CC ANALYSIS STUDIES Andy Blake Cambridge University Fermilab, September 2006.

CC ANALYSIS STUDIES

Andy BlakeCambridge University

Fermilab, September 2006

Overview

Andy Blake, Cambridge University CC talk, slide 2

• Have started to look at some CC analysis issues.

– Validation of R1.24.

– Study the current reconstruction + analysis.

• Long term goal is to optimize the measurement of sin2223.

– Need to accurately resolve the size of the oscillation dip.

– Need a clean event sample. (good CC/NC separation, good energy resolution, few reconstruction errors etc…)

• This talk is divided into the following topics: (I) energy reconstruction. (II) CC/NC separation.

Data Selection

Andy Blake, Cambridge University CC talk, slide 3

• Far Detector beam Monte Carlo SR ntuples (R1.24c). ~75,000 beam events.

• Apply simple pre-selection to reconstructed events: 1 event per snarl (avoid split events for now). >10 digits per event (below tracking threshold). signature of CC interaction is reconstructed track.

• Define fiducial volume: 50 cm from edge of detector. 40 cm from centre of coil hole. 5 planes from beginning of detector 5 planes either side of super-module gap. 20 planes from end of detector.

(I) Energy Reconstruction

Event Reconstruction

Andy Blake, Cambridge University CC talk, slide 5

Muon Reconstruction:

Define longest track to be primary track.

FC events: both vertex and end inside fiducial volume. use momentum from range.

PC events: vertex inside fiducial volume. end not inside fiducial volume. use momentum from curvature.

Shower Reconstruction:

Collect up all sub-showers close to the event vertex. Add in unassigned strips not along primary track.

Neutrino Energy:

neutrino energy = muon momentum + shower energy.

Shower Reconstruction

Andy Blake, Cambridge University CC talk, slide 6

associated withhadronic shower

associated with muon track

a sub-shower is associated with vertex shower if: Zshw-Zevt<0.5m OR E>0.5 GeV OR shower not on track.

Shower Reconstructionan unassigned strips are associated with vertex shower if: PHeast+PHwest>200 ADCs AND strips not adjacent to track.

( Approximate 10,000 SigCor ~ 1 GeV for these strips).

Andy Blake, Cambridge University CC talk, slide 7

Reconstruction Efficiencies

Andy Blake, Cambridge University CC talk, slide 8

MUON TRACK RECONSTRUCTION SHOWER RECONSTRUCTION

# events with tracks

# eventsefficiency =

# events with shower energy

# events with tracksefficiency =

Reconstruction efficiencies for CC events (true vertex inside fiducial volume):

Reconstruction Efficiencies

Event (1) P = 500 MeV

Event (2) P = 500 MeV

Event (3) P = 700 MeV Event (4) P = 600 MeV

Examples of CC events without reconstructed tracks: blue line = true muon direction

Muon Momentum From Range

Andy Blake, Cambridge University CC talk, slide 10

0-1 GeV

1-2 GeV

2-3 GeV

3-4 GeV

4-5 GeV

bias towards high energiesdue to track reconstruction

(N.B: used to be much worse!)

bias towards low energiesdue to track containment

and showers at end of track.

momentum reconstruction consistent with ~5% error.

reco - true muonmomentum for

1 GeV wide bins:

MUON MOMENTUM FROM RANGE

Muon Momentum From Range

R1.18.2 R1.24c

• The bias at low muon energies has always existed!

• Caused by over-tracking and/or mis-tracking (track finder prefers longer tracks)

• Bias is reduced by the introduction of the new track finder in R1.24c.

• Bias is worse in shower-like events, so will be correlated with PID parameter.

Andy Blake, Cambridge University CC talk, slide 11

Muon Momentum from Range

Andy Blake, Cambridge University CC talk, slide 12

true muon direction = blue line

(3) wrong track picked (4) wrong track picked

(2) track extended past vertex(1) track deviates off course

Muon Momentum from Curvature

Andy Blake, Cambridge University CC talk, slide 13

0-1 GeV

1-2 GeV

2-3 GeV

3-4 GeV

4-5 GeV

reco - true muonmomentum for

1 GeV wide bins:

MUON MOMENTUM FROM CURVATURE

Muon momentum resolution is ~ 10-20%,but there are low/high energy tails

Muon Momentum from Curvature

Feed down of high energy neutrinos into low energy bins:

Andy Blake, Cambridge University CC talk, slide 14

PC events with: Preco < 3 GeV

Out-lying tail of high energy muons reconstructed with a low energy

(true muon momentum)

Muon Momentum from Curvature

Andy Blake, Cambridge University CC talk, slide 15

(1) (2) (3)

