co systematic erro · erro rs, D i = T + σ i r i with P (r) = e − r 2 / 2 √ 2 π. With...

15
α s F A A H a f A a A F a ^ H Experimental Input Parton Distributions: Nonperturbative parametrization at Q 0 DGLAP Evolution to Q Hard Scattering: (Perturbatively Calculable) = F λ A x, m Q , M Q a f a A x, m µ F λ a x, Q µ , M Q O Q χ

Transcript of co systematic erro · erro rs, D i = T + σ i r i with P (r) = e − r 2 / 2 √ 2 π. With...

  • Parton

    Distributionsand

    their

    Uncertainties

    JonPumplin

    DPF2002{W

    illiamfburg5/25/02

    CTEQ6PDFanalysis(J.Pumplin,D.Stump,W.K.Tung,

    J.Huston,H.Lai,P.Nadolsky[hep-ph/0201195])

    includenew

    datasets

    includecorrelatedsystematicexperimentalerrors

    evaluateuncertaintiesoftheresult:

    {

    EigenvectorPDFsetstomapuncertainties

    {

    Lagrangemultiplierresults

    UniversalPDFinterface:LesHouchesAccord

    Results:

    {

    W

    andZproduction

    {

    parton-partonluminosities

    {

    gluonandquarkdistributions

    Measuresofuncertainty

    {

    Measurementof

    αs

    {

    Statisticalbootstraps

    Outlook

    1

    Overview

    ofQCD

    GlobalAnalysis

    FA

    A

    H

    a

    fA aA

    Fa ^

    H

    Experim

    entalInput

    Parton D

    istributions:N

    onperturbative parametrization at Q

    0D

    GLA

    P E

    volution to Q

    Hard S

    cattering:(P

    erturbativelyC

    alculable)=

    FλA

    (x,mQ,MQ

    )=

    ∑afaA

    (x,mµ

    )⊗

    ^Fλa

    (x,Qµ,MQ

    )+

    O

    ((�Q

    )2)

    Sourcesofuncertainty:

    1.Experimentalerrorsincludedin

    χ

    2

    2.Unknownexperimentalerrors

    3.Parametrizationdependence

    4.Higher-ordercorrections&

    LargeLogarithms

    5.PowerLaw

    corrections(\highertwist")

    FundamentaldiÆculties:

    1.Goodexperimentsrununtilsystematicerrors

    dominate;andthemagnitudeofsystematic

    errorsinvolvesguesswork.

    2.Systematicerrorsofthetheoryandtheir

    correlationscannotevenbeguessed.

    2

  • *, W, Zq

    gq

    qq

    *, W, Z

    g qqq

    *, W, Z

    (dir)

    DIS

    DY

    Dir.P

    h.

    Jet Inc.

    e N N N N

    p N Nk Np N

    p N Nk Np Np p

    SLA

    CB

    CD

    MS

    NM

    C, E

    665H

    1, ZE

    US

    CD

    HS, C

    HA

    RM

    CC

    FR

    CH

    OR

    US

    E605, E

    772N

    A51

    E866

    CD

    F, D0

    WA

    70, UA

    6E

    706

    CD

    F, D0

    CD

    F, D0

    Experim

    ental Input

    3

    Kinematicregioncoveredbydata

    A

    widevarietyofdataaretiedtogetherbythe

    DGLAP

    renormalizationgroupevolutionequation.

    Consistency{orlackthereof{betweenthe

    experimentscanbeobservedonlybyapplyingQCD

    totiethem

    togetherinaglobal�t.

