Sagittarius A* High Energy X-ray Flare Properties During NuSTAR Monitoring of the Galactic Center from 2012 to 2015

TL;DR

This study analyzes ten high-energy X-ray flares from Sagittarius A* observed by NuSTAR between 2012 and 2015, revealing spectral properties and supporting a synchrotron emission mechanism.

Contribution

First comprehensive analysis of hard X-ray flare spectra from Sagittarius A* using NuSTAR data over multiple years.

Findings

Flares have photon indices between 2.0 and 2.8.

No spectral hardening observed with increased brightness.

Quiescent state spectrum and upper luminosity limit established.

Abstract

Understanding the origin of the flaring activity from the Galactic center supermassive black hole, Sagittarius A*, is a major scientific goal of the NuSTAR Galactic plane survey campaign. We report on the data obtained between July 2012 and April 2015, including 27 observations on Sgr A* with a total exposure of ~ 1 Ms. We found a total of ten X-ray flares detected in the NuSTAR observation window, with luminosities in the range of $L_{3-79~keV}$~$(0.2-4.0) \times 10^{35}~erg~s^{-1}$. With this largest hard X-ray Sgr A* flare dataset to date, we studied the flare spectral properties. Seven flares are detected above 5{\sigma} significance, showing a range of photon indices ({\Gamma} ~ 2.0-2.8) with typical uncertainties of +/-0.5 (90% confidence level). We found no significant spectral hardening for brighter flares as indicated by a smaller sample. The accumulation of all the flare spectra in 1-79 keV can be well fit with an absorbed power-law model with {\Gamma}=2.2+/-0.1, and does not require the existence of a spectral break. The lack of variation in X-ray spectral index with luminosity would point to a single mechanism for the flares and is consistent with the synchrotron scenario. Lastly, we present the quiescent state spectrum of Sgr A*, and derived an upper limit on the quiescent luminosity of Sgr A* above 10 keV to be $L_{Xq, 10-79 keV}$ < $(2.9{\pm}0.2) \times 10^{34}~erg~s^{-1}$.

Figures

Tables

Table 1. Table 1 NuSTAR  Galactic Center Observations During 2012 to 2015 and Simultaneous X-ray Observations

NuSTAR Obs				Joint Obs
Target	obsID	Start(UTC)	Exp	Instrument	obsID	Start(UTC)	Exp
Sgr A⋆	30001002001	2012-07-20 02:11:07	154.2 ks	Chandra	13842	2012-07-21 11:52:48	191.7 ks
Sgr A⋆	30001002003	2012-08-04 07:56:07	77.1 ks	Chandra	13852	2012-08-04 02:36:57	156.6 ks
Sgr A⋆	30001002004	2012-10-16 18.31:07	49.6 ks	Chandra	13851	2012-10-16 18:48:57	107.1 ks
Mini-survey	40010001002	2012-10-13 06:41:07	23.9 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Mini-survey	40010002001	2012-10-13 19:21:07	24.2 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Mini-survey	40010003001	2012-10-14 09:56:07	24.0 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Mini-survey	40010004001	2012-10-15 00:31:07	24.0 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29	30001002006	2013-04-26 01:01:07	37.2 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29	80002013002	2013-04-27 06:16:07	49.8 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29	80002013004	2013-05-04 17:31:07	38.6 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29	80002013006	2013-05-11 14:26:07	32.7 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29 w/T1*	80002013008	2013-05-18 17:36:07	39.0 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29 w/T1	80002013010	2013-05-27 10:16:07	37.4 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29	80002013012	2013-06-14 09:56:07	26.7 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29	80002013014/6	2013-06-07 04:16:07	29.5 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29 w/T2**	80002013018	2013-07-31 01:56:07	22.3 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29 w/T2	80002013020	2013-08-08 15:01:07	12.0 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29 w/T2	80002013022	2013-08-09 09:01:07	11.2 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
SGR J1745$-$29 w/T2	80002013024	2013-08-13 00:06:07	11.7 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Sgr A⋆ w/T2	30001002008	2014-06-18 02:21:07	33.1 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Sgr A⋆ w/T2	30001002010	2014-07-04 10:36:07	61.3 ks	Chandra	16597	2014-07-05 02:14:47	16.5 ks
Sgr A⋆ w/T2	30002002002	2014-08-30 19:45:07	59.8 ks	XMM-Newton	0743630201	2014-08-30 19:20:01	33.9 ks
				XMM-Newton	0743630301	2014-08-31 20:23:30	26.9 ks
				Chandra	16217	2014-08-30 04:49:05	34.5 ks
Sgr A⋆ w/T2	30002002004	2014-09-27 17:31:07	67.2 ks	XMM-Newton	0743630401	2014-09-27 17:30:23	33.5 ks
				XMM-Newton	0743630501	2014-09-28 21:01:46	39.2 ks
Sgr A⋆ w/T2	30002002006	2015-02-25 23:41:07	29.2 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Sgr A⋆ w/T2	30002002008	2015-03-31 04:41:07	25.7 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Sgr A⋆ w/T2	30002002010	2015-04-01 06:31:07	14.4 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Sgr A⋆ w/T2	30002002012	2015-04-02 08:21:07	13.1 ks	$\cdots$	$\cdots$	$\cdots$	$\cdots$

Table 2. Table 2 NuSTAR  Flares and simultaneous detection by Chandra / XMM-Newton

NuSTAR				Joint Obs
Flare	Start (UT)	Coverage(s)	Significance($\sigma$)	Instrument	Start(UT)	Duration(s)	Significance($\sigma$)
Nu1 (J20)	2012-07-20 12:15:21	920	5	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Nu2 (J21-1)	2012-07-21 01:45:15	1238	7	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Nu3 (J21-2)	2012-07-21 06:01:12	3099	20	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Nu4	2012-08-05 08:20:17	1319	2	Chandra	2012-08-05 07:41:54	3623	3
Nu5	2012-10-15 01:11:10	822	3	$\cdots$	$\cdots$	$\cdots$	$\cdots$
Nu6 (O17)	2012-10-17 19:50:08	1249	20	Chandra	2012-10-17 19:35:09	5900	11
Nu7 (VB3)	2014-08-30 23:44:15	1215	14	XMM	2014-08-30 23:42:08	2727	10
Nu8 (B3)	2014-08-31 04:23:41	1104	8	XMM	2014-08-31 04:31:35	1469	6
Nu9 (B4)	2014-09-01 01:08:17	2175	5	XMM	2014-09-01 00:43:38	4359	15
Nu10 (B5)	2014-09-29 06:06:55	6273	2	XMM	2014-09-29 06:06:55	7655	6

Table 3. Table 3 Power-law model for the Chandra and NuSTAR data of flare Nu6.

Parameters	Value
$N_H$ (1023 cm-2)	${1.5}_{-0.2}^{+0.3}$
$\Gamma$	${2.06}_{-0.16}^{+0.19}$
Flux (${10}^{-11}$ erg cm2 s-1)	$6.2\pm 0.6$
${\chi }_{\nu }^2$ (DoF)	0.94 (57)

Table 4. Table 4 NuSTAR  Flares and simultaneous detection by Chandra / XMM-Newton

Flare	$\Gamma$	$F_{abs,3-79\mathrm{keV}}$	$L_{3-79\mathrm{keV}}$	Flare strength
		(${10}^{-11}$ erg cm-2 s-1)	(${10}^{35}$ erg s-1)
Nu1 (J20)	$2.6\pm 0.9$	${0.7}_{-0.3}^{+0.6}$	${0.7}_{-0.3}^{+0.4}$	${18}_{-8}^{+13}$
Nu2 (J21-1)	$2.8\pm 0.6$	$0.9\pm 0.3$	$0.9\pm 0.3$	${25}_{-8}^{+13}$
Nu3 (J21-2)	$2.3\pm 0.2$	$2.2\pm 0.4$	$2.1\pm 0.3$	${35}_{-7}^{+10}$
Nu6 (O17)	$2.1\pm 0.2$	$4.4\pm 0.7$	$4.0\pm 0.5$	${54}_{-11}^{+14}$
Nu7 (VB3)	$2.3\pm 0.2$	$2.4\pm 0.3$	$2.3\pm 0.3$	${43}_{-9}^{+11}$
Nu8 (B3)	$2.2\pm 0.5$	$1.8\pm 0.3$	$1.7\pm 0.3$	${34}_{-9}^{+10}$
Nu9 (B4)	$2.2\pm 0.6$	${0.8}_{-0.3}^{+0.5}$	${0.7}_{-0.2}^{+0.4}$	${15}_{-7}^{+11}$
Nu5	$3\pm 2$	$0.5\pm 0.3$	$0.6\pm 0.3$	${18}_{-9}^{+11}$
Nu4	$2\pm 2$	${0.4}_{-0.2}^{+0.3}$	${0.3}_{-0.2}^{+0.3}$	$4_{-3}^{+4}$
Nu10 (B5)	$3\pm 3$	$\le 0.15$	$\le 0.2$	$\le 6$

Full text

1. Introduction

Sagittarius A*⋆* (Sgr A*⋆), located at the Galactic nucleus of the Milky Way Galaxy, is one of the most underluminous supermassive black holes (SMBH) known. The current quiescent bolometric luminosity of Sgr A⋆* is $L\simeq 10^{36}\rm erg~{}s^{-1}$, which is roughly eight orders of magnitude lower than its Eddington luminosity of a $4\times 10^{6}M_{\sun}$ black hole (Narayan et al., 1998; Ghez et al., 2008). However, there has been observational evidence indicating that Sgr A*⋆* could have been much brighter in the past (e.g. Zhang et al., 2015; Ponti et al., 2013 and the reference therein). As the closest SMBH to Earth (Reid & Brunthaler, 2004), Sgr A*⋆* is an ideal laboratory to study accretion processes of quiescent black hole systems (Falcke & Markoff, 2013).

The X-ray emission of its quiescent state comes from an optically thin thermal plasma with $kT\sim 2$ keV that extends out to the Bondi radius about $10^{5}$ times the gravitational radii ($r_{B}\sim 10^{5}r_{g}$) (Quataert, 2002; Baganoff et al., 2003; Wang et al., 2013). The X-ray quiescent state of Sgr A*⋆* is punctuated by flares lasting up to a few hours (e.g. Baganoff et al., 2001; Porquet et al., 2003; Dodds-Eden et al., 2009; Trap et al., 2011; Neilsen et al., 2013, 2015; Degenaar et al., 2013; Barrière et al., 2014; Ponti et al., 2015). During the flares, the X-ray luminosity of Sgr A*⋆* increases by a factor of up to a few hundred over the quiescent level (Porquet et al., 2003; Nowak et al., 2012). Fast variability with timescales of a few hundred seconds (Porquet et al., 2003; Nowak et al., 2012; Barrière et al., 2014) suggests a compact emission region within a few gravitational radii from the black hole ($r_{g}/c=20\rm~{}s$). Therefore, flares hold the key to probe the physical conditions in the immediate vicinity of the SMBH.

After a decade of intense Sgr A*⋆* monitoring, there still remain many puzzles regarding the origin of the flaring activity (e.g. see review by Genzel et al., 2010). Two distinctively different classes of models have been proposed as the origin of the flares: electron acceleration processes (Markoff et al., 2001; Liu et al., 2004; Yuan et al., 2004; Dodds-Eden et al., 2010; Dibi et al., 2014), and transient events in the Sgr A*⋆* accretion flow (Tagger & Melia, 2006; Broderick & Loeb, 2005; Yusef-Zadeh et al., 2006; Eckart et al., 2006b; Trap et al., 2011; Zubovas et al., 2012). The flare models mentioned above invoke two types of radiation meachanisms for the X-ray flares: 1) synchrotron emission (with cooling break, or SB model) where the NIR to the X-ray emission is generated from one population of electrons; 2) inverse Compton (IC) emission where the NIR emitting electrons up-scatter the NIR synchrotron emission itself (i.e. synchrotron self-Compton, SSC) or the sub-mm photons from the environment (external Compton, EC). Recent multi-wavelength observations of a bright Sgr A*⋆* flare indicate synchrotron emission with a cooling break and an evolving high energy cut-off as the most likely mechanism (Ponti et al., 2017).

Dozens of Sgr A*⋆* X-ray flares have been observed so far, mainly by Chandra, XMM-Newton and Swift. As different flare radiation models predict different spectral shapes, the spectral properties of these flares carry vital information for us to understand the radiation mechanisms and ultimately the physical processes behind the flares. Recent studies discussed whether the flare spectral shapes depend on the luminosities (Porquet et al., 2003; Nowak et al., 2012; Degenaar et al., 2013). During the Chandra Sgr A*⋆* X-ray Visionary Project (XVP), thirty-nine X-ray flares were detected in 2–8 keV (Neilsen et al., 2013). Data in this relatively narrow bandwidth did not provide evidence for X-ray color differences between faint and bright flares. The analysis of the XMM-Newton data confirms this result in the 3–10 keV energy band; however, it suggests spectral evolution within each flare (Ponti et al., 2017).

The flare spectrum beyond 10 keV has the potential to help distinguish between the synchrotron-type model (which predicts a single power-law spectrum) and the IC-type model (which instead predicts an X-ray spectrum with curvature). Using the 3–79 keV data obtained by NuSTAR in 2012, Barrière et al. (2014) for the first time reported different spectral indices between two flares, with a harder spectrum detected for the brighter flare at 95% confidence level. However, due to limited statistics and a limited number of flares, neither emission mechanism could be ruled out. While the SB model has been preferred for its more physical parameters (Dodds-Eden et al., 2009; Barrière et al., 2014), Dibi et al. (2016) shows some challenges to this model through the first statistical study of flare models using Chandra observations. More X-ray flares detected in the broad X-ray band with good statistics need to be accumulated to answer these unsolved questions.

Aiming at building a large database of X-ray flares of different luminosities, durations, and spectra, NuSTAR has been monitoring Sgr A*⋆* through the Galactic Center observing campaign since its launch in 2012. In this paper we report on the NuSTAR Galactic Center observing campaign, and our Sgr A*⋆* flare study results using data obtained from 2012 to 2015. We searched for X-ray flares from all 27 Galactic Center observations with Sgr A*⋆* in the field of view (FoV), totaling $\sim 1$ Ms of exposure time. Besides the four flares reported in Barrière et al. (2014), six more Sgr A*⋆* hard X-ray flares were detected, resulting a total of ten NuSTAR flares, seven simultaneously detected by Chandra or XMM-Newton. Using the largest broadband X-ray flare database by far, we investigated the spectral properties for all the flares. The paper is organized as follows. In Section 2, we introduce the NuSTAR Galactic Center observation campaign. In Section 3, we present the data reduction. We demonstrate the flare search results in Section 4. In section 5, we present the spectral properties for Sgr A*⋆* flares and quiescent state, which are discussed in Section 6.

2. NuSTAR Galactic Center Observing Campaign

Sgr A*⋆* is a key target of the NuSTAR Galactic Center campaign. The first Sgr A*⋆* observation was initiated in 2012 July as a coordinated observation campaign with Chandra and Keck. Three NuSTAR Galactic Center observations resulted in 375 ks total exposure time, during which four bright flares with X-ray luminosity in the range of $L_{3-79\rm~{}keV}=(0.73-3.97)\times 10^{35}\rm erg~{}s^{-1}$ were detected by NuSTAR up to 79 keV (Barrière et al., 2014). The bright flare detected in October 2012 was simultaneously detected by Chandra, while no X-ray flare was covered by the Keck observation window. The Sgr A*⋆* region was also covered by four out of six pointings ($\sim 25$ ks exposure each) of the NuSTAR Galactic Center mini-survey conducted in October 2012 (Mori et al., 2015).

In 2013, major X-ray observatories, including Chandra, XMM-Newton and Swift, conducted long Sgr A*⋆* observing campaigns in order to investigate potential variation in Sgr A*⋆* X-ray activity caused by pericenter passage of the very red Br$\gamma$ object called G2 (Gillessen et al., 2012; Witzel et al., 2014). A recent study of all 150 XMM-Newton and Chandra Galactic Center observations over the last 15 years reports a significant increase in the number and average luminosity of bright flares happening after the pericenter passage of G2 (Ponti et al., 2015). It is still uncertain whether this variation is due to clustering of bright flares observed during more frequent monitoring or increased accretion activity induced by G2. The outburst of SGR J1745$-$29 (Kennea et al., 2013; Mori et al., 2013; Rea et al., 2013), a transient magnetar only 2.4*′′* from Sgr A*⋆, triggered further observations of the Galactic Center region in 2013. Later in 2013, two X-ray transients, CXOGC J174540.0$-$290005 and AXJ 1745.6$-$2901, went into outburst at different times (see ATELs 5095, 5074, 5226, 1513). NuSTAR allocated a total of $\sim 380$ ks to monitor these Galactic Center transient phenomena in 2013. These observations were dominated by the bright X-ray transients, thus making it impossible for NuSTAR to characterize even the brightest Sgr A⋆* flares.

As the magnetar SGR J1745$-$29 became less dominant, another 100 ks NuSTAR observation was allocated to a multi-wavelength Sgr A*⋆* observation campaign coordinated with Chandra and Spitzer in summer 2014. A third multi-wavelength campaign (NuSTAR  XMM-Newton  SINFONI-VLT and VLBA) was performed after the pericenter passage of G2 (see Ponti et al., 2017). A summary of all the 27 NuSTAR observations with Sgr A*⋆* in the FoV are listed in Table 1.

3. Data Reduction

3.1. NuSTAR

We analyzed all the existing NuSTAR Galactic Center observations with Sgr A*⋆* in the FoV, resulting in 27 observations with a total exposure of $\sim 1$ Ms. We reduced the data using the NuSTAR Data Analysis Software NuSTARDAS v.1.3.1. and HEASOFT v.6.13, filtered for periods of high instrumental background due to SAA passages and known bad detector pixels. Photon arrival times were corrected for on-board clock drift and precessed to the Solar System barycenter using the JPL-DE200 ephemeris. For each observation, we registered the images with the brightest point sources available in individual observations, improving the astrometry to $\sim 4^{\prime\prime}$. We used a source extraction region with 50*′′* radius centered on the radio position of Sgr A*⋆* at R.A.=266.41684*∘, Decl.=–29.00781∘* (J2000) (Reid & Brunthaler, 2004). Then we extracted 3-30 keV light curves in 300 s bins with deadtime, PSF, and vignetting effect corrected. For all 27 observations we examined the data obtained by both focal plane modules FPMA and FPMB, and made use of those not heavily contaminated by ghost-rays from distant bright X-ray sources.

To derive the NuSTAR flare spectra, we used the same source region we adopted when extracting the light curves to extract both the source and background spectra. The source spectrum was extracted from the flaring intervals determined by the flare search method (see Section 4), The background spectrum was extracted from off-flare intervals for each flare in the same observation. Spectra of FPMA and FPMB were combined and then grouped with a minimum of $3\sigma$ signal-to-noise significance per data bin, except the last bin at the high-energy end for which we require a minimum significance of $2\sigma$.

