An alternative scheme to estimate AstroSat/LAXPC background for faint sources
aa r X i v : . [ a s t r o - ph . I M ] F e b J. Astrophys. Astr. (0000) :
An alternative scheme to estimate AstroSat / LAXPC background for faintsources
Ranjeev Misra , Jayashree Roy and J. S. Yadav Inter-University Centre for Astronomy and Astrophysics, Pune, 411007, India. Department of Physics, Indian Institute of Technology, Kanpur, 208016, India * Corresponding author. E-mail: [email protected] received ; accepted
Abstract.
An alternative scheme is described to estimate the layer 1 LAXPC 20 background for faint sources where the sourcecontribution to the 50-80 keV count rate is less than 0.25 counts / sec (15 milli-crabs or 6 × − ergs / s / cm ). Weconsider 12 blank sky observations and based on their 50-80 keV count rate in 100 second time-bins, generatefour template spectra which are then used to estimate the background spectrum and lightcurve for a given faintsource observation. The variance of the estimated background subtracted spectra for the 12 blank sky observationsis taken as the energy dependent systematic uncertainty which will dominate over the statistical one for exposureslonger than 5 ksecs. The estimated 100 second time bin background lightcurve in the 4-20 keV band with a 3%systematic error matches with the blank sky ones. The 4-20 keV spectrum can be constrained for a source with flux ' ∼ ff erent blank sky observations, threeAGN sources (Mrk 0926, Mrk 110, NGC 4593) and LMC X-1 are shown. Keywords.
AstroSat / LAXPC—Instrument Background—Calibration
1. Introduction
The sensitivity of the LAXPC instrument (Antia et al.,2017) onboard AstroSat (Yadav et al., 2016;Agrawal et al., 2017) to extract spectral and longterm temporal information of faint sources dependscritically on how well the background of the instrumentis characterized. The background variation is primarilydue to the changing response of the instrument to avarying local charged particle distribution. The cosmicX-ray background from the ∼ .
25 square degree fieldof view contributes less than 10% of the observedbackground and hence its cosmic variance does notcontribute significantly to the background variation.The standard method (Antia et al., 2017) to estimatethe background involves blank sky spectra obtained asa function of latitude and longitude of the satellite. Fora given science observation, a blank sky observation ischosen which is typically one that is closest in time tothe science observation. The background is estimatedas that expected for the latitude and longitude coveredduring the science observation based on the blank skyobservation after taking into account gain variationbetween the blank sky and source observations. This method provides background spectra which di ff er fromthe true background by roughly 3% and has beenextensively used for science analysis. This systematicuncertainty exceeds the statistical one for exposureslonger than 5 ksecs.It is prudent to have di ff erent and independentmethods to estimate the background to provide con-fidence on the scientific results obtained. Here wedescribe such a scheme which assumes that for faintsources the detected flux in the high energy band (50-80 keV) can be attributed to the background alone andhence can act as a proxy to measure the backgroundlevel as a function of time. As shown in this work,the technique is applicable to sources that contributeless than 15 milli-Crab of flux in the 50-80 keV band.The scheme named as “Faint source background esti-mation” has been incorporated in the LAXPC softwarelaxpcsoft available at the AstroSat Science SupportCell . It has been used for scientific analysis of severalfaint sources ( e.g. LMC X-1 (Mudambi et al., 2020),RGB J0710 +
591 (Goswami et al., 2020; Yadav et al., http: // astrosat-ssc.iucaa.in / ?q = laxpcData © Indian Academy of Sciences 1
J. Astrophys. Astr. (0000) :
10 12 14 16 18 20 22 24 26 13 14 15 16 17 18 19 20 21 22 C oun t R a t e i n - k e V band Count Rate in 50-80 keV band
Figure 1 . Count rates for the 4-20 keV and 50-80 keVplotted against each other for LAXPC 20 Layer 1. Thetime-scale for integration is 100s.
