Sub-MeV spectroscopy with AstroSat-CZT Imager for Gamma Ray Bursts
Tanmoy Chattopadhyay, Soumya Gupta, Vidushi Sharma, Shabnam Iyyani, Ajay Ratheesh, N. P. S. Mithun, E. Aarthy, Sourav Palit, Abhay Kumar, Santosh V Vadawale, A.R. Rao, Varun Bhalerao, Dipankar Bhattacharya
JJ. Astrophys. Astr. (0000) :
Sub − MeV spectroscopy with AstroSat − CZT Imager for Gamma RayBursts
Tanmoy Chattopadhyay , Soumya Gupta , Vidushi Sharma
2, 3 , Shabnam Iyyani , AjayRatheesh , N. P. S. Mithun , E. Aarthy , Sourav Palit , Abhay Kumar , Santosh VVadawale , A.R. Rao , Varun Bhalerao and Dipankar Bhattacharya Kavli Institute of Astrophysics and Cosmology, 452 Lomita Mall, Stanford, CA 94305, USA The Inter-University Centre for Astronomy and Astrophysics, Pune, India Department of Physics, KTH Royal Institute of Technology, AlbaNova, 10691 Stockholm, Sweden Dipartimento di Fisica, Universit`a di Roma Tor Vergata, Via della Ricerca Scientifica 1, I-00133 Roma, Italy INAF Istituto di Astrofisica e Planetologia Spaziali, Via del Fosso del Cavaliere 100, 00133 Roma (RM), Italy Dipartimento di Fisica, Universit`a La Sapienza, P. le A. Moro 2, 00185 Roma, Italy Physical Research Laboratory, Ahmedabad, Gujarat, India Indian Institute of Technology Bombay, Mumbai, India Tata Institute of Fundamental Research,Mumbai, India * Corresponding author. E-mail: [email protected], [email protected] received —; accepted —
Abstract.
Cadmium Zinc Telluride Imager (CZTI) onboard
AstroSat has been a prolific Gamma-Ray Burst (GRB)monitor. While the 2 − pixel Compton scattered events (100 −
300 keV) are used to extract sensitive spectroscopicinformation, the inclusion of the low-gain pixels ( ∼
20% of the detector plane) after careful calibration extendsthe energy range of Compton energy spectra to 600 keV. The new feature also allows single-pixel spectroscopy ofthe GRBs to the sub-MeV range which is otherwise limited to 150 keV. We also introduced a new noise rejectionalgorithm in the analysis (‘Compton noise’). These new additions not only enhances the spectroscopic sensitivityof CZTI, but the sub-MeV spectroscopy will also allow proper characterization of the GRBs not detected by
Fermi .This article describes the methodology of single, Compton event and veto spectroscopy in 100 −
600 keV for theGRBs detected in the first year of operation. CZTI in last five years has detected ∼
20 bright GRBs. The newmethodologies, when applied on the spectral analysis for this large sample of GRBs, has the potential to improvethe results significantly and help in better understanding the prompt emission mechanism.
Keywords.
AstroSat —CZT Imager—sub-MeV spectroscopy—Gamma Ray Burst.
1. Introduction
Cadmium Zinc Telluride Imager (hereafter CZTI) onboard
AstroSat (Singh et al ., 2014; Paul, 2013), India’sfirst dedicated Astronomical satellite, has been demon-strated as a prolific Gamma Ray Burst (GRB) monitor,since the launch of
AstroSat (Rao et al ., 2016; Chat-topadhyay et al ., 2019). CZTI is one of the two hardX-ray detectors sensitive in 20 −
150 keV. The instru-ment employs an array of CZT detectors, each 40 mm ×
40 mm × . Each detector is further segmented spatiallyto 256 pixels with a pitch of ∼ ∼ ◦ . Details of the payload design and function are given in Bhalerao et al . (2017) and Rao et al . (2016). Atenergies beyond 100 keV, the increasing transparencyof the collimators and the supporting structure enablesCZTI to work as an all-sky monitor. Because of thisall-sky sensitivity, CZTI instrument since the launch of AstroSat has been working as an e ffi cient GRB monitorwith around ∼
83 GRB detections per year .In last one year, we have explored a number of newtechniques in the spectral analysis for bright ON-axissources like Crab and Cygnus X-1 (Chattopadhyay etal. (2021), under preparation). We also identified anumber of possible improvements in the AstroSat massmodel for better spectroscopic and polarimetry analysis http: // astrosat.iucaa.in / czti / ?q = grb © Indian Academy of Sciences 1 a r X i v : . [ a s t r o - ph . H E ] F e b J. Astrophys. Astr. (0000) : for these sources. Implementation of these new tech-niques (listed below) will yield a significant improve-ment in the overall spectro-polarimetric sensitivity ofGRBs detected by the CZTI. • After the launch of
AstroSat , ∼
20 % of the CZTIpixels were found to have electronic gains sig-nificantly lower than the ground calibrated gainvalues. Majority of these pixels now possessesgain around 2 − ∼
60 keV for X-rayphoton detection but are also sensitive to photonsof much higher energies up to ∼
800 keV. We re-fer to these pixels as low-gain pixels, which wereoriginally excluded from any scientific analysis.However, after a careful and detailed analysis ofthe events from these pixels, here we attempt toinclude these pixels to increase the spectroscopicenergy ranges for GRBs. • From the detector plane histogram (DPH) imagesof the valid Compton scattered events, we furtheridentify the noisy pixels giving rise to 2-pixelevents. Filtering out this ‘Compton noise’ whichis otherwise not removed from the standard noiserejection algorithm.The new techniques allow us to explore the capa-bility of CZTI as a sub-MeV GRB spectrometer. Inthe standard CZTI analysis pipeline, the prompt emis-sion spectroscopy of the bursts are limited only in 100 −
200 keV whereas even with the 2-pixel Compton scat-tering events, the spectroscopy can only be extended upto ∼
350 keV. With the utilization of the low-gain pix-els, the spectroscopy of the GRBs are now extendedall the way up to ∼ ff erentways the CZTI instrument provides spectroscopic in-formation for the GRBs − (1) 1-pixel or single pixelevents from CZT detectors in 100 −
900 keV, (2) 2-pixel or Compton scattering events from CZT detec-tors in 100 −
700 keV which are used to extract po-larization information and (3) four CsI-Veto detectorsbelow the CZTI sensitive in 100 −
500 keV. We usethe
AstroSat mass model to generate the e ff ective areaas a function of energy and response matrix for eachof these spectroscopic techniques and perform broad-band spectral analysis along with Fermi and
Swift − BATdata. Proper spectral fits and constraining the spectralparameters critically depend on the correct estimationof response matrix elements which are di ff erent for dif-ferent GRB direction with respect to the satellite ori-entation. Although the mass model has been validated and tested in detail using imaging method (Mate et al., As-troSat satellite indirectly tests the mass model further.This also helps in identifying the shortcomings in someparts of the mass model and quantifying those fromthe spectral fits. CZTI sub-MeV spectroscopy is par-ticularly valuable for those GRBs which are detectedby
AstroSat and Niel Gehrels
Swift
BAT but not by
Fermi , as it allows us to constrain the spectral parame-ters including the peak energy in the energy range (15 −
900 keV), which otherwise generally is not possiblebecause of the narrow energy range of BAT. In this ar-ticle, we explore CZTI as a sub-MeV spectroscopy andreport the spectroscopic measurements for the elevenbright GRBs detected in the first year of CZTI operationwith the implementation of these new developments forthe entire burst time interval which is obtained usingthe Bayesian block technique on the GRB single eventdata. The new techniques and the burst selection meth-ods are described in section 2. In section 3., we describethe spectroscopy methods followed by broadband spec-tral analysis in section 4.. While this article primarilyoutlines the methodologies of sub-MeV spectroscopyfor GRBs, we plan to apply the new techniques to asample of ∼
20 bright GRBs detected in last five yearsof operation of
AstroSat .
