EVIDENCE OF ELEVATED X-RAY ABSORPTION BEFORE AND DURING MAJOR FLARE EJECTIONS IN GRS 1915+105

Astrophysical black holes reveal themselves to astronomers through their interaction with adjacent gas. There are two observed aspects to this phenomenon: the thermal glow from the viscously heated accreting gas and the bipolar ejection of plasma that can propagate very close to the speed of light. Both behaviors are seen in solar mass black holes (microquasars) in the Milky Way as well as supermassive black holes in active galactic nuclei (AGNs). The relationship between the accretion and jet formation has been intensely studied (e.g., Fender et al. 2004), but is still not well understood. Microquasars have a logistical advantage for astronomers since significant evolution of the system can happen on short timescales (ms to days). Thus, direct associations between accretion and outflow diagnostics can be discerned, in principle, within a human lifetime (Rodriguez et al. 2008; Miller-Jones et al. 2012). In contrast, AGNs vary on timescales of ∼1–10⁶ yr. Thus, the chance of gathering enough information to make statistically significant correlations between accretion and outflow observable quantities in a human lifetime is drastically diminished. Hence, there is an appeal to using microquasars as time-compressed laboratories to study black hole and quasar physics. Substantial progress has been made in this regard for GRS 1915+105, which is well known for ejecting radio "blobs" out to large distances with apparent superluminal speeds (Mirabel & Rodriguez 1994; Fender et al. 1999; Dhawan et al. 2000). Punsly & Rodriguez (2013a, hereafter PR13) analyzed large RXTE data sets and demonstrated that empirical relationships exist between the accretion states and major flare ejections (MFEs). In particular, the X-ray luminosity is highly elevated in the last hours preceding major ejections and it is correlated with the power required to eject the discrete blobs of plasma at relativistic velocity. Second, the X-ray luminosity was found to be highly variable during MFEs, yet the time-averaged X-ray luminosity during the ejection event is correlated with that just before the MFE and is of a similar (but perhaps a slightly lower) magnitude. However, the ejections are brief and occur at unpredictable times. Thus, all the relevant X-ray data are serendipitously gathered during wide-field monitoring. Therefore, we have no pointed X-ray observations near or at the crucial instant of ejection and therefore no spectral data to constrain the accretion state. Ironically, in this circumstance, the rapid evolution of the microquasars impedes our understanding of the black hole accretion system. Fortunately, the MAXI observatory typically observes GRS 1915+105 every 1.5 hr and it has the ability to gather spectral data from 1–20 keV (Matsuoka et al. 2009). For very bright states (as occurs during major ejections), single-epoch observations can provide a useful spectrum. For lower luminosity states, with slower evolution, one can bin multiple observations over a few hours to achieve sufficient signal to noise for a useful spectrum.

Being thusly motivated to understand the X-ray spectral state before and during major ejections, we have used the current monitoring program with the RATAN-600 radio telescope to find evidence of major ejections (i.e., large radio fluxes and a steep radio spectrum). Using a (limited) library of high temporal resolution X-ray light curves of major ejections from PR13, we estimate the time frame for flare ejection from the MAXI light curve. We downloaded the MAXI spectral data temporally adjacent to this epoch and looked for trends in the spectral fits and the raw number counts. Our findings, based on the two largest radio flares in the first eight months of 2013, are presented here.

We were granted radio monitoring time on the RATAN-600 radio telescope at 4.8, 8.2, and 11.2 GHz for both the first and second half of 2013. These observations have been carried out in a current monitoring program of the microquasars with RATAN-600 (e.g., Trushkin et al. 2008). The cryogenically cooled receivers and antenna were calibrated daily with secondary calibrators 3C286 and PKS1345+12. High-luminosity, steep-spectrum radio flares originating from GRS 1915+105 have historically been associated with superluminal ejections (Rushton et al. 2010). Major flares begin as an optically thick flux increase that becomes optically thin at lower and lower frequency as they evolve over the next few hours. The temporal spacing of our monitoring program—one measurement per day—is far too large to estimate the time of the major ejection that is associated with the radio flare. From PR13 and Figure 1, ejection initiation and end times require temporal resolution 100 times finer, on the order of 0.25 hr or less. However, strong flares are luminous for at least one day, so it is adequate for identifying major ejections (Rodriguez & Mirabel 1999; Miller-Jones et al. 2005). We define major flares as radio flares with a flux density S_ν > 100 mJy at frequency 4.8 GHz, with a steep spectral index, α > 0.5 (S_ν ∼ ν^−α). We have identified five such major flares in the first eight months of 2013.

