Source catalog generation

1
Source catalog generation
Jean Ballet, CEA Saclay
GSFC, 29 June 2005
Aim Build the LAT source catalog (1, 3, 5 years)

• Four main functions
• Find unknown sources over the whole sky (output
  list of positions).
• Localize sources. Output list of precise
  positions (and uncertainties)
• Characterize sources (significance, flux,
  spectrum, variability). This is no different from
  studying already known sources, and can be done
  using the likelihood method.
• Identify sources (find counterparts in existing
  external catalogs).

2
Catalog production pipeline
Remove space-ground transmission artifacts
Location LAT ISOC
Pipeline processing
Level 1 Database Reconstructed events Calibration
data Ancillary data
LAT Raw Data
Level 0 Database Raw events
Interstellar model (U5)
U6
U4
Location CEA Saclay
Source search
List of sources
Photon maps
Maximum likelihood (A1)
Exposure maps
A2 Source identification
LAT catalog (D5) List of sources and
identifications
List of sources Characterization
Catalog Database (D6)
3
Catalog pipeline. Sequence
Aim Implement automatic loop to find and
characterize the sources

• Minimal features
• Detect sources over current diffuse model
• Get a precise position
• Run Likelihood on all sources to get precise
  flux, spectrum and significance
• Split into time pieces to get light curve
• Run sourceIdentify to get counterparts

• Task scheduling tool (like OPUS) for distributing
  work over CPUs
• Simple database for bookkeeping and for the
  source lists

Associated product sensitivity maps in several
energy bands, or tool to provide minimum
detectable flux as a function of spectral index
and duration
4
Catalog pipeline. Schedule

1. Identify candidate source search algorithms. Done
2. Define evaluation criteria. Not yet concluded.
3. Build pipeline prototype. Good progress.
4. Evaluate candidate algorithms. Done on DC1.
5. First selection of source search algorithm. Not
   done yet. Wait until DC2.
6. Define processing database. By end 2005
7. Integrate pipeline elements (including flux
   history, identification). 2006.
8. Ready end 2006 (for DC3)

5
All-sky source search. Algorithms

• Wavelet algorithm developed at Perugia (F.
  Marcucci, C. Cecchi, G. Tosti)
• Wavelet algorithm developed at Saclay (MR_filter
  by J.L. Starck, presented in September 2004). On
  DC1, was quite comparable to 1.
• Optimal filter PSF/(1BKG) in Fourier space
  (myself , presented in September 2004). On DC1,
  did not find as many sources as 1 or 2 but
  returns source significance.
• Multichromatic wavelet method developed by S.
  Robinson and T. Burnett. Tries using the PSF at
  the energy of each photon. On DC1, did not find
  as many sources as 1 or 2 but still interesting
  for localization.
• Aperture photometry tried by T. Stephens as a
  reference. On DC1, did not find as many sources
  as the other methods but useful exercise.
• Voronoi algorithm proposed by J. Scargle. Was not
  ready for DC1.

6
Source localization

• Done locally (for each source in turn)
• Typical algorithm (like SExtractor) uses a
  smoothed map as input, and interpolates to find
  the maximum.
• Another possibility would be to use a different
  algorithm (like the multichromatic wavelet) to
  localize sources once they are detected.
• We need to provide a precision on source position

• Building TSmaps for all sources is certainly VERY
  CPU intensive.
• The precision depends mostly on the source
  spectrum (estimated by likelihood) and the source
  significance. This can be computed once and for
  all, and simulations can tell whether secondary
  parameters (like background shape) are important.

7
Pipeline prototype

• We have started with the likelihood step (most
  time consuming)
• Basis provided by J. Chiang (catalogAnalysis
  package, sourceAnalysis Python script). Chains
  event selection, exposure map generation and
  likelihood run.
• Use the OPUS task scheduler
• Try using OPUS in a minimal way (do not decompose
  too much) in order to facilitate portability.

• Standard region of interest (20 radius) contains
  many sources (several tens)
• Optimizers have trouble converging over so many
  parameters.
• Reduce area over which sources are fitted, taking
  nearby sources (fixed) from run on nearby region
  of interest.
• In addition, call likelihood in several steps
  (first diffuse emission, then add bright sources,
  then add faint ones).

8
What we plan to provide at DC2 kickoff
Run all-sky source detection Run likelihood over
all detections of first step Provide source list
(FITS format for the catalog, but not all columns
will be filled)

• Source_Name (just a number for internal
  reference), RA, Dec, Lon, Lat from all-sky source
  detection
• Test_Statistic, Flux (gt 100 MeV), Unc_Flux,
  Spectral_Index, Unc_Spectral_Index from likelihood

• What we will try to add during DC2
• Error box or ellipse
• Counterparts (using sourceIdentify)

9
Source catalog generation
Jean Ballet, CEA Saclay
GSFC, 29 June 2005
Catalog generation is on the way !
Several open points

1. Should we provide fluxes in a different band
   (starting at higher energy than 100 MeV) to
   optimize the relative error ?
2. Do we need a separate step for source
   localization and position error ?
3. Should we implement additional cuts on the data
   (e.g. on off-axis angle) ?