LAT and GBM Joint Spectral Analysis

This procedure provides all information needed to do the joint analysis with GBM (BGO and NaI) and LAT data.

Notes: Xspec is used as a spectral analysis tool in this procedure because it can handle multiple data sets and response matrices for the same spectral model, making joint analysis of GBM and LAT data possible. The current version of Xspec works on Windows under Cygwin.

Prerequisites

• Linux rh9, gcc 3.2

• ScienceTools v6r0p5, or later

• dataSubselector v3r6, or later

• evtbin v0r10p1, or later

• rspgen v0r6p0, or later  

• XSPEC (see Xanadu Data Analysis for X-Ray Astronomy)

Notes:

• Data files were obtained with observationSim, and the sky model includes:

• The Crab Pulsar

• All 3EG sources

• The extra galactic diffuse emission

• The GRB under study is the GRB simulated with the GRBobs model

• A few hundreds seconds (500) of simulated sky with observationSim (v5r8p3 ) were run, and the resulting FITS files are:

• test_events_0000.fits (Containing the LAT events)

• test_scData_0000.fits (Containing S/C information)

If you plot the "TIME" column of the first file you will find something like this, showing the LAT event data with the simulated bursts at 100 s.

The simulation produces also the following files which are needed by the GBM simulators:

• GRBOBS_050718001_DEF.txt
• GRBOBS_050718001_PAR.txt
• GRBOBS_050718001_GBM.txt

Note: These files are not needed for the analysis !!! The GBM team will take these files and produce the GBM output FITS file.

The GBM output files which you will need are located at:

http://www.slac.stanford.edu/exp/glast/workbook/sciToolsData/
latGbmJtSpectralAnal/GBM_BN050718001

You are now ready to start the analysis.

Steps:

1. Select LAT data

2. Bin the LAT data

3. Bin the GBM data

4. Compute the response matrix for the LAT

5. Use XSPEC for spectral analysis

1. LAT binned spectral analysis

2. Combined GBM+LAT spectral fit

1) Extract the LAT data (dataSubselector)

Go to dataSubselector executable directory, and:

This will produce the FITS file: GRB_050718001_LAT.fits

Notice that the procedure requires the "true" Ra and the Dec and the starting time of the GRB. You can use fv to inspect the original "LAT events" file for estimating them.

2) Binning the LAT extracted data (evtbin)

Go to the evrbin executable dir and:

This will produce the PHA1 FITS file: GRB_050718001_LAT.pha

3) Binning the GBM data (evtbin)

Let's start with the BGO 1 detector (label B1):

And then let's compute the NaI 6 detector (which is the one with more counts):

Now you  have also GRB_050718001_B1.pha and GRB_050718001_N6.pha PHA1 file.

Notice that the new evtbin allows NONE as spacecraft data!

4) Compute the response matrix for the LAT data (rspgen)

Using XSPEC for spectral analysis

You now have all the files needed to perform a spectral analysis using xspec. These include:

Notice that the LAT does not have any background file since it has a negligible background !

a) Spectral Binned analysis for LAT data: fitting the high energy component with a power law model

Another test that can be easily done is the LAT only spectral analysis, fitting the High Energy part of the GRB050718001 spectrum with a power law.

This is the xspec sequence of commands:

Notice that the power law index is in agreemant with the input value 2.25, with a good chi square!

b) LAT and GBM joint spectral fitting: the first GRB viewed in 6 decades of energy !

Start xspec, and set the plot device to x-window:

Load the data...

...the responses...

... and the backgrounds for the 2 GBM detectors:

Now you can plot the raw data (counts vs channels):

The results should be the following:

The color code is:
White: BGO data (B1)
Red: NaI data (N6)
Green: LAT data (LAT)

Set the x-axis in energy instead of channels:

Suggestion from Valerie Connaughton: BGO detector's effective area starts from 30 keV, but we now that the BGO operational mode is such that it electronically works at gains that enhance the high energy counts, at the expense of the low energy counts (measured by the NaI detector). Counts below ~150 keV should be thrown away. (first dataset)
Vice versa, we should throw out the NaI counts above 1MeV. (second dataset)

This is the result:

Now you can load the 4 parameters 'Band' model (grbm):

The high energy power law perfectly agree with the input parameter, and the reduced chi square is very good indicating the goodness of fit, and that the effect of the GBM data in constraining the parameters of the model.

The graphical output is the following:

Important!!
For the non x-spec users (like myself) the way to plot the results of xspec can be not obvious!
The data (dot with error bars) are NOT normalized by the effective area of the detectors, so that they represents the "raw counts" (divided by the exposure and by the energy bin width.
The model (in this case the Band function) is represented "folded" with the DRM. This means that in the above plot, the curve is ONLY ONE MODEL, as seen by different detectors. There is no reason to expect a continuity between one detector and the other. The only thing the observer should care is how well the data lie on the fit and how good is the chi-square (or other goodness estimators)!

For plotting purposes, you can choos to rebin the data, setting the minimum significance (in sigma) and the maximum number of bins which will be merged. My choice was to rebin only BGO and NaI data:

to produce this output: