Reconstruction

1
Reconstruction

• Muon tracks (Zenith, Azimuth, vertex, time)
• Cascades (Vertex, Time, Energy)

• First guesses
• Simple ! Fast !
• Used as a seed to more complicated algorithms
• Fast analysis for online filtering
• Higher level reconstruction
• Based on log-likelihood method
• Ice-properties, Cherenekov cone important
• More accurate, more time consuming

2
Line fit

• First guess for tracks
• Input hit time (ti), hit location (ri),
  Amplitude (Ai)
• Ignores Cherenkov cone, ice properties
• Light travelling in velocity V
• minimze c2 to obtain velocity V and vertex r
• Magnitude of V is the mean speed of the light
  propogating through a 1-dimensional detector
  projection.
• Large V long thin tracks, Low V Cascades,
  high energy tracks

3
Dipole Fit

• Sort all hits by time
• Calculate global dipole moment by averaging all
  indivdual dipole moments
• Different step sizes. Best performance using N/2.

4
Direct Walk

• Selecting hits most likely produced by direct
  photons.
• Depends on in-code magic numbers that correspond
  to the detector geometry
• Reconstructing track-elements based on pattern
  recognition - What OMs correspond to what track
  element
• If several good tracks were found average within
  clusters

5
JustAnotherMuonSearch

• 4 steps
• Hit Selection
• Cluster Finder first guess candidate track
• Fast Fitter Narrow down first guesse solutions
• Fit Quality Choose 1 / 2 best track
• Not fully implemented in Icetray yet.

6
CascadeFirst

• Primary first guess algorithm for cascades.
• Calculates the center of gravity of the hit OMs.
• The COG of the hits gives the first guess vertex
  position.

7
Tensor of Inertia

• First guess algorithm geared for cascades
• Treats the amplitudes of OMs as virtual masses,
  and forms an Inertia Tensor of these masses
• The eigenvector corresponding to the smallest
  eigenvalue approximates direction of cascade.
• Ratio of eigenvalues tells you how cascade-like
  event is.

8
Loglikelihood Reconstructions

• Computationally more expensive than first
  guesses, but more accurate.
• What particle would most likely give rise to an
  observed hit OM pattern in an event?
• You need the probability of a photon arriving at
  a time t at an OM, coming from a particle
  location. (Particle type is called the
  hypothesis , a muon or cascade for example )
• This probability is given by the various Pandel
  functions in AMANDA and IceCube.
• Llh reconstructions require a first guess
  reconstruction (called a seed)
• Llh sends seed to a minimizer, which wiggles
  particle parameters around until the likelihood
  is maximized.

9
llh reco modules in IceTray

• cscd-llh and muon-llh-reco
• Cscd-llh fits the vertex point and time of the
  cascade. (4 free parameters)
• Muon-llh-reco fits a vertex point and the
  direction of the muon. (5 free parameters)
• Both use Pandel functions to calculate timing
  probabilities.
• Minimizer provided by ROOTs TMinuit. (SIMPLEX
  and MIGRAD)

10
Using the LLH modules

• A LLH module requires the following things, set
  in the steering file
• Which first guess you use
• The input data-readout type (RecoHits vs.
  RecoPulses)
• There are many other configurable parameters.
  See documentation for details.

11
Future Reconstruction Plans Algorithms

• ReNew cascade first guess algorithm using
  waveforms
• Wf-llh-reco loglikelihood reconstruction of
  cascades and muons using waveforms.
• Ophelia High Energy muon tau reconstruction.
• Rime Reconstruction of Events using special
  wave forms feature.
• Photorec numerical PDF based on Photonics more
  accurate than Pandel.

12
Reconstruction chain
Save results
Read Event
Prepare data
Reconstruct
Read event files. Fill Geometry, Calibrate (if
necessary)
Process the data to get hits or pulses. Choose
strings, bad OM, LC modes .
Choose your favorite algorithm and go for it !
Create a root tree with the reco results, and
analyze! .

• Amasim f2k files
• IceSim Binaries
• Data (Binary)

13
Reco Series

• AmandaAnalogReadout LE, TE, TOT, ADC
• Reco pulse hit_ID, time, Amplitude, width
• Reco hit hit_ID, time

time Amplitude Width
t
14
IceTray survival guide

• Check the manuals
• Change the log level in log4cplus
• Add debug info throughout the code.
• Use log_debug(Entering loop with d and
  f,a,theta)
• Or even coutltltI am alive zenltltzenithltlt\n
• Dont forget to compile the module after adding
  the comment
• Dump the run output to a file and follow the
  messages and crash points
• Save and document a working steering file
  versions used.
• Try and try and retry
• Ask

15
Instructions for todays exercise are
here http//icecube.wisc.edu/hagar/icetray/rec
o-exercise.html Bon Appetite !