plans

MG5 much more advantageous:

23

MGv5 to the rescue

Extreme programming and python leads to modular and flexible code

Test routines makes maintenance easy

Designed from the beginning for being fast and user-friendly.

Output in many formats and automatic writing of HELAS subroutines.

Short term plans:MadLoop @ MGv5 with the recycling optimizations

Get rid of ALL existing constraints.

Further term plans:Automate generation UV and R2 terms in FeynRules.

Recursion relations and computation of color ordered amplitudes.


Bridging Theory with Experiment

Fabio Maltoni

on 2011.05.09 at PHENO 2011 in Madison

Pheno 2011 - Madison Fabio Maltoni

BRIDGING THEORY WITH EXPERIMENTS

Fabio MaltoniCP3-UCLouvain

PHENO 2011May, 9th

Monday 9 May 2011


! !ab!X(x1, x2, "S(µ2

R),Q2

µ2

F

,Q2

µ2

R

)!X =!a,b

" 1

0

dx1dx2 fa(x1, µ2

F )fb(x2, µ2

F )

Two ingredients necessary:

1. Parton Distribution functions (from exp, but evolution from th).

2. Short distance coefficients as an expansion in αS (from th).

Master QCD formula

Monday 9 May 2011


• For low multiplicity include higher order terms in our fixed-order calculations (LO→NLO→NNLO...) ⇒

• For high multiplicity use the tree-level results

First way:

!ab!X = !0 + "S!1 + "2

S!2 + . . .

Comments:1. The theoretical errors systematically decrease.2. Pure theoretical point of view. 3. A lot of new techniques and universal algorithms have been developed. 4. Final description only in terms of partons and calculation of IR safe observables ⇒ not directly useful for simulations

TH

HOW WE (USED TO) MAKE PREDICTIONS?

Monday 9 May 2011

Wine and Cheese at FNAL May 6 2011 Fabio Maltoni

NLO BASICSNLO contributions have three parts

Virtual part

!�� !

md(d)!V +

! "# $

Real emission part

!

m+1d(d)!R+

!

md(4)!B

Born

The cost of a new prediction at NLO can easily exceed 100k$.

Loops have been for long the bottleneck of NLO computations

Virtuals and Reals are each divergent and subtraction scheme need to be used (Dipoles, FKS, Antenna’s)

A lot of work is necessary for each computation

Monday 9 May 2011


LOOP TECHNIQUES

modified by the speaker

Monday 9 May 2011


BEST EXAMPLE: MCFM Downloadable general purpose NLO code [Campbell & Ellis+ collaborators]

☞ ~30 processes

☞ First results implemented in 1998 ...this is 13 years worth of work of several people (~4M$)

☞ Cross sections and parton-level distributions at NLO are provided

☞ One general framework. However, each process implemented by hand.

Final state Notes Reference

W/Z

diboson(W/Z/γ)

photon fragmentation, anomalous couplings

hep-ph/9905386,arXiv:1105.0020

Wbb massless b-quarkmassive b quark

hep-ph/9810489arXiv:1011.6647

Zbb massless b-quark hep-ph/0006304

W/Z+1 jet

W/Z+2 jetshep-ph/0202176,hep-ph/0308195

Wc massive c-quark hep-ph/0506289

Zb 5-flavour scheme hep-ph/0312024

Zb+jet 5-flavour scheme hep-ph/0510362

Final state Notes Reference

H (gluon fusion)

H+1 jet (g.f.) effective coupling

H+2 jets (g.f.) effective couplinghep-ph/0608194,arXiv:1001.4495

WH/ZH

H (WBF) hep-ph/0403194

Hb 5-flavour scheme hep-ph/0204093

ts- and t-channel (5F),top decay included

hep-ph/0408158

t t-channel (4F)arXiv:0903.0005,arXiv:0907.3933

Wt 5-flavour scheme hep-ph/0506289

top pairs top decay included

Monday 9 May 2011


Sherpa’s artist

Events at hadron colliders

High Q2

Monday 9 May 2011


• Describe final states with high multiplicities starting from 2 →1 or 2 →2 procs, using parton showers, and then an hadronization model.

Second way:

EXP

Comments:

1. Fully exclusive final state description for detector simulations2. Normalization is very uncertain3. Very crude kinematic distributions for multi-parton final states 4. Improvements are only at the model level.

most known and used : PYTHIA, HERWIG

HOW WE (USED TO) MAKE PREDICTIONS?

Monday 9 May 2011


SM STATUS A FEW YEARS AGOpp→ n particles

complexity [n]1 32 54 6 87 9 10

accuracy [loops]

0

1

2 fully exclusive

fully inclusive

parton-level

Monday 9 May 2011


• Workload is tripled!

