Repeated X-Ray Flaring Activity in Sagittarius a G

Repeated X-ray Flaring Activity in Sagittarius A G. Bélanger, A. Goldwurm, Fulvio Melia, Farah Yusef-Zadeh, P. Ferrando, Delphine Porquet, Nicolas Grosso, Robert Warwick

To cite this version:

G. Bélanger, A. Goldwurm, Fulvio Melia, Farah Yusef-Zadeh, P. Ferrando, et al.. Repeated X-ray Flaring Activity in Sagittarius A. The Astrophysical Journal, American Astronomical Society, 2005, 635, pp.1095-1102. hal-00008003

HAL Id: hal-00008003 https://hal.archives-ouvertes.fr/hal-00008003 Submitted on 17 Aug 2005

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. ccsd-00008003, version 1 - 17 Aug 2005 60208 UK rnbe France Grenoble, Z871 USA 85721, AZ ai,France Paris, [email protected] n vlto faceindss h assade and causes the disks, accretion formatio of the evolution including stud phenomena, and the related supermas- for and centre’s laboratory accretion wonderful Galactic of a the provides Thus, hole black sive horizon. event the of n rnin vnsmnfse ihnamr adu of handful mere a dur- within trans- ( produced manifested almost radii photons Schwarzschild is events X-ray that for transient and envelope object, IR ing thin this high-energy optically to of an parent regions to innermost points the it probing for guise hole. i black exist massive that the conditions intriguing surrounding the environment itself the about is questions faintness more many This raises 2003). al. et (Schödel aatcnce,SrA svr i n hnsa esthan less at shines bright and and dim active very typical is more 10 A* wav Sgr unlike accessible excellent nuclei, addition, the galactic and of In all detail in great bands. resolution with spectral th and environment, as spatial nearby well as its i to characteristics, of al us radiative allows (Schödel et proximity A*’s object relative Sgr Its such vestigate 2003). closest holes al. the et black Ghez is supermassive 2003; and of universe existence ev- the compelling the in most for the date with to us includi provides idence A* this, Sgr for that reasons fact several astrophy the are the in There many community. for ical curiosity and fascination of source o nojc fisms,nwtogtt be to thought now mass, its of object an for 7 5 4 3 6 2 1 n e,ti anns a culyb lsigi dis- in blessing a be actually may faintness this yet, And aitru * h lc oea h aatccne,i a is center, Galactic the at hole black the A*, Sagittarius rpittpstuigL using typeset Preprint J 2005 − eateto hsc n srnm,Uiest fLeicest of University Astronomy, and Physics of Department a-lnkIsiuefu xrtreticePyi,Mu Physik, für extraterrestrische Max-Planck-Institute ntemxed ehrh srpriuee omlge 11 Cosmologie, et Astroparticule recherche de Unité mixte eateto hsc n twr bevtr,University Observatory, Steward and Physics of Department eateto hsc n srnm,Nrhetr Univers Northwestern Astronomy, and Physics of Department evc ’srpyiu,DAPNIA d’Astrophysique, Service 9 aoaor ’srpyiu eGeol,UiesteJos Université Grenoble, de d’Astrophysique Laboratoire Mla&Flk 01 ie h digo luminosity Eddington the times 2001) Falcke & (Melia  iwta h aoiyo aigeet edt ehr n not and hard be to tend headings: events non-the Subject flaring of of ks. flar population majority 5 second important the and about that an first second view of the the for presence and 1.9 the ks and suggests 1.5 2.5 about about of lasted indeces flare photon first The 31. n hs bevtosw eetdtoflrsfo g *with A* Sgr from fo flares 2004, two in detected out ( we carried w observations nucleus Here these X Several Galactic ing object. the this Galaxy. from of our observed of observations been center already the have at features understandi located complete hole more black a sive of development our to essential L .B´ G. netgtn h pcrladtmoa hrceitc o characteristics temporal and spectral the Investigating X ∼ 27  9 × A 10 T E tl mltajv 6 v. emulateapj style X .INTRODUCTION 1. 34 r , 12 S r s erg ≡ .G A. 2 lc oepyis—Glx:cne aay ulu X-ray — nucleus Galaxy: — center Galaxy: — physics hole black eto -as binaries X-rays: — neutron GM / DSM  − EETDXRYFAIGATVT NSGTAISA* SAGITTARIUS IN ACTIVITY FLARING X-RAY REPEATED 1 bv h uecn uioiy n n20 ac 1adano and 31 March 2004 on one luminosity: quiescent the above ) / / E,911GfsrYet,Fac;be- France; Gif-sur-Yvette, 91191 CEA, c 2 rruhy9 roughly or , / 22 , 12 / 04 .M F. , Rcie 05Jnay2;Rvsd20 uy28) July 2005 Revised 27; January 2005 (Received ihD871 Germany D-85741, nich M  lc eteo,75005 Berthelot, place r ecse E 7RH, LE1 Leicester er, ≈ fAioa Tucson, Arizona, of p-ore,38041 eph-Fourier, 3.4 3 t,Easo,IL Evanston, ity, .F P. , × × 10 ff 10  csof ects 11 6 05Jnay27 January 2005 cm) ABSTRACT and M ose ng n- e- s- n y ⊙ n . , 12 .G N. , htw o osdrt etefis -a counterpart X-ray first the be to consider now (Bagano we what setsaeo hssuc a hrceie yapower-law a by characterized was index source photon this of state escent of aefo hssuc (Bagano source this from flare eie eklmnst of luminosity peak a derived the iigprino aefo g *oe oe90s They s. 900 of some index over photon A* Sgr flare from a flare th found of a detection of re the portion and reported rising (2003a) detected al. been et have varyin Goldwurm features, with ported. spectral A*, Sgr and of scales direction add time two the day, from this flares To bright level). tional quiescent the above 45 factor (or ece eklmnst of luminosity peak a reached hog ycrto efcmtn(S)(i ei 2001; component Melia & quiescent Marko (Liu this (SSC) self-compton to a synchrotron rise through peak synchrotron give the wavelengths produce sub-mm that electrons of population bu 0k,hdahre htnidxof index photon harder a had ks, 10 about eetwvlnts n h rcse htgv iet thes review). to a for rise 2001 give Falcke that & Melia processes (see the a and an radiation wavelengths, magnetic between ferent strong relationship the the fields, in gravitational particles charged from of characteristics acceleration emission the reconnection, magnetic h -ascmn rmSgtaisA SrA)is A*) (Sgr A* Sagittarius from coming X-rays the f n ece upiigyhg ek21 e luminosity keV 2–10 peak high surprisingly of a reached and atdruhy27k,hdasf pcrlidx( index spectral soft This a (2003). had ks, 2.7 al. roughly et lasted di Porquet quite by was presented flare a flare that of is report A* recent Sgr most from The quiescence). above 25 factor h lc oe(i,Ptoin n ei 04,within 2004), Melia near and flare magnetic Petrosian, a (Liu, in hole electrons black of the from acceleration 2002a), Melia swift & (Liu a accretion of enhancement sudden lumino quiescent source’s ity. the above 160 factor precedented ntefl ftesm year, same the of fall the In nery20,the 2001, early In -a aigatvt nSrA ol eidcdb a by induced be could A* Sgr in activity flaring X-ray  L L go h msinmcaim nti supermas- this in mechanisms emission the of ng mleetosdrn h vn n uprsthe supports and event the during electrons rmal Chandra oa xouetm fnal 0 s Dur- ks. 500 nearly of time exposure total a r X X rsn h eut ftolong two of results the present e eyluminous. very epciey hshr htnidxstrongly index photon hard This respectively. e ryflrswt ayn uain n spectral and durations varying with flares -ray = = ff 4 ek21 e uioiisaot4 times 40 about luminosities keV 2–10 peak (3.6 (2.2 .P D. , ff ta.2001). al. et h obndfi nteEi pcr yield spectra Epic the on fit combined The ta.20)t h ai oreSrA.Tequi- The A*. Sgr source radio the to 2003) al. et ± ±  aepoo ne ihnteerr,adthey and errors, the within index photon flare = Γ 0.4) 0.4) ff 2 × × rn rmtetopeiu vns o it for events, previous two the from erent . 7 5 .W R. , 10 10 − + 0 1 Chandra . . 35 33 9 3 n eie –0kVluminosity keV 2–10 derived a and r s erg r s erg  :osrain stars: — observations s: L Chandra − X − ff L = Γ hro 04August 2004 on ther 1 1 X = hscrepnst nun- an to corresponds This . ta.20) hsflr lasted flare This 2001). al. et ti huh httesame the that thought is It . -a bevtr detected Observatory X-ray 6 .Y F. , = (5.4 0.7 (1.0 XMM-Newton ± ±  eetdtefis X-ray first the detected .,cmail with compatible 0.6, ± 1.0) 0.1) -Z = Γ  × × 10 = Γ 1.3 10 7 34 ± 2.5 35 r s erg .,and 0.6, r s erg ± dif- t − 0.3), 1 the − (a s- i- d g e e 1 - t 2 a jet (Markoff et al. 2001), or perhaps through some other nonthermal process within a radiatively inefficient accretion TABLE 1   flow (Yuan, Quataert, & Narayan 2003 and 2004). Further- O L more, short time scale X-ray and IR flares could be much ObsID Obs Start time Obs End time Duration more difficult to detect if the luminosity of SgrA* were (UT) (UT) (ks) even slightly larger in which case the thermal SSC emis- 788 2004-03-28T14:57:08 2004-03-30T04:44:00 112.6 sion would possibly dominate over the non-thermal emission 789 2004-03-30T14:46:36 2004-04-01T04:35:49 123.4 (Yuan, Quataert, & Narayan 2004). The recent detection of 866 2004-08-31T03:12:01 2004-09-01T16:45:58 130.1 SgrA* in the near-IR band in both the quiescent and flar- 867 2004-09-02T03:01:39 2004-09-03T16:35:35 131.4 ing states (Genzel et al. 2003), and the more recent detection of a correlated increase in near-IR and X-ray flux (Eckart et al. 2004), provide further important constraints on building a comprehensive model of this source. Although relatively few 866 and 867, between 2004 August 31 and September 2 for instances of detected flares from SgrA* have been reported, a total exposure time of ∼ 490ks, during which the Epic Mos there appears to be two “types” of flares—the soft and hard— and PN cameras were in PrimeFullWindow and PrimeFull- and a model that can naturally explain both of these types of WindowExtended modes, respectively, during epoch1, and in events is presented in Liu, Petrosian, & Melia (2004). PrimeFullWindow during epoch2. The log of the observa- We here present the results of two long XMM-Newton ob- tions is presented in Table1. We will use epoch1 to refer servations of SgrA* and the Galactic center in which we de- to the observation period spanning ObsID 788 and 789, and tected two flares that rose to a factor of 40 abovethe quiescent epoch2 for the one spanning ObsID 866 and 867. luminosity from the direction of the central black hole, and We generated event lists for the Mos 1, Mos 2, and PN several other, less significant events that could also be coming cameras using the emchain and epchain tasks of the XMM- from SgrA*. Both of these observations were conducted con- Newton Science Analysis System v6.1.0. These were sub- currently with Integral in order to investigate possible links sequently filtered and used to construct images in two en- in the manifestation of a sudden enhanced accretion, mag- ergy bands: 0.5–2 keV, and 2–10 keV. The filter on the netic reconnection, or other types of events in the hard X- event pattern in imaging mode, (PATTERN < = 12 for Mos, and ray (2–12 keV) and soft gamma-ray (20–120 keV) domains. PATTERN < = 4 for PN), ensures that only events created by Furthermore, a set of other observations of SgrA* covering X-rays and free of cosmic ray contamination are selected. Ar- the radio, sub-millimeter and infrared frequencies, coordi- tifacts from the calibrated and concatenated datasets, as well nated with the 2004 XMM-Newton large project, were also as events near CCD gaps or bad pixels are rejected by setting planned and partly performed, and two new transient radio (FLAG==0) as a selection criterion. A further selection on the sources were discovered with the VLA (Bower et al. 2005). maximum count rate in the 10–12 keV range (18 cts/s) for Results from the Integral observations (Bélanger et al. 2005) Mos and 12–14 keV range (22 cts/s) for PN was applied to and from the overall multiwavelength campaing will be re- exclude all periods of increased charged particle flaring activ- ported elsewhere. The two main flares, which we will refer to ity. This stringent selection criterion was only applied to the as the factor-40 flares hereafter, occurred on time scales be- image construction. tween 2500–5000s and have similar spectral characteristics. The heavy absorption in the direction of the Galactic center, These may be added to the statistics of detected flares from prevents photons from this region having energies below 1.5– SgrA* and add weight to the interpretation that there are in- 2 keV from reaching us. For this reason, we used the 0.5– deed two types of flares, bright and soft or not-so-bright and 2keV images of foreground stars to identify counterparts for hard; and that the majority of these events are of the latter astrometric corrections, and the 2–10keV image to determine kind. A larger sample of such events is essential for construc- the flare centroid. tring a more robust classification—and thus identification—of the causes for this flaring activity in SgrA*. Furthermore, a 2.1. Astrometry brightening of the region around SgrA* by a factor of ∼ 2 in We identified three Tycho-2 and two Chandra sources with the 2–10keV band with respect to all previous XMM-Newton positional uncertainties of 0′′.025 and 0′′.16, respectively 8. Of observations of the Galactic center is evident in both datasets. these calibration sources, three were in the central Mos CCD The astrometry-corrected fitted centroid of this emission co- ′′ and two just outside of it. This is an essential point in the as- incides with the position of a bright transient lying about 3 trometric corrections because the positional uncertainty of a south of SgrA* and discovered by Chandra in the summer Mos CCD relative to another is ∼ 1′′.5 and this therefore de- of 2004 (Muno et al. 2005a). Here we will only briefly com- termines the minimum systematic positional uncertainty over ment on the emission characteristics of this new source, CX- the whole field of view. However, if we have at least three cal- OGC J174540.0-290031, during the XMM-Newton observa- ibration sources in the central CCD, then we can reduce this tions and refer the interested reader to Porquet et al. (2005) systematic uncertainty for sources in that CCD. and Muno et al. (2005b) for further details. The astrometric correction was done by first running The structure of this paper is as follows: In §2 we describe edetect chain usingthe 0.5–2keVimagestoget a list ofthe the observations and the analysis methods used to reduce the detected sources with their fitted position and associated sta- data. In §3, we present the results of the analysis, and in §4, tistical uncertainty. Then, using eposcorr that optimizes the we discuss their possible implications and interpretations. 8 The astrometric sources are: Tycho-2 6840-20-1, 6840- 2. OBSERVATIONS AND METHODS OF ANALYSIS 666-1, 6840-590-1, CXOGC J174545.2-285828 and CX- The XMM-Newton satellite observed the Galactic center as OGC J174607.5-285951. Tycho-2 positions were taken from http://www.astro.ku.dk/˜cf/CD/data/catalog.dat and astro- part of a large project during revolutions 788 and 789, be- metric precision from Hog et al. (2000). Chandra positions are from Muno tween 2004March 28 and April 1, and then duringrevolutions et al. (2003) and errors are from Baganoff et al. (2003a) 3

(Yusef-Zadeh et al. 1999). The light curves for each observa- TABLE 2 tion were made from the event files of the Epic instruments;   A C Mos 1, Mos 2 and PN, using the times of the first and last ObsID 789 ObsID 866 events overall to determine the bounds used to define the temporal bins. We constructed a rate set for each event file, cal- R.A. offset ...... −2.67 ± 0.60 0.88 ± 0.61 culated the effective bin time defined as the sum of the good- Decl. offset...... 0.35 ± 0.56 0.35 ± 0.46 Rotation ...... −0.20 ± 0.30 −0.05 ± 0.25 time-intervals (GTIs) within a bin, applied the GTI correc- RMS before correction 3.22 1.40 tions by multiplying each rate by the ratio of nominal bin time RMS after correction . 0.59 0.88 to effective bin time, and finally summed these corrected rates N. — All quantities are given in arc seconds. The column for all temporally coincident bins. ObsID 789 lists the parameter values for Mos 2 and ObsID 866 for To correct for backgroundfluctuations even though they are Mos 1. No significant rotational offset was detected in either case and therefore none was applied in the analysis. generallyassumed to be negligible on a source extraction zone as small as 10′′, we extracted a background light curve from an annular region with inner and outer radii of 10 and 500 arcsec around SgrA*. These backgroundrates were also GTI- correlation between the positions of the calibration sources correctedand then rescaled to the area of the source extraction and their X-ray counterparts allowing for a displacement in zone before being subtracted from the source rates. (The max- R.A. and Decl. as well as rotation, we found the boresight imum background count rate we found for any given bin was corrections for the three Epic instruments. The Mos cameras of the order of 10% of the source count rate). have smaller pixels, and therefore a finer angular resolution This procedure yielded the background subtracted, GTI- than the PN instrument. We therefore used the Mos images corrected light curves. The GTI correction was particularly for our astrometric study of the obervations during which the important for the data of ObsID789 during which high lev- flares occured, namely