当前位置:首页 >> >>

XMM-Newton Observations of the Ultraluminous Nuclear X-ray Source in M33

Astronomy & Astrophysics manuscript no. 2003˙303 (DOI: will be inserted by hand later)

February 2, 2008

XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33?
L. Foschini1 , J. Rodriguez2,3 , Y. Fuchs2 , L. C. Ho4 , M. Dadina1 , G. Di Cocco1 , T. J.-L. Courvoisier3,5 , G. Malaguti1

arXiv:astro-ph/0312118v2 5 Dec 2003


2 3 4


Istituto di Astro?sica Spaziale e Fisica Cosmica (IASF) del CNR, Sezione di Bologna, Via Gobetti 101, 40129 Bologna (Italy) CEA Saclay, DSM/DAPNIA/SAp (CNRS FRE 2591) F–91191 Gif–sur–Yvette Cedex, France ? INTEGRAL Science Data Centre, Chemin d’Ecogia 16, CH–1290 Versoix, Switzerland The Observatories of the Carnegie Institution of Washington, 813 Santa Barbara Street, Pasadena, CA 91101 (USA) Observatory of Geneva, 51 chemins des Mailettes, 1290 Sauverny, Switzerland

Received 10 September 2003; accepted 3 December 2003 Abstract. We present observations with XMM-Newton of M33 X-8, the ultraluminous X-ray source (L0.5?10 keV ≈ 2 × 1039 erg/s) closest to the centre of the galaxy. The best-?t model is similar to the typical model of Galactic black holes in very high state. Comparison with previous observations indicates that the source is still in a very high state after about 20 years of observations. No state transition has been observed even during the present set of XMM-Newton observations. We estimate the lower limit of the mass of the black hole > 6 M⊙ , but with proper parameters taking into account di?erent e?ects, the best estimate becomes 12 M⊙ . Our analysis favours the hypothesis that M33 X–8 is a stellar mass black hole candidate, in agreement with the ?ndings of other authors. In addition, we propose a di?erent model where the high luminosity of the source is likely to be due to orientation e?ects of the accretion disc and anisotropies in the Comptonized emission. Key words. X-rays: binaries — X-rays: galaxies — Galaxies: individual: M33

1. Introduction
About 20 years ago, with the early observations of nearby spiral galaxies by the Einstein satellite, a new class of intermediate luminosity (LX = 1039 ? 1040 erg/s) X– ray sources was discovered (cf Fabbiano 1989). These sources, later de?ned as ultraluminous X–ray sources (ULX, Makishima et al. 2000), were immediately intriguing, since one of the proposed model is that they could be intermediate mass black holes (102 ? 104 M⊙ ) accreting at sub–Eddington rates, the missing link between stellar mass X–ray binaries and active galactic nuclei (see Miller & Colbert 2003 for a review on intermediate mass black holes and their relationship with ULX). However, there are also other explanations available, which do not require a new class of object. According to these models, the ULX are stellar mass X–ray binaries, but either with truly super–Eddington accretion rate (e.g. Watarai
Send o?print requests to: L. Foschini e-mail: foschini@bo.iasf.cnr.it ? Based on public observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the USA (NASA).

et al. 2000, Begelman 2002), or with sub–Eddington rate, but with some type of collimated emission, either simply anisotropic (King et al. 2001) or relativistic (K¨rding et o al. 2002, Georganopoulos et al. 2002) to increase the observed luminosity. The threshold to de?ne an ULX is now generally set to 1039.0 erg/s, without any reference to the physical mechanism responsible for this value (see Miller & Colbert 2003 for a review). Surveys of ULX (e.g. with ROSAT Colbert & Ptak 2002, with XMM–Newton Foschini et al. 2002, with Chandra Colbert et al. 2003) can give gross information about these sources, their statistical properties, their relationships with the host galaxy. However, to improve the understanding of these sources, a detailed study of nearby ULX with high signal–to–noise data are needed. M33 (NGC 598) is one of the nearest spiral galaxies (d = 795 kpc). Classi?ed as SA(s)cd, it has an inclination angle of 55? (Ho et al. 1997). Since the ?rst observations with the Einstein satellite (Long et al. 1981), it was clear that the central source (M33 X-8) had particular features (luminosity in the 0.2 ? 4 keV energy range of about 1039 erg s?1 , soft spectrum, excess of absorption along the line of sight) suggesting that the source is somewhat di?erent


