Analysis of the 3C279 DC2 field

1
Analysis of the 3C279 DC2 field

• Rita Sambruna,
• Davide Donato,
• Dave Davis
• (GSFC)

2
Content

• We used the DC2 data to extract and analyze light
  curves and spectra from the 3C279 field (3C279,
  3C273, AGN 1)
• But first, we ran some simulations to gauge the
  performance of gtlikelihood and address a few
  questions
• A series of scripts were written to run in
  automatic the extraction procedure

3
What we expect GLAST will do for Blazars like
3C279

• Measure spectral shape in various states
• Measure inter- and intra-day flux and spectral
  variability
• Measure intervening and intrinsic absorption

Wehrle et al. 1996
4
Part I Understanding gtlikelihood

• Questions we wanted to address
• What is the error on the flux of a source if the
  wrong spectral model is used (e.g., power law
  instead of broken powerlaw)? How well can the
  input model be recovered?
• How well can the spectral parameters be
  determined as a function of flux?
• Can we determine if we are using the wrong
  spectral model for an individual source?

5
Simulations

• Individual source, bright faint (1/10 in flux)
• Source with a nearby (2.5º) companion
• - bright/bright (same flux)
• - faint/faint (same flux)
• - bright/faint (1/10 in flux)
• - faint/fainter (1/4 in flux)

6
Input Parameters

• 25 days simulation divided in 5 equally spaced
  bins.
• Bright source model
• bin 1 and 2 PL2.5, Flux0.4 and
  0.2 (arbitrary units)
• bin 3 PL11.5 PL22.5 Eb2 GeV,
  Flux0.6
• bin 4 and 5 PL1.5, Flux0.2 and
  0.3
• Companion model
• bin 1 and 2 PL1.5
• bin 3, 4 and 5 PL2.5

7
Summary of Results

• Unbinned Analysis
• a1) Error on flux (fractional distance from true
  value with true model)

DRMNGB
MINUIT
0-55 10-400
8
Summary of Results

• Unbinned Analysis
• a2) Error on flux (fractional distance from true
  value with WRONG model)

DRMNGB
MINUIT
5-60 30-500
The wrong model overestimates the flux
9

• b) Errors on spectral parameters (distance from
  true value)
• The BknPL model
• DRMNGB does not move from initial guess for
  break
• Indices within 0-50
• The PL Model
• DRMNGB Index within 0-100
• MINUIT index within 0-40

Bright Isolated source
10

• c) Determine spectral model
• For a bright source our simulations show that
• the wrong model yields suspicious values for
  some parameters and/or weird errors

11
Part II 3C 279 and its field
Catalog gives 12 sources In the 20 deg field
12
Analysis Objectives

• Extract light curves and spectral indices with 1
  day resolution for 3C279, 3C273, and AGN1
• 3C279 search for intra-day flux and spectral
  variability
• Hardness ratios
• Spectrum
• Rebinning the light curve

13
The Scripts

• Based only on Unix commands on a tcsh shell (can
  be easily converted also for a bash shell)
• Script 1 for each time-bin and energy-bin
• Time Space
• gtselect few sec
  tens kb
• gtlivecube few sec
  7.4 Mb
• gtexpmap 3 min
  1.2 Mb
• gtlikelihood 20-100 sec
  few kb
• Script 2
• reads results of gtlikelihood and
  converts them in readable ASCII files for plots,
  separating the values of different sources and
  different energy bins, for each source

14
Procedure

• First DRMNGB to find the parameter space
• Then MINUIT from the DRMNGB solution
• Data cut at lt 100 MeV (negligible effects)

15
1-day light curve 3C 273
GLAST-LAT DC2 source catalog (V1) at ASDC
16
1-day light curve 3C 279
GLAST-LAT DC2 source catalog (V1) at ASDC
17
Spectral variability
AGN1
3C273
3C279
18
DRMNGB
?
?
Large errorbars on several datapoints
19
MINUIT
No errorbars on several datapoints
20
Search for intra-day spectral variations

• General Methodology (RXTE, ASCA, XMM, etc.)
• Usually people extract a count rate light curve
  on a minimum timescale corresponding to a given
  S/N, and then search for flux variability
  rebinning the curve on various timescales (e.g.,
  with lcurve)
• DC2 data
• Run gtselect, gtlivecube, gtexpmap, and
  gtlikelihood for EVERY desired time and energy
  scale
• NOTE only relatively long timescales can be
  probed as the uncertainties on spectral
  parameters increase with decreasing S/N ?
  Hardness Ratios

21
3C279 Inter-day variability
First step Time-bin size of 3 hours for the
last 11 days
22
Hardness ratios (3 hrs)

• Light curves (gtlikelihood) in energy bands
• H 1-300 GeV
• S 100 MeV 1 GeV
• Ratio from lcurve

23
Rebinning the light curve
3 hours, from likelihood
8 hours, from likelihood
24
A spectrum
25
Conclusions 3C279

• Nearby (3 deg) variable source AGN1
• Likelihood highest fidelity for light curve
  spectrum (AGN1 contaminates 3C279)
• 3C279 shows flux variability on timescales of a
  few hours to gt 1 day
• No clear spectral variability trend (likelihood
  and Hardness ratios)
• AGN1 variable days flux/index
• 3C273 some flux variability

26
About gtlikelihood

• Powerful high-fidelity tool
• Relatively easy to use
• Long execution times for expo maps, cubes and
  large data volume generated
• MINUIT does not always generate errors
• DMRNGB often large errors
• More extensive documentation would help user
  avoid common pitfalls

MORE AS DATA ANALYSIS PROGRESSES
27
Old slides
28
EGRET High and Low states
TO BE DONE Run more realistic simulations
using fluxes and spectral indexes derived from
theoretical models
29

• d) Binned vs. Unbinned
• We used the binned likelihood analysis (DRMNGB
  optimizer) for 2 cases faint/faint and
  faint/fainter
• gtlikelihood does NOT give errors (Integral,
  Index, Energy break)
• The values for all the quantities do NOT change
  dramatically from the initial guesses
• In some cases the TS value is negative
• Are we doing something wrong?

30
Systematics

• We still do not understand why
• Some datapoints for relatively large flux have
  large error bars
• Large errors on index do not translate into large
  error on Integral
• We hope to find an answer using MINUIT optimizer