Exploring Sub-MeV Sensitivity of AstroSat-CZTI for ON-axis Bright Sources
Abhay Kumar, Tanmoy Chattopadhyay, Santosh V Vadawale, A.R. Rao, Soumya Gupta, Mithun N.P.S., Varun Bhalerao, Dipankar Bhattacharya
JJ. Astrophys. Astr. (0000) :
Exploring Sub-MeV Sensitivity of AstroSat − CZTI for ON − axis BrightSources Abhay Kumar , Tanmoy Chattopadhyay , Santosh V Vadawale , A.R. Rao , SoumyaGupta , Mithun N. P. S. , Varun Bhalerao and Dipankar Bhattacharya Physical Research Laboratory, Navrangpura, Ahmedabad, 380009, India. Indian Institute of Technology, Gandhinagar, 382355, India. Kavli Institute of Astrophysics and Cosmology, 452 Lomita Mall, Stanford, CA 94305, USA The Inter-University Centre for Astronomy and Astrophysics, Pune, India Tata Institute of Fundamental Research, Mumbai, India Indian Institute of Technology Bombay, Mumbai, India * Corresponding author. E-mail: [email protected], [email protected] received –; accepted –
Abstract.
The Cadmium Zinc Telluride Imager (CZTI) onboard
AstroSat is designed for hard X-ray imagingand spectroscopy in the energy range of 20 - 100 keV. The CZT detectors are of 5 mm thickness and hence havegood e ffi ciency for Compton interactions beyond 100 keV. The polarisation analysis using CZTI relies on suchCompton events and have been verified experimentally. The same Compton events can also be used to extend thespectroscopy up to 380 keV. Further, it has been observed that about 20% pixels of the CZTI detector plane havelow gain, and they are excluded from the primary spectroscopy. If these pixels are included, then the spectroscopiccapability of CZTI can be extended up to 500 keV and further up to 700 keV with a better gain calibration in thefuture. Here we explore the possibility of using the Compton events as well as the low gain pixels to extend thespectroscopic energy range of CZTI for ON-axis bright X-ray sources. We demonstrate this technique using Crabobservations and explore its sensitivity. Keywords.
AstroSat —CZT-Imager—sub-Mev spectroscopy—Crab.
1. Introduction
The Cadmium Zinc Telluride Imager (hereafter CZTI)onboard
AstroSat , India’s first dedicated astronomysatellite (Paul 2013; Singh et al . 2014), is primarilydesigned for hard X-ray coded mask imaging and spec-troscopy in the energy range of 20 - 100 keV (Bhalerao et al . 2017). It consists of 64 Cadmium Zinc Telluride(CZT) detector modules having 5 mm thickness and 4cm × ×
16 array of pixels (2.5 mm × AstroSat ,about 20% of the CZTI pixels (therefore ∼
20% of thetotal 924 cm geometric area) were found to have a rela-tively lower gain than that of the spectroscopically goodpixels, which makes these pixels sensitive to higher en-ergy photons ( ∼
70 - 1000 keV for a gain shift > ffi cient de-tection e ffi ciency up to 1 MeV. Motivated by this, wetried to explore the possibility of including these low gain pixels in the analysis to enhance the spectroscopicsensitivity of CZTI up to the sub-MeV region.The Coded mask spectrum generated using thestandard pipeline is restricted up to 100 keV, where thebackground is simultaneously obtained from the codedmask imaging. Above 100 keV, the 0.5 mm thick tanta-lum mask becomes increasingly transparent, along withthe collimators and CZTI support structures. To ob-tain the high energy spectra above 100 keV in the ab-sence of simultaneous background measurements re-quires a careful selection of blank sky observations.Background flux depends on multiple factors like thespacecraft’s geometric location in orbit, orbital preces-sion of the satellite, and the time spent within the highbackground South Atlantic Anomaly (SAA) region inan orbit. These contribute to a systematic modulationin the flux along the satellite’s orbit, hence making thebackground subtraction quite challenging.Another challenge is to calibrate these pixels in theabsence of any mono energetic lines at energies above © Indian Academy of Sciences 1 a r X i v : . [ a s t r o - ph . I M ] F e b J. Astrophys. Astr. (0000) :
