Deeply Virtual Compton Scattering JLab Franck Sabati CEA

Deeply Virtual Compton Scattering @ JLab Franck Sabatié CEA Saclay On behalf of the Hall A and Hall B collaborations Pacific Spin ’ 07 - Vancouver August 1 st 2007 Introduction Non-dedicated measurements E 00 -110 experiment in Hall A E 03 -106 experiment in Hall A E 00 -110 prelim. deep-p 0 results E 1 -DVCS experiment in Hall B Summary Pacific Spin ’ 07 Exclusive’ 07

GPDs from Theory to Experiment x+x x-x GPDs g* g t Theory Handbag Diagram Factorization theorem states: 2. The GPDs enter then DVCS as an integral over x: Q 2 and large amplitude In the suitable asymptotic limit, at the x. B and t fixed - GPDs appear in real part throughthe a PP integral over is x the leading handbag diagram contribution tox=x DVCS. - GPDs appear in the imaginary part but at the line 1. Needs to be checked !!! g* g Experiment but it’s not so simple… g* g Physical process Collins, Freund

Experimental observables linked to GPDs Experimentally, DVCS is undistinguishable with Bethe-Heitler However, we know FF at low t and BH is fully calculable Using a polarized beam on an unpolarized target, 2 observables can be measured: Ji, Kroll, Guichon, Diehl, Pire, … At JLab energies, |TDVCS|2 supposed small

Into the harmonic structure of DVCS |TBH|2 Interference term hadronic plane e-’ e- g* leptonic plane Belitsky, Mueller, Kirchner j g p BH propagators j dependence

Tests of scaling 1. Twist-2 terms should dominate s and Ds 2. All coefficients have Q 2 dependence which can be tested!

Special case of the asymmetry The asymmetry can be written as: Pro: easier experimentally, smaller RC, smaller systematics Con: direct extraction of GPDs is model- (or hypothesis-) dependent (denominator complicated and unknown) It was naturally the first observable extracted from non-dedicated experiments…

Published non-dedicated JLab/Hall B results on ALU and AUL JLab/Hall B - E 1 JLab/Hall B - Eg 1 ALU AUL PRL 97, 072002 (2006) PRL 87, 182002 (2001) Both results show, with a limited statistics, a sin j behavior (necessary condition for handbag dominance) In the ALU result, models (VGG) tend to over-estimate the data

E 00 -110 experimental setup and performances • 75% polarized 2. 5 u. A electron beam • 15 cm LH 2 target • Left Hall A HRS with electron package • 11 x 12 block Pb. F 2 electromagnetic calorimeter • 5 x 20 block plastic scintillator array Dt (ns) for 9 -block around predicted « DVCS » block Vertex resolution Pbeam=75. 32% ± 0. 07% (stat) 1. 2 mm

E 00 -110 kinematics The calorimeter is centered on the virtual photon direction 50 days of beam time in the fall 2004, at 2. 5 m. A intensity

Analysis – Looking for DVCS events HRS: Cerenkov, vertex, flat-acceptance cut with R-functions Calo: 1 cluster in coincidence in the calorimeter above 1 Ge. V With both: subtract accidentals, build missing mass of (e, g) system

Analysis – po subtraction effect on missing mass spectrum Using p 0→ 2 g events in the calorimeter, the p 0 contribution is subtracted bin by bin After p 0 subtraction

Analysis – Exclusivity check using Proton Array and MC Using Proton-Array, we compare the missing mass spectrum of the triple and double-coincidence events. The missing mass spectrum using the Monte-Carlo gives the same position and width. Using the cut shown on the Fig. , the contamination from inelastic channels is estimated to be under 3%. Normalized (e, p, g) Monte-Carlo triple coincidence events (e, g)X – (e, p, g)

Difference of cross-sections PRL 97, 262002 (2006) Twist-2 Twist-3 Corrected for real+virtual RC Corrected for efficiency Corrected for acceptance Corrected for resolution effects Checked elastic cross-section @ ~1% Extracted Twist-3 contribution small ! New work by P. Guichon !

Q 2 dependence and test of scaling <-t>=0. 26 Ge. V 2, <x. B>=0. 36 Twist-2 Twist-3 No Q 2 dependence: strong indication for scaling behavior and handbag dominance Twist 4+ contributions are smaller than 10%

Total cross-section PRL 97, 262002 (2006) Corrected for real+virtual RC for efficiency for acceptance for resolution effects Extracted Twist-3 contribution small ! but impossible to disentangle DVCS 2 from the interference term

DVCS on the neutron in JLab/Hall A: E 03 -106 LD 2 target 24000 fb-1 x. B=0. 36, Q 2=1. 9 Ge. V 2 MODEL-DEPENDENT Ju-Jd extraction

Deep-p 0 electroproduction in JLab/Hall A Same data as E 00 -110, but: - 2 g requirement in the calorimeter at p 0 mass, - ep 0 X at proton mass. counts Fit to data 0 180 j 360

Cross-sections for deep-p 0 production in JLab/Hall A VGG Laget +rediff. Laget -Low-t cross-section largely overshoots GPD & JML models. -TT term is large, suggesting a large transverse component. -But p 0 cross-section ratio for proton and deuteron targets is found ~ 1 More data needed ! PRELIMINARY

E 1 -DVCS with CLAS : a dedicated DVCS experiment in Hall B Beam energy: Beam Polarization: Integ. Luminosity: ~5. 8 Ge. V 75 -85% 45 fb-1 Mgg (Ge. V 2) Inner Calorimeter + Moller shielding solenoid ~50 cm

E 1 -DVCS kinematical coverage and binning W 2 > 4 Ge. V 2 Q 2 > 1 Ge. V 2

E 1 -DVCS : Asymmetry as a function of x. B and Q 2 <-t> = 0. 18 Ge. V 2 <-t> = 0. 30 Ge. V 2 <-t> = 0. 49 Ge. V 2 <-t> = 0. 76 Ge. V 2 Accurate data in a large kinematical domain Integrated over t

E 1 -DVCS : ALU(90°) as a function of |t| + models VGG twist-2+3 VGG twist-2 JM Laget

E 1 -DVCS : Cross-sections over a wide kinematical range Ph. D Thesis H. S. Jo (IPNO) 0. 09<-t<0. 2 0. 6<-t<1 0. 2<-t<0. 4 M I L 1<-t<1. 5 E PR 0. 4<-t<0. 6 Y R A IN 1. 5<-t<2

Summary (DVCS) DVCS BSA (Hall B/CLAS): q Data in a large kinematical range and good statistics. It will give a very hard time to models, to fit the whole range in Q 2, x. B and t. q Data seem to favor JM Laget model at low-t and VGG model at high t. DVCS Cross-section difference (Hall A): q High statistics test of scaling: Strong support for twist-2 dominance q First model-independent extraction of GPD linear combination from DVCS data in the twist-3 approximation q Upper limit set on twist-4+ effects in the cross-section difference: twist>3 contribution is smaller than 10% DVCS Total cross-section (Hall A): q Bethe-Heitler is not dominant everywhere q |DVCS|2 terms might be sizeable (more on this in J. Roche’s talk)

Observables and their relationship to GPDs The cross-section difference accesses the imaginary part of DVCS and therefore GPDs at x = x The total cross-section accesses the real part of DVCS and therefore an integral of GPDs over x

Deep r 0 data largely overshoots GPD models at low W New !