    Allexperimentsthatusehadronsintheinitialstate

    {Tevatron,LHC,andnon-acceleratorexperiments{

    requirethepartondistributionsfortheiranalysis.4

  • SelectionofData

    CTEQ5

    CTEQ6

    #

    sys

    #

    sys

    BCDMS

    µp

    168

    no

    BCDMS

    µp

    339

    yes

    BCDMS

    µd

    156

    no

    BCDMS

    µd

    251

    yes

    H1

    ep172

    no

    H1a

    ep•

    104

    yes

    H1b

    ep•

    126

    yes

    ZEUS

    ep

    186

    no

    ZEUS

    ep•

    229

    yes

    NMC

    µp

    104

    no

    NMC

    µp

    201

    yes

    NMC

    µp/µn

    123

    no

    NMC

    µp/µn

    123

    yes

    CCFR

    F

    2

    νN

    87

    no

    CCFR

    F

    2

    νN

    159

    yes

    CCFR

    F

    3

    νN

    87

    no

    CCFR

    F

    3

    νN

    87

    no

    E605

    pp

    DY

    119

    no

    E605

    pp

    119

    no

    NA51

    pd/pp

    DY

    1

    no

    NA51

    pd/pp

    1

    no

    E866

    pd/pp

    DY

    15

    no

    E866

    pd/pp

    15

    no

    CDF

    W

    11

    no

    CDF

    W

    11

    no

    CDF

    jet

    33

    yes

    CDF

    jet

    33

    yes

    D�jet

    24

    yes

    D�Jet

    •90

    yes

    New

    Data

    (Directphotondataarenotusedbecauseof

    uncontrolledsystematic\

    kT

    "e�ects,whichneed

    resummation)

    5

    CTEQ6Globalanalysis

    Inputfrom

    Experiment:

    •∼

    2000datapointswith

    Q>

    2GeV

    from

    e

    ,

    µ

    ,

    ν

    DIS;leptonpairproduction(DY);lepton

    asymmetryin

    W

    production;high

    pT

    inclusive

    jets;

    αs (M

    Z

    )from

    LEP

    Inputfrom

    Theory:

    NLO

    QCD

    evolutionandhardscattering

    Parametrizeat

    Q

    0:

    A

    0

    xA

    1

    (1 −

    x

    )

    A

    2

    (1+

    A

    3

    xA

    4)

    •s

    =

    �s

    =

    0

    .4(�u

    +

    �d)/

    2at

    Q

    0;nointrinsic

    b

    or

    c

    Constructe�ective

    χ

    2global= ∑

    expts

    χ

    2n

    :

    •χ

    2globalincludestheknownsystematicerrors

    Minimizing

    χ

    2globalyields\BestFit"PDFs.

    Variationof

    χ

    2globalinneighborhoodofthe

    minimum

    de�nesuncertaintylimits.

    Estimateuncertaintyasregionofparameter

    spacewhere

    χ

    2

    2(BestFit)+

    T

    2

    with

    T≈

    10.

    (Quitedi�erentfrom

    Gaussianstatisticsbecauseof

    unknowncorrelatedsystematicerrorsintheoryand

    experiments{asmeasuredbyinconsistencybetween

    experiments).

    6

  • CommentonParametrization

    For

    d

    val ,

    uval ,or

    g

    ,weuse

    xf

    (x,Q

    0)=

    A0

    xA

    1

    (1 −x

    )

    A

    2

    eA

    3

    x

    (1+

    eA

    4x

    )

    A

    5

    Thiscorrespondsto

    ddx

    ln( x

    f

    )=

    A1

    x−

    A

    2

    1−x

    +

    c

    3+

    c

    4

    x

    1+

    c

    5

    x

    i.e.,weadda1:1Pad�eform

    tothesingulartermsof

    thetraditional

    A

    0

    xA

    1

    (1 −x

    )

    A

    2

    parametrization.

    A

    suÆciently

    exibleparametrizationisimportant;

    butforconvergence,theremustnotbetoomany

    \

    atdirections."Forthatreason,someofthe

    parametersarefrozenforsome

    avors.

    (TomeasureasetofcontinuousPDFfunctionsat

    Q

    0

    onthe

    basisofa�nitesetofdatapointswouldappeartobean

    ill-posedmathematicalproblem.However,thisdiÆcultyisnot

    sosevereasmightbeexpectedsincetheactualpredictionsof

    interestthatarebasedonthePDFsarediscretequantities.In

    particular,�ne-scalestructurein

    x

    inthePDFsat

    Q

    0

    tendto

    besmoothedoutbyevolutionin

    Q

    .Theycorrespondto

    at

    directionsin

    χ

    2

    space,sotheyarenotaccuratelymeasured;but

    theyhavelittlee�ectontheapplicationsofinterest.)

    7

    MSU/CTEQ

    uncertaintymethodsa

    i

    aj

    2 - contours

    2-dim illustration of the

    neighborhood of the globalm

    inimum

    in the 16-dim parton

    parameter space

    LX

    ...

    HessianMatrixMethod:

    eigenvectorsof

    errormatrixyield40sets

    {S ±i }

    thataredisplaced

    \up"or\down"by�

    χ

    2

    =

    100from

    thebest�t.

    Geterrorbysum

    ofsquaresandconstruct

    extremePDFsforanyobservable;orsimplylook

    atextremesfrom

    the40sets.

    LagrangeMultiplierMethod:

    Track

    χ

    2

    as

    functionof

    F

    (e.g.

    σW

    )byminimizing

    χ

    2

    +

    λF

    .

    Yieldsspecial-purposePDFsthatgiveextremes

    of

    σW

    ,or〈y〉

    forrapiditydistributionof

    W

    ,or

    σfor

    t �t

    production;or

    σt

    �t ( √s

    =

    14TeV)/

    σt

    �t ( √s

    =

    2TeV),or

    MW

    mass

    measurementerror,...

    8

  • Hessian(ErrorMatrix)method

    Classicalerrorformulae

    χ

    2

    =∑ij

    (ai −

    a

    (0)

    i

    )(H

    )

    ij (aj −

    a

    (0)

    j

    )

    (�

    F

    )2=

    χ2 ∑ij

    ∂F

    ∂ai

    (H−

    1)

    ij∂F

    ∂aj

    Hessianmatrix

    H

    isinverseoferrormatrix.

    Directapplicationfailsbecauseofextreme

    di�erencesinvariationof

    χ

    2fordi�erentdirections

    inthespaceof�ttingparameters(\steep"and

    \

    at"directions),asshownbyahugerangeof

    eigenvaluesof

    H

    :

    Eigenvalues of H

    essian matrix

    9

    Convergenceproblemsaresolvedbyaniterative

    methodthat�ndsandrescalestheeigenvectorsof

    H

    ,leadingtoadiagonalform

    χ

    2

    =

    ∑i

    z

    2i

    (�

    F

    )2

    = ∑i (

    F

    (S

    (+)

    i

    )

    −F

    (S

    (−

    )

    i

    ) )

    2

    where

    S

    (+)

    i

    and

    S

    (+)

    i

    arePDFsetsthataredisplaced

    alongtheeigenvectordirections.Theiterative

    procedureisavailableinFORTRAN

    at

    http://www.pa.msu.edu/ ∼

    pumplin/iterate/

    10

  • χ

    2

    andSystematicErrors

    Thesimplestde�nition

    χ

    20

    =

    N∑i

    =

    1

    ( Di −

    Ti )

    2i

    D

    i

    =

    data

    Ti

    =

    theory

    σi

    =

    \expt.error"

    isoptimalforrandom

    Gaussianerrors,

    Di

    =

    Ti

    +

    σi r

    i

    with

    P

    (r

    )=

    e −r

    2/

    2

    2

    π.

    Withsystematicerrors,

    Di

    =

    Ti (a

    )+

    αi r

    stat,i

    +

    K∑k

    =

    1

    rk β

    ki .

    The�ttingparametersare

    {aλ }

    (theoreticalmodel)and

    {rk }

    (correctionsforsystematicerrors).

    Publishedexperimentalerrors:

    •αi

    isthe`standarddeviation'oftherandom

    uncorrelated

    error.