ObsID789 and 866. The results of this els of background solar flare particles caused the saturation of study are listed in Table2 where we give the offsets and root buffersandthusthe loss of datain the PN camerabyswitching mean square (rms) dispersion in the positions of the astromet- it to counting mode for short time periods and so generating ric sources with respect to their reference positions before and several short GTIs. In such instances, the count rate must be after correction. estimated on longer time scales and thus larger binning. Although we performed this procedure for each individual For all light curves, we excluded the bins that correspond camera as well as for the merged Mos event list we used the to time periods that do not have simultaneous coverage by all Mos 2 results for our analysis of the first flare (ObsID789) three instruments, and those with very high count rates seen in and the Mos 1 for the analysis of the second (ObsID866) only one of the instruments; identified by looking at the ratio since these had the smallest dispersion after boresight cor- of the Mos to PN count rates. All light curves presentedin this rections. We found that the merged Mos image systemati- paper are combined Epic light curves (Mos 1 + Mos 2 + PN). cally yielded smaller offsets in both coordinates and that the rms dispersion remained in the range 0′′.9 to 1′′.2 before and 2.3. Spectral extraction after boresight corrections were applied. This behaviour is The source and background spectra for each flare were ex- expected since the astrometric precision derived from calibra- tracted over a circular region with a radius of 10′′ centered tions and that rely upon the knowledge of the position of one ′′ on SgrA*, and in each case, the backgrounds were inte- camerawith respectto theother is stated as 1 .2 (Kirsch 2005). grated over about 50 ks during “quiescent” periods — ex- So even though we can generally expect smaller offsets and cluding flares and eclipses — as done by Goldwurm et al. dispersions before corrections when using the merged Mos 2003 and Porquet et al. 2003. This ensures that the spectrum image as this naturally averages the astrometry of both Mos is composed of almost only flare photons such that the con- cameras, the minimum dispersion after corrections will al- ′′ tribution from the underlying nearby emission is negligible. ways be around 1 .2. Unusually large dispersions can be The time windows over which each flare spectrum was ex- caused by the “splitting” of one or more of the astrometric tracted were optimized for maximum signal to noise ratio and sources by a column of bad pixels as was seen in Mos 2 dur- are [197161000:197163000]for the 2004 March 31 event and ing ObsID866. Astrometric corrections maximize our ability [210335000:210337500] for the 2004 August 31 flare, given to locate sources and to distinguish a flaring source from its as XMM-Newton time in units of seconds (see Figures4 and closest known neighbour as is our intention here. 6). These intervals correspond to 2000 and 2500s respectively We found that the change in position due to corrections of ′′ and do not cover the entire event. up to 2.7 in a given coordinate has negligible effects on the Since the source photons represent between about 35 and light curve and spectra that are constructed by integrating over ′′ 50% of the total counts, and that these are in all cases quite a 10 radius around the source. This is primarily due to the low in number, we used the C-statistic to fit and derive the shape of the instrument point spread function which is char- model parameters. In the context of this statistic, the source acterized by a very narrow peak and broad wings. As long as spectra do not have to be rebinned. However, the method re- the peak is included in the extraction region then variations quires a moderately well defined backgroundwith at least 5–8 due to a shifted centroid are completely negligible. Each in- counts per bin and we therefore grouped these using grppha dividual event list is treated separately to ensure proper good- such that each bin contains a minimum of 10 counts. To en- time-interval corrections. sure coherence between the source and background spectra, we applied exactly the same grouping to the source spectra. 2.2. Light curve construction Regrouping in larger bins, with 20 counts per bin for exam- The source extraction region is defined as a circular ple, gives parameter values that are statiscally consistent with area of radius 10′′ centrered on SgrA*’s radio position: the ones derived from the ungrouped data set. (J2000.0) R.A. = 17h45m40s.0383, Decl. = −29◦00′28′′.069 The redistribution and ancillary response matrices were 4 generated for each source spectrum using the rmfgen and 3.1. Flare Analysis arfgen SAS tasks. The position of the flaring source was determined by run- 3. RESULTS ning edetect chain on an image of the central region constructed by selecting events in a temporal window correspond- Figures1 and 2 show the epoch1 and epoch2 light curves ing to the duration of the flare, and applying the astrometric respectively. Two large flares and a few smaller ones are ap- corrections to the fitted position of the most significant de- parent in the second portionof Figure1 and the first portionof tection. The resulting flaring source position for the 2004 Figure2. More detailed views of the flaring periods are shown March 31 event was found to be: R.A. = 17h45m40s.08, decl. in Figures4 and 6. = −29◦00′28′′.48, J2000 with an uncertainty in the position of During epoch1, the flaring activity occurs towards the end 1′′.12, calculated by combining the statistical uncertainty on the observation where a large event is