Les recherches de nouvelles particules au LHC : quel role pour la … · 2009. 3. 24. · Les...

Les recherches de nouvelles particules au LHC :quel role pour la QCD ?

Gavin Salam

LPTHE, CNRS and UPMC (Univ. Paris 6)

Rencontre de Physique des Particules 2009Ecole Polytechnique, Palaiseau

24 mars 2009

Avec des exemples tires de travaux avec Butterworth, Davison & Rubinet Cacciari, Rojo & Soyez

QCD & Searches, G. Salam (p. 2)

Introduction An often-repeated argument

LHC collides quarks and gluons

Quarks and gluons interact strongly → huge QCD

backgrounds

Therefore we will need to rely on our understanding ofQCD in order to make discoveries at LHC.

True, false, or only half the story?

Introduction An often-repeated argument

LHC collides quarks and gluons

Quarks and gluons interact strongly → huge QCD

backgrounds

Therefore we will need to rely on our understanding ofQCD in order to make discoveries at LHC.

True, false, or only half the story?

Introduction Contrasting views (1)

It must be true, otherwise why would there be such a large effortdevoted to LHC-QCD calculations?

◮ Parton shower Monte Carlo Generators Pythia, Herwig, Sherpa

◮ LO tree-level calculations Alpgen, Madgraph, Sherpa, ...

◮ NLO calculations ∼ 50 people

◮ NNLO calculations Higgs, W/Z, next step jets

◮ All-orders calculations resummations, SCET

◮ Parton Distribution Functions (PDFs) CTEQ, MSTW, NNPDF, . . .

Introduction Contrasting views (2)

Healthy scepticism

[...] unless each of the background components can be separatelytested and validated, it will not be possible to draw conclusions fromthe mere comparison of data against the theory predictions.

I am not saying this because I do not believe in the goodness of ourpredictions. But because claiming that supersymmetry exists is far tooimportant a conclusion to make it follow from the straight comparisonagainst a Monte Carlo.

Mangano, 0809.1567

Introduction This talk

Try to examine the question of howmuch QCD matters, how much it can

help with searches.

Introduction What kinds of searches?

mass peak

Signal

QCDprediction

New resonance (e.g. Z ′) where you see alldecay products and reconstruct an invari-ant mass

QCD may:

◮ swamp signal

◮ smear signal

leptonic case easy; hadronic case harder

mass edge

Signal

QCDprediction

New resonance (e.g. R-parity conservingSUSY), where undetected new stable par-ticle escapes detection.

Reconstruct only part of an invariant mass→ kinematic edge.

QCD may:

◮ swamp signal

◮ smear signal

high−mass excess

Signal

QCDprediction

Unreconstructed SUSY cascade. Study ef-fective mass (sum of all transverse mo-menta).

Broad excess at high mass scales.

Knowledge of backgrounds is crucial isdeclaring discovery.

QCD is one way of getting handle on back-ground.

Introductiond

σ / d

Signal ?

CONTINUE(briefly) HERE

Signal ?

STARTHERE

Introductiond

σ / d

Signal ?

STARTHERE

Introductiond

σ / d

Signal ?

STARTHERE

Introduction

Before Starting

Introduction

The most pervasive role of QCD at LHC

Every single paper that comes out from the ATLAS and CMS ppphysics programmes will involve the use of one or more QCD-basedparton-shower Monte Carlo event generators: Pythia, Herwig orSherpa.

For simulating physics signals.

For simulating background signals.

For simulating pileup.

As input to simulating detector respone.

Introduction Pythia

Predicting QCD

Predicting QCD SUSY example: gluino pair production

Signal

Signal Background

jet jet

Predicting QCD SUSY searches: what excesses?

SUSY ≃ factor 5−10 excess

Predicting QCD How accurate is QCD?

αs ≃ 0.1

That implies LO QCD should beaccurate to within 10%

It isn’t

Rules of thumb:

LO good to within factor of 2

NLO good to within scaleuncertainty

Anastasiou, Melnikov & Petriello ’04

Anastasiou, Dissertori & Stockli ’07

αs ≃ 0.1

It isn’t

Rules of thumb:

Anastasiou, Melnikov & Petriello ’04

Anastasiou, Dissertori & Stockli ’07

αs ≃ 0.1

It isn’t

Rules of thumb:

Predicting QCD Control samples

We don’t have NLO for the back-ground (e.g. 4 jets + Z, a 2 → 5process).

Only LO (matched with parton show-ers). How does one verify it?