• Long delays due to localized expertise and error prone. Painful validations are necessary at each step.

• It leads to a proliferation of private MC tools/sample productions impossible to maintain, document and reproduce on the mid- and long- term.

• Just publications is a very inefficient way of communicating between TH/PHENO/EXP.

BSM TH/EXP INTERACTIONS : THE OLD WAY

Monday 9 May 2011


1. have the possibility of making collider studies for any BSM theory by knowing the Lagrangian (and benchmarks).

2. that our EXP/TH results could be directly used by the TH/EXP colleagues.

3. have the accuracy of a NLO prediction with the flexibility of parton shower/hadronization

4.have the above to for ANY SM background as well as for ANY BSM signals.

5. have them all available at the touch of a button.

“GAP ANALYSIS” (2003 CA)

We would like to:

OK? REACTIONS?

Monday 9 May 2011

Pheno 2011 - Madison Fabio MaltoniMonday 9 May 2011


QCD AND MC (SIMPLIFIED) PROGRESS

2003

20112008 2009

Matching:ME+PS

NLOwPS

New Looptechniques

BSM framework

Fully Automatic NLOwPS

Monday 9 May 2011


1. hadron-level description1. parton-level description

Matrix Element Shower MC

2. fixed order calculation 2. resums large logs

4. valid when partons are hard and well separated 4. valid when partons are

collinear and/or soft5. nedeed for realistic studies

Approaches are complementary: merge them!

3. quantum interference exact 3. quantum interference through angular ordering

5. needed for multi-jet description

Difficulty: avoid double counting

[Mangano][Catani, Krauss, Kuhn, Webber][Frixione, Nason, Webber]

.

ME with PS

Monday 9 May 2011


Merging fixed order with PS

SHERPA

...

...

PS →

ME ↓

Double counting of configurations that can be obtained in different ways (histories). All the matching algorithms (CKKW, MLM,...) apply criteria to select only one possibility based on the hardness of the partons. As the result events are exclusive and can be added together into an inclusive sample. Distributions are accurate but overall normalization still “arbitrary”.

[Mangano][Catani, Krauss, Kuhn, Webber]

Monday 9 May 2011


W+JETS FROM TEVATRON TO LHC

) [GeV]minTJet Transverse Energy (E

0 50 100 150 200

[pb]

T)d

ET

/dE

!(d

min

TE"

-210

-110

1

10

210

CDF Run II Preliminary n jets#) + $e%(WCDF Data -1dL = 320 pb"W kin: 1.1&| e

' 20[GeV]; |# eT E

30[GeV]# $T]; E2 20[GeV/c# WT M

Jets: |<2.0'JetClu R=0.4; |hadron level; no UE correction

LO Alpgen + PYTHIA normalized to Data!Total

jetst1

jetnd2

jetrd3

jetth4

Working Amazingly well!

Monday 9 May 2011


NLOWPSProblem of double counting becomes even more severe at NLO* Real emission from NLO and PS has to be counted once* Virtual contributions in the NLO and Sudakov should not overlap

Current available (and working) solutions: MC@NLO [Frixione, Webber, 2003; Frixione, Nason, Webber, 2003] - Matches NLO to HERWIG angular-ordered PS. - Some events have negative weights. - Available also for Pythia now. - Automation [Frederix, Frixione, Torrielli] POWHEG [Nason 2004; Frixione, Nason, Oleari, 2007] - Is independent from the PS. It can be interfaced to PYTHIA or HERWIG. - Generates only positive unit weights. - Can use existing NLO results via the POWHEG-Box [Aioli, Nason, Oleari, Re et al. 2009]

- Used by HELAC [Kardos, Papadopoulos, Trocsanyi 1101.2672] and SHERPA [Hoeche,Krauss, Schooenner, Siegert, 1008.5399]

Monday 9 May 2011


SM STATUS : SINCE 2007pp→ n particles

complexity [n]1 32 54 6 87 9 10

accuracy [loops]

0

1

2fully exclusive

fully inclusive

parton-level

Monday 9 May 2011


SM STATUS : SINCE 2007pp→ n particles

complexity [n]1 32 54 6 87 9 10

accuracy [loops]

0

1

2

fully exclusive and automatic

fully exclusive

fully inclusive

parton-level

Monday 9 May 2011


Cost saving

Programs are modular and computations based on elements that can be systematically and extensively checked. Trust can beeasily built.

Robustness

One framework for all. Available to everybody for an unlimited set of applications for all. Suitable to EXP collaboration.