    •βki

    isthe`standarddeviation'ofthe

    k

    th(completely

    correlated!)systematicerroron

    Di .

    11

    Totakeintoaccountthesystematicerrors,wede�ne

    χ′2( a

    λ ,rk )

    =

    N∑i

    =

    1 (Di − ∑

    krk β

    ki −

    Ti )

    2

    α

    2i

    + ∑k

    r

    2k ,

    andminimizewithrespectto

    {rk }

    .Theresultis

    r̂k

    = ∑k′ (A

    1 )kk′ B

    k′,

    (systematicshift)

    where

    Akk′

    =

    δkk′+

    N∑i

    =

    1

    βki β

    k′i

    α

    2i

    Bk

    =

    N∑i

    =

    1

    βki ( D

    i −Ti )

    α

    2i

    .

    The

    r̂k 's

    dependonthePDFmodelparameters

    {aλ }

    .Wecan

    solveforthem

    explicitlysincethedependenceisquadratic.

    Wethenminimizetheremaining

    χ

    2( a

    )withrespecttothe

    modelparameters

    {aλ }

    .

    •{aλ }

    determine

    fi (x

    ,Q

    20 ).

    •{r̂k }

    arearetheoptimal\corrections"forsystematic

    errors;i.e.,systematicshiftstobeappliedtothedata

    pointstobringthedatafrom

    di�erentexperimentsinto

    compatibility,withintheframeworkofthetheoretical

    model.

    12

  • ComparisontoData

    ComparisonoftheCTEQ6M

    �ttodatawith

    correlatedsystematicerrors.

    dataset

    Ne

    χ

    2eχ

    2e/N

    e

    BCDMSp

    339

    377.6

    1.114

    BCDMSd

    251

    279.7

    1.114

    H1a

    104

    98.59

    0.948

    H1b

    126

    129.1

    1.024

    ZEUS

    229

    262.6

    1.147

    NMC

    F2p

    201

    304.9

    1.517

    NMC

    F2d/p

    123

    111.8

    0.909

    D�

    jet

    90

    69.0

    0.766

    CDFjet

    33

    48.57

    1.472

    Otherdatasets:

    CCFR

    ν

    DIS

    (150/156)

    E605

    Drell-Yan

    (95/119)

    E866

    Drell-Yan

    (6/15)

    CDF

    W-leptonasymmetry

    (10/11)

    13

    CTEQ6M

    �ttoZEUS

    dataatlow

    x

    510

    50100

    5001000

    Q2�G

    eV2�

    1

    1.5 2

    2.5

    F2�x,Q2��offset

    x�0.000161

    x�0.000253

    x�0.0004

    x�0.000632 x�

    0.0008

    x�0.00102

    x�0.0013

    x�0.00161

    x�0.0021

    x�0.00253

    x�0.0032

    x�0.005

    x�0.008

    ZE

    US

    datalow

    xvalues

    Thedatapointsincludetheestimatedcorrections

    forsystematicerrors.Thatistosay,thecentralvalues

    plottedhavebeenshiftedbyanamountthatisconsistentwith

    theestimatedsystematicerrors,wherethesystematicerror

    parametersaredeterminedusingotherexperimentsviathe

    global�t.

    Theerrorbarsarestatisticalerrorsonly.

    14

  • CTEQ6M

    �ttoZEUS

    dataathigh

    x

    10100

    100010000

    Q2�G

    eV2�

    0

    0.25

    0.5

    0.75 1

    1.25

    1.5

    1.75

    F2�x,Q2��offset

    x�0.013x�

    0.021

    x�0.032

    x�0.05

    x�0.08

    x�0.13

    x�0.18

    x�0.25

    x�0.4

    x�0.65

    ZE

    US

    datahigh

    xvalues

    Thedatapointsincludetheestimatedcorrections

    forsystematicerrors.

    Theerrorbarsarestatisticalerrorsonly.

