Presentation is loading. Please wait.

Presentation is loading. Please wait.

Likelihood analysis of small diffuse sources Riccardo Rando Elisa Mosconi, Omar Tibolla DC2 Kickoff Meeting – SLAC, 1-3 March 2006.

Similar presentations


Presentation on theme: "Likelihood analysis of small diffuse sources Riccardo Rando Elisa Mosconi, Omar Tibolla DC2 Kickoff Meeting – SLAC, 1-3 March 2006."— Presentation transcript:

1 Likelihood analysis of small diffuse sources Riccardo Rando Elisa Mosconi, Omar Tibolla DC2 Kickoff Meeting – SLAC, 1-3 March 2006

2 DC2 Kickoff, SLAC, 1-3 March 2006 2 Outline + Intro ● A simple source ● A simple analysis ● A more challenging source ● Other ideas ● Conclusions Small sources are those that cover a small fraction of the sky In this talk I will consider sources with a diameter of a few degrees Issues are the precision of diffuse response calculation within the likelihood pkg

3 DC2 Kickoff, SLAC, 1-3 March 2006 3 Simulation Let's consider a small squarish source, side 2 degrees A uniform spectrum index  =-1.6 (MapSource) Intensity ~1/6 Crab from 20 MeV up MapSource reads in map only for relative values: all pixels =1 within source boundaries, 10 -9 elsewhere Observation: 1 month Let's make things simple: it is placed at RA,DEC=0,0 : low (galactic) bkg

4 DC2 Kickoff, SLAC, 1-3 March 2006 4 MapSource From ST confluence page: <SpectrumClass name="MapSource" params="17.,2.1,$(FLUXROOT)/sources/gas_gal.fits,30.,2e5"/> ● Total photon flux from the map (#/m^2/s) = 17. ● Photon spectral index = 2.1 ● Fits file (plate-carree only) = $(FLUXROOT)/sources/gas_gal.fits ● Minimum photon energy (MeV) = 30. ● Maximum photon energy (MeV) = 2e5 File is a flat snapshot of the source Map is a projection of the sky, pixels are bins where integrated flux must be specified in ph/cm 2 /s/sr Anyhow, in the MapSource spectrum one sets the overall normalization, so only relative intensities are relevant

5 DC2 Kickoff, SLAC, 1-3 March 2006 5 FITS file RVAL1 = -179.75 CRPIX1 = 1.00 CTYPE1 = 'RA ' CUNIT1 = 'deg ' CRVAL2 = -89.75 CRPIX2 = 1.00 CTYPE2 = 'DEC ' CUNIT2 = 'deg ' CDELT1 = 0.5 / increment on x CDELT2 = 0.5 / increment on y

6 DC2 Kickoff, SLAC, 1-3 March 2006 6 Looking at the data Let's see at the data, in 0.5x0.5 deg 2 bins, RA-DEC projection Linear scale Somehow (WT, Poisson counts) one would have to obtain the image size/shape. Very simply, I cut the simulated events below 1 GeV, and here's the result 2 deg Binning 0.25 deg = no bkg Shape is pretty clear Model source map (1 on source, ~0 elsewhere) ~2400 photons in 1 month model

7 DC2 Kickoff, SLAC, 1-3 March 2006 7 Likelihood Now you know the drill... selection (from 30 MeV) livetimecube, expmap, diffresp, fit (see likelihood tutorial) Fixed extragalactic and galactic diffuse Try with both free & fixed spectral index X0fixed: Prefactor: 5104.12 +/- 123.639 Index: -1.6 Scale: 100 Npred: 2206.8 X0: Prefactor: 5212.46 +/- 243.929 Index: -1.56128 +/- 0.0169117 Scale: 100 Npred: 2426.05 This is pretty straightforward Maybe a little low? (-2.5  ) What about this?? (units: 10 -9 ph/cm 2 /s/MeV ) How does it relate to 0.03 ph/m 2 /s from 20 MeV?

