THE IDENTIFICATION OF MAXI J1659–152 AS A BLACK HOLE CANDIDATE

Black hole X-ray binaries have been studied since the early 1970s. It is now well established that the X-ray spectral and variability properties of these sources are strongly correlated, which is most clearly seen in the transient blackhole X-ray binaries (BHTs). While there is a great variety in the observed outburst behavior of BHTs (even for single sources), their outbursts typically proceed along "q"-shaped tracks in hardness–intensity diagrams (HIDs; see, e.g., Homan & Belloni 2005; Dunn et al. 2010), which are traced out in a counterclockwise manner; we note that similar tracks are also traced out by neutron star transients (e.g., Tudose et al. 2009; Linares 2009). The various branches of these tracks correspond to distinct spectral states. Not all sources show all possible states and the time they spend in each state can differ significantly.

There are various conventions for describing the spectral (and variability) states of BHTs; see, e.g., reviews by Homan & Belloni (2005) and Remillard & McClintock (2006). In this Letter, we follow the convention used in Belloni (2010), which is based on the work by Belloni et al. (2005) on GX 339-4. This source is often used as a template for the outburst evolution of BHTs, owing to its well-defined q-shaped HID tracks, and the fact that it shows behavior that is common to many other systems.

When GX 339-4 goes into outburst, its intensity is low and spectrum is hard (low–hard state or LHS). As the intensity increases, the spectrum remains hard, until the source makes a transition to the intermediate state (IMS), where the hard color starts to decrease at a rather constant intensity. The IMS can be divided into a hard and a soft IMS (HIMS and SIMS, respectively) depending on the spectral and variability characteristics observed. The transition from the LHS is always first to the HIMS. GX 339-4 often shows several transitions between HIMS and SIMS until the hardness decreases even further and the source reaches the so-called high-soft state (HSS), where subsequently the hardness remains approximately constant as the intensity eventually decreases. At some point during this decrease hardness increases again and the source transits from the HSS via the IMS back to the LHS, and returns to quiescence.

A BHT exhibits various types of quasi-periodic oscillations (QPOs) and broadband noise components, whose properties are strongly correlated with spectral state (see, e.g., Klein-Wolt & van der Klis 2008). The power spectrum of the LHS is characterized by strong broadband variability (0.01–100 Hz fractional rms amplitude up to 50%). During the HIMS so-called type-C QPOs are observed (see Wijnands et al. 1999; Remillard et al. 2002; Casella et al. 2005, for QPO-type definitions), often with strong harmonic content and accompanied by strong broadband variability (fractional rms up to 30%). In the SIMS various types of variability are observed: power spectra with type-B QPOs (also with strong harmonic content) or (weaker) type-A QPOs, but also power spectra with weak peaked noise and/or QPO features that have been poorly characterized. All SIMS power spectra have in common that their broadband variability is (considerably) weaker than in the LHS and HIMS. Type-C QPOs are typically observed between 0.1 and 10 Hz, whereas type-A and type-B QPOs are generally confined to the 4–8 Hz range. In the HSS variability reaches its minimum strength (a few percent rms); sometimes very weak QPOs above 10 Hz are seen (Homan et al. 2001).

In this Letter, we present a study of Rossi X-ray Timing Explorer (RXTE) observations of MAXI J1659–152 (henceforth J1659). This source was discovered with the Swift Burst Alert Telescope (Barthelmy et al. 2005) on 2010 September 25 and was initially thought to be a gamma-ray burst (GRB 100925A; see Mangano et al. 2010). Later, it was suggested by Kann (2010) to be a new Galactic X-ray transient due to its persistent X-ray emission; this was confirmed by Negoro et al. (2010) with MAXI observations. J1659 has also been detected in the radio (van der Horst et al. 2010a), optical (Jelinek et al. 2010), and submillimeter bands (de Ugarte Postigo et al. 2010). The aim of this Letter is to discuss the results of the aperiodic timing analysis and color evolution from RXTE observations, based on which we conclude that J1659 is a black hole candidate (Kalamkar et al. 2010).