Feed down of high energy neutrinos into low energy bins for PC events:

Ptrue = 9.9 GeVPreco = 0.5 ± 0.1 GeV

Ptrue = 8.4 GeVPreco = 1.8 ± 0.8 GeV

Ptrue = 14.3 GeVPreco = 1.5 ± 0.5 GeV

Shower Energy

Andy Blake, Cambridge University CC talk, slide 16

0-1 GeV

1-2 GeV

2-3 GeV

3-4 GeV4-5 GeV

Reconstructed energy peaks in correct place and resolution is consistent with 55%/√E

reco - true showerenergy for 1 GeV wide energy bins:

RECONSTRUCTED SHOWER ENERGY

Neutrino Energy

Andy Blake, Cambridge University CC talk, slide 17

0-1 GeV

1-2 GeV

2-3 GeV 3-4 GeV

4-5 GeV

RECONSTRUCTED NEUTRINO ENERGY

reco - true neutrinoenergy for 1 GeV wide energy bins:

Neutrino Energy

Andy Blake, Cambridge University CC talk, slide 18

R1.24c

R1.18.2

Neutrino Energy

Andy Blake, Cambridge University CC talk, slide 19

Look at feed down of high energy neutrinos into low energy bins:

All events with Ereco < 3 GeV

The feed down from high energyPC events well below 1% level.

Energy Resolution

Andy Blake, Cambridge University CC talk, slide 20

FC events: E = 5% P 55%/√Eshw

PC events: E = q/p/(q/p)2 55%/√Eshw

define an approximate resolution function for FC and PC CC events:

FC PC

Neutrino Energy Resolution

Andy Blake, Cambridge University CC talk, slide 21

Divide up events by estimated neutrino energy resolution:

Energy Resolution

Andy Blake, Cambridge University CC talk, slide 22

E/E < 15% 15% < E/E < 30%

30% < E/E < 60% E/E > 60%

oscillations close to zero!

m2 = 2.74 eV2

sin22 = 1.0

Note: NCbackgroundnot included!

(II) CC/NC Separation

CC/NC Separation (1)

Andy Blake, Cambridge University CC talk, slide 24

Start with standard PID variables.

Track Planes Track PH / Event PH Track PH / Track Planes

CC

NC

CC/NC Separation (1)

Andy Blake, Cambridge University CC talk, slide 25

100% CC entries

Form the PID using standard prescription:

CC

NC

CC/NC Separation (1)

Andy Blake, Cambridge University CC talk, slide 26

3 variables

PID = - 0.2

CC/NC Separation (2)

Andy Blake, Cambridge University CC talk, slide 27

“Track-like” planes(number of planes with little shower activity)

Error in track fit | Q/p | / Q/p

(test of consistency with muon track fit)

CC

NC

Choose some additional variables: (i) Reasonable physics motivation. (ii) Good separation of CC and NC events. (iii) Fairly similar in Near and Far Detector (e.g. can’t use timing).

50 planes

CC

NC

CC/NC Separation (2)

Andy Blake, Cambridge University CC talk, slide 28

5 variables

3 variables

PID = - 0.2

CC/NC Separation (3)

Divide up PID distributions by track length. (i) track planes < 25 (ii) 25 < track planes < 50 (iii) track planes > 50

Use the difference in the shape of the PID distributions as a function oftrack length to enhance the CC/NC separation at low neutrino energies.

(i) ~100% of tracks with >50 planes are CC events. (ii) distributions of track PH / event PH change markedly for short tracks. (iii) track-like planes + fit error provide some separation at low energies.

Andy Blake, Cambridge University CC talk, slide 29

CC/NC Separation (3)

Andy Blake, Cambridge University CC talk, slide 30

CC NC

separation from pulse height has almost all gone

PID variablesfor muon tracks spanning less than 25 planes:

Some separationfrom track-like

planes

some separationfrom track curvature

CC/NC Separation (3)

Andy Blake, Cambridge University CC talk, slide 31

5 variables

3 variables

5 variables + separation by event length

PID = - 0.2

Events with: Ereco < 3 GeV

Summary

Andy Blake, Cambridge University CC talk, slide 32

• Reconstruction appears to be in pretty good shape.

– Energy reconstruction is accurate with good resolution.

– Some small problems and biases, but very hard to handle.

– Only a small number of outlying events fall into oscillation region.

• Oscillation dip is better resolved by dividing events by resolution.

– This may improve the measurement of sin2223.

•Some improvement possible in CC/NC separation.

– ~10% improvement in selection efficiency at low energies.

• Lots more work for me to do!