100 keV to estimate their gains. A careful calibra-tion of these pixels have been attempted in a compan-ion paper by Chattopadhyay et al . 2020 in this is-sue, and spectroscopy up to 900 keV is explored forGamma Ray Bursts (GRBs). Compared to the ON-axissources, spectroscopy of GRBs is relatively easy be-cause of the availability of background (from the im-mediate pre-GRB and post-GRB observations) and thesignificantly higher signal to noise ratio for the GRBs.On the other hand, the ON-axis bright astrophysicalsources are fainter, and hence longer exposure observa-tions are required for su ffi cient detection, which leadsto more instrumental, charged particle, and cosmic X-ray background contributions.This paper outlines the methodology of sub-MeVspectroscopy with CZTI for bright ON-axis sources.We utilize the 2-pixel Compton events, which are usedto extract polarimetry information of the X-ray pho-tons (Chattopadhyay et al . 2014, 2019; Vadawale etal . 2015, 2018) to enhance the spectroscopic capabil-ity. The selection of background observations and thebackground subtraction used for polarimetric measure-ments is described in detail here, with an emphasis onspectroscopy. We utilize the low gain pixels to extendthe spectroscopic energy range of the instrument wellbeyond the standard limit. With the inclusion of thelow gain pixels, Compton spectroscopy was also ex-tended to 500 keV. We use the AstroSat mass modelin GEANT4 (Agostinelli et al . 2003) to generate thespectral response.Here we carry out broadband spectroscopy of Crabusing standard spectroscopic events (30 - 100 keV), 2-pixel Compton events including low gain pixels (100 -500 keV), and also explored the 1-pixel events includ-ing low gain pixels (100 - 700 keV) to establish the sub-MeV spectroscopy methods and at the same time tryto constrain the spectral parameters of Crab at higherenergies. The X-ray emission coming from Crab Neb-ula can be divided into three parts: pulsed point sourcefrom the neutron star, synchrotron emission poweredby charged particles coming from the centrally locatedpulsar, and the large di ff used emission region in theNebula. Toor & Seward (1974) fitted the Crab spec-tra in 2 - 60 keV using a power law and concludedthat it could be used as a standard calibration sourcedue to its steady nature. It is believed that it appears tobe steady on a time scale of a few years because mostof the emission is from the di ff used expanding ejecta,which is extended. There is, however, no theoreticalbasis that the pulsed emission will be steady. Di ff erentmodels have been tried to explain the stable behaviorand emission mechanism of Crab. It is often describedby a power law model (Kirsch et al . 2005; Kuiper et al .2001). A broken power law is also used to describe the slope evolution, with a break around 100 keV (Strick-man et al . 1979; Ling & Wheaton 2003). Massaro etal . (2000) and Mineo et al . (2006) used a single curvedpower law with a variation of the slope with Log(E) todescribe the spectra. The INTEGRAL / SPI data were fitwith a broken power law (Jourdain & Roques 2008) andalso by the Band model generally used for GRBs (Band et al . 1993; Jourdain & Roques 2020). There is no gen-eral agreement on the best overall spectral model so far.Further, the absolute flux, emission mechanism, andthe cause of Crab’s stability are also not well known.The extended bandwidth of CZTI using the Comptonand single spectrum, including low gain pixel events,can help us understand the spectra of the Crab. It isto be noted that CZTI is also sensitive to polarisation,and hence a simultaneous measurement of polarisationalong with spectroscopy can help us in a better under-standing of the Crab emission mechanism and the causeof the stability in its emitted flux.In section 2., a brief description of the observationsand analysis procedure is given. The results obtainedare presented in section 3.. Finally, in section 4., we dis-cuss the sub-MeV sensitivity of CZTI and future plans.
2. Observations and analysis procedure
There are many observations of Crab over the past fiveyears of operation of
AstroSat . We have selected thoseobservations with su ffi cient exposure and also having asuitable background observation. The selection of ap-propriate background observation and subtraction is animportant part of the analysis. The CZTI support struc-ture becomes increasingly transparent above 100 keV.Therefore, background measurements can be a ff ecteddue to the presence of bright X-ray sources within70 ◦ of the pointing direction of CZTI. The Crab andCygnus X-1 are two bright sources which should beavoided during the background observation. Since boththe sources are located almost opposite to each otherin the sky, it is possible to find a good region, awayfrom these two sources, for the background observa-tions. The background observation should also be closeto the source observation time to avoid the error in somelong-term secular variations in the background behav-ior. There are a few such observations that satisfy thecriterion of the background selection. Based on theseconsiderations, we have finally selected five observa-tions between 2015 and 2017 for further analysis. De-tails of the Crab observations and the correspondingbackground observations selected for the present anal-ysis are given in Table 1. . Astrophys. Astr. (0000) : Table 1.
Summary of Crab and blank sky observations.
Crab Blank SkyObsID Date Exposure ObsID Date Exposure RA DEC(yyyy / mm / dd) (ks) (yyyy / mm / dd) (ks) (deg) (deg)9000000096 2015 / /
12 41 9000000276 2016 / /
16 64 183.48 22.89000000252 2016 / /
07 60 9000000276 2016 / /
16 64 183.48 22.89000000406 2016 / /
31 114 9000000404 2016 / /
29 64 228.21 -9.099000000964 2017 / /
14 78 9000000974 2017 / /
22 51 183.48 22.89000000970 2017 / /
18 123 9000000974 2017 / /
22 51 183.48 22.8
Crab RA: 83 . ◦ and DEC: 22 . ◦ Single and Compton event selection
CZTI is a pixelated detector. The scientific data anal-ysis of CZTI is done with two types of CZTI events:1-pixel or single pixel events and 2-pixel events. Thesingle pixel events registered in the CZTI are consid-ered as the true 1-pixel events if there is no event reg-istered within 100 µ s time window on either side of thesingle event. The events in the CZTI are time stampedat every 20 µ s (Bhalerao et al . 2017) and any twoevents occurring within 20 µ s coincidence time win-dow in two pixels are considered as true 2-pixel event(Chattopadhyay et al . 2014). The standard single pixelmask-weighted spectra in 30 - 100 keV (hereafter EB1)is generated following standard pipeline software avail-able at the AstroSat science support cell (ASSC) .After the launch of AstroSat , it is observed thatCZTI had low gain pixels of about 20% of the detectorplane. These pixels are found to have a shift in gain (gs)by a factor of 1.5 − / gs); hence the name low gain pixels. Because thesepixels now record higher energies, the spectroscopicrange of CZTI with single pixel events can be extendedup to 700 keV with the inclusion of low gain pixels. De-tailed characterisation methods of low gain pixels havebeen discussed in the companion paper by Chattopad-hyay et al . 2020. Above 100 keV single event spectra,including the low gain pixel events, need to be analysedoutside of the standard pipeline because they are not in-cluded in the standard pipeline of CZTI. The initial pro-cessing is to remove the intervals of high background http: // astrosat-ssc.iucaa.in / during the passage of South Atlantic Anomaly fromeach observation and removal of the noisy (pixel hav-ing counts more than five sigma above mean value) orspectroscopically bad pixels (energy resolution is poorcompared to the normal pixels). This procedure is validfor both the single and 2-pixel event spectra. After se-lecting the single pixel events, energy deposited in eachpixel is used to generate the single pixel spectrum in-cluding low gain pixels in the 100 - 700 keV (hereafterEB2). We have binned the events into 60 channels with10 keV energy bin sizes in the 100 - 700 keV energyrange.Above 100 keV, the 5 mm thick CZTs have su ffi -cient e ffi ciency for Compton interactions. Polarisationanalysis in the 100 - 380 keV range depends upon suchCompton events (Chattopadhyay et al . 2014; Vadawale et al . 2015). These Compton events are used here todo spectroscopy in the 100 - 380 keV range. After in-corporating the low gain pixels, Compton spectroscopycan be further extended to 500 keV. For the Comptonevent selection, we follow the 2-pixel Compton eventselection criteria, as discussed in Chattopadhyay et al .2014. The readout logic in the CZTI is such that it readsevents from one module at a time. If two events areregistered in two di ff erent pixels in the same module,then it is possible that two events would get two di ff er-ent time stamps. Hence, all the events occurring withinthe coincidence time window of 40 µ s are selected forthe analysis. These events are further filtered throughCompton kinematics criteria to enhance the signal tonoise ratio. After the selection of the Compton events,the sum of the energy deposited in the scattering and theabsorption pixel is used to generate the 2-pixel Comp-ton events spectrum including low gain pixels in the100 - 500 keV (hereafter EB3). We have binned the J. Astrophys. Astr. (0000) : data at 10 keV energy bin size in the 40 channels rang-ing from 100 keV to 500 keV.2.2
Background subtraction: Phase match method
The primary sources of background in the CZT detec-tor are the Cosmic X-ray background, the Earth’s X-rayAlbedo, and the locally produced X-rays due to Cos-mic ray interactions. The Compton background in thedetector is due to the Compton scattering of these back-ground X-rays. In addition to this, a small part of theCompton background consists of chance coincidenceevents within 40 µ s coincidence time window. × × × × Time (s) C oun t s × × Time (s) C oun t s Time (s) χ × × Time (s) C ou n t s Figure 1 . Top panel shows the light curves of Crab (ObsId96) on the left and background (ObsId 276) on the right,both fitted with fifth degree polynomials (red curve for Craband green curve for background). Bottom left figure showsthe variation of χ as a function of shift in the Crab lightcurve with respect to the background, to determine the timefor ‘phase matching’. Bottom right figure shows the ‘phasematched’ background and Crab polynomials, in the greencurve and red, respectively. The blue curve represents thebackground where Crab observation is available. The background events from the blank sky obser-vations are filtered through the same selection criteriaas the source, as discussed in section 2.1. The observedbackground counts show a prominent orbital variationas well as a diurnal variation depending on the geomet-ric location of the spacecraft in the orbit, orbital preces-sion, and the location and duration of the SAA region in orbit. All these contribute to a systematic modula-tion (see the top panel of Figure 1) in the flux along theorbit of the satellite within the duration of the obser-vation. Because of the modulating flux, it is importantto select similar portions of the Crab and backgroundorbits based on the spacecraft’s ground tracks (latitudeand longitude) and use them for further analysis. Butthis puts a strict condition on the selection of the orbitsand leaves a short usable exposure of Crab and back-ground observations. An alternate method developedfor background subtraction is to match the phase of thebackground and Crab light curves (the ‘phase matchmethod’), used for the polarisation measurements ofCrab (Vadawale et al . 2018). In the phase matchmethod, first, the Crab and background light curvesare fitted with an appropriate higher-order (here 5 th or-der) polynomial. Then the Crab and background lightcurves are matched by sliding the Crab polynomial overthe background polynomial every 10 seconds and esti-mating the best match by minimising the χ . Withinthe matched region of the two polynomials, the back-ground is taken only for those time regions for whichthere is a source observation (see bottom right of Figure1). From the background’s phase-matched region, wecalculate the ratio of the average background count rateto the count rate in the phase matched region (‘correc-tion factor’) and multiply that to the total backgroundexposure to calculate an e ff ective background exposure.The use of e ff ective exposure automatically takes careof the di ff erent phases