dvcs analysis for hera i
DESCRIPTION
DVCS analysis for HERA I. Laurent Schoeffel (Saclay) and Laurent Favart (ULB). Referees : Beate Naroska and Victor Lenderman. Motivations Status of present measurements and improvements Trigger/Run selections Definition of main physical cuts BH Sample analysis (calibartion of LAR,…) - PowerPoint PPT PresentationTRANSCRIPT
DVCS analysis for HERA ILaurent Schoeffel (Saclay) and Laurent Favart (ULB)
MotivationsStatus of present measurements and improvements
Trigger/Run selectionsDefinition of main physical cutsBH Sample analysis (calibartion of LAR,…)DVCS sample Cross sections and discussion
2 main purposes / previous H1 analysis : increase statistics and measure thet slope of the DVCS x-sections => decrease both data and models uncertainties => discrimination between models?
Referees : Beate Naroska and Victor Lenderman
*+p +p’
Bj (x+)P+ (x-)P+ with P+=(p+p’)/2 and ~ xBj/2
Factorization
*
With HGeneralized Partons Dist. (GPD) k k(x+,…) *k(x-,…) d k |k(xBj,…)|² d’ for DIS
Limit 0 : q(x,bT) = FT [H(x,0,t)] => GPDs Fourier Transform of impact parameter
dependent partons densities
=> Transverse q/g dist. in coordinate space longitudinal q/g dist. in momentum space
Motivations
t=(p-p’)²
Limit t0 :
q-q correlations
Meson like Dist. Amplitude (DA)
DGLAP
ERBL
standard parton densities
Identical PDF but different correlations
dDVCS/dxdQ²dt ~ 3 x/ Q6 [CFGPD]² ebt ~ 4 3 x/ Q6 ²DIS ebt
For this analysis : we have written a new DVCS MC which calcultates x-sections+… from GPDs (at NLO ; Q² evolution of GPDs)
Improvements :1996;1997;1999e+;2000 => larger statistics (46.5 pb-1) => two samples Q²=4 GeV² and Q²=8 GeV² => 2 dependences (W) W
Better treatment of the Forward detectors (FMD;PRT) : new PRT reweight
New treatment of the Zvtx in case of DVCS sample => for the first time : measurement of d/dt exp(-B|t|) => important gain on the theoretical uncertainty
MILOU MC (for DVCS) extended at NLO treatment of proton dissociation
Preliminary analysis for EPS03 : 2000 data with 26 pb-1
Status (prel.) and main improvements
DVCS : e+ in SPACAL et (LAr) [Signal Sample]
BH with same topology as DVCS
BH with e+(LAR) [Control Sample]
Analysis strategy
Triggers selection1996 : S3 corrected from ineff.(Q²) // B. Clerbaux Ph.D. Run selection [157877-166249] B. Clerbaux Ph.D. selection FMD noise [160123-161157] excluded Lumi = 3.5 pb-1
1997 : S3 ( = 100% after fid. cuts in SPACAL) Run selection // R. Stamen Ph.D. (e.g. FMD readout pb [<184257] excluded) Lumi = 8 pb-1
1999-2000 : Triggers S0 and S3 (S3 : L2 condition SPCL_R30) We use S0 inside SPCL_R30 and S3 otherwise ( = 100% after fid. cuts) Run selection : we reject noisy PRT periods for 00-99 periods in 1999 with ~25% CJC off [257637-261338] (=> we reject 99 MB period) high prescales LumiEFF (00) = 26 pb-1 (LTOT = 37 pb-1) LumiEFF (99) = 9 pb-1 (LTOT = 10 pb-1) 96-00 => Lumi = 46.5 pb-1
1996 : Runs [160123;161157] noise in L2
1997 : Runs <184257 Readout pb in all layers
1999e+ : OK
2000 : OK
Run selection : FMD noise/readout
Noise in FMD :with random trigger events no cluster in LAr with E>0.5 GeV…
