1. Techniques for fitting di-muon spectra in d-Au collisions at  s NN = 200 GeV - First observation of the  ’ at RHIC (?) David Silvermyr, ORNL for the PHENIX collaboration Quark Matter ’04 Oakland, January 2004

2. QM 2004D. Silvermyr, PHENIX2 Outline Motivation - why interest in quarkonia ? PHENIX Experiment intro Dimuon spectra –Combinatorial background subtraction –Fit functions –Correlated non-quarkonia contributions –Line-shape fitting Summary and Outlook

3. QM 2004D. Silvermyr, PHENIX3 Motivation Interest in quarkonia Heavy quark production, and in particular J/ , is one of the more prolific QGP probes. The original idea was that, due to screening in the dense matter, J/  would be less likely to form, resulting in a clear suppression signal. However, there are several competing effects that are involved in charmonium production in dense matter. To resolve these ambiguities, species and energy scans, as well as studying additional observables, is needed. Interest in  ’ One such additional/interesting helpful probe is the  ’(2S) at 3.7 GeV/c 2, i.e. a state close to the J/  (1S) in the charmonium family. Of the suggested observables, I’ll just mention looking at the ratio between  ’ and J/  as a function of centrality, and see if they are suppressed similarly/start melting together/ or not. See e.g. ref. nucl-th/0303030 for more info, and PRL 84 (2000) 3256 [FNAL E866], PLB 553 (2003) 67 [NA50] for studies made with fixed target p-A collisions.. This presentation deals with the more modest topic: if we are able to observe the  ’ at all, in the latest RHIC run (2003, dAu), and what kind of uncertainties are involved in such an analysis.

4. QM 2004D. Silvermyr, PHENIX4 Fixed-target Data : FNAL These fixed-target p-A experiments had of the order of a 10 4  ’ per studied target (A).. Rather similar A-scaling (  values) and x F dependence was observed for  ’ and J/ .

5. QM 2004D. Silvermyr, PHENIX5 Fixed-target Data : SPS NA50 fixed-target also studied p-A interactions, as well as A-A. A rather constant ratio of  ’ over J/  of ~1.6 % was found in p-A. Note, however, that the “extra” suppression in A-A interactions do seem different for the two states! NA50: p-A A-A

6. QM 2004D. Silvermyr, PHENIX6 PHENIX Two forward muon spectrometers Tracking, momentum measurement with cathode strip chambers  ID with penetration depth / momentum match Two central electron/photon/hadron spectrometers: Tracking, momentum measurement with drift chamber, pixel pad chambers e ID with E/p ratio in EmCAL + good ring in RICH counter. Two sets of forward- rapidity detectors for event characterization Beam-beam counters measure particle production in 3.0<|  |<3.9. Luminosity monitor + vertex determination. Zero-degree calorimeters measure forward-going neutrons. Correlation gives centrality

7. QM 2004D. Silvermyr, PHENIX7 Reconstructing Muons Some of the general muon analysis requirements are: Likely muon candidate: Depth in Muon Identifier (MUID): at least to gap 2 [halfway through MUID]. Matching between MUID roads and MUTR hits (at station 3) within 20 cm. Track quality: Reduced chi2 for each track < 20 Pair quality: Both muons into the same arm – otherwise there is very high mass tail. Agreement of bend plane two-track closest approach (vertex) to event vertex within 25 cm. Throughout this poster the North arm triggered dAu data sample from 2003 is used as an example. No acceptance and efficiency corrections are made anywhere.

8. QM 2004D. Silvermyr, PHENIX8 Raw Data The basic problem is just to extract the charmonium signal from unlike-sign (+-) dimuon pairs, when comparing with a combinatorial background estimate of like-sign (++ and --) dimuons. We can thus divide this up in two parts: I) Estimating the combinatorial background contribution II) Having a good fit function for the needed signal components  ’ (at 3.7 GeV)?

9. QM 2004D. Silvermyr, PHENIX9 I) Estimating the comb. background The classic approach to estimate the combinatorial background part of the unlike-sign combinations N +-, when N +-, N ++ and N –- have the same acceptance, is 2*sqrt (N ++ * N -- ) This is strictly only valid when + and – have the same p T distributions; a reasonable start approximation. One can do this bkg estimate either with bin-by-bin counting or using fit functions (which works better at low statistics). For the rest of this poster, bin-by-bin counting is used. With the present statistics we have assigned a 5% syst. error for the number of J/  due to uncertainties in this subtraction. Since the  ’ over J/  ratio is expected to be of the order of a few %, this systematic obviously needs to be reduced significantly before a real  ’ result can be obtained.. Another approach is to use an event mixing technique (presently under investigation).