of source and background obser-vations. For example, if the source and backgroundsare observed in the same phase, the multiplication fac-tor will be close to 1. For longer source observations(exposure (cid:29) background exposure), it is divided intomultiple parts and for each part of the source obser-vations, we calculate the correction factor in the waydescribed above and then the final correction factor iscalculated as the weighted average.It is to be noted that the phase match method onlyensures that the source (Crab in this case) and blank skybackground data are taken for the same orbital phase(spacecraft orbit). It is not for the background spec-tral modeling. The background scaling is done to thephase matched background region to best mimic theactual background during the source observation. Be-cause the background spectra are obtained from obser-vations of “blank sky” with no other hard X-ray sourcesin the FOV (confirmed from BAT catalog), the spectrafor the used backgrounds are expected to be the same.To demonstrate this, we selected a CZTI observationfrom UVIT catalog (ObsID 1008) such that there areno other bright X-ray sources in the FOV of CZTI (callthis ‘CU’). We then used the polynomial method to dophase match between CU and background data ObsID . Astrophys. Astr. (0000) : Channel Number −2 −1 C o un t R a t e Channel Number R a t i o Channel Number −0.010.000.01 R e s i d u a l Channel Number −2 −1 C o un t R a t e Channel Number R a t i o Channel Number −0.010.000.01 R e s i d u a l Figure 2 . The top figure shows the phase matched 2-eventCompton spectra (EB3) of the blank sky observation CU(red) and background B (blue) from 100 keV to 500 keV.The ratio of the two spectra is shown in the middle paneland the residuals of the two are shown in the bottom panel.The bottom figure shows similar plots for the low gain pixelspectra (EB2) from 100 keV to 700 keV .974 (call this ‘B’), which is used in the analysis and cor-rect for total flux in B for the phase of CU. Because CUis essentially a blank sky background for CZTI, similarflux and spectra for both B and CU after phase correc-tion is expected.We found identical flux for both B and CU (see Fig-ure 2), signifying that the polynomial method is capa-ble of finding the common phase and scale the flux ac-cordingly. The spectra are also found to be identical,justifying the underlying assumption that the blank skyobservations for CZTI with the predefined selection cri-teria yield a similar photon energy distribution. 2.3
Spectral Response
The response for EB1 is generated using the standardpipeline of the
AstroSat and for both EB2 and the EB3using the GEANT4 simulation of the
AstroSat massmodel. Details of the
AstroSat mass model and it’svalidation has been discussed in Chattopadhyay et al .2019 and Mate et al . 2020, in this issue. We sim-ulated the mass model for 56 mono-energies rangingfrom 100 keV to 2 MeV (at every 20 keV up to 1 MeVand 200 keV till 2 MeV) for 10 photons. For eachenergy, the distribution of deposited energy in CZTIpixels is computed at 1 keV binning for each pixel.The CZTI pixel-level LLD (Lower Level Discrimina-tor), the ULDs (Upper Level Discriminator), list ofnoisy and dead pixels obtained from the actual obser-vational data are applied to the simulation data. For2-pixel Compton events, the sums of the energies in thecorresponding two pixels are used to obtain the totaldeposited energy, while for single event response, to-tal absorbed energy for a given incident photon is used.We applied the same criterion of event selection, as dis-cussed in section 2.1. We then convolve the 1 keV binsby a Gaussian of 8 keV Full Width at Half Maximum(FWHM). We have not noticed any significant increasein FWHM with energy for CZTI pixels during groundcalibration. Therefore, FWHM is kept constant acrossthe energy. It is to be noted that the response for EB1is generated using µτ and charge di ff usion based lineprofile model (Chattopadhyay et al . 2016).
3. Results
Single and Compton event spectra (EB1 and EB3)
We use the broken power law model to fit the Crabcurved spectra in 30 - 500 keV. It has been long usedto explain the Crab spectra with break energy at 100keV (Strickman et al . 1979; Ling & Wheaton 2003).