We analyzed all 65 RXTE Proportional Counter Array (Jahoda et al. 2006) observations of J1659 taken between 2010 September 28 (MJD 55467) and 2010 November 8 (MJD 55508), after which the source was not observed due to solar viewing constraints. The intensity and colors were obtained using the Standard 2 data from all the Proportional Counter Units (PCUs) active during the observations. The intensity is defined as the count rates in the 2.0–20.0 keV band; hard color as the count rate ratio of the 16.0–20 keV and 2.0–6.0 keV bands. Count rates are background corrected and normalized using the Crab Nebula observations. This method of normalization (Kuulkers et al. 1994) is based on the assumption that the Crab Nebula is constant in intensity and hard color. It is used to correct for differences between PCUs (see, e.g., van Straaten et al. 2005, for details).

The timing analysis was done with the Event mode data in the 2–60 keV range. Power spectra were generated for each observation by averaging the fast Fourier transform power spectra of continuous 128 s intervals, with a Nyquist frequency of 4096 Hz and a lowest frequency of 7.8 mHz. Periods of dipping activity in the X-ray light curves, as reported by Kuulkers et al. (2010), were not excluded from our timing analysis. The Poisson noise spectrum is estimated with the analytical function of Zhang et al. (1995) and subtracted from this average power spectrum. The resulting power spectrum is expressed in source fractional rms normalization (van der Klis 1989), using the average background rate during the observation. Each power spectrum was fit with a multi-Lorentzian (1–5 components) function and (when necessary) a power law.

We performed a spectral analysis for only a few selected observations, including those cases where HID location and timing properties were inconclusive with respect to the spectral state of the source. Spectra were extracted from the Standard 2 mode data of PCU2, using HEASOFT v6.10. The spectra were background subtracted, corrected for dead time, and a systematic error of 0.6% was applied. Fits were made with XSPEC v12.6.0 (Arnaud 1996) between 3.0 and 40.0 keV, with phenomenological models consisting of a simple accretion disk component (diskbb), a (cutoff) power law, and a Gaussian line (fixed at 6.4 keV). Reported fluxes are in the 2–20 keV band and are corrected for interstellar absorption, which was fixed to 1.7 × 10²¹ atoms cm⁻², the average value in the direction of J1659 (Kalberla et al. 2005).

3.1. Light Curves and Color Evolution

The top panel of Figure 1 shows the 1.5–20 keV light curve from MAXI. It includes the early rise and late decay, which were not covered by our RXTE data set (see below). Panels B–E of Figure 1 show the RXTE 2–20 keV light curve, hardness curve, time evolution of fractional rms amplitude (0.01–100 Hz), and QPO frequency evolution, respectively. The outburst showed a relatively rapid rise; when it was first observed with RXTE, three days after discovery, it already had a flux of 6.6 × 10⁻⁹ erg cm⁻² s⁻¹. The rise was followed by an extended plateau during which some irregular intensity variations (timescale of a few days) were observed (hereafter referred to as "flares"). During two observations in this flaring phase (MJD 55488.02 and 55490.11), we observed several rapid (∼20 s), almost step-function-like changes in the count rate from the source. The changes were on the order of 30%, lasted ∼100–1500 s, and were not accompanied by changes in the spectral hardness. Similar phenomena have been observed in, e.g., GX 339-4 and XTE J1859+226 (Miyamoto et al. 1991; Casella et al. 2004) and have been referred to as "flip-flops." For both observations the high and low count rate levels were separated and treated as different observations for the remainder of our analysis (and are plotted separately in Figures 1 and 2). The flux reached its maximum during a flare on October 8 (MJD 55477) at a value of 1 × 10⁻⁸ erg cm⁻² s⁻¹. A sudden radio flux decrease was reported for that day (van der Horst et al. 2010b). From the intensity plateau the source evolved toward a slow decay, after which it presumably returned to quiescence. The flux during the last RXTE observation was 1.4 × 10⁻⁹ erg cm⁻² s⁻¹.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 2(a) shows the HID. The source started in the upper-right corner and traversed its track in a counterclockwise manner, ending in the lower-right corner. Hysteresis is observed in the spectral evolution, with the hard-to-soft spectral evolution occurring at a count rate ∼3 times higher than the soft-to-hard spectral evolution. While the spectral evolution during the early and late phases of the outburst was largely monotonic, the source exhibited frequent back-and-forth motion in spectral hardness during the flaring phase of the outburst.

Since RXTE did not observe J1659 during the first three days of its outburst, we used MAXI (Matsuoka et al. 2009) data to infer the spectral evolution during the early rise. We plot the MAXI HID in Figure 2(c). Overall, the path traced out is similar to that seen in the RXTE HID. The first part of the MAXI HID (i.e., the right vertical branch and subsequent left turn) was not covered by RXTE. A comparison with HIDs of other BHTs suggests that during the first three days of the outburst, J1659 evolved through parts of the LHS and HIMS.