Zoom In Zoom Out Reset image size

The correlation of X-ray luminosity with flare energy and power shown in PR13 indicates that we should look at the strongest flares in order to get adequate statistics in our MAXI data. Of these five steep-spectrum flares, two had S_ν > 200 mJy at 4.8 GHz. The first (hereafter Flare 1) was detected on MJD 56314.349 and the second (Flare 2) on MJD 56352.245. The observations adjacent to these flares are summarized in Table 1. In the following, we consider the MAXI spectral data for these flares. The other steep-spectrum radio flares that were not suitable for spectral study with MAXI are discussed in the Appendix.

We downloaded the data from the MAXI pipeline around the times of the strong major radio flares listed in Table 1. We aim to accomplish two goals with these data. The first goal is to estimate the dates that the major ejections occurred. Second, we implement time-resolved MAXI spectroscopy to determine the evolution of the X-ray spectrum preceding and during major ejections.

3.1. Estimating the Date of the Ejection Episode of Flare 2

3.2. Time-resolved Spectroscopy of Flare 2

Based on our estimates of the ejection episodes above, we use the MAXI spectral data to look for spectral evolution before and during ejection (i.e., we are looking for evidence of the physical mechanism responsible for ejection). Since the sensitivity of MAXI is low (low effective area), binning of the data over several times the temporal resolution is often required in order to constrain our spectral fits. This greatly reduces the time resolution except at strong peaks. The results obtained from the spectral fits with a model consisting of an absorbed power law (tbabs*powerlaw in XSPEC terminology) with abundances obtained from Wilms et al. (2000) to the MAXI data from 1–20 keV are listed in Table 2. The first column represents the range of dates in each bin. The second column is the fitted column density of hydrogen, N_H, responsible for the effects of absorption in the spectrum. The third column is the photon index of the power-law fit to the data, Γ. The next two columns are the absorbed flux and the unabsorbed (intrinsic) flux that deconvolves the effects of N_H, respectively. The last column is the reduced χ² figure of merit of the fit. The errors on the fitted parameters are at 90% confidence and the errors on the flux are at 68% confidence (1σ). There are some evident trends in Table 2 for the 56352.245 flare. Obviously, the luminosity increases before the ejection. However, we also see Γ steepen and this steep spectrum seems to continue during the luminosity variations that occur during the ejection. The uncertainty in the photon index in Table 2 makes this phenomenon less than statistically significant. However, we also see an increase in N_H with more statistical significance and we explore this below in Figure 2. The data in Figure 2 (from Table 2) are suggestive of a steady increase in N_H during the 15 hr preceding the ejection. However, the errors are large and do not warrant such a strong statement without a statistical analysis. To this end, note that the solid red line in Figure 2 is the fit to the data from 56350.85 up to 56351.54 using the method of least squares with uncertainty in both variables (Reed 1989). Equations (1) and (2) are the fits from 56350.85 to 56351.54 and 56351.66, respectively:

$$\begin{eqnarray} N_{{\rm H}} &=& (6.48 \pm 0.92)\; (\mathrm{DATE} - 56350.65)\nonumber\\ &&+ (5.27 \pm 0.57){\rm DATE} \le 56351.54, \end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray} N_{{\rm H}} &=& (7.49 \pm 1.67)\; (\mathrm{DATE} - 56350.65)\nonumber\\ && + (4.83 \pm 1.20), {\rm DATE} \le 56351.66. \end{eqnarray} \tag{ 2 }$$

The dashed lines in Figure 2 indicate the maximum and minimum slopes consistent with the data at the 1σ level based on Equation (1). The slope and the corresponding uncertainty in Equation (1) indicate an increase in N_H at the ∼6.5σ level preceding the flare.