    15

    �4

    �2

    02

    4�

    i

    0 20 40 60 80

    100

    N

    ZE

    US

    (a)Histogram

    ofresidualsfortheZEUSdata.The

    curveisaGaussianofwidth1.

    �4

    �2

    02

    4�D

    i �T

    i ��Αi

    0 20 40 60 80

    N

    ZE

    US

    (b)A

    similarcomparisonbutwithoutthecorrections

    forsystematicerrorsonthedatapoints.

    16

  • �4

    �2

    02

    4�

    i

    0 20 40 60 80N

    NM

    CF2p

    (a)Histogram

    ofresidualsfortheNMC

    data.

    �4

    �2

    02

    4�D

    i �T

    i ��Αi

    0 20 40 60 80

    N

    NM

    CF2p

    (b)A

    similarcomparisonbutwithoutthecorrections

    forsystematicerrorsonthedatapoints.

    17

    �2

    �1

    01

    23

    r �

    1 2 3 4

    1 2 3 4Z

    EU

    Sshifts

    SystematicshiftsfortheZEUSdata

    (10systematicerrors)

    �2

    �1

    01

    23

    r �

    1 2 3 4

    1 2 3 4N

    MC

    shifts

    SystematicshiftsfortheNMC

    data

    (11systematicerrors)

    18

  • CDF

    inclusivejetcrosssection

    -50 0 50

    100(D-T)/T

    CT

    EQ

    4M

    Statistical E

    rrors o

    nly

    -50 0 50

    100C

    TE

    Q4H

    J

    -50 0 50

    100

    50100

    150200

    250300

    350400G

    eVJet T

    ransverse E

    nerg

    y

    MR

    ST

    100200

    300400

    pT�G

    eV�

    �0.2

    �0.1 0

    0.1

    0.2

    0.3

    0.4�corrected

    data�

    theory��theory

    CD

    Finclusive

    jet

    Recallthattheseinclusivejetcrosssectionmeasurements

    providedthe�rstmajorstimulustothestudyofPDF

    uncertainties{inparticular,theuncertaintiesassociatedwith

    choicesmadeintheform

    ofparametrizationsat

    Q

    0.

    19

    CDF

    Inclusivejets{systematicerrors

    -20 0 20(a) H

    igh

    PT H

    adro

    n resp

    on

    se

    -20 0 20(b

    ) Lo

    w P

    T Had

    ron

    respo

    nse

    -20 0 20(c) E

    nerg

    y Scale S

    tability

    -20 0 20(d

    ) Frag

    men

    tation

    -20 0 20(e) U

    nd

    erlying

    Even

    t

    -20 0 20(f) N

    eutral P

    ion

    Resp

    on

    se

    -20 0 20

    100200

    300400

    (g)C

    alorim

    eter Reso

    lutio

    n

    Percentage change in cross section

    -20 0 20

    100200

    300400

    (h) N

    orm

    alization

    Tran

    sverse En

    ergy (G

    eV)

    kr̂k

    1

    0

    .511

    2

    0

    .816

    3

    0

    .022

    4

    1

    .347

    5

    1

    .307

    6

    0

    .089

    7

    0

    .222

    8

    xxx

    20

  • W

    rapiditydistributions

    Ourmethodsallow

    ustocalculatetheextreme

    predictionsduetoPDF

    uncertaintyforwhatever

    quantityisofexperimentalinterest.

    Forexample,extremesof

    σW

    ,〈y〉,〈y

    2〉

    for

    W

    productionatFNAL

    {relevantfor

    MW

    measurement:

    Samecurvesaftersubtractingcentralvalues...

    21

    Uncertaintyofthegluondistribution

    Uncertaintybands(envelopeofpossible�ts)forthe

    gluondistributionat

    Q

    2

    =

    10GeV2.

    Thecurvescorrespondto

    CTEQ5M1(solid)

    CTEQ5HJ(dashed)

    MRST2001(dotted)

    Ironically,thedi�erencesbetweentheseis

    comparabletotheestimateduncertainty!