8 DC2 Kickoff, SLAC, 1-3 March 2006 8 Making sense of prefactor We measured a prefactor of (5.2±0.2)×10 -6 ph/(cm 2 s MeV) Spectral index set to -1.56 This should be 1 if the map in the model file were a true representation of the source flux from 20 MeV! I put 1 arbitrarily in all source pixels, though. Hence, the map (2x2 deg 2, 1 ph/cm 2 /s/sr, 1.22 10 -3 ph/cm 2 /s total) is off by a factor 2.29·10 -3 Correcting the map value gives 2.29·10 -3 x 1.22·10 -3 = 2.79·10 -6 /cm 2 /s This means 2.8·10 -2 ph /m 2 /s, or 0.028 ph/m 2 /s So we recover an integral flux of (0.028±0.001) ph/(m 2 s) The true value was 0.03!! Looks OK 

9 DC2 Kickoff, SLAC, 1-3 March 2006 9 Uniformity? Assess source uniformity, in emission and spectral index?? Divided the source model into 2 2x1 deg 2 pixels: is the source uniform L/R? Do new diff.resp. and likelihood Look ~ uniform Split again: 4 pixels (UL,UR,DL,DR)... LR TOT X0left: Prefactor: 5863.79 +/- 1042.95 Index: -1.61463 +/- 0.0520546 Scale: 100 Npred: 939.53 X0right: Prefactor: 4931.39 +/- 601.263 Index: -1.5385 +/- 0.0356165 Scale: 100 Npred: 1486.85 I L =0.016 ± 0.003 ph/m 2 /s  L = -1.61 ± 0.05 I R =0.013 ± 0.002 ph/m 2 /s  R = -1.54 ± 0.04 

10 DC2 Kickoff, SLAC, 1-3 March 2006 10 More complicated Intensity spectral index vary along source All events, bins 0.5 deg Above 1 GeV, bins 0.25 deg X1: Prefactor: 7080.25 +/- 294.092 Index: -1.69109 +/- 0.0167291 Scale: 100 Npred: 2607.8 X1upright: Prefactor: 5471.25 +/- 1154.14 Index: -1.61747 +/- 0.062207 Scale: 100 Npred: 710.497 X1downright: Prefactor: 16465.4 +/- 1692.57 Index: -1.90758 +/- 0.0375423 Scale: 100 Npred: 1369.86 X1upleft: Prefactor: 2174.37 +/- 708.43 Index: -1.46277 +/- 0.0854636 Scale: 100 Npred: 236.447 X1downleft: Prefactor: 2976.89 +/- 725.928 Index: -1.47621 +/- 0.0663137 Scale: 100 Npred: 313.46

11 DC2 Kickoff, SLAC, 1-3 March 2006 11 More complicated : results I=0.005 g=1.6 I=0.01 g=1.7 I=0.015 g=1.8 I=0.01 g=1.7 I UL =0.003 ± 0.001 ph/m 2 /s  DL = -1.46 ± 0.08 I DL =0.004 ± 0.001 ph/m 2 /s  DL = -1.48 ± 0.07 I DR =0.024 ± 0.002 ph/m 2 /s  DR = -1.91 ± 0.04 I UR =0.007 ± 0.002 ph/m 2 /s  UR = -1.62 ± 0.06 Statistics is probably the limiting factor (2670 photons in source ~500 photons for DL,UR) True values measured

12 DC2 Kickoff, SLAC, 1-3 March 2006 12 Other ideas I investigated a region where galactic contribution is negligible When diffuse bkg is small/negligible another strategy can be useful to gain insight on the source characteristics before trying to model it = Select events from a certain energy up, to limit effect of PSF Make a countmap, with n B bins in RA & DEC, n E bins in energy Make exposure map with identical binning in RA,DEC,energy (Several energy bins are required to have a good approx.) Divide fits files bin by bin using, e.g., farith Map of “true flux” is obtained Select regions, plot log(Ng) vs log(E) Fit and obtain spectral index I found values within 5-10% of likelihood fit 1 3 2 1 2 3 /

13 DC2 Kickoff, SLAC, 1-3 March 2006 13 Conclusions ● Likelihood seems to work adequately with small diffuse sources ● Sources can be examined and compared with a given model ● Some more advanced investigation is feasible ● Still to assess impact of diffuse backgrounds

14 DC2 Kickoff, SLAC, 1-3 March 2006 14 Troubles with obsSim There seems to be a slight inconsistency on how obsSim and likelihood read the input fits file: MapSource was defined so that pixels fall at 179.5―180.0, 180.0―180.5... Apparently likelihood does this, but obsSim adds an offset I take the “real” data for the diff source only and split them left and right, to count photons. Compare with npred Offset on the DEC direction seems absent 180.0


Download ppt "Likelihood analysis of small diffuse sources Riccardo Rando Elisa Mosconi, Omar Tibolla DC2 Kickoff Meeting – SLAC, 1-3 March 2006."

Similar presentations


Ads by Google