preceeded by two the fitted position (0′′.44), the boresight correction (0′′.82), and smaller ones. In epoch2, the large flare and its precursor oc- the rms dispersion in the astrometric sources after correction cur near the start of the first observation. In the latter, the main (0′′.62). This position is at 0′′.66 from the radio position of event is closely followed by a small peak, and another statisti- SgrA* and 2′′.58 from the new transient CXOGC J174540.0- cally significant flare closer to the end of ObsID866 (marked 290031, the closest known X-ray source to the central black by an arrow in Figure2). Two small peaks at the very end hole. This permits us to unambiguously associate the flaring of the same observation have significances just above 3σ. In source with SgrA*. addition, there is a distinct periodic eclipse clearly seen over For the 2004 August 31 flare, the position of the flar- more than four complete cycles in the course of ObsID866 ing source was found to be: R.A. = 17h45m40s.14, decl. = and also present in ObsID867 even though it is not as evi- −29◦00′28′′.40, J2000 with a positional uncertainty of 1′′.31. dent. This is 0′′.52 from SgrA* and 2′′.98 from CXOGC J174540.0- In the next section, we present a detailed survey of the re- 290031, therefore the association of this flaring source with sults obtained for the most significant flares, followed by an the central black hole is once again unambiguous. The cen- analysis of the temporal characteristics of the light curves. troid of the 2–10keV emission in the non-flare period of Ob- We show that the two factor-40 flares are coming from the sID866 is located just 0′′.71 from CXOGC J174540.0-290031 direction of SgrA* and not from the transient binary system but 2′′.38 from SgrA*. This strongly suggests that this source CXOGC J174540.0-290031 to which the eclipses can be at- does indeed contribute a large portion of the X-ray flux in this tributed (see Porquet et al. 2005). band but that it is without a doubt not the flaring source. As an example, we show in the left hand side panel of Figure3, an image of the Galactic center during the August 31 flare composed of events selected from the same temporal window as the one used to build the flare spectrum (see Figure6). In black, the position of SgrA* is crossed and labeled, that of CXOGC J174540.0-290031 is simply marked by a cross for clarity, and the circles show the uncertainty on the position after boresight corrections. In green, the cross marks the fitted position of the brightest source detected in the flare image, and the circle indicate the 68 and 90% confidence regions derived by combining the statistical error on the fit (0′′.60), the error from the boresight correction (0′′.77), and the rms (0′′.88) that we take as a systematic uncertainty, On the right hand side, we see the non-flare image of the same re- F. 1.— Light curve in the 2–10 keV energy range for a circular region of radius 10′′ centered on Sgr A* binned in 500 second intervals for the epoch 1 gion constructed by selecting all event used to build the back- observations. MJD 53092 corresponds to 2004 March 28. ground spectrum. The labeling scheme is the same as in the left panel. Each pixel is 1′′.1 in size matching the sky projected size of the camera’s physical pixels. A wavelet filter was applied to smooth the image for presentation purposes only. It is evident that the centroid of the emission during the flare is significantly shifted towards Sgr A* and that the transient source CXOGC J174540.0-290031 can be excluded as the flaring source at the 90% confidence level. In the non- flare period, the emission is very closely centered on the transient binary — the central black hole’s closest known, hard X-ray emitting neighboor — and in this case, SgrA* can be excluded at more than the 90% confidence level as the source of this emission. The error circles on the fitted centroid of the non-flare image are smaller because the statistical uncertainty on the fit is 0′′.22 compared with 0′′.60 for the flaring source. Combining this value with the same boresight correction un- F. 2.— Light curve in the 2–10 keV range for a circular region of radius certainty and rms as was used in the analysis of the flare image 10′′ centered on Sgr A* binned in 500 second intervals for the epoch 2 ob- yields a total uncertainty of 1′′.19. servations. The arrow marks an isolated peak with statistical sigificance of about 5σ. MJD 53248 corresponds to 2004 August 31. The first flaring period occurred on 2004 March 31 (Fig- ure4) and contains one of the two factor-40 flares detected 5

Sgr A* Sgr A*

F. 3.— Smoothed Mos 2 image of the Agust 31 flare with an exposure time of 2000s (left panel) and average non-flare image with an exposure of about 50ks (right). The boresight corrected position of Sgr A* is labeled and that of CXOGC J174540.0-290031 is marked by a cross. The black circles indicate the 1σ uncertainty on the position due to the astrometric correction. The green cross and circles mark the fitted emission centroid and the associated uncertainty at the 68 and 90% confidence levels. over the course of all four pointings, preceded by two smaller ones. The most prominent flare took place around 23:05 and peaked at 0.76cts/s estimated on the basis of 500s time bins. This event lasted ∼ 2.5ks from rise to fall. The PN (black), Mos 1 (red) and Mos 2 (green) spectra for this flare are shown in figure5.

F. 5.— Epic spectra during the Sgr A* flare of 2004 March 31. The spectra are shown with the best fit absorbed power-law model and the corresponding residuals. The fit is done in the range 2–10 keV. The PN spectrum is shown in black and the Mos 1 and Mos 2 are in red and green respectively.

