DVCS analysis at HERA and implications for COMPASS

1
DVCS analysis at HERAand implications for COMPASS
Laurent Schoeffel CEA Saclay (SPP) All analysis
done in collaboration with Emmanuel Sauvan CPPM -
Marseille
Short introduction Experimental analysis leading
to DVCS DVCS cross sections and interpretation
at HERA Implications for COMPASS
2
What do we test at HERA ? (I) inclusive
measurements of F2(x,Q²) and Gluon density
ep?eX
Saturation line ln Qs²(Y)
BFKL
DGLAP
3
What do we test at HERA ? (II) Physics invariant
along any lines // saturation line gt New
fundamental scale ? Qs²(x) Q0²(x/x0)-0.3
?tot (x,Q²) ?tot (?Q²/Qs²) with ?tot 4?²? F2/Q²
Also verified for any process at low x (including
DVCS) (L.S., C.Marquet 2006)
All F2 measurements scale in only one variable!
?Q²/Qs²
4
What do we test at HERA ? (III) Emission of
gluons local/non-local (?) in kT towards the high
density limit
At low x (lt0.01) more and more gluon are produced
Pgg1/x and populate the proton core Possible
recombination effects (high density of gluons)
gt non lineatity in evolution equation ?
5
2 major interests of exclusive processes
? Better constraints on the gluon density ?
Spatial transverse distribution
bulk
pion cloud
Gluons radiation from q,g (in p)

• F(x,-?T²?²) ? d²? ei(?.?) F(x,? ?²)
• Fg,q appears in amplitude of VM, DVCS prod.
• gt lt?²gt 4 d/dt (F(x,t)/F(x,0))
• constant ? ln(1/x) (at low x)

6
Spatial transverse distributions (I)

• Measurements of t dependences of exclusive
  processes
• gt Spatial distribution of partons F(x,? ?²)
•  new information  which provides an
  essential
• feature of the nucleon structure
• BCA for DVCS at COMPASS is very promising in this
  respect
• Crucial measurements for normalisation of
  cross-section prediction.
• If we assume an exponential form d?/dt b ?
  exp(bt)
• gt b gives the normalisation for any theoretical
  predictions...

Present status at HERA Ex b 6 GeV-2 gt
lt?²gt2b0.46 fm²
DVCS is a key process for all this
7
Spatial transverse distributions (II)
Parametrisation(Q²M²) bb(Q²)0.6
14/(Q²M²)0.26 1 gt Common description of all
results
8
Goal measurement of the process e p ?
e ? Y Yp
2 EM clusters SpaCal (large scattered angle)
and LAr (central) No other activity in
calorimeters (above noise level) Forward
detectors (FMD, PRT) used to TAG an event with MY
gt 2 GeV
e
e
920 GeV
27.5 GeV
?
All plots are presented for HERA II data 2004
and/or 2005
35 pb-1
95 pb-1
9
measurement of the process e p ? e ?
Y kinematics
Double Angle (DA) method
10
Comments on the kinematics

• The kinematics is over-constrained
• 1. Measurement from scattered lepton
• 2. Measurement from  angles 
• For DVCS (?e,??) or inclusive process, ep?eX,
  (?e,?h)
• In addition to formulae of the last slide, we
  need
• ptDA 2Ee,beam /(tg(?e/2) tg(?h/2))
• EDA ptDA/sin(?e)

The idea is to use the DA kinematics to benefit
from the good resolution in angles (CJC) gt
essential work on alignments
11
Some important issues in the analysis
Alignements of trackers / calorimeters Dead
zones of detectors Calibrations Efficiencies of
selection cuts Triggers (uncovered in this
talk) Understanding the forward part
Main analysis done on large statistic
samples Inclusive neutral current
ep?e(SpaCal) X Compton (2 EM clusters
in SpaCal)
Then, move to the specific DVCS analysis
12
Alignements X,Y SpaCal / CJC
with an inclusive selection ep?e(SpaCal)X
13
Alignements X,Y SpaCal / CJC
14
Alignements Z SpaCal / CJC
Essential to get a good measurement of the
variable t(p-p)²
2005 data vs MC
15
Checking the EM Calibration of the SpaCal
NC ep ? e (SpaCal) X
Ratio of (Ee/Eda)Data / (Ee/Eda)MC
A first calibration is done using events from the
kinematic peak.
2004 data vs MC
Eda (GeV)
Similar result for 2005 data
16
Checking the HAD Calibration of the LAr
NC ep ? e (SpaCal) X (HAD. Final state h)
2004 data vs MC
Similar result for 2005 data
17
Control plots for the NC inclusive selection
NC ep ? e (SpaCal) X
Data 2004 vs MC
After all corrections Kinematic variables xBj,
yQ²/(s.xBj) Good description within the
systematic errors
18
F2
F2 determination from 2004 data
Data 2004
gt alignements, calibrations,
efficiencies are done correctly!
Data/theory
1.05
1
0.95
19
We can control the luminosity determination of
the experiment (time)
gt Good understanding of the detector
simulations
20
Now we can move to the DVCS analysis e p ? e
? Y
e
?
2 examples of new specific studies gt Response of
the forward detectors Over calibration of
clusters at low EM energy (Elt5 GeV) in the LAr
21
Understanding the forward part

