IC22 Point Source Analysis with Unbinned Maximum Likelihood
IC-22 Point Source Analysis with Unbinned Maximum Likelihood C. Finley, J. Dumm, T. Montaruli 2008 May 2
Overview 1. Basic Set of Cuts for point-source quality sample using Level 2 processing • 14 days ( 12. 6 days livetime) of Level 2 Filtered data § • 1000 files of Level 2, muon-filter Nugen E-1 § • (Sept. 14, 15, 16, 17, 18, 20, 21, 22, 23, 24, 25, 27, 28, 29) Dataset 753 1000 files of Level 2, muon-filter Nugen E-2 § (optimal for E-2 source sim. ) (optimal for high statistics Atm. Nu) Dataset 768 2. First look at adding energy term to the likelihood in IC-22 (similar to method of J. Braun for AMANDA-II) 2008 Madison C. Finley 2
Muon Filter + Pandel Zenith>90 Cut 2008 Madison C. Finley 3
Paraboloid Sigma Cut 2008 Madison C. Finley 4
Pandel Reduced Log Likelihood Cut 2008 Madison C. Finley 5
Split Hit Series Cut: Pandel zen 1>80 and zen 2>80 2008 Madison C. Finley 6
NDir. C Cut 2008 Madison C. Finley 7
Pandel SDir Cut 2008 Madison C. Finley 8
Maximum Likelihood Analysis Part I: Point Spread Function By now the maximum likelihood expression is familiar for point source searches: where the source PDF is a Gaussian with width given by the paraboloid sigma uncertainty estimate: and the background PDF depends only on zenith angle: 2008 Madison C. Finley 9
Sensitivity with Analysis-Level Cuts Preliminary IC-22 Sensitivity for 250 d ( ~ 20 atm. Nu events per day ) E-2 average sensitivity 0 = 1. 7 10 -11 Te. V-1 cm-2 s-1 2008 Madison C. Finley 10
Sensitivity with Analysis-Level Cuts IC-9 Sensitivity IC 22 Sensitivity E-2 average sensitivity 0 = 1. 7 10 -11 Te. V-1 cm-2 s-1 2008 Madison C. Finley 11
Sensitivity with Analysis-Level Cuts E-3 E-2 Sensitivity est. from Gent meeting, based on atm. Nu background simulation only 2008 Madison C. Finley 12
Maximum Likelihood Analysis Part II: Energy Now, want to add energy term to the Likelihood function, to weight higher energy events with greater significance: 2008 Madison C. Finley 13
Maximum Likelihood Analysis Part II: Energy Now, want to add energy term to the Likelihood function, to weight higher energy events with greater significance: 2008 Madison C. Finley 14
Energy PDFs Start with simplest energy estimator: NChan P (Ei | =2) P (Ei | =3) Patm (Ei) 2008 Madison C. Finley 15
gamma Maximum Likelihood fit to n. Src and Gamma: Examples + + n. Src gamma n. Src Examples: Simulated E-2 source at declination +30° + + Left: 15 events injected (cross) Right: 30 events injected (cross) 1 -sigma and 2 -sigma contours are shown for best fit to number of source events n. Src and spectral index gamma n. Src 2008 Madison C. Finley 16
gamma Maximum Likelihood fit to n. Src and Gamma: Examples + + n. Src gamma n. Src + n. Src Below: Simulated E-3 source at declination +30° Left: 15 events injected; Right: 30 events injected 1 -sigma and 2 -sigma contours are shown for best fit to number of source events n. Src and spectral index gamma 2008 Madison C. Finley 17
Effect of Energy Term on Discovery Potential Simulated E-2 source at declination +30°: 5 -sigma Discovery potential (Power 50%): without energy term in likelihood: 6. 1 10 -8 Ge. V-1 cm-2 s-1 (E/Ge. V)-2 (mean number of source events: 15) with energy term in likelihood: 2008 Madison 4. 2 10 -8 Ge. V-1 cm-2 s-1 (E/Ge. V)-2 (mean number of source events: 10. 5) C. Finley 18
Summary Basic cuts on Level 2 parameters and basic maximum likelihood analysis yield point source sensitivity ≈ 7 x better from IC-9. Many significant improvements coming soon: • Level 3 reconstructions • MPE reconstruction: better high energy efficiency, better angular resolution talk by J. Dumm) • Time residuals of MPE should improve efficiency of NDirect cut (or we may find an alternative cut for best high energy efficiency) (see Adding energy term to likelihood: • Test with new energy estimators in level 3 (better than Nch estimate) • Reasonable to expect at least 30% improvement in discovery potential for hard spectra. 2008 Madison C. Finley 19
- Slides: 19