1. John Womersley Direct Photons John Womersley Fermilab CTEQ Summer School, Madison June 2002 Mehr licht!

2. John Womersley Hadron-hadron collisions Complicated by –parton distributions — a hadron collider is really a broad-band quark and gluon collider –both the initial and final states can be colored and can radiate gluons –underlying event from proton remnants fragmentation parton distribution parton distribution Jet Underlying event Photon, W, Z etc. Hard scattering ISR FSR

3. John Womersley Motivation for photon measurements As long as 20 years ago, direct photon measurements were promoted as a way to: –Avoid all the systematics associated with jet identification and measurement photons are simple, well measured EM objects emerge directly from the hard scattering without fragmentation –Hoped-for sensitivity to the gluon content of the nucleon “QCD Compton process” 

4. John Womersley In the meantime... Jet measurements have become much better understood Lower photon cross sections and ease of triggering on EM objects lead to photon data being at much lower E T than typical jet measurements –Turn out to be susceptible to QCD effects at the few GeV level that Photons have not been a simple test of QCD and have not given input to parton distributions, and they continue to challenge our ability to calculate within QCD

5. John Womersley Photon Signatures of New Physics Important to understand QCD of photon production in order to reliably search for –Higgs H   is a discovery channel at LHC –Gauge mediated SUSY breaking  0   G, photon + ME T signatures –Technicolor Photon + dijet signatures Diphoton resonances –Extra dimensions Enhancement of  pp   at high masses (virtual gravitons)

6. John Womersley Photon identification Essentially every jet contains one or more  0 mesons which decay to photons –therefore the truly inclusive photon cross section would be huge –we are really interested in direct (prompt) photons (from the hard scattering) –but what we usually have to settle for is isolated photons (a reasonable approximation) isolation: require less than e.g. 2 GeV within e.g.  R = 0.4 cone This rejects most of the jet background, but leaves those (very rare) cases where a single  0 or  meson carries most of the jet’s energy This happens perhaps 10 –3 of the time, but since the jet cross section is 10 3 times larger than the isolated photon cross section, we are still left with a signal to background of order 1:1.

7. John Womersley Event topology Simplest process:  pp   + jet –Photon and jet are back-to-back in  and balance in E T Experimentally we find that at about one third of the photon events have a second jet of significant E T –Higher order QCD processes jet  Back to back in parton-parton center of mass jet  boosted into lab frame

8. John Womersley Photon candidate event in DØ Run 1 Photon Recoil Jet

9. John Womersley Triggering The greatest engineering challenge in hadron collider physics To access rare processes, we must collide the beams at luminosities such that there is a hard collision every bunch crossing –396 ns in Run 2 = 2.5 MHz We cannot write to tape (or hope to process offline) more than about 50 events per second –Trigger rejection of 50,000 required in real time with minimal deadtime and high efficiency for physics of interest

10. John Womersley Photon Triggers Example of how this works in DØ: Level 1 (hardware trigger) –Requires E T > threshold in one trigger tower of the EM calorimeter (    = 0.2  0.2) –Total accept rate ~ 10 khZ; can allow ~ 1 kHz for electron and photon triggers Level 2 (Alpha CPU, processing the trigger tower information) –Requires EM fraction cut and isolation cuts –Rejection ~ 10 Level 3 (Linux farm, processing the full event readout) –Clusters    = 0.1  0.1 cells with better resolution –Applies shower shape and isolation cuts –Rejection ~ 20

11. John Womersley Thresholds and prescales Relatively high cross section processes like photons, with steeply falling cross sections, will be accumulated using a variety of thresholds with different prescales A very simple example: –EM cluster > 5 GeVaccept 1 in 1000 –EM cluster > 10 GeV accept 1 in 50 –EM cluster > 30 GeV accept all Then “paste” the cross section together offline: ETET # events 5 10 30  1000  50  1 ETET Cross section 5 10 30

12. John Womersley Photon candidates: isolated electromagnetic showers in the calorimeter, with no charged tracks pointed at them –what fraction of these are true photons? Signal Background Signal and Background Experimental techniques in Run 1 DØ measured longitudinal shower development at start of shower CDF measured transverse profile at start of shower (preshower detector) and at shower maximum    00 Preshower detector Shower maximum detector

13. John Womersley Photon purity estimators CDF DØ Each E T bin fitted as sum of: = photons = background w/o tracks = background w/ tracks

14. John Womersley Photon sample purity CDF DØ

15. John Womersley Angular distributions The dominant process producing photons Should be quite different from dijet production: Can we test this?