Shaposhnikov & Yamaoka (2010) reported that during the softest part of the outburst J1659 reached the HSS. We performed a spectral fit of the observation with the lowest spectra hardness (MJD 55489.26) to verify this. We obtain an excellent fit (reduced χ² = 0.96, for 62 dof) with a 0.8 keV disk blackbody and a power-law component with index 2.2 (plus a weak Gaussian at 6.4 keV). The disk component contributed ∼60% to the 2–20 keV flux. Belloni (2010) does not provide a spectral definition of the HSS, but following the definition of the soft (or thermal dominant) state by Remillard & McClintock (2006), which uses 75% as a lower limit, we conclude that the source did not reach the soft state during this outburst. This behavior is similar to, e.g., the 2000 outburst of XTE J1550–564 (Miller et al. 2001).

3.2. Variability Properties

Figure 2(b) shows the relation between hard color and broadband rms. Such a diagram can be helpful in identifying different types of power spectra (Belloni et al. 2005; Fender et al. 2009). At high hard colors, hardness and rms are correlated, but at low hard colors (<0.5) there is considerable scatter.

Based on the overall strength of the broadband variability, the spectral hardness, and the shape of the power spectra (noise/QPOs), we divided our RXTE observations into five groups (see symbols in Figures 1 and 2). For the classification of the PDS and the QPOs, we made use of the classification schemes of Remillard et al. (2002) and Casella et al. (2005). Representative power spectra are shown in Figure 3 and are referred to as PDS 1–5.

Zoom In Zoom Out Reset image size

3.2.1. Group 1: Type-C QPOs

The Group 1 PDS have strong 0.01–100 Hz variability (∼23%), a strong QPO (varying between 1.6 and 8.5 Hz), at times with a harmonic and a sub-harmonic. They were observed during the spectrally hard parts of the outburst (hard color > 0.45; see open circles in Figures 1 and 2). The frequency is strongly anti-correlated with spectral hardness (Figure 1(e)). The QPO frequencies and Q values, and the fractional rms amplitude in the 0.01–100 Hz frequency range are typical of PDS exhibiting type-C QPOs, as observed in the HIMS (no phase lags or energy dependence of the QPOs were measured). Hence, we identify these QPOs as type-C.

3.2.2. Group 2: Type-B QPOs

The Group 2 PDS have 0.01–100 Hz variability <10% rms and power-law noise as opposed to the strong variability and peaked noise of the Group 1 PDS. These PDS are observed in a fairly narrow range of spectral hardness, have QPOs with frequencies between 2.6 and 4.1 Hz, at times accompanied by a harmonic and a sub-harmonic, and cluster in a small patch of points in the hardness–rms diagram. In Figure 4, we plot the 0.01–100 Hz fractional rms amplitude versus frequency of these and the PDS with type-C QPOs, as well as the PDS with the unclassified QPOs (see below). What we observe in Figure 4 is very similar to the results of Casella et al. (2004), indicating that these are type-B QPOs, which classify these observations as SIMS. Fractional rms amplitude anti-correlates with frequency for type-C QPO, while type-B QPOs do not follow such an anti-correlation. Note that after the detection of the first two type-B QPOs (at 3.9 Hz and 4.1 Hz), the spectrum temporarily hardened during the next two observations, and type-C QPOs (at 6.0 and 6.9 Hz) were observed, (see Figure 1) followed again by type-B QPOs (3.9 and 3.5 Hz). This indicates rapid SIMS/HIMS transitions (Wijnands et al. 1999) over a period of ∼6 days.

Zoom In Zoom Out Reset image size

3.2.3. Groups 3 & 4: Weak Power

In the soft part of the outburst most power spectra do not show clear indications for QPOs. Such power spectra were observed during two time intervals. During the first time interval (MJD 55486–55590), the source intensity moved up and down on a daily basis—this is also the period during which the flip-flops were observed (see Section 3.1). Inspection of the individual power spectra revealed subtle differences between the low count rate and high count rate observations. PDS3 is the averaged power spectrum of the five high count rate level observations (weak1 in Figures 1 and 2); it is consistent with a single power law with no distinguishable features. The low count rate observations show indications for an additional broad bump around 7–8 Hz. Such a feature is also seen for the observations in the second interval with weak power (MJD 55496.5–55498.5). PDS4 is the averaged power spectrum of these nine observations (weak2 in Figures 1 and 2). The PDS4-like power spectra occurred at slightly lower hardness than the PDS3 ones. Power spectra similar to PDS3 and PDS4 were observed in GX 339-4 (Belloni et al. 2005) and were grouped together with power spectra that show type-A (not observed here, see below) and type-B QPOs to define the SIMS. Therefore, also based on this definition J1659 did not reach the HSS.

