22
1 Latest CC analysis developments • New selection efficiencies: – Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts – Attempt to optimise PID cut • 5 year plan: – Sensitivities calculated year by year • Discrimination between osc. models – Standard oscillations vs nu decoherence D. A. Petyt May ‘03

1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

Page 1: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

1

Latest CC analysis developments

• New selection efficiencies:– Based on C++ reco + PDFs rather than old

(Fortran+reco_minos) cuts– Attempt to optimise PID cut

• 5 year plan:– Sensitivities calculated year by year

• Discrimination between osc. models– Standard oscillations vs nu decoherence

D. A. Petyt May ‘03

Page 2: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

2

PDF-based CC/NC separation• Idea is to replace old cut-based efficiencies with more modern/realistic values

• New selection uses likelihood method described at South Carolina

• Currently using simplest implementation – 3 1D pdfs.

CC

NC

Page 3: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

3

PID results (from SC)

CC

NC

Cut on the PID parameter to distinguish between CC/NC – the precise value of this cut will be a trade-off between purity and efficiency

Page 4: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

4

Sample ‘CC-like’ distributions

Visible energy (0-30 GeV)

PID>-0.5

PID>0.1PID>-0.1

PID>-0.3

Key

Perfect ID

CC

NC

Page 5: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

5

Optimising the PID cut

• Many possible ways to do this. The method I chose attempts to optimise the oscillation signal by examining– Overall significance of the effect (units of – Size of the dip at oscillation maximum (units of – Evidence for a rise at low energy (units of

• Optimisation will depend on how well NC background is understood. – I assume that I can make a full subtraction of the mis-identified

NC events, but assign an additional error that is proportional to the number of NC events in each bin of Evis

• Optimisation will be dependent on m2. – The plots I show here are for m2=0.0025 eV2

– Have looked at m2=0.0016 eV2. Optimum cut seems similar but statistical errors are large

Page 6: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

6

PID cut =-0.5, m2=0.0025

Page 7: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

7

Finding the optimum

• Define a quality factor that is the product of these three quantities.

• Optimum selection between –0.1<PID<0.1

• Optimium PID cut increases with increasing uncertainty on NC background

• I chose PID>–0.1 – try to maximise CC efficiency

Page 8: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

8

CC selection efficiencies as a fn. of true E and y

True neutrino energy (0-30 GeV)

Page 9: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

9

NC inefficiencies

Factor of 3 lower than previous cuts

True neutrino energy (0-30 GeV)

Page 10: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

10

Comparison of old and new efficiencies

Unoscillated CC spectra – No NC background

Page 11: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

11

5 year plan• I have taken the numbers for the

‘default’ 5 year plan that was sent around several weeks ago and have calculated year-by-year sensitivities with the following inputs:– le and semi-me,he beams from

NuMI-L-783– Efficiencies based on PID cut– Assume “VLE=LE”– Ignore anti-neutrino running (don’t

know fluxes)– Statistical errors only –

recommendations for beam (and other) systematics?

• This is a literal interpretation of 5 yr plan:– Do we want more generic N1020

p.o.t plots instead/as well? Some of these were shown at SC.

Page 12: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

12

Page 13: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

13

Page 14: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

14

Page 15: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

15

Page 16: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

16

Page 17: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

17

Discriminating against exotic models• Several exotic models of neutrino mixing exist which can provide a

reasonable fit to the SK -like data.• Nu decay is ruled out by SK due to the non-observation of sterile. Nu

decoherence, which does not involve activesterile transitions, is still a possibility.

Page 18: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

18

Oscillations vs decoherence

• This analysis motivated by hep-ph/0303064, which shows how well SK+K2K data can discriminate against the decoherence hypothesis

• General survival probability (osc+decoherence):

– Comments:• Pure oscillations: • Pure decoherence: m2• In pure decoherence scenario: P

• I have performed a fit to these 3 parameters, assuming standard oscillations with m2=0.0025 eV2 and compared the results to those obtained for SK/K2K in hep-ph/0303064

E

LmeP E

L

2cos1

2

2sin1

22

2 2

Page 19: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

19

Fit result – 5 year plan, year 1

Shown are the three 2D projections of the 3-dimensional allowed volume

SK allows m20 at 99% C.L. MINOS does not

Page 20: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

20

Fit result – 5 year plan, year 5

Page 21: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

21

Constraint on decoherence parameter

Page 22: 1 Latest CC analysis developments New selection efficiencies: –Based on C++ reco + PDFs rather than old (Fortran+reco_minos) cuts –Attempt to optimise

22

Plots for 5yr plan/response to PAC questions• Two basic questions:

– What plots are required?

– What inputs (beams, systematics…)

• 5 year plan plots:– Specific run plans and/or generic sensitivity vs POT plots?

– Low m2 scenario (using VLE beam, if available)?

• PAC questions:– Q1 – ‘5 sigma dip’: precise definition? Easy to do but implies low p.o.t. even for

low m2

– Q2,3 – ‘95% exclusion of nu decay/3 preference of osc. over decay’: between best-fit osc and best-fit decay? Probably also implies low p.o.t. (see decoherence results)

• Inputs:– More up-to-date spectra (esp. le) available?

– Recommended systematic uncertainties on beam flux?

– Treatment of mis-identified NC: assume perfect subtraction or full subtraction with some error. What error?