F. 4.— Zoom on the flare of 2004 March 31 with time bins of 500s where 22 −2 the first factor-40 flare occured. Its significance is about 10σ. The red lines with a column density of NH = (8.3 ± 2.5) × 10 cm . The delineate the time window used in the spectral extraction. Two other smaller average unabsorbed flux during the flare was found to be flares are marked by the green arrows and have significances of abotu 5 and (6.35 ± 0.45) × 10−12 ergcm−2 s−1. Since the count rate at the 4σ respectively. MJD 53092 corresponds to 2004 March 28. peak is about twice the average, we estimate that the maximum flux is also about twice the average and so the peak We fitted each spectrum individually and also simultane- 2–10keV luminosity is around 10 × 1034 erg s−1 at a distance ously. The low statistics limits the reliability of the Mos re- of 8kpc; a factor of 45 above quiescence 9. sults and for this reason we present the best fit parameters The black body was best fit with a temperature of 1.9 ± for the PN data as well as for the combined data set in Ta- 0.3 keV and absorbing column of (4.1 ± 1.7) × 1022 cm−2. ble4. Three models were tested: absorbed power-law, black Finally, fitting a bremsstrahlung model resulted in an absorb- body and bremsstrahlung. All give satisfactory fits. We used tion column of 7.8 ± 1.9 × 1022 cm−2 — similar to that of the pegpwrlw model in XSpec v11.3.1 in which the the total the power-law — and a temperature around 35keV. The er- unabsorbedflux over the range of the fit is used as the normal- ror range on the temperature, however, is huge and therefore ization. This allows the photon index and normalization to be not constraining at all. The large upper bounds in the error fit as independent parameters in the model and to derive the of the bremsstrahlung temperature probably reflect the fact uncertainty on the flux directly from this normalization. that there is no statistically significant break in the spectrum. For the pegged power-law model, the best fit values of the combined data set are a photon index of Γ= 1.5 ± 0.5 9 We refer to both events as factor-40 flares for simplicity. 6

The best fit parameter values, fluxes and luminosities for the power-law and black-body models are given in Table4. TABLE 3      The respective occurrence times, peaks and durations of the B X- ± / two small flares that preceded it are: 17:59, 0.51 0.04 cts s 2004 March 31 2004 August 31 (∼ 5σ) and 1500s for the first, and 19:48, 0.45 ± 0.04 cts/s UTC XMM time (s) UTC XMM time (s) (∼ 4σ) and 1000 s for the second. These two events are Start time ...... 22:48 1.9716050+08 09:58 2.1033350+08 marked by green arrows in Figure4 Unfortunately, the short Peak time ...... 23:09 1.9716175+08 11:05 2.1033750+08 duration and lower signal-to-noise ratio of these events pre- End time ...... 23:22 1.9716250+08 11:22 2.1033850+08 vents us from doing their spectral analysis. Duration (s) . . . . 2 500 5 000 The second major flare was on 2004 August 31 and is Peak rate (cts/s) 0.760 ± 0.048 0.617 ± 0.035 Mean rate (cts/s) 0.303 ± 0.002 0.293 ± 0.002 shown in detail in Figure6. The large flare is preceded very Deviation (σ)... 9.6 9.7 closely by what we will call a double-peaked precursor that N. — Error bars correspond to the 68% confidence interval. Count rates and lasted about the same time as the flare itself. Furthermore, we times are estimated using 500s time bins. see a ∼ 30% changein flux overabout900s or 15min between the two peaks of the precursor which constrains the emitting region to less than 2 AU at a distance of 8kpc. The first peak of the precursor occurred at around 9:00 and the flare that fol- This flare was modeled in the same way as was done for lowed rose to its maximum at about 11:05 with a count rate the March 31 flare and the detailed spectral fitting results of 0.62 ± 0.03 cts/s for 500 s bins. Both the precursor and are listed in Table4. The combined spectrum was best fit the flare lasted ∼ 5000s and so the whole flaring period had with an absorbed power-law of photon index Γ= 1.9 ± 0.5 and 22 −2 a duration of about 10000 s from rise to fall. The narrow column density NH = (12.5 ± 3.4) × 10 cm . For the black peak following the flare occurred at 13:01 and the last flare body model, the temperature was 1.8 keV and the absorbing 22 −2 occurred the next day, 2004 September 1, at 6:48. Figure7 column density (7.1 ± 2.3) × 10 cm . The bremsstrahlung shows the PN (black), Mos 1 (red), Mos 2 (green) spectra of column density was found to be 11.3 ± 2.5 with a tempera- the flaring event with the best fit absorbed power-law model ture around 15keV but once again the error range is so large and residuals. that this value cannot be constrained. The low statistics sev- erly limit our ability to detect spectral variations during the flares. The average 2–10keV luminosity of this flare is around 4.3 × 1034 erg s−1 (see Table4) and as in the case of the March 31 flare, we estimate the peak luminosity to be about twice the average, and thus reached a maximum intensity around 40 times the quiescent luminosity of SgrA*. A summary of the basic temporal features of the two factor- 40 flares are givenin Table3. This includes the start, peak and end times of the flares given in UTC and XMM-Newton time in seconds to facilitate reference to the data. We also listed the peak count rate and the deviation from the mean in units of sigma.