16. John Womersley Transformation to photon-jet system Lab pseudorapidity of photon Lab pseudorapidity of jet  * = CM pseudorapidity  BOOST of CM relative to lab Central calorimeter coverage jet    BOOST ** cos  * = tanh  *

17. John Womersley cos  * = tanh  * CM pseudorapidity  * Photon p T Lines of minimum and maximum p* p* = p T cosh  * Use multiple regions to maximize statistics; paste distribution together using overlapping coverage Want uniform coverage in CM variables while respecting physical limits on detector  coverage and trigger p T min p T from trigger  min p*

18. John Womersley Angular distributions

19. John Womersley Photons as a probe of quark charge Inclusive heavy flavor production “sees” the quark color charge: While photons “see” the electric charge: Charm (+2/3) should be enhanced relative to bottom (-1/3)

20. John Womersley CDF photon + heavy flavor Use muon decays; p T of muon relative to jet allows b and c separation Charm/bottom = 2.4  1.2 Cf. 2.9 (PYTHIA) 3.2 (NLO QCD)

21. John Womersley Control sample using same dataset –identify  0 (= jet) instead of photon: gg  QQ events Charm/bottom ~ 0.4

22. John Womersley An idea for the future Use  tt  events to measure the electric charge of the top quark –How do we know it’s not 4/3? Baur et al., hep-ph/0106341

23. John Womersley Photon cross sections at 1.8 TeV DØ, PRL 84 (2000) 2786 CDF, submitted to Phys. Rev. D QCD prediction is NLO by Owens et al.

24. John Womersley (data – theory) / theory DØ, PRL 84 (2000) 2786 QCD prediction is NLO by Owens et al., CTEQ4M What’s going on at low E T ? CDF, submitted to Phys. Rev. D ±12% normalization statistical errors only

25. John Womersley “k T smearing” Gaussian smearing of the transverse momenta by a few GeV can model the rise of cross section at low E T (hep-ph/9808467) 3 GeV of Gaussian smearing PYTHIA style parton shower (Baer and Reno) Account for soft gluon emission CDF data  1.25

26. John Womersley Why would you need to do this? NLO calculation puts in at most one extra gluon emission  In PYTHIA, find that additional gluons add an extra 2.5–5 GeV of p T to the system 10 GeV  2.6 GeV “k T ” 50 GeV  5 GeV “k T ”

27. John Womersley Fixed target photon production Even larger deviations from QCD observed in fixed target (E706) again, Gaussian smearing (~1.2 GeV here) can account for the data

28. John Womersley Photons at HERA ZEUS data agrees well with NLO QCD –no need for k T ? ZEUS 96-97 Have to include this “resolved” component

29. John Womersley ZEUS measurement of photon-jet p T

30. John Womersley A consistent picture of k T W = invariant mass of photon + jet final state

31. John Womersley Is this the only explanation? Not necessarily... Vogelsang et al. have investigated “tweaking” the renormalization, factorization and fragmentation scales separately, and can generate shape differences This is not theoretically particularly attractive

32. John Womersley Contrary viewpoints Aurenche et al., hep- ph/9811382: NLO QCD (sans k T ) can fit all the data with the sole exception of E706 “It does not appear very instructive to hide this problem by introducing an extra parameter fitted to the data at each energy” E706 Ouch!

