CC analysis progress

• This talk:– A first attempt at calculating CC energy sensitivity

using the Far Mock data MC files with full reconstruction.

– Quite a few changes from Five year plan plots:• “Real” events used rather than fast MC with histogrammed

flux and y distributiuons• Reconstructed track and shower energies used rather than

Gaussian-distribution resolution functions• Nuclear effects accounted for including intranuclear re-

scattering• Neugen 3 used with flatter y-distribution

– This analysis is not optimised yet – should be able to improve energy resolution and event selection efficiencies

D.A. Petyt June 2004

Event ID – PDFs (top row – CC, bottom row – NC)

Require that all events contain at least 1 reconstructed track

Event length (planes) Track p.h./event p.h track p.h./plane (siglin ADC)

Event ID – PID parameter

NC

CC

)()()(

)()()(

332211

332211

ncncncnc PPPP

PPPP

)(ln)(ln ncPPPID

Selection efficiencies

Cut here

Event reconstruction efficiencies are included in these numbers

CC efficiencies – Reco and PID chain

Should be better in R1.8

First 5 efficiency plots are sequential, last plot is cumulative

Energy reconstruction

• Ereco=EEhad

• Use track fitter to provide muon momentum: range for contained tracks, 1/(q/p) for exiting tracks (min 3m track length)

• How to estimate Ehad?– Test some ideas on NC

events:• Ehad = Shower ph*scale

factor?• Ehad = Event ph * scale

factor

– Using event ph works better: ph/GeV distribution has 15% higher mean and narrower width

True NC events

Number of reco tracks and showers

Fraction of event p.h. in reco shower(s)

Shower p.h./Ehad (siglin ADC units)

Event p.h./Ehad (siglin ADC units)

Showers in CC events• Attempt to use same algorithm

to estimate shower energies in CC events

• Must account for pulse height contribution of muon track in shower region.

• Can’t use SR trk.ph.siglin, as this includes shower ph for strips in which tracks and showers overlap

• Estimate track pulse height as trk.ndigit*295, where scaling factor is determined from long muons

• Subtract estimated track p.h. from event ph to yield shower energy

• Scaling factor obtained: 10500 siglin ADC/GeV

True CC events

Numerator: shw p.h.

Denominator: event ph – estimated track ph.

Shower p.h./Ehad (siglin ADC units)

Estimated shower p.h./Ehad (siglin ADC units)

Resolution functions

Asymmetric tails caused by shower fluctuations

Small negative bias. Tails mostly due to q/p errors

All (reco-true)/true

P reconstructed high - Exiting muon with little or no curvature

P reconstructed high - Exiting muon with kink

Ehad reconstructed high – high ph digits

Ehad reconstructed high – lego plot of snarl 81

Comparison of true and reconstructed E spectra

True CC events True NC events

Identified CC-like events: PID >-0.3

TrueReconstructed

Energy spectra and spectrum ratios

Unoscillated

NC

Oscillated

Identified CC-like events: m2=0.002, sin22=1, scaled to 16e20 p.o.t

No NC subtraction in ratio plot

Unoscillated

NC

Oscillated

Identified CC-like events: m2=0.002, sin22=1, scaled to 16e20 p.o.t

Energy spectra and spectrum ratios

NC contribution subtracted with 20% error

Allowed regionsm2=0.002 eV2, sin22=1, 16e20 p.o.t

“Rise” in spectrum ratio not seen here

Allowed regions - 2

m2=0.00295 eV2, sin22=1, 16e20 p.o.t

“Rise” in spectrum ratio is observed here

Comparison of old and new

5 year plan analysis

Comparison of old and new

This analysis

What happened to the low energy rise?

I can think of 4 likely reasons:• Energy smearing due to nuclear/mass/pion absorption

effects in NEUGEN. Already saw some evidence of this when I compared toy MC event generator with NEUGEN last June

• Additional smearing caused by non-Gaussian tails in resolution functions. Andy Culling’s Ehad method should help here

• Toy MC used a binned FarDet energy spectrum with linear interpolation between bins. Will tend to over-estimate event rate in lowest energy (0-1 GeV) bin

• Track finding efficiency is lower, especially at low E

What happened to the low energy rise?

Track finding efficiency, old and new

• Track finding efficiency is much higher in the old ntuple. I suspect that this is because the reconstruction was configured with cosmic parameters which contain much looser cuts on what constitutues a “track” plane. Tracks are also found in most NC events, but these can be suppressed with the PID cut.

• “NEW” efficiency is set to improve with R1.8 (low p tuning by Niki, bug fixes by Jim)

Old (NEUGEN2) MC file, tracker run with cosmic mu parameters?

Mock Data MC files, tracker run with R1.7

QEL p distribution

All events Reco track PID cut

How many of these can we recover?

What next?

• Re-do analysis with R1.8, which will contain a number of reco improvements. How much does this help?

• Look into incorporating Andy Culling’s Ehad prescription to improve shower energy resolution

• Look at events that do not have a reconstructed track. Are there usable CC events in this sample? How can we extract them?