3.2. Chandra

Chandra observed Sgr A*⋆* 38 times at high spectral resolution with the HETGS during the 2012 XVP campaign (Neilsen et al., 2013). Three of these observations were coordinated with the NuSTAR pointings; the details of the overlapping observations are listed in Table 1. For the present analysis, we used the same Chandra data extraction as Neilsen et al. (2013). Briefly, this involves processing with standard tools from the ciao software package (v.4.5), identifying photons dispersed by the transmission gratings using the diffraction equation, and extracting events from a small extraction region (a 2.5 pixel radius circle for the zeroth order photons and 5 pixel wide rectangular strips for the first order dispersed photons) to limit the background. Finally, we extracted 2–8 keV light curves in 300 s bins.

For the spectral analysis, we used the same extraction region as for the light curves to create zeroth order and first order grating spectra and responses. Since we are interested in the flares, we extracted spectra for the on-flare and off-flare time intervals separately, using the off-flare periods as background spectra to be subtracted. To account for pileup in the zeroth order spectra, we used the pileup kernel developed by Davis (2001), although the pileup parameter is poorly constrained by the data.

3.3. XMM-Newton

We reduced the XMM-Newton data using version 13.5.0 of the XMM-Newton SAS software. We extracted the source photons from a circular region with 10″ radius centered on Sgr A*⋆. For each flare we extracted source photons during the time window defined by the Bayesian block routine, adding 200 s before and after the flare. Background photons have been extracted from the same source regions by selecting only quiescent periods. The count rate of even the brightest Sgr A⋆* flares are below the pile-up count rate threshold of 2 cts s*-1*, providing XMM-Newton the key advantage of being able to collect pile-up free, and therefore unbiased, spectral information even for the brightest flares. For more details of the XMM-Newton data reduction, see Ponti et al. (2015, 2017).

4. Flare Search

4.1. Flare Search Methods

For the NuSTAR observations, we applied Bayesian block analysis to the combined FPMA and FPMB light curves as described in Barrière et al. (2014). The Bayesian block analysis addresses the problem of detecting and characterizing local variance in the light curves, e.g. transient phenomena (Scargle et al., 2013). This Bayesian statistics based method represents the signal structure as a segmentation of the time interval into blocks (or subintervals) separated by change points. The statistical properties of the signal change discontinuously at the change points but are constant within one block. Therefore, the time range of the observation is divided into blocks, where the count rate is modeled as constant within errors. This analysis has been by far one of the most popular methods for detecting and characterizing Sgr A*⋆* X-ray flares (Nowak et al., 2012; Neilsen et al., 2013; Ponti et al., 2015; Mossoux et al., 2015).

We used the Bayesian block analysis algorithm as described by Scargle et al. (2013). The dynamic programming algorithm employs a Monte Carlo derived parametrization of the prior on the number of blocks and finds the optimal location of the change points. The number of change points is affected by two input parameters: the false positive rate $fpr$, which quantifies the relative frequency with which the algorithm falsely reports detection of change points with no signal present, and the prior estimate of the number of change points, $n_{\rm cp-prior}$. For the NuSTAR data, we adopted the same parameters as used in Barrière et al. (2014), i.e. $fpr=0.01$ and a geometric prior $n_{\rm cp-prior}=4-\log(fpr/0.0136~{}N^{0.478})$, where $N$ is the total number of events.

The same Bayesian block analysis algorithm was modified to read XMM-Newton events files and applied to all the XMM-Newton observations as well, as described in Ponti et al. (2015). For the Chandra observations, both direct fits (with one or more Gaussian components superimposed on a constant background) and Bayesian block analysis were adopted for the Chandra X-ray light curves to detect and characterize X-ray flares, as described in detail in Neilsen et al. (2013) and Ponti et al. (2015). The properties of the detected Chandra flares are not sensitive to the detection algorithm.

4.2. Flare Detection Results

As our NuSTAR X-ray flare database gets larger, from now on we name all the flares in chronological order, along with other publication names, if any. Table 2 lists the name, start time, duration and detection significance for the ten flares as detected by NuSTAR, and as detected by Chandra or XMM-Newton if there is a simultaneous observation.

4.2.1 2012 Joint Sgr A*⋆* Observing Campaign and Mini-survey: Six Flares Detected

For the three 2012 NuSTAR Sgr A*⋆* observations (ObsID 30001002001, 30001002003, 30001002004), the Bayesian block analysis led to detection of four bright X-ray flares from Sgr A*⋆* (for details see Barrière et al., 2014). Three out of the four bright flares were detected in a row within $\sim 20$ hrs from 2012 July 20 to July 21, named as flares Nu1(J20), Nu2(J21-1) and Nu3(J21-2) with durations of $\sim 920$ s, $\sim 1238$ s and $\sim 3099$ s respectively. The baseline count rate of the Sgr A*⋆* region in 3–79 keV is $0.59\pm 0.01$ cts s*-1* (all count rates given with 1$\sigma$ error bars). The baseline emission is dominated by faint X-ray point sources and diffuse emission around Sgr A*⋆, while the instrument background contributes $<5\times 10^{-3}\rm~{}cts~{}s^{-1}$. During the flares, the count rate in the same source region reaches $0.73\pm 0.03\rm~{}cts~{}s^{-1}$ for flare J20, $0.80\pm 0.03\rm~{}cts~{}s^{-1}$ for Nu2(J21-1), and $1.05\pm 0.02\rm~{}cts~{}s^{-1}$ for Nu3(J21-2). The fourth bright flare, noted as Nu6(O17), reported in Barrière et al. (2014) was simultaneously detected by Chandra and NuSTAR on 2012 October 17. This bright flare results in a significant detection level of $\geq 10\sigma$ for both X-ray observatories. Compared with the full profile of this flare obtained by Chandra, NuSTAR captured the peak $\sim 1249$ s of the flare. The NuSTAR flare peak count rate reaches $1.20\pm 0.02\rm~{}cts~{}s^{-1}$, while the baseline emission maintains at the same level as in the 2012 July observation ($0.59\pm 0.01$ cts s-1*).