INTEGRAL / SPI has also shown the spectral fitting us-ing broken power law up to sub-MeV region (20 keV- 1 MeV) (Jourdain & Roques 2008). We have anal-ysed all the selected observations and then fitted theresultant spectra (EB1 and EB3) simultaneously us-ing const × bknpower in XSPEC (Arnaud 1996) freez-ing break energy at 100 keV while the other parame-ters (photon indices) are tied across the spectra. To ac-count for the cross calibration and di ff erences betweenthe di ff erent spectra, a constant was multiplied to themodel. It was fixed to one for EB1 and left free to varyfor others. The EB1 below 30 keV and above 100 keVis ignored due to calibration issues. No systematic hasbeen added to the EB1 and EB3. Spectral fitting for oneof the five observations (ObsID 406, 114 ks) is shown in J. Astrophys. Astr. (0000) :
Table 2 . Comparison of fitted parameters between
INTEGRAL / SPI and
AstroSat for broken power law. The errors arereported for 90% confidence interval. The mean value of the parameters of all the observations for di ff erent combinations ofdata is mentioned in the bottom row of each block. Instrument Spectra ObsID
PhoIndx PhoIndx E break Norm Flux (30 − keV ) χ / dof(keV) 10 − (erg / cm s)INTEGRAL 2 . + . − . . + . − . ∗ . + . − . . EB EB . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . EB EB . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . EB , EB EB . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . / . + . − . . + . − . ∗ . + . − . . + . − . INTEGRAL / SPI flux in the 30 - 100 keV is estimated using the model parameters given in (Jourdain & Roques2008, 2020). ∗ is for fixed parameters. . Astrophys. Astr. (0000) : Figure 3. The values of the fitted parameters for all thefive observations along with the
INTEGRAL / SPI resultsare given in Table 2. The low energy slope (
PhoIndx
PhoIndx
2) are well con-strained and consistent with the
INTEGRAL / SPI (Jour-dain E., Roques J. P. 2008) values within errors (seeFigure 4). The contour plots of
PhoIndx
Norm for all the five observations are shown in Figure 5. Theinputs for the contour plot are generated using the chaincommand in
XSPEC after getting the best fit parameters,and the corner plots are generated using the python cor-ner module (Foreman-Mackey 2017). The corner plotsshow that the value of
PhoIndx . + . − . and thatof Norm is 6 . + . − . , which is ∼
31% smaller than the
Norm for
INTEGRAL / SPI . −5 −4 −3 no r m a li z ed c oun t s s − k e V − ∆ χ Energy (keV)
Figure 3 . Broadband spectra of Crab (ObsID 406, 114 ks)fitted with a broken power law. The black, red colors areused for EB1, and EB3 respectively.
CZTI sensitivity is also studied by looking at theflux variation of Crab with time. Figure 6 comparesthe Crab flux covering the period between 2015 Nov12 and 2017 Jan 18. Each data point corresponds tothe flux obtained in each observation, and the dottedline represents the flux measured by
INTEGRAL / SPI .The flux from ObsID 96 is much lower than the val-ues found for the other ObsIDs. If we exclude the datafrom ObsID 96, the remaining four measurements arewithin 5% of the mean value. We note here that thebackground for this ObsID is measured more than twomonths before the Crab observation, whereas for all theother ObsIDs, the background is measured within tendays of the respective Crab observation. The e ff ect ofsecular variations in the background on the flux mea-surements needs to be investigated further.
96 252 406 964 970ObsID1.01.52.02.53.03.5 E n e r g y s l o p e Figure 4 . Best fit parameter of all the selected observationsobtained after simultaneous fitting of EB1 and EB3 usingbroken power law model. Red diamonds represent higherenergy slope (
PhoIndx
2) and the blue squares represent lowenergy slope (
PhoIndx
PhoIndx
PhoIndx . . . . Norm . . . P h o I n d x . . . PhoIndx2
Figure 5 . Corner plot of the phoIndx
Norm for allthe five ObsIDs plotted together for EB1 and EB3. Di ff erentcolors of contours represents di ff erent observations as shownin legend. The peak value of the PhoIndx . + . − . andthe Norm is 6 . + . − . Low Gain spectra
As discussed earlier, 20% of the pixels in CZTI arefound to have lower gains as compared to the spectro-
J. Astrophys. Astr. (0000) :
96 252 406 964 970ObsID0.40.60.81.01.21.4 F l u x ( − e r g s c m − s − ) Figure 6 . Estimated flux of the selected observations afterfitting EB1 and EB3 simultaneously using broken powerlaw . The blue diamond represents the CZTI value of theCrab flux. The dashed-dot horizontal line represent theINTEGRAL value of the Crab flux in the correspondingenergy band. scopically good pixels, which makes these pixels sen-sitive to higher energy photons ( ∼
70 - 1000 keV for again shift > −5 −4 −3 no r m a li z ed c oun t s s − k e V − ∆ χ Energy (keV)
Figure 7 . Broadband spectra of the Crab (ObsID 406, 114ks) fitted with a broken power law. The black, green and redcolors are used for EB1, EB2 and EB3 respectively.
