1. Accelerator neutrino interactions in the MINOS Experiment Krzysztof Wojciech Fornalski Warszawa 3.XII.2007

2. Krzysztof W. Fornalski2 Overview of the talk zMINOS experiment zneutrino interactions zcuts method of selection zRange Searching method zresults zsummary

3. Krzysztof W. Fornalski3 MINOS experiment

4. Krzysztof W. Fornalski4 First informations zMinos zMinos - mythical Greek king of Creta, son of Zeus and Europe zMINOS zMain Injector Neutrino Oscillation Search- a long-baseline neutrino oscillation experiment yNeutrinos from the 120 GeV pulsed proton beam on graphite target yTwo detectors (at FermiLab and Soudan)- USA yThe MINOS experiment measures the neutrino oscillation between the detectors zExperiment started in 2005 zFirst results at the beginning of 2006

5. Krzysztof W. Fornalski5 Two underground detectors - Near (ND) and Far Detector (FD) zND at FNAL, FD at Soudan Mine, 735 km from FNAL zND is to measure the beam composition and energy spectrum zFD is to search for evidence of oscillations between detectors zFD is ~2x bigger than ND Veto Shield Coil

6. Krzysztof W. Fornalski6 Detectors construction zNear & Far detectors are functionally identical zThey share the same basic detector technology and granularity: yIron/Scintillator tracking calorimeters y2.54cm thick magnetized steel planes = 1.2T y1cm thick scintillator planes  Alternate planes rotated by  90 o (U,V) zthanks to this rotation we can find the 3D representation of an event UVUVUVUV Steel Scintillator Orthogonal strips

7. Krzysztof W. Fornalski7 Neutrino interactions

8. Krzysztof W. Fornalski8 Types of interactions  Neutral Current (NC) - Z boson exchange  N ->  X (X - hadronic cascade)  Charged Current (CC) - W bosons exchange  N -> l  X (X - hadronic cascade, l - charged lepton)

9. Krzysztof W. Fornalski9 Neutral Current or Charged Current?  CC Event NC Event e CC Event UZ VZ We have  with long  track + hadronic shower short, with typical EM shower profile no  No electron track - short radiation length short event, often diffuse the neutrino is “deflected”! 3.5m 1.8m2.3m

10. Krzysztof W. Fornalski10 CC to NC separation zMINOS can measure oscillations based on muon neutrinos which interact as CC  It is necessary to separate CC  from NC events zthe biggest difficulty: low energy events Events (energy) histogram - based on Monte Carlo events

11. Krzysztof W. Fornalski11 Main goals zcreation a CC from NC Monte Carlo events separation zanalysis of two methods: cuts method and Range Searching zmodifications of the second method Simple examples of CC  (right) and NC (left) events

12. CC/NC Separation - cuts method

13. Krzysztof W. Fornalski13 Cuts method zSimple one-dimensional method of CC/NC separation zthree independent variables from STNP (Standard NTuple) files (ROOT) z~70 000 Monte Carlo events

14. Krzysztof W. Fornalski14 Cuts method - the variables zSignal from the first 25% active planes to biggest cascade signal znumber of active planes (length of the event) zsignal of the longest reconstructed track

15. Krzysztof W. Fornalski15 Cuts on variables zSimple cuts on all variables to separate CC from NC events zif an event is not clasified in one cut, it is taken by next variable and so on... an example of one cut the results of cuts method will be shown later

16. CC/NC Separation -Range Searching method

17. Krzysztof W. Fornalski17 Range Searching - the algorithm zmulti-dimensional method zcreation of the density CC/NC map based on the MonteCarlo model zone cut on the discriminant variable Overview:

18. Krzysztof W. Fornalski18 Range Searching -first step zcreate a model - binary tree contained 220 000 Monte Carlo CC&NC events zit is necessary to have many thousands of model events -> better density zeach event is described in 3D space by 3 variables - all coordinates are values of the variables zin this example there are the same three variables as in the cuts method

19. Krzysztof W. Fornalski19 Range Searching - simple 2D example zexample events in 2D variable space to separate signal to background (fig. a) zalgorithm tests the events one by one by comparing the V region around the event

20. Krzysztof W. Fornalski20 Range Searching - simple 2D example zV region (volume) around event (being analyzed) is estimated on the basis of the Monte Carlo model zV contains a certain number of signal (CC) and background (NC) MC events

21. Krzysztof W. Fornalski21 Range Searching - simple 2D example zthanks to number of CC and NC events in V it is possible to find the density D of the events (fig. b) zthis density determines if the event being analyzed is CC or NC zOne cut on D enable to separate CC from NC (fig. c)

22. Krzysztof W. Fornalski22 Range Searching - real MINOS variables zAfter creating a Monte Carlo model, the next step is to find the best volume of V zV should not contain too small or too large number of events -> worse resolution zbest V: 15% of the characteristic region of variable zSeparation power:

23. Krzysztof W. Fornalski23 Range Searching - discriminant variable D zby using RS algorithm we can find values of the D variable of all checking events zone cut can separate CC from NC events Events (D)

24. Krzysztof W. Fornalski24 Range Searching - modifications zevent length ; track signal to all event signal ; all event signal to event length; (left figure) zlength of the longest track minus length of the biggest cascade ; longest track signal to event length ; number of reconstructed tracks; (right figure) the use some other groups of variables:

25. Krzysztof W. Fornalski25 Range Searching - modifications zsize of V is changing zwhen the signal or background events number are less than 25 in one V, then V increase zbetter resolution modification of the original algorithm:

26. Krzysztof W. Fornalski26 Results of all presented methods

27. Krzysztof W. Fornalski27 Purity and efficiency of the NC & CC selection zto check the correctness of the selection, it is necessary to calculate the purity and efficiency of the selection: zPur = N true / (N true + N false ) zEff = N true / N all where N true is a number of correctly selected events, N false of incorrectly selected events, N all is the number of all events (CC or NC)

28. Krzysztof W. Fornalski28 Purity (right figures) and efficiency (left figures) of CC (up fig.) and NC (down fig.) selections. Big black points - Range Searching method. Small violet points - cuts method Eff CC Pur CC Eff NC Pur NC

29. Krzysztof W. Fornalski29 Purity and efficiency. Blue points - modified Range Searching method. Red points (CC only) - official MINOS Pur and EFF. Official MINOS NC selection does not exist yet! Eff CC Pur CC Eff NCPur NC

30. Krzysztof W. Fornalski30 Separation power zDefined earlier:

31. Krzysztof W. Fornalski31 Summary  MINOS can measure oscillation based on CC zgood CC from NC separation is necessary ztest of two major separation CC/NC methods: Range Searching better than cuts methods zRange Searching method can be modified to achieve better results zRange Searching can be used in other experiments, in which you need to separate some signal from some background

32. Krzysztof W. Fornalski32 THANK YOU! fornalski@knf.pw.edu.pl www.fuw.edu.pl/~minos