L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33

from an active galactic nucleus (Trinchieri et al. 1988). The authors suggested the possibility that M33 contains a new type of X-ray binary system. Later on, ASCA (Takano et al. 1994) observations extended up to 7 keV and strengthened the early results of Trinchieri et al. (1988). The best-?t model was composed of a multicolour disc (MCD) plus a power law at high energies, consistent with that of Galactic black holes in their high state. However, Schulman & Bregman (1995), based on ROSAT observations, conclude that the probability of such an unusual X-ray binary close to the centre of M33 is very small. Another point which makes M33 X-8 an unusual source is the steadiness of its ?ux, except for a modulation of ? 20% with a period of 106 days (Dubus et al. 1997). This discovery strengthened the hypothesis of a binary system, since the modulations can be due to the precession motion of the accretion disc (cf. Maloney et al. 1996). It is important to add that there is a lack of information at wavelengths other than X-rays for the source, since the source is located in a crowded region, so that it is di?cult to ?nd the right counterpart or the companion star. The recent increase of interest for ultraluminous Xray source phenomenon gave new light to the study of M33 X-8. Indeed, since the spatial resolution of Einstein, ROSAT, and ASCA were not su?cient to rule out the possibility of a small o?set of the source from the optical centre, Makishima et al. (2000) suggested that X-8 could be an ULX. However, Chandra observations put the tightest constraints on the position of X-8 (Dubus & Rutledge 2002). The authors found a possible counterparts at radio wavelengths: it was identi?ed with the point source n. 102 discovered by (Gordon et al. 1999) with the VLA at 20 and 6 cm. In the near-IR, M33 X–8 is at the 2MASS position of the nucleus (2MASS J01335089 + 3039365) within 0.6′′ , which corresponds to about 2.3 pc at the distance of 795 kpc (Dubus & Rutledge 2002). The hypothesis of an active galactic nucleus (AGN) in the centre of M33 is inconsistent with the upper limits on the central black hole mass obtained from the velocity dispersion measurements of the nuclear region: Kormendy & McClure (1993) gave an upper limit of 5×104 M⊙ , by using the Canada-France-Hawaii Telescope. Recently, Gebhardt et al. (2001) set, with the Hubble Space Telescope (HST), an upper limit to only 1500 M⊙. Moreover the 106 days periodicity is not consistent with the AGN hypothesis. The possibility that M33 X-8 is an ULX is the best explanation, as already suggested by Makishima et al. (2000), although the source is very close to the centre of M33. We present a detailed analysis of observations of the source M33 X-8 with XMM-Newton. This work is organized as follows: after the introduction (Sect. 1), the X-ray data reduction and analysis are described in the Sect. 2. Section 3 deals with the observations of the nuclear region of M33 in the near-IR and radio wavelengths. The interpretation of the X-ray data is divided in the Sections 4 and 5: the evaluation of the mass of the compact object is

Power (Lheahy units)

Frequency (Hz)

Fig. 1. Power density spectrum of M33 X-8 build from all the XMM-Newton observations, except for ObsID 0102640101, which showed instrumental noise. extensively dealt with in the ?rst part, while the second discusses the main characteristics of the source.

2. XMM-Newton observation and data reduction
A set of observations of the central region of M33 is available in the XMM-Newton Public Data Archive (see Table 1), with the nucleus in several position angles (onaxis and o?-axis). For the processing, screening, and analysis of the data from the EPIC MOS1 and MOS2 cameras (Turner et al. 2001) and PN camera (Str¨ der et al