Analysis of the dielectron continuum in AuAu 200

Analysis of the dielectron continuum in Au+Au @ 200 Ge. V with PHENIX Alberica Toia for the PHENIX Collaboration • Physics Motivation • Analysis Strategy (765 M events) – Cuts • Single electron cuts • Electron pair cuts: remove hit sharing – Spectra: Foreground, Background (mix events), Subtracted – Efficiency / Acceptance • Cocktail & theory comparison 1

Physics Motivation: em probes e- e+ Freeze-out Ex p an si o n time Hadronization QGP Thermaliztion Hard Scattering space Au electro-magnetic radiation: g, e+e-, m+mrare, emitted “any time”; reach detector unperturbed by strong final state interaction Au 2

e+e- Pair Continuum at RHIC Expected sources • Light hadron decays – Dalitz decays p 0, h – Direct decays r/w and f • Hard processes – Charm (beauty) production – Important at high mass & high p. T – Much larger at RHIC than at the SPS • Cocktail of known sources – Measure p 0, h spectra & yields – Use known decay kinematics – Apply detector acceptance – Fold with expected resolution Possible modifications Chiral symmetry restoration continuum enhancement modification of vector mesons thermal radiation charm modification exotic bound states esuppression (enhancement) e+ R. Rapp nucl-th/0204003 3

Electron Identification PHENIX optimized for Electron ID • track + • Cherenkov light RICH + • shower EMCAL Charged particle tracking: DC, PC 1, PC 2, PC 3 and TEC Excellent mass resolution (1%) All charged tracks RICH cut Real Net signal p e- Background Energy-Momentum Pair cuts (to remove hit sharing) PC 3 e+ DC PC 1 magnetic field & tracking detectors 4

Combinatorial Background Which belongs to which? Combinatorial background g e+ eg e+ ep 0 g e+ e. PHENIX 2 arm spectrometer acceptance: d. Nlike/dm ≠ d. Nunlike/dm different shape need event mixing like/unlike differences preserved in event mixing Same normalization for like and unlike sign pairs RATIO -- --- Foreground: same evt --- Background: mixed evt BG fits to FG 0. 1% 5

Combinatorial Background • Different independent normalizations used to estimate sys error – Measured like sign yield – Event counting: Nevent / Nmixed events – Poisson assumption: N± = 2√N++N— – Track counting: ‹N±› = ‹N+›‹N-› • All the normalizations agree within 0. 5% Systematic uncertainty: 0. 25% e+ e po e+ e - --- Foreground: same evt --- Background: mixed evt 6

Photon conversion rejection g e+e- at r≠ 0 have m≠ 0 (artifact of PHENIX tracking) • effect low mass region • have to be removed Conversion removed with orientation angle of the pair in the magnetic field r ~ mee Photon conversion --- without conversion --- with conversion beampipe air Support structures Mass [Ge. V/c 2] 7

Subtracted spectrum BG normalized to Measured like sign yield Integral: 180, 000 above p 0: 15, 000 All the pairs Combinatorix Signal Green band: systematic uncertainty • Acceptance • Efficiency 8 • Run-by-run

Signal to Background • Very low signal to background ratio in the interesting region main systematic uncertainty ssignal/signal = s. BG/BG * BG/signal 0. 25% large!!! Yellow band: error on combinatorial background normalization Green band: other systematics 9

A closer look at resonances Agreement with other analyses phi A. Kozlov, K. Ozawa J/psi Upsilon? ? ? H. Pereira, T. Gunji 10

Cocktail comparison • Data and cocktail absolutely normalized • Cocktail from hadronic sources • Charm from PYTHIA Predictions are filtered in PHENIX acceptance • Good agreement in p 0 Dalitz • Continuum: hint for enhancement not significant within systematics • What happens to charm? • Single e pt suppression • angular correlation? ? ? • LARGE SYSTEMATICS!11

Data/cocktail Measurement [10 -5 counts/event] Predictions [10 -5 counts/event] 0. 15 -0. 7 Ge. V/c 2 17. 8 ± 3. 8 ± 1. 50 12. 3 1. 1 -2. 5 Ge. V/c 2 0. 67 ± 0. 50 ± 0. 11 1. 16 12

Comparison with theory • calculations for min bias • QGP thermal radiation included • Systematic error too large to distinguish predictions • Mainly due to S/B • Need to improve HBD R. Rapp, Phys. Lett. B 473 (2000) R. Rapp, Phys. Rev. C 63 (2001) 13 R. Rapp, nucl/th/0204003