2. Estimating LAXPC background
We consider twelve blank sky observations that arelisted in Table 1. For these observations, the lightcurvein 100 seconds for LAXPC 20 layer 1, were com-puted in energy bands 4-20 keV and 50-80 keV bands.The count rates of these two energy bands are plottedagainst each other in Figure 1. Most of the data liewithin 50-80 keV count rate of 14 to 19 counts / sec. Wefind that it is prudent to consider data only in this range,and divide the range into four parts corresponding to14-15, 15-16, 16-17 and 17-19 counts / sec. The averagespectra corresponding to these selections are shown inFigure 2. The spectra have been normalized such that50-80 keV flux levels are nearly equal, in order to high-light the di ff erent spectral shapes at low energies . Wenote that the spectral shapes are di ff erent at low ener-gies and treat these four spectra as templates for esti-mating the background spectrum and lightcurve for asource.The procedure for estimating the background for asource is as follows:1) Collect the 50-80 keV lightcurve in 100 second timebins.2) Select GTIs based on the count rate of the 50-80 keVenergy range being between 14 and 19 counts / s. Due to gain variation, the channel to energy conversion for theblank sky observations are di ff erent. Here, we consider one obser-vation spectrum as the reference and we interpolate the other spectrasuch that all of them have the same energy bin channels as that ofthe reference. C oun t s pe c t r u m ( C oun t s / s e c / k e V ) Energy (keV)
Figure 2 . Average Blank sky spectrum corresponding towhen the 50-80 keV count rates are in the range 14-15,15-16,16-17 and 17-20 counts / sec. The spectra have beennormalized such that 50-80 keV flux levels are nearly equal,in order to highlight the di ff erent spectral shapes at low ener-gies. These four spectra are used as templates to estimate thebackground spectrum and lightcurve of a source observation.
3) For each time bin of the 50-80 keV, use the observedcount rate to estimate the complete background energyspectrum using the four templates. The four templatesare assigned to the mid point of their count rate in 50-80keV, i.e. 14.5, 15.5, 15.5 and 18 c / s. The templates arethen interpolated to obtain the corresponding spectrumappropriate for the observed count rate.4) Integrate the estimated background energy spectrumfor each time bin over the desired energy band.5) Combine the estimated background energy spectrato estimate the time averaged background spectrum.To test the e ffi cacy of the method, the backgroundspectra was estimated for each of the blank sky ob-servations using the method described above. The es-timated background was then compared with the ob-served spectrum and at each energy bin. The twelveblank sky observations were used to get the standarddeviation of the estimated background and the observedspectra as a function of energy. The standard devia-tion in units of counts / sec / keV are shown as a func-tion of energy in Figure 3. This standard deviation canbe used as the energy dependent systematic error onan estimated background spectrum obtained from thismethod. The standard deviation is compared with a typ-ical background spectrum shown in Figure 3 and hencethe systematic error on the background is of the order ofa few percent. Thus, the systematics attained from thismethod are of the same order as that from the standardtechnique. The standard deviation of the backgroundsubtracted count rates in 4-20 keV band for the 12 blank . Astrophys. Astr. (0000) : sky observations is around 0.3 c / s, indicating a 3-sigmadetection of a source to be ∼ / s in this energy band.Also shown is the 1 milli-crab source spectrum, whichreveals that although the source count rate is a factor offew below the background, it should be detected usingthis method. The systematics at 25 keV is comparableto the source flux from a 1 mCrab source. For com-parative reference, the typical Poisson noise levels foran exposure of 5 and 50 ksecs are shown. Note that thesystematic error dominates over the Poisson one for ex-posures longer than 5 ksecs. The software includes thiserror in the background spectrum file.The estimated background lightcurve in any energyband is based on the count rate in the 50 -80 keV band.The deviation of the estimated lightcurve from the truebackground is due to the systematic limitations of thetechnique and the typical Poisson error of the 50-80keV band count rate in 100 second time bin which is ∼ . / sec and which is assumed to be only from thebackground. Hence this scheme is limited to caseswhen the source count rate is significantly less thanthe template separation rate. Thus it is applicable forsources with count rate < .