We have analysed all the selected observations andthen fitted the resultant spectra (EB1, EB2, and EB3)simultaneously using the same constraints as describedin section 3.1. An 8% systematic error, however, havebeen added to EB2 to take into account of the uncertain-ties in the background estimation and gain calibration. The systematic error is added to the data till the resid-uals are uniformly distributed across zero and the spec-tral fit is acceptable. Spectral fitting for one of the fiveobservations (ObsID 406, 114 ks) is shown in Figure 7.The fitted parameters are given in Table 2. The low en-ergy slope (
PhoIndx
1) is well constrained and close to
INTEGRAL / SPI (Jourdain & Roques 2008) value. Thehigher energy slope (
PhoIndx ∼ INTEGRAL / SPI reported values, thoughthe errors in the parameter are quite large. The influ-ence of subtle background spectral variations and pos-sible gain non-linearity at higher energies (the CZT de-tectors are not calibrated outside the energy range of 10to 150 keV) perhaps can lead to the somewhat flatterspectra above 100 keV. Though we are getting a flat-ter spectrum, we get consistent flux within 10-20% ofthe mean value (see Figure 8). This gives enough con-fidence that there is a possibility of extending CZTIbandwidth up to 700 keV with better energy calibrationof the low gain pixels.
96 252 406 964 970ObsID0.50.60.70.80.91.01.1 F l u x ( − e r g s c m − s − ) Figure 8 . Estimated flux of the selected observations afterfitting EB1, EB2, and EB3 simultaneously using brokenpower law. The blue diamond corresponds to Crab flux inthe 100 - 700 keV energy band (EB2). The red dashed-dothorizontal lines represent the mean value (0 . + . − . ) of theCrab flux of all the five selected observations added togetherin the corresponding energy range.
4. Discussion and conclusions
In this article, we have attempted to explore the sensi-tivity of CZTI in the sub-MeV region and have outlineda methodology of sub-MeV spectroscopy using Crabobservations. For this purpose, we have used the sin-gle pixel mask-weighted spectral data in the 30 - 100keV energy range (EB1), 2-pixel Compton events in- . Astrophys. Astr. (0000) : cluding low gain pixels in the 100 - 500 keV energyrange (EB3), and the single pixel events including lowgain pixels in the 100 - 700 keV energy range (EB2).For this work, we used the calibration parametersobtained by Chattopadhyay et al . 2020 (in this issue)for similar work to explore the sub-MeV spectroscopicsensitivity for Gamma ray Bursts (GRBs). The advan-tage in the case of GRBs is the higher signal strengthand, in particular, the availability of simultaneous back-ground events before and after the burst. In the caseof persistent sources, the unavailability of simultaneousbackground spectra makes the selection of proper blanksky flux and its subtraction extremely important (sec-tion 2.2), particularly when the signal to noise ratio isrelatively low. For this work, the spectral response wasgenerated using a simple Gaussian energy distributionfor simplicity (section 2.3), which we plan to improvelater with the use of a more physical line profile modelbased on charge trapping and di ff usion (Chattopadhyay et al . 2016).We applied these techniques for spectral analysisof the Crab, where we used a broken power law (bkn-power in XSPEC ) for the spectral fitting. The spectralfits show su ffi cient flux sensitivity of CZTI to carry outspectroscopy for ON-axis bright sources (see Table 2)up to 500 keV, and it can be extended up to 700 keVwith better gain calibration. We find that the overallflux measured by CZTI in 100 - 700 keV band (EB2)is consistent throughout the observations within 20%from the mean value. For the single event and Comp-ton event spectra in 30 - 500 keV (EB1 and EB3), thelow energy slope ( PhoIndx