33. John Womersley Isolated  0 cross sections Proponents of k T point out that  0 measurements back up the k T hypothesis (plots from Marek Zielinski) –WA70  0 data require k T to agree with QCD (unlike WA70 photons) –  /  0 ratio in E706 agrees with theory, in WA70 does not Aurenche et al. claim the opposite (hep-ph/9910352) –all  0 data below 40 GeV compatible, unlike photon data (E706) – “seems to indicate that the systematic errors on prompt-photon production are probably underestimated”

34. John Womersley Aurenche et al. vs. E706

35. John Womersley Resummation Predictive power of Gaussian smearing is small –e.g. what happens at LHC? At forward rapidities? The “right way” to do this should be resummation of soft gluons –this works nicely for W/Z p T, at the cost of introducing parameters Catani et al. hep-ph/9903436 Threshold resummation Fixed Order Laenen, Sterman, Vogelsang, hep-ph/0002078 Threshold + recoil resummation: looks promising Threshold resummation: did not model E706 data very well

36. John Womersley Fink and Owens resummed calculations hep-ph/0105276 E 706 data DØ data Agreement with data is pretty good Does require 2 or 4 non-perturbative parameters to be set

37. John Womersley Photons at  s = 630 GeV At the end of Run 1, CDF and DØ both took data at lower CM energy Central region data are qualitatively in agreement and show a k T -like excess at low E T CDF DØDØ

38. John Womersley But... When the UA2 data (also at 630 GeV) is added, it reinforces the impression of a deficit at large x T What’s happening here? Can I really ignore the data normalization in making all these comparisons with k T ?

39. John Womersley Is it just the PDF? New PDF’s from Walter Giele can describe the observed photon cross section at the Tevatron without any k T, and predict the “deficit” CDF (central) DØ (forward) Blue = Giele/Keller sets Green = MRS99 set Orange = CTEQ5M and L Not all of Walter’s PDF sets have this feature: it depends on what data are input

40. John Womersley Anything similar in other final states? b cross section at CDF and at DØ Data continue to lie ~ 2  central band of theory b B centralforward Cross section vs. |y| p T > 5 GeV/c p T > 8 GeV/c

41. John Womersley DØ b-jet cross section at higher p T Differential cross section Integrated p T > p Tmin from varying the scale from 2μ O to μ O /2, where μ O = (p T 2 + m b 2 ) 1/2 New

42. John Womersley (data – theory)/theory

43. John Womersley b-jet and photon production compared DØ b-jets (using highest QCD prediction) 0 - 0.5 0.5 1.0 1.5 CDF photons  1.33 DØ photons Data – Theory/Theory Photon or b-jet p T (GeV/c)

44. John Womersley Diphoton production Rate is very small: few hundred events in Run I (p T > 12 GeV) But interesting because –final state kinematics can be completely reconstructed (mass, p T and opening angle of  system) –background to H   at LHC NLO calculations available

45. John Womersley DØ diphoton measurements Find that we need NLO QCD to model the data at large p T (small  ), but NLO calculation is divergent at p T = 0 (  =  ) Need a resummation approach (RESBOS) or showering Monte Carlo (PYTHIA) or ad hoc few-GeV k T smearing p T ~ 3 GeV  pTpT

46. John Womersley Latest NLO diphoton calculation Binoth, Guillet, Pilon and Werlen, hep-ph/0012191 Shoulder at 30 GeV in calculation is a real NLO effect (contribution opens up with both photons on same side of the event)

47. John Womersley Photons: final remarks For many years it was hoped that direct photon production could be used to pin down the gluon distribution through the dominant process: Theorist’s viewpoint (Giele): “... discrepancies between data and theory for a wide range of experiments have cast a dark spell on this once promising cross section … now drowning in a swamp of non-perturbative fixes” Experimenter’s viewpoint: it is an interesting puzzle, and we like solving interesting puzzles –data  NLO QCD –k T remains a controversial topic –experiments may not all be consistent –resummation looks quite good, but how predictive is it? –what is the role of the PDF’s? 

48. John Womersley Run 2 Missing E T + di-em Candidate EM1EM2 E T = 27.4 GeV  = 0.52  = 3.78 Loose match with a low-p T track E T = 26.0 GeV  = 1.54  = 5.86 No track match ME T = 34.3 GeV; M(diEM) = 53 GeV  +ME T is a signature of gauge-mediated SUSY-breaking