25 counts / s which trans-lates to 15 milli-crabs or 6 × − ergs / s / cm in the 50- 80 keV band.The estimated background lightcurves and spectrafor LAXPC 10 can also be estimated using the sametechnique. However, since LAXPC 10 has an higherbackground with larger uncertainty than LAXPC 20,it is recommended that LAXPC 20 should be primar-ily used for such analysis and LAXPC 10 results to betaken as a corroboration.
3. Verification and Examples
To illustrate the method, lightcurves and spectra weregenerated for three blank sky observations from 2017,2018 and 2019 which were not part of the observationsused to obtain the templates, three Active Galactic Nu-clei (Mrk 0926, Mrk 110 and NGC 4593), and for theextra-galactic X-ray binary LMC X-1. The lightcurveswere generated for a time bin of 100s and for 4-20 keVenergy range.The left panels of Figures 4 and 5, shows the to-tal lightcurve (i.e. source with background marked asSrc + Bkg), the estimated background lightcurve and thesubtraction of the two for the blank sky observations C oun t s / s e c / k e V Energy (keV)
Typical background spectrumMilli Crab spectrumStandard deviation for background fitsTypical background poisson noise for 5 ksec exposureTypical background poisson noise for 50 ksec exposure
Figure 3 . The standard deviation of the estimated back-ground as compared to blank sky observations. Alsoshown for comparison are a typical blank sky spectrum, thespectrum for a 1 milli-crab source and the typical Poissonlevel for the blank sky spectrum. from 2017 and 2019. The 2019 blank sky observationshows increased count rate for two times just before thesatellite entered the SAA. If the two increases are re-moved from the data then the resultant lightcurves aresimilar to ones obtained for the 2017 data (middle panelof Figure 5). While the reason for these higher countsrates is not clear, such variations just before entry intoSAA should be treated with caution in science analysis.These are most likely due to local variation / fluctuationsin geomagnetic field. This is to emphasize that propertime-selection by inspection is required to obtain re-liable results. The right panels of Figures 4 and 5,show the residuals of the observed spectra over thebackground for the 2017 and 2019 blank sky observa-tions. Although the residuals are shown for a wide en-ergy band, note that the technique is only valid in 4-30keV range. This shows that the systematics included inthe background spectra are adequate and no significantresiduals are seen in the 4-30 keV band. Residuals ofthe order of 0.02 counts / sec / keV should be attributed tosystematics. Since a 3% uncertainty has been includedin the estimated background lightcurve, the error on thebackground subtracted lightcurve is a combination ofthe Poisson noise in the observed lightcurve and thisbackground uncertainty. Table 2 lists the average countrate, the variance ( σ ), the standard deviation ( √ ( σ )),the expected variance ( σ EV ) and the expected stan-dard deviation ( √ ( σ EV )) for the background subtractedlightcurves binned in 100 and 1000 seconds. For thesethree blank sky observations the average count rate iswithin the 0.3 c / s deviation found for the twelve blanksky used as the template for the technique. The stan- J. Astrophys. Astr. (0000) : dard deviation ( √ ( σ )) and the expected one ( √ ( σ EV ))are similar. This implies that the background estima-tion technique can be applied to study lightcurves in100 to 1000 second binning over a time-span of ∼ / s, should beconsidered as evidence of variability.The background subtracted spectra for the AGNsources were fitted using a power-law and for LMC X-1 a disk emission model and power-law was used. Fig-ure 6 shows the background subtracted spectrum alongwith the expected background spectrum (top panel) andresiduals (bottom panel) for Mrk 0926. We emphasizethat the spectra here are shown for the full energy rangeof 4-80 keV for illustration, and science analysis shouldbe limited to 30 keV, since the higher energy 50-80 keVspectrum has been used to calibrate the