upgrade of bbfit v3r6 v4r0 c. reed development of method s. galata implementation into seatray...

Upgrade of BBfitv3r6 v4r0

C. Reeddevelopment of method

S. Galata implementation into SeaTray

J. B. Implementation into online (standalone) version

24 . 11 . 2008 C P P M 2

Progress with BBFit

Corey Reed

Antares Physics Analysis WorkshopRome, November, 2008

BB3D 2-Line Degeneracy BBC Full Fit Resolution Plans

24 . 11 . 2008 C P P M 3

2-line Degeneracy2-lines used in fit

Two tracks have same chi2 in BBFit

Same zenith

Different azimuth

BB3D 2-Line Degeneracy BBC Full Fit Resolution Plans

24 . 11 . 2008 C P P M 4

2-line DegeneracyPredict time, angle of Cherenkov photonsLoose time residual cut30 ns, to check only direct hitsLook all hits (not just those used in fit)

Variable Selection EfficiencySum of hit charge * angl. wt.61.3%Sum of angl. accpt. wt. 67.7%Number of hits in cut 69.9%Total charge of hits in cut 70.2%

BB3D 2-Line Degeneracy BBC Full Fit Resolution Plans

New hit selection + Final fit step

Take result from original BBfit or mirror solution if betterNew hit selection within 20nsecNew fit with M-estimator : time residuals only

Whole procedure was originally implemented in ArkWas thought to work well only with full geometry

(wrong assumption !)

Implementation into online BBfit

• For Nline==2– Mirror resolution method applied

• For Nline>1– New hit selection + M-estimator applied

• Ntuple modification (BBbatch)– tcosth,tphi,tcterr,tphierr replaced– New fields added : nlim,nhim,ampm,chim

• Event Display– New hit selection and final fit displayed

BBfit v3r6 linesLines with T3

BBfit v4r0 linesLines added from new hit selection

Track Fit quality• Multi-line sample (nline>1 && nhit > 5) Ag-2008• Toffset_line + MC(2nsec) included

Data/MC = 1.3

Data 665MC 609MC 40

Improvement of angular resolution

• Example: atmospheric neutrinos (12L 3pe)• online BBfit, nominal geometry

Nline == 2 Nline > 2v3r6 4.32ºv4r0 1.79º MC 1.85º

v3r6 0.95ºv4r0 0.64º

Angular resolution (median)

• Important improvement with final fit step

• Full geometry only contributes 0.1º

v3r6 v4r0 full geom v3r6 v4r0 full geom

Number of Lines used in fit

• 12 lines 3pe period

nline nlinm

Classic BBfit After new hit selection

Number of storeys (hits) used in fit

• 12 lines 3pe period

<nhit>=11.5 <nhim>=18.9

Classic BBfit After new hit selection

Total charge of used hits

• 12 lines 3pe period

<amp>=69 p.e. <ampm>=79 p.e.

Classic BBfit After new hit selection

Effective area ()• Comparison of strategies

Q<1.4

Nline>1&&Nhit>5

AAfit (C.B.) BBfit

Loose cut Point Source analysis 90% neutrino purity

Effective area ()• Comparison of strategies

Crossover at ~5TeV

BBfit very bad at high energies

BBfit better at low energies if same purity requires

“Loose cut” curves agreeSanity check

Dashed: BBfitSolid: AAfit

Effective area ()

• Modified quality cut in BBfit• Recover HE part which was lost due to

distance penalty in fit qualityNline > 1Q < 1.3+(0.04*NDF)^2Q < 1.4

Same efficiency at low E

Important gain at high E

Effective area ()

Crossover at ~20TeV

BBfit better at low energies if same purity requires

BBfit with similar performance at high E

Dashed: BBfitSolid: AAfit

Effective area

• Watch out for differences between &

Full NeutrinoDashed Antineutrino

Elevation angles 2008

• Purity unchanged by modified cut

Data 665MC 609MC 40

Data 624MC 588MC 40

Q < 1.4 Q < 1.3+(0.04*NDF)^2

Angular resolution

• Enery dependent cut improves PSF for M-estimator fit

• Median unchanged

Q < 1.3+(0.04*NDF)^2Q < 1.3+(0.04*NDF)^2Q < 1.4Q < 1.4

v3r6 v4r0 full geom v3r6 v4r0 full geom

0.69º

0.43º