J. Huston QCD Backgrounds to New...

TeVU 8/29/02J. Huston

See http://www.pa.msu.edu/~huston/tevu/tevu.pdf

QCD Backgrounds to New Physics

J. Huston

…thanks to Weiming Yao, John Campbell, Bruce KnutesonNigel Glover for transparencies


but first some commercials


Run 2 Monte Carlo Workshop

Transparencies, videolinks to individual talksand links to programscan all be found athttp://www-theory.fnal.gov/runiimc/

orhttp://thpc20.fnal.gov/runiimc/

I’ll be referring to someof these programs inthe course of my talk


Les Houches

Two workshops on “Physics at TeVColliders” have been held so far, in1999 and 2001 (May 21-June 1)

Working groups on QCD/SM, Higgs,Beyond Standard Model

See web page:

http://wwwlapp.in2p3.fr/conferences/LesHouches/Houches2001/

especially for links to writeups from 1999and 2001

QCD 1999 writeup (hep-ph/0005114)is an excellent pedagogical reviewfor new students

QCD 2001 writeup (hep-ph/0204316)is a good treatment of the state ofthe art for pdfs, NLO calculations,Monte Carlos

Les Houches 2003 will have more ofa concentration on EW/top physics


Les Houches 2001 Writeups

The QCD/SM Working Group: SummaryReport

◆ hep-ph/0204316

The Higgs Working Group: Summary Report(2001)

◆ hep-ph/0203056

The Beyond the Standard Model WorkingGroup: Summary Report

◆ hep-ph/0204031


Les Houches 2001

200th bottle of wineconsumed at theworkshop


Other useful references (for pdfs)

LHC Guide to Parton Distributions and Cross Sections, J. Huston;http://www.pa.msu.edu/~huston/lhc/lhc_pdfnote.ps

The QCD and Standard Model Working Group: Summary Report from LesHouches; hep-ph/0005114

Parton Distributions Working Group, Tevatron Run 2 Workshop; hep-ph/0006300

A QCD Analysis of HERA and fixed target structure functiondata, M. Botje; hep-ph/9912439

Global fit to the charged leptons DIS data…, S. Alekhin; hep-ph/0011002

Walter Giele’s presentation to the QCD group on Jan. 12; http://www-cdf.fnal.gov/internal/physics/qcd/qcd99_internal_meetings.html

Uncertainties of Predictions from Parton Distributions I: the Lagrange MultiplierMethod, D. Stump, J. Huston et al.;hep-ph/0101051

Uncertainties of Predictions from Parton Distributions II: the Hessian Method, J.Pumplin, J. Huston et al.; hep-ph/0101032

Error Estimates on Parton Density Distributions, M. Botje; hep-ph/0110123

New Generation of Parton Distributions with Uncertainties from GlobalQCD Analysis (CTEQ6); hep-ph/0201195


Other references…

…to the ability to calculate QCD backgrounds …to the need to do so


Monojets in UA1

UA1 monojets (1983-1984)

◆ Possible signature of new physics(SUSY, etc)

◆ A number of backgrounds wereidentified, but each was noted asbeing too small to account for theobserved signal

▲ pp->Z + jets

|_ νν

▲ pp->W + jets

|_ τ + ν

|_ hadrons + ν

▲ pp->W + jets

|_ l + ν

▲ pp->W + jets

|_ τ + ν

|_ l + ν

jet

…but the sum was not

◆ “The sum of many small things is abig thing.” G. Altarelli

Can calculate from first principles orcalibrate to observed cross sectionsfor Z->e+e- and W->eν

Ellis, Kleiss, Stirling PL 167B, 1986.


Signatures of New Physics

Ws, jets, γs, b quarks, ET

…pretty much the same as signatures for SM physics

How do we find new physics? By showing that its not old physics.

◆ can be modifications to the rate of production

◆ …or modification to the kinematics, e.g.angular distributions

Crucial to understand the QCD dynamics and normalization of bothbackgrounds to any new physics and to the new physics itself

Some backgrounds can be measured in situ

◆ …but may still want to predict in advance, e.g. QCD backgrounds toH->γγ

For some backgrounds, need to rely on theoretical calculations, e.g. ttbbbackgrounds to ttH


Theoretical Predictions for New (Old) Physics

There are a variety of programs available forcomparison of data to theory and/or predictions.

◆ Tree level

◆ Leading log Monte Carlo

◆ NnLO

◆ Resummed

Important to know strengths/weaknesses of each.

In general, agree quite well…but beforeyou appeal to new physics, check theME. (for example using Comphep)Can have ME corrections to MC or MC corrections to ME. (in CDF->HERPRT)

Perhaps biggest effort…include NLO MEcorrections in Monte Carlo programs…correct normalizations. Correct shapes. NnLO needed for precision physics.

Resummed description describes soft gluoneffects (better than MC’s)…has correctnormalization (but need HO to get it); resummedpredictions include non-perturbative effectscorrectly…may have to be put in by hand in MC’sthreshold kT

W,Z, Higgsdijet, direct γ

b space(ResBos)

qt space

Where possible, normalize to existing data.…in addition, worry about pdf, fragmentation uncertainties


W + Jet(s) at the Tevatron

Good testing ground for parton showers,matrix elements, NLO

Background for new physics

◆ or old physics (top production)

Reasonable agreement for the leadingorder comparisons using VECBOS (butlarge scale dependence)

-1.5

-1

-0.5

0

0.5

1

0 10 20 30 40 50 60 70 80 90 100

Systematic uncertainties

Leading Jet ET (GeV)

dσ

(W +

≥1

jets

) / d

ET

(D

ata −

QC

D)

/ QC

D

CDF PRELIMINARY

CDF Data (108 pb-1)0.4 Jet Cones, |ηd| < 2.4

DYRAD NLO QCD Predictionswith jet smearingMRSA/

Qr2 = Qf

2 = (0.5 MW)2

Qr2 = Qf

2 = MW2

Qr2 = Qf

2 = (2.0 MW)2

Good agreement with NLO (and smaller scale dependence) for W + >= 1 jet


W + jets

For W + >=n jet production, typicallyuse Herwig (Herprt) for additionalgluon radiation and for hadronization