Similar studies for PRT
2000 : PRT noise (from X. Janssen)
1996 : FMD noise runs [160123;161157] => excluded
1997 : FMD readout pb for Runs<184257 and PRT for runs [190131;193331]
1999e+ : PRT noise for runs [246729;247590] { periods in 1999 with ~25% CJC off [257637-261338]}
2000 : PRT noise for runs [263535;263744] and [265400;265900]
FMD noise files extracted for this run selection (X. Janssen)
Forward Detectors quality selection
Analysis main selection
(0) Fiducial cuts in SPACAL (/year) to ensure trigger efficency ~ 100%
(1) 2 EM clusters 96-00 : ESPA>15 GeV
and 96/97 : PtLAR> 1 GeV ; 99-00 : PtLAR>1.5GeV (2) E(third cluster in LAR)<0.5 GeV
(3) Zlar>-150cm (central or fwd LAR) LAR [25;145] deg. (z/) cuts (/year)
(4) FMD and PRT : (FMD01≤1, FMD012≤2) (96 : PRT0126=0 ; 97 : PRT012=0 ; 99/00 : PRT01234=0) Note : noise file for FMD and PRT reweight // X. Janssen analysis
(5) Tracks : BH = 1 linked track (DTRA or DTNV if no vertex fitted track is found). For DVCS : no DTRA nor DTNV
Vertex simulationsBH SAMPLE
BH (elastic+inelastic from COMPTON MC)
Normalization/Lumi from COMPTON iscorrect within 2% ; fluctuations ~ 5%
Dilepton (GRAPE) + (DIFFVM) ~ 4.5% (99-00) and 6.5% (96-97) of the control sample
ZVTX is presented after reweighting
Calibrations (LAR)
Calibration EM cluster energy / tracks
Correction function for energy reconstructed in data and MC / year of analysis
BH SAMPLE
before correction (00)
after correction
MC
EclLAR-Etrack (GeV)
EclLAR (GeV)=> agreement cl/tr better than 0.5% well described by the MC
Calibrations LARClusters/Tracks
The calibration of the LARCl/tracks is correct
The calibration of SPACAL can be checked with EMPZand intercalibration with Me
Control plots 96-00
DVCS : vertex treatment
No track => no vertexTo calculate kinematics we use the nominalvalue of ZVTX
Elastic DVCS MC (MILOU) normalized / nb events
Elastic BH contribution (COMPTON)normalized to the lumi of COMPTONjustified by the control sample analysis
Proton-dissociation DVCS [MILOU] + BH (above the green line) [COMPTON]=> important contribution : for BH we keep the normalisation from COMPTON and for MILOU we normalize to a sample of noTAG events
Note : at low y (DVCS sample), MILOU (in BH mode) ~ elastic COMPTON
DVCS : Proton Dissociation TAG events Sample
BH contribution MILOU in proton-dissociation mode is generated with the t slope [d/dt~eBt] B=1.5+/-0.5 GeV-2
=> FMD or PRT signal
Elastic contribution
The Coplanarity=|e-| ~ flat for DVCS-Pdiss // low “B” value vs elastic
nb pair of hits FMD L0+1+2
Fraction of proton-dissociation in thenoTAG sample :16% for 96-97 (+/-8%)10% for 99-00 (+/-5%)
Same conventions as before
DVCS : calibrations checks
DVCS : coplanarity =|e-|
Broadening of the distributionw.r.t. BH control sample=> more events have larger |t| values / BH sample also reflected on Me or |t| spectra
steeper fall for BH
BH sample
Shapes for the
In the LAR we have a good descriptionof the shape estimators for the realphoton => the H1 simulation with the new cluster shape parametrization is doing well !