3.2. Timing Analysis F. 6.— Light curve of the 2004 August 31 flare binned in 500 s intervals. The cyclical decrease in flux observed during ObsID866 Green arrows point out local peaks with respective significances of about 4, 5, 10 and 4σ. We show the base level of the light curve before and after the (first portionof Figure2) occurswith a period of about 8h and flare for reference. The drop in flux at the base of the precursor corresponds is most probably due to an eclipse in the binary system CX- to the first of five eclipses in the binary system CXOGC J174540.0-290031 OGC J174540.0-290031. A complete analysis of the XMM- detected during ObsID 866. MJD 53248 corresponds to 2004 August 31. Newton obsevations of this transient source are presented by Porquet et al. (2005), and the results of the Chandra observations by Muno et al. (2005b). A spectral density analysis of the X-ray data from SgrA* was performed and since XMM-Newton’s spatial resolution severely limits our ability to distinguish the quiescent emission of SgrA* from that of its surroundings, temporal struc- tures in the X-rays emission detected by XMM-Newton from the Galactic center can only be attributed to SgrA* with confidence if seen during a flare, when the flux rises substantially above the quiescent emission. Therefore, periodic features seen in the base level of the light curve cannot readily be attributed to SgrA*. Furthermore, a period detected during a flare can only be confidently attributed to SgrA* if it is not present in the rest of the light curve. Since the range of frequencies and resolution in a peri- odogram are primarily functions of the total length of the observation and sampling or bin time, only the second of the two F. 7.— Epic spectra during the Sgr A* flare of 2004 August 31 fitted in factor-40 flares allowed us to perform a meaningful search for the range 2–10keV. The colours are the same as in Figure 5. periodic modulations. This event lasted about 10ks including 7 the precursor. This is about 4 times longer than the first. hole, producing either a two-component distribution charac- The first analysis revealed a periodic structure that could be terized by a broken power-law, with steep low-energy and flat significant but the proper treatment of the signal that includes high-energy components, or a modified thermal distribution white and possibly a component of red noise, requires a de- with a high-energy enhancement (see, e.g. Liu, Petrosian, & tailed analysis that is quite complex and that we have begun Melia 2004; Yuan, Quataert, & Narayan 2003 and 2004). In but will present elsewhere upon completion. either of these cases, the lower energy electrons produce the near-IR emission and the high-energy distribution produces 4. DISCUSSION AND CONCLUSION the hard X-rays. Further, the rather short time scale (∼ 2500– We observed the neighborhood of SgrA* for more than 4500s) and hard spectral index (Γ= 1.6) of the flares would, three consecutive days from 2004 March 28 to April 1, and according to this model, favor magnetic reconnection as the an additional three consecutive days from 2004 August 31 engine of the event. to September 3. Both observations were interrupted for only In either of these two scenarios — an accretion instability, ≈ 30ks and during each of these observation periods we de- or a magnetic reconnection event — one could expect to see tected flaring activity composed in each case of a large flare a modulation in the light curve, mirroring the underlying Ke- accompanied by smaller ones. The two main flares are posi- plerian period of the emitting plasma. Indications of the pres- tionally coincident with SgrA* and, as Baganoff et al. (2001), ence of such a periodic modulationduring a near-IR flare were we associate these with the supermassive black hole at the presented by Genzel et al. (2004)and during an X-ray flare by center of the Milky Way, given our current knowledge of Ascenbach et al. (2002). These results must be confirmed by the region surrounding this source. These flares, both with similar detections in other events seen in the near-IR and X- peak luminosities about 40 times that of the assumed qui- ray bands. A search for such a semi-periodic modulation in escent level, exhibit similar spectral characteristics: a hard the longer of the two factor-40 flares from SgrA* presented photon index of Γ ≈ 1.5–1.9 and a column density NH ≈ 8– here is under way and the results will be reported elsewhere. 13 × 1022 cm−2. These parameter values are more akin to The accretion instability scenario predicts a strong corre- those of the first two detected flares (Baganoff et al. 2001; lation between the sub-mm/IR and the X-ray photons with Goldwurm et al. 2003), than to those of the very bright flare no or little change in the millimeter flux density. The two- reported by Porquet et al. (2003). component synchrotron model predicts possible, though not There were a total of three flares with significance greater necessary, correlations between X-ray and near-IR flares, with than 3σ, corresponding to a factor of 15 above SgrA*’s qui- larger X-ray amplitudes in the case of simultaneous flaring, escent level in ObsID789 and the same number in ObsID866, important variations in the spectral slopes of the X-ray flares if we consider the precursor to be a separate flaring event. Al- compared with those in the near-IR flares, and small ampli- though we cannot determine the position of the smaller flares tude variability in the radio and sub-millimeter wavebands. accurately and thus cannot attribute them to SgrA* with cer- It is on this last point that we could distinguish the accre- tainty, if these are indeed coming from the central black hole tion induced X-ray flare model from the two-component syn- then, irrespective of the temporal distribution of these flares, chrotron model. our estimate of their average occurrence rate is about 1 per day. This estimate agrees with those based on Chandra observations performed between 1999 and 2002 (Baganoff 2003b). Note that flares from SgrA* appear to occur in clusters. G. Bélanger would like to thank Jean Ballet, Monique Ar- In light of the recently detected near-IR spectrum of SgrA* naud, Jean-Luc Sauvageau and Anne Decourchelle for their (Genzel et al. 2003), and of the constraints that it imposes on kind help with several aspects of the XMM-Newton analy- the various emission models for this source, the flares we de- sis, and Michael Muno, Régis Terrier and Matthieu Renaud tected could have resulted from a sudden increase in accre- for several useful discussions. G. B. is grateful to the ref- tion accompanied by a reduction in the anomalous viscosity, eree whose comments were very pertinent and lead to a finer which would point to thermal bremsstrahlung as the X-ray analysis and deeper understanding of the statistical methods emission process. This would give rise to a hard spectrum, un- used in this task. G. B. acknowledges the financial support like the softer one expected from a synchrotron self-Compton from the French Space Angency (CNES). This work is based process (Liu & Melia 2002a). on observations obtained with XMM-Newton, an ESA science Another possibility is that the flares arose from the quick mission with instruments and contributions directly fundedby acceleration of electrons in the accretion flow near the black ESA member states and the USA (NASA).