Can also start off with n-1 jets andgenerate additional jets using Herwig

10-1

1

10

10 2

10 3

0 20 40 60 80 100 120 140 160 180 2

(a)

(b)

(c)

+ CDF Preliminary Data (108 pb-1)

− VECBOS + HERWIG Fragmentation+ Detector Simulation

CTEQ3M, Q2 = < pT >2

(Normalized to data)

Jet ET (GeV)

Eve

nts

/ 5 G

eV

(a) ET2 (W + ≥2 Jets Data)

VECBOS W + 1 Jet

(b) ET3 (W + ≥3 Jets Data)

VECBOS W + 2 Jets

(c) ET4 (W + ≥4 Jets Data)

VECBOS W + 3 Jets

10-1

1

10

10 2

10 3

0 20 40 60 80 100 120 140 160 180 200

(a)

(b)

(c)

(d)

+ CDF Preliminary Data (108 pb-1)

− VECBOS + HERWIG Fragmentation+ Detector Simulation

CTEQ3M, Q2 = < pT >2


Jet ET (GeV)

Eve

nts

/ 5 G

eV

(a) ET1 (W + ≥1 Jet)

(b) ET2 (W + ≥2 Jets)

(c) ET3 (W + ≥3 Jets)

(d) ET4 (W + ≥4 Jets)


More Comparisons (VECBOS and HERWIG)

Start with W + n jets from VECBOS

0

100

200

300

400

500

0 1 2 3 4 5

+ CDF PreliminaryData (108 pb-1)

− VECBOS + HERWIG+ Detector SimulationCTEQ3M, Q2 = < pT >(Normalized to data)

Eve

nts

/ 0.5

W + ≥2 Jets DataVECBOS W + 1 Jet

0

20

40

60

80

100

0 1 2 3 4 5∆Rjj of Leading Two Jets

Eve

nts

/ 0.5


0

100

200

300

400

500

0 1 2 3 4 5 6

+ CDF PreliminaryData (108 pb-1)

− VECBOS + HERWIG+ Detector SimulationCTEQ3M, Q2 = < pT >2


Eve

nts

/ 0.5

W + ≥2 Jets

0

20

40

60

80

100

0 1 2 3 4 5 6∆Rjj of Leading Two Jets

Eve

nts

/ 0.5

W + ≥3 Jets

•Start with W + (n-1) jets from VECBOS


More Comparisons

Start with W + n jets from VECBOS Start with W + (n-1) jets fromVECBOS

1

10

10 2

10 3

0 50 100 150 200 250 300 350 400

+ CDF Preliminary Data (108 pb-1)− VECBOS + HERWIG Fragmentation

+ Detector SimulationCTEQ3M, Q2 = < pT >2


Eve

nts

/ (10

GeV

/c2 )


10-1

1

10

10 2

0 50 100 150 200 250 300 350 400Mjj of Leading Two Jets (GeV/c2)

Eve

nts

/ (10

GeV

/c2 )

W + ≥3 Jets DataVECBOS W + 2 Jets

1

10

10 2

10 3

0 50 100 150 200 250 300 350 400

+ CDF Preliminary Data (108 pb-1)− VECBOS + HERWIG Fragmentation

+ Detector SimulationCTEQ3M, Q2 = < pT >2


Eve

nts

/ (10

GeV

/c2 )

W + ≥2 Jets

10-1

1

10

10 2

0 50 100 150 200 250 300 350 400Mjj of Leading Two Jets (GeV/c2)

Eve

nts

/ (10

GeV

/c2 )

W + ≥3 Jets


When good Monte Carlos go bad

Consider W + jet(s) sample

Compare data (Run 0 CDF) toVECBOS+HERPRT (Herwigradiation+hadronization interfaceto VECBOS) normalized to W+Xjets

Starting with W + 1 jet rate indata, Herwig predicts 1 W+ >4 jetevents; in data observe 10

…factor of ~2 every jet

◆ very dependent on kinematicsituation, though

▲ jet ET cuts

▲ center-of-mass energy

▲ etc

#events >1 jet >2 jets >3 jets >4 jet

pT>10 GeV/c

Data 920 213 42 10

VECBOS + HERPRT (Q=<pT>)

W + 1jet 920 178 21 1

W + 2jet ----- 213 43 6

W + 3jet ----- ----- 42 10

VECBOS+HERPRT(Q=mW)

W + 1jet 920 176 24 2

W + 2jet ---- 213 46 6

W + 3 jet ---- ----- 42 7


Factors get worse at the LHC

factor of 50


Why?

Some reasons given by the experts (Mangano, Yuan,Ilyin…)◆ Herwig (any Monte Carlo) only has collinear part of matrix element for gluon

emission; underestimate for the wide angle emission that leads to widelyseparated jets

◆ phase space: Herwig has ordering in virtualities for gluon emission while thisis not present in exact matrix element calculations; more phase space forgluon emission in exact matrix element calculations

◆ in case of exact matrix element, there are interferences among all of thedifferent diagrams; these interferences become large when emissions takeplace at large angles (don’t know a priori whether interference is positive ornegative)

◆ unitarity of Herwig evolution: multijet events in Herwig will always be afraction of the 2 jet rate, since multijet events all start from the 2-jet hardprocess

▲ all “K-factors” from higher order are missed.


Tree Level Calculations

Leading order matrixelement calculationsdescribe multi-bodyconfigurations betterthan parton showers

Many programs existfor calculation of multi-body final states attree-level

◆ References: see Run 2MC workshop

CompHep

◆ includes SM Lagrangian and several othermodels, including MSSM

◆ deals with matrix elements squared

◆ calculates leading order 2->4-6 in the finalstate taking into account all of QCD andEW diagrams

◆ color flow information; interface exits toPythia

◆ great user interface

Grace

◆ similar to CompHep

Madgraph

◆ SM + MSSM

◆ deals with helicity amplitudes

◆ “unlimited” external particles (12?)