1) and higher energy slope(
PhoIndx
2) agree reasonably with the INTEGRAL / SPIresult. However, for the combined spectral fit over allthree bands (EB1, EB2, EB3), the spectral parametersare not fully consistent with the results reported by IN-TEGRAL / SPI with the high energy slope (PhoIndx2)showing a possible flattening. This could be either dueto spectral calibration or incorrect background estima-tion at higher energies above 400 keV.We note that the measured flux by CZTI is lowerby ∼
31% than that estimated by
INTEGRAL / SPI . How-ever, it should also be noted that in general, it is quitedi ffi cult to make a comparison of flux measurements ofthe two instruments, particularly in hard X-rays, dueto the di ffi culty in measurement of the absolute e ff ec-tive areas of various instruments. Considering the er-rors in our flux measurements ( ∼ ∼ ff erence in the flux measurements of Crab made byCZTI as compared to INTEGRAL / SPI. We plan to do across-calibration with simultaneous NuSTAR observa-tions and try to understand the e ff ective area calibration of CZTI in a future work.To summarize, we find that CZTI has su ffi cientspectral sensitivity in the sub-MeV region for the ON-axis sources. The parameters are well constrained in the30 - 500 keV range, but the spectra become flatter above500 keV. In future, we plan to develop better back-ground subtraction methods, and at the same time, weattempt to develop a detailed background model using5-years of CZTI data. Moreover, we plan to investigatethe gain of the low gain pixels in more detail; in partic-ular, we look for various high energy background linesby proton induced radio-activation as shown by Odaka et al . 2018. With the better characterization of the lowgain pixels and background subtraction methods, CZTIis expected to provide sensitive spectroscopic informa-tion for various hard X-rays sources such as Cygnus X-1 and other bright sources and add to the better under-standing of the emission mechanisms in these sources. Acknowledgements
This research is supported by the Physical ResearchLaboratory, Ahmedabad, Department of Space, Gov-ernment of India. We acknowledge the ISRO ScienceData Archive for
AstroSat
Mission, Indian Space Sci-ence Data Centre (ISSDC) located at Bylalu for pro-viding the required data for this publication and Pay-load Operation Center (POC) for CZTI located at In-ter University Centre for Astronomy & Astrophysics(IUCAA) at Pune for providing the data reduction soft-ware.
References [1] Agostinelli S. et al . 2003 , NIMPA, 506, 250[2] Arnaud K. A. 1996 , ASPC, 101, 17[3] Band D. et al . 1993 , ApJ, 413, 281[4] Bhalerao V. et al . 2017 , JApA, 38, 31[5] Chattopadhyay T. et al . 2016 , SPIE, 9905,99054D[6] Chattopadhyay T. et al . 2014 , ExA, 37, 555[7] Chattopadhyay T. et al . 2019 , ApJ, 884, 123[8] Foreman-Mackey D. 2017 , ascl.soft,ascl:1702.002[9] Jourdain E., Roques J. P. 2008 , int..work, 143[10] Jourdain E., Roques J. P. 2020 , ApJ, 899, 131[11] Kirsch M. G. et al . 2005 , SPIE, 5898, 22
J. Astrophys. Astr. (0000) : [12] Kuiper L. et al . 2001 , A&A, 378, 918[13] Ling J. C., Wheaton W. A. 2003 , ApJ, 598, 334[14] Massaro E. et al . 2000 , A&A, 361, 695[15] Mineo T. et al . 2006 , A&A, 450, 617[16] Odaka H. et al . 2018 , NIMPA, 891, 92[17] Paul B. 2013 , IJMPD, 22, 1341009[18] Singh K. P. et al . 2014 , SPIE, 9144, 91441S[19] Strickman M. S. et al . 1979 , ApJL, 230, L15[20] Toor A., Seward F. D. 1974 , AJ, 79, 995[21] Vadawale S. V. et al . 2015 , A&A, 578, A73[22] Vadawale S. V. et alet al