• FMD and PRT are used to caracterise an elastic
  evt (Yp)
• 
  / proton-dissociative evt ( large 
  MY)
• gt We do not TAG a scattered proton
• but we can TAG the Y system (with MY gt 2 GeV)
  with a good eff.
• Then, we correct the cross-section(Y) to the
  elastic(p) with a systematic
• uncertainty of 50 of the correction gt error of
  5 on ?DVCS

Tagging eff. for a pdis evt
Log10(MY/1GeV)
22
BH sample (control plots)
e
?
DATA 2005

• Good understanding of the sample
• Essential as the BH component
• 40/50 of the DVCS sample
• ep ?e(SpaCal) ?(LAr) Y

BH
23
DVCS sample (control plots)
DATA 2005
DATA 2005
DVCS elastic
DVCS pdis
BH component
Data 2004 L34 pb-1 320 DVCS evts Data
2005 L95 pb-1 870 DVCS evts
DATA 2004
DATA 2005
24
DVCS cross sections ?(?p??p)
Published data (HERAI H1 and ZEUS) and
preliminary results for 2004
25
What about the Interference ? (I)

• d?/d? DVCSBHIntcos(?)
• In our measurement, we  integrate  over ?
• gt the contribution of the interference term ? 0
  ?!
• In the cross sections we present, this is what
  is assumed
• But, Fiducial cuts could distort the cos(?)
  dependence gt non zero value ?
• Interesting pb at the frontier of HERA/COMPASS in
  term of analysis 
• Then, we compute ? in the Belitsky frame (Fig.)
• Boost in the proton rest frame (Ep,beam920 GeV)
• Rotate the axis system to have the z axis along
  the ?

26
What about the Interference ? (II)
2005 e- data L 95 pb-1

• All the MC are without interference and the sum
  is fine vs Data
• 1. The interference contribution is low (if any)
• 2. The ? spectrum is symetric gt no distortion
  from analysis Fiducial cuts
• gt any cos(?) term will give a null
  contribution after integration
• missing in the published/present analysis
  a systematic error on this
• 3. Can we understand this behaviour ?

27
What about the Interference ? (III) MC study

• For the same Lumi as in the Data (e-)
• Generation of BHDVCSInt(f)
• BH in violin
• without inteference (f0)
• sum black points
• with interference (f2)
• sum filled histogram
• 2. We smear energies and angles
• with resolutions from Data
• note the structure lt Boost (Ep)

Note same kin. cuts as in the analysis histo
similar fiducial cuts to reproduce the acceptance
28
What about the Interference ? (IV) Good
understanding of the ? spectrum
For 95 pb-1 in e and e- collisions
MC study
Perspective Measurement of the BCA ?
29
Back to DVCS cross sections ?(?p??p)t(p-p)²
1996-2000 analysis
d?/dt d?/dtt0 exp(-bt) b6.02 /-
0.35 /- 0.39 GeV-2 best description gt bb(Q²)
(see introduction)
Normalisation of cross sections 1/b (gt b
value included in MC and predictions gt absolute
normalisation)
30
A word on the impact of GPDs (I)
There are 2 main graphs to compute for DVCS
process

• DVCS Amplitude ? dx 1/ (x- ? i?) - 1/(x?
  i?) ?eq²Hq(x,?,t) axial O(m/Q)
• gt Im(A(?p)) 4?²?/W² ?eq²HqS(?,?,t)
  with ?xBj/(2-xBj)
• Re(A (?p)) P.V.
• Note At NLO, the gg graph contributes
  with Im(ANLO) Im(ALO) 0.7

