THE MURMUR OF THE HIDDEN MONSTER: CHANDRA'S DECADAL VIEW OF THE SUPERMASSIVE BLACK HOLE IN M31

It is known that several bright X-ray sources are present in the nuclear region of M31, challenging a clean isolation of the emission from M31* (Garcia et al. 2000; Li et al. 2009). We first examine the HRC observations, since they have the better angular resolution (PSF FWHM ≈ 04). Figures 1(a) and (b) show an image of the nuclear region, stacking all HRC observations on the original pixel scale of 01318. Four X-ray sources are immediately identified within 4'' of the nucleus. Following Garcia et al. (2005) and Li et al. (2009), we label these sources in Figure 1(a) as P2 for the source positionally coincident with the nucleus, N1 for the source just ∼05 northeast of P2, SSS for the source ∼15 south of N1, and S1 for the source ∼2'' farther south. The positional coincidence between P2 and the nucleus is demonstrated in Figure 1(b) in which Hubble Space Telescope (HST)/Advanced Camera for Surveys (ACS) intensity contours of the optical double nuclei of M31 (Lauer et al. 1993) are shown. We have registered the HRC and HST/ACS images by matching common extranuclear sources, resulting in an uncertainty of ∼02 (Garcia et al. 2010). The other three sources are most likely stellar objects (Kong et al. 2002; Di Stefano et al. 2004; Li et al. 2009).

Zoom In Zoom Out Reset image size

Visual inspection of individual HRC observations reveals that all four sources show some degree of flux variation, with P2 varying most significantly. This can be illustrated by dividing the HRC data into two subsets, the first half consisting of observations taken before 2006 (Figure 1(c); a total exposure of 256 ks) and the second half consisting of observations taken since 2006 (Figure 1(d); a total exposure of 315 ks). Using the three extranuclear sources as a reference, it is obvious that P2 has become significantly brighter since 2006. We note that two more sources can be identified in Figure 1. The fainter one only appears before 2006, located immediately east of SSS; the brighter one only appears since 2006, located at the upper right (northwest) corner of the images. Neither of these two transient sources affects the following analysis and hence will not be discussed further.

We now turn to the ACIS observations. An image stacking all 58 ACIS observations is shown in Figure 2(a), on a pixel scale of 0123 (i.e., one-fourth of the original ACIS pixel size) to take advantage of the sub-pixel positioning information that results from telescope dithering and to approximately match the HRC pixel scale. With a PSF FWHM of ∼06, the four sources are again clearly identified. Visual inspection of individual ACIS observations consistently reveals the flux variability in P2. Most dramatically, in an observation taken on 2006 January 6 (ObsID 7136, with an exposure of 5 ks), P2 appears to be by far the brightest source in the field (Figure 2(b)), indicating that an outburst has been fortuitously captured. We similarly present two subsets of the ACIS data, i.e., observations taken before 2006 (Figure 2(c); a total exposure of 149 ks) and since 2006 (but excluding ObsID 7136; Figure 2(c); a total exposure of 152 ks). Again, the two subsets show a marked difference in the brightness contrast between P2 and the three extranuclear sources.

Zoom In Zoom Out Reset image size

To obtain a more quantitative view of the flux variability, we construct a light curve for P2. Due to the proximity of the four sources, in particular between N1 and P2, it is crucial to account for the mutual PSF scattering. We adopt a two-dimensional fitting procedure, similar to that employed by Li et al. (2009), to simultaneously determine the fluxes for all four sources in each observation. Specifically, we take the following steps.