Below we report two new flares detected from the 2012 Galactic Center observation campaign. First, to search for fainter flares, we compared the NuSTAR observations with the simultaneous Chandra observations. In the coordinated 2012 Chandra observations (ObsID 13842, 13852, 13851), the direct fit algorithm detected seven flares, which was further confirmed by the Bayesian block analysis method (Table 1, Neilsen et al., 2013). By comparing the duration of these seven Chandra flares and the NuSTAR observation good time intervals (GTIs), we found two more flares covered by the NuSTAR observations. For one of the two flares, merely $\sim 100$ s of exposure time is covered by the NuSTAR GTIs, resulting in poor statistics for any meaningful analysis. We therefore exclude this flare from our study. The other faint flare was detected by Chandra on 2012 August 5 with a $\sim 3\sigma$ detection. The NuSTAR GTIs of the observation 30001002003 partly covered this flare, resulting in a marginal detection ($\sim 2.5\sigma$). While the Sgr A*⋆* region baseline emission remains the same as in 2012 July ($0.59\pm 0.01$ cts s*-1*), the NuSTAR 3–79 keV count rate of this flare is $0.64\pm 0.02$ cts s*-1*. Because of its low count rate relative to the baseline count rate, flare Nu4 is not significant in the NuSTAR data alone.

We also searched for Sgr A*⋆* flaring activities using the observations from 2012 NuSTAR Galactic Center Mini-survey (Mori et al., 2015). Four of the six observations have the Sgr A*⋆* region included in the FoV (ObsID 40010001002, 40010002001, 40010003001, 40010004001). We performed the Bayesian block analysis on these four observations, following the procedures described in Barrière et al. (2014). An increase of Sgr A*⋆* X-ray flux is detected at $\sim 3.3\sigma$ significance level on 2012 October 15 (hereafter flare Nu5). During 2012 October, the Sgr A*⋆* baseline emission count rate is $0.57\pm 0.01\rm~{}cts~{}s^{-1}$, consistent with that of 2012 July, while the count rate during flare Nu5 is $0.80\pm 0.07$ cts s*-1*. There were no joint observations of the Galactic Center during the Mini-Survey, so we have no additional constraints on the properties of the flare.

4.2.2 2013 NuSTAR Galactic Center Transient Observations: No Flares Detected

When the magnetar SGR J1745-29 (merely 2.4*′′* away from Sgr A*⋆) went into outburst in 2013 April with a peak flux of $F_{1-10\rm~{}keV}\sim 2\times 10^{-11}\rm~{}erg~{}cm^{-2}~{}s^{-1}$, the Sgr A⋆* source region was dominated by the X-ray emission from the magnetar (e.g. Mori et al., 2013; Rea et al., 2013; Ponti et al., 2015). The severe contamination from the magnetar prevents a clear detection and clarification of even bright X-ray flares for observations 30001002006 to 80002013024 (see Table 1). During the magnetar monitoring campaign, flare detections further suffered from PSF wing contamination from two nearby X-ray transients CXOGC J174540.0$-$290005 and AXJ 1745.6$-$2901, which went into outburst in 2013 May and July respectively (see Section 2). The baseline emission from the Sgr A*⋆* area was therefore highly variable due to contamination from the three bright X-ray transients. A routine flare search via Bayesian block analysis on the $\sim 380$ ks Galactic Center observations conducted in 2013 found no significant Sgr A*⋆* flaring activity, as NuSTAR was not sensitive to flares with luminosities lower than 50 times the Sgr A*⋆* quiescent luminosity during this period.

4.2.3 2014 Joint Observing Campaign:

No X-ray Flares Detected

During the 100 ks Sgr A*⋆* observations coordinated with Chandra and Spitzer(obsID 30001002008, 30001002010 for NuSTAR; obsID 16597 for Chandra) the X-ray flux of the magnetar SGR J1745-29 had dropped to $F_{1-10\rm~{}keV}\sim 2\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$, allowing adequate characterization of Sgr A*⋆* X-ray flares. In the 16.5 ks Chandra observation (obsID 16597), we found no Sgr A*⋆* flaring activity via a direct light curve fit. Since the X-ray transient AX J1745.6$-$2901 was still bright in our observation, it increased the NuSTAR baseline count rate to $0.84\pm 0.02$ cts s*-1*, which is $\sim 50\%$ higher than in 2012. Due to the increased baseline emission from the transient, we can only say that there were no flares with luminosities above 20 times the quiescent luminosity during this campaign. Around 2014 June 18 UT 09:24, Sgr A*⋆* flaring activities were detected by Spitzer, but we found no X-ray counterpart for this flare. The Spitzer flare characteristics will be discussed elsewhere.

4.2.4 2014-2015 Joint Observing Campaign:

Four Flares Detected

Four X-ray flares were simultaneously detected by XMM-Newton and NuSTAR in 2014 fall (obsID 30002002002, 30002002004 for NuSTAR; obsID 0743630201, 0743630301, 0743630401, 0743630501 for XMM-Newton). Three out of the four flares were detected in a row within $\sim 26$ hrs on 2014 August 30, August 31 and September 1, hereafter flare Nu7, Nu8 and Nu9. XMM-Newton was able to capture the full flare profile for all three flares (Ponti et al., 2015, 2017). However, due to interruptions caused by Earth occultations, NuSTAR GTIs only captured the rising half (1215 s) of flare Nu7, 518 s of the rising stage of flare Nu8, and half of flare Nu9 (see Figure 1).

This is the second time that multiple flares are detected by NuSTAR roughly within one day, which could suggest that bright flares tend to take place in clusters, as also indicated by previous flare studies (Porquet et al., 2008; Ponti et al., 2015). The transient source AXJ 1745.6$-$2901 continued to stay in outburst, therefore continuing to contaminate the Sgr A*⋆* region. During the 2014 Fall NuSTAR observation, the baseline emission from the Sgr A*⋆* region was $0.78\pm 0.02$ cts s*-1*, about 30% higher than that in 2012. XMM-Newton also detected a fainter X-ray flare on 2014 September 29. The NuSTAR observation in the same time range results in a $2\sigma$ detection (hereafter Nu10).

5. Flare Spectral Properties

5.1. The Brightest X-ray Flare Detected by NuSTAR

Flare Nu6 is the brightest X-ray flare detected by NuSTAR. It was simultaneously detected by both NuSTAR and Chandra. While Chandra captured the full flare lasting $\sim 5900$ s (Neilsen et al., 2013), NuSTAR only captured the peak $\sim 1249$ s of the flare, mainly due to interruption by Earth occultation.

