G/G from high-pT events in SMC

•Determination of ∆G/G from Photon Gluon Fusion•Analysis in Leading Order where it can be separated•based on simulations with LEPTO•Search for sample with high PGF contribution•application for DIS region, SMC data with Q2 >1GeV2

E.Rondio

for Spin Muon Collaboration (SMC)

Sołtan Institute for Nuclear Studies

Warsaw, Poland

Workshop on Hadron Structure and Spectroscopy, Paris, March 1st to 3rd 2004

History

• Idea proposed by R.D.Carlitz, J.C.Collins and A.H.Mueller, Phys.Lett.B 214, 229 (1988)

• Revisited by A.Bravar,D.von Harrach and A.Kotzinian, Phys.Lett.B 421, 349 (1998)

• Method used in HERMES for photoproductionHERMES, A.Airapetian et al., Phys.Rev.Lett.84, 2584 (2000)

• Here application for DIS region, SMC data with Q2 >1GeV2

SMC, B.Adeva et al.., submitted to Phys.Rev.D, hep-ex/0402010

QCDCQCDCLL

LPLPLL

PGFPGFLL

lhhXlN

RaRaq

q

RaG

ΔGA

G/G evaluation from measured asymmetry

where: AlNlhhX measured asymmetry,

q/q approximated by A1/D asymmetry N,

aLL partonic asymmetry,

R fraction of contributing processes

Applicability and restrictionsSplitting between processes only in LO >>> when higher order effects expected to be important

it can not be used >>> here scale dependence was checked and found

small, so no clear signal of such strong dependenceUsing information which is not an observable (which type of interaction given event is) >>> so it has to be taken from simulation >>> the above makes analysis model dependent (using Lepto or eg. Pythia can give different results) but … a tool to check reliability is comparison of data with

MC Spin effects do not have to be simulated >>>measurement is independent of assumptions about

polarized parton distributions and spin effects in fragmentation

Why events with high-pT hadrons ?

PGF

signal

LP QCDC

• Two high-pT hadrons more likely in QCDC and PGF because in LP source of pT only fragmentation in PGF and QCDC in addition pT from hard scattering

background

Target: butanol, ammonia –

proton d-butanol - deuteron

Beam:

µ+ 190 GeV

Pµ= -0.78±0.03

Measured asymmetry:

lhhXlNTμ fAPP

NN

NN

where: beam, target

Selected events cover following x, y, Q2

region

xBj xBj

yQ2

[GeV]

Conditions on hadrons in the final state

2 hadrons: pT> 0.7GeV, z>0.1, xF>0.1

(no electron contamination observed after these cuts)

Event selection for asymmetry

vertex in target half, beam through full target length, stable conditions

Kinematic cuts and regions: Q2>1GeV2, 0.4<y<0.9, acceptance for and h

Statistics after selections

proton deuteron

81 178 75 266

below 0.5% of the inclusive sample

Monte Carlo studies

→ studies for DIS µp interactions at 190 GeV→ LEPTO simulations, Q2 1 GeV2

→ detector and reconstruction effects• geometrical acceptance for hadrons• simulations of trigger conditions• looses in reconstruction (chamber efficiencies)• smearing for scattered µ and hadrons (1/p, angles)• secondary interaction in target for hadrons

→conditions in MC generation scale for hard processes (syst.errors only)

cut-off’s in matrix element calculation parameters of symmetric fragmentation function

Data and Monte Carlo agree at the level of 10-25%

To be used for selections of PGF and ∆G evaluation

Data and Monte Carlo comparison

Event kinematicsSensitive to trigger mixture, smearing

Hadron variablesSensitive to smearing and MC generation (ff)

Data

MC

Simulation of exp. conditions

Sensitive to details of target:

position, angle

Good description after inclusion

of hadron secondary interactions

Modification of fragm. function

a=0.5, b=0.1 (stand.)

zbma Tezzzf /1 2

)1()(

Contribution of PGF processFor SMC experimental

conditions Lepto at generation level RPGF = 8% events with two hadrons

(phad>5GeV) RPGF = 12% additionally pT

had > 0.7 GeV RPGF = 24%

How to get more? Two methods tried:• kinematical selections

(cuts) and • Neural Network

classification (NN)

The criteria to judge the selection:

PGF(in)

PGF(out)Efficiency

PGF(out)QCDC(out)LP(out)

PGF(out)Purity

Several variables tried

Opposite charges of hadrons –

small effect, 1/3 events lost

Azimuthal angle between hadrons

– no improvement

Best - ∑p2T

Cuts on hadron variables

Neural network

• input layer: event kinematics (x, y, Q2) and hadron variables (E1,2, pT1,2, charge, azimuthal angle between pT of two selected hadrons), • best way to use correlations• output layer: single unit number within range (0,1)

NN response Architecture: multi-layer feed-forward configuration

Neural Network responsenumber within range <0,1.> events at high values of NN response are more likely to

be PGF

PGF enriched sample

selected by setting the threshold

on the NN response

NN treshold

Processes

contributions

for two selection

method

PDG

QCDC

LO

PGF

LO

QCDC

Best result of cut

selection based

on pT2

compared to NN

Asymmetry AlNlhhX

Systematic uncertainties:

•False asymmetries from acceptance variation

•Calculation of radiative effects (unpolarized and polarized part)

Effect due to restricted phace space

•Polarization of beam and target

•Target material

Selection Proton AlNlhhX Q2

Deuteron AlNlhhX Q2

pT2 0.0180.0710.010 7.07 0.054

0.0930.008 7.91

NN 0.0300.0570.010 3.30 0.070 0.0770.010

4.00Interpretation of A lN→ lhhX in terms of ∆G/G requires

additional information from MC simulation.

AlNlhhX

pT0.7GeV pT22.5GeV2

NN0.26

Results on Asymmetry

Input for calculation of ∆G/G

∆q/q approximated by A1·D

neglecting PGF contribution in inclusive

A1 measurements,

ok. only if RPDG(incl)<< RPDG(selected)

From other measurements:

A1 asymmetry taken from fit

to all experimental data

f(x)=xa·(1-ebx)+c ,

Q2 dependence neglected

proton

deuteron

Hermes

Hermes

Input for calculation of ∆G/G From MC simulations:

• aLL calculated in POLDIS

aLLLP 0.8

aLLQCDC 0.6

aLLPGF -0.44

• fractions of processes Selection RLP RQCDC RPGF

pT22.5GeV2 26% 42% 32%

NN 0.26 38% 30% 32%

Important consistency between data and MC

Statistical precision of ∆G/G

Gluon polarization

Separately for proton and deuteron

∆G/G determined for a given fraction of nucleon momentum carried by gluons η

Selection G/G (G/G)stat genPGF

pT22.5GeV2 -0.07 0.40 0.09

NN 0.26 -0.20 0.29 0.07

Average value final SMC result on

∆G/G =-0.200.290.11

SMCHermes

NNpT1

2+pT22

comparison

• Difference < 2 σ

• Different process DIS vs. Photoproduction

• Factor 2 difference

in ηgluon

Systematic uncertainty on ∆G/G Contribution to the systematic

due to uncertainty on parameters used in MC :

• sensitivity to fragmentation, • cutoffs in matrix elements calculations• scale dependence (2Q2,Q2/2),

Changes in RPGF < 5%

Similar effect for

pT of faster hadron

Error source uncertainty on ∆G/G

Precision of A1 fit 0.026Scale change Q2/2 ; 2Q2 0.010Fragmentation function 0.034Cutoffs in matrix elements

0.008

err. from MC and A1 0.053Syst.error from Alhh 0.062

Total 0.115

+20%R / -20%R 0.067 / 0.100

+4% aLL / -4% aLL 0.015 / 0.017

Systematic uncertainty on ∆G/G

Changing only R or aLL

Summary• The method of ∆G/G evaluation from asymmetry for

events with high-pT hadrons was applied to SMC data in DIS region

• Results obtained for cut selection and neural network ∆G/Gstat. sys. -0.07 ± 0.40 ± 0.11 cut ∑pT

2

∆G/Gstat. sys. -0.20 0.29 0.11 NNpoints to rather small value of gluon polarization

• precision of ∆G/G limited by the statistical error, • systematic error controlable (and can be reduced for

high statistics by precise data/MC comparison)

• Improvement on accuracy of ∆G/G in future: COMPASS at CERN, RHIC at BNL, E161 at SLAC