• 1.  
  From the circumnuclear sources used for calibrating the astrometry (Section 3), we further select relatively isolated ones (∼10), i.e., those without neighboring sources within 4''. For each observation, we obtain a local PSF image by stacking the selected sources after centroiding. The average PSF for the summed image (be it over a given detector, a given energy range, or a given epoch) is then obtained by stacking individual PSF images of the constituent observations.
• 2.  
  With the CIAO tool Sherpa, we fit the summed image, using a model consisting of four point-like sources and a constant local background. Each source is represented by a delta function, whose center and normalization are free parameters. This model is convolved with the average PSF. Only the central 4'' from the nucleus is considered in the fit (outlined by the circle in Figure 1(a)). Given the limited counts in each unbinned pixel, the best fit is obtained by minimizing the C-statistic (Cash 1979). The fit is robust in determining the source centroids, the chief goal of this step.
• 3.  
  For each observation, we repeat the above fitting procedure, fixing the centroids of all four sources as determined from the summed image. The fit then allows us to determine the flux and its uncertainty, corrected for the local effective exposure, for each source in each observation. Such a procedure not only maximizes the counting statistics, but also properly accounts for the propagation of the dominating error term arising from the mutual PSF scattering.

The resultant decadal light curve of P2 (Figure 3) shows two prominent features. First, an outburst, preceded by a 6 year epoch of quiescence, appears on 2006 January 6 during which the flux of P2 reaches a value of (76.5 ± 5.9) × 10⁻³ counts s⁻¹. Second, after the outburst, P2 apparently enters a more active state, characterized by frequent flux variations. Summing the observations taken before 2006 and repeating the above fitting procedure, we obtain an average flux of (0.3 ± 0.1) × 10⁻³ counts s⁻¹. Similarly, for the observations since 2006 (but excluding ObsID 7136), we find an average flux of (6.9 ± 0.3) × 10⁻³ counts s⁻¹. Therefore, the observed flux of the outburst is ∼250 times the quiescent level before it, whereas the long-term average flux increases by a factor of ∼23 after the outburst. For comparison, the three extranuclear sources⁶ show a ratio of fluxes since and before 2006 of ∼1.1 (N1), ∼0.7 (SSS), ∼1.5 (S1), respectively, all indicating a nearly constant long-term flux; their maximum-to-mean flux ratios are ∼3–5, also suggesting that the flux variation seen in P2 is exceptional.

Zoom In Zoom Out Reset image size

It is impractical to perform a spectral analysis for P2, due to the limited number of counts in most observations and the contamination of scattered photons from the other sources whose spectral information is also uncertain. Instead, by dividing the ACIS counts into a soft band and a hard band, we estimate the hardness ratio, defined as HR = (I_{0.5–2 keV} − I_2–8 keV)/(I_{0.5–2 keV} + I_2–8 keV), to be −0.11 ± 1.30, −0.76 ± 0.04, and −0.50 ± 0.04 before, during, and after the outburst, respectively. While in the first epoch HR is essentially indeterminate, it appears that the emission during the outburst is softer than that in the last epoch. For comparison, a Galactic foreground-absorbed (N_H = 7 × 10²⁰ cm^-2) power-law spectrum with a photon index of 1.8 (2.6) would result in HR = −0.50(−0.76). We thus convert the observed count rates into 0.5–8 keV intrinsic luminosities of 0.2, 43, and 4.8×10³⁶ erg s^-1 for the three epochs, by assuming a power-law spectrum with a photon index of 1.8, 2.6, and 1.8, respectively.

We find no statistically significant intra-observation flux variation for the outburst. We note that a typical 5 ks long observation may probe flux variations arising from a region of ∼5 × 10⁻⁵ pc in diameter, roughly four times the Schwarzchild radius of the SMBH. Unfortunately, ObsID 7136 is the most isolated observation: no other observation was taken before or after it within several months. Therefore, we cannot rule out that P2 stayed at high flux levels for up to months between 2005 and 2006. The fact that we have found only one outburst in 98 observations indicates a ∼1% duty cycle for outbursts of this kind.