◆ color flow information

◆ not much user interfacing yet

Alpha + O’Mega

◆ does not use Feynman diagrams

◆ gg->10 g (5,348,843,500 diagrams)


Monte Carlo Interfaces

To obtain full predictability for atheoretical calculation, would liketo interface to a Monte Carloprogram (Herwig, Pythia, Isajet)

◆ parton showering (additional jets)

◆ hadronization

◆ detector simulation

Some interfaces already exist

◆ VECBOS->Herwig (HERPRT)

◆ CompHep->Pythia

A general interface accord wasreached at the 2001 LesHouches workshop

All of the matrix elementprograms mentioned will output4-vector and color flowinformation in such a way as tobe universally readable by allMonte Carlo programs

CompHep, Grace, Madgraph,Alpha, etc, etc

->Herwig, Pythia, Isajet


Les Houches and Monte Carlos

Much of the time during meetingwas spent developing a genericprocess interface from matrixelement to Monte Carloprograms

This interface allows:◆ arbitrary hard subprocesses to

be plugged intoshower/hadronizationgenerators.

CompHEP

Grace Herwig

MadGraph z Isajet

VecBos Pythia

Wbbgen

◆ ->Les Houches accord (#1)

“Les Houches” User ProcessInterface

for Event Generators

hep-ph/0109068

E. Boos, M. Dobbs, W. Giele, I. Hinchliffe, J. Huston,

V. Ilyin, J. Kanzaki, K. Kato, Y. Kurihara,

L. Lonnblad, M. Mangano, S. Mrenna, F. Paige, E. Richter-Was,

M. Seymour, T. Sjostrand, B. Webber, D. Zeppenfeld

Possible because one or more authors from

each of these programs was present

at Les Houches

◆ Matt Dobbs has been the front man for

coordinating the disputes/discussions

◆ literally hundreds of email exchanges


Universal Interface

This interface will allow for amore complete predictability forME programs

◆ parton showering (additional jets)

◆ hadronization

◆ detector simulation

Some specialized interfacesalready exist

◆ VECBOSzHerwig (HERPRT)

◆ WbbgenzHerwig

◆ CompHepzPythia

This interface should supercedethem.

Specialize in the ‘generic’ parts of the event.

f(x,Q2) f(x,Q2)PartonDistributions

HardSubProcess

PartonCascade

Hadronization

Decay

+Minimum BiasCollisions


Interface

Provides information on parton 4-vectors,mother-daughterrelationships, spins/helicities andcolor flow

◆ also points to intermediate particleswhose mass should be preserved inthe parton showering

Not intended as a replacement forHEPEVT

◆ addresses communication betweenevent generators only, not betweenevent generators and the outsideworld

Partonic information is in 2 Fortrancommon blocks

◆ run info

◆ specific event info

Interface Structure

<<Container for RUN related information>>common /HepRUP/

+paramter MAXPUP: integer = 100+IDBMUP(2): integer+EBMUP(2): double+PDFGUP(2): integer+PDFSUP(2): integer+IDWTUP: integer+NPRUP: integer+XSECUP(MAXPUP): double+XERRUP(MAXPUP): double+XMAXUP(MAXPUP): double+LPRUP(MAXPUP): integer

<<Container for EVENT related information>>common /HepEUP/

+parameter MAXNUP: integer = 500, max num particle entries+NUP: integer = number entries this event+IDPRUP: integer = process id+XWGTUP: double = event weight+SCALUP: double = scale [GeV]+AQEDUP: double = QED coupling for this event+AQCDUP: double = QCD coupling for this event+IDUP(MAXNUP): integer = particle id+ISTUP(MAXNUP): integer = particle status+MOTHUP(2,MAXNUP): integer = pointer to parents+ICOLUP(2,MAXNUP): integer = particle (anit)color indices+PUP(5,MAXNUP): double = particle momentum, energy, mass+VTIMUP(MAXNUP): double = particle invariant lifetime+SPINUP(MAXNUP): double = spin vector angle (usually +1,-1)

<<called by SHG to for HepRUP info>>subroutine UPINIT()

<<called by SHG for HepEUP info>>subroutine UPEVNT()

integer MAXPUP parameter ( MAXPUP=100 ) integer IDBMUP, PDFGUP,PDFSUP, IDWUP, NPRUP, LPRUP double precision EBMUP,XSECUP, XERRUP, XMAXUP common /HEPRUP/ IDBMUP(2), EBMUP(2), PDFGUP(2),PDFSUP(2), + IDWTUP, NPRUP, XSECUP(MAXPUP), XERRUP(MAXNUP), + XMAXUP(MAXNUP), LPRUP(MAXPUP)

(Specialized for each matrix element)


Subroutines

Each stage (run and event)associated with own subroutine,called from the showergenerator, where information isplaced in the respective commonblock, based on output from thematrix element generator

Subroutine names (in Pythia 6.2)are:

◆ UPINIT

◆ UPEVNT

◆ note no PY prefixes

Other authors should use thesame convention

Interface Structure










Interface Structure


Unweighting

Shower generator can unweightevents from matrix elementgenerator, mix differentsubprocesses from matrix elementgenerator, or just read eventsstraight from a file

◆ if unweighting/mixing is needed thenshower generator needs info aboutsubprocess cross sections and/ormaximum weights

If extra information is needed forspecific user implementation, thenimplementation-specific commonblock has to be created

Note that a lot of the technicalitiesare intended for ME/MC authors, notfor users; in most cases, thesedetails will be invisible to the casualuser

MAXUP: maximum number of different processes to be interfaced at one time


Run related information

Each stage (run and eventassociated with own subroutine)

Run subroutine◆ IDWTUP: master switch indicating

how the event weights (XWGTUP)are interpreted (some examplesbelow)

▲ +1: events are weighted on inputand SHG is asked to produceevents with weight +1 on output

▲ -1: same as above but eventweights may be either positive ornegative; SHG will produceevents with weights +1 or -1 onoutput

▲ +3: events are unweighted oninput so SHG only asks for nextevent

▲ -3: same as above but eventweights may be either +1 or -1




(Specialized for


Event related information

NUP: number of particle entries for thisevent

IDPRUP: ID of the process for this event

XWGTUP: event weight

IDUP: particle ID (non-physical particlesassigned IDUP=0)

