1. PHENIX Heavy Flavor Measurement by Single Leptons in p+p, d+Au and Au+Au Fukutaro Kajihara CNS, University of Tokyo For the PHENIX Collaboration Heavy Flavor Productions & Hot/Dense Quark Matter RIKEN BNL Research Center Workshop December 12, 2005 at Brookhaven National Laboratory

2. 2/20 Outlines Physics motivation Heavy quark measurement in the PHENIX –Analysis techniques of signal extraction Summary of heavy quark measurement in PHENIX Recent Results –p+p and d+Au –Total Charm Yield in Au+Au –Energy loss of charm quark –Charm flow Summary and outlook

3. 3/20 Physics Motivation Why do we measure heavy quarks (charm/bottom)? In p+p collisions: Important test of pQCD. Can pQCD predict charm production? Base line analysis for d+Au and Au+Au In d+Au collisions: Study of “cold” nuclear matter effect (Cronin effect, shadowing, G(x)) In Au+Au collisions: Medium modification effects (energy loss, collective flow) Important baseline of J/  analysis (recombination model)

4. 4/20 Heavy quark measurement in the PHENIX Central Arm (East) Muon measurement: 2 Forward Arms (South/North) 1.2<|  |<2.4,  98% of produced hadrons can be cut by absorber. Muon Arm (North) PHENIX Detector Complex Central Magnet Electron measurement: 2 Central Arms (East/West) |  |<0.35,  =  arms Radiation length < 0.4%X 0  rejection > 10 4 (AuAu) Measure the leptons from semileptonic decay of heavy quark

5. 5/20 Analysis Techniques of signal extraction Heavy flavor electron We have 2 kinds of subtraction method of backgrounds.  conversion and   Dalitz decays are main source of background electrons –Cocktail subtraction Simulate all background electrons –Converter subtraction Extract background electrons by additional photon converter (1.7%X 0 ) Prompt muon Decay muon (from charged K/  ) are main background. –They can be subtract by vertex dependent analysis. There is small Hadron background. –We can estimate the amount by simulation and subtract. Brass converter around beam pipe

6. 6/20 Summary of Heavy Quark Measurement in PHENIX (1) Run1130 GeV Au+AuElectron Cocktail Published PRL88 192303 The first open charm measurement Run2200 GeV Au+AuElectron Converter Published PRL94 082301 Total  cc is scaling with binary collision 200 GeV Au+AuElectron Cocktail PRL accepted nucl-ex/0510047 Energy loss of charm was discovered 200 GeV Au+AuElectron Flow Published PRC72 024901 Electron flow was discovered 200 GeV p+pElectron Cocktail PRL accepted hep-ex/0508034 Base line. Comparison with pQCD 200 GeV p+pMuonPreliminary QM05 We have already measured the following data.

7. 7/20 Summary of Heavy Quark Measurement in PHENIX (2) Run3200 GeV d+AuElectron, Cocktail/ConverterPreliminary QM04 200 GeV d+AuMuonPreliminary QM05 200 GeV p+pElectron, Cocktail/ConverterPreliminary QM05 Run4200 GeV Au+AuElectron, CocktailPreliminary QM05 200 GeV Au+AuElectron, ConverterComing soon 200 GeV Au+AuElectron FlowPreliminary QM05 62.4 GeV Au+AuElectron ConverterPreliminary ICQGP Run5200 GeV Cu+CuElectron Converter/CocktailComing soon 200 GeV p+pElectron Converter/CocktailComing soon 200 GeV p+pMuonComing soon We are preparing to publish the following new data.

8. 8/20 Result from p+p and d+Au

9. 9/20 Heavy flavor electron in Run2 p+p at  s NN = 200 GeV PRL accepted recently. (hep-ex/0508034). Data spectrum of Run 2 p+p seems to be the same or more harder than FONLL prediction. This data give very important base to study medium effects in heavy quark production. We need to analyze data of Run3 and Run5 p+p for the extension in higher p T region.

10. 10/20 Prompt  in Run2 p+p at  s NN = 200 GeV PHENIX Preliminary Preliminary We obtained the same result in the rapidity region (1.2<|  |<2.4). It seems that there is little rapidity dependence of heavy quark production in p+p. Red: Run2 heavy flavor electron Blue: Run2 prompt muon

11. 11/20 Heavy flavor electron in Run3 d+Au at  s NN = 200 GeV The Run3 d+Au data are shown as invariant cross section per binary collision cross section for the MinBias d+Au collision. The Run3 d+Au data are shown as invariant cross section per binary collision cross section for the MinBias d+Au collision. In low p T region, Run3 data almost agrees with Run3 p+p data within error bars. Minimum Bias

12. 12/20 Prompt  in Run3 d+Au at  s NN = 200 GeV Suppression(?) in d going direction Enhancement(?) in Au going direction d Au South North Beam direction