    Theuncertaintiesofquarkdistributions(notshown)are

    smallerthanthisgluonuncertainty,becausetheDIS

    measurementsaresensitivetothesquareofthequarkchargein

    leadingorder.TheuncertaintiesofallPDFsdecreasewith

    increasing

    Q

    {\convergentevolution"

    22

  • M

    easurem

    entof

    αs

    We�ndthattheCTEQ6analysisisnicelyconsistent

    withtheWorldAveragedeterminationof

    αs (M

    Z

    ).

    Butitisnotpreciseenoughtoimprovethatvalue.

    0.090.1

    0.110.12

    0.130.14

    0.15Α

    s �MZ�

    0.090.1

    0.110.12

    0.130.14

    0.15PD

    G2000

    summ

    ary

    AverageH

    adronicJetse�

    e

    e�e�

    eventshap

    Fragmentation

    Zw

    idthepeventshap

    PolarizedD

    IS

    DIS

    taudecays

    Lattice

    Ydecay

    23

    χ

    2

    versus

    αS

    (MZ

    )forindividualdatasetsinCTEQ6

    0.1080.118

    0.1280 5 10 15 20

    CD

    Fjet�33�

    0.1080.118

    0.128

    0 2 4 6 8 10C

    DFw�11�

    0.1080.118

    0.128�

    2 0 2 4 6E

    866�15�

    0.1080.118

    0.128

    0 20 40 60D

    0jet�90�

    0.1080.118

    0.128�

    5 0 5 10 15C

    CFR

    3�87�

    0.1080.118

    0.1280 5 10 15 20 25

    E605�119�

    0.1080.118

    0.128

    0 1 2 3 4N

    A51�1�

    0.1080.118

    0.128

    0 10 20 30N

    MC

    r�123�

    0.1080.118

    0.128�

    5 0 5 10 15N

    MC

    rx�13�

    0.1080.118

    0.128

    0 10 20 30C

    CFR

    2�69�

    0.1080.118

    0.128�

    10 0 10 20 30 40H

    1b�126�

    0.1080.118

    0.128�

    10 0 10 20 30 40 50Z

    EU

    S�229�

    0.1080.118

    0.128

    0 20 40 60N

    MC

    p�201�

    0.1080.118

    0.128

    0 20 40 60B

    CD

    MSp�339�

    0.1080.118

    0.128�

    20 0 20 40 60B

    CD

    MSd�251�

    0.1080.118

    0.128�

    5 0 5 10 15 20 25 30H

    1a�104�

    24

  • Measurmentof

    αS

    (MZ

    ):

    Ifassume�

    χ2

    =

    1criterionineachexperiment,the

    experimentsareinconsistent.

    Ourerrorestimate( T

    =

    10)is

    αS

    (MZ

    )=

    0

    .1165

    ±0

    .0065

    Thiscorrespondstosomewhatconservative

    assumptions{perhapstobethoughtofasan

    e�ective\2

    σ

    "limit.Henceitiscomparabletothe

    MRST

    limitbasedon

    T

    =

    5.

    0.090.1

    0.110.12

    0.130.14

    0.15alpha

    S

    0.090.1

    0.110.12

    0.130.14

    0.15�Χ

    2�1

    rangesfrom

    GA

    BC

    DM

    Sp

    BC

    DM

    Sd

    H1a

    H1b

    ZE

    US

    NM

    Cp

    NM

    Cr

    CC

    FR2

    CC

    FR3E605CD

    Fw

    E866

    D0jet

    CD

    Fjet

    MR

    STvalue

    25

    SimilarsituationforW

    andZ

    crosssections

    W

    henastrict�

    χ

    2

    =

    1criterionwasappliedto

    self-consistentsubsetsoftheexperiments,the

    subsetswerenotconsistentwitheachother.