ISTUP: status code

◆ -1: incoming particle

◆ +1: outgoing particle

◆ -2: intermediate space-like propagatordefining an x and Q2 which should bepreserved (DIS-specific)

◆ +2: intermediate resonance, mass shouldbe preserved

▲ recoil from parton shower needs to beabsorbed by particles in the event

◆ +3: intermediate resonance, fordocumentation only

◆ -9: incoming beam particles




each matrix element)


Event info

MOTHUP(2,I): index of first and lastmother

◆ For decays, daughter particles willonly have 1 mother

◆ For 2->n, daughter particles will have2 mothers

Color flow: specific choice of colorflow for a particular event is oftenunphysical, due to interferenceeffects, but SHGs require specificcolor state from which to beginshower

◆ ICOLUP(1,I): integer tag for colorflow line passing through color of theparticle

◆ Integer tag fro color flow line passingthrough anti-color of tag




each matrix element)


Example (gg->gg)

1

2

3

4

501

503

502 504

I ISTUP(I) IDUP(I) MOTHUP(1,I) MOTHUP(2,I) ICOLUP(1,I) ICOLUP(2,I)1 –1 21 (g) 501 5022 –1 21 (g) 502 5033 +1 21 (g) 1 2 501 5044 +1 21 (g) 1 2 504 503

initial/final state

particle code

mother/daughter relationships

color flow


Consider ttbar production

t and tbar given ISTUP=+2, whichinforms SHG to preserve theirinvariant masses when showeringand hadronizing the event

Intermediate s-channel gluon hasbeen drawn, but no entry becausecannot be distinguished from t-channel

Definition of color or anti-color linedepends on orientation of graph

◆ define color and anti-color accordingto physical time order

◆ quark will always have color tagICOLUP(1,I) filled, but never its anti-color tag ICOLUP(2,I); reverse foranti-quark; gluon has info in bothtags

Example: hadronic tt production

1

24

t 7b

8W

3t

5b

6W

501

502502

503

503

I ISTUP(I) IDUP(I) MOTHUP(1,I) MOTHUP(2,I) ICOLUP(1,I) ICOLUP(2,I)1 –1 21 (g) 0 0 501 5022 –1 21 (g) 0 0 503 5013 +2 –6 (t) 1 2 0 5024 +2 6 (t) 1 2 503 05 +1 –5 (b) 3 3 0 5026 +1 –24 (W−) 3 3 0 07 +1 5 (b) 4 4 503 08 +1 24 (W+) 4 4 0 0

The t and t are given ISTUP=+2, which informs the SHG to preserve their invariant masseswhen showering and hadronizing the event. An intermediate s-channel gluon has been drawnin the diagram, but since this graph cannot be usefully distinguished from the one with at-channel top exchange, an entry has not been included for it in the event record.

The definition of a line as ‘color’ or ‘anti-color’ depends on the orientation of the graph.This ambiguity is resolved by defining color and anti-color according to the physical timeorder. A quark will always have its color tag ICOLUP(1,I) filled, but never its anti-color tagICOLUP(2,I). The reverse is true for an anti-quark, and a gluon will always have informationin both ICOLUP(1,I) and ICOLUP(2,I) tags.

Note the difference in the treatment by the parton shower of the above example, and anidentical final state, where the intermediate particles are not specified:

9


q q

e e

1

3

2

54

501 501

Another example:little pink elephant exchange

I ISTUP(I) IDUP(I) MOTHUP(1,I) MOTHUP(2,I) ICOLUP(1,I) ICOLU1 –1 –2 (u) 0 0 0 502 –1 2 (u) 0 0 501 03 +2 0 (pink elephant) 1 2 0 04 +1 11 (e−) 3 3 0 05 +1 –11 (e+) 3 3 0 0


What Les Houches doesn’t do

Specify the exact formof output format

◆ nominally details aresupposed to be invisibleto casual user

Specify correct Q scalefor parton showering

◆ imagine that W + 5 jetevents probe moredeeply into shower thanW + 1 jet events

▲ would need lower scale

Interface Structure










The proper way

The proper way of taking care of this would be togenerate parton showers starting at full hard scalebut vetoing those emissions that populate samephase space as exact ME

◆ See for example, Krauss, Catani, Kuhn andWebber, hep-ph/00109231

◆ Sjostrand argued for this type of specification atLes Houches but was out-shouted

◆ Herwig doesn’t need this information since it usesthe color flow to start the showering

Need to look for the next best solution


Comparisons of ME and MC calculations

Tree level ME predictions should agreewith each other (modulo αs choices, etc)

Parton showering Monte Carlos takeslightly different roads, though and maynot necessarily agree

◆ e.g. gg->H

Through implementation of colorcoherence, Herwig reproduces the LLterms and part of NLL terms

◆ see C. Balazs, J. Huston, I. Puljak,PRD; Ditto + S. Mrenna, LesHouches 2001 + paper in preparation

Pythia approximates color coherence andhas approximate agreement with logstructure

More global comparisons in progress

0 10 20 30 40 50 60 70 80 90 1000

0.2

0.4

0.6

0.8

1

pT (GeV)0 10 20 30 40 50 60 70 80 90 100

dp

T (

pb

/Gev

)⁄

dσ

0

0.2

0.4

0.6

0.8

1

0 20 40 60 80 100 120 140 160 180 20010

-3

10-2

10-1

1

pT (GeV)0 10 20 30 40 50 60 70 80 90 100

dp

T (

pb

/Gev

)⁄

dσ

0

0.2

0.4

0.6

0.8

1 gg → H + X at LHCmH = 125 GeV, CTEQ5M, √s = 14 TeVcurves: ResBos NLL: B(1,2), A(1,2), C(0,1) LL: B(1), A(1), C(0)histograms: PYTHIA 5.7 PYTHIA 6.136 PYTHIA 6.203 HERWIG 6.3

pT (GeV)0 20 40 60 80 100 120 140 160 180 200

dp

T (

pb

/Gev

)⁄

dσ

10-3

10-2

10-1

1


Logs that we know and love

A1, B1 and (a bit of) A2 areeffectively in Monte Carlos(especially Herwig)

A1,A2 and B1 for Higgsproduction are in currentoff-the-shelf version ofResBos

◆ …as are C0 and C1 whichcontrol the NLO normalization

The B2 term has recentlybeen calculated for gg->H

C. Balazs, Higgs production with soft-gluon effects, les Houches ’99 14


The need for higher order…

0

0.2

0.4

0.6

0.8

1

0 0.5 1 1.5 2

LONLO

NNLO

arX

iv:h

ep-p

h/01

0606

96

Jun

2001


What would we like?