13. 13/20 Results from Au+Au Total charm yield Energy loss of charm Charm elliptic flow

14. 14/20 Mass dependence of the energy loss  Recent theories propose energy loss of charm quark is similar to light quarks. (Armesto et al, PRD 71, 054027, 2005 ; M. Djordjevic et al., PRL 94, 112301, 2005.)  2001, proposed “dead cone” effect (phase space limitation reduces gluon emission) suggests smaller energy loss of charm PHENIX measured large   and   suppression in Au+Au collisions at  s NN = 200 GeV. R AA (   ) ~ 0.2! Light quarks lose their energy by gluon emission in QCD matter mainly. The next interest is “how about heavy quark ?”.

15. 15/20 Total Charm Yield in Run2 Au+Au at  s NN = 200 GeV Binary scaling works well for total charm yield dN e /dy is fit to AN coll   = 0.938+/-0.075+/-0.0018 Coming soon: High statistic data in Run4 Au+Au S.S. Adler, et al., PRL 94 082301

16. 16/20 R AA of heavy flavor electron in Run2 Au+Au at  s NN = 200 GeV Theory curves (1abc) from N. Armesto, et al., PRD 71, 054027 (2ab) from M. Djordjevic, M. Gyullasy, S.Wicks, PRL 94, 112301 PRL accepted recently (nucl-ex/0510047). (1c) q_hat = 14 GeV 2 /fm (1b) q_hat = 4 GeV 2 /fm (1a) q_hat = 0 GeV 2 /fm (2a) dN g / dy = 1000 (2b) dN g / dy = 3500 Clear evidence for strong medium effects!

17. 17/20 Heavy flavor electron in Run4 Au+Au at  s NN = 200 GeV High statistic data come from Run4 Au+Au Data for each centrality is compared with Run3 p+p data (black line). Suppression depends on centrality. R AA is shown in the next page. We are preparing the high p T spectrum (up to p T = 10 GeV/c).

18. 18/20 R AA of heavy flavor electron in Run4 Au+Au at  s NN = 200 GeV We can see strong suppression even for heavy quark (charm). The matter is so dense that even heavy quarks are stopped. The data provides a strong constraint on the energy loss models. We are preparing the high p T R AA (up to p T = 10 GeV/c). (3) q_hat = 14 GeV 2 /fm (2) q_hat = 4 GeV 2 /fm (1) q_hat = 0 GeV 2 /fm (4) dN g / dy = 1000 Theory curves (1-3) from N. Armesto, et al., PRD 71, 054027 (4) from M. Djordjevic, M. Gyullasy, S.Wicks, PRL 94, 112301

19. 19/20 Open Charm Flow in Run4 Au+Au at  s NN = 200 GeV v 2 (D) = v 2 (  ) v 2 (D) = 0.6 v 2 (  ) v 2 (D) = 0.3 v 2 (  ) Theory: Greco, Ko, Rapp: PLB 595 (2004) 202 Significant anisotropy is observed for heavy flavor electron. v2 has good agreement of charm flow assumption below p T < 2.0 GeV/c In high p T region (p T > 2 GeV/c), v2 is reduced. (b quark contribution?) If v2(D) = a*v2(  ), v2(D) is non-zero value and 60% of v2(  ).

20. 20/20 Summary and Outlook Summary For p+p collisions: Heavy flavor electron in mid-rapidity and prompt muon in forward-rapidity have a similar spectrum. FONLL calculation are smaller than the each measured spectrum. For d+Au collisions: Binary scaling of heavy flavor electron works well in mid-rapidity in low p T region. For Au+Au collisions: Binary scaling of total charm yield works well. Nuclear modification factor R AA shows a strong suppression at high p T region. v2 of heavy flavor electron was found. If v2(D) = a*v2(  ), a=60%. Outlook High p T electron study (to p T =10 GeV/c) for p+p/Au+Au at  s NN =200 GeV. Prompt muon spectrum in Au+Au at  s NN =200 GeV. High statistic Cu+Cu analysis.

21. 21/20 Back Up Slide

22. 22/20 Electron Signal and Background [Photonic electron] … Background Conversion of photons in material Main photon source:    →  In material:  → e + e - (Major contribution of photonic) Dalitz decay of light neutral mesons    →  e + e - (Large contribution of photonic) The other Dalitz decays are small contributions. Direct Photon (is estimated as very small contribution) [Non-photonic electron] … Signal and minor background. Heavy flavor decays (the most of all non-photonic) Weak Kaon decays K e3 : K ± →   e ± e ( 1.0 GeV/c) Vector Meson Decays  J  → e + e -  (< 2-3% of non-photonic in all p T. ) Thermal electron (is estimated as very small contribution) PHENIX has 2 photonic e subtraction methods: Cocktail/Converter method.

