Atmospheric Neutrino Event Reconstruction

Atmospheric Neutrino Event Reconstruction

Andy BlakeCambridge University

June 2004

Introduction

Reconstruction- Track/shower finding- Track fitting (fast measurement of track curvature)

Analysis Modules- Raw Digit Dump. (raw digits, TPMT hits, dead chips etc…)- Cand Digit Dump. (positions, times, pulseheights, fibre lengths etc…)- Cand Track/Shower Dump. (track/shower parameters, analysis variables etc…)- Event Display.

Testing- Run over most of the data.- Used in Caius’ analysis.- Fast ( ~100 ms / event ).

AtNuReco

This Talk

• Direction Reconstruction → Timing Resolution

• Charge Reconstruction → Separating /

_

Direction Reconstruction

Up-Going Events

Direction-Finding Algorithm:- consider distance vs time for track- force fits with β = ± 1- calculate RMS about each fit- RMSdown-RMSup > 0 for up-going tracks.

up-goingneutrinos

UP-GOING EVENT !

MC/data Comparisons

Timing Resolution

Timing resolution:

Data = 2.75 nsMC = 2.40 ns

RMSdown for stopping muons:

Try to understand this discrepancy: - refractive index - time walk - timing calibration

Refractive Index

pmpm

p0p

m0m

LL nctct

nLctct

nLctct

Lm Lp

tm tp

t0 muon

nreco=1.75 ndata=1.82 nMC=1.73

Use double-ended strips on muon tracks:

Timing Calibration

2

LLnctctΔt pmpm

measured time difference - expected time difference between strips ends between strip ends

define:

Mean Δt tests goodness of

calibration

RMS Δt measuresintrinsic timing

resolution

Use measured refractive indices

Timing Calibration

~3000 entries per plane

September 2003 data

Overall calibration good to <0.5 ns

Poor calibration in last two crates

Monte Carlopeaks at -0.3 ns(weird east-west

asymmetry)

Mean Δt ( → test calibration constants )

Timing Calibration

September 2003 data

MC slightlybetter than data.

RMS Δt ( → measure intrinsic timing resolution )

Using n=1.82 instead of n=1.75improves resolution by ~0.2 ns.

Time Walk

Timing resolution depends on size of signal:

RMS Δt(set Qm ≈ Qp)

Time Walk

… but timing fits are charge-weighted so this effect gets suppressed.

Signal rise time depends on size of signal:

Timing Resolution

2cal

2PH

2

2222

trk ΔtΔtc

ΔLΔnΔtΔt

Monte Carlo : 2.4 ns

2.45 ns 0.9 ns

0.25 ns2.45 ns

2.7 ns 0.3 nsData :

Total resolution

Intrinsicresolution

Refractive index Calibration

=

=

Try to reproduce tracking resolutions by combining individual errors:

use ΔL ~ 4m

TimeWalk

0.7 ns

0.5 ns

Timing Drift

~5% degradation in 6 months

Timing Drift

September 2003 data

RMS Δt ( → intrinsic timing resolution )

2% systematic variation across detector?

Timing resolution per plane -SM1 slightly worse than SM2?

Timing DriftCurrent Timing Calibration:

• Overall calibration has degraded (~0.3 ns → ~0.4 ns)• some structure + large displacements have developed.

Current Calibration

• Timing structure lines up with VARC + crate boundaries.

Current Calibration

• Large displacements are due to hardware changes.• VFB swaps, VARC swaps, dynode threshold adjustments etc…

Up-Going Events

• Correlations between regions of poor calibration and up-going tracks !

2.5 kT-yrs data

PC up-going tracksafter timing cuts

Up-Going EventsUp-going candidate: planes 118 – 128 - real or badly calibrated ?

Up-Going EventsUp-going candidate: planes 438 – 447 - real or badly calibrated ?

Conclusions• Far Det timing resolution is ~2.5ns - Monte Carlo and data agree to <5%. - Resolution in data degraded by calibration + larger time walk + choice of refractive index.

• Timing calibration vital for analysis of up-going atmospheric neutrinos.

- Calibration constants are drifting over time. - Hardware changes cause significant shifts. - Need to correct for these changes.

Charge Reconstruction

Charge Reconstruction

• Charge-Finding Algorithm:- need to measure curvature of muon track in magnetic field.- split track into overlapping 15 plane segments and parametrize each track segments using quadratic fit.- calculate Q/p and ΔQ/p for each segment.- combine the measurements to give an overall value for Q/p and ΔQ/p.

pQ

ks1.0

pQ

pQ

: k loss energy linear assuming

|Bp|0.3

Bpds

pd

CpQ

:charge give to rearrange

dsdp

pBp0.3Qds

pd

: by given momentum muon

0

2

loss

) 2.30.25

0.59correction gap airC

direction, track measuredp (

ˆ

ˆ

ˆˆ

ˆˆ

Charge Separation

2.0ΔQ/PQ/P

for 90% exceeds separation theΔQ/PQ/P

usingout carried separation charge

Comparison with SRCompare: SR tracker + SR fitter (total efficiency = 86%) AtNu tracker + AtNu fitter (total efficiency = 89%)

(atmos CC, >10 track planes, passed track fitter)

~20%

~30%slightly bettercharge separation

for AtNu

Comparison with SR

AtNu SR

AtNu 89% 90%

SR 84% 86%

FITTING

TRACKING

AtNu providesslightly better tracking

for atmos nu events.

SR provides slightly better fittingfor atmos nu events.

Look at all combinations :

Comparison with SRCharge separation in cosmic muons:

AtNu better at lower energies

SR better at higher energies

Comparison with SR

Conclusions

• AtNu charge reconstruction in good shape. - Algorithm is fast and robust. - Performs well at low energies, → ideal for FC atmospheric neutrino analysis. - Efficiency starts to drop away for E >20 GeV.

• Momentum from curvature good to ~30%. - plan to refine assumptions made in algorithm.