2. New Techniques in the Spectrum Analysis
In this section, we describe the new techniques imple-mented in the spectral analysis compared to that dis-cussed in (Chattopadhyay et al ., 2019). In the previouspolarimetry reports on
AstroSat
GRB by Chattopad-hyay et al . (2019); Chand et al . (2018, 2019); Sharma et al . (2019), we utilized only 75 −
80 % of the CZTIcollecting area consisting only the ‘spectroscopicallygood’ pixels. A fraction of CZTI pixels are found tohave lower gains (gain value 3 − ∼ Characterization of low-gain pixels
From the detector plane histogram (DPH) of the on-board data, it was seen that the count rate in somespatially clustered pixels were significantly lower com-pared to the mean count rate (see Figure 1). Even . Astrophys. Astr. (0000) :
DET X D E T Y Figure 1 : The detector plane histogram (DPH) of allthe CZT detector quadrants for the obsID: 9000000618(data from 2016-August). The lower count rates de-tected in a fraction of the pixels are seen as patches inthe DPH which is because of the relatively higher gainvalues of the pixels. The color bars indicate the countrate.though most of the pixels are found in clusters, thereare instances of isolated pixels as well. These pixelsalso did not show the alpha tagged line at 60 keV fromthe on board calibration source
Am, indicating thatthe gain has shifted at least by a factor of two or three(hereafter we refer to these pixels as low-gain pixels).The reason for the shift is unknown, however since noshift was seen in the laboratory measurements duringcalibration and appeared right after the launch, mechan-ical stress during the launch is thought to be one of thepossibilities.From the light curve analysis from the low-gainpixels with di ff erent time bins, we found that the countrates detected by these pixels are of Poissonian natureand therefore the detected events are not spurious pixelnoise and could be real X-ray events. Because thesepixels consist almost 20 % of the CZTI active area, weexplored the possibility of characterizing the pixels indetail. In absence of any mono-energetic line at higherenergies to calculate the correct gain for the low-gainpixels, we compare the overlapping region of the con-tinuum spectrum of these pixels and the spectroscopi-cally good pixels. For that purpose we first fitted thegood pixel spectra for each of the 64 detector moduleswith an empirical model in 45 −
180 keV range usingthree Gaussian (Tantalum k α line at 54 keV, a bumpstructure around 65 keV source of which is unknownand a Tellurium activation line at around 88 keV) and abroken power law with a break energy around 140 keV
60 80 100 120 140 160 180
Energy (keV) C o un t s / b i n ( 55.70 ) ( 87.58 ) ( 143.50 ) Observed DataStatistical Modelspectral features
Figure 2 : Continuum spectra from the spectroscopicallygood pixels (in blue) for one of the detector modules(data taken from July 2016). The spectrum is fitted withan empirical model (red) consisting of three Gaussian(1. Tantalum line at 54 keV, 2. a bump structure nearthe Tantalum line and 3. an arbitrary line around 90keV which is most likely a proton induced backgroundfeature (Odaka et al ., 2018)) and a broken power law(break energy around 140 keV which denotes the onsetof falling detection e ffi ciency for the 5 mm thick CZTdetectors). This template has been used to compare thespectra of the low-gain pixels to estimate their gains(see text for more details).as shown in Figure 2. The break energy denotes theonset of falling detection e ffi ciency for a 5 mm thickCZT detector and therefore can also be used to calibratethe low-gain pixels along with the continuum compar-ison. The strong line around 88 keV seen in the spec-tra is supposed to originate from high energy particleinduced Tellurium activation ( m T e with half life of9.17 × seconds (Odaka et al ., 2018)). We also see ahint of a line feature around 145 keV which could alsobe from activated Tellurium ( m T e with half life of4.96 × seconds). Because of the large half life of theisotopes, we see the lines even far from the SAA regionwhere the activation is supposed to take place (Odaka et al ., 2018). Since the number of good pixels vary ineach module, the count rate was normalised by the totalnumber of good pixels in that module.In order to have su ffi cient statistics in the spec-tra of both good and low-gain pixels, we took a longone month data ( ∼ J. Astrophys. Astr. (0000) :
60 80 100 120 140 160 180 C o un t s / b i n (55.86) (87.53) (141.85) Observed DataStatistical Modelspectral features
60 80 100 120 140 160 18010 (55.86) (87.53) (141.85) Gain = 0.99
60 80 100 120 140 160 180 C o un t s / b i n (55.57) (87.04) (142.50) Observed DataStatistical Modelspectral features
60 80 100 120 140 160 18010 (55.57) (87.04) (142.50) Gain = 1.0
60 80 100 120 140 160 180 C o un t s / b i n (54.88) (86.64) (144.08) Observed DataStatistical Modelspectral features
60 80 100 120 140 160 18010 (54.88) (86.64) (144.08) Gain = 1.03
Energy (keV)
Figure 3 : Spectra of three Low-Gain pixels of type I be-fore and after applying gain correction given in left andright panel respectively. top- pixel number 161 frommodule 4, middle- pixel number 248 from module 5,and bottom- pixel number 45 from module 13. Thered lines are the empirical models used to compare theoverlapping region of 45 −
180 keV of the low gainpixels. After comparison, the fitted gain shift factors (amultiplication factor to the ground calibrated gain) arefound to be between 0.8 and 1.5 for the type I low gainpixels.fitted module wise good pixel models were then usedto compare the spectra of each of the low-gain pixels in100 −
180 keV range and reduced χ values were cal-culated by varying a multiplication factor to the groundcalibrated gain of the low-gain pixels in the range of 0.8 − −
1. Low-Gain pixels type I: These pixels were foundto have gain shift factor between 0.8 and 1.5 andare seen to have spectral features like the Tan-talum line, the 90 keV background and spectralbreak at around 140 keV as also seen in the goodpixel spectra. Left column of Figure 3 shows the comparison of the spectra (shown in blue)with the good pixel model (shown in red) forthree pixels whereas the right panel shows thesame comparison after correcting for the gains ofthe pixels. These pixels were previously identi-fied as low-gain pixels in CALDB. Since we findthem to have gains very close their ground cal-ibration values, we plan to calibrate them withthe on board calibration source for further valida-tion, details of which will be presented elsewhere(Mithun et al., in prep.).2. Low-Gain pixels type II: These pixels were foundto have relatively higher gain shift values be-tween 1.5 and 5.0. Comparison of the continuumspectra in 100 to 180 keV before and after gaincorrection is shown in Figure 4.3. Low-Gain pixels type III: For a fraction of pixelswe could not get satisfactory fit in the common100 −
