THE X-RAY POINT-SOURCE POPULATION OF NGC 1365: THE PUZZLE OF TWO HIGHLY-VARIABLE ULTRALUMINOUS X-RAY SOURCES

Ultraluminous X-ray sources (ULXs) are bright compact off-nuclear point sources detected in nearby galaxies (e.g., Makishima et al. 2000; Fabbiano et al. 2001; Swartz et al. 2004). For sub-Eddington accretion rates in the absence of beaming, their bolometric luminosities of a few ×10³⁹–10⁴⁰ erg s⁻¹ require black hole masses M_BH ∼ 10–400 M_☉, implying the existence of intermediate-mass black holes (IMBHs). Despite increasing solid evidence for the existence of supermassive black holes in sizable bulge galaxies (M_BH > 10⁵ M_☉; e.g., Greene & Ho 2007, and references therein) and the numerous examples of stellar-mass black holes (SMBHs) in our Galaxy and beyond, there are few convincing candidates for black holes with masses in the intermediate regime. The term SMBHs, as used here, refers to black holes produced at the end stage of normal stellar evolution. Observationally, SMBHs are constrained to have masses below ∼20 M_☉ (e.g., Remillard & McClintock 2006, and references therein; see also Fryer & Kalogera 2001; Orosz et al. 2007). Theoretical studies show that compact remnants of up to 60–80 M_☉ could form under special circumstances (assuming solar abundances and initial masses of up to about 200 M_☉; e.g., Heger et al. 2003; Yungelson et al. 2008) and we adopt this lower limit¹ as our definition of an IMBH: a compact remnant with mass between 100 and 10⁵ M_☉. The existence of IMBHs presents challenges in terms of formation and feeding mechanisms leading some researches to prefer alternative explanations for ULX emission (e.g., King 2006; Yungelson et al. 2008).

A simple explanation for the observed high X-ray luminosity of an ULX is a misidentified background active galactic nucleus (AGN), whose chance coincidence with a nearby galaxy and low optical emission contrives to suggest a more unusual object. A recent example is one of the most luminous ULXs found to date in the galaxy MCG-03-34-63 (Miniutti et al. 2006), which was later identified with a background AGN (G. Miniutti 2008, private communication). A more intriguing alternative for the origin of the ULX emission than a background AGN is super-Eddington accretion from a high mass X-ray binary (HMXB) with a SMBH primary. If the emission is beamed (e.g., Poutanen et al. 2007; Körding et al. 2002; Georganopoulos et al. 2002) and the outflow or jet axis points close to out line of sight, emission of up to ∼10⁴⁰ erg s⁻¹ can be explained without the need to invoke IMBHs. Super-Eddington and superluminal accretors have been observed in our Galaxy (e.g., V4641, SS433, GRS 1915+105; Revnivtsev et al. 2002; Begelman et al. 2006; Mirabel & Rodriguez 1994) and according to King (2006) could be responsible for the apparent super-Eddington emission from the majority of ULXs. Sources with X-ray luminosities in excess of 10⁴⁰ erg s⁻¹, however, are much rarer and more likely to host IMBHs. Indeed Swartz et al. (2004) found that the ULX luminosity function for star-forming galaxies requires a broken power-law fit with break luminosity of 10⁴⁰ erg s⁻¹, consistent with the idea of a change in the nature of the ULX population above this luminosity.

Luminous ULXs are more commonly found in star-forming galaxies, leading researchers to suppose that the ULX phenomenon is associated with recent star formation and pointing toward a HMXB origin for the ULX emission. Rappaport et al. (2005) present a detailed analysis of HMXB evolution models together with a binary population synthesis code to study the theoretical expectations for ULX populations of star-forming galaxies under the assumption of a HMXB origin. They conclude that the majority of ULXs with L_X ≲ 10⁴⁰ erg s⁻¹ can be explained in terms of binary systems containing SMBHs. Consequently, from both theoretical and empirical perspective, sources with X-ray luminosities in excess of 10⁴⁰ erg s⁻¹ remain the most puzzling examples of ULX emission.

In this paper we characterize the X-ray point-source population of the giant barred spiral galaxy NGC 1365. We show that it is compatible with the population of high mass X-ray binaries expected for galaxies with the star formation rate of NGC 1365. Komossa & Schulz (1998) discussed the first ULX in NGC 1365, NGC 1365 X1, using ROSAT High Resolution Imager (HRI) data. The source was also noticed by Iyomoto et al. (1997) in an Advanced Satellite for Cosmology and Astrophysics (ASCA) observation. At the time of discovery this was one of the most luminous ULXs known, which, together with the unusual properties of NGC 1365 itself, motivated our subsequent 2002 Chandra follow-up observation. In this paper, we focus our attention on the X-ray flux and spectral variability of the two brightest ULXs in NGC 1365, which reach X-ray luminosities in excess of 10⁴⁰ erg s⁻¹, and are among the most luminous ULXs known to date (for comparison samples, see Makishima et al. 2000; Fabbiano et al. 2001; Swartz et al. 2004; Devi et al. 2007). Adding optical and infrared (IR) data to the detailed X-ray modeling of Chandra and XMM-Newton observations, we evaluate the plausibility of the different emission scenarios for the two ULXs. The paper is organized as follows: In Section 2, we present the X-ray point-source population of NGC 1365. The X-ray spectral and flux variability of NGC 1365 X1, together with the available optical/IR data are analyzed in Section 3. We discuss the X-ray and optical data of the new ULX, NGC 1365 X2, in Section 4, and study in detail the plausibility of the different emission mechanisms for both ULXs in Section 5, followed by a short summary in Section 6. Throughout this paper we assume a luminosity distance of D_L = 21.2 Mpc to NGC 1365, and the standard concordance cosmology of Spergel et al. (2003). All Chandra and XMM-Newton data used in this paper were reprocessed using CIAO version 3.4 and HEASOFT version 6.2 (SAS ver. 7.1) starting with the event 1 files for Chandra and repeating the basic pipeline processing for XMM-Newton followed by the standard calibration and filtering operations in each case. All quoted flux and luminosity uncertainties are 90% confidence limits, unless noted otherwise.

NGC 1365 (R.A., decl. = 03:33:36.40, −36:08:25.7; distance modulus² of 31.6 mag) is a luminous giant barred spiral galaxy with a high star formation rate (SFR ∼12 M_☉ yr⁻¹ for L_FIR = 6.8 × 10¹⁰L_☉; Lonsdale & Helou 1985; Kennicutt 1998). Motivated by the discovery of NGC 1365 X1 with ROSAT, we observed NGC 1365 with Chandra in 2002, in order to follow up the luminous ULX and to characterize the X-ray point-source population of the galaxy. The 2002 Chandra observation shows diffuse extended nuclear emission and 23 off-nuclear point sources³ in the central ∼6' × 6' region. The point sources were detected using the wavdetect algorithm with source significance higher than 3σ (Freeman et al. 2002) as implemented in CIAO version 3.4. Spectra were extracted for each off-nuclear source from a 2'' radius aperture and fit with a simple power-law model assuming no intrinsic absorption above the Galactic value using Cash statistics. The positions, spectroscopic-fit parameters, and model-flux estimates of all off-nuclear point sources are presented in Table 1. Between 2004 and 2006, NGC 1365 was the target of extensive XMM-Newton (PI: G. Fabbiano) and Chandra (PI: G. Risaliti) campaigns aimed at unravelling the highly obscured and variable active galactic nucleus with Chandra and studying the ULX NGC 1365 X1 with XMM-Newton.

The left panel of Figure 1 shows the 2002 Chandra background-subtracted and smoothed image of the inner 6' × 6' (∼36kpc × 36kpc; compare this to the B = 25^m arcsec⁻² isophotal radius of 30 kpc). The cumulative number of X-ray point sources with luminosities greater than a given value are shown as a histogram in the right panel. The X-ray luminosities of local star-forming galaxies are proportional to their SFRs (see Gilfanov et al. 2004, and references therein) as the numbers of HMXBs, which dominate the extended emission in local star-forming galaxies, increases with increasing SFR. Correcting for the SFR, Grimm et al. (2006) have derived a simple analytical expression which predicts the differential number of X-ray point sources N for a given point-source luminosity, L₃₈ (the luminosity in units of 10³⁸ erg s⁻¹): dN/dL₃₈ = 3.3 × SFR × L^−1.61₃₈. We plot the integral form of this relation in the right panel of Figure 1 with a dashed line. The expected number of X-ray point sources in NGC 1365 agrees well with the observed histogram for bright sources (log L_X > 38.5), suggesting that the majority of the X-ray point sources are likely associated with the galaxy. The shaded histogram in the right panel of Figure 1 is a rough estimate of the possible contamination from background AGNs for each luminosity bin, which we compute using the minimum 2–10 keV luminosity for each bin and the log N − log S relation of Hasinger et al. (2001) as detailed in Section 5.1.1. We expect at most 2 of the 16 sources brighter than 10³⁹ erg s⁻¹ in the 6' × 6' area to be background AGNs. We mention in passing that there is no correlation between the measured point-source fluxes and their projected distances from the center of NGC 1365.

Zoom In Zoom Out Reset image size

The positions of the 26 X-ray point sources on an optical image of NGC 1365 are shown in Figure 2. Despite the fact that many of the X-ray point sources coincide with the minor and major spiral arms, only four out of the 26 have an optical counterpart within ∼2''. Three of the four optical counterparts are extended sources, most likely H ii regions, while the remaining southernmost source with an optical counterpart (X15 in Table 1) is a point source. The optical counterpart (in the case of X2) and upper limits (in the case of X1) are discussed in Sections 4 and 3.3, respectively. The remaining three sources with optical counterparts, X17, X15, and X13 (see Table 1) have X-ray fluxes of F_{0.3–10 keV} = 2.9 × 10⁻¹⁴ erg s⁻¹ cm⁻², F_{0.3–10 keV} = 9.8 × 10⁻¹⁴ erg s⁻¹ cm⁻², and F_{0.3–10 keV} = 4.6 × 10⁻¹⁴ erg s⁻¹ cm⁻², and power-law spectral indices of Γ = 1.1 ± 0.6, Γ = 1.5 ± 0.2, and Γ = 1.1 ± 0.5 (assuming absorption equal to the Galactic value). The USNO-B 1.0 (Monet et al. 2003) counterparts of X17 and X15 have R-band magnitudes of R1 = 18.4 and R2 = 19.2 (X17) and R1 = 19.2 (X15).

Zoom In Zoom Out Reset image size

NGC 1365 X1 (03:33:34.61, −36:09:36.6) is the most luminous ULX in NGC 1365. It was discovered by Komossa & Schulz (1998) to be highly variable on long timescales (a factor of 10 within six months), with a peak luminosity reaching ∼5 × 10⁴⁰ erg s⁻¹ in the 2–10 keV band (an ASCA measurement by Iyomoto et al. 1997, updated to D_L = 21.2 Mpc; note that Soria et al. 2007 found a somewhat lower value by reanalyzing the ASCA spectrum). X1 is ∼74'' south from the NGC 1365 nucleus (equivalent to a projected distanc of ≈7.5 kpc) behind an average Galactic column of 1.4 × 10²⁰ cm⁻². It is one of the brightest ULXs known, and one of the few ULXs with extensive X-ray (seven Chandra and five XMM-Newton pointings since 2002, in addition to older ASCA and ROSAT data) as well as extensive optical/IR coverage from the Very Large Telescope (VLT), the Hubble Space Telescope (HST), and the Two Micron All Sky Survey (2MASS). NGC 1365 was the target of a recent Chandra campaign to observe the rapid changes of its obscured active nucleus. Table 2 below gives a summary of these six Chandra pointings as well as the 2002 Chandra and 2003–2004 XMM-Newton (five pointings) observations. We reprocessed all data using CIAO version 3.4.1.1 and HEASOFT version 6.1.2; the effective exposure times of each pointing, after corrections for flaring and chip positions, are listed in the last column of Table 2.

3.1. X1: X-ray Variability

The soft (0.2–5 keV) and hard (5–10 keV) band light curves of X1 are shown in Figure 3, with the right panels focusing on the recent Chandra data. The ULX source underwent a rapid brightening in both soft and hard bands with the softband count rate apparently slightly leading the hard band. After varying by less than a factor of ∼3 (5) in the soft (hard) band over two years, the ULX source flared over the 13 day Chandra observation period, changing by a factor of ∼ 5 (10) in the soft (hard) band. Most of the observed counts in both the Chandra and XMM-Newton exposures are in the soft band below 5 keV. The bottom panels of Figure 3 show the variation of the fraction of hard band photons (using equivalent ACIS-S counts) with time. Note that the source appears to harden during the Chandra flare, but as a whole the contribution of the 5–10 keV emission to the total is small, typically 2%–4%. Despite XMM-Newton's higher effective area above 5 keV (∼ a factor of 3.5 more counts for an EPIC-pn MED filter at the same flux level) there are too few counts in all but the two longest XMM-Newton observations above 5 keV to obtain good constraints on any hardband components. We return to this point in Section 3.2 below.

Zoom In Zoom Out Reset image size

Figure 4 shows the long term 0.5–10 keV X-ray variability of NGC 1365 X1, including the earlier ROSAT and ASCA observations.⁴ The X-ray luminosity of NGC 1365 X1 is typically a few times 10³⁹ ergs s⁻¹, but varies between 2 × 10³⁹ ergs s⁻¹ and (4–6) × 10⁴⁰ ergs s⁻¹, or a factor of ≲30 over an eight year period. Note that this large variation on a timescale of years is coupled with much quicker (on the order of days) variation by a factor of 6 during the 2006 April Chandra flare. We searched within the longest (with non flaring intervals >10 ks) XMM- Newton and Chandra exposures for variability on scales of minutes to hours but found no statistically significant changes (the probabilities of constant rates were larger than 10% in all cases). This is consistent with previous results on samples of ULXs, the majority of which show no short-term variability (e.g., only 5%–15% of Chandra ULXs are variable on short timescales; Swartz et al. 2004).

