THE SPECTRAL AND TEMPORAL PROPERTIES OF TRANSIENT SOURCES IN EARLY-TYPE GALAXIES

A TC is usually defined to be a source that either appears or disappears, or is visible for only a limited amount of "contiguous" time during observations, where the flux ratio between the peak "on-state" emission and the non-detection upper limit, or "off-state" is greater than a certain value (usually between 5 and 10; e.g., Williams et al. 2008). However, using only this ratio as a discriminator can lead to overestimating the number of transients by including lower luminosity sources, which have poorly constrained L_X values. Following B08 and B09, we instead use the Bayesian model developed by Park et al. (2006; see Section 2.4 of B08 for details), which estimates the uncertainties in the ratio, as well as lower bounds.¹¹ We classify sources with a lower bound ratio >10 as TCs and sources with ratios between 5 and 10 as PTCs. We note here that a small number of Galactic sources have been observed to vary by ratios >10 but do not go into periods of quiescence, e.g., 4U 1705-440 (Homan et al. 2009) and 4U 0513-40 (Maccarone et al. 2010b), where we particularly note that this latter source has been confirmed to be an ultracompact binary (Zurek et al. 2009).

From the multi-epoch observations of the galaxies presented, the Bayesian analysis leads to the detection of 18 TCs and PTCs, listed in Table 2. In this table, Column 1 indicates the source numbering used in this paper and Column 2 indicates the original source number as presented in each catalog paper (B08, B09, S08, or "new" for sources detected in the NGC 4278 cycle 11 observations). Columns 3 and 4 provide R.A. and decl. of each source, Column 5 indicates the observation ID(s) in which the source was detected. Column 6 provides net counts for that pointing (0.3–8.0 keV), Column 7 the mode ratio, Column 8 the lower-bound ratio from the Bayesian analysis, Column 9 marks the source as a TC or PTC, and Column 10 the long-term light curve behavior (see Section 2.2 for an explanation). In Column 11, additional notes for each source are provided and Column 12 provides g-band magnitude (Vegamag) optical upper limits for sources that have no detected optical counterpart. These upper limit values are the 3σ flux values centered on the R.A. and decl. values in Columns 3 and 4. In B08 there was no optical coverage for source A8 but from recent Hubble Space Telescope (HST) observations it can now be confirmed that this source is coincident with a GC (with a separation <01), with g- and z-band (Vegamag) magnitudes of 23.4 mag and 21.9 mag, respectively. This corresponds to a color of 0.8 (ABMAGs), which indicates that the cluster is blue.¹² The long-term light curves are shown in Figure 1 where the top, middle, and bottom panels present the TCs and PTCs in NGC 3379, NGC 4278, and NGC 4697, respectively (in Figure 2 individual long-term light curves are presented for the four additional sources that are not presented in B08 or B09).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In summary, we observe five TCs and three PTCs in NGC 3379, three TCs and five PTCs in NGC 4278, and one TC and one PTC in NGC 4697. Using the HST WFPC2 and Advanced Camera for Surveys observations of these galaxies we find that three transients reside in GCs, and 14 in the galaxy stellar field. The source A3 in the NGC 3379 field is found to be coincident with a background object; we only list this source and report its properties for completeness.

The long-term light curves (Figure 1) show three TCs with "on states" of at least three months in NGC 3379. Of these sources, one has a maximum outburst time of nine months (A1), while the other two (A6 and A8) could have maximum outburst times up to four years and five months, respectively. In NGC 4278, only source B7 was observed in two epochs (both in 2010 March), with a minimum outburst time of ∼5 days and unconstrained maximum outburst time; the other TCs and PTCs were detected in single epochs, giving minimum outburst times of hours. Upper limits to the duration of the outbursts ranging from months to a few years can be placed on seven TC/PTCs, while the remaining seven were detected at the beginning or end of the monitoring observations and therefore the duration of the outburst is unconstrained. In Table 2, Column 10 lists our estimated outburst durations.

Source A5 is denoted as only being in outburst in one observation in Table 2, although in B08 it was reported as being detected in three out of five observations. However, A5 was also flagged as a possible double source in B08, a conclusion supported by our further investigation (see Figure 3). The counts from observation 1 arise almost exclusively from within the top white circular region shown in this figure, whereas in the two subsequent observations the counts are centered in the lower green circle. The luminosity from observations 2 and 3 is more than a factor of 15 lower, suggesting that this secondary source is a variable lower flux object.

Zoom In Zoom Out Reset image size

Short-term variability was investigated for all sources with net counts >20 in a single observation, using both the Kolmogorov–Smirnov test and the Bayesian blocks method (Scargle 1998; see Section 2.4 of B08 for more details). Of the 18 transient sources presented in this paper, only A5 and A8 exhibited detectable short-term variability during their outburst.

The intraobservation variability of these two sources was investigated in more detail by extracting the light curves from the "on" observations (and the results are discussed in Sections 3.5.1 and 3.3, respectively). This analysis was performed with the CXC CIAO software suite (v4.2)¹³ and HEASOFT (v5.3.1), where barycenter corrected binned light curves were extracted with the CIAO tool dmextract. The binning for these two sources was selected to provide well-constrained values, allowing variability to be identified, while remaining fine enough to ensure that any changes in flux were not masked by the selected binning scheme. Counts were extracted from circular source region files with radii as in B08. Background counts were extracted from a large elliptical source-free region, located within the overlapping area covered in all observations.

Hardness ratios (HRs) and colors were derived for each source, as described in Section 2.3 of B08. They are listed in Table 3 along with log L_X. These luminosity values were derived from the net count rate, assuming a power-law shape of Γ = 1.7 and Galactic N_H (see Section 2.1 in B08 for further details).

Spectral extraction was performed for the "on" observations for the TC/PTC sources, using the CIAO tool psextract. We used the same circular extraction regions as in B08 or B09; for the TC/PTC discovered in our new analysis, the radii are: 25 (B7), 30 (B8), 22 (C1), and 30 (C2); background counts were extracted from surrounding annuli, with outer radii two to three times larger, depending on the presence of nearby sources. In cases where the source region was found to overlap with a nearby source, the extractions region radius was reduced (to a minimum of 15) and the area of the overlapping source was excluded.

The source spectra were fitted in XSPEC (v12.5.1). We used only data between 0.3 and 8.0 keV to avoid the calibration uncertainties at the low energies and the high cosmic-ray background above 8.0 keV. All the spectra were fitted to two models that describe the properties of X-ray binary spectra well (see, e.g., review by Remillard & McClintock 2006): the multicolor disk blackbody (DISKBB in XSPEC; hereafter referred to as DBB) and the power-law (PO) models. For each model all parameters were allowed to vary freely, with the exception of the absorption column N_H, described by the X-ray absorption model tbabs, which, in cases where the best-fit value was below that of the Galactic absorption, was frozen to the Galactic value.¹⁴ In instances where there were sufficient counts to bin the data to at least 20 counts per bin (allowing Gaussian error approximation to be used), the minimum χ² method was used to fit the data. The Cash statistic (Cash 1979) was otherwise used, however, this statistical method has the disadvantage that it does not provide a goodness-of-fit measure like χ².

Table 4 summarizes the best-fit models, with the three galaxies listed separately. Column 1 contains the source number, Column 2 the observations used in each fit, Column 3 the net source counts from the spectral extraction, Column 4 indicates which spectral model was used, Column 5 the fit statistic (χ² and number of degrees of freedom ν, or the C-statistics—indicated by C), and Column 6 the null hypothesis probability (or goodness when using Cash). Columns 7–9 present the best-fit values of the fit parameters with 1σ errors for each interesting parameter (F denotes that the value was frozen): N_H, Γ for the power-law model, and kT_in the temperature of the innermost stable orbit of an accretion disk, for the multicolor disk model. Column 10 indicates the intrinsic value of L_X and Column 11 the luminosity range from the lowest (1σ) lower bound to the highest (1σ) upper bound of each observation included in the joint fit.¹⁵

Table 4. Summary of the Best-fit Parameters of the Source Spectra

Source	ObsID	Net Counts	Model	χ2/ν	$P_{\chi ^2}$	Parameters			LX × 1038 erg s−1	LX Range
						NH × 1020	Γ	kTin	(0.3–8.0 keV)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
A1	7074	153.5	Power law	238C	44%C	20.3+8.3−7.4	2.59+0.33−0.31	...	4.89	4.12–5.39
			DBB	236C	93%C	2.8F	...	0.64+0.08−0.07	2.42	1.48–2.62
A2	7073	42.2	Power law	191C	77%C	2.8F	1.70F	...	0.79	0.48–1.06
			DBB	188C	29%C	2.8F	...	1.00F	0.62	0.41–0.83
A3	7074	31.4	Power law	148C	54%C	2.8F	1.70F	...	0.66	0.43–1.00
			DBB	149C	67%C	2.8F	...	1.0F	0.53	0.27–0.75
A4	7076	87.3	Power law	247C	77%C	2.8F	1.64+0.19−0.19	...	2.14	1.68–2.52
			DBB	247C	85%C	2.8F	...	1.16+0.26−0.22	1.74	1.08–1.98
A5	1587	60.0	Power law	111C	95%C	75+31−22	6.1+1.7−1.0	...	139	108–170
			DBB	108C	82%C	30+18−14	...	0.22+0.06−0.04	4.42	0.81–5.42
			BB	108C	60%C	19+19−15	...	0.19+0.04−0.04	2.46	0.00–2.89
Flare	1587	41.9	Power law	114C	54%C	2.8F	2.42+0.29−0.26	...	14.11	10.46–25.72
			DBB	92C	69%C	30.7F	...	0.22 +0.03−0.02	29.11	20.10–33.18
			DBB	99C	16%C	2.8F	...	0.33+0.06−0.05	9.46	4.15–10.01
			BB	90C	89%C	19.2F	...	0.19+0.02−0.02	16.50	10.41–20.00
			BB	94C	73%C	2.8F	...	0.22+0.02−0.03	9.42	5.69–12.72
A6	7073	78.3	Power law	43C	12%C	20+7−6	8.19+0.94−0.73	...	6.26	1.91–10.12
	and 7074		DBB	87C	11%C	8+9−6	...	0.09+0.02−0.01	1.54	0.54–2.82
			BB	94C	2%C	2+5−2	...	0.09+0.02−0.01	1.04	0.00–8.31
			BB	94C	4%C	2.8F	...	0.09+0.02−0.01	1.12	0.622–1.67
A7	1587	17.8	Power law	75C	67%C	2.8F	1.70F	...	0.75	0.49–1.20
			DBB	77C	79%C	2.8F	...	1.00F	0.61	0.34–0.92
A8	7073	521.2	Power law	20.5/23	0.61	2.8F	1.11+0.08−0.07	...	14.83	13.87–16.23
			DBB	27.2/23	0.25	2.8F	...	2.60+0.34−0.46	13.97	5.01–15.52
B1	7079	222.2	Power law	2.80/8	0.09	1.1+5.7−1.1	1.59+0.20−0.14	...	6.33	5.90–7.46
			Power law	3.33/9	0.10	1.8F	1.60+0.16−0.15	...	6.34	5.63–7.43
			DBB	6.45/9	0.07	1.8F	...	1.28+0.12−0.23	5.36	2.97–5.97
B2	7079	100.1	Power law	184C	66%C	16.6+12.2−9.2	3.25+0.72−0.56	...	1.87	1.81–2.33
			DBB	181C	72%C	1.8F	...	0.41+0.07−0.05	1.03	0.66–1.07
B3	7079	163.1	Power law	224C	66%C	29.8+9.6−8.8	3.21+0.50−0.45	...	9.57	7.33–9.95
			DBB	220C	60%C	4.9+4.7−4.9	...	0.55+0.13−0.07	2.57	1.67–2.77
B4	7077	53.1	Power law	173C	64%C	1.8F	1.7+0.23−0.22	...	1.22	0.90–1.64
			DBB	166C	71%C	1.8F	...	0.79+0.18−0.12	0.84	0.42–1.05
B5	7080	20.5	Power law	114C	76%C	1.8F	1.7F	...	0.63	0.38–1.17
			DBB	112C	52%C	1.8F	...	1.0F	0.53	0.23–0.91
B6	4741	313.1	Power law	12.60/12	0.40	9.4+4.1−5.1	2.00+0.14−0.18	...	23.22	22.28–25.91
			DBB	18.5/13	0.14	1.8F	...	0.82+0.09−0.09	14.62	10.13–15.19
B7	11269	87.1	Power law	317C	95%C	1.8F	1.3+0.35−0.30	...	2.53	1.86–3.80
	and 12124		DBB	317C	97%C	1.8F	...	1.16+1.01−0.34	1.63	1.13–2.13
B8	11269	60.38	Power law	143C	24%C	1.8F	2.40+0.33−0.20	...	1.73	1.39–2.36
			DBB	131C	55%C	8.8+10.3−8.6	...	0.36+0.09−0.07	1.55	0.54–1.64
			BB	131C	67%C	1.8F	...	0.28+0.03−0.03	1.11	0.82–1.39
C1	4730	86.9	Power law	100C	50%C	82.2+3.6−22.1	9.87+0.03−9.87	...	11590	unconstrained
			DBB	69C	57%C	110+44−31	...	0.06+0.02−0.01	16557	8160–18670
			BB	67C	58%C	100+32−31	...	0.06+0.01−0.01	5752	4400–7076
C2	4730	15.8	Power law	91C	86%C	2.1F	1.7F	...	0.40	0.17–0.69
			DBB	92C	83%C	2.1F	...	1.00F	0.31	0.15–0.58

Notes. Errors are given to 1σ. These spectral models are presented and discussed in Sections 2.4 and 3.

Download table as:  ASCIITypeset image

Figure 4 shows that the majority of TC/PTCs have both HR and colors consistent with those of the persistent LMXBs in the galaxies. Five of these sources have insufficient counts to allow even simple spectral modeling (A2, A7, B5, C2, and the background source A3, which will not be discussed further; see Table 4) and therefore single-component models with canonical values have been applied (Γ = 1.7 or kT_in = 1.0 keV; Remillard & McClintock 2006). Six more sources (A1, A4, B1, B4, B6, and B7) have spectra well described by single-component models. Four of these (A4, B1, B4, and B7) are well fitted with a PO model with Γ ∼1.7 and Galactic N_H. A1 instead requires a large value of N_H in the PO fit. Based on the simulations of Brassington et al. (2010), this result may favor disk emission. The single-component PO model of source B6 also shows an elevated value of N_H; however, the two-dimensional errors on this value indicate that the best-fit N_H is consistent with Galactic absorption (Figure 5).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Sources B2 and B3 have colors and HR values softer than the persistent LMXB population (Figure 4). Their spectra are well fitted by both a single PO and a single DBB model. However, the PO model in both cases requires an elevated value of N_H, indicative of the source containing a significant disk component (Brassington et al. 2010). In the case of B2, the best-fit value of N_H from the DBB model tends to zero when left free to vary, indicating that there is likely a non-thermal component as well as disk emission in this source (Brassington et al. 2010). When freezing the absorption column to the Galactic value, the best-fit disk temperature is 0.41 keV, with an X-ray luminosity of 1.0 × 10³⁸ erg s⁻¹. The spectrum of source B3 is well described by a DBB model with an absorption value more than twice that of the Galactic absorption value (albeit poorly constrained), indicating that more than 75% of the emission is arising from the disk component, with a temperature of 0.55 keV.

Four further sources (A5, A6, B8, and C1; see Figure 4 to compare with the persistent sources) all have little emission above 2 keV and are therefore super soft or quasi-soft sources (SSS or QSS, respectively). In all cases, the PO fits yielded extremely steep power-law indices (>6), reflecting the very soft nature of the emission. Their spectra were also fitted with blackbody (BB) models, as reported in Table 4.

When simple single-component models did not provide an adequate description of the source spectrum, we used more complex models, as discussed below for sources C1 (Section 3.4.1) and A8 (Section 3.3).

One of the clear results from this work is that TC/PTC sources are predominantly found in the field with only 3 sources out of the 17 TC/PTCs (excluding the background source) coincident with a GC.¹⁶ These numbers support the theory which suggests that the LMXB field population largely arises from relatively detached systems evolved from native field binaries and as such will exhibit transient accretion for >75% of their lifetime (Piro & Bildsten 2002). This is as opposed to transient GC-LMXBs, which have a more difficult formation channel (as discussed in Section 3.2).

From the detailed spectral analysis presented in this paper we have been able to determine that out of the 14 confirmed field sources, 10 exhibit colors and spectra consistent with typical emission from LMXBs, with power-law (Γ 1.3–2.0) or disk (kT_in 0.4–1.0 keV) emission and on-state intrinsic luminosities L_X ranging from 4 × 10³⁷ erg s⁻¹ to 2.32 × 10³⁹ erg s⁻¹ (see Table 4). The spectra of these sources are consistent with those of neutron star (NS) or black hole (BH) binaries in either a hard state (HS) or thermally dominant (TD) state (Remillard & McClintock 2006).

Only one of these sources has a luminosity significantly above the Eddington limit for a 1.4 M_☉ NS (B6) and is a strong BH candidate. A further three sources that have been determined to be in a TD state (A1, B2, B3) exhibit inner-disk temperatures that are softer than the spectra typically observed in Galactic NS-LMXBs (kT_in 1–2 keV), indicating that they could be BH candidates. A further three field sources (A4, B4, and B7) have been determined to be in an HS, but have X-ray luminosities ⩾10% of the Eddington limit. This high accretion is not observed in NS-LMXBs in an HS and suggests that the primary of these binaries may be BHs. However, due to the quality of our data, while an HS is the preferred interpretation, we cannot rule out any of these sources emitting in a TD state, and therefore cannot exclude any of these sources from being NS-LMXBs. The remaining three sources have low-count data and have L_X values derived from canonical models, therefore these sources could be NS or BH-LMXBs. A further soft source (B8) has been classified as a QSS but is also consistent with a NS-LMXB (or white dwarf (WD) LMXB; see Section 3.5). This results in a ratio of BH to NS-LMXBs in transient field sources ranging from at least 10% up to 90% of sources. Such an unconstrained ratio is consistent with a study of the transient population of M31 (Williams et al. 2006) which suggests that the majority of the sources that they identify to be LMXBs contain BHs.

3.1.1. Comparison with PS Models of Field LMXBs

Three of the TC/PTCs presented in this work are in GCs. It has been suggested that there is a trend for GCs to have a higher fraction of ultracompact binaries, compared to the field LMXB population (Deutsch et al. 2000; Heinke 2010). However, for sources brighter than ∼ × 10³⁷ erg s⁻¹ such objects are expected to be persistent (Lasota et al. 2008; Bildsten & Deloye 2004) and are therefore not the transient sources presented here. Instead these sources could be NS with main sequence or red giant donors, where population synthesis models of the formation of LMXBs in GCs predict that these systems, in addition to ultracompact binaries, also contribute to the GC-LMXB population (Ivanova et al. 2008). Such sources were shown to exhibit transient emission from 60% to all of their mass-transfer lifetimes. We therefore suggest that A2, a source with low counts, which has therefore been described with a canonical PO spectrum, is likely to be a NS-main sequence or NS-red giant system with L_X ∼8 × 10³⁷ erg s⁻¹ (although NS-red giants are less favored as Ivanova et al. 2008 predicted that their numbers are low when considering their formation rates and short lifetimes).

The second GC-LMXB, B1, has a spectrum that is well described by both a PO and a DBB model; however, the best-fit L_X from the PO model indicates that the source is emitting greatly above the 2%–4% Eddington luminosity that is typically seen from an LMXBs in an HS (Maccarone 2003) and the DBB model is therefore preferred.¹⁷ The best-fit values from the DBB model determine an inner disk temperature of kT_in∼ = 1.3 keV and L_X = 5.4 × 10³⁸ erg s⁻¹. Therefore, because of this high luminosity, which is above the Eddington limit for a NS binary (although could be close to the limit of a heavy (2–3 M_☉; Kalogera & Baym 1996) NS, or a 1.4 M_☉ NS with an He or C/O donor), it is possible that B1 is a BH binary instead of a NS-main-sequence system.

The third GC-LMXB in this sample is A8 (presented in Section 3.3), which could be a BH-ULX undergoing a period of super-Eddington accretion.

The detection of a BH binary in a GC is thought to be rare, since in tidally captured persistent sources BHs are likely to be expelled from the GC (Spitzer 1969); they could be transients with very low DCs from an exchange interaction (Kalogera et al. 2004). More recent work has indicated that a significant number of BH can be retained in GCs (Mackey et al. 2007; Moody & Sigurdsson 2009). In particular, Maccarone et al. (2011) point out that once a number of BHs are ejected from their host GC the ratio of the masses between the heavy (BHs) and light (non-BH) components falls below the critical value and the Spitzer instability criterion is no longer met, therefore not all BHs will be expunged.

Observationally, the existence of accreting BHs in GCs has been supported by the detection of five luminous (above the NS Eddington limit), variable sources (Maccarone et al. 2007; Brassington et al. 2010; Shih et al. 2010; Irwin et al. 2010; Maccarone et al. 2011). Using the TC/PTC criteria defined in this paper, four of these binary systems are persistent and could be ultracompact binaries with either a BH (Gnedin et al. 2009) or NS with super-Eddington accretion and rather mild beaming (King 2011).¹⁸ The BH-GC candidates A8 and B1 exhibit large flux variability, with L_X >1 × 10³⁸ erg s⁻¹, therefore ruling out genuine disk instability transients. They are also unlikely to be the mildly beamed super-Eddington NS-LMXBs suggested by King (2011).

Instead, A8 and B1 could represent BHs in a GC that have been formed through an exchange interaction, resulting in transient behavior (Kalogera et al. 2004). In Barnard et al. (2011) this formation channel has been advocated to explain a recurring transient in M31, which has been observed with an outburst luminosity ∼L_Edd and therefore could be a BH-GC candidate (although a NS binary cannot be ruled out). In their work they also suggest that the source could have formed through a complex interaction of a triple system resulting in a BH–WD binary, as first theorized by Ivanova et al. (2010). Further, additional observations of the first strong BH-GC candidate (first presented in Maccarone et al. 2007) have led to the suggestion that this source is also a triple system, with the inner binary comprising a BH and WD (Maccarone et al. 2010a).

The properties of the host clusters for the five BH-GC candidates are presented in Maccarone et al. (2011), where they demonstrate that this small population suggests that BH binaries are favored in massive "red" GCs. The properties of the GCs hosting A8 and B1 are also massive with g = 23.4 and V = 21.6, respectively. However, both of these sources have color values indicating that these GCs are blue (although B6 has a (V − I) of 1.00 ± 0.02 which is only marginally classified as a "blue" low metallicity cluster: Fabbiano et al. 2010; blue clusters have V − I ⩽ 1.05). This further indicates that the mass of the GC is important in the formation of GC-LMXBs in these environments. However, we do not comment on the influence of GC colors in the formation of BH-GCs, as the metallicity correlation presented in Maccarone et al. (2011) was only suggestive.

The GC source A8 in NGC 3379 is very luminous (Table 4) and variable (Section 2.2). The photometric parameters are consistent with those of the persistent LMXB population (Figure 4), and a single-component PO model provides a statistically acceptable description of the spectrum, with Γ  = 1.1 (Table 4). However, there is a statistically significant deficiency in the spectrum at ∼1.0 keV as well as excess emission <0.7 keV (Figure 6). These features cannot be modeled with either a composite DBB plus PO model, additional neutral absorption components, or a thermal plasma model. A single-component power-law model, modified by an ionized intrinsic absorber (absori), was fitted to the spectrum. The results of this fit are presented in Table 5, where the best-fit values from an XSPEC model with tbabs×tbabs×absori×po (Column 2) are shown with one of the neutral absorption components being frozen to Galactic N_H.

Zoom In Zoom Out Reset image size

Table 5. Best-fit Parameters of A8 for the Whole Spectrum and the Spectra from the High- and Low-flux States

Parameters	Whole	High-flux	Low-flux
	Spectrum	Spectrum	Spectrum
(1)	(2)	(3)	(4)
Net counts	521.2	366	130
χ2/ν	13.8/20	13.2/13	303C
$P_{\chi ^2}$	0.84	0.04	44%C
NH (1021 cm−2)	0.43+0.54−0.43	1.79+0.99−1.26	0.34+0.68−0.39
Γ	1.45+0.17−0.23	1.89+0.35−0.41	1.04+0.24−0.22
kpo(10−5)	1.2+0.8−0.4	3.7+3.1−1.7	...
Tabs (K)	105F	105F	...
Zabs (Z☉)	1F	1F	...
ξabs (erg cm s−1)	133+113−39	149+79−52	...
NHabs (1022 cm−2)	2.33+2.23−1.46	4.95+2.84−2.25	...
LX (1039 erg s−1)	1.59	2.75	0.89
LX Range (1039 erg s−1)	1.39–1.93	1.97–3.14	0.80–1.03

Notes. Presented are values from the tbabs * tbabs * absori * power-law model with the quoted errors given to 1σ. The difference in counts between the "whole" observation and the high- and low-flux spectra arises from the short time gap between the two extracted spectra. This model is discussed in detail in Section 3.3.

Download table as:  ASCIITypeset image

A8 is one of the two TC/PTC sources exhibiting short-term variability, where the count rate falls from >0.01 counts s⁻¹ to <0.005 counts s⁻¹ ∼42 ks into the observation (see Figure 7). The spectrum of the high flux state requires an ionized absorption component. The best-fit values of the absori power-law model are presented in Column 3 of Table 5, and the spectrum is presented in the left-hand panel of Figure 8. These values are similar to those determined for the spectrum from the full observation, except that the best-fit Γ is 1.89 and N_H is larger (although both parameters are consistent with the values derived from the whole observation when considering the uncertainties). The derived X-ray luminosity for the source in this higher flux state is 2.8 × 10³⁹ erg s⁻¹, in the Ultraluminous X-ray Source (ULX) range.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The spectrum of the lower state is instead adequately described by a single-component PO model (right-hand panel, Figure 8), with N_H consistent with Galactic (Column 4 of Table 5). However, the best-fit photon index is very flat, similar to the value derived when fitting the whole spectrum to a single-component model. This could indicate that ionized absorption may still be present, although there are too few counts to be able to investigate it. The HRs, binned to 5000 s over this period, are constant. From this model a luminosity of 8.9 × 10³⁸ erg s⁻¹ is derived.

This source was also detected in the third observation of NGC 3379 (ObsID 7074), although it only has ∼9 net counts, too few to perform any meaningful spectral analysis. B08 estimated an X-ray luminosity of L_X = 2.5 × 10³⁷ erg s⁻¹. This luminosity then fell to an upper limit of 2 × 10³⁷ erg s⁻¹ in the subsequent pointing, taken three months later.

The parameters determined from the model during the period of high flux emission are similar to those reported for the NGC 1365 flaring ULX by Soria et al. (2007). In A8, the properties of the spectrum (see Table 5) suggest that the source is in some sort of outflowing phase, similar to the decline period of the ULX in NGC 1365. During this phase the spectrum is fitted with an ionized absorption column of ∼5⁺³₋₂ × 10²² cm⁻² and an ionization parameter of ξ ≈150⁺⁷⁹₋₅₂ erg cm s⁻¹, largely consistent with the parameters determined in Soria et al. (2007), of ∼1⁺²₋₁ × 10²² cm⁻² and 111⁺¹⁸⁴₋₈₃, respectively, where they also determine Γ ∼1.9 and neutral absorption column ∼2 × 10²¹ cm⁻². The lower flux spectrum, following the outflow phase reported here, indicates that there has been a three-fold reduction in L_X over the period of 12 hr (assuming a PO model), similar to the factor of ∼2 seen after three days in NGC 1365.

Soria et al. (2007) suggest that such properties could be analogous to flares observed in Galactic X-ray binaries in a steep-power-law state, where the outflow commences after the source exceeds the Eddington limit. Alternatively, some Galactic binaries have been observed to remain in a power-law spectrum HS during short outbursts (Yu & Yan 2009). It has been suggested that some sources with such HS events could reach flux levels which classify them as ULX during their short outburst events (Yu & Yan 2009), as we find in the case of A8. This further strengthens the suggestion of Yu & Yan (2009) that some ULXs that stay in the HS during a flaring event harbor stellar-mass compact stars. We note that Galactic BH binaries have been observed to typically proceed through hysteresis cycles (Remillard & McClintock 2006) meaning that the HS event should be followed by a TD state, which is not observed in the subsequent pointing three months later. However, these canonical state transitions observed for Galactic binaries are often not observed in ULXs (e.g., Feng & Soria 2011 and references therein).

SSS are sources with little or no emission above 1 keV, with typical X-ray luminosities between 1 × 10³⁶ and 1 × 10³⁸ erg s⁻¹. Following the discovery of SSS in the Large Magellanic Cloud (Long et al. 1981), van den Heuvel et al. (1992) proposed a model of quasi-steady nuclear burning on the surface of a WD accreting matter from a Roche lobe-filling companion with high accretion rates, with emission described by a simple absorbed blackbody (BB). Further observations have revealed a heterogeneous class of objects with sources detected in both early- and late-type galaxies, in the field and in GCs, and with a range of temporal behavior, and luminosities as high as L_X >1 × 10³⁹ erg s⁻¹, in the ULX range (e.g., Carpano et al. 2007; Di Stefano & Kong 2003; Fabbiano et al. 2003). It has been suggested that these sources could be WDs in super-Eddington outbursts, stellar-mass BHs, or even intermediate-mass BHs (e.g., Di Stefano et al. 2004).

A6 was observed in the second and third observations of NGC 3379 (ObsID 7073 and 7074). The HR and colors do not change between pointings; consequently the spectra from these two observations were jointly fitted with all parameters apart from the normalization tied between observations. Of the three models, the PO does not provide a physically realistic description of the spectrum, determining a best-fit photon index of ∼8, and can be discounted. The best-fit parameters between the DBB and BB model are very similar, with an inner disk temperature of 91 ± 15 eV, and an absorption column consistent with the Galactic value for the BB model, and an N_H of more than twice the Galactic value for the DBB model, although within errors this is consistent with Galactic N_H. These best-fit models yield L_X = 1.1 × 10³⁸ erg s⁻¹ (BB) and L_X = 1.5 × 10³⁸ erg s⁻¹ (DBB), consistent within the uncertainties.

The spectrum is that of a "classical" SSS. Due to its transient nature it is possible that the observed X-ray emission arises from a classical nova (CNe) where episodic thermonuclear explosions on the surface of the WD lasting from months to years can appear as a transient SSS (Pietsch et al. 2007). From X-ray/optical studies of M31 and M33, Pietsch et al. (2005) suggested that these sources are a major class of SSS.

3.4.1. The Extremely Luminous SSS C1

The second SSS, C1, is even softer than A6, with an inner disk temperature of 60 ± 10 eV and a large absorption column (∼1 × 10²² cm⁻²), yielding an intrinsic extremely high X-ray luminosity of 5.8 × 10⁴¹ erg s⁻¹. The spectrum is shown in Figure 9, along with 1σ and 2σ contour parameters for the BB model. Even using the 1σ lower limits of the inner disk temperature and absorption column of this fit, the luminosity would still be extreme, 5.6 × 10⁴⁰ erg s⁻¹. An NS atmosphere model applied to this spectrum also resulted in a large best-fit luminosity >1 × 10⁴¹ erg s⁻¹.

Zoom In Zoom Out Reset image size

We further investigated if this source could be a foreground object (e.g., a flaring M dwarf) by fitting a thermal plasma model (APEC) to the spectrum. This resulted in a best-fit temperature ∼0.3 keV and a flux of ∼1.1 × 10⁻¹⁴ erg s⁻¹ cm⁻² (corresponding to L_X ∼ 2 × 10³⁸ erg s⁻¹ at the distance of NGC 4697). By assuming that C1 is a star it is possible to calculate the L_X/L_bol ratio by combining the photometric g-band upper limit for this source with the semi-empirical stellar spectral energy distributions of Kraus & Hillenbrand (2007) to provide L_bol, with L_X values scaled from the best-fit X-ray luminosity derived from the APEC model. The resulting value, L_X/L_bol > 0.35, is much higher than values that are typically seen in stars, where in a recent large survey (Wright et al. 2011) L_X/L_bol values were 1 × 10⁻³ with a dispersion of an order of magnitude. Therefore, C1 is unlikely to be a foreground object unless it was observed during an incredibly rare flaring event for which the X-ray luminosity increased by more than an order of magnitude. To determine the likelihood of observing such a flare, we compare to the studies of Feigelson et al. (2004) and Wright et al. (2010), who both studied the properties of stellar X-ray sources in deep, high Galactic latitude fields. From a total of 71 sources and a cumulative exposure of 26.68 Ms they observed only two stars with flaring events with peak amplitudes greater than a factor of 10. From their detection thresholds and the relative exposure times we estimate that we should detect ∼5 stars in the observation of NGC 4697. This equates to a cumulative exposure of 660 ks on these stars, and therefore a probability of observing a significantly large flare in one of these stars of 0.05. This seems highly unlikely and therefore we conclude that the emission from C1 does not arise from a flare star.

C1 was previously discussed in S08 (Section 4.4), who, after determining that a background active galactic nucleus is unlikely, suggest an intermediate mass BH with a mass range of ∼(7 × 10³–10⁶) M_☉, although they note that the formation of such a large BH outside of the nucleus or a GC is a challenge for current BH formation theory.

X-ray luminosities in excess of 10⁴¹ erg s⁻¹ have only been reported for two previous SSS; ULX-1 in M101 (Kong & Di Stefano 2005) and a luminous SSS transient in NGC 4631 (Carpano et al. 2007). Both of these galaxies are late-type star-forming systems, unlike NGC 4697. These sources, when described by a single-component BB model, exhibited similar properties to C1, with soft X-ray spectra (kT_in ∼ 40–150 eV), high intrinsic absorption, and resulting extremely large L_X. However, in both cases, alternate models have been suggested resulting in less extreme luminosities (Mukai et al. 2005; Liu 2009; Soria & Ghosh 2009; Carpano et al. 2007). For C1, even though a single-component model is statistically acceptable, there is a possible excess in the residuals around 0.8 keV (see Figure 9). Following Carpano et al. (2007) an additional Gaussian line was included with the BB model in the fit (we also attempted to use an absorption edge to model the spectrum, as presented in Soria & Ghosh (2009) for the NGC 4631 source, but were unable to constrain the model parameters). The values obtained from the BB+Gaussian model are presented in Table 6 and the spectrum and the 1σ and 2σ contours are shown in Figure 10. The best-fit parameters indicate a lower intrinsic absorption, consistent with the Galactic value within errors, and a temperature of 70 eV (compared to 60 eV in the single-component BB model). The peak energy of the Gaussian component, ∼0.7 keV, is likely to arise from a blend of unresolved emission lines (e.g., O viii; Ness et al. 2005) or there could alternatively be absorption edges, as seen in Soria & Ghosh (2009). The resulting 0.3–8.0 keV luminosity, 3.4 × 10³⁸ erg s⁻¹, although poorly constrained, is three orders of magnitude lower than that from the simple BB model. Although in excess of the emission from a hydrogen burning 1.35 M_☉ WD system, this luminosity could arise from an extreme super Eddington event (a fireball scenario; Soria & Ghosh 2009).

Zoom In Zoom Out Reset image size

Table 6. Best-fitting Parameters of SSS C1

Parameters	Best-fit Values
(1)	(2)
Cstat	52C
Goodness	10%
NH (1021 cm−2)	0.85+0.31−0.85
kTin (eV)	70+33−49
EL (keV)	0.71+0.04−0.09
σL (keV)	0.12+0.04−0.01
LX (1038 erg s−1)	3.40
LX Range (1038 erg s−1)	0.00–3.90

Notes. Presented are values from the tbabs * BB +Gauss model with the quoted errors given to 1σ. This spectral model and subsequent interpretation of the source is presented in Section 3.4.1.

Download table as:  ASCIITypeset image

The 2XMM survey reports a source within <2'' of C1. Given its soft HRs (e.g., HR2 = −0.95 ± 0.08¹⁹) and the lower spatial resolution of XMM-Newton, it is possible that this emission arises from C1. The XMM-Newton observation was taken in 2003 July (compared to 2004 August for the "on" Chandra observation) yielding an X-ray luminosity of 1.30 × 10³⁸ erg s⁻¹, from the 2XMM catalog (compared to 4.72 × 10³⁸ erg s⁻¹ for the model derived from the Chandra observation over the 0.2–12 keV energy range). If this observed emission is from C1, it suggests that the source underwent two outbursts, with a recurrence time of ∼1 year. From simulations Starrfield et al. (2004) estimated that the hydrogen layer involved in the surface nuclear burning would have a mass of <1 × 10⁻⁶ M_☉ and that steady surface hydrogen burning (for a 1.35 M_☉ WD) only occurs for accretion rates below 1 × 10⁻⁶ M_☉ yr⁻¹. If the outbursts of C1 are a consequence of this fireball scenario this implies accretion rates >1 × 10⁻⁶ M_☉ yr⁻¹ and therefore the hydrogen shell could be replenished within one year. In fact, Soria & Ghosh (2009) estimate that the SSS in NGC 4631 has an accretion rate of around 1 × 10⁻⁵ M_☉ yr⁻¹, which could therefore replenish the shell within a month.

However, this interpretation should be treated with caution due to the large uncertainties in the best-fit parameters. Furthermore, the inclusion of the Gaussian component, possibly related to the photospheric expansion period, is speculative, although it serves the purpose of lowering the intrinsic luminosity from the extreme value of >5 × 10⁴¹ erg s⁻¹ to a more typical SSS value of 3.4 × 10³⁸ erg s⁻¹.

QSS are systems with little or no emission above energies of 2 keV, with temperatures between 100 eV and 350 eV (Di Stefano & Kong 2004). QSS are too hot to be WDs, unless there is significant upscattering of photons emitted by the WD or the emission emanates from a limited portion of the surface. Alternative scenarios include QSS as NS or BH binary systems (Di Stefano et al. 2010; and SNRs; Orio 2006). The results of the spectral analysis put B8 and A5 in this category (see Table 4).

Source B8 was observed in one of the cycle 11 Chandra observations and has a maximum outburst time of ∼3 years. Spectral modeling determines a temperature of 280 and 360 eV and X-ray luminosity ∼(1.1–1.5) × 10³⁸ erg s⁻¹ (depending on if a DBB or a BB model is preferred). Such values are consistent with the properties of previously observed QSS.

3.5.1. The Flaring QSS A5

In the case of the flaring source A5, the best-fit DBB model yields inner disk temperature of 220 eV, with intrinsic absorption of ∼3 × 10²¹ cm⁻². Both the temperature and absorption column values from the BB model are lower than the DBB model, although they are consistent within errors, and L_X = 2.5 × 10³⁸ erg s⁻¹. The outburst data (of ∼3 ks; Figure 11) yield similar best-fit parameters to those determined from the full observation. However, due to the reduced statistics it was necessary to fix the absorption column; this was frozen to the Galactic value as well as the best-fit column density of 31 × 10²¹ cm⁻² and 19 × 10²¹ cm⁻² for the DBB and BB models, respectively.

Zoom In Zoom Out Reset image size

Although the average best-fit luminosity of A5 is L_X = 2.5 × 10³⁸ erg s⁻¹, during the flare (see Table 4), the luminosity reaches 2.9 × 10³⁹ erg s⁻¹ for the high absorption column and 9.5 × 10³⁸ erg s⁻¹ for the Galactic absorption value for the DBB model; the corresponding values for the BB model are 1.7 × 10³⁹ erg s⁻¹ and 9.4 × 10³⁸ erg s⁻¹.

This ∼3000 s flaring event is not typical of QSS (e.g., Di Stefano et al. 2004; Liu 2011). Instead it could arise from a flare star and, to investigate this, a thermal plasma model was also fitted to the spectrum (e.g., Wargelin et al. 2008). This model resulted in a best-fit temperature of 1 keV and a flux of ∼5 × 10⁻¹⁴ erg s⁻¹ cm⁻² (corresponding to L_X = ∼1 × 10³⁹ erg s⁻¹ at the distance of NGC 3379). Using the g-band upper limit derived for A5 (see Section 2.1) and following the same method described in Section 3.4.1, L_X/L_bol ∼ 1. The most luminous stellar flares ever observed reach only L_X/L_bol ∼ 0.1 (Güdel et al. 2004) making this an unfeasibly large value for a stellar source and making the possibility that this is a flaring star highly unlikely.

Instead this behavior is suggestive of a superburst from a NS. These events were first observed with BeppoSAX (Cornelisse et al. 2000) and exhibit outbursts of ∼3000–10, 000 s, with luminosities ∼(2–3) × 10³⁸ erg s⁻¹ and energetics ∼1 × 10⁴² erg. Because of the large total energy released from these events, these bursts are thought to be powered by carbon burning instead of hydrogen and helium (see Cumming & Bildsten 2001; Cooper & Narayan 2005 and references therein), although it has also been suggested that this emission could arise from runaway electron capture (Kuulkers et al. 2002).

The spectra of superbursts are well described by a BB model, with temperatures ∼2 keV. For A5, when only the outburst spectrum is fitted, a temperature of ∼200–300 eV is determined. However, as we only model the high flux component of the spectrum, the luminosity of the source is L_X = (1.7–2.9) × 10³⁹ erg s⁻¹ when intrinsic absorption (constrained from the spectrum of the whole observation) is included. When N_H is fixed to the value of Galactic absorption L_X = 9.4 × 10³⁸ erg s⁻¹, with a lower limit of 4 × 10³⁸ erg s⁻¹. From these parameters the fluence of this source could be as high as 9.5 × 10⁴² erg (with a lower limit of 1.3 × 10⁴² erg). This emission is more luminous and energetic than the reported superburst sources, although both of these parameters are consistent within errors. However, the significant difference between A5 and superburst sources arises from the best-fit temperature, where A5 is much cooler. Under the assumption that the source is emitting isotropically a lower limit of R > 5 × 10⁷ cm is derived. This value indicates that this source is larger than a NS and therefore cannot be a superburst.

Instead, given the implied radius (which has a maximum limit of R < 3.4 × 10⁸ cm), this emission could arise from a WD. If such a high fluence event is powered by a thermo-nuclear processes this source could be a helium nova explosion. Such an object was first considered by Kato et al. (1989), where the case of a degenerate WD accreting helium from its helium-rich companion was presented. In this scenario, for accretion rates ∼ × 10⁻⁹ to  × 10⁻⁶ M_☉ yr⁻¹, when sufficient helium has been accumulated along a shell on the surface of the WD, a thermo-nuclear reaction, or helium shell flash, occurs resulting in a nova outburst. In 2000, Kato et al. (2000) reported the discovery of such an object, V445 Puppis. This type of outburst event has also been predicted to emit in the X-ray, although possibly not until the post-outburst stage, before which the wind phase may lead to absorption of the soft X-rays (Iben & Tutukov 1994; Ashok & Banerjee 2003). However, the period of X-ray emission is expected to be similar to that of classical novae, which lasts for periods >months (Pietsch et al. 2007), therefore, an X-ray outburst of ∼3 ks indicates that this emission does not arise from a helium nova.

Another possibility is that this emission arises from a very short period system such as a double WD binary, where these short periods imply that mass transfer is driven by angular momentum loss via gravitational radiation. This would allow for the high X-ray luminosity observed and suggests a mass of 0.4–0.9 M_☉ for the accreting white dwarf in A5 (L_X = (4–9.4) × 10³⁸ erg s⁻¹). Assuming that the mass transfer is driven by gravitational radiation gives donor masses slightly smaller than the accretors. Taking slightly smaller WD mass pairs gives a somewhat smaller average luminosity, allowing for some intrinsic variability, although the cause of this is unclear. A possible complication is that the Eddington luminosity may be depressed below the usual value for this type of accretion, since for CO-dominated compositions the Kramers opacity can exceed Thomson at high temperatures (Peacock et al. 2012). Although "opacity" here refers to the Rosseland mean and is therefore not strictly applicable for considering radiation pressure, a fuller exploration of this type of model would be needed.