Bruce Knuteson’s wishlist from the Run 2 Monte Carlo workshop

…all at NLO


What are we likely to get?


MCFM (Monte Carlo for FemtobarnProcesses) J. Campbell and K.Ellis

Goal is to provide a unifieddescription of processes involvingheavy quarks, leptons and missingenergy at NLO accuracy

There have so far been three main applicationsof this Monte Carlo, each associated with adifferent paper.

◆ Calculation of the Wbb background to a WHsignal at the Tevatron.

R.K.Ellis, Sinisa Veseli, Phys. Rev. D60:011501(1999), hep-ph/9810489.

◆ Vector boson pair production at the Tevatron,including all spin correlations of the boson decayproducts.

J.M.Campbell, R.K.Ellis, Phys. Rev. D60:113006(1999), hep-ph/9905386.

◆ Calculation of the Zbb and other backgrounds to aZH signal at the Tevatron.

J.M.Campbell, R.K.Ellis, FERMILAB-PUB-00-145-T, June 2000, hep-ph/0006304.

The last of these references contains the mostdetails of our method.


Higgs backgrounds using MCFM


Wbbar and Zbbar

0

0.5

1

1.5

2

2.5

0 20 40 60 80 100 120 140 160

dσ /

dMbb

[fb/

GeV

]

Mbb [GeV] (with Gaussian smearing of 10 GeV)

pp− → νe e+ bb− + X

√s=2 TeV MRS98

NLO

LO


Recent example of data vs Monte Carlo

There is a discovery potential atthe Tevatron during Run 2 for arelatively light Higgs (especially ifHiggs mass is 115 GeV)

…but small signal to backgroundratio makes understanding ofbackgrounds very important

CDF and ATLAS recently wentthrough similar exercisesregarding this background

◆ CDF using Run 1 data

◆ ATLAS using Monte Carlopredictions

ν

b

b-

ν


Data vs Monte Carlo


Sleuth strategy

Consider recent major discoveries inhep

◆ W,Z bosonsCERN 1983

◆ top quark Fermilab 1995

◆ tau neutrino Fermilab 2000

◆ Higgs Boson? CERN 2000

In all cases, predictions weredefinite, aside from mass

Plethora of models that appear dailyon hep-ph

Is it possible to perform a “generic”search?

Transparencies from Bruce Knuteson talk at Moriond 2001


W2j

We consider exclusive final statesWe consider exclusive final states

We assume the existence of standard object definitions

These define e, _, ττττ, γγγγ, j, b, ET, W, and Z fi

All events that contain the same numbers of each of theseobjects belong to the same final state

Step 1: Exclusive final statesSleuth Bruce Knuteson

e_ET

Z4j

eET jj eE

T 3jW3j eeγγγγeγγγγγγγγ

ZγγγγWγγγγγγγγ

__jj e_ET j

γγγγγγγγγγγγ ___eee


Results

Results agree well with expectationNo evidence of new physics is observed

DØ data

Search for regions of excess (more data Search for regions of excess (more data events than expected from background)events than expected from background) within that variable space within that variable space

probability to be SM


Fragmentation Uncertainties: Higgs->γγ and

Backgrounds

One of the most useful searchmodes for the discovery of theHiggs in the 100-150 GeV massrange at the LHC is in the twophoton mode

Higgs->γγ has very large

backgrounds from QCD sources

◆ Diphoton production

◆ γπo and πoπo production; jetsfragmenting into very high z πo’s

With excellent diphoton massresolution, can try to resolveHiggs “bump”

Still important to understand levelof background

10000

12500

15000

17500

20000

105 120 135

mγγ (GeV)

Eve

nts

/ 2 G

eV

0

500

1000

1500

105 120 135

mγγ (GeV)

Sign

al-b

ackg

roun

d, e

vent

s / 2

GeV


Diphoton Backgrounds in ATLAS

Again, for a H->γγ search at the

LHC, face irreducible backgrounds

from QCD γγ and reducible

backgrounds from γπo and πoπo

◆ in range from 70 to 170 GeV

jet-jet cross section is estimated

to be a factor of 2E6 times the γγ

cross section and γ-jet a factor

of 8E2 larger

◆ Need rejection factors of 2E7 and

8E3 respectively

◆ PYTHIA results seem to indicate

that reducible backgrounds are

comfortably less than reducible

ones

◆ …but how to normalize PYTHIApredictions for very high z fragmentation ofjets; fragmentation not known well at highz and certainly not for gluon jets

ATLAS detector and physics performance Volume IITechnical Design Report 25 May 1999

from higher-order corrections and from uncertainties in jet fragmentation. In order to reducethese backgrounds to a level well below that of the irreducible γγ continuum, rejection factors of2x107 and 8x103 are required.

These rejection factors have been realisticallyevaluated by using large samples of fully sim-ulated two-jet events, as described inSection 7.6. The rejection factors were then ap-plied to estimate the cross-sections for the re-ducible jet-jet and γ-jet backgrounds relative tothe irreducible γγ-background. The results areshown in Figure 19-2 as a function of the two-photon invariant mass mγγ. After applying thefull photon identification cuts from the calo-rimeter and the Inner Detector, the residualjet-jet and γ-jet backgrounds are found to be atthe level of approximately 15% and 20%, re-spectively, of the irreducible γγ backgroundover the mass range relevant to theH → γγ search.