Zoom In Zoom Out Reset image size

3.2. X1: Spectroscopic Fits

We extracted spectra from 2'' radius (Chandra) or 15'' radius (XMM-Newton) regions⁵ around NGC 1365 X1 and selected background regions free of point sources at similar distance from the NGC 1365 nucleus to minimize contamination. The spectral fits were performed using Cash statistics for individual spectra with less than 300 counts (two Chandra and three XMM-Newton observations marked "C-stat" in Table 3) using 15 count binning for the remaining five Chandra and 20 count binning for the two longest XMM-Newton observations. In order to test specific models of X-ray emission we attempted to fit all spectra with two simple models: (1) power-law fits, including intrinsic absorption; (2) multicolor disk (MCD) plus power-law emission models with absorption equal to the Galactic value. The second model is physically motivated, representing the optically thick thermal X-ray emission from the accretion disk (diskbb model in XSPEC) as well as a harder, corona comptonized component (represented by a simple power-law model in XSPEC). The spectral fits for each of the Chandra and XMM-Newton observations for the two different models are given in Tables 3 and 4, respectively.

Table 3. NGC 1365 X1: Absorbed Power-Law Fits

ObsID (1)	F2–10 keV (2)	F0.5–2 keV (3)	Γ (4)	NH (5)	Data/Method (6)	stat/DoF (7)
3554	6.6+1.5−3.7	7.4+2.2−4.9	2.2 ± 0.2	20 ± 7	ACIS-S/C-stat	80/103
6871	16.5+3.8−1.9	21.0+3.6−2.6	1.9 ± 0.1	11 ± 4	ACIS-S/χ2	31/34
6872	34.0+5.5−3.5	24.6+4.3−2.1	1.9 ± 0.1	20 ± 3	ACIS-S/χ2	36/49
6873	42.7+6.2−4.0	19.0+2.0−3.0	1.6 ± 0.1	13 ± 3	ACIS-S/χ2	50/48
6868	35.4+5.8−6.9	17.0+3.4−4.6	1.6 ± 0.1	14 ± 4	ACIS-S/χ2	36/42
6869	20.1+3.0−4.8	12.4+2.7−3.6	1.8 ± 0.2	13 ± 4	ACIS-S/χ2	18/30
6870	5.1+1.5−1.4	6.1+1.9−1.6	2.0 ± 0.1	13 fixed	ACIS-S/C-stat	101/120
0151370101	4.6+1.3−2.0	1.7+0.6−0.7	1.6 ± 0.2	8 ± 5	PN+MOS/C-stat	300/332
0151370201	1.5+0.8−0.7	3.2+2.2−1.7	1.6 ± 0.2	8 fixed	PN+MOS/C-stat	89/74
0151370701	2.6+0.6−0.6	1.5+0.9−0.6	2.5 ± 0.2	10 fixed	PN+MOS/C-stat	151/183
0205590301	6.9+1.4−0.8	6.1+1.2−0.6	2.06 ± 0.08	10 ± 2	PN+MOS/χ2	144/142
0205590401	7.1+1.1−1.1	6.7+1.0−0.8	2.06 ± 0.08	7 ± 2	PN+MOS/χ2	107/107

Notes. The fluxes are quoted in units of 10⁻¹⁴ erg s⁻¹ cm⁻². All flux errors are 90% confidence limits; all other parameter errors are 1σ errors. (1) Observation ID; (2) absorption corrected flux in the 2–10 keV band; (3) absorption corrected flux in the 0.5–2 keV band; (4) power-law photon index; (5) intrinsic absorption column density in units of 10²⁰ cm⁻²; (6) data and fitting method used; (7) χ²/DoF or C-stat/DoF of the fit.

Download table as:  ASCIITypeset image

Table 4. NGC 1365 X1: MCD-Power-Law Fits

ObsID (1)	Fdiskbb (2)	F0.5–2 keV (3)	F2–10 keV (4)	kT (5)	K (6)	N (7)	stat/DoF (8)
3554	4.6+0.3−1.2	4.2+0.2−2.6	9.3+0.3−1.0	0.63 ± 0.15	0.014 ± 0.011	6.5E-6	83/106
6871	20+1.2−2.4	12+0.5−2.5	19+1.2−2.6	0.86 ± 0.07	0.017 ± 0.005	8.9E-6	31/35
6872	25+1.0−2.2	14+0.9−4.1	35+1.1−3.7	1.13 ± 0.10	0.0072 ± 0.026	17E-6	49/50
6873	23+1.0−4.0	14+0.8−5.0	41+0.9−7.0	1.17 ± 0.13	0.0054 ± 0.0027	23E-6	49/49
6868	22+0.9−6.2	12+1.0−6.2	34+1.1−9.6	1.22 ± 0.13	0.0046 ± 0.023	17E-6	34/43
6869	14+0.9−3.0	8.8+0.5−3.1	20+0.7−3.1	0.97 ± 0.11	0.0072 ± 0.0033	11E-6	18/31
6870	6.5+0.5−3.7	3.4+0.5−2.8	5.6+0.5−2.7	0.95 ± 0.13	0.0036 ± 0.0025	2.2E-6	105/120
0151370101	2.6+0.3−1.0	1.8+0.3−1.1	4.6+0.3−1.1	1.02 ± 0.20	0.0012 ± 0.0014	2.7E-6	297/333
0151370201	1.8+0.3−0.7	1.2+0.3−0.8	3.3+0.3−0.7	1.04 ± 0.52	0.0007 ± 0.0023	1.9E-6	88/74
0151370701	2.8+0.3−1.0	2.0+0.2−1.4	2.0+0.2−0.3	0.46 ± 0.07	0.028 ± 0.017	1.3E-6	151/183
0205590301	6.0+1.0−1.2	4.5+0.2−0.3	7.3+1.9−2.1	0.65 ± 0.04	0.015 ± 0.003	4.4E-6	151/142
0205590401	6.7+0.5−0.6	5.3+0.2−0.4	8.1+0.2−0.3	0.52 ± 0.03	0.041 ± 0.008	5.5E-6	99/108

Notes. The fluxes are quoted in units of 10⁻¹⁴ erg s⁻¹ cm⁻². (1) Observation ID; (2) flux of the MCD component in the 0.0001–10 keV band; (3) total flux in the 2–10 keV band; (4) Total flux in the 0.5–2 keV band; (5) maximum MCD-fit temperature in keV; (6) normalization of the MCD component, K = (R_in/D)²cos θ, where R_in is the apparent inner accretion-disk radius in km, D is the distance to the source in units of 10 kpc and θ is the inclination of the disk; see Kubota et al. (1998); (7) 1 keV normalization of the Γ = 1 power-law component in units of photons keV⁻¹ cm⁻² s⁻¹; (8) χ²/DoF or C-stat/DoF of the fit; the type of the fit for each observation (χ² or C-stat) is the same as that given in Table 3. We assume no intrinsic absorption above the Galactic value of 1.4 × 10²⁰ cm⁻² for all fits. All flux errors are 90% confidence limits; all other parameter errors are 1σ errors.