As in previous studies (Garcia et al. 2005, 2010; Li et al. 2009), our identification of P2 as the X-ray counterpart of M31* rests on positional coincidence. The uncertainty of the registration among the X-ray, optical, and radio images is ∼01–02, corresponding to a linear size of ≲1 pc, and the chance coincidence of an interloping X-ray source is only ≲1% (Garcia et al. 2010). On the other hand, within 1 pc of the SMBH, members of the eccentric stellar disk (Lauer et al. 1993; Tremaine 1995) almost certainly produce some X-ray emission, which is however difficult to reliably quantify due to the underlying extremely high stellar density. Indeed, the absolute luminosities of the outburst and the subsequent individual detections are only modest (≲a few 10³⁷ erg s^-1), and by face value, can be attributed to an X-ray binary, e.g., containing an accreting stellar-mass black hole. Most known Galactic black hole binaries show outbursts at a substantial fraction of their L_Edd, and in rare cases, some of them also show complex flux variability (e.g., McClintock & Remillard 2006). Hence, before addressing the nature of the outburst in the context of M31*, we note that a stellar origin of P2 cannot be completely ruled out.

The detected outburst adds M31* to a short list of non-active SMBHs that have shown strong X-ray flux variations. Early ROSAT observations revealed a handful of X-ray outbursts associated with the nuclei of optically non-active galaxies (cf. Komossa 2002), which are interpreted in terms of tidal disruption of a star passing by the SMBH (Rees 1988). The absolute luminosities of these outbursts are ∼$10^{42}\hbox{--}10^{44}{\rm \,\rm erg\, s^{-1}}$, a substantial fraction of L_Edd of the associated SMBH. Moreover, these outbursts are transient rather than recurrent. These are to be contrasted with the M31* outburst, whose luminosity is ∼10^−8.5 L_Edd and which is followed by recurrent flux variations, albeit with smaller amplitudes. Hence, the M31* outburst is unlikely to be triggered by tidal disruption event.

The outburst is more reminiscent of the strongest X-ray flares seen in Sgr A* (Porquet et al. 2003, 2008). The M31* outburst and Sgr A* flares are similar in strength (≳100 times relative to the quiescent level), spectral softness, and low incidence rate. Hence, it is reasonable to speculate that the outburst shares the same physical origin with the Sgr A* flares. In this regard, the individual detections since the outburst are likely flares of smaller amplitudes, again similar to the smaller, but more frequent, flares seen in Sgr A*. However, it appears that the outburst either has triggered or signaled the smaller flares, and the post-outburst mean flux level became more than 10 times the mean flux before the outburst. No such trend has been reported in the long-term X-ray emission from Sgr A*.

At present, the physical origin of the Sgr A* flares remains elusive and most proposed models are phenomenological (e.g., Yuan et al. 2004; Liu et al. 2006; Maitra et al. 2009). In a magnetohydrodynamical model analogous to that developed for the phenomena of coronal mass ejection in our Sun (Lin & Forbes 2000), Yuan et al. (2009a) interpreted the flares as episodic ejection of relativistic plasma blobs ("episodic jets") inflated by magnetic field reconnection in the inner region of the accretion flow. Relativistic particles are accelerated in the reconnection current sheet and are responsible for the X-ray flares. After Sgr A*, M31* is only the second SMBH found to show flaring emission. A comparative study of the two SMBHs should yield important constraints to the flare modeling. Interestingly, M31* and Sgr A* are also the two least active SMBHs known. This refreshes the issue of whether flares are fundamentally related to the lowest accretion rates of LLAGNs, as discussed by Yuan et al. (2004).

Regardless of the exact physical process that produces the active X-ray emission from M31*, it is expected that the average radio emission from M31* has also substantially increased since 2006, by analogy to the associated X-ray/radio flares in Sgr A* (e.g., Marrone et al. 2008). This is also suggested by the so-called fundamental plane of black hole activity (Merloni et al. 2003; see also Yuan et al. 2009b), which empirically relates the X-ray luminosity, radio luminosity, and mass of an accreting black hole. We investigated archival VLA observations taken since 2006, but found no useful constraints on the radio variability of M31*, due to poor data quality. New coordinated EVLA observations with much enhanced sensitivities will shed light on the origin of the X-ray flares in M31*. Simultaneous X-ray/radio detection of variability in P2 will help rule out an X-ray binary, whose radio flux is expected to be well below the sensitivity limit of EVLA.