Zoom In Zoom Out Reset image size

Table 2. Parametric Fits to MAXI Data Near Major Flare Ejections

Date	NH	Γ	Observed Flux	Intrinsic Flux	Reduced
			1–10 keV	1–10 keV	χ2(dof)
(MJD)	(1022 cm−2)		(10−8 erg s−1 cm2)	(10−8 erg s−1 cm2)
56312.32 ± 0.16	$8.2^{+2.0}_{-1.7}$	2.7 ± 0.2	1.56 ± 0.06	5.05 ± 0.19	1.40(79)
56312.64 ± 0.10	$10.5^{+3.6}_{-3.0}$	$3.1^{+0.4}_{-0.3}$	1.68 ± 0.15	9.50 ± 0.85	1.20(42)
56312.84 ± 0.04	$10.9^{+2.5}_{-2.1}$	3.0 ± 0.3	3.10 ± 0.20	15.10 ± 0.80	0.90(66)
56313.06 ± 0.01	$16.7^{+4.4}_{-3.5}$	$3.2 ^{+0.4}_{-0.3}$	5.20 ± 0.40	42.80 ± 5.00	1.35(41)
56313.16 ± 0.04	$9.7^{+4.1}_{-3.1}$	$2.7^{+0.4}_{-0.3}$	3.20 ± 0.30	12.40 ± 1.60	0.90(27)
56350.85 ± 0.20	$6.4^{+1.7}_{-1.4}$	2.4 ± 0.2	1.31 ± 0.07	3.36 ± 0.18	0.88(75)
56351.20 ± 0.14	$9.3^{+2.0}_{-1.7}$	2.8 ± 0.2	1.63 ± 0.09	6.28 ± 0.37	1.27(68)
56351.46 ± 0.07	$10.1^{+2.3}_{-2.0}$	2.8 ± 0.2	3.00 ± 0.20	12.20 ± 0.82	0.90(62)
56351.54 ± 0.01	$11.2^{+3.2}_{-2.7}$	2.9 ± 0.3	3.90 ± 0.30	18.00 ± 1.39	0.80(43)
56351.59 ± 0.01	$9.7^{+4.7}_{-3.4}$	$3.2^{+0.6}_{-0.5}$	1.4 ± 0.30	8.70 ± 2.60	0.90(19)
56351.66 ± 0.01	$14.2^{+3.0}_{-2.5}$	3.2 ± 0.3	4.70 ± 0.40	33.40 ± 2.84	1.00(55)
56351.83 ± 0.13	$11.0^{+1.6}_{-1.4}$	$2.8^{+0.2}_{-0.1}$	2.2 ± 0.10	14.30 ± 0.65	0.99(117)

Download table as:  ASCIITypeset image

Even thought this result is statistically robust, we investigate whether it is a manifestation of the degeneracy between Γ and N_H in the parametric spectral fits. Within a power-law model, a modest low-energy flux is only consistent with a large Γ if N_H is large as well. This is a consequence of the power-law assumption. To this end, we use a non-parametric diagnostic, the ratio of counts per second (cps) in a MAXI bin from 3.0–6.0 keV (the spectral peak) to cps in the lowest MAXI energy bin 2.0–2.7 keV. Of course, the lowest energy bin is the most susceptible to the effects of absorption and the cps will be low and our statistics will suffer accordingly. Thus, considerable binning in the time domain is required to improve the statistics. The advantage of this non-parametric method of analysis is that it does not depend on Γ and the power-law assumption. This removes the parametric degeneracy and replaces it with a ratio that is purely an empirical number. The blue data points indicate that the 2.0–2.7 keV cps are increasingly more suppressed relative to the 3.0–6.0 keV cps as the ejection is approached. This non-parametric analysis supports the power law-based deduction that N_H increases before the major ejection.

3.3. MAXI Observations of the Major Flare 1

Even though Flare 1 is larger than Flare 2, the MAXI data suffer from some inopportune gaps in temporal coverage and are inferior for our purposes. Figure 3 plots N_H from the spectral fits in Table 2, the MAXI 2–20 keV light curve, and the SWIFT 10–50 keV light curve. The estimated range of possible ejection times, 56313.86–56314.06, are based on the six criteria listed above in Section 3. There was a large gap in MAXI coverage that envelopes most of the ejection episode. The statistics are poor because most of the luminous bins are lost due to the MAXI gap near ejection (when the fluxes were the highest). Furthermore, the X-ray cps decreases very rapidly after achieving a peak near 56314.0. Thus, large bins are required to obtain adequate statistics. The combination of wide ranges of dates in the bins and what appears to be a steep increase in N_H after 56313.8 results in an insufficient number of bins to resolve the increase. Thus, we cannot claim a large statistically significant increase in N_H preceding launch. The increase in N_H in Figure 3 before the MFE episode is consistent with the increase seen prior to the MFE that is responsible for Flare 2.

Zoom In Zoom Out Reset image size

In this article, we examined time-resolved MAXI X-ray spectroscopy of two major flares of GRS 1915+105 found with RATAN-600 radio monitoring during the first 8 months of 2013. The time-resolved spectroscopy indicates that N_H increases as the X-ray luminosity increases before a major ejection and appears to remain elevated until the end of the ejection episode. This behavior was seen convincingly for Flare 2 and appears to be consistent with poorer data from Flare 1.

For the flares considered here, we determined that N_H increased by ≈5–8 × 10²² cm⁻² over a time frame that begins a few hours before the initiation of the ejection and appears to extend well into the ejection episode. The increase of N_H can be cast in light of the increase in the intrinsic absorption due to the column density located within the GRS 1915+105 binary system, N_H(intrinsic). There is a significant component of the absorbing column due to the line of sight across the Galaxy, an extrinsic component, N_H(extrinsic). It was estimated in Chapuis & Corbel (2004) that N_H(extrinsic) ≈ 3.5 × 10²² cm⁻². The discussion in Belloni et al. (1997) indicates a slightly higher stable interstellar absorption column, N_H(extrinsic) ≈ 4.5 × 10²² cm⁻². From Table 2, this means that N_H(intrinsic) ≈ 0.0–1.5 × 10²² cm⁻² 12–15 hr before the ejection episode and increases by a factor >10 by the time of the ejection.