Download table as:  ASCIITypeset image

The addition of an intrinsic absorber substantially improved the power-law model fits for 9 of the 12 observations shown in Table 3. For the remaining three observations with limited photon statistics, we fix the intrinsic absorption to the value estimated for the well constrained observation closest in time. The variability of the spectral parameters for the power-law fits are shown in Figure 5. The required intrinsic absorption, N_H,i, ranges between (7 ± 2) and (21 ± 3) × 10²⁰ cm⁻², a variation by a factor of ≲3. Similar increase in the absorbing column with increasing flux was observed in the HMXB Cygnus-X3 with INTEGRAL (Hjalmarsdotter et al. 2004). The estimated power-law photon indices vary between 1.6 ± 0.1 and 2.5 ± 0.2, with the majority of fits having Γ = 1.9–2. From Figure 5, the spectral index appears to flatten with increasing flux during the Chandra flare of April 2006 (indicating a harder spectrum) while the intrinsic absorption appears to remain constant over the same 13 day period. Note that the spectral parameter variation in general needs to be interpreted with caution, due to the known degeneracy between Γ and N_H,i in power-law fits, as well as the lower sensitivity of Chandra at high energies (see also the discussion in Appendix A3 of Portegies Zwart et al. 2004).

Zoom In Zoom Out Reset image size

Figures 6 and 7 show the Chandra and XMM-Newton fits for the second model, representing the direct accretion disk emission plus corona-comptonized component. Statistically these fits are indistinguishable from the absorbed power-law fits presented above (compare the last columns of Tables 3 and 4). In fact if we only consider the 0.3–5 keV emission, the MCD fits are adequate and no power-law component is necessary in any of the 12 observations. Considering the full 0.3–10 keV band, the MCD model alone provides good fits to all but the three longest XMM-Newton observations, for which the addition of a power-law component improves the fit with higher than 99% confidence. As noted in Section 3.1 the Chandra and (to a lesser extent) XMM-Newton effective areas result in insufficient counts above 5 keV for all but the two longest XMM-Newton exposures. The additional power-law component required by these spectra has a photon index of Γ ≈ 1. Fitting the longest exposure, XMM-Newton 0205590301, we obtain the normalization of this additional power-law component and proceed to include a Γ = 1 power-law component in all MCD fits, scaling the normalization of the power law according to the observed 5–10 keV count rate.⁶ The resulting MCD plus power-law fit parameters are listed in Table 4 and their variability with time and X-ray flux is shown in Figure 8. These fits allow us to estimate both the maximum color temperature of the accretion disk and the total luminosity emitted by the disk and to compare these to IMBH and supercritically accreting SMBH models in Section 5.2. In this model, the accretion-disk emission contributes 30–60% of the total flux in the 0.3–10 keV band and the maximum color temperature increases with increasing flux.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Stobbart et al. (2006), who studied the highest signal-to-noise (S/N) XMM-Newton spectra of 13 ULXs, concluded that neither simple power-law models nor simple MCD models provide adequate fits to the highest S/N spectra of ULXs. Better representations of their data are found using cool (∼0.2 keV) or warm (∼1.7 keV) disk plus power-law models, combination of cool blackbody plus warm MCD ("dual thermal") models (bbody + diskpn in XSPEC), or MCD plus comptonized-corona models (diskpn+eqpair in XSPEC). The authors caution that depending on the choice of model, the inferred disk temperatures of their sample would cluster either at the "cool" or "warm" end of a luminosity–temperature diagram, and be interpreted as evidence of the presence of IMBHs (if cool) or SMBHs (warm). Even the best spectra (e.g., over 10⁴ counts with XMM-Newton) were inadequate to distinguish between these models. Thus, disk-temperature measurements based on spectra with less than a few thousand counts could give misleading results, especially when using the "wrong" spectral model. In addition, Stobbart et al. (2006) suggest that the evidence for a distinct curvature in the 2–10 keV spectra of ULXs (seen as a break in the power-law emission in 8/13 of their sources) can be used as a red flag against the interpretation of the hard band spectrum as emission from an optically thin corona. A broken power-law model above 2 keV is not required by any of our individual observations nor by the joint fit to the highest S/N XMM-Newton and/or Chandra spectra of X1. Only the two longest XMM-Newton observations of X1, however, have ≳ 2000-counts spectra, which could be responsible for the nondetection of curvature above 2 keV. We therefore caution that the interpretation of the MCD plus power-law model fits of X1 presented in Section 5.2 depends on the assumption that the hard band spectrum is well represented by a simple power law arising in an optically thin corona, while the soft emission arises directly from the accretion disk.

3.3. X1: Optical/IR Observations

X1, whose X-ray position is known to better than 1'' based on the Chandra images, was not detected on any of the optical or IR images available. Nevertheless, we use these images to obtain sensitive limits to the R, I, J, H, and K Johnson–Cousins band magnitudes of the ULX counterpart. The details of the lower-limit estimates are presented below and summarized in Table 5. We correct all optical and IR images astrometrically using the USNO-B catalog stars (Monet et al. 2003) in the immediate vicinity of the ULX source position.

Table 5. NGC 1365 X1: Optical and IR Limits

Telescope (1)	Filter (2)	Texp (3)	Aperture (4)	MJD (5)	ULX Limit (6)	east star (7)	west star (8)
VLT FORS-2	R-special	30	15	52736	R > 23.3	18.5 ± 0.3	19.4 ± 0.3
HST WFPC2	F814W	10	08	49732	I > 21.4	16.81	17.56
HST WFPC2	F814W	10	08	49732	I > 21.6	16.96 ± 0.03	17.71 ± 0.05
HST WFPC2	F555W	100	08	51220	V > 24.1	19.8 ± 0.1	20.4 ± 0.1
2MASS	J	...	30	51602	J > 17.2	15.18 ± 0.07	16.32 ± 0.17
2MASS	H	...	30	51602	H > 16.2	14.56 ± 0.06	15.79 ± 0.21
2MASS	K	...	30	51602	K > 15.9	14.50 ± 0.10	15.32 ± 0.20

Notes. (1) and (2): telescope/instument configuration and filter used in each observation; (3) exposure time, T_exp, in seconds; (4) aperture radius used to measure source fluxes; the background was measured locally within an aperture with twice the area of the source aperture; (5) modified Julian Date (MJD) of the observation; (6) ULX limit in the standard Johnson–Cousins filters; (7) and (8): magnitudes of the two stars (to the east and west of the ULX position, respectively; see Figure 9) used for magnitude calibration in each Johnson–Cousins filter.