23. 23/20 Heavy flavor electron in Run3 p+p at  s NN = 200 GeV/c Data spectrum of Run 3 p+p seems to be more harder than PYHIA prediction. This data give very important base to check nuclear and density modification. We need to analyze data of Run3 and Run5 p+p for the extension in higher p T region.

24. 24/20 Photonic Subtraction-Cocktail Method Sources of photonic electron are known well. Most sources of photonic electron were measured in PHENIX. Decay kinematics and photon conversions can be reconstructed by detector simulation. Then, subtract “cocktail” of the background electrons from the inclusive spectrum Advantage is small statistical error. The method is applied for Run3 p+p (d+Au) and Run4 Au+Au data.

25. 25/20 Photon Converter (Brass: 1.7% X 0 ) We can determine photonic electron yield by the radiation length (X 0 ) of material amount. We know precise X 0 of each detector material, but don’t the total effective value (+ air etc.). However, we can measure the photonic electron yield by inserting of converter. Then, the photonic electron is subtracted from inclusive. Advantage is small systematic error even in low p T region. The method is used for Run3 analysis. Photonic Subtraction-Converter Method N e Inclusive electron yield Material amounts:  0 1.1% 1.7% Dalitz : 0.8% X 0 equivalent radiation length 0 With converter Conversion in converter W/O converter 0.8% Non-photonic Conversion from known material ? % Photonic

26. 26/20 Heavy flavor electron V 2 Photonic v2: –p T < 1.0 GeV/c – converter subtraction method (open symbols) –p T > 1.0 GeV/c – calculation from measured v 2 (  ) (black curve) Non-photonic v2: Inclusive v2 – Photonic v2 * R PI Small conversion material (<0.4%) and a control measurement with converter are essential for the extraction Inclusive e v 2 Photonic e v 2 Signal electron All B.G.

27. 27/20 Electron Trigger in the PHENIX EMCal - RICH Trigger (ERT) EMCal - RICH Trigger (ERT) Trigger threshold ; Trigger threshold ; EMCal Deposit energy > 600 or 800 MeV. RICH Cherenkov photon > 3 p.e. For Trigger rate reduction; Minimum bias event selection (Beam-Beam Counter (BBC) Trigger) is used. Geometrical hit matching; RICH4x4 tiles & PbGl 4x4 tiles RICH4x4 tiles & PbSc 3x3 tiles Trigger Efficiency ~ 70 - 80 %

28. 28/20 Systematic Error Estimation Systematic uncertainties in Rsim ■ Systematic uncertainties in Rsim Conversion itself; 3% Conversion itself; 3% Acceptance difference between converter/non-converter run; less than 4%. Acceptance difference between converter/non-converter run; less than 4%. eID efficiency; 2%. eID efficiency; 2%. ■ Systematic uncertainties in Acceptance && eID Acceptance; 10% (acceptance agreement between the data and simulation) Acceptance; 10% (acceptance agreement between the data and simulation) Momentum smearing correction was not applied; 10% Momentum smearing correction was not applied; 10% eID ; 2 %. eID ; 2 %. ■ Systematic uncertainties in ERT Electron Trigger Efficiency for ERT data. The difference between simulation and real data; 15 %. The difference between simulation and real data; 15 %. ■ Vector Meson decay contribution is very small and ignored. ■ Kaon decay contribution is a few and ignored. The total systematic error (<19%) is determined by adding the errors from Rsim and the errors in the acceptance and the eID efficiency (the ERT trigger efficiency) in squares.

29. 29/20 R  (the ratio of photonic electrons in converter/non-converter run) We know material thickness (as radiation length: X 0 ) in the PHENIX. It is a very good approximation that pT spectrum shape of conversion electrons are essentially identical to the spectrum shape of electrons from   Dalitz decays. It is a very good approximation that pT spectrum shape of conversion electrons are essentially identical to the spectrum shape of electrons from   Dalitz decays. Empirical fitting curve for Rsim 1.68 + 0.19 tanh (0.79 pT) R  ~ 1.87 at high pT limit Statistical error only We can estimate Rsim based on the GEANT simulation with the PHENIX geometrical environment. Comparison with Real data; Rcn Comparison with Real data; Rcn – Rcn: the ratio of electron yield in converter run to non-converter run. – Rcn should be close to Rsim in the low pT limit. – The difference between Rsim and Rcn indicates non-photonic source in data. – Rcn should be 1.0 at high pT limit because only non-photonics are included.