    Thetrueerroristhereforeconsiderablylargerthan

    χ

    2

    =

    1wouldimply.

    26

  • New

    waystomeasureconsistencyof�t

    (WorkinprogresswithJohnCollins)

    Keyidea:Inadditiontothe

    Hypothesis-testingcriterion�

    χ

    2∼√

    2

    N

    weusethestronger

    Parameter-�ttingcriterion�

    χ

    2∼

    1

    Theparametersherearerelativeweightsassignedto

    variousexperiments,ortoresultsobtainedusing

    variousexperimentalmethods.Examples:

    Plotminimum

    χ

    2i

    vs.

    χ

    2tot −

    χ

    2i

    ,where

    χ2i

    isone

    oftheexperiments,oralldataonnuclei,orall

    dataatlow

    Q

    2,...

    or•

    PlotbothasfunctionofLagrangemultiplier

    u

    where

    (1−

    u

    2i

    +

    (1+

    u

    )(χ

    2tot −

    χ

    2i

    )isthe

    quantityminimized.

    Canobtainquantitativeresultsby�ttingtoamodel

    withasinglecommonparameter

    p

    :

    χ

    2i

    =

    A

    +

    (p

    sin

    θ )

    2

    ⇒p

    =

    sin

    θ

    χ

    2not

    i

    =

    B

    +

    (p−

    S

    cos

    θ )

    2

    ⇒p

    =

    cos

    θ

    Thesedi�erby

    1,i.e.,by

    S

    \standarddeviations"

    27

    NM

    C D

    2/H2

    NM

    C D

    2/H2

    S =

    2.6

    BC

    DM

    S D

    2B

    CD

    MS

    D2

    S =

    7.6

    Fitsto8oftheexperimentsintheCTEQ5analysis

    Expt

    1

    2

    3

    4

    5

    6

    7

    8

    S

    2

    .7

    3

    .3

    3

    .3

    4

    .2

    5

    .3

    7

    .6

    7

    .4

    8

    .3

    tan

    φ0

    .56

    0

    .54

    0

    .99

    0

    .86

    0

    .71

    1

    .14

    0

    .65

    0

    .39

    28

  • Application:Uncertaintiesof

    luminosityfunctionsatLHC

    102

    103

    √s (GeV

    )

    Luminosity function at LH

    C

    Fractional Uncertainty

    0.1

    0.1

    0.1

    -0.1

    -0.1

    -0.2

    0.2

    Q-Q

    --> W

    +

    Q-Q

    --> W

    - G-G

    000 ≈≈≈ ≈

    −−

    ±

    50200

    500

    ^

    Notethatonecomponentofthe

    uncertaintyinpredictingtheHiggs

    productioncrosssectionatLHC

    is

    anuncertaintyof ∼

    8%

    duetoPDF

    uncertainty.

    29

    Outlook

    Partondistributionsoftheprotonareincreasinglywell

    measured.

    Usefultoolsareinplacetoestimatetheuncertaintyof

    PDFsandtopropagatethoseuncertaintiestophysical

    predictions.

    TheLesHouchesAccordinterfacemakesiteasytohandle

    thelargenumberofPDFsolutionsthatareneededto

    characterizeuncertainties.(hep-ph/0204316)

    Workonre�ningtheknowledgeofthe\Tolerance

    Parameter"

    T

    isunderway

    {

    Collins&

    Pumplin[hep-ph/0105207]

    {

    Statisticalbootstrapmethods

    Improvementsinthetreatmentofheavyquarke�ectsare

    inprogress.

    FermilabrunIIdataandHERA

    IIdatawillprovidethe

    nextmajorexperimentalstepsforward.

    PartonDistributionFunctionsareamajoravenue

    towardunderstandingthefundamental

    nonperturbativephysicsoftheproton.Theyarealso

    acrucialprerequisiteforprecisionStandardModel

    studiesandNew

    Physicssearchesathadroncolliders

    andexperimentswithhadrontargets.

    30