Download table as:  ASCIITypeset image

The left panel of Figure 9 shows a 30 s VLT FORS2 (Very Large Telescope Focal Reducer and low-dispersion Spectrograph) R-Special filter⁷ observation of NGC 1365 within 50'' of the X-ray position of X1. Smaller region cutouts around the X1 X-ray position in two of the HST filters are shown in the right panel of Figure 9. The ULX X-ray position coincides with a faint spiral extension; two faint sources are seen 37 south and 4'' northeast of the X-ray position, respectively. Considering the large separations, neither of these sources is a likely optical counterpart, but the optical counterpart of the ULX was clearly fainter than both sources during the observation. The two bright stars ∼16'' east and ∼13'' west of the ULX in Figure 9 have USNO-B 1.0 R-band magnitudes of 18.5 ± 0.3 and 19.4 ± 0.3, respectively. Judging by the flux ratios between those stars and the faint sources ∼4'' south and northeast from the ULX X-ray position, the faint sources have R-band magnitudes of ∼22.4 and ∼22.3, respectively. We chose a local background region a few arcsec northwest of NGC 1365 X1 (but still within the diffuse spiral arm emission) and aperture radii of 15 for all flux measurements. The R-band magnitude limit for the ULX emission is R > 23.3.

Zoom In Zoom Out Reset image size

Two HST WFPC2 observations of the central region of NGC 1365 (see the right panel of Figure 9) were taken in 1995 with the F814W (10 s) and F555W filters (100 s). The HST F814W filter is close to the Johnson–Cousins I-band. Using the I magnitudes of the same two bright stars (east and west of the ULX source position in Figure 9) as photometric calibrators, we estimate a lower limit to the apparent magnitude at the ULX position of I > 21.4 (see the second row of Table 5). If we use the independent HST WFPC2 photometric calibration,⁸ we obtain the values given in the third row of Table 5, and a ULX source position lower limit of I > 21.6, in agreement with the lower apparent-magnitude limit obtained using bright-star calibration. The lower limit to the V-band apparent magnitude at the ULX position given in the fourth row of Table 5 was estimated in similar fashion using the 100 s F555W-filter HST exposure.

2MASS coverage of the region around the ULX source position is also available. We use relative photometry (using the same bright stars east and west of the source) to estimate the H, J, and K band lower limits. The results, using 3'' radius source aperture measurements, appear in the last three rows of Table 5. We comment on the implications of all optical/IR apparent magnitude limit results in Section 5.1.2.

NGC 1365 X2 (03:33:41.85, −36:07:31.4) is a highly variable X-ray point source in the north tip of the eastern spiral arm of NGC 1365 (see Figure 2). Based on the highest luminosity reached during one of the 2006 April Chandra observations, which exceeds 4 × 10⁴⁰ erg s⁻¹ in the 0.5–10 keV band, the object is among the most luminous ULXs known.

We extracted spectra for the X2 source from all Chandra observations as described in Section 3.2 for NGC 1365 X1. Because of the proximity of X2 to other X-ray point sources (the closest one is only 9'' south of X2, the second closest 20'' west) combined with the larger XMM-Newton point spread function and the faintness of X2 during the XMM-Newton observations, we do not attempt to extract spectra/upper limits from these exposures. We attempted to fit the Chandra spectra with both power-law and simple MCD models. Assuming Galactic absorption, we find the simple power-law or MCD fits are unacceptable and intrinsic absorption is necessary in both cases. The absorbed power-law fits are summarized in Table 6. Due to the small number of counts, we use the Cash-statistic fit for all but one of the spectral fits—the highest S/N spectrum (Obs.ID. 6870), which has 535 counts in the 0.3–10 keV band, was binned to 15 counts per bin and fit using χ² statistic. We show the power-law fit and the 3σ confidence contours for the photon index and the column density of the intrinsic absorber for the highest S/N fit of X2 in the top panel of Figure 10. For all spectra with less than ∼200 counts in the 0.3–10 keV band, we fix the power-law spectral index to Γ = 1.3, the value obtained for the two spectra with the largest numbers of counts. A cold absorber of N_H,i = 6–9 × 10²¹ cm⁻² is required by all fits, including the MCD fits.

Zoom In Zoom Out Reset image size

The absorbed MCD fits are statistically indistinguishable from the absorbed power-law fits presented above. For all April 2006 Chandra observations the maximum color temperature estimated from the MCD fits was around 2 keV, with large uncertainties (1σ uncertainties ∼0.3–0.6 keV). The 3σ confidence contours for two interesting parameters for the highest S/N MCD fit are shown in the middle panel of Figure 10. The addition of a comptonized component (e.g., a power law) does not improve the fits. We caution that, considering the small number of total counts and even smaller number of counts above 5 keV, it is conceivable that the absorbed MCD model is inappropriate, and the disk-temperature constraints are unreliable. ULXs and X-ray binaries whose MCD fits result in unusually high disk temperatures similar to those obtained for X2 have been known since Makishima et al. (2000). Makishima et al. (2000) suggested that the unusually high temperatures are naturally explained by emission from a disk around a rotating black hole. The bottom panel of Figure 10 shows a multitemperature blackbody model fit to the highest S/N X2 spectrum which assumes a thin, steady-state, general relativistic accretion disk around a rotating (Kerr) black hole (Li et al. 2005). Both the MCD and the kerrbb fits shown in Figure 10 are acceptable (χ²/DoF = 25/31 and χ²/DoF = 24/32, respectively) and we comment on the results in Section 5.2.2.

X2 falls within a 1995 ROSAT HRI observation. Singh (1999) claimed a 3.5σ detection with 0.0020 ± 0.0006 cps (F_{0.5–10 keV} ∼ 1.3 × 10⁻¹² erg s⁻¹ cm⁻² assuming a power law model with Γ = 1.3 and N_H,i = 9 × 10²¹ cm⁻²). We reanalyzed the ROSAT HRI observation and found only a very faint source with 0.00053 ± 0.00026 cps at that position (∼2σ); we use this as an upper limit to the X-ray count rate of X2 at this epoch, arriving at an obscuration corrected flux of F_{0.5–10 keV} < 3.5 × 10⁻¹³ erg s⁻¹ cm⁻² for a power-law model with Γ = 1.3 and N_H,i = 9 × 10²¹ cm⁻². X2 is one of the three sources that were not detected during the 2002 Chandra observation with an upper limit of CR<0.0004 cps or F_{0.5–10 keV} < 0.9 × 10⁻¹⁴ erg s⁻¹ cm⁻².

Super or hypernovae can reach X-ray luminosities of over 10⁴⁰ erg s⁻¹, but they are expected to be steady sources on short timescales and to decline in luminosity slowly over periods of months to years. The strong variability of both NGC 1365 X1 and NGC 1365 X2 over extended periods of time (factors of 10 over years) and their short time flares argue against supernova emission and imply a compact accreting source in both cases. In addition, neither of the two ULXs discussed here has the thermal emission-line dominated spectrum from shock-heated plasmas which is characteristic of supernovae. Below we discuss the IMBH, super-Eddington SMBH, and background AGN possibilities for the two NGC 1365 ULXs, focusing on their spectral characteristics, luminosity variability, and optical/IR-to-X-ray ratios.