Common procedure (roughly):

◮ Get control sample at low pt

◮ SUSY should be small(er)contamination there

◮ Once validated, trust LOprediction at high-pt

Predicting QCD Is this safe?

A conservative QCD theory point of view:

It’s hard to be sure: since we can’t calculate Z+4 jets beyond LO.But we would tend to think it is safe, as long as control data

are within usual factor of two of LO prediction

Illustrate issues with toy example: Z+jet production

◮ It’s known to NLO and a candidate for “first” 2 → 2 NNLO

∼ e+e− → 3 jets, NNLO: Gehrman et al ’08, Weinzierl ’08

◮ But let’s pretend we only know it to LO, and look at the pt distributionof the hardest jet (no other cuts — keep it simple)

q (=jet)

example based on

background work for

Butterworth, Davison,

Rubin & GPS ’08

Predicting QCD Toy data, control sampledσ

pt,jet [GeV]

Z + jet cross section (LHC)

µ2 = mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

toy data

100000

40 60 80 100

stage 1: get control sample

Check LO v. data at low pt

◮ normalisation off by factor 1.5(consistent with expectations)

So renormalise LO by K-fact

◮ shape OKishDon’t be too fussy: SUSY

could bias higher pt

pt,jet [GeV]

µ2 = mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

toy data

LO x K-fact (1.5)

100000

40 60 80 100

pt,jet [GeV]

µ2 = mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

toy data

LO x K-fact (1.5)

100000

40 60 80 100

Predicting QCD Toy data, high ptdσ

pt,jet [GeV]

µ2 = mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

LO x K-fact (1.5)

100000

0 200 400 600 800 1000

stage 2: look at high pt

◮ good agreement at low pt , byconstruction

◮ excess of factor ∼ 10 at highpt

◮ check scale dependence of LO[NB: not always done except

e.g. Alwall et al. 0706.2569]

still big excess

pt,jet [GeV]

µ2 = mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

LO x K-fact (1.5)

toy data

100000

0 200 400 600 800 1000

still big excess

pt,jet [GeV]

µ2 = {mZ2 + p2

t,Z, ⟨pt,jets⟩2, HT2, MZ

kt alg., R=0.7

CTEQ6M

MCFM 5.2

LO x K-fact (1.5)

toy data

LO scale dep

100000

0 200 400 600 800 1000

still big excess

Predicting QCD What’s in the toy data?

Is it:

◮ QCD + extra signal?

◮ just QCD? But then where does a K -factor of 10 come from?

Here it’s just a toy illustration. In a year or two it may be for real:

◮ Do Nature / Science / PRL accept the paper?

Discovery of New Physics at the TeV scaleWe report a 5.7σ excess in MET + jets production that is consistentwith a signal of new physics . . .

◮ Do we proceed immediately with a linear collider?It’ll take 10–15 years to build; the sooner we start the better

◮ At what energy? It would be a shame to be locked in to the wrong energy. . .

Predicting QCD What’s in the toy data?

Is it:

◮ QCD + extra signal?

◮ just QCD? But then where does a K -factor of 10 come from?

Here it’s just a toy illustration. In a year or two it may be for real:

◮ Do Nature / Science / PRL accept the paper?

Discovery of New Physics at the TeV scaleWe report a 5.7σ excess in MET + jets production that is consistentwith a signal of new physics . . .

◮ Do we proceed immediately with a linear collider?It’ll take 10–15 years to build; the sooner we start the better

◮ At what energy? It would be a shame to be locked in to the wrong energy. . .

Predicting QCD Open the box. . .dσ

pt,jet [GeV]

µ2 = {mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

LO x K-fact (1.5)

toy data

LO scale dep

100000

0 200 400 600 800 1000

Unlike for SUSY multi-jetsearches, in the Z+jet case wedo have NLO.

Once NLO is included the excessdisappears

The “toy data” were just the

upper edge of the NLO band

Hold on a second: how doesQCD give a K-factor O (5 − 10)?

NB: DYRAD, MCFM consistent

pt,jet [GeV]

µ2 = {mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

LO x K-fact (1.5)

toy data

LO scale dep

100000

0 200 400 600 800 1000

pt,jet [GeV]

µ2 = {mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

LO x K-fact (1.5)

toy data

LO scale dep

100000

0 200 400 600 800 1000

Predicting QCD Why the large K -factor?

Leading Order

q (=jet)

αsαEW

Next-to-Leading Order

Z q (=jet)

g (=jet)

α2sαEW ln2 pt

LHC will probe scales well above EW scale,√

s ≫ MZ .QCD and EW effects mix, EW bosons are light.

New logarithms (enhancements) appear.