180 keV range even for the maximum gainshift values. These pixels are ignored from anyfurther analysis.We carried out the analysis for each of the five yearsof CZTI data (normally June / July of each year depend-ing on the data availability) to check for repeatabilityor any possible time evolution in the obtained gain val-ues for the type I and type II low-gain pixels. We usethe gain list from the year of detection of a given GRB(note for this paper we use gain list of the year 2016 asall the GRBs analyzed here are detected in 2016). In fu-ture, we plan to characterize the low-gain pixels (partic-ularly the type II pixels) using various particle-inducedradioactivation background lines (Odaka et al ., 2018)for further verification. It might be possible to furtherverify the gain values by looking at the Crab pulse pro-file from these pixels and calculate the ratio of pulsedfractions in two pulses as they are known to be energydependent. We also plan to validate the gain values ofthe type I pixels by investigating the alpha tagged spec-tra from these pixels.In order to boost the confidence in the use of low-gain pixels, we attempted to reconstruct the Crab pulseprofile using these pixels after gain correction. Figure 5shows the pulse profile of Crab pulsar in low-gain pix-els from all the four CZTI quadrants during a ∼
78 ksobservation on 14 th Jan 2017. This further verifies thatthe events from these pixels are genuine X-ray eventsand not random noise. We used the crab ephemeris atMJD 57769.0 from Lyne et al . (1999). The events arefolded from
AstroSat time of 222220803.426 seconds.The background is subtracted by the counts in the o ff pulse region and the pulse profile is normalised by themaximum peak counts for the purpose of visualization.We could also detect the GRBs in these pixels as shown . Astrophys. Astr. (0000) :
60 80 100 120 140 160 180 C o un t s / b i n (54.88) (86.64) (144.08) Observed DataStatistical Modelspectral features
130 140 150 160 170 18010 (144.08) Gain = 2.54
60 80 100 120 140 160 180 C o un t s / b i n (55.57) (87.04) (142.50) Observed DataStatistical Modelspectral features
100 120 140 160 18010 (142.50) Gain = 2.8
60 80 100 120 140 160 180 C o un t s / b i n (55.57) (87.04) (142.50) Observed DataStatistical Modelspectral features
100 120 140 160 18010 (142.50) Gain = 4.9
Energy (keV)
Figure 4 : Same as Figure 3 but for three type II Low-Gain pixels (top- pixel number 223 from module 13,middle- pixel number 0 from module 5, and bottom-pixel number 1 from module 5). For type II low-gainpixels, we found the fitted gain shift factors between1.5 and 4.in Figure 6 for GRB 160821A. Since the number oflow-gain pixels vary in di ff erent quadrants, the countrate is normalized by the total number of low-gain pix-els in a quadrant. Detection of astrophysical sources inlow-gain pixels therefore presents a strong case in usingthem for future spectral analysis.2.2 Compton Noise
CZTI has already been demonstrated as a sensitive ON-axis and GRB polarimeter in 100 −
350 keV in Vadawale et al . (2018) and Chattopadhyay et al . (2019) respec-tively, where the Compton scattered events are usedto generate the azimuthal angle histogram. The sameCompton events can be used in spectroscopy of theGRBs. These events are selected through strict Comp-ton kinematics criteria − • identify the adjacent 2-pixel events from 20 µ scoincidence window both from the spectroscopi-cally good pixels and low-gain pixels, Pulse Phase N o r m a li s e d I n t e n s i t y low gain pixelsnormal pixels Figure 5 : The pulse profile of Crab pulsar in low-gainpixels (blue) of all the CZT quadrants after gain correc-tion. For comparison the pulse profile in the spectro-scopically good pixels are plotted against it (red). • impose criteria of ratio of the energies depositedin two pixels between 1 and 6 in order to filterout the noisy chance events. This is motivated bythe fact that in a true Compton scattering event,the electron recoil energy deposited in one of thepixels, is much lower than the scattered photonenergy deposited in the other pixel.In spite of the strict selection criteria, there is still asignificant amount of overlapping noise events. Neigh-bouring pixels can flicker at time scales lower than thecoincidence time window of 20 µ s causing some ofthese events to permeate into the Compton event se-lection and thus causing instrumental artifacts in themodulation curve. A DPH showing outliers in 2-pixelevents is shown in Figure 7. These events can beidentified as outliers from the DPH of neighbouring 2-pixel events and can be removed from further analysis.Threshold for an outlier is kept at four sigma and threesigma from the mean for normal and low-gain pixelsrespectively. Due to the di ff erence in count rate be-tween the side and corner pixel double events, we iden-tify the noisy pixels in the side and corner pixels sep-arately. When a pixel is identified as noisy, no eventsfrom that pixel is considered for further analysis. Fur-ther details on the Compton noise analysis can be foundin Ratheesh et al., Selection of burst interval
In this work, the spectrum analyses are conducted onthe time integrated emission of the bursts. The time in-terval corresponding to the integrated emission is cho-
J. Astrophys. Astr. (0000) :
Time (sec) C o un t s s e c p i x e l Q0Q1Q2Q3
Figure 6 : Light curve of GRB 160821A in the low-gain pixels with corrected gains. Di ff erent colors rep-resent the four di ff erent CZTI quadrants as indicatedinside the plot. The time axis is plotted from AstroSat time 209507728 seconds (marked as zero). Each CZTIquadrant is shadowed by di ff erent degree for each GRBaccording to its location with respect to the spacecraftgiving rise to unequal flux levels in di ff erent CZTIquadrants.sen by employing the Bayesian block algorithm (Scar-gle, 1998; Scargle et al ., 2013; Burgess, 2014) of timebinning on the single pixel event data of the bursts. Theblock with the minimum probability density value cor-responding to the background region is taken as theguide to decide the start and stop times of the integratedemission. The onset time of the first block with theprobability density greater than that of the backgroundwhich is closer to the onset time of the burst and the endtime of the last block after which the background con-tinues are considered as the start and stop times of thetime interval of integrated emission respectively (Fig-ure 17 in Appendix).In the next section, we describe the methodology ofspectroscopy using 1-pixel and 2-pixel CZTI events andCsI-veto detected events followed by broadband spec-troscopy results for the eleven bright GRBs detected in2015 −
3. Methodology for Spectroscopy
Spectroscopic response
The 2D spectral responses for CZTI 1-pixel, 2-pixel and CsI / veto spectroscopy are generated using DET X D E T Y Figure 7 : Detector plane histogram of the neighbouring2-pixel Compton events for the 3 rd CZTI quadrant. Theplotted data belongs to obsID: 9000000618 (data from2016-August). The colorbar indicates count rate. Thebrighter spots in the image correspond to the Comp-ton noisy events arising from noisy neighboring pixels.These events are removed from further analysis.GEANT4 simulation. Here we outline the basic stepsof response generation. Response is computed usingGEANT4 mono-energy simulations of the full