3.2.4. Group 5: Unclassified QPO

For three power spectra that showed somewhat broad QPOs (Q ∼ 1.6–3.5), we had difficulty assigning a classification—they are marked as unclassified in Figures 1–4. The observations were among the shortest (800 s) and the power spectra were of lower quality. However, the locations of these observations in Figure 4 (and also Figure 2(b)) suggest that these QPOs may be type-B QPOs with a lower frequency (∼1.9 Hz) and also lower Q-value, which would classify these observations as SIMS.

We did not see indications for type-A QPOs in our observations. These are typically accompanied by similarly shaped noise as type-B QPOs, but show no strong harmonic content (as seen in PDS2). Also in the rms–frequency plot they fall in a different location than type-B QPOs (Casella et al. 2005), which argues against the QPOs in the type PDS5-like power spectra being of type-A.

3.2.5. Search for High-frequency QPOs

High-frequency QPOs in the the 50–450 Hz range have been reported for a handful of BHTs (McClintock & Remillard 2006). We inspected all individual power spectra in the total 2–60 keV band, as well as in the 5.7–60 keV band, but no indications for such QPOs were found. We also inspected the averaged power spectra of the different groups (with groups 2 and 5 added together). Again, no significant QPOs were found, with a 2.1σ measurement of a peak at ∼160 Hz (Q ∼ 4, ∼4% rms) in the 2–60 keV power spectrum of PDS 4 constituting our most significant detection.

We analyzed all the RXTE observations of MAXI J1659–152 taken during its 2010 outburst. Spectral states were identified using the HID in combination with the aperiodic variability properties. In the HID J1659 traced out a counterclockwise loop, similar to those seen in black hole and neutron star transients (see Section 1). The types of power spectral features we identified (in particular type-B and type-C QPOs), and the way they evolve along the HID track, indicate that J1659 is a black hole candidate. The spectral decomposition of the softest spectrum (i.e., diskbb + power-law, without the need for second thermal component; Lin et al. 2007), the radio loudness in the early part of the outburst (Paragi et al. 2010), the flip-flops seen in two observations, and the sudden (∼day) drop in the rms (i.e., PDS3) at low hardness are all characteristic of this type of systems (Dunn et al. 2010; Miyamoto et al. 1991; Fender et al. 2009), strengthening our identification of J1659 as a black hole candidate.

Based on spectral and variability properties, we find that J1659 moved from a HIMS to a SIMS and back again (at lower intensity), without reaching the HSS (or the soft or thermal dominant state as defined by Remillard & McClintock 2006). Our RXTE observations started after the source had left the LHS (as indicated by MAXI observations) and ended before it had returned to that state.

The observations in the softest part of the SIMS had weak variability and revealed two types of power spectra, one consistent with a power law and one with an additional bump around 7–8 Hz. We note that similar groups of observations with weak variability were observed in the SIMS of GX 339-4 by Belloni et al. (2005). Fender et al. (2009) found that radio flares and subsequent quenching of the radio flux often occur in a time interval of a few days before and after the time of sudden drops in the rapid X-ray variability (identified as a distinct zone in their rms–hardness diagrams). In J1659 a similar drop occurred on MJD 55481, a few days after the quenching of the radio flux as reported by van der Horst et al. (2010b), consistent with other BHTs.

Optical observations constraining the mass of the compact object are required to confirm the black hole nature of J1659. From dipping episodes in the X-ray light curves, the orbital period has been proposed to be 2.4–2.5 hr (Kuulkers et al. 2010; Belloni et al. 2010). This makes the source very interesting as, if confirmed, this would make it the black hole binary with the shortest known orbital period.

This research has made use of the MAXI data provided by RIKEN, JAXA, and the MAXI team and has also made use of data obtained from the High Energy Astrophysics Science Archive Research Center (HEASARC), provided by NASA's Goddard Space Flight Center. M. Linares acknowledges the support from NWO Rubicon Fellowship. P. Casella acknowledges funding via a EU Marie Curie Intra-European Fellowship under contract no. 2009-237722.