Wide accessibility

Trade human time and expertise spent on computing one process at the time with time on physics and pheno.

AUTOMATION

Monday 9 May 2011



2003

20112008 2009

Matching:ME+PS

NLOwPS

New Looptechniques

BSM framework


Monday 9 May 2011


For the calculation of one-loop matrix elements, several methods are now established :

•Generalized Unitarity (ex. BlackHat, Rocket,...)[Bern, Dixon, Dunbar, Kosower, hep-ph/9403226 + ....; Ellis, Giele, Kunszt 0708.2398, +Melnikov 0806.3467]

•Integrand Reduction (ex. CutTools, Samurai) [Ossola, Papadopolulos, Pittau, hep-ph/0609007; del Aguila, Pittau, hep-ph/0404120; Mastrolia, Ossola, Reiter, Tramontano, 1006.0710]

•Tensor Reduction (ex. Golem)[Passarino, Veltman, 1979; Denner, Dittmaier, hep-ph/0509141, Binoth, Guillet, Heinrivh, Pilon, Reiter 0810.0092]

NEW LOOP TECHNIQUES

Monday 9 May 2011


GUINNESS WR NLO CALCULATIONS

W+4 jets tt+2jets[Berger et al., 1009.2338] [Bevilacqua et al., 1002.4009]

HELAC-PHEGAS + CutTools

Both based on unitarity methods and recursive relations for trees.

Monday 9 May 2011


Slide from L. Dixon

Monday 9 May 2011



2003

20112008 2009

Matching:ME+PS

NLOwPS

New Looptechniques

BSM framework


Monday 9 May 2011


Idea

TH

Aut. Feyn. Rules

Any amplitude

Any x-sec

partonic events

PHENO

Lagrangian

Pythia

EXP

Detec. Sim.

Data

PGS

Pythia

Paper

Feyn. Rules

Amplitudes

x secs

Paper

New MC

Paper

New Pythia

Amps 2→2


Monday 9 May 2011


Idea

TH

Aut. Feyn. Rules

Any amplitude

Any x-sec

partonic events

PHENO

Lagrangian

Pythia

EXP

Detec. Sim.

Data


Monday 9 May 2011


Idea

Lagrangian

FeynRules

ME Generator

Signal & Bkg

Events

PS+Had

Detect. Sim.

Data

PGS

TH EXP

Papers

BSM TH/EXP INTERACTIONS : THE NEW PATH

Monday 9 May 2011


Idea

Lagrangian

FeynRules

ME Generator

Signal & Bkg

Events

PS+Had

Detect. Sim.

Data

๏ One path for all ๏ Physics and software validations streamlined๏ Robust and efficient Th/Exp communication๏ It works top-down and bottom-up

PGS

Complete automatization LO (and now NLO) calculations available, including merging with the parton shower in multi-jet final states, for SM as well as for BSM physics. Automatization of NLO is very promising now...

TH EXP

Papers

BSM TH/EXP INTERACTIONS : THE NEW PATH

Monday 9 May 2011


THE FEYNRULES PROJECT

• Public database, user driven, easy legacy

• All automatic ME generators supported

• Unprecedented validation and robustness

• It can be systematically improved/extended

• Superfield notation, higher spin-particles, more...

• User driven easy new functionalities (eg, NLO)

Monday 9 May 2011


Example: BSM multijet final states

pp→X6 +jets pp→Graviton (ADD&RS) +jets

[Alwall 2011] [de Aquino, Hagiwara, Qiang Li, FM, 2011]

Monday 9 May 2011


QCD AND MC PROGRESS

2003

20112008 2009

Matching:ME+PS

NLOwPS

New Looptechniques

BSM framework


Monday 9 May 2011


MadGraph

THE JOINT VENTURE

MC@NLOCutTools

http://amcatnlo.cern.ch

FKS

FKS

Modular structure:

• MadLoop or External Tool (via Binoth LH accord)

• MadFKS

• MC@NLO counterterms

• Interfaced to Herwig (Pythia in progress)

Monday 9 May 2011




AUTOMATIC NLO IN SM MADFKS+MADLOOP

Running time: Two weeks on a 150+ node cluster

Proof of efficient EPS handling with ttZ.

Successful cross-check against known results (and bugs found in other NLO codes Zjj, W+W+jj)

Total cross sections at the LHC for 26 sample procs

Very loose cuts just when needed

[Hirshi, Frederix, Frixione, FM, Garzelli, Pittau, Torrielli, 1103.0621].