Leading Order

q (=jet)

αsαEW

q (=jet)

g (=jet)

α2sαEW ln2 pt

Leading Order

q (=jet)

αsαEW

q (=jet)

g (=jet)

α2sαEW ln2 pt

Predicting QCD Cross-checkdσ

pt,jet [GeV]

µ2 = {mZ2 + p2

kt alg., R=0.7

CTEQ6M

MCFM 5.2

NLOptZ NLO

100000

0 200 400 600 800 1000

Plot distribution for ptZ .

This selects events in which theZ is the hardest object.

Kills diags with EW double-logs.

NLO is well-behaved.

Predicting QCD Where does this leave us?

◮ Excess ≡ New Physics, iff you are really, really sure you understandbackgrounds;

◮ Control samples may not be good enough cross-check

◮ Plain LO QCD can be misleading, understanding the physics is crucialThis can be non-trivial even in simplest of cases

But can help you choose good observables

◮ NLO provides a powerful cross check — and progress is being made inmulti-jet case, e.g. W + 3jet calculation @ NLO

BlackHat: Berger et al. ’08, ’09

Rocket: Ellis, Kunszt, Giele, Melnikov & Zanderighi ’08, ’09

◮ What about MLM, CKKW matching for combining different tree-levelcontributions? Designed to avoid deficiences of Parton Showers

But does more — a sort of “LO++”. Still, not NLO

In this case it actually gets most of the answer

[checked by de Visscher and Maltoni]

Predicting QCD

Viewing QCD

(A subject that has seen much less

attention than “predicting QCD,”

but may be just as relevant)

Predicting QCD

Consider case of mass peaks — but bear in mind thatother kinematic structures are fundamentally related.

Signal ?

Predicting QCD Some peaks are easy — QCD not needed

e.g. resonance → ℓ+ℓ−, or big broad resonance to jets

Dijet Mass (GeV)0 1000 2000 3000 4000 5000 6000 7000

CMS Simulation

Dijet Mass (GeV)0 1000 2000 3000 4000 5000 6000 7000

Bhatti et al (for CMS), study of dijet mass resonances (q∗), 0807.4961

Predicting QCD Observability may depend on parameters

RS KK resonances, from Frederix & Maltoni, 0712.2355

Cases where QCD has the most to contribute are those that are borderline

Predicting QCD

Basic question:Can we make kinematic “structures” emerge more clearly?

Predicting QCD

Which particles should onechoose in order to best

reconstruct the resonance?

Predicting QCD

How should onedefine the “jets”?

Predicting QCD

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

SISCone, R=0.3, f=0.75Qw

f=0.12 = 54.3 GeV

Predicting QCD

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

f=0.12 = 28.8 GeV

Predicting QCD

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

f=0.12 = 21.3 GeV

Predicting QCD

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

f=0.12 = 30.6 GeV

Predicting QCD

dijet mass [GeV]

qq, M = 2000 GeVarX

iv:0810.1304

1900 2000 2100

f=0.12 = 54.3 GeV

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

f=0.12 = 28.8 GeV

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

f=0.12 = 21.3 GeV

dijet mass [GeV]

qq, M = 2000 GeV

arXiv:0810.1304

1900 2000 2100

f=0.12 = 30.6 GeV

Choice of jet-definition

has significant impact

Predicting QCD Recent progress with jets

◮ Definitions that are sensible within QCD and experiments (infraredsafety, speed) Cacciari & GPS ’05; GPS & Soyez ’07

Cacciari, GPS & Soyez ’08

◮ Understanding analytically the interplay between jet definitions & QCDDasgupta, Magnea & GPS ’07

Cacciari & GPS ’07; Cacciari, GPS & Soyez ’08

◮ Monte Carlo studies to verify consequences for experimentAnastasiou et al ’08; Nojiri & Takeuchi ’08

Cacciari, Rojo, GPS & Soyez ’08

Thaler & Wang ’09

◮ Jet tools for events with hierarchies of scalesButterworth, Davison, Rubin & GPS ’08

Brooijmans ’08; Krohn, Thaler & Wang ’08

Kaplan, Rehermann, Schwartz & Tweedie ’08

Almeida et al. ’08

Predicting QCD Recent progress with jets

◮ Definitions that are sensible within QCD and experiments (infraredsafety, speed) Cacciari & GPS ’05; GPS & Soyez ’07

Cacciari, GPS & Soyez ’08

◮ Understanding analytically the interplay between jet definitions & QCDDasgupta, Magnea & GPS ’07

Cacciari & GPS ’07; Cacciari, GPS & Soyez ’08

◮ Monte Carlo studies to verify consequences for experimentAnastasiou et al ’08; Nojiri & Takeuchi ’08

Cacciari, Rojo, GPS & Soyez ’08

Thaler & Wang ’09

◮ Jet tools for events with hierarchies of scalesButterworth, Davison, Rubin & GPS ’08

Brooijmans ’08; Krohn, Thaler & Wang ’08

Kaplan, Rehermann, Schwartz & Tweedie ’08

Almeida et al. ’08

Rather than go into detail on these subjects myself, I’llhand over to Mathieu Rubin so that he can tell you about

one of the most unexpected of the results.