AstroSat mass model at specific θ and φ viewing angles for eachGRB ( θ and φ for a given GRB are provided by either Swift or Fermi ). The mono-energetic lines for simu-lations were selected between 100 keV and 2 MeV atevery 20 keV till 1 MeV and at every 100 keV in 1 − pho-tons for each energy) in order to have a statisticallysignificant energy distribution in CZTI for each mono-energetic line. The simulation file contains informa-tion of total seven interactions or steps for each incidentphoton (x, y, z-position of interactions in CZTI and de-posited energy in each interaction, see Chattopadhyay et al . (2014)) in CZTI modules. We add up the energiesfrom all the interactions happening within a pixel of2.5 mm × ) based routine outside the GEANT4. We apply thesame CZTI pixel-level LLD (Lower Level Discrimina-tor) values in the simulation data whereas the ULDs(Upper Level Discriminator) were computed from ac-tual observational data for each module and is appliedto simulation data accordingly. From this event list, the1-pixel and 2-pixel events are separated and processed Research Systems, Inc. (1995). IDL user’s guide : interactive datalanguage version 4. Boulder, CO :Research Systems . Astrophys. Astr. (0000) :
50 0 50 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB151006A
50 0 50 100 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160106A C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160131A
50 0 50 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160325A C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160509A
20 0 20 40 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160607A
20 0 20 40 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160623A
50 0 50 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160703A
580 600 620 640 +2.57e6 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160802A
100 150 200 C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160821A
20 0 20 40
Time (s) C o m p t o n E v e n t s C o un t s S i n g l e E v e n t s C o un t s GRB160910A
Figure 8 : GRB light curves from CZTI 1-pixel (red) and2-pixel events (black points) for the 11 GRBs. The timeintervals of the bursts are obtained from the Bayesianblock analysis on the 1-pixel CZTI light curves asshown by the vertical dashed lines. The ‘zero’ denotedin the time-axis stands for the trigger time reported bythe
Fermi -GBM. di ff erently for final response generation. For 1-pixelevents, the distribution of deposited energies is calcu-lated at a bin size of 1 keV from 0 keV to 1000 keV(total 1000 bins) for each of the 56 mono-energies.It is to be noted that Geant4 simulation takes careof all types of interactions with appropriate probabil-ities including photoelectric, Compton, Rayleigh in-side CZTI and photons scattered from the spacecraftor other surrounding payloads to CZTI. Because ofthese multiple interactions and scattered events fromsurrounding materials, the distribution of deposited en-ergy in CZTI is broad and non-gaussian. However,the large number of photon simulation gives su ffi cientstatistics to obtain the correct energy distribution in thefull range of 100 bins for all 56 mono-energies. The 2Dmatrix (56 × AstroSat mass model sim-ulation data where we only consider events and asso-ciated energies deposited in the CsI detectors to esti-mate the deposited energy distribution. It is to be notedthat we use a µτ and charge di ff usion based line profilemodel (Chattopadhyay et al ., 2016) for mask weightedresponse below 150 keV, whereas for this work, we usea simple Gaussian model for simplicity.3.2 Single Pixel (1-Pixel) Spectroscopy
Because the CZTI surrounding structures and the colli-mators become increasingly transparent above 100 keV,the spectral analysis of the GRBs starts from 100 keVand extends to 900 keV after incorporating the low-gainpixels. Detection e ffi ciency of a 5 mm CZT drops be-low 10 % above 1 MeV resulting in a low signal to noiseratio at those energies. The single pixel events are se-lected such that there are no other events reported in100 µ s time window on either side of the event. En-ergies deposited in all such events in the full burst re-gion (interval obtained from Bayesian block analysis)are used to generate the spectrum with a 10 keV bin-ning.The 1-pixel light curves and the selected time inter-vals are shown in Figure 8 (red solid lines). The back-ground spectrum is constituted by selecting at least 300seconds of time window from the pre and post-burst re- J. Astrophys. Astr. (0000) : gions.We quantify the systematics in the 1-pixel spectraldata arising due to the uncertainties and inaccuraciesin the AstroSat mass model and the CZTI detector viathe analysis of the spectral data of GRBs detected atdi ff erent incoming orientations. We use Band model(Band et al ., 1993) to fit the spectra while keeping thepower law indices ( α and β ) and peak energy ( E peak )frozen at the values reported by either the Konus Wind or Fermi spectral analysis, and the normalisation of theBand model is left free.For the di ff erent GRBs, listed in Table 1, detectedat di ff erent orientations including most of the incomingangles around the spacecraft, once obtaining a “best”fit, we find unresolvable discrepancies between calcu-lations and the data, i.e., systematic errors. Withoutknowing or assuming the origin of these features, wecharacterize the e ff ect by adding systematic errors in anincremental fashion until we achieve a uniform residualwith a reduced χ <
2. We, therefore, add 10 −
15 %systematic to the 1-pixel spectroscopic data for all theGRBs to take care of the inaccuracies in the
AstroSat mass model.For example, the spectral fits with the respectiveresiduals obtained for the GRB 160623A (left) andGRB 160802A (right) that are detected on either sideof the spacecraft are shown in the Figure 9 where 1-pixel spectra in 100 −
900 keV obtained from the fullburst region are shown in red crosses.3.3
Compton (2-pixel) Spectroscopy
The Compton spectroscopy is carried out in the energyrange of 100 −
700 keV since above 700 keV there isno su ffi cient Compton scattering e ffi ciency of CZTI de-tectors. The 2-pixel Compton events are identified fromadjacent pixel events within 20 µ sec coincidence win-dow with an additional Compton kinematics criteria ofratio of two deposited energies between 1 and 6 (alsodiscussed in Chattopadhyay et al . (2014)). The ener-gies from the two events recorded are added up to getthe total energy and therefrom the spectrum with a binsize of 10 keV.The systematics involved in the Compton spectraldata are assessed using the same methodology adoptedfor 1-pixel spectral data (section 9). The spectral fitsand the residuals obtained for the Compton spectra ofGRB 160623A and GRB 160802A in 100 −
700 keVare shown in red data points in Figure 9. Similar to The CZTI spectral data fit is considered to be reasonable when(a) the obtained residuals are roughly randomly distributed aroundzero, (b) the reduced chi-sq χ < Konus-wind and