background.Similar results were obtained for the other sources andthe residuals are of the order of 0.02 counts / sec / keV asexpected from the blank sky observations.Table 2 lists the properties of the background sub-tracted lightcurves for the sources and the flux in the4-20 keV band. Mrk 0926 and Mrk 110 are clearly de-tected with a background subtracted average count rateof 8 and 6 c / s (Table 2), but show no variability withthe observed standard deviation of the same order asthe expected one. On the other hand NGC 4593 has alower count rate of 4 c / s but shows slight evidence ofvariability with a r.m.s of √ ( σ − σ EV ) ∼ . / s. TheAstroSat observation of extra galactic source LMC X-1has been reported by Mudambi et al. (2020). It showsa count rate of 19 c / s with clear evidence of variabil-ity with r.m.s of ∼ . / s. These results show that aLAXPC observation of a source with a 4-20 keV bandflux ' × − ergs / s / cm (i.e. ' / s or ' ∼ . / scan be detected which translates to a fractional r.m.s of10% for a 5 milli-crab source.
4. Summary and Discussion
We have presented an alternate scheme to estimatethe background spectrum and lightcurve for AstroSatLAXPC 20 based on using the detected count rate in the50-80 keV as a measure of the background. A softwarethat incorporates the scheme is available at AstroSatScience Support Cell . The software also can com-pute the estimated background lightcurves and spectra http: // astrosat-ssc.iucaa.in / ?q = laxpcData for LAXPC 10 using the same technique. However,since LAXPC 10 has an higher background with largeruncertainty than LAXPC 20, it is recommended thatLAXPC 20 should be primarily used for such analysisand LAXPC 10 results to be taken as a corroboration.The scheme can be perhaps be improved by explor-ing possibilities to better estimate the background rate.This includes considering only single events to see ifthe correlation between the low and high energy countsin the blank sky spectra is tighter. The correlation be-tween the low and high energy count rate does vary fordi ff erent blank sky observations and one can examineif any satellite system parameter can be used to predictthe variation. A related idea would be to study the cor-relation as a function of latitude and longitude to seeif there is any predictable trend. These improvementsmay lead to a background estimation closer to the Pois-son limit for a 30 ksecs exposure.The scheme is applicable to faint sources wherethe source contribution to the 50-80 keV count rate isless than 0.25 counts / sec (15 milli-crabs or 6 × − ergs / s / cm ) and is limited to the energy range 4-30 keV.The systematic uncertainty in the background spectrawill dominate over the statistical error for exposureslarger than 5 ksecs and is of the same order as thatfrom the standard technique. Thus, it will be prudentto confirm spectral results using both techniques. Thetechnique allows for background lightcurve estimationon time-scales larger than 100 seconds. The spectrumof a ∼ ' × − ergs / s / cm ) can be constrained by LAXPCobservations. Since the systematics dominate the Pois-son statistics for exposures greater than 5 ksecs, thesensitivity of the instrument to measure the spectrumof a source does not improve for exposures longer than ∼
30 ksecs. Variability of a lightcurve binned at 100seconds can be detected for a level greater than 1 c / s inthe 4-20 keV band, which translates to 10% fractionalr.m.s of a 5 milli-crab source. Since the backgroundlightcurve is estimated using the observed variabilityseen in the high energy band, it is expected to morereliable than the standard technique. Thus, using thistechnique, LAXPC can be used to study both the spec-tral and temporal properties of sources with flux greaterthan 5 milli-crab. Acknowledgements
This publication uses the data from the AstroSatmission of the Indian Space Research Organization(ISRO), archived at the Indian Space Science Data Cen-tre (ISSDC). We thank members of LAXPC instrumentteam for their contribution to the development of the . Astrophys. Astr. (0000) :
Table 1.