5.1. Background AGNs?

We argue that both NGC 1365 X1 and NGC 1365 X2 are unlikely to be background AGNs. Based on their positions −X1 coincides with one of the diffuse weak spiral arms south of the NGC 1365 nucleus, while X2 coincides with an H ii region at the tip of the main northeast spiral arm, as well as the negligible number of background sources as bright as these ULXs expected for a similar-sized random patch of sky, X1's unusual X-ray-to-optical/IR ratios and its small intrinsic X-ray absorbing column, an IMBH or a supercritically accreting SMBH are more likely in both cases. We cannot rule out that X1, which has no optical counterpart, is not an optically/IR obscured AGN with little X-ray obscuration with certainty, but the association of NGC 1365 X2 with the H ii region in the northeast arm of the galaxy makes it extremely unlikely that the source is a background AGN.

5.1.1. logN − logS Argument

According to the log N − log S Lockmann Hole results of Hasinger et al. (2001), we expect less than 0.2 sources per square degree with F_{0.5–10 keV} > (5–6) × 10⁻¹³ ergs s⁻¹. Consequently, less than 2 × 10⁻⁴ sources with flux similar or higher than that of the maximum flux of X1 or X2 are expected within 2' of the center of NGC 1365. This, together with the fact that X1 is located in one of the faint spiral-arm extensions of NGC 1365 and X2 coincides with an H ii region at the tip of the NE arm, makes it unlikely that either of the sources is a background AGN.

5.1.2. X-ray-to-Optical/IR Ratios

NGC 1365 X1

As noted in Section 3.3, any optical counterpart of the X1 source is fainter than R > 23.3. This limit is consistent with any stellar mass companion, including the brightest supergiants. Computing the X-ray-to-optical ratio, log [X/O] = log F_X + 0.4R + 5.61, we find lower limits between 1.5 and 2.7 for the full range of X-ray fluxes observed for X1. Typical X-ray-to-optical ratios for AGNs range between −1 < log [X/O] < 1 (e.g., Maccacaro et al. 1988; Hornschemeier et al. 2001), with the upper range corresponding to blazars. About 90% of all AGNs have log [X/O] < 2. Since the observed lower limit to the X-ray-to-optical ratio for X1 is smaller than this value for only 3 of the 12 X-ray observations, we consider it highly unlikely that this ULX is a background AGN. For comparison, the eight ULXs which were found to be background AGNs (or a narrow line galaxy in one case) by Gutiérrez & López-Corredoira (2007) have 0.04 < log [X/O] < 0.6. The two additional X-ray point sources from Table 1 with optical counterparts in the USNO-B 1.0 catalog have log [X/O] ∼ −0.4 (X17) and log [X/O] ∼ 0.3 (X15). X15, whose optical counterpart is a point-source, is likely a background AGN.

We estimated a limit of I > 21.6 (F_I < 6 μJy at 0.9 μm) for the I-band magnitude of NGC 1365 X1. The X-ray-to-I-band ratio, log [X/I] = log [F_{0.5–2 keV}/F_I], has a lower limit ranging between −0.1 and 1.7 for the full range of observed 0.5–2 keV fluxes, with a most likely lower limit between 0.4 and 0.7, indicating that the source could be a very rare optically faint/X-ray bright background AGN at a redshift higher than 2 (see Figure 4 of Brandt & Hasinger 2005). If that was the case, however, the measured flux would imply an X-ray luminosity of ∼10⁴⁶ erg s⁻¹. According to the X-ray luminosity function of Hasinger et al. (2005) type 1 (unobscured) AGNs with such high luminosities are rare: ∼10⁻⁹ Mpc⁻³. The analysis of the X-ray background (e.g., Gilli et al. 2007) further shows that the ratio of obscured to unobscured AGNs is ≲1 for luminous AGNs, implying that there are ≲200 AGNs in total of comparable luminosities between redshifts of 1.9 and 2.1 in the whole universe, which together with the large variability of X1, uncharacteristic for luminous AGNs, makes it very unlikely that X1 is a luminous redshift higher than 2 AGN.

Using the 2MASS observations, the K_S > 15.9 limit of X1 corresponds to $F_{K_{{{{\rm S}}}}}<0.29$ mJy at 2.2 μm, indicating a lower limit of the X-ray-to-K_S flux ratio of ∼ 0.2 (ranging between 0.04 and 0.6 for the observed X-ray range). Wilkes et al. (2002) studied the X-ray properties of red 2MASS AGNs. According to their Figure 1, typical red 2MASS AGNs have X-ray-to-K_S flux ratios less than 0.05, while the z < 1 broad-line AGNs form Elvis et al. (1994) have X-ray-to-K_S flux ratios between 0.05 and 0.7. The lower limits to the X-ray-to-K_S flux ratio for NGC 1365 X1 is consistent with an unobscured broad-line AGN, but this interpretation is inconsistent with the observed optical limits.

NGC 1365 X2

In the optical image, X2 is associated with a bright extended emission region in the eastern spiral arm of NGC 1365 (see Figure 11), 86'' south from the NGC 1365 nucleus (equivalent to ≈8.8 kpc), which was classified as an H ii region by Hodge (1969). Using the USNO-B 1.0 catalog's R-band magnitudes of the optical counterpart, R1 = 13.25 and R2 = 13.75, we estimate an X-ray-to-optical ratio of −2 < log [X/O] < −1, depending on the X-ray flux level. The H ii region has been studied in detail by Roy & Walsh (1997) and Ravindranath & Prabhu (2001). It is a typical large star-forming region with a diameter of 670 pc, M_V = −12.4, and an Hα line luminosity of 1.6 × 10³⁹ erg s⁻¹ (S. Ravindranath 2008, private communication). Roy & Walsh (1997) studied the optical spectrum of the H ii region (number 33 in their Table 2) in comparison with other H ii regions in NGC 1365. The observed narrow-line ratios suggest an ionization-bound nebula which is photoionized by massive stars. There is no evidence that the X-ray emission of NGC 1365 X2 affects the observed line ratios, which is consistent with its interpretation as an X-ray binary (whose energy input would be too small to affect the observed narrow line-ratios of a giant nebula).

Zoom In Zoom Out Reset image size

According to the X-ray observations, X2 contributes about 5 × 10⁴⁹ ionizing photons per second, and the Hα luminosity emitted as a consequence of this ionizing flux should be about 5 × 10³⁷ erg s⁻¹ (see Section 5.3 of Soria et al. 2005, and references therein for the details of the computation). This is a factor of 30 fainter than the measured Hα line luminosity of the H ii region, confirming the conclusion that X2 does not contribute substantially to the photoionization of the surrounding nebula.