DVCS 96-00
, backgrounds are takeninto account (DIFFVM) :1% for 99-003.5% for 96-97
=> 2 samples 96-97 at low Q² 99-00 at larger Q²
96-97 :2 GeV² < Q² < 20 GeV²30 GeV < W <120 GeV|t| < 1 GeV-2
with
<Q²> = 4 GeV²<W> = 71 GeV
99-00 :4 GeV² < Q² < 80 GeV²30 GeV < W <140 GeV|t| < 1 GeV-2
with
<Q²> = 8 GeV²<W> = 82 GeV
Cross sections determination
calculated BH,
0)(iradgen
1)(iradrec
pdissbckgevtsDVCS
W
Q²d
dσ
LW
Q²Δ
N
Nε
NNN
W
Q²d
dσ
n2
a2p*γ2p*γ22
)Q
1(Ayy),(Qσ withy),(Qy)σ,Γ(Q
dydQ
dσ
Cross-section (ep) :
Related to the *p cross-section via the flux factor :
=> we can compute *p and determine a and n iteratively…
Note : we come back later on the cross-section in t
5% inefficency for the BDC in 99e+ for Q² < 20 GeV² ; 3.5% CJC mixing in 97 ; trigger efficency(Q²) for 96
The use of correction factors Nrec/Ngen is jsutified by the good understandingof DATA/MC we have presented (detector resolutions correctly described,…)
Systematics
Substraction of proton dissociation : 8% (96-97) ; 5% (99-00) Correction factor : <A(Q²;W)/A>~5% [<A(t)/A>~7% with a max of 12% -last bin-] note : <A(Q²;W)/A>~5% = 4% (varying the t-slope) 3% PRT reweighting effect Determination of a and n : => / = 7% (96-97) ; 5% (99-00) Substraction of the BH contribution : 5% => / = 3% in the last W bin Luminosity measurement : 1.5% Noise from FMD 1.5% (96-97) ; 0.1% (99-00) CJC noise : 2% ZVTX position : 1.5 cm (to take into account the treatment of ZVTX) => / < 4.5% Electron energy : 1% => / < 3% Photon energy : 2% => / < 3% Electron angle : 1.3 mrad => / < 4% Photon angle : 3 mrad => / < 3%
We can combine these measurements w.r.t. luminosities of both sample => one global (Q²) for 46.5 pb-1
Q² x-sections
Correction factors (Q² bins in GeV²) =>
99-00 [4;6.5] : 0.16 => 1 without fid. cuts[6.5;11] : 0.34[11;20] : 0.67[20;30] : 0.60[30;80] : 0.58
96-97 [2;4] : 0.08 => 0.96 without fid. cuts[4;6.5] : 0.21 => 0.92 id.[6.5;11] : 0.39[11;20] : 0.48
Good agreement with EPS03 prel.
Q² x-sectionsComparisons with previousresults
There’s a difference of about 10%w.r.t. 97 analysis difference in thecalculation of the correction factors.In 97 analysis, Ngen was calculatedwith the cut on PT;LAR included, NOT in the present analysis…
with the PT;LAR cut
Here again, we can combine these measurements w.r.t. luminosities of both sample => (W)
W dependence for the 2 values of Q²
W x-sectionsCorrection factors (W bins in GeV) =>
99-00 [30;60] : 0.31[60;80] : 0.37[80;100] : 0.47[100;120] : 0.35[120;140] : 0.18 => 0.97 without fid. cuts
96-97 [30;60] : 0.16 => 0.94 without fid. cuts[60;80] : 0.23 => 0.95 id.[80;100] : 0.18 => 0.96 id.[100;120] : 0.09 => 1 id.
Good agreement with EPS03
W x-sectionsComparions with previousresults
with the PT;LAR cut
Same comment as before forthe comparison with 97 analysis
Cross section(t)
tBn2
a2p*γ
e)Q
1(y
dt
t)y,,(Qdσ
For cross section determination, same methode as before with :
=> two values of B for the 2 samples (different Q²)
Note that we can not combineto the same value of Q² asB “may” depend on Q²
To determine a global B value,we just combine the 2 samplesw.r.t. luminosities=> B=5.84 +/- 0.80 GeV-2
at Q² = 6 GeV²
Correction factors (|t| bin in GeV²)99-00 [0;0.2] : 0.26 => 0.92 without fid. cuts[0.2;0.4] : 0.49[0.4;0.6] : 0.62[0.6;1.0] : 0.6996-97 [0;0.2] : 0.13 => 0.87 without fid. cuts[0.2;0.4] : 0.21 => 0.97 id.[0.4;0.6] : 0.36[0.6;1.0] : 0.71
(Q²)/models
Error band is calculated withB=5.84 +/- 0.80 GeV-2
QCD predictions : good agreementwith CTEQ parametrization forthe diagonal part of the input for GPDs =>HS,V,g(x,) QS,V,g(x) : DGLAPthe “internal” skewing is generated by the NLO evolution.-in ERBL domain H is continuous and + 2 first “polynomial sum rules”
Dipole models : DD overestimatesthe data ; FM is reproducing nicely the measurements
(W)/models
Error band is calculated withB=5.84 +/- 0.80 GeV-2Again a good agreement isfound with CTEQ for QCDmodels and with FM for dipole inspired models.