Monday 9 May 2011


NLO results known (but no public code) for scalar Higgs since some time. No results for pseudoscalar known. First fully automatic results for both H and A [Frederix, Frixione, Hirschi, FM, Pittau, Torrielli,1104.5613].

ttH/ttA ZZ→4l (W→eν)bb

FIRST APPLICATIONS

Monday 9 May 2011


NLO calculation includes γ*/Z interference, full spin correlations and single resonant diagrams.Equivalent at pure NLO to MCFM.

MadLoop (CutTools)+MadFKS+MC@NLO+Herwig6

FIRST APPLICATIONS


Monday 9 May 2011


Several NLO results available since some time but all with approximations (ie, mb=0 or no spin correlations). No approximation here and NLO equivalent to recent [Badger, Campbell, Ellis, 1011.6647].

Solid: w/ spin correlations Dashed: w/o spin correlations


FIRST APPLICATIONS

Monday 9 May 2011


SM STATUS : SINCE LAST WEEKpp→ n particles

complexity [n]1 32 54 6 87 9 10

accuracy [loops]

0

1

2 fully exclusive and automatic

fully exclusive

fully inclusive

parton-level

aMC@NLO (MadLoop+MadFKS+MC@NLO)

Monday 9 May 2011


pp→ n particles

complexity [n]1 32 54 6 87 9 10

accuracy [loops]

0

1

2

fully exclusive

fully inclusive

parton-level

fully exclusive and automatic:

near future

done

BSM STATUS AND OUTLOOK

Monday 9 May 2011


CONCLUSIONS✦ The need for better description and more reliable predictions for SM

processes for the LHC has motivated a significant increase of theoretical and phenomenological activity in the last years, leading to several important achievements.

✦ A new generation of tools and techniques is now available. Full Automation

of Accurate (NLO) computations at fixed order as well as their the matching to parton-shower has been proven for the SM.

✦ Amazingly efficient, flexible and robust BSM simulation chain available and being continuously improved. Same level of sophistication as SM processes can be attained. Both top-down and bottom-up approaches included.

✦ EXP/TH interactions enhanced by a new framework and not limited anymore by the burden of heavy/long and inefficient calculations...

A

A

A

Monday 9 May 2011


CREDITS

• Thanks to all the MadGraph team/collaborators/friends for continuous and exciting collaborations

• Thanks to the MC community for always fruitful collaborations.

• The material (and very often the presentation itself) shown in this talk is the work of many people, including Claude Duhr, Stefano Frixione, Valentin Hirshi, Rikkert Frederix, Johan Alwall,...

Monday 9 May 2011

How shall we enjoy LHC physics ?

Tools like MadGraph will not only help

experimentalists but also theorists to

contribute to the new physics discovery.

For those of you who wish to master basic tools to simulateLHC physics, an excellent school is organized in Kyoto thisSeptember.

2011 IPMU School and Workshop on Monte CarloTools for LHC

5-10 September 2011

Yukawa Institute for Theoretical PhysicsKyoto University

Excellent lecturers Frank Krauss, Bryan Webber, GavinSalam, Torbjorn Sjostrand, including two from the Mad-Graph team, Rikkert Frederix and Claude Duhr, will givelectures and do tutoring.

Application deadline is Jun/20. Please apply through thewebpagehttp://www.hep.phy.cam.ac.uk/theory/mcschool11/

Following the series of annual schools on Monte Carlo event generators for LHC, held at Durham (2007), Debrecen (2008), Lund (2009) and Karlsruhe

(2010), the 2011 school will be held in conjunction with a mini-workshop at

Yukawa Institute for Theoretical Physics, Kyoto University

5 - 10 September 2011

The School will provide a four day course of training in the physics and techniques used in modern Monte Carlo event generators via a series of lectures and practical sessions. The school

is aimed at advanced doctoral students and young postdocs. It will be followed by a two day mini-workshop on the latest developments in event generators and their applications at LHC.

Please check the indico server for updated information on travel, accommodation, programme etc.

2011 IPMU School and Workshop on Monte Carlo Tools for LHC

1/4 ページMonte Carlo Tools for LHC School and Workshop

2011/06/06http://www.hep.phy.cam.ac.uk/theory/mcschool11/

Topics to be covered in the Workshop: There will be a mixture of invited and submitted contributions, to be decided later, on Monte Carlo issues relevant to the LHC.

Applications, travel and other information Applications must be received by 20 June 2011. Applications are particularly encouraged from women and other under-represented sections of the community. Please note that enrollment will be limited to 60 participants. There is no registration fee but accommodation will be booked at a downtown hotel at a cost to participants of approximately US$80 per night. Lunches at economical prices will be taken at the Kyoto University campus cafeteria. A limited number of bursaries will be available to cover local costs for participants in financial need. The application form as well as further information about the School and Workshop can be obtained from the indico server.