5.2. L-T Diagram Results

5.2.1. NGC 1365 X1

Panel A of Figure 12 shows the luminosity–temperature (L–T) diagram for NGC 1365 X1. The estimated MCD temperatures and accretion disk luminosities follow a L ∝ T^α relation, with α = 2.8^+1.1_−0.6. Panel B of Figure 12 displays the theoretical tracks for 20, 40, and 60 M_☉ black hole primaries, accreting at sub-Eddington (colored red) or super-Eddington rates (colored blue; Poutanen et al. 2007). The X1 relation agrees with the L ∝ T⁴ relation expected for a black hole primary with mass between 40 M_☉ and 60 M_☉, but the observed maximum luminosities during the Chandra flare of 2006 April (L_X > 10⁴⁰ erg s⁻¹) require super-Eddington accretion rates and/or beaming. In the Poutanen et al. (2007) model, the observed luminosities can reach $L \approx L_{{{{\rm Edd}}}}(1+0.6\ln {\textrm{\.m}})$ where $\textrm{\.m}$ is the accretion rate in units of Eddington.⁹ The maximum color temperature of the disk is $T_{{{{\rm c,max}}}}=f_{{{{\rm c}}}}\times 1.6\times (m^{-1/4})\times (1-0.2\textrm{\.m}^{-1/3})$ keV, where m is the black hole mass in units of M_☉, and f_c is the color-correction factor, f_c = 1.7 (T_max is the maximum temperature of the advective disk; see Equation (36) of Poutanen et al. 2007). A 40–60 M_☉ black hole primary is consistent with all XMM-Newton and Chandra points in Figure 12 within 3σ of their estimated uncertainties if the accretion rate varies between 0.1 and 16 of the Eddington rate in the Poutanen et al. (2007) model. The measured temperatures during the Chandra flare, are systematically high relative to the super-Eddington 60 M_☉ model (by about 30% or ∼3σ). A variation of the color-correction factor with accretion rate for 1< $\textrm{\.m}$ <16 could account for this discrepancy.

Zoom In Zoom Out Reset image size

We note that a less massive primary is problematic not only because it fails to explain the observed temperature–luminosity pairs in Figure 12, which could, as noted in Section 3.2, be affected by the specific model fit assumptions. A 10 M_☉ black hole primary, for example, is also problematic because the required beaming factors or super-Eddington accretion rates (a beaming of over a factor of 30 or, alternatively, higher than 10⁻⁶M_☉ available for accretion in ∼ 2 days during the Chandra flare in the Poutanen et al. 2007 model) are uncomfortably large.

We conclude that the L–T diagram of X1 is consistent with the expectations for a 40–60 M_☉ black hole primary accreting at 0.1 <$\textrm{\.m}$ < 16 from a massive companion star, offering no support for the IMBH hypothesis.

5.2.2. NGC 1365 X2

As described in Section 4, the X2 spectra are consistent with absorbed power-law, or absorbed MCD fits, and the current data are of insufficient quality to distinguish between these or more complex models (e.g., models including a combination of a MCD and a comptonized component). The MCD fits to the X2 spectra are inadequate to constrain the maximum color temperature and inner radius of emission for each individual observation and assuming a simple MCD model could lead to unreliable temperature estimates. As a result, the following interpretation of the results depends critically on the assumption that X2 is in a disk-dominated state and the MCD or kerrbb models are appropriate.

Assuming that X2 is in a disk dominated state and the MCD model is appropriate, the disk temperature estimates are too high for a sub-Eddington accretion around a nonrotating black hole (T_c,max < 1.7 keV). The Poutanen et al. (2007) model with a SMBH primary can reproduce the observed disk temperatures for M < 5 M_☉, but cannot reproduce the maximum luminosity (a factor of ≳ 50 higher), since in that model the luminosity exceeds Eddington by a factor of $1+0.6\ln {\textrm{\.m}}$, which is less than a factor of 5 for $\textrm{\.m}<10^3$. Other supercritical accretion models (e.g., Ohsuga et al. 2005) also predict luminosities which exceed the Eddington by factors of a few, up to 10–15. If we would believe the high accretion disk temperatures, Makishima et al. (2000) have suggested a solution—a rotating black hole would naturally explain the higher disk temperatures. For X2 a maximally rotating black hole (a = 0.998) with disk inclination fixed to θ = 85 (close to edge on), we obtain M = 210 ± 70 M_☉ with an accretion rate $\textrm{\.M}=(1.8\pm 0.3)\times 10^{20}$ g s⁻¹ ($\textrm{\.m}\approx 0.4\hbox{--}0.5$), and an intrinsic absorption of N_H,i = 6 ± 2 × 10²¹ cm⁻². The bolometric luminosity of this model is ∼(1–2) × 10⁴⁰ erg s⁻¹, similar to the kerrbb-fit value of 3.5^+0.4_−1.0 × 10⁴⁰ erg s⁻¹. Consequently the X2 spectra are consistent with emission from a disk around a maximally rotating black hole, for a large range of IMBH masses, M ∼ 80–500 M_☉. The S/N however is not sufficient to distinguish those models from simple absorbed power laws, and the evidence provided by this spectroscopic fit is only suggestive.

5.3. Implications of the X-ray Variability

The X-ray spectra of classical X-ray binaries initially get softer as their flux increases, which is not observed for either NGC 1365 X1 or X2. If anything, the X-ray spectrum of NGC 1365 X1 gets harder in the higher flux states. Other ULXs are known to show this behavior on shorter timescales which was suggested to arise in cool-accretion disk plus optically thick corona models of ULX emission as the corona is heated in the high flux states (e.g., NGC 5204 X1; Roberts et al. 2006). Soria et al. (2007) interpret the hardening with increased flux as the effect of an additional ionized-absorption component—a flux decrease in the 0.5–1.5 keV region which looks like a continuum hardening in power-law fits with neutral absorption.

From Figure 4, NGC 1365 X1 appears to be persistently luminous in the X-ray band (typically a few times 10³⁹ erg s⁻¹) over decades with occasional short (∼ days), but not very powerful (≲ a factor of 10) flares. NGC 1365 X2, on the other hand, has more transient behavior—it was not detected before 2006 April, when it suddenly reached unobscured luminosities of a few times 10⁴⁰ erg s⁻¹, varying on a timescale of days (Figure 13). Assuming either a SMBH or an IMBH binary, the persistent high state of X1 can be interpreted as evidence of a massive donor star (≳ 20–30 M_☉) supplying a large, stationary, but not necessarily steady (considering the Chandra flare on timescale of days) accretion disk. As noted in Section 5.2 above, the 2006 Chandra flare of X1 is too small to be associated with a disk instability, as observed in SXTs in LMXBs, which result in accretion rate and luminosity increases of factors of ∼10⁴.

Zoom In Zoom Out Reset image size

Kalogera et al. (2004) studied the long term transient behavior of SMBHs and IMBHs due to thermal-viscous disk instability for ULXs in young stellar populations. For the most likely donors in such environments, with masses in excess of 5 M_☉, they inferred a minimum black hole mass for transient behavior, which is ≳50 M_☉. Based on this argument and the observed variability patterns of the two ULXs, it is more likely that NGC 1365 X2, which is a transient source associated with an H ii region, contains an IMBH.

5.4. Free-Floating Supermassive Black Hole?

We present 26 X-ray point sources detected in a series of Chandra observations of the giant spiral galaxy NGC 1365. The cumulative distribution of sources brighter than a given X-ray flux is compatible with predictions based on the star formation rate of the galaxy, as expected if high-mass X-ray binaries dominate the extended X-ray emission of star-forming galaxies (e.g., Gilfanov et al. 2004; Grimm et al. 2006).

