210 likes | 404 Views
CC/NC SEPARATION STUDY. Andy Blake Cambridge University Friday February 23 rd 2007. Introduction. Have developed a PID for CC/NC separation PID is calculated using a likelihood technique, and extends the “standard” PID algorithm by incorporating some new PDFs and
E N D
CC/NC SEPARATION STUDY Andy Blake Cambridge University Friday February 23rd 2007
Introduction • Have developed a PID for CC/NC separation • PID is calculated using a likelihood technique, and extends the • “standard” PID algorithm by incorporating some new PDFs and • accounting for their change in shape as a function of energy. • Code committed to “MadAbID” class in Mad package. • Documentation also available in doc-db #2720. • Will outline method and present some results in this talk. Andy Blake, Cambridge University CC/NC Separation, slide 2
PID Variables • Variables for CC/NC separation • Event Topology Variables. • – percentage pulse height in track. • – pulse height per track plane. • – number of track like planes. • – percentage error in track fit. • Event Kinematics Variables. • – reconstructed y. • Event Energy and Charge. • – number of track planes. • – reconstructed charge. • Relative CC/NC Normalization. Test consistency with muon track. Test consistency with CC interaction. Incorporate CC/NC spectral information. Andy Blake, Cambridge University CC/NC Separation, slide 3
PID Definition • PID calculated from the product of several 1D and 2D PDFs as follows: Andy Blake, Cambridge University CC/NC Separation, slide 4
PID Variables: (I) Track Topology TRACK-LIKE PLANES vs TRACK PLANES Andy Blake, Cambridge University CC/NC Separation, slide 5
PID Variables: (I) Track Topology TRACK PH / TRACK PLANES vs TRACK PLANES Andy Blake, Cambridge University CC/NC Separation, slide 6
PID Variables: (I) Track Topology TRACK PH / EVENT PH vs TRACK PLANES Andy Blake, Cambridge University CC/NC Separation, slide 7
PID Variables: (I) Track Topology TRACK FIT ERROR vs TRACK PLANES Andy Blake, Cambridge University CC/NC Separation, slide 8
PID Variables: (II) Event Kinematics RECO Y vs RECO E Andy Blake, Cambridge University CC/NC Separation, slide 9
PID Variables: (III) CC/NC Spectrum TRACK PLANES RECONSTRUCTED CHARGE CC/NC NORMALIZATION Andy Blake, Cambridge University CC/NC Separation, slide 10
PID Study • Cedar MC Ntuples. • – Far Detector (generate PDFs using 8.7e22 PoTs • calculate PIDs using 2.9e22 PoTs). • – Near Detector (generate PDFs using 8.5e18 PoTs • calculate PIDs using 2.5e18 PoTs). • Event Selection. • – reconstructed track. • – successful track fit. • – contained track vertex. • Construction of PDFs. • – PDFs constructed assuming no oscillations. • – 2D PDFs normalized to remove any spectral information. • (i.e. divide out the shape of the energy spectrum). • – as a final step, try pre-selecting events with PCC=1. • (i.e. create second set of PDFs for events with PCC<1). Andy Blake, Cambridge University CC/NC Separation, slide 11
PID Results (Far Detector) Andy Blake, Cambridge University CC/NC Separation, slide 12
Purity vs Efficiency (Far Detector) Standard PID: track PH/ event PH. track PH/ track planes. track planes. Incorporate: track-like planes. track fit error. track charge. reconstructed Y. (replace trk.ph/evt.ph). pre-selection. N.B: maximum values of purity*efficiency indicated by stars Andy Blake, Cambridge University CC/NC Separation, slide 13
Purity vs Efficiency (Far Detector) Standard PID cut approximately here! Andy Blake, Cambridge University CC/NC Separation, slide 14
PID Results (Near Detector) Andy Blake, Cambridge University CC/NC Separation, slide 15
PID Results (Low Energies) E < 3 GeV 3 < E < 6 GeV 6 < E < 9 GeV E > 9 GeV Andy Blake, Cambridge University CC/NC Separation, slide 16
Summary • Incorporation of new variables into PID calculation improves purity and • efficiency of CC/NC separation in both detectors over all energies. • This will hopefully improve the sensitivity of the oscillation analysis! • Possible improvements to this method: • – Separate fully and partially contained events. • – Separate neutrinos and anti-neutrinos. • – Incorporate other topology variables (e.g. pulse height profile of event). • – Incorporate another kinematic variable (e.g. x distribution). Andy Blake, Cambridge University CC/NC Separation, slide 17