Slajd 1

1
NEW RESULTS ON ?G/G FROM COMPASS EXPERIMENT AT
CERN
Jan Pawel Nassalski Soltan Institute for Nuclear
Studies, Swierk / Warsaw, Poland On behalf of
COMPASS Collaboration
2
momentum 160 GeV intensity 2108 µ/spill
(4.8s/16.2s) longitudinal polarization -76
Beam

LHC
SPS
µ
N
6LiD, longitudinal polarization 53, -50 NH3
(2007)
Target
3
COMPASS setup
Kinematic coverage
From 2006 larger acceptance (new target
solenoid). Also 3-cell target.
4
?G/G(x) determined from Photon-Gluon Fusion events
1. Open charm cross section difference in charmed
meson production scale set by (2mc) 2 ?
small background ? experimentally difficult
2. High-pT hadron pairs cross section difference
in 2-hadron production scale set by Q2 gt
1GeV2 (this talk) or pt2 (quasi-real
photoproduction) ? large statistics ? large
background ? analysis is MC dependent
Analysing power
Gluon polarization
COMPASS analysis in LO
5
?G/G(x) from open charm events
Charmed meson reconstruction

• 60 cm thick 6LiD target cells ? no charm decay
  vertex reconstruction,
• 
  only invariant mass used,
• K/p identification in RICH important,
• cuts on D0 decay angle and zD.

Use D tagging
? cut on
6
?G/G(x) from open charm events
Event weighting
Pt target polarisation Pµ beam
polarisation aLL analysing power f dilution
factor S number of signal events B number of
background events
Each event is assigned a weight
aLL parametrised using Neural Networks
trained on MC AROMA 81 correlation
7
?G/G(x) from open charm events
Event weighting
D0
D
S and B determined from fitting mass distribution
by Gaussians and an exponentials.
raw
large variation over kinematic region
S - parametrised in terms of 10 variables
describing event kinematics and the RICH
response.
weighted
8
?G/G(x) from open charm events
Results
Systematic errors
9
?G/G(x) from high-pt events
Data selection

• Cuts on inclusive variables (x)
• Q2 gt1 GeV2,
• 0.1 lt y lt 0.9.
• Cuts on hadronic variables
• 2 charged hadrons required
• pt1,2 gt 0.7 GeV,
• xF1,2 gt0, z1,2 gt0, z1 z2 lt0.95,
• m2h1,h2 gt 1.5 GeV2.
• 500k events after selection.

(x) COMPASS results from quasi-real
photoproduction 2002-2003 data have been
published in PLB 633 (2006) 25-32.
10
?G/G(x) from high-pt events
Determination of ?G/G
The analysis is done in LO (NLO partially taken
by using parton shower in MC), resolved photon
neglected. Use measured asymmetries ALL2h and
ALLinc they have different contributions
from 3 processes


PGF
11
?G/G(x) from high-pt events
Determination of ?G/G
Event weighting

• For each event we determine
• weight w f D Pµ ß,
• Acorr.

? f, D and Pµ are obtained from the data.
All R (fractions of processes), aLL ( analysing
powers), xC and xg are obtained from inclusive
and high-pt Monte Carlo (LEPTO with MRST2004LO
and JETSET fragmentation).
? It is important to have good agreement between
data and MC.
12
?G/G(x) from high-pt events
Data vs. MC for inclusive variables
xBj
Q2
y
13
?G/G(x) from high-pt events
Data vs. MC for hadronic variables
leading hadron
subleading hadron
pt1
pt2
pt12 pt22
After tuning of JETSET fragmentation
14
?G/G(x) from high-pt events
Parametrisation using Neural Networks

• R, aLL, xC and xg are given by parameterizations
  obtained from Neural Networks
• trained on MC using following input
• xBj and Q2 for inclusive sample,
• xBj ,Q2,pl1,2,pt1,2 for high-pt sample.

Average values for high-pt events
15
?G/G(x) from high-pt events
Results
Systematic errors
16
?G/G(x) from COMPASS
New (not published)
COMPASS high-pT
Hermes high-pT
COMPASS open charm
17
SUMMARY and PROSPECTS

• New, preliminary results from open charm and
  high-pt events
• indicate ?G/G ? 0 for xg ? 0.1.
• Prospects for new results from open charm
  events
• - include data from 2007,
• - do NLO analysis.
• Prospects for new results from high-pt events
• - include data from 2006 and 2007,
• - use one high-pt hadron,
• - determine ?G/G in 2-3 bins of xg.

18
SPARES
19
?G/G(x) from open charm events
Event weighting
large variation over kinematic region
S and B determined from fitting mass distribution
by Gaussians and an exponential.
S - parametrised in terms of 10 variables
describing event kinematics and the RICH
response
20
?G/G(x) from high-pt events
Parametrisation using Neural Networks

• R, aLL, xC and xg are given by parameterizations
  obtained from Neural Networks
• trained on MC using following input
• xBj and Q2 for inclusive sample,
• xBj ,Q2,pl1,2,pt1,2 for high-pt sample.

Output variables of NN o1 and o2, where
Average values for high-pt events
21
(No Transcript)