The increase in N_H has two obvious plausible physical interpretations.

• 1.  
  The increase represents accreting gas at relatively high latitudes, since the line of sight is estimated at 65°–70° to the accretion disk normal from superluminal ejection kinematics (Mirabel & Rodriguez 1994; Fender et al. 1999). The additional high-latitude accretion component is then a part of the triggering (and sustaining) mechanism of the major ejections.
• 2.  
  The enhanced absorbing column represents an outflowing wind from the accretion disk that begins before a major ejection and continues during the ejection.

There is not enough data to conjecture fully on these alternatives from a physical perspective. However, the first alternative is consistent with three-dimensional numerical simulations of accretion onto rotating black holes. A small class of simulations was found to be consistent with the correlation of the power of the ejection and the elevated X-ray luminosity: namely, simulations that produced an ergospheric disk jet about a black hole with spin parameter a > 0.984. In the simulations, the jet power and the X-ray luminosity can be strong only during episodes of elevated accretion (Punsly & Rodriguez 2013b). In order to understand if the behavior of N_H seen here is endemic to large flares as opposed to anecdotal, we plan on continuing the RATAN monitoring to find more major flares. In particular, because of the correlation of radio luminosity before and during ejections and the X-ray flux found in PR13, stronger radio flares will have larger corresponding X-ray fluxes around the ejection episode. Thus, an extremely powerful flare (i.e., larger than 500 mJy at 4.8 GHz) will have a large MAXI cps and smaller bins can be used, thereby improving the time resolution and the statistics.

This research has made use of MAXI data provided by RIKEN, JAXA, and the MAXI team. B.P. thanks Tatehiro Mihara of the MAXI team for helping with the data acquisition and analysis. J.R. acknowledges funding support from the French Research National Agency: CHAOS project ANR-12-BS05-0009 (http://www.chaos-project.fr) and financial support from the UnivEarthS Labex program of Sorbonne Paris Cité (ANR-10-LABX-0023 and ANR-11-IDEX-0005-02). S.A.T. is thankful to the Russian Foundation for Base Research for support (grant N12-02-00812). We were also very fortunate to have a referee who offered many useful comments.

In this Appendix, we describe the other steep-spectrum radio flares from the first 8 months of 2013 with >100 mJy flux density at 4.8 GHz, S_ν(4.8 GHz) > 100 mJy. None of these flares have suitable X-ray data for time-resolved spectroscopy.

• 1.  
  56439.87 is the date of the RATAN-600 detection of the first of a series of flares that culminated in the large flare on 56352.25 (Flare 2) that was described at length in the text. RATAN-600 measured S_ν(4.8 GHz) ≈ 140 mJy and the spectrum was steep. Unfortunately, the MAXI scan stopped on GRS 1915+105 and did not complete when MAXI turned off its high voltage in order to protect X-ray detectors in a high-radiation region. The process removed the data since the scan did not complete (T. Mihara 2013, private communication).
• 2.  
  56445.25 is the date of the RATAN-600 detection of S_ν(4.8 GHz) ≈ 120 mJy and the spectrum was steep. Unfortunately, MAXI did not cover GRS 1915 during the flare (T. Mihara 2013, private communication).
• 3.  
  On 564543.97, RATAN-600 measured S_ν(4.8 GHz) = 175 ± 15 mJy and the spectrum was steep. The 2–20 keV MAXI light curve that was sampled every 0.06 days was in the range of 1.7–2.8 cps near the time of flare launch (the local maximum). This is less than half of what we found near the ejection period for Flare 2 in Figure 1 and about 1/3 of the count rate for Flare 1 in Figure 3 with the same bin size in the time domain. Thus, the count rate is too low for time-resolved spectroscopy as the uncertainty in the fits would be enormous due to the low-number statistics. This could be compensated for by approximately doubling the bin size in the time domain, but this would not provide meaningful time resolution. This modest flare highlights the importance of isolating the strongest radio flares for MAXI time-resolved spectroscopy.
• 4.  
  The flare on 56487.87 detected by RATAN-600 was not formally steep: S_ν(4.8 GHz) = 163 ± 16 mJy, S_ν(8.2 GHz) = 160 ± 16 mJy and S_ν(11.2 GHz) = 107 ± 11 mJy. It is a steep spectrum from 8.2 GHz to 11.2 GHz, but a flat spectrum between 4.8 GHz and 8.2 GHz. It seems likely that S_ν(4.8 GHz) would be larger than 100 mJy and a steep spectrum a few hours later if the radi- emitting plasma was in a state of rapid expansion. Unfortunately, like the 56439.87 radio flare, the MAXI scan stopped on GRS 1915+105 and did not complete. The process removed the data since the scan did not complete (T. Mihara 2013, private communication).