Introduction to Monte Carlo Event Generators (Frank Krauss, IPPP, Durham)Introduction to Jet Finding and Jetography (Gavin Salam, CERN, Paris and Princeton)Introduction to MadGraph/MadEvent (Rikkert Frederix, Zurich)Beyond Standard Model Monte Carlo with FeynRules (Claude Duhr, IPPP, Durham)Matching Fixed Order and Parton Shower Generators (Bryan Webber, Cambridge)Modelling of Minimum Bias Events and Underlying Events (Torbjörn Sjöstrand, Lund) Tutorials and Hands-on Experience: Herwig++ (David Grellscheid, Durham) Pythia 8 (Torbjörn Sjöstrand, Lund) Sherpa (Marek Schönherr, Dresden) MadGraph/MadEvent (Rikkert Frederix, Zurich) FeynRules (Claude Duhr, Durham)



Contact and enquiries Enquiries should be sent to the organisers.

Local Organizers:

Kaoru Hagiwara, Shigeki Matsumoto, Takeo Moroi, Mihoko Nojiri (KEK/IPMU)

Organizing Committee:

David Grellscheid, University of Durham Hitoshi Murayama, Director, IPMU

Michael Seymour, University of Manchester Peter Skands, CERN

Bryan Webber, University of Cambridge

Sponsored by IPMU



event home page | view: | focus on: | details: | manage | IPPP style - - all days - - - - all sessions - - contribution login

2011 IPMU School and Workshop on Monte Carlo Tools for LHC

from Monday 05 September 2011

(09:00) to

Saturday 10 September 2011 (12:30)

at Yukawa Institute for Theoretical Physics (YITP)

Description: Following the series of annual schools on Monte Carlo event generators for LHC, held at Durham (2007), Debrecen (2008), Lund (2009) and Karlsruhe (2010), the 2011 School will be held in conjunction with a mini-workshop at the Yukawa Institute for Theoretical Physics, Kyoto University. school: 5 - 8 September 2011 workshop: 9/10 September 2011

Monday 05 September 2011 | Tuesday 06 September 2011 | Wednesday 07 September 2011 | Thursday 08 September 2011 | Friday 09 September 2011 | Saturday 10 September 2011 |

Monday 05 September 2011 top

09:00 Introduction to Monte Carlo event generators 1 (1h00') Frank Krauss (IPPP, University of Durham)

10:00 Introduction to jet finding and jetography 1 (1h00') Gavin Salam (LPTHE Paris) 11:00 Coffee break

11:30 Introduction to MadGraph/MadEvent 1 (1h00') Rikkert Frederix (University of Zurich) 12:30 Lunch break

13:45 Poster session 1 (45')

14:30 Event generators tutorial/discussion 1 (3h00')

Tuesday 06 September 2011 top


10:00 Introduction to jet finding and jetography 2 (1h00') Gavin Salam (LPTHE Paris) 11:00 Coffee break

11:30 Introduction to MadGraph/MadEvent 2 (1h00') Rikkert Frederix (University of Zurich)

1/3 ページ2011 IPMU School and Workshop on Monte Carlo Tools for LHC (05-10 September 2011)

2011/06/06http://conference.ippp.dur.ac.uk/conferenceOtherViews.py?view=ippp&confId=309

12:30 Lunch break



Wednesday 07 September 2011 top


10:00 Matching fixed order and parton shower generators 1 (1h00') Bryan Webber (Cambridge University) 11:00 Coffee break

11:30 Modelling of minimum bias and underlying events (1h00') Torbjörn Sjöstrand (Lund U.) 12:30 Lunch break



Thursday 08 September 2011 top


10:00 Matching fixed order and parton shower generators 2 (1h00') Bryan Webber (Cambridge University) 11:00 Coffee break

11:30 Beyond Standard Model Monte Carlo with FeynRules (1h00') Claude Duhr (IPPP) 12:30 Lunch break


14:30 FeynRules/MadGraph tutorial/discussion (3h00')

Friday 09 September 2011 top

09:00->11:00 Workshop session 1 11:00 Coffee break

11:30->12:30 Workshop session 2 12:30 Lunch break

2/3 ページ2011 IPMU School and Workshop on Monte Carlo Tools for LHC (05-10 September 2011)

2011/06/06http://conference.ippp.dur.ac.uk/conferenceOtherViews.py?view=ippp&confId=309

on recent developements in madgraph - 大阪大学 · 2011-06-08 · on recent developements in...

Documents