30. 30/20 Non-Photonic Electron Component The right figure shows Non-photonic electron yield. In this calculation, Minimum Bias data is used in Pt 1.2 GeV/c, Electron triggered data is used (Red points). Red points shows QM04 result (PHENIX Preliminary). New points are well consistent with Red points. Red points shows QM04 result (PHENIX Preliminary). New points are well consistent with Red points. The right figure shows non-photonic electron component. The right figure shows non-photonic electron component. PHENIX Work in progress Statistical error only Ed 3 N/dp 3 [GeV -2 c 2 ] MBERT (N(e + ) + N(e - ))/2 pT [GeV/c] PHENIX preliminary With Full Statistics of Run3 d-Au data Unit: [GeV -2 c 3 ].  acc : Acceptance,  eff : Efficiency. Here, N(Pt) = (R(Pt)*A(Pt) – C(Pt))/(Rsim(Pt) - 1).

31. 31/20 Photonic Electron Component The right figure shows Photonic electron yield. In this calculation, Minimum Bias triggered data (Black points) and electron triggered data (Red points) are used. These new results are compared with the photonic cocktail calculation with the PHENIX simulation framework. The black curve is the total of photonic components. These new results are compared with the photonic cocktail calculation with the PHENIX simulation framework. The black curve is the total of photonic components. The right figure shows photonic electron. PHENIX work in progress Statistical error only pT [GeV/c] Ed 3 N/dp 3 [GeV -2 c 2 ] (N(e + ) + N(e - ))/2 MBERT With Full Statistics of Run3 d-Au data Unit: [GeV -2 c 3 ]  acc : Acceptance,  eff : Efficiency Here, P(Pt) = (C(Pt) – A(Pt))/(Rsim(Pt) - 1)

32. 32/20 N MVD : the number of pair in 60-100 [MeV]. (MVD + Photon Converter) N BP : the number of pair in 0-40 [MeV]. (Dalitz + Beam pipe) Error of double ratio R  (pair:real data)/R  (pair:simulation) as the R  systematic error. Result: R  (pair:real data)/R  (pair:simulation) ~ 3 % R  Evaluation with Mass of e+e- Conversion Pairs Simulation Real (MB) With converter Without converter With converter Without converter Dalitz Beam Pipe Dalitz Beam Pipe MVD (Converter) Combinatorial background is subtracted

33. 33/20 Photonic Electron Simulation Statistical error only (N(e + ) + N(e - ))/2 MBERT Cocktail Not only   but also   was simulated to reproduce photonic electrons and calculate R  (total 400M events were generated with or without converter)  R  (  ) and R  (  ) are combined. R  (  ) and R  (  ) are combined. Systematic error derived from 10 % error of  ratio is only 0.5 %. Systematic error derived from 10 % error of  ratio is only 0.5 %. Photonic electrons from other sources are negligible. Photonic electrons from other sources are negligible. Total photonic electron yield well agrees with the result from cocktail simulation. Total photonic electron yield well agrees with the result from cocktail simulation. Ed 3 N/dp 3 [GeV -2 c 2 ]

34. 34/20 Photonic and Non-photonic electrons –Photonic electron sources (background): Photon conversion, Dalitz decay (    ’  etc.), –Non-photonic electron sources: A few % of Kaon, vector meson decays Charm and Beauty decay (Signal) Converter subtraction method –A(Pt) : The inclusive electron yield in the non-converter run. –P(Pt) : The yield of photonic component –N(Pt) : The yield of non-photonic component –C(Pt) : The inclusive electron yield in the converter run. –R  (Pt) : The ratio of the photonic electron yield in the converter run to non-converter run. From these variables, –A(Pt) = P(Pt) + N(Pt). –C(Pt) = R  (Pt)*P(Pt) + N(Pt). P(Pt) and N(Pt) are determined as follows; –P(Pt) = (C(Pt) – A(Pt))/(R  (Pt) - 1). –N(Pt) = (R(Pt)*A(Pt) – C(Pt))/(R  (Pt) - 1). Run3 d+Au Analysis R  is a very important value to extract non-photonic electron yield. Photon Converter (Brass: 1.68% X 0 )

35. 35/20 Result from Run3 d-Au at sqrt(s NN ) = 200 GeV Detector simulation of the photon conversion from   and   was updated. The Run3 d-Au data in the right figure are shown as invariant cross section per binary N-N collision cross section by scaling with  pp /(Ncoll = 8.5) for the MinBias d+Au collision. The Run3 d-Au data in the right figure are shown as invariant cross section per binary N-N collision cross section by scaling with  pp /(Ncoll = 8.5) for the MinBias d+Au collision. Run3 data almost agrees with both Run2 or Run3 p-p data within error bars and seems to indicate no strong modification. Run3 data almost agrees with both Run2 or Run3 p-p data within error bars and seems to indicate no strong modification. Run2 p+p data Systematic error Scaled Run3 d+Au data Work in progress PHENIX Preliminary Min. Bias data ERT data Without Systematic Errors