Conclusions

Conclusions Conclusions

We’ve seen examples where doing the QCD “right” makes a big difference.

From first part: it’s clear that relative O (αs) (“the details”) in QCDpredictions (NLO) may be more than just a luxury refinement.

Part of the motivation for the large calculational effort in the field

Crucial in building confidence in our understanding of any LHC “excess”

From second (brief!) part: QCD at LHC it not just about calculatingbackgrounds. Learning to “view” events properly can have a major impact.

QCD can guide us in making good choices

A much smaller field — but several groups making progress

Crucial in order to maximise LHC’s sensitivity to new physics

Common theme: LHC will probe a broad range of scales: from below EWscale, to 1.5 orders of magnitude above it.

EXTRAS

Large K -factors

Mangano, 0809.1567

Not matched

But see 2-jet ≃ 1-jet, which is sign of problems

Is there a larger excess when plottedv. MET (∼ ptZ )?

Is this because Eff.Mass (∼ pt,jet) isenhanced in bkgd, but MET is not?

EXTRAS

Large K -factorsAnother example: b-jet production

pt [GeV]

Tevatron

|y| < 0.7, R = 0.7, µR = µF = Pt

MC@NLO

0 2 4 6 8

50 500 100

K factor

pt [GeV]

|y| < 0.7, R = 0.7, µR = µF = Pt

100 1000 0 2 4 6 8 10 12

pt [GeV]

50 500 100

1/2 Pt < µR = µF < 2 Pt

pt [GeV]

50 500 100

FCRFEX

scale dep.

pt [GeV] 100 1000

1/2 Pt < µR = µF < 2 Pt

channels

pt [GeV] 100 1000

1Herwig 6.5

Banfi, GPS & Zanderighi, ’07

Les recherches de nouvelles particules au LHC : quel role pour la … · 2009. 3. 24. · Les...

Documents

ANNONCE DU RESULTAT POSITIF, ROLE DE LENTOURAGE ANNONCE DU RESULTAT POSITIF, ROLE DE LENTOURAGE

Appel d’Offres N° IT-2524/ST/LHC

Une Brève Histoire des Particules - indico.lal.in2p3.fr · Une Brève Histoire des Particules : de l’Atomisme au LHC en passant par ACO . 2016 . Nicolas Arnaud ... → Tri en temps

CS ADC. SON ROLE CS ADC SON ROLE SON ORGANISATION

PARTICULES RADIOACTIVES

Particules de type virus (PTV) et particules sous-virales

The PS Complex as Proton Pre-Injector for the LHC – …cern.ch/blas/Publications/2000-03.pdf · · 2003-08-29The PS Complex as Proton Pre-Injector for the LHC ... 3.4.5 Characteristics

Particules et Interactions

APRÈS L’EXPOSITION « LE GRAND COLLISIONNEUR – LHC », LA

Nouvelle physique entre Cosmologie et le LHC : … · Nouvelle physique entre Cosmologie et le LHC : Axions, Neutrinos et Z’ M. Elmer To cite this version: M. Elmer. Nouvelle physique

LHC et physique des particules : derniers résultats et nouveaux … · 2011-07-26 · LHC et physique des particules : derniers résultats et nouveaux enjeux . Conférence de presse

The LHC Collimation Systemlhc-collimation.web.cern.ch/lhc-collimation/files/lhc-mac-sep02f.pdf · The LHC Collimation System R. Assmann, CERN-SL ... Mandate: Study beam dynamics and

Si le Higgs existe, il sera produit dans les collisions de protons du LHC et il pourra être détecté par ATLAS en identifiant ses désintégrations en particules

Distributed Monitoring Infrastructure for Worldwide LHC

Role de l_auditeur

Utilisation d’ EDMS en BE Mécanique au travers du projet LHC

Les particules élémentaires.pdf

Accurate predictions for Higgs production at the LHC: the VBF

Particules - Le Quartz

Les détecteurs du LHC (ATLAS et les autres )