10. QM 2004D. Silvermyr, PHENIX10 II) Fit Function Components J/  and  ’ - Gaussians, masses in agreement with PDG, widths in agreement with experimental resolution. Drell-Yan (DY) and Open Charm contributions (DD) - Single exponentials, with slopes from the data. - Not always included; varying number of included exponentials from zero (none) to two (both).. The masses, widths, and slopes are free parameters in the fits, with the exception that the J/  and  ’ widths are constrained to be the same, and the scale between their masses is constrained to be identical with PDG values.

11. QM 2004D. Silvermyr, PHENIX11  and  ’ Plot shows comb. background-subtracted spectra. Fit includes the extra Gaussian for the  ’ also. No extra background component is included, nor DY or DD contributions.

12. QM 2004D. Silvermyr, PHENIX12 Simulations We generate Drell-Yan and Open-Charm spectra using PYTHIA. We then run these (and J/ ,  ’ simulations too) through the PHENIX GEANT code and through the full offline reconstruction, to get all resolution and efficiency effects accounted for correctly.

13. QM 2004D. Silvermyr, PHENIX13 Including DD and DY One can also include contributions from open charm (DD) and Drell- Yan (parameterization based on simulations as initial guesses), with slopes fitted from the tails of the signal spectra. Note that the free-fitted slopes do not quite match our expectations.. (DD) (DY)

14. QM 2004D. Silvermyr, PHENIX14 Single Exponential If one, instead of using exponentials for each of open charm (DD) and Drell-Yan, just uses a single exponential, the following result is obtained. Slope value lies between the two on the previous slide. Both the J/  and  ’ counts are reduced.

15. QM 2004D. Silvermyr, PHENIX15 Line-shape fitting An alternative approach to the free fits is to use a method that fits the full data and mc histograms (i.e. whole simulation curves, not parameterizations) simultaneously. The whole available mass range is also used in the fits. For the plot below, the like-sign and unlike-sign histograms are not scaled, but the mc histograms are allowed to be arbitrarily scaled individually and have their sum account for the difference between unlike- and like-sign.

16. QM 2004D. Silvermyr, PHENIX16 Result Comparisons Method# J/  #  ’  ’/ J/   2 Gaussians960 +- 32 53 +- 11 5.5 +- 1.2 2 Gaussians + Exp 766 +- 31 35 +- 94.6 +- 1.2 2 Gaussians + 2 Exp 828 +- 32 50 +- 9 6.0 +- 1.1 Line-shape Fitting 784 +- 5 25 +- 33.2 +- 0.4 As a ‘fit-stability’ check, we compare the results obtained with the different methods, summarized in the table below. The quoted errors are just the statistical errors reported from the fits. The systematic errors are not estimated but are believed to be large.. Thus, improvements both regarding the combinatorial background subtraction and fit methods, and more simulation studies, are needed before any real results can be obtained! Caution!! When using the alternative fitted comb. background subtracting technique, ratio values of about 2% were obtained instead of around 5% as shown here!!

17. QM 2004D. Silvermyr, PHENIX17 Summary and Outlook Accounting for combinatorial background –Standard unlike sqrt method does rather well - default –Other methods also used as an add’l handle on systematics Signal extraction –Gaussians for  and  ’ –Exponentials for add’l correlated contributions: DD and DY. Other functions used for comparisons. Line-shape fitting –Promising method. Possible improvements to come regarding stability for modified fit ranges. Also, systematic errors need to be estimated. Result so far –Ratio  ’/  on the few % level consistent with expectations, but exact values very dependent on e.g. comb. bkg subtraction method. More studies needed before a possible physics result. Look forward to future runs with higher luminosity where  ’ studies should be more rewarding!

18. QM 2004D. Silvermyr, PHENIX18 PHENIX Collaboration

19. QM 2004D. Silvermyr, PHENIX19 Acknowledgements [..not planned to include on actual poster..] Many individuals have worked on various aspects of fitting PHENIX Run-3 muon data, in particular: F. Fleuret, S. Gadrat, G. Roche, R. de Granier Cassagnac, J. Gosset, J. Burward-Hoy, M. Leitch, J. Nagle, S. Kelly, C. Zhang - Thanks everyone!