Note that with the extraction ofd/dt we have a stronger constraint on B (compared toprevious analysis)
=> possible discrimination between parametrizations/ models…
Analysis summary
Measurement of the DVCS x-sections at HERA I (e+) => L = 46.5 pb-1
=> increase of the statistics (factor 4)=> 2 bins in Q² as a function of W 2 bins in Q² as a function of t
All simulations are done with the new cluster shape parametrization
Noise files for FMD and new reweight of PRT
Better treatment of the Z vertexProton dissociation included in MILOU (for DVCS) and SOPHIA for COMPTON=> t slope measurement for DVCS process possible (never measured before)
On the above plot, the 2 best models (with b(Q²))
=> * DVCS is a precise NLO QCD calculation * GPDs models using only dynamicaly generated skewing from classic PDFs leads to very good agreement in shape and normalization with DVCS
* Large sensitivity to the ERBL domain (factor 5)
DRAFT distributed within 1 week
Conclusions
Comments on BH in MILOU MC
Complete formula for BH = d [A/P()+Bcos()/P()+Ccos(2)/P()]With P() = 1+Xcos() and X~2(2-y)/(1-y)[-t(1-y)/Q²]
For integrals convergence we need X<1 which is the case (“most of thetime”) due to t/Q²<<1
1. when y->1 (BH sample), X can be very close to 1- (if >1, then the event is not considered) =>d cos(2)/(1+Xcos()) is very unstable numerically => source of the differences between BH new MC and COMPTON=> neglecting terms [B] and [C] gives an agreement of up to 5% with COMPTON
2. at low y (for the DVCS sample), X<<1 => very good agreement up to 3%
Calibrations LARClusters/Tracks
Effect of fiducial cuts (z/) in LAr
Z (LAr)
(LAr)
LAr Control plots Px;Py;Pz
Third Cluster (E3<0.5 GeV)
E3 (energy of thethird cluster…)
Sum of Ecl except the 2 EM clusters
Control plots LarDVCS “elastic”(noTAG sample)
Control plots LarDVCS “elastic”(noTAG sample)
Control plotsFor P-dissociation (TAG sample) :
Third cluster E3 in the “elastic” case (noTAG sample) : 96-00
Kinematicsdetermined from DA
Q² (GeV²)
W (GeV)
-t=|PTe+PT|² (GeV²)
(rad)
Ee(ini)=EMPZ/2 : DVCS generated with MILOU (elastic) : no cut is calculatedWith gen. variables
DVCS generated with MILOU (elastic)with all cuts applied : is calculatedwith rec. variables (EMPZ also…)=> LARGE fluctuations (factor ~2)!
Need a boost+rotation in the Belitsky frame=> Ei(k) must be known precisely=> we take Ei=EMPZ/2 to take into account QED radiation effects
Measuring the
Ee(ini)=27.5 GeV
EMPZ (GeV) GEN/REC for the Signal Sample BH MC (green) and DVCS MC (black) (normalized to nb of events)
The behaviour is similar for the BH contribution to the Signal Samplewith even larger fluctations
=> impossible to extract an asymetry [coming from the BH/DVCS interference] !
DATA/MC for 99-00 period [Ee(ini)=EMPZ/2]
The use of BST ?
For one part of the 00 sample ~10 pb-1
we can use the vertex reconstructedfrom BST instead of ZNOM
We have checked that it does not bringimprovements for DVCS events w.r.t. thepresent analysis
Differences at large t are covered by thesystematics on ZVTX and A(t)