The Chandra data does not show spectral evolution within this flare, so we jointly fitted the 1249 s NuSTAR flare peak spectrum in 3–79 keV and the $\sim 5900$ s Chandra full flare spectrum in 0.5–9 keV. We used the Interactive Spectral Interpretation System v.1.6.2-19 (Houck & Denicola, 2000), setting the atomic cross sections to Verner et al. (1996) and the abundances to Wilms et al. (2000). The joint spectrum is well-fit by a simple absorbed power-law, with the dust scattering taken into account for the Chandra spectra (Tbabsdustscatpowerlaw) (Baganoff et al., 2003; Neilsen et al., 2013). We did not use the dust scattering model for the NuSTAR spectra, because with a large extraction region, the photons scattering into and out of the line of sight compensate with each other (Barrière et al., 2014). The best-fitted photon index is $\Gamma=2.06^{+0.19}_{-0.16}$ with an absorption column density $N_{\rm H}=(1.5^{+0.3}_{-0.2})\times 10^{23}~{}{\rm cm}^{-2}$ (Table 3). Both the photon index and the column density are consistent with those derived from NuSTAR spectrum alone ($\Gamma=2.04^{+0.22}_{-0.20}$, $N_{\rm H}=(1.7^{+0.7}_{-0.6})\times 10^{23}\rm~{}cm^{-2}$, Barrière et al., 2014), though better constrained. The spectrum with the best-fit absorbed power-law model for the flare is shown in Figure 2. The 0.5–79 keV unabsorbed flare peak flux is $F_{X}=(6.2\pm 0.6)\times 10^{-11}$ erg cm*-2* s*-1*, corresponding to a luminosity of $L_{X}=(4.7\pm 0.5)\times 10^{35}$ erg s*-1* assuming the distance to the Galactic Center is 8.0 kpc (Reid & Brunthaler, 2004). This is by far the brightest X-ray flare detected by NuSTAR and one of the brightest flares detected by Chandra.

5.2. Spectral Properties of All Ten Flares

We analyzed the X-ray spectra of all the X-ray flares detected by NuSTAR, jointly with either Chandra or XMM-Newton when available. We extracted the source spectra from the flare time ranges, and the background spectra from off-flare time ranges. We first focused on the seven flares that are detected with $>5\sigma$ detection significance, i.e. flares Nu1, Nu2, Nu3, Nu6, Nu7, Nu8 and Nu9. The first set of four flares (Nu1, Nu2, Nu3, Nu6) were detected in 2012 Fall, when no X-ray transient in the Galactic Center was detected. Among them, flare Nu6 was simultaneously detected by Chandra. The second set of three flares (Nu7, Nu8, Nu9) was detected jointly by NuSTAR and XMM-Newton in 2014 fall, during which AX J1745.6$-$2901 was still in outburst and increased the Sgr A*⋆* off-flare baseline emission by $\sim 30\%$ through PSF contamination (see section 4.2.4). Therefore, varying baseline emission is an aspect of our data set. In order to make a fair comparison of the flare spectral shapes, below we first examined two factors that could affect joint fitting of all seven flares: 1) AX J1745.6$-$2901 PSF contamination; 2) absorption column density.

First, we checked how the contribution from the transient AX J1745.6$-$2901 would affect the measurements of the 2014 flares Nu7, Nu8 and Nu9. We investigated the light curve and the spectrum of the transient AXJ 1745.6$-$2901 during the flare and the off-flare time ranges in the 2014 observation (obsID: 30002002002) where the second set of three bright flares were detected. Throughout this observation, the transient does not demonstrate significant variation except for eclipses. The 3–79 keV count rate in the 30*′′* region centered on AX J1745.6$-$2901 maintains at $2.00\pm 0.02$ cts s*-1*, while during the eclipse the count rate dropped to $0.34\pm 0.02$ cts s*-1*. No eclipse coincides with any of the three flares. Therefore, when selecting background spectra during the off-flare time range, we excluded eclipses. Next, we compared the spectra of AX J1745.6$-$2901 during and off the flares. Both can be well fit with a simple absorbed power-law model, yielding $N_{\rm H}=(1.8\pm 0.2)\times 10^{23}$ cm*-2* and $\Gamma=1.77\pm 0.03$ with an absorbed 3–79 keV flux of $F_{3-79\rm~{}keV}\sim 9.5\times 10^{-11}$ erg cm*-2* s*-1*. The absorbed 3–79 keV flux during and off the flares was constant at $F_{3-79\rm~{}keV}\sim 9.5\times 10^{-11}$ erg cm*-2* s*-1*. Therefore, the PSF contamination from AX J1745.6$-$2901 within the Sgr A*⋆* region does not have significant variation during and off the flares, and thus can be treated as a constant contribution to the baseline spectrum. However, this elevated off-flare baseline emission from the Sgr A*⋆* region in 2014 (30% higher than in 2012) does cause larger error bars for spectral properties of flares Nu7, Nu8 and Nu9.

Second, we investigated whether the absorption column density $N_{\rm H}$ varies from 2012 to 2014. We fit the two sets of NuSTAR flare spectra separately with an absorbed power-law model, and found that the best-fit values of the absorption column density for each set are consistent with each other, resulting in $N_{\rm H}=(1.7^{+0.7}_{-0.6})\times 10^{23}~{}$cm*-2* for the first set of spectra and $N_{\rm H}=(1.7^{+0.9}_{-0.8})\times 10^{23}\rm~{}cm^{-2}$ for the second set of spectra. Therefore, here we can safely assume that the absorption column density did not vary with time (see Ponti et al. 2017; Jin et al. 2017 for more details).

After investigating the above factors, we proceeded to joint spectral fitting of the seven bright flares. We use the same model as described in Section 5.1, i.e. Tbabspowerlaw for the NuSTAR flare spectra and Tbabsdustscatpowerlaw for the Chandra and XMM-Newton flare spectra. The absorption column density values $N_{\rm H}$ are tied among all the spectra. The photon indices of the spectra associated with the same flare obtained by different instruments are tied; the photon indices of different flares are independent. The power-law normalization are set free. We then performed a joint fit of the seven bright X-ray flares using all available data, resulting in a good fit with $\chi_{\nu}^{2}=1.02$ with DoF of 295. The resultant column density is $N_{\rm H}=(1.6\pm 0.2)\times 10^{23}~{}$cm-2*. Table 4 lists the corresponding best-fit photon index, flux and luminosity for each flare. We also calculated the strength $S$ of each flare, which is defined as the ratio of the unabsorbed 2–10 keV flare flux and the quiescent state when $F_{q}=(0.47^{+0.04}_{-0.03})\times 10^{-12}$ erg cm*-2* s*-1* (Nowak et al., 2012).

