Post on 13-Jan-2016
Kate HusbandCambridge University, UK
Development of an Online Filter for Selection of Cascade-like Events in IceCube
Supervisor: Eike Middell
Signals
• IceCube looks for high-energy astrophysical • E.g. from supernova remnants• Also detects atmospheric ’s
• Primary cosmic rays produce muons which is the main background to IceCube.
IceCube
• 1km3 high energy, telescope at S Pole
• Contains DOMs with PMT which record amp. & time of photons
• Ice contains dust particles and air bubbles so photons scatter
Events: Cascade-like
• interact through weak force only
• Quarks produce cascade of Cherenkov photons
• Photon intensity decays rapidly
• Muons can also knock out e- resulting in em cascade
Events: Track-like
• Muons move close to c • Don’t lose much energy
in producing Cherenkov light
• Only arrive from above detector as decay quickly
• Hence long straight downwards tracks
Multi- Vs. Single-Peak
• If triggered, DOM records continuous waveform
1. Try and fit pulses of different times and amp. into waveform = MULTI-PEAK (mp)
OR
2. Find no. of photons from total amplitude and give all photons the same time = SINGLE-PEAK (sp)
AIMs
• Online filter must separate multiple background events from signal within constraints.
• Constraints: limited bandwidth & CPU power
• Aim to improve efficiency by looking at possible new single-peak cut variables
A. ToI• Each DOM is given a
mass proportional to amount of charge it records
• Centre of gravity of distribution is used as approx. to interaction vertex
• Good cut variable as spherical ~0.3, tracks ~0.
B. LineFit Velocity
• Fits a straight line through triggered DOMs
• Var(time) is not a good cut variable
• Charge weighting doesn’t increase separation
C: Cumulative Charge
• Cumulative sum of charge of each DOM over time
• Take time at x% of total charge to get tx
• Single-peak
• ~t75 found to give largest separation
D: Charge Propagation
• Decay of charge-energy ratio different
• Total charge ≈energy and centre of gravity ≈ interaction vertex
• Single-peak
• Mean on new axis for each event weighted with distance is new variable: MCD
Optimization
• New filter: MCD+ToI (sp): 73% e
• Previous filter: vel+ToI (mp): 71.7% e (smaller detector)
Energy Efficiency
• New filter is slightly better at low energies
• Blue: vel +ToI (mp), Red: MCD+ToI (sp)
Summary
• LineFit velocity cannot be replaced by variance of time
• New single-peak cut variables: t75 and MCD
• New filter suggested with MCD>61.4, ToI>0.14 passing 73.0% sig. for 1% bkd.
• New filter has slightly better low energy efficiency.