31
A word on the impact of GPDs (II) Freund
hep-ph/0306012
Modeling GPD at low scale HS,V,g(x,?) ?
QS,V,g(x) (Ji notations) at an intital scale 1
GeV² and ? dependence generated by the GPD
evolution (DGLAPERBL) which is equivalent
to fS,V,g(?,?,t) QS,V,g(?) hS,V,g (?,?) FS,V,g
(t) with bprofile??
LO gt NLO AG lt0 (-30
tot. Amplitude)
Correction for the Re() part of about 10/20
(dispersion rel.)
?(?p??p) ImA(?p??p)² (1?²) / (16? b)
32
DVCS cross sections ?(?p??p) Comparisons with
GPDs prediction (NLO)
HS,V,g(x,?) ? QS,V,g(x) _at_ 1 GeV² PDF used ?
CTEQ6M ? Dependence generated via the GPD
evolution
Good description of the Q² and W dependences!
33
DVCS cross section prediction (I) What happens
if we forget about GPDs and use PDFs ?
It means we would identify ImA(?p??p) and
ImA(?p??p) Im(A(?p??p)) 4?²?/W² ltegt²
GPDS(?,?) with ?xBj/2 Im(A(?p??p)) 4?²?/W²
ltegt² PDFS(xBj)
This can be checked experimentally
Im(A(?p??p)) 4?²?/Q² F2 (more precisely
the transverse part FT

as the ? is transverse for DVCS-) Im(A(?p??p
)) 16?b ?DVCS /(1?²)1/2 and we have
measured F2 (or FTF2-FL) and ?DVCS
34
DVCS cross section prediction (II) What happens
if we forget about GPDs and use PDFs ?
R Im(A(?Tp??p))/ Im(A(?Tp??Tp))
As we observe that R is not equal to 1 ! R 2 gt
factor 4 on the cross section prediction
35
Direct check on GPDPDF functions
(a.u.)
gt PDF
H
?0.1 gt xBj0.2 GPD(x?,?)2.PDF(xBj)
x
36
DVCS cross section prediction (III) Can we
refine the use PDFs ?

• Using PDFS(?xBj/2) ?
• which means that we would write
• GPDS(?,?) ? PDFS(?)
• This is not correct and even worst at low x
• in practice GPD(?,?) gt PDF(?) and it is much
  greater at low x
• and there is no simple way to take into account
  the gg graph!
• gt The use of GPDs is essential
• and the sensitivity to the exact form
• of the GPD param. is large!

37
DVCS cross section prediction (IV)
Reminder of our model HS,V,g(x,?) ? QS,V,g(x)
_at_1 GeV² and ? dependence generated by the GPD
evolution (DGLAPERBL) Freedom in the choice of
PDF
PDF _at_ 1 GeV² MRST2001 CTEQ
gt sensitivity to GPD param.
38
Implications for COMPASS (GPDs) (I) Using
Freund modeling of GPDs

• Reminder
• CA Re(Amplitude)
• P.V.(Coef?GPDs)
• At xBj0.1 (for ex.)
• We observe a strong
• Sensitivity to
• LO/NLO
• Twist 3
• Exact GPD param.
• gt In the direct line of
• HERA studies with
• better constraints on
• quarks GPDs as the
• impact of Hg 0 here!

integrated over t
39
Implications for COMPASS (GPDs) (II)
In Fixed Target kin. CA is increasing with
x and quite stable in Q² (previous slide) gt
Better sensitivity at x0.1 or 0.2
At HERA kin. CA is decreasing with x! This
comes from the effect of the Re(Gluon
Amplitude) which enters into the  game  and is
larger at lower x
40
Implications for COMPASS (GPDs) (III)

• Reminder
• SSA Im(Amplitude)
• GPDs(?,?)
• We observe a strong
• Sensitivity to
• LO/NLO
• Exact GPD param.
• Again, important
• measurement to improve
• the modeling of GPDs

integrated over t
41
Conclusion

• Measurement of a nb cross section
• A. We need to cover a lot of large stat. analysis
• With E.Sauvan, we have provided directly 6 /
  16 H1 prel. for ICHEP
• B. Even then, refinements are not ended!
• gt Interesting adventure for an analyst
• What for ? (physics point of view)
• We are measuring a new function ? PDF!
• It is not debatable that we have established the
  impact of GPDs and
• shown the sensitivity of the present
  measurements, with the
• second round to be done at COMPASS ( JLab)!