Faint flares with strengths less than 30 times the Sgr A*⋆* quiescent flux (Nu1, Nu2 and Nu9, black in Figure 3) have best-fit photon indexes of $\Gamma=(2.2$–$2.8)\pm(0.6-1.0)$; flares with strengths higher than 30 times while lower than 50 times the Sgr A*⋆* quiescent flux (Nu3, Nu7 and Nu8, red in Figure 3) have best-fit photon indexes of $\Gamma\sim 2.3\pm(0.2$–$0.5)$. The brightest flare, Nu6, with strength $\sim 54$ times the Sgr A*⋆* quiescent flux, has the hardest spectrum with a photon index of $\Gamma=2.06\pm 0.17$ (green in Figure 3). To investigate whether brighter flares possess harder spectra, we performed a linear fit to the flare photon index over their strength. We found that the data can be best fit with a linear function $\Gamma=(-0.016\pm 0.010)\times\Gamma+(2.9\pm 0.5)$, with a slope of $a=-0.016\pm 0.010$ and $\Gamma_{0}=2.9\pm 0.5$ (the error bars are given in $1\sigma$ significance level). Given the low significance ($<2\sigma$)of the slope, our results is consistent with no hardening spectra for brighter flares. While the best-fit spectral hardening is $\Delta\Gamma=-0.6$ for flares with strengths from $S=18$ to $S=54$, a spectral hardening of $|\Delta\Gamma|>1.7$ and a spectral softening of $|\Delta\Gamma|>0.5$ can be excluded. A Spearman rank correlation test results in $P>0.10$ (with Spearman’s $\rho$ of $-0.36$), confirming that no strong correlation has been found. Therefore, our current flare dataset does not show an obvious correlation between flare spectral shape and luminosity, although such dependence cannot be excluded. This result is consistent with previous works (Porquet et al., 2008; Nowak et al., 2012; Neilsen et al., 2013; Degenaar et al., 2013; Ponti et al., 2017).

As we now have a larger sample of flares detected in a broad X-ray energy band, we investigated whether the flare spectra require any curvature or spectral breaks by accumulating all the flares. We fit the seven flares with the same absorbed power-law model and parameter settings as discussed above, except that the photon indices $\Gamma$ of all the data sets are now tied with each other. This results in an equally good fit, with $\chi_{\nu}^{2}=1.01$ for DoF of 301. We derived a best fit column density of $N_{\rm H}=(1.5\pm 0.2)\times 10^{23}$ cm*-2* and the flare photon index of $\Gamma=2.2\pm 0.1$. A spectral break is not required by this data set. An energy break below 20 keV can be ruled out by the data. We thus conclude that there is no evidence for a spectral break with this larger flare spectrum sample.

For the three flares with detection significance lower than $5\sigma$ (due to low luminosity or limited time coverage), we tried a joint fitting with absorbed power-law models using Cash statistics (Cash, 1979). While fixing $N_{\rm H}$ to $1.5\times 10^{23}~{}$cm*-2*, the photon indices of the three flares cannot be well constrained, resulting in $\Gamma=(2-3)\pm 3$. All three flares possess luminosity less than 20 times the quiescent level.

5.3. NuSTAR Sgr A*⋆* Quiescent State Emission within a 30*′′* Radius Region

In order to provide an upper limit to the Sgr A*⋆* quiescent state emission above 10 keV, we also measured the spectrum of the baseline emission of the 30*′′* radius region centered on Sgr A*⋆, when the supermassive black hole was in its X-ray quiescent state. The baseline spectrum was extracted from the 2012 Sgr A⋆* observation with Sgr A*⋆* flares removed, during which no X-ray transient activity was detected within the NuSTAR FoV. The source is regarded as an extended source when running the NuSTAR pipeline.

The baseline X-ray emission within 30*′′* radius of Sgr A*⋆* comes from various types of sources, including the supernova remnant Sgr A East, star clusters like IRS 13 and IRS 16, numerous X-ray point sources including G359.95$-$0.04 (a PWN candidate), and local X-ray diffuse emission (Baganoff et al., 2003). Due to the complexity of the baseline emission components, we used a phenomenological model to fit the spectrum. The model we used is a combination of two thermal plasmas, a power-law and a Gaussian representing the 6.4 keV neutral Fe line, all subject to absorption Tbabs*(apec1+apec2+gaussian+power-law), resulting in $\chi_{nu}^{2}=1.01$ with DoF of 199 (see Figure 4). The absorption column density is $N_{\rm H}=(1.7\pm{0.3})\times 10^{23}\rm~{}cm^{-2}$. The best-fit values for the temperature and abundance of the two apec models are $kT_{1}=1.16^{+0.18}_{-0.17}\rm~{}keV$ with $z_{1}=2.1^{+0.7}_{-0.5}$ and $kT_{2}=7.2^{+2.1}_{-1.6}\rm~{}keV$ with $z_{2}=1.5^{+0.9}_{-0.6}$. The photon index of the power-law is $\Gamma=1.64^{+0.17}_{-0.19}$. The total absorbed flux in 2–10 keV and 10–79 keV are measured as $F_{abs,2-10}=(2.8\pm 0.1)\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$ and $F_{abs,10-79}=(3.7\pm 0.1)\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$. The corresponding unabsorbed fluxes in these two energy bands are $F_{unabs,2-10}=(8.2\pm 0.1)\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$ and $F_{unabs,10-79}=(3.8\pm 0.1)\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$. Therefore, we derive the upper limit to the quiescent luminosity of Sgr A*⋆* above 10 keV as $L_{q,10-79\rm~{}keV}=2.9\times 10^{34}$ erg s*-1*.