Fermi spectral analysis. the 1-pixel spectra, we find reasonable fit to the dataand agreement with
Fermi norm by adding a systematicof 10 −
15 % uniformly throughout the energy range,100-700 keV. Therefore, this systematic is added to the2-pixel spectral data of all the GRBs.3.4
CsI (or Veto) spectroscopy
There are four CsI(Tl) scintillator detectors (each 167mm ×
167 mm in size and 2 cm in thickness) belowCZTI quadrants to veto the high energy particle in-duced background events reported in both CZTI andCsI detectors (Bhalerao et al ., 2017; Rao et al ., 2016).The veto detectors were initially not meant for spec-troscopy. However since the detectors possess su ffi -cient detection e ffi ciency in the sub-MeV region, weexplored the possibility of using them for spectroscopyto enhance the overall spectroscopic sensitivity. Theexisting CZTI pipeline provides the veto spectrum atevery second. We employ the available data to generatespectrum for each Veto detector in a similar way thatis used for CZTI single pixel events. However, we donot use the poorly calibrated 4 th Veto quadrant for spec-troscopy. It is to be noted that the Veto spectrum con-sists of all interactions in the CsI detectors and di ff erentfrom the Veto tagged events where the both CZTI andVeto are triggered due to simultaneous events recordedin those detectors.For all the GRBs detected from the rear side ofthe spacecraft, we find the observed spectra to be flat-ter than the response folded model. An example isshown in the top-left spectral plot of Figure 10 forGRB 160623A which is detected at θ of ∼ ◦ . Wefind an identical systematic trend in all the back sideGRBs. However, we do not attribute the systematic tothe mass model as CZTI 1-pixel and 2-pixel spectralfits for back side GRBs do not show such systematictrend in the residuals. On the other hand, the trend issignificantly lower in Veto detectors for the front sideGRBs. Therefore we believe that this systematic isoriginated in the CsI detectors but primarily for detec-tions from back side. CsI detectors are scintillator de-tectors where the scintillation light is collected by thePMTs (2 PMTs for each of the 4 CsI detectors). Atlower energies ( ∼
100 keV), the number of scintillationphotons generated is lower than that at higher energies.Given the fact that there are only two PMTs to col-lect the scintillation photons, the detection probabilityof the GRB photons at lower energies is expected tobe relatively low. We also note that the detectors wereinitially not meant for spectroscopy and therefore thenumber of readout photo-multiplier tubes and opticalcoupling between the crystal and the photo-multipliertubes (PMTs) were not optimized to enhance the detec-tion probability. The light collection e ffi ciency might . Astrophys. Astr. (0000) : Figure 9 : Left: Example of count spectra (upper panel) and residuals (lower panel) obtained for the 1-pixel (red)and 2-pixel (black) CZTI events for one back Side GRB (GRB 160623A detected at angle ∼ ◦ , left panel) andone front side GRB (GRB 160802A, viewing angle ∼ ◦ , right panel). We fit the spectra with Band model keepingthe spectral parameters frozen at the values reported in literature to check for the consistency in spectral shapewith Fermi and
Konus-Wind .be significantly compromised for events happening inthe back side of CsI because of the absence of opticalreflecting coating on the back surface and a relativelyhigher level of cover shielding on the back side nearthe PMTs (light collecting area is relatively lower onthe back side).To take care of this, we multiply the photon de-tection probability (represented by an empirical term,1 − e − Energy / E ) to the model (same as multiplying tothe CsI detector response) to mimic for an energy de-pendent systematic where the value of E − depends onthe location of transient observed with respect to CZTI.For the front side GRBs (example shown in the bot-tom panel of Figure 10 for GRB 160802A) i.e theta < ◦ the value of E − is found to be around 0.01 keV − which gives 90 % detection probability at ∼
200 keV,whereas for the orthogonal GRBs i.e 90 ◦ < theta < ◦ , the value of E − comes out to be around 0.008keV − . For the back side GRBs, value of E − is foundto be around 0.0045 keV − signifying poor detectionprobability (90 % detection probability at ∼
600 keV).Since we get similar values of E − for front, backand orthogonal GRBs, we plan to incorporate the ex-ponential feature observed in the Veto detectors in theresponse itself. We also include an additional 5 % sys-tematic in the data in case of back side GRBs.
4. Results: Broadband Joint Spectroscopy of GRBs
With a fair assessment of the systematics present inthe CZTI and Veto spectral data, we now conduct thebroadband joint spectral analyses involving the spectraldata from
Fermi , Niel Gehrels
Swift
BAT, along with CZTI data including the single, Compton and Veto forthe time integrated emission of di ff erent GRBs. Weanalyse the time integrated spectrum of 10 GRBs thatwere detected by CZTI in the first year of its operation(2015 − Fermi spectral data includes two bright sodiumiodide (NaI) detectors with source angle less than ( < ◦ ) and the brightest bismuth germanate (BGO) de-tector (Gruber et al ., 2014). In case of GRB151006A,GRB160509A and GRB160821A, the low energyLarge Area Telescope (LLE) data are also used. The Fermi spectral files are extracted using Fermi BurstAnalysis GUI v. 02-03-00p33 (gtburst ). The SwiftBAT spectral files are prepared by the standard method-ology .The spectral analyses are performed using the X-Ray Spectral Fitting Package ( XSPEC , Arnaud 1996)version: 12.11.0 and have followed chi-square statis-tics. Both BAT and CZTI spectral files are compatiblewith Gaussian statistics, however, the GBM and LLEfiles are consistent with Pgstat wherein the backgroundand signal are assumed to be Gaussian and Poissonianrespectively. Therefore, using Heasoft Ftool
GRPPHA ,we rebinned both GBM and LLE spectral files such thateach energy channel contains a minimum of 20 pho-tons.The spectral fit results and the respective residu-als obtained for the best fit empirical functions like https: // fermi.gsfc.nasa.gov / ssc / data / analysis / scitools / gtburst.html https: // swift.gsfc.nasa.gov / analysis / threa ds / bat threads.html https: // heasarc.gsfc.nasa.gov / ftools / caldb / help / grppha.txt J. Astrophys. Astr. (0000) :
Figure 10 : Top left: the count spectra (upper panel) and their respective residuals (lower panel) obtained forthe three quadrants of the Veto detectors (black: quadrant A, red: quadrant B and green: quadrant C) for GRB160623A (detected from the back side of CZTI). We see a systematic trend in the residuals possibly due to lowerdetection probability by the scintillators around 100 keV which improves at higher energies; top right: Same asthe left figure but after implementing an energy dependent correction (1 − e − Energy / E ) where E − = . − (see text for more details); bottom: same as the top figure but for GRB 160802A (detected from the front side)after implementing the energy dependent correction with a higher value of E − = .