Z → ee background

h

Figure 19-2 Expected ratios of the residual reduciblejet-jet and γ-jet backgrounds to the irreducible γγ-con-tinuum background as a function of the invariant massof the pair of photon candidates at high luminosity.

0

0.1

0.2

0.3

0.4

80 100 120 140 160

γ-jet/γγ

jet-jet/γγ

mγγ (GeV)

Rat

io


Different models predict different high z fragmentation

Backgrounds to γγ production inHiggs mass region arise fromfragmentation of jets to high z πo’s

◆ Pythia and Herwig predict verydifferent rates for high z

◆ all fragmentation is not equal

Example of a background that canbe measured in situ, but nice to beable to predict the environmentbeforehand

DIPHOX (see Run 2 MC workshop)program can calculate γγ, γπo, andπoπo cross sections to NLO

◆ comparisons underway to Tevatrondata

gg->γγ and qqbar->γγ at NNLO maybe available soon

10-3

10-2

10-1

1

10

10 2

(1/N

Jet)

dN/d

x E(c

h)

Y topology

Quark jetsGluon jets

DELPHI

00.20.40.60.8

11.21.4

0 0.2 0.4 0.6 0.8

Gluon / Quark

xE(ch)

Glu

on/Q

uark

10-3

10-2

10-1

1

10

10 2

(1/N

Jet)

dN/d

x E(c

h)

Jetset 7.4

00.20.40.60.8

11.21.4

0 0.2 0.4 0.6 0.8

Gluon / Quark

xE(ch)

Glu

on/Q

uark

10-3

10-2

10-1

1

10

10 2

(1/N

Jet)

dN/d

x E(c

h)

Ariadne 4.08

00.20.40.60.8

11.21.4

0 0.2 0.4 0.6 0.8

Gluon / Quark

xE(ch)

Glu

on/Q

uark

10-3

10-2

10-1

1

10

10 2

(1/N

Jet)

dN/d

x E(c

h)

Herwig 5.8C

00.20.40.60.8

11.21.4

0 0.2 0.4 0.6 0.8

Gluon / Quark

xE(ch)

Glu

on/Q

uark

B. Webber, hep-ph/9912399


PDF Uncertainties

What’s unknown aboutPDF’s

◆ the gluon distribution

◆ strange and anti-strangequarks

◆ details in the {u,d} quarksector; up/downdifferences and ratios

◆ heavy quarkdistributions

Σ of quark distributions (q + qbar) is well-

determined over wide range of x and Q2

◆ Quark distributions primarily determinedfrom DIS and DY data sets which havelarge statistics and systematic errors infew percent range (±3% for 10-4<x<0.75)

◆ Individual quark flavors, though may haveuncertainties larger than that on the sum;important, for example, for W asymmetry

information on dbar and ubar comes atsmall x from HERA and at medium x fromfixed target DY production on H2 and D2

targets

◆ Note dbar≠ubar

strange quark sea determined fromdimuon production in ν DIS (CCFR)

d/u at large x comes from FT DYproduction on H2 and D2 and leptonasymmetry in W production


Example:Jets at the Tevatron

Both experiments compare to NLO QCDcalculations

◆ D0: JETRAD, modified Snowmassclustering(Rsep=1.3, µF=µR=ETmax/2

◆ CDF: EKS, Snowmass clustering(Rsep=1.3 (2.0 in some previouscomparisons), µF=µR=ETjet/2

•In Run 1a, CDF observed an excess in thejet cross section at high ET, outside the range of the theoretical uncertainties shown


Similar excess observed in Run 1B


Exotic explanations

Transverse Energy

Cross section for various compositeness scales ΛC

1

1.5

2

2.5

3

3.5

4

4.5

5

0 50 100 150 200 250 300 350 400 45

Rat

io

Λ1300/Λ∞

Λ1400/Λ∞

Λ1500/Λ∞

Λ1600/Λ∞

Λ1700/Λ∞

Data 92/Λ∞

CDF Preliminary

Only statistical errors are plotted


Non-exotic explanations

Modify the gluon distribution at high x


Tevatron Jets and the high x gluon

Best fit to CDF and D0 central jet cross sections provided byCTEQ5HJ pdf’s

50 100 150 200 250 300 350 400pT

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Incl. Jet : pt7 * dσ/dpt

(Error bars: statistical only)

14% < Corr. Sys. Err. < 27%

Ratio: Prel. data / NLO QCD (CTEQ5M | CTEQ5HJ)

norm. facor :CTEQ5HJ: 1.04CTEQ5M : 1.00

CDF

Data / CTEQ5MCTEQ5HJ / CTEQ5MCDF Data ( Prel. )CTEQ5HJCTEQ5M

100 200 300 400 pT

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Incl. Jet : pt7 * dσ/dpt

Error bars: statistical only

8% < Corr. Sys. Err. < 30%

Ratio: Data / NLO QCD (CTEQ5M | CTEQ5HJ)

χ2 = norm. factor :CTEQ5HJ: 25/24 1.08CTEQ5M : 24/24 1.04

D0

Data / CTEQ5MCTEQ5HJ / CTEQ5MD0 dataCTEQ5HJCTEQ5M


D0 jet cross section as function of rapidity

10-1

1

10

10 2

10 3

10 4

10 5

10 6

10 7

10 8

0 100 200 300 400 500

ET (GeV)

⟨⟨⟨⟨ dσσσσ /

/// dE

T dηηηη⟩⟩⟩⟩

(fb/

GeV

) 0.0 ≤≤≤≤ ||||ηηηη|||| <<<< 0.5 0.5 ≤≤≤≤ ||||ηηηη|||| <<<< 1.0 1.0 ≤≤≤≤ ||||ηηηη|||| <<<< 1.5 1.5 ≤≤≤≤ ||||ηηηη|||| <<<< 2.0 2.0 ≤≤≤≤ ||||ηηηη|||| <<<< 3.0

DØ PreliminaryDØ Preliminary√√√√↵↵↵↵��

�� ====

−−−−

��1

pb 92 Ldt √√√√↵↵↵↵��

�� ====

−−−−

��1

pb 92 LdtRun 1BRun 1B

Nominal cross sections & statistical errors only

-1

0

1

50 100 150 200 250 300 350 400

-1

0

1

50 100 150 200 250 300 350 400

-1

0

1

50 100 150 200 250 300 350 400 450 500

-1

0

1

50 100 150 200 250 300 350 400 450 500

-1

0

1

50 100 150 200 250 300 350 400 450 500

JETRADµ=ET

max/2

CTEQ4HJ provides bestdescription of data

How reliable is NLO theory inthis region?K-factors?


Chisquares for recent pdf’s

PDF χχχχ2222 χχχχ2222//// dof Prob

CTEQ3M 121.56 1.35 0.01

CTEQ4M 92.46 1.03 0.41

CTEQ4HJ 59.38 0.66 0.99

MRST 113.78 1.26 0.05

MRSTgD 155.52 1.73 <0.01

MRSTgU 85.09 0.95 0.631

10

10 2

10 3

10 4

10 5

10 6

10 7

50 100 150 200 250 300 350 400 450 500

0.0 ≤ |η| < 0.50.5 ≤ |η| < 1.01.0 ≤ |η| < 1.51.5 ≤ |η| < 2.02.0 ≤ |η| < 3.0

•For 90 data points, are the chisquaresfor CTEQ4M and MRSTgU “good”?•Compared to CTEQ4HJ?


D0 jet cross section

CTEQ4 and CTEQ5 had CDFand D0 central jet cross sectionsin fit

Statistical power not greatenough to strongly influence highx gluon

◆ CTEQ4HJ/5HJ required aspecial emphasis to be given tohigh ET data points

Central fit for CTEQ6 is naturallyHJ-like

χ2 for CDF+D0 jet data is 113 for

123 data points

dataΗ theory theory versus pT GeVerror bars Η stat.Η syst.

100 200 300 400 500

Η 0.5

0

0.5

1

1.52ΗΗΗ3

100 200 300 400 500

Η 0.5

0

0.5

1

1.51.5ΗΗΗ2

100 200 300 400 500

Η 0.5

0

0.5

1

1.51ΗΗΗ1.5

100 200 300 400 500

Η 0.5

0

0.5

1

1.50.5ΗΗΗ1

100 200 300 400 500

Η 0.5

0

0.5

1

1.50ΗΗΗ0.5

Figure 19: Comparison between theory and the D0 jet data. The error bars are statistical

and systematic errors combined.

33


PDF Uncertainties: included in CTEQ6M sets inLHAPDF

Use Hessian technique (T=10)


Gluon Uncertainty

Gluon is fairly well-constrained up to an x-value of 0.3

New gluon is stiffer thanCTEQ5M; not quite as stiffas CTEQ5HJ


Luminosity function uncertainties at the Tevatron

102

√s (GeV)

Luminosity function at TeV RunII

Frac

tiona

l Unc

erta

inty

0.1

0.1

0.1

-0.1

-0.1

-0.1

0.1

Q-G --> Z

0

0

0

^

≈

≈ ≈

≈

200 405020

-0.1

0

≈

Q-G --> W+

Q-G --> W-

Q-G --> γ*

≈

102

√s (GeV)

Luminosity function at TeV RunII

Frac

tiona

l Unc

erta

inty

0.1

0.1

0.1

-0.1

-0.1

-0.2

0.2

Q-Q --> W+ (W-)

Q-Q --> γ* (Z)

G-G

0

0

0

^

≈

≈ ≈

≈−

−

0.3

0.4

200 4005020

±


Luminosity Function Uncertainties at the LHC

102 10

√s (GeV)

Luminosity function at LHC

Frac

tiona

l Unc

erta

inty

0.1

0.1

0.1

-0.1

-0.1

-0.1

0.1

0

0

0

0

-0.1

Q-G --> W+

Q-G --> W-

Q-G --> Z

Q-G --> γ*

^50 500200

≈ ≈

≈≈

≈≈

102 103

√s (GeV)

Luminosity function at LHC

Frac

tiona

l Unc

erta

inty

0.1

0.1

0.1

-0.1

-0.1

-0.2

0.2

Q-Q --> W+

Q-Q --> W-

G-G

0

0

0

≈

≈ ≈

≈−

−

±

50 200 500

^


Effective use of pdf uncertainties

PDF uncertainties are important both for precision measurements (W/Z crosssections) as well as for studies of potential new physics (a la jet cross sections athigh ET)

Most Monte Carlo/matrix element programs have “central” pdf’s built in, or caneasily interface to PDFLIB

Determining the pdf uncertainty for a particular cross section/distribution mightrequire the use of many pdf’s

◆ CTEQ Hessian pdf errors require using 33 pdf’s

◆ GKK on the order of 100

Too clumsy to attempt to includes grids for calculation of all of these pdf’s withthe MC programs

->Les Houches accord #2◆ Each pdf can be specified by a few lines of information, if MC programs can perform

the evolution

◆ Fast evolution routine will be included in new releases to construct grids for each pdf

NB: pdf uncertainties make most sense in the context of NLOcalculations; current MC programs are basically leading order and LOpdfs should be used when available

◆ NNB: CTEQ6L is a leading order fit to the data but using the 2-loop αs, since somehigher order corrections are in MC programs like Pythia, Herwig, etc


Les Houches accord #2

Using the interface is aseasy as using PDFLIB (andmuch easier to update)

First version has CTEQ6M,CTEQ6L, all of CTEQ6error pdfs and MRST2001pdfs

See pdf.fnal.gov (and talkby Walter Giele at thisconference)

call InitiPDFset(name)

◆ called once at the beginningof the code; name is the filename of external PDF file thatdefines PDF set

call InitPDF(mem)

◆ mem specifies individualmember of pdf set

call evolvePDF(x,Q,f)

◆ returns pdf momentumdensities for flavor f atmomentum fraction x andscale Q


The Big Idea

Reminder: the big idea:◆ The Les Houches accords will be

implemented in all ME/MC programsthat experimentalists/theorists use

◆ They will make it easy to generatethe multi-parton final states crucial tomuch of the Run 2/HERA/LHCphysics program and to compare theresults from different programs

◆ experimentalists/theorists can allshare common MC data sets

◆ They will make it possible togenerate the pdf uncertainties for anycross sections


Les Houches accords

Les Houches accord #1 (ME->MC)

◆ accord implemented in Pythia 6.2

◆ accord implemented in CompHEP

▲ CDF top dilepton group has beengenerating ttbar events withCompHEP/Madgraph + Pythia

◆ accord implemented in ALPGEN

▲ hep-ph/0206293

◆ accord implemented in Madgraph

▲ MADCUP:http://pheno.physics.wisc.edu/Software/MadCUP/}.