Two of the X-ray point sources, NGC 1365 X1 and NGC 1365 X2, are highly ultraluminous, with peak luminosities exceeding (3–5) × 10⁴⁰ erg s⁻¹. Both ULXs are most likely accreting compact objects (with SMBH or IMBH primaries), since both the super/hypernova origin and the background or "free floating" AGN interpretations appear highly unlikely. Both sources are among the most luminous ULXs known and as such, among the most likely candidates for IMBH binaries. Their luminosities and optical counterparts (a lower limit for X1 and a giant H ii region for X2) are consistent with an accreting IMBH in both cases; assuming no beaming and accretion rate equal to the Eddington rate (computed from the highest Chandra luminosities), the black hole masses are ∼250 M_☉ in the case of X1 and ∼340 M_☉ in the case of X2.

Remillard & McClintock (2006) defined three outburst states of SMBHs, based on data of Galactic black hole binaries: (1) the thermal or accretion-disk dominated state (with disk fraction higher than 80%; previously known as the High/Soft state), (2) the hard (1.4 < Γ < 2.1) or power-law dominated state (the Low/Hard state), and (3) the steep power law (SPL) state, in which both disk emission (typically less than 50%, but up to 80%) and a steep power law (Γ>2.4) contribute (the Very high state). Active nuclei with supermassive black holes are typically observed in the hard state, in which case a radio jet is often present in the case of X-ray binaries as well as low-luminosity AGNs. It is not clear whether accreting IMBHs will exhibit the three states observed in Galactic BH binaries.

The observed X-ray spectra for X1 and X2 are consistent with power-law models including intrinsic absorption, as well as thermal accretion-disk (plus a power law in the case of X1, which contributes about 50% of the flux) emission models. The 12 Chandra and XMM-Newton spectra of X1 are equally well fit by power-law emission modified by intrinsic neutral absorption and a multicolor disk (MCD) emission plus about equal contribution from a comptonizing corona (reminiscent of the SPL/Very high state). The MCD fits to the X-ray spectra of X1 imply maximum color temperatures which are too high for the black hole masses inferred from the maximum blackbody luminosity (∼100 M_☉), if we assume isotropic emission and sub-Eddington accretion rates, suggesting that beaming and supercritical accretion also play a role. Assuming that the MCD plus power-law fits provide an accurate representation of the spectra, a black hole mass primary in the 40–60 M_☉ range is consistent with the observed L–T diagram for X1.

In the case of X2, the observed X-ray spectra are consistent with intrinsically absorbed power law, MCD, and maximally-rotating IMBH fits and the insufficient photon statistics prevents us from obtaining reliable accretion-disk temperature estimates. The MCD temperature estimates obtained (∼2 keV) are unrealistically high for even the supercritically accreting SMBH models, suggesting that the simple power-law model or the rotating black hole model provides more viable explanations for the observed X-ray spectra. A rapidly-rotating (a ∼ 0.998) IMBH with M = 80–500 M_☉ is consistent with the highest S/N Chandra spectrum of X2 and can reproduce the observed luminosities. An IMBH interpretation for X2 is also consistent with the theoretical expectations of Kalogera et al. (2004), who inferred a minimum black hole mass for transient behavior of M > 50M_☉ for a binary with a donor star more massive than 5M_☉.

Absorbed power-law models provide equally good fits to the X-ray spectra of X1 and X2. The results of the power-law model fits for X1 and X2 are reminiscent of the hard (formally, Low/Hard) state of Galactic black hole binaries. The power-law slopes deduced for X1 (1.6 < Γ < 2.2) and X2 (Γ = 1.3 ± 0.2) are consistent with those of Galactic black hole binaries in the hard state, where the comptonized component dominates; the observed luminosities are however very high, suggesting that super-Eddington accretion modes and or beaming/collimation also play a role. The maximum luminosities reached ((3–5) × 10⁴⁰ erg s⁻¹) are factors of 10–20 higher than those of the ≲20 M_☉ black hole X-ray binaries studied in our Galaxy, pushing the limits of the current supercritical accretion models and requiring very high beaming/collimation.

The most likely black hole masses are probably around 40–60 M_☉ in the case of X1, and 80–500 M_☉ in the case of X2, although most of the arguments for such high masses come from the thermal components of the spectral fits, which could be unreliable. We consider masses of ≲20 M_☉ less likely also because they are harder to reconcile with typical expectations for the supercritical/beamed emission models available.

While this paper was in preparation, Soria et al. (2007) published an analysis of NGC 1365 X1 and concluded that the spectra are best fit by absorbed power-law components in all cases, and an accretion disk component contributes at most 10%–25% of the emission of X1 in select higher S/N spectra. The authors suggest that this is compatible with a ≈50–150 M_☉ mass black hole, where the accretion disk is colder (0.2–0.4 keV) and the inner parts of the accretion flow are in the form of an outflow or a Comptonizing corona, which is responsible for the power-law emission. The current data, despite their high photon statistics, are insufficient to distinguish between different scenarios without the assumption of a specific theoretical model (e.g., power-law emission with or without contribution from the accretion disk and intrinsic absorption) followed by a targeted interpretation of the results. Higher S/N X-ray spectra and better constraints on the short-term variability of X1, as well as better theoretical understanding of the most luminous ULXs (which will allow us to select the proper theoretical models to fit), will allow us to distinguish those scenarios in the future. It is reassuring that despite our different interpretations of the X-ray spectra of X1, we reach similar conclusions about the possible black hole mass of X1.

The optical counterpart of X2, a luminous H ii region in the spiral arm of NGC 1365, is the most likely place to find an IMBH—both in terms of formation mechanisms for IMBHs, which require very dense environments, and in terms of the availability and capture of a suitable donor star to supply the accretion material. The lack of an optical counterpart of X1 makes the association with an IMBH more problematic, but not implausible; recent theoretical work suggests that for primordial IMBHs (one of three possible formation mechanisms, all of which require a dense environment), most IMBHs with masses less than 1000 M_☉ would be ejected from the dense globular cluster where they formed (Holley-Bockelmann et al. 2007). Consequently, the lack of an optical counterpart for X1 does not automatically preclude an IMBH primary. The X-ray variability of X2, which appears to be a transient source, is also statistically consistent with an IMBH interpretation.

We thank Hans Ritter and Andrea Merloni for illuminating discussions of the properties of X-ray binaries. We are grateful to Swara Ravindranath for sharing with us her data of the H ii region whose optical position coincides with NGC 1365 X2 and suggesting we consult Roy & Walsh (1997) for the optical spectrum. We thank Li-Xin Li for his help with the implementation of the kerrbb model. We thank our referee, Roberto Soria, for his careful reading of the manuscript and insightful comments. This work is based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the US (NASA). In Germany, the XMM-Newton project is supported by the Bundesministerium für Wirtschaft und Technologie/Deutsches Zentrum für Luft- und Raumfahrt (BMWI/DLR, FKZ 50 OX 0001) and the Max-Planck Society. It is also based on an observation obtained with the Chandra X-ray telescope (a NASA mission). This project has been partly funded by the DFG Priority Programme 1177 "Galaxy Evolution".