01 keV − . . Astrophys. Astr. (0000) : Band function (
Band ) and cuto ff power law ( CPL ) arereported in Table 1 and shown in Figure 11, 12 and14 respectively. We find that residual obtained forCZTI spectral data are consistent with those obtainedfor
Fermi . The residuals are found within 3 σ for CZTIdata.The small energy window of Swift
BAT (15 −
150 keV) generally does not allow us to constrain the E peak of the spectrum in cases where there is only BATdetection. In case of GRB 160607A, GRB 160703Aand GRB 160131A, where Fermi detections were notavailable, we conducted the spectral analysis using
Swift
BAT along with CZTI data. We demonstrate thatthe usage of CZTI data extending until 900 keV allowsus to well constrain the E peak of the spectrum. Thee ff ective area correction factor obtained between BATand CZTI are shown in Figure 13, where constant forBAT is frozen to unity. The energy flux estimated in therange of 10 − ff erentdetectors, we have tied the spectral parameters of allthe detectors including the normalisation of the spec-tral model. The di ff erence in count rates in di ff erentdetectors are taken care of by including the e ff ectivearea correction factor along with the spectral model thatis used to analyse the data. To estimate the e ff ectivearea correction factor between the di ff erent detectors,we multiply an energy independent constant factor tothe spectral model during the fitting process. The e ff ec-tive area correction factor obtained between Fermi andthe di ff erent datasets of CZTI except in GRB 160131A,GRB 160607A and GRB 160703A where the valuesare obtained with respect to the Swift
BAT are shownin Figure 13. On average, the normalization estimatesof the empirical function fits done to the single, Comp-ton, Veto 1, 2, and 3 data are found to vary around 20%,55%, 55%, 40% and 40% of the normalization estimateof the brightest
Fermi
NaI detector respectively. Whilewith respect to the BAT detector, the normalization esti-mates for the single, Compton, Veto 1, 2, and 3 events,vary around 40%, 140%, 85%, 20% and 70% respec-tively. In certain GRBs, we observe low normalizationsfor Compton and Veto data which results in an e ff ectivearea correction factor >
2. The cause of such cases arebeing studied.For GRB 151006A and GRB 160325A, both
Fermi and BAT observations are available. So, in these GRBs,we conduct a joint spectral analysis of BAT and CZTIdata and then compare the spectral fit results withthat obtained using
Fermi
GBM data alone. We areable to ascertain the α , E peak and normalization valueswhich are reasonably consistent with Fermi
GBM re-sults, within 90% error limits (Table 2 and Figure 14). This further endorse the capability of CZTI as a sub-MeV spectrometer along with BAT to determine theGRB spectrum.We note here that being opaque below 100 keV,CZTI spectrum alone cannot measure the GRB spec-tral parameters fully. On the other hand, if we assumecanonical values for the power law indices ( α = − β = − .
5) of the Band function, we can constrainthe E peak and normalisation of the spectrum. In cer-tain cases, the E peak estimates are found to lie close tothe edge or outside the energy window of CZTI (e.gGRB160509A and GRB160821A). In Figure 15, theenergy fluxes estimated in the energy range 100 keV − Fermi data only spectral fits (where allthe fit parameters are left free) of the di ff erent bursts.We find the CZTI flux estimates are consistent within2 σ scatter around the line denoting CZTI energy fluxis equivalent to Fermi flux.
5. Summary and future plan
CZT-Imager on board
AstroSat has been a prolific GRBmonitor with around detection of nearly 83 GRBs peryear. I In this article, we explored the spectroscopicsensitivity of CZTI in the sub-MeV region by attempt-ing spectroscopic analysis for some of the bright GRBsdetected in the first year (October 2015 − Spetember2016) of
AstroSat operation. The improvement in thespectroscopic sensitivity has been possible because of(1) inclusion of the low-gain CZTI pixels after a thor-ough calibration which consists of around 20 % of theCZTI detection area, (2) identification and removal of2-pixel noisy events. Both the methods improve theS / N of the bursts significantly and in particular the low-gain pixels enable the spectroscopy all the way up to900 keV (1-pixel Compton spectroscopy: 100 − −
900 keV) . We alsoutilize the CsI (or Veto) detectors for spectroscopy in100 −
500 keV to enhance the overall sensitivity.In Section 4., we performed joint
Fermi and
As-troSat (and BAT wherever available) spectral analysisfor 10 out of the eleven first year GRBs (except GRB160623A where a concurrent observation with
Fermi was not available) in the full burst region. We are ableto obtain spectral fit parameter values that are in closeagreement with those obtained in solo
Fermi analysis. The scatter is the standard deviation of the Gaussian fit to the dis-tribution of the displacement of the CZTI measured flux from the
Fermi flux and is found to be σ = . J. Astrophys. Astr. (0000) :
Table 1 : Spectral fit results of the analysis of the time integrated emission of the bursts in the sample using theCZTI,
Fermi and Niel Gehrel
Swift
BAT data.GRB name T start T stop α β E peak / E cut log (Flux) Chi-square / Model Fit Other(s) (s) (keV) (erg / cm / s) DOF InstrumentsGRB151006A -1.11 30.21 − . + . − . − . + . − . + − − . ± .
005 664 . /
772 Band GBM + LAT + BATGRB160106A 0.41 47.26 − . ± . − . + . − . + − − . ± .
005 520 . /
697 Band GBMGRB160131A 11.95 49.63 1 . ± .
03 – 407 + − − . + . − . . /
436 Cuto ff pl BATGRB160325A -1.31 46.41 − . + . − . − . + . − . + − − . ± .
004 717.06 /
768 Band GBM + BATGRB160509A 7.63 22.06 − . ± . − . ± .
01 279 + − − . ± − .
001 896 . /
709 Band GBM + LATGRB160607A -0.72 14.96 0 . + . − . − . ± .
05 108 + − − . ± .
003 250.34 /
420 Band BATGRB160703A -2.24 28.31 − . + . − . − . + . − . + − − . ± .
004 264.04 /
423 Band BATGRB160802A -0.36 17.84 0 . ± .
02 - 263 + − − . ± .