▲ MADGRAPH 2: within a few weeks

◆ work proceeding on Herwig; in release 6.5Sept 2002 (private version made availablefor debugging)

◆ Implemented in Grace

◆ in AcerMC:hep-ph/0201302

Les Houches accord #2 (pdfs inME/MC)

◆ version of pdf interface has beendeveloped

▲ available athttp://pdf.fnal.gov

◆ commitment for beingimplemented in MCFM

◆ commitment for beingimplemented in your name here


LO vs NLO

In some cases, differencebetween LO and NLO pdfs maybe larger than NLO uncertaintieson pdfs, e.g. the gluondistribution at high x

Ambiguity in LO pdfs◆ use 1-loop αs or 2-loop αs

▲ CTEQ6L uses 2-loop

▲ now have a 1-loop 6L inLHAPDF

▲ difference between the two isprobably a better estimate ofuncertainty than choosing a pdffrom another set

Will think about the idea of LOpdf uncertainties


Using pdf uncertainties

Currently need to run 40 pdfs tocalculate pdf uncertainties

Could be time-consuming (andunnecessary)

Think of 3 steps of decreasinglaziness

◆ look at plots of pdf luminosityuncertainty for given massand subprocess (and rapidity)

◆ use pdfs that are extremes fora particular subprocess

▲ particular eigenvectors probethat direction

◆ generate full cross sectionusing central pdf and storethe ratio of pdf luminositiesfor the 40 error pdf’s

theoretical model with many parameters, as well as its own sources of inherent uncertaintie

In [9, 10, 11], we formulated two methods, the Hessian and the Lagrange, which enab

the systematic characterization of the neighborhood of the global minimum of the χ2 function

These allow a systematic way of assessing the compatibility of the data sets in the framewor

of the theoretical (PQCD) model [36], as well as of estimating the uncertainties of the PDF

and their physical predictions if the experimental inputs are assumed to be compatibl

within certain practical tolerance. The basic ideas are summarized in the accompanyin

illustration (adapted from [10]):

(a)Original parameter basis

(b)Orthonormal eigenvector basis

zk

Tdiagonalization and

rescaling bythe iterative method

ul

ai

2-dim (i,j) rendition of d-dim (~20) PDF parameter space

Hessian eigenvector basis sets

ajul

p(i)

s0s0

contours of constant 2global

ul: eigenvector in the l-directionp(i): point of largest ai with tolerance T

s0: global minimum p(i)zl

The behavior of the global χ2 function in the neighborhood of the global minimum in th

PDF parameter space is encapsulated in 2Np sets of eigenvector PDFs (with Np ∼ 20 bein

the number of initial PDF parameters), obtained by an iterative procedure to diagonaliz

the Hessian matrix and to adjust the step sizes of the numerical calculation to match th

natural physical scales. This procedure efficiently overcomes a number of critical obstacles

encountered when applying standard tools to perform error propagation in the global χ

minimization approach. Details are given in [9, 10]

The analysis done in this work make full use of this method. The resulting 2Np +

PDF sets, consisting of the best fit and the eigenvector basis sets, allow us to calculate th

best estimate, and the range of uncertainty, of the PDF themselves as well as any physic

3For our analysis, there are ∼ 2000 data points from ∼ 16 different experiments of very different systematics and a wide range of precision.

4These are due to difficulties in calculating physically meaningful error matrices, by finite differences,the face of: (i) vastly different scales (∼ 106−7) in different, a priori unknown, directions in the high dimensioparameter space; and (ii) numerical fluctuations due to integration errors in the theoretical (PQCD) modcalculation as ell as round offs


Workshop in Cambridge (and at Tevatron)

Right before Amsterdam, there was a workshop held in Cambridge on TeV-scale Physics

◆ http://www.hep.phy.cam.ac.uk/theory/webber/camws.html

Original idea of workshop was to examine the implications of the several hundred pb-1 ofdata available from CDF and D0

Instead emphasis was more on ME/MC tools, especially NLO MC’s

◆ MC@NLO available now for diboson production, and soon will follow with moresubprocesses

◆ also talks by John Collins, Steve Mrenna, Dave Soper on their ideas for NLO MCs

◆ Most of the talks I’ve linked to the website:http://www.pa.msu.edu/~huston/cambridge_tevscale/cambridge_program.html

May be followup workshop in Durham this winter (January)

Also, there will be a 1-day mini-workshop with CDF/D0/theory at Fermi on issues of:

◆ Common MC tunings

◆ ME/MC comparisons and what else is needed

◆ Practical pdf uncertainties

Friday Oct 4 in 1-West


Conclusions

Great opportunity at Run 2 at theTevatron for discovery of new physics;even better opportunity when the LHCturns on

In order to be believeable, we mustunderstand the QCD backgrounds to anynew physics

◆ Don’t rely totally on Monte Carlosand certainly not on one Monte Carloalone

◆ In the words of Ronald Reagan,“Trust but Verify”, if possible,theoretical predictions/formalismswith data

▲ existing Run 1 data/Run 2 data

▲ background” data to be taken at theLHC

If no data, then verify with more completetheoretical treatments

Many new tools/links between old toolsare now being developed to make this jobeasier for experimenters

Hopefully, we’ll find many more of thetype of the event on the right to try them

t

Date post:	13-Jun-2020
Category:	Documents
Upload:	others
View:	4 times
Download:	0 times

J. Huston QCD Backgrounds to New...

Documents