REFERENCES Baganoff, F. K. et al. 2001, Nature, 413, 45 Markoff, S. et al. 2001, A&A, 379, L13 Baganoff, F. K. et al. 2003, ApJ, 591, 891 Melia et al. 2001, ApJ, 554, L37 Baganoff, F. K. 2003, AAS/High Energy Astrophysics Division, 7 Melia, F. & Falcke, H. 2001, ARA&A, 39, 309 Bélanger, G. et al. 2005, ApJ, in press Muno, M. et al. 2003a, ApJ, 589, 225 Bower, G. et al. 2005, ApJ, in press (astro-ph/0507221) Muno, M. et al. 2005a, ApJ, in press (astro-ph/0412492) Eckart, A., et al. 2004, A&A, 427, 1 Porquet, D. et al. 2003, A&A, 407, L17 Genzel, R. et al. 2003, Nature, 425, 934 Porquet, D. et al. 2005, A&A, in press Ghez, A. M., et al. 2003, ApJ, 586, L127 Schödel, R. et al. 2003, ApJ, 596, 1015 Goldwurm, A. et al. 2003a, ApJ, 584, 751 Yuan, F., Quataert, E., & Narayan, R. 2003, ApJ, 598, 301 Hog, E., et al. 2000, A&A, 363, 385 Yuan, F., Quataert, E., & Narayan, R. 2004, ApJ, 606, 894 Liu, S. & Melia, F. 2001, ApJ, 561, L77 Yusef-Zadeh, F., Choate, D. & Cotton, W. 1999, ApJ, 518, L33 Liu, S. & Melia, F. 2002a, ApJ, 566, L77 Zhao, J.-H., Bower, G. C., & Goss, W. M. 2001, ApJ, 547, L29 Liu, S. & Melia, F. 2002b, ApJ, 573, L23 Liu, S., Petrosian, V. & Melia, F. 2003, ApJ, 611, L101 Kirsch, M. 2005, XMM-SOC-CAL-TN-0018 8 913 731 936 740 / / / / bins / 883 886 691 1086 ) C 1 4 3 8 7 5 7 9 4 − ...... 0 0 0 0 0 0 0 1 + + − − + + − − 6 9 6 6 . . . . erg s [2–10] 2 2 4 3 X 34 L ) (10 5 4 9 4 7 2 9 6 ...... 1 0 0 0 0 0 1 0 1 erg − ed over the range from 2 to 10 keV + + − − + + − − s 6 6 8 0 12 . . . . 2 [2–10] − 2 2 3 4 − X F alization corresponds to the dimensionless 5 9 8 5 6 6 8 2 ensity and the luminosity is calculated using ...... Black Body 1 0 0 0 1 1 0 1 ) cm + + − − + + − − 5 − 9 7 4 4 . . . . 5 5 7 7 4 3 3 2 2 5 3 3 ...... 0 0 0 0 0 0 0 0 + + − − + + − − 1 8 9 9 ’Instr.’ refers to the instrument whose data was used in the . . . . kT Norm (10 2 1 1 1 ) (keV) (10 2 3 3 7 1 7 3 5 9 ......  − 2 2 1 2 1 2 1 1 F + + − − + + − − H 5 1 1 4 cm . . . . N  22 X- 913 7 731 6 936 4 740 3 / / / / bins (10 TABLE 4 / 883 883 689 1082   e interval. Flux density for the Power-law model is normaliz C ) C 1 2 8 8 5 9 3 7 7 − ...... 1 0 0 0 0 0 0 0 + − + − + + − − 3 4 9 1 . . . . erg s  [2–10] 4 3 4 5 S e listed flux values are absorbed with the associated column d X 34 8kpc to the Galactic center. In the Black Body model, the norm L ) (10 2 4 2 4 3 4 4 9 ...... 1 0 0 0 0 0 0 0 0 erg − + − + − + + − − s 7 7 0 0 12 . . . . 2 − [2–10] 2 2 4 4 − to the distance in units of 10kpc squared. In the first column, Power-law X 1 F cm (10 − 5 5 5 5 5 6 4 6 ...... 0 0 0 0 0 0 0 0 erg s + − + − + + − − 9 3 5 7 . . . . Γ 39 1 1 1 1 ) 4 8 4 5 . . . . 2 5 3 1 8 . . . . 3 2 3 2 − 2 3 2 2 + − + − + + − − 5 0 H 3 0 . . cm . . N 22 . — Errors on fitted parameters correspond to the 68% confidenc  Date (Instr.) (10 N March 31 (All) 8 March 31 (PN) 8 August 31 (All) 12 August 31 (PN) 10 and corresponds to the total unabsorbed flux in this range. Th the power-law normalization and is derived for a distance of fit and ’All’ refers to the combination of PN, Mos 1 and Mos 2. ratio of the luminosity in units of 10