002 743 . /
690 Cuto ff pl GBMGRB160821A 113.47 159.87 − . ± . − . ± .
02 860 + − − . ± . . /
699 Band × Highecut GBM + LATGRB160910A 3.27 15.04 − . + . − . − . + . − . + − − . + . − . /
680 Band GBMThe errors are reported for 68% confidence interval. The references for the burst detection in di ff erent instru-ments are provided. GRB151006A - GBM (Roberts & Meegan, 2015), LAT (Ohno et al ., 2015), BAT (Cummings et al ., 2015); GRB160131A - Fermi-GBM trigger number 473813134; GRB160325A - GBM (Roberts, 2016),LAT (Axelsson et al ., 2016), BAT (Lien et al ., 2016a); GRB160509A - GBM (Roberts et al ., 2016), LLE (Ko-cevski & Longo, 2016); GRB160607A - BAT (Lien et al ., 2016b); GRB160703A - BAT (Lien et al ., 2016c);GRB160802A - GBM (Bissaldi, 2016); GRB160821A - GBM (Stanbro & Meegan, 2016), LAT (McEnery et al .,2016); GRB160910A - GBM (Veres & Meegan, 2016) Table 2 : The Band model fit comparison between BAT + CZTI and
Fermi alone analysis of the bursts GRB151006A and GRB 160325A. The errors are reported for 90% confidence interval.GRB name Band Parameters BAT BAT + CZTI FermiGRB151006A α − . + . − . − . + . − . − . + . − . β − . + − . − . + . − . − . + . − . E peak (keV) 288 + − + − + − Norm . + . − . . + . − . . + . − . χ red α − . + . − . − . + . − . − . + . − . β − + e − − . − . + . − . − . + . − . E peak (keV) 137 + − + − + − Norm . + . − . . + . − . . + . − . χ red . Astrophys. Astr. (0000) : − − − − no r m a li z ed c oun t s s − k e V − GRB 151006A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
NaI 0NaI 3BGO 0LLE CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3BAT
Band
NaI 2NaI aBGO 1CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3 − − − − no r m a li z ed c oun t s s − k e V − GRB160106A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV) − − − no r m a li z ed c oun t s s − k e V − GRB 160325A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
NaI 6NaI 0BGO 1CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3BAT
NaI 3NaI 0BGO 0LLE CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3 − − − − no r m a li z ed c oun t s s − k e V − GRB 160509A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
NaI 2NaI aBGO 1CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3 − − − − no r m a li z ed c oun t s s − k e V − GRB160802A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
CPL
BandHighect
NaI 6NaI 7BGO 1LLE CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3 − − no r m a li z ed c oun t s s − k e V − GRB160821A100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
NaI 5NaI 1BGO 0CZTI SingleCZTI ComptonVeto Q1Veto Q2Veto Q3 − − − − no r m a li z ed c oun t s s − k e V − GRB 160910A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
Figure 11 : The count spectra (upper panel) and the respective residuals (lower panel) obtained for the broadbandjoint spectral analysis consisting of
Fermi + CZTI data ( + BAT data in cases where it is available) for GRB151006A, GRB 160106A, GRB 160509A, GRB 160325A, GRB 160802A, GRB 160821A and GRB 160910Aare shown.
J. Astrophys. Astr. (0000) :
BAT CZTI - single CZTI - ComptonVeto Q1 Veto Q2 Veto Q1 − − − − − no r m a li z ed c oun t s s − k e V − GRB 160131A10020 50 200 500 − ( da t a − m ode l ) / e rr o r Energy (keV)
Band
BAT CZTI - single CZTI - ComptonVeto Q1 Veto Q2 Veto Q1 − − − − − no r m a li z ed c oun t s s − k e V − GRB160607A10020 50 200 500 − − ( da t a − m ode l ) / e rr o r Energy (keV)
Band
BAT CZTI - single CZTI - ComptonVeto Q1 Veto Q2 Veto Q1 − − − − − no r m a li z ed c oun t s s − k e V − GRB160703A10020 50 200 500 − ( da t a − m ode l ) / e rr o r Energy (keV)
Figure 12 : The count spectra (upper panel) and the respective residuals (lower panel) obtained for the joint spectralanalysis consisting of Niel Gehrels
Swift
BAT + CZTI data using the spectral model Band function for the burstsGRB 160131A (top left), GRB 160607A (top right) and GRB 160703A (bottom middle) are shown. Here wedemonstrate that for bursts without
Fermi detections, the usage of CZTI data extending until 900 keV along withBAT, enables us to constrain the E peak of the GRB spectrum. . Astrophys. Astr. (0000) : G R B A G R B A G R B A G R B A G R B A G R B A G R B A G R B A G R B A G R B A E ff e c t i v e a r e a c o rr e c t i o n f a c t o r SingleComptonVeto Q1Veto Q2Veto Q3
Figure 13 : The e ff ective area correction factors obtainedfor the di ff erent CZTI datasets: single (black square),Compton (blue diamond), Veto Q1 (yellow circle), VetoQ2 (purple triangle) and Veto Q2 (red star) with re-spect to the brightest NaI detector of Fermi for di ff erentGRBs in the sample except for GRB 160131A, GRB160607A and GRB 160703A where the values are ob-tained with respect to Swift BAT detector are shown.This provides an independent validation of the AstroSat mass model, thereby boosting the confidence in thespectral analysis of the CZTI GRBs. Spectral valida-tion of the mass model and availability of CZTI spectraup to 900 keV also allows to explore spectral study ofthe GRBs detected only by Swift-BAT and CZTI butnot by
Fermi . This aspect has been particularly demon-strated in the case of GRB151006A and GRB160325Awhere we find reasonably consistent spectral fit valuesfor BAT + CZTI in comparison to solo
Fermi data anal-ysis of these bursts. Thus, the satisfactory spectral fitsobtained in 15 −
900 keV (15 −
150 keV from BAT,100 −
900 keV from CZTI) for GRB 160607A, GRB160131A and GRB 160703A demonstrates the impor-tance of CZTI sub-MeV spectroscopic capability par-ticularly to characterize the GRBs that are not detectedby
Fermi . We also identify possible systematics in-volved in the mass model and attempt to quantify them( <
15 %) in the front and rear sides of the spacecraft.This paper primarily describes the new methods ofsub-MeV spectroscopy with CZTI. We are continuingto refine these methods further, and will extensively testthem against a much larger sample of bright GRBs de-tected by CZTI in the last five years. We also plan toexplore the feasibility of using the CZT detectors andthe CsI detectors in Compton camera configuration to enhance the spectroscopic sensitivity of the instrument.From a preliminary analysis, we could successfully de-tect the GRBs in the veto-tagged events (Compton scat-tered photons from CZT detectors which are absorbedby the CsI detectors) after applying Compton scatteringkinematic conditions. We plan to use the
AstroSat massmodel to generate response matrix for the veto-taggedevents.
Appendix A
Plots for the Bayesian block analysis conducted on sin-gle event data of the GRBs are shown here.
Acknowledgements
This publication uses data from the
AstroSat missionof the Indian Space Research Organization (ISRO),archived at the Indian Space Science Data Centre(ISSDC). CZT-Imager is built by a consortium of In-stitutes across India including Tata Institute of Fun-damental Research, Mumbai, Vikram Sarabhai SpaceCentre, Thiruvananthapuram, ISRO Satellite Centre,Bengaluru, Inter University Centre for Astronomy andAstrophysics, Pune, Physical Research Laboratory,Ahmedabad, Space Application Centre, Ahmedabad:contributions from the vast technical team from allthese institutes are gratefully acknowledged.We acknowledge the use of Vikram-100 HPC at thePhysical Research Laboratory (PRL), Ahmedabad andPegasus HPC at the Inter University Centre for Astron-omy and Astrophysics (IUCAA), Pune.This research has also made use of data ob-tained through the High Energy Astrophysics ScienceArchive Research Center Online Service, provided bythe NASA / Goddard Space Flight Center.