Details of the twelve blank sky observations used to generate the template spectra
Target Observation R.A. Decl. Date ExposureID (deg) (deg) (ksecs)Sky-9 75 50 G05 156T09 9000000604 237.37 47.10 2016 Aug 16 32.0Sky-5 T01 132T01 9000000636 57.37 -47.10 2016 Aug 30 27.4Sky-6 C01 015T01 9000000668 7.65 12.55 2016 Sep 15 43.0Sky-10 G06 115T01 9000000734 321.22 -48.68 2016 Oct 16 35.9Sky-6 C02 011T01 9000000850 7.65 12.55 2016 Dec 03 39.4Sky-3 C02 003T01 9000000924 129.48 -27.89 2016 Dec 24 32.1Abell3535 A02 108T01 9000001024 194.45 -28.49 2017 Feb 11 49.1Sky-9 75 50 G07 044T09 9000001334 237.37 47.10 2017 Jun 24 32.9Sky 4u1626 G07 049T02 9000001354 250.00 -70.00 2017 Jul 04 22.4Sky-8 C02 021T01 9000001482 237.39 70.35 2017 Aug 21 39.5Sky-9 75 50 G08 046T09 9000001600 237.37 47.10 2017 Oct 11 28.9Blank Sky 5 255-50 A04 198T01 9000001708 57.37 -47.10 2017 Nov 21 35.6
Table 2 . The average counts, variance and expected variance of the background subtracted lightcurves and the flux in the4-20 keV band for sources used to verify the scheme.
Source R.A. Decl. Date Exposure Time Bins Average Variance ( σ ) √ σ Expected √ σ EV FluxVariance ( σ EV ) 4-20 keV(deg) (deg) (ksec) (sec) c / s ( c / s ) ( c / s ) ( c / s ) ( c / s ) × − ergs cm − s − Blank Sky-8 183.48 22.80 2017 Jan 10 3.98 100 -0.05 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± S r c + B k g B k g C oun t s / s −1012 S r c Time (s)
105 20 50−0.0500.05 C oun t s / s e c / k e V Energy (keV)
Figure 4 . Left panels shows the lightcurves of the blank sky observations from 2017. Top panel shows the total source withbackground count rate, middle panel shows the background and the bottom panel shows the background subtracted countrate from the source. Blank spaces between the lightcurves pertains to the SAA passages. Right figure shows the residuals ofthe energy spectrum.
J. Astrophys. Astr. (0000) : S r c + B k g B k g C oun t s / s S r c Time (s) 14151617 S r c + B k g B k g C oun t s / s −101 S r c Time (s)
105 20 50−0.0500.05 C oun t s / s e c / k e V Energy (keV)
Figure 5 . Left figure shows the lightcurves of the blank sky observations from 2019. Top panel of left figure shows thetotal source with background count rate, middle panel shows the background and the bottom panel shows the backgroundsubtracted count rate from the source. Middle figure shows the same lightcurves when the two increases due to entry of thesatellite in SAA passages are removed from the data. Right figure shows the residuals of the energy spectrum. S r c + B k g B k g C oun t s / s S r c Time (s) −3 C oun t s / s e c / k e V
105 20 50−0.0500.05 C oun t s / s e c / k e V Energy (keV)
Figure 6 . Left panel shows the lightcurves of the source Mrk 0926 (Top panel: total source with background countrate, Middle panel: the background and Bottom panel: the background subtracted count rate). Right panel shows the en-ergy spectrum along with the expected background spectrum (top) and residuals (bottom) after fitting with a power-law model. . Astrophys. Astr. (0000) :
LAXPC instrument. This research has made use ofsoftware provided by the High Energy AstrophysicsScience Archive Research Center (HEASARC), whichis a service of the Astrophysics Science Division atNASA / GSFC. The authors thank the referee, Keith MJahoda, for suggestions and comments which substan-tially improved the manuscript.