For comparison, the unabsorbed 2–10 keV flux of Sgr A*⋆* in quiescence measured by Chandra is $F_{2-10}=(0.47^{+0.05}_{-0.03})\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$, contributing to only 5% of the unabsorbed 2–10 keV flux in the 30*′′* radius region of Sgr A*⋆* we measured using NuSTAR. The thermal apec components of the spectrum mainly originates from supernova heating of the interstellar medium, coronally active stars, and non-magnetic white dwarfs (Perez et al., 2015 and references therein). These thermal components becomes negligible towards 20 keV, as shown in Figure 4. The high-energy X-ray emission above 20 keV is dominated by the PWN candidate G359.95$-$0.04 (Wang et al., 2006) and a newly discovered diffuse component dominating above 20 keV, which is likely an unresolved population of massive magnetic CVs with white dwarf masses $M_{WD}\sim 0.9M_{\sun}$ (Revnivtsev et al., 2009; Mori et al., 2015; Perez et al., 2015; Hailey et al., 2016). We compared the measured 20–40 keV Sgr A*⋆* quiescence flux with that of G359.95$-$0.04 and the hard X-ray diffuse emission. Based on the analysis of Wang et al. (2006) on G359.95$-$0.04, its extrapolated 20–40 keV flux falling in the NuSTAR HPD circle (30*′′) is $F_{20-40,PWN}=(0.3\pm 0.1)\times 10^{-12}$ erg s-1*. According to the hard X-ray diffuse emission spatial distribution model (Perez et al., 2015), the 20–40 keV flux of this diffuse component in the inner 30*′′* around Sgr A*⋆* is $F_{20-40,d}=(0.8\pm 0.1)\times 10^{-12}$ erg s*-1*. The sum of the PWN and the hard X-ray diffuse emission 20–40 keV flux is therefore $F_{20-40,PWN+d}=(1.1\pm 0.1)\times 10^{-12}$ erg s*-1*, which is very close to the 20–40 flux of the inner 30*′′* region $F_{20-40}=(1.16\pm 0.05)\times 10^{-12}\rm~{}erg~{}cm^{-2}~{}s^{-1}$ as measured using NuSTAR, leaving about 5% flux from other sources. Therefore, the high-energy flux is dominated by the contribution from the PWN candidate G359.95$-$0.04 and the hard X-ray diffuse emission. It is reasonable to estimate that the contribution of Sgr A*⋆* is also close to 5% above 20 keV, as it is in 2–10 keV.

6. Summary and Discussion

Using the $\sim 1$ Ms NuSTAR Galactic Center observations from 2012 Fall to 2015 Spring, we searched for flaring activity from the supermassive black hole Sgr A*⋆* via Bayesian block analysis and compared our data to simultaneous X-ray observations by Chandra and XMM-Newton to identify additional fainter events. NuSTAR has so far captured a total of ten X-ray flares up to 79 keV. This has allowed us to study the Sgr A*⋆* flare spectral properties with a larger flare sample in a broad X-ray energy band.

Seven flares were significantly detected at $\geq 5\sigma$ confidence, with 3–79 keV luminosities ranging from $L_{3-79\rm~{}keV}\sim(0.7$–$4.0)\times 10^{35}$ erg s*-1*, corresponding to a factor of 15–54 above the quiescent luminosity of Sgr A*⋆* (Table 4). Four out of the seven bright X-ray flares were simultaneously detected with Chandra or XMM-Newton. Three flares are detected at lower significance due to low luminosities or limited time coverage by NuSTAR.

Whether there is spectral dependence on luminosity is important in discriminating and constraining both the flare radiation mechanism and understanding the physical processes behind it. Systematic studies of Sgr A*⋆* flare data obtained by Chandra, XMM-Newton, and Swift have shown no evidence for spectral/color differences among flares with different luminosities (Nowak et al., 2012; Degenaar et al., 2013; Neilsen et al., 2013). By virtue of the broadband spectroscopy with NuSTAR, Barrière et al. (2014) for the first time reported a brighter flare Nu6 (O17) with a harder spectrum than a fainter flare Nu2 (J21-1). However, with a larger NuSTAR flare dataset, we find this trend is detected below $2\sigma$, i.e. suggesting no significant spectral hardening for brighter flares (Figure 3). A spectral hardening of $|\Delta\Gamma|>1.7$ can be excluded for flares with strengths from $S=18$ to $S=54$. As there is no strong evidence for varying spectral index from flare to flare, we accumulated all the NuSTAR flare spectra (with joint Chandra/XMM-Newton spectra when available) and fit with the same model. A simple power-law with $\Gamma=2.2\pm 0.1$ provided a good fit to our current data, requiring no spectral curvature/spectral break. The lack of variation in X-ray spectral index with luminosity and the lack of evidence for spectral curvature would point to a single radiation mechanism for the flares and is consistent with the synchrotron scenario, though the SSC model cannot be ruled out. We note that a recent multi-wavelength study of bright flares reports a tentative detection of spectral evolution during bright flares (Ponti et al., 2017), which needs to be further tested. Since all ten flares reported in this work are only partly captured by the NuSTAR GTIs, we are not able to verify this result using the NuSTAR dataset.

Lastly, we show the spectrum of the inner 30*′′* of the Galaxy when Sgr A*⋆* is in quiescence. While the thermal components become negligible above $\sim 20$ keV, a non-thermal component starts to dominate. This is similar to the spectra from two regions at radii $r\approx 1^{\prime}$-2*′* to the southwest and northeast of Sgr A*⋆* (Perez et al., 2015), where the dominant sources above 20 keV are likely to be an unresolved population of massive magnetic CVs. For the inner 30*′′* region, the dominating sources above 20 keV include not only the contribution from this massive CV population, but also a bright PWN candidate G359.95$-$0.04. We estimate that the Sgr A*⋆* quiescence flux contributes to about 5% of the 20–40 keV flux from the 30*′′* region measured by NuSTAR. The upper limit of the Sgr A*⋆* 10–79 keV luminosity is $L_{q,10-79}=2.9\times 10^{34}$ erg s*-1* when the whole signal from the inner 30*′′* is integrated.

This work was supported under NASA Contract No. NNG08FD60C, and made use of data from the NuSTAR mission, a project led by the California Institute of Technology, managed by the Jet Propulsion Laboratory, and funded by the National Aeronautics and Space Administration. We thank the NuSTAR Operations, Software and Calibration teams for support with the execution and analysis of these observations. This research has made use of the NuSTAR Data Analysis Software (NuSTARDAS) jointly developed by the ASI Science Data Center (ASDC, Italy) and the California Institute of Technology (USA). S.Z. ackowledges support by NASA Headquarters under the NASA Earth and Space Science Fellowship Program - Grant “NNX13AM31H”. J.N. acknowledges funding from NASA through the Einstein Postdoctoral Fellowship, grant PF2-130097, awarded by the CXC, which is operated by the SAO for NASA under contract NAS8-03060. G.P. acknowledges support by the Bundesministerium für Wirtschaft und Technologie/Deutsches Zentrum für Luft- und Raumfahrt (BMWI/DLR, FKZ 50 OR 1604) and the Max Planck Society. We thank all members of the Chandra Sgr A*⋆* XVP collaboration (http://www.sgra-star.com).