References
Arnaud, K. 1996, in Astronomical Data Analysis Soft-ware and Systems V, Vol. 101, 17Axelsson, M., Bissaldi, E., Desiante, R., & Longo, F.2016, GRB Coordinates Network, 19227Band, D., Matteson, J., Ford, L., et al . 1993, The As-trophysical Journal, 413, 281Bhalerao, V., Bhattacharya, D., Vibhute, A., et al . 2017,Journal of Astrophysics and Astronomy, 38, 31Bissaldi, E. 2016, GRB Coordinates Network, 19754,1
J. Astrophys. Astr. (0000) : k e V ( P ho t on s c m − s − k e V − ) GRB 151006A 10020 50 − ( da t a − m ode l ) / e rr o r Energy (keV) k e V ( P ho t on s c m − s − k e V − ) GRB 151006A10020 50 200 500 − ( da t a − m ode l ) / e rr o r Energy (keV) k e V ( P ho t on s c m − s − k e V − ) GRB 151006A10 100 1000 10 − ( da t a − m ode l ) / e rr o r Energy (keV)
BAT only BAT +CZTI Fermi GBM
GRB151006A GRB151006A GRB151006ABAT CZTI - single CZTI - ComptonVeto Q1 Veto Q2 Veto Q1
NaI 0NaI 3BGO 0 k e V ( P ho t on s c m − s − k e V − ) GRB 160325A 10020 50 − ( da t a − m ode l ) / e rr o r Energy (keV) k e V ( P ho t on s c m − s − k e V − ) GRB 160325A10020 50 200 500 − ( da t a − m ode l ) / e rr o r Energy (keV) k e V ( P ho t on s c m − s − k e V − ) GRB 160509A10 100 1000 10 − − ( da t a − m ode l ) / e rr o r Energy (keV)
BAT only BAT +CZTI Fermi GBM
GRB160325A GRB160325A GRB160325A
Figure 14 : For GRB 151006A (top panel) and GRB 160325A (bottom panel), we demonstrate that the analysis ofthe BAT data alone (top left and bottom left) does not allow us to ascertain the spectral peak and β of the Bandfunction fit to the data. However, when using CZTI data along with BAT (middle plot of top and below panel), wefind that we can constrain the E peak of the spectrum which is reasonably consistent with that determined in solo Fermi
GBM analysis (top right and bottom right). The fit parameter values are given in Table 2. C Z T I E n e r g y f l u x , l o g ( F l u x ) Figure 15 : Above the CZTI flux estimates (Y) done forthe spectral fits done to CZTI data alone versus the en-ergy flux estimates done for
Fermi (X) alone spectralfits are shown. The red (blue) shaded region marks the1 σ (2 σ ) scatter of the distribution of points around theY = X line shown in dotted red line.
Time (sec) C o un t s s e c Q0Q1Q2Q3
Figure 16 : Detection of GRB 160821A in the veto-tagged events. Di ff erent colours stand for di ff erentCZTI quadrants. The time-axis is plotted from AstroSat time 209507728 seconds (marked as zero) onward. . Astrophys. Astr. (0000) :
Burgess, J. M. 2014, MNRAS, 445, 2589Chand, V., Chattopadhyay, T., Oganesyan, G., et al .2019, The Astrophysical Journal, 874, 70Chand, V., Chattopadhyay, T., Iyyani, S., et al . 2018,The Astrophysical Journal, 862, 154Chattopadhyay, T., Vadawale, S. V., Rao, A. R., et al .2016, in SPIE, Vol. 9905, Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Se-ries, 99054DChattopadhyay, T., Vadawale, S. V., Rao, A. R.,Sreekumar, S., & Bhattacharya, D. 2014, Experi-mental Astronomy, 37, 555Chattopadhyay, T., Vadawale, S. V., Aarthy, E., et al .2019, The Astrophysical Journal, 884, 123Cummings, J. R., Barthelmy, S. D., Gehrels, N., et al .2015, GRB Coordinates Network, 18410, 1Gruber, D., Goldstein, A., von Ahlefeld, V. W., et al .2014, The Astrophysical Journal Supplement Series,211, 12Kocevski, D., & Longo, F. 2016, in Eighth HuntsvilleGamma-Ray Burst Symposium, Vol. 1962, 4092Lien, A. Y., Barthelmy, S. D., Cummings, J. R., et al .2016a, GRB Coordinates Network, 19234, 1—. 2016b, GRB Coordinates Network, 19506Lien, A. Y., Barthelmy, S. D., Cenko, S. B., et al .2016c, GRB Coordinates Network, 19648Lyne, A., Pritchard, R., & Roberts, M. 1999McEnery, J., Racusin, J., & Longo, F. 2016, GRB Co-ordinates Network, 19831, 1Odaka, H., Asai, M., Hagino, K., et al . 2018, Nu-clear Instruments and Methods in Physics ResearchA, 891, 92Ohno, M., Bissaldi, E., Vianello, G., Kocevski, D., &Longo, F. 2015, GRB Coordinates Network, 18406,1Paul, B. 2013, International Journal of Modern PhysicsD, 22, 41009Rao, A. R., Chand, V., Hingar, M. K., et al . 2016, TheAstrophysical Journal, 833, 86Roberts, O. J. 2016, GRB Coordinates Network, 19224Roberts, O. J., Fitzpatrick, G., & Veres, P. 2016, GRBCoordinates Network, 19411, 1 Roberts, O. J., & Meegan, C. 2015, GRB CoordinatesNetwork, 18404, 1Scargle, J. D. 1998, The Astrophysical Journal, 504,405Scargle, J. D., Norris, J. P., Jackson, B., & Chiang, J.2013, The Astrophysical Journal, 764, 167Sharma, V., Iyyani, S., Bhattacharya, D., et al . 2019,The Astrophysical Journal Letter, 882, L10Singh, K. P., Tandon, S. N., Agrawal, P. C., et al . 2014,in Society of Photo-Optical Instrumentation Engi-neers (SPIE) Conference Series, Vol. 9144, Societyof Photo-Optical Instrumentation Engineers (SPIE)Conference SeriesStanbro, M., & Meegan, C. 2016, GRB CoordinatesNetwork, 19835, 1Vadawale, S. V., Chattopadhyay, T., Mithun, N. P. S., et al . 2018, Nature Astronomy, 2, 50Veres, P., & Meegan, C. 2016, GRB Coordinates Net-work, 19901, 1
J. Astrophys. Astr. (0000) :
GRB151006AGRB160131A
GRB151006AGRB160131A
GRB 160509AGRB 160325A
GRB160106A
GRB 160509AGRB 160325A
GRB 160623AGRB 160607A
GRB 160623AGRB 160607A
GRB 160910AGRB 160703A
GRB160821AGRB160802A
GRB160821AGRB160802A
GRB 160910AGRB 160703A
Figure 17 : The Bayesian block binning of the single event CZTI light curve of the bursts are shown above in blacksolid lines. The time interval of the integrated emission of each burst is marked by the vertical dotted lines on therespective plots. The red dashed horizontal line marks the background level. The basic light curve is plotted inthe background in pink colour. We note that here the 0 marks the start of the T90