A Hadron Blind Detector (HBD) for PHENIX signal electron Cherenkov blobs partner positron needed for rejection e- e+ q pair opening angle ~1 m S/B ~ 100 x • Irreducible charm background • S/B increased by factor 100 I. Ravinovich • Dalitz rejection via opening angle – Identify electrons in field free region – Veto signal electrons with partner HBD concept: – windowless CF 4 Cherenkov detector – 50 cm radiator length – Cs. I reflective photocathode – Triple GEM with pad readout 14 Construction/installation 2005/2006

Summary & Outlook • First dielectron continuum measurement at RHIC – Despite of low signal/BG – Thanks to high statistics – Good understanding of background normalization • Measurement consistent with cocktail predictions within the errors – Improvement of the systematic uncertainty • HBD upgrade will reduce background great improvement of systematic and statistical uncertainty “The most beautiful sea hasn't been crossed yet. And the most beautiful words I wanted to tell you I haven't said yet 15. . . “ (Nazim Hikmet)

Backup 16

Single electron cuts • Event cut: – zvertex <= 25 • Single electron cuts: – Pt: – Ecore – Match PC 3 & EMC • PC 3 (Phi+z) • EMC (Phi+z) = >= 150 Me. V – 20 Ge. V 150 Me. V < < 3 sigma – Dispmax < – N 0 min >= – dep >= showers NOT removed) – chi 2/npe 0 < – Quality = 5 (ring displacement) 3 tubes -2 sigma (overlapping 10 63, 51, 31 17

Pair cuts DC ghosts (like sign): fabs(dphi) < 0. 1 rad fabs(dz) < 1. 0 cm RICH ghosts (like and unlike sign) Post Field Opening Angle < 0. 988 --- Foreground: same evt --- Background: mixed evt like Cos(PFOA) 18

Systematic error • Systematic error of simulation – Acceptance difference between real/simulation is less than: 3%. – Single e e. ID efficiency difference between real/simulation is less than 8. 8%. • Dep < 1%, emcsdphie < 1%, emcsdze < 1%, n 0 < 7%, chi 2/npe 0 < 1%, disp < 5%. • Systematic error of real data – Stability of acceptance: 5% – Stability of e. ID efficiency: 5% • Other correction factor – Embedding efficiency < 10% (Run 2 7%). • Background Normalization 19

Acceptance filter • Decoupling acceptance – efficiency corrections • Define acceptance filter (from real data) • Correct only for efficiency IN the acceptance • “Correct” theory predictions IN the acceptance • Compare ACCEPTANCE FILTER charge/p. T q 0 Roughly parametrized from data f 0 z vertex 20

Efficiency 2 sets of simulations of dielectron pairs • White in mass (0 -4 Ge. V) • White in p. T (0 -4 Ge. V) • Vertex(-30, 30), rapidity (1 unit), phi (0, 2 p) • Linearly falling mass (0 -1 Ge. V) • Linearly falling p. T (0 -1 Ge. V) • Vertex(-30, 30), rapidity (1 unit), phi (0, 2 p) 2 D efficiency corrections: Mass vs p. T 21

Single e distribution: Poisson 0 -10% 10 -20% 20 -30% 30 -40% 22

Normalization of combinatorial background Same normalization used for like and unlike sign pairs 4 different (independent) normalizations: • Akiba : Nevent / Nmixed events • Hemmick: N± = 2√N++N— • Zajc: ‹N±› = ‹N+›‹N-› • Drees: 0. 5*( (Integral. FG++(0. 15 -4 Ge. V) / Integral. BG++(0. 15 -4 Ge. V)) + (Integral. FG-- (0. 15 -4 Ge. V) / Integral. BG-- (0. 15 -4 Ge. V)) ) Normalization factors [10 -2] • akiba: 8. 74 • hemmick: 8. 69 • zajc: 8. 71 • drees: 8. 70 All normalizations agree within 0. 5% upper limit: Integral in charm region=0 Normalization factor 8. 75 23

The unfiltered calculations • black: our standard cocktail • red : hadronic spectrum using the VACUUM rho spectral function • green: hadronic spectrum using the IN-MEDIUM rho spectral function • blue : hadronic spectrum using a rho spectral function with DROPPING MASS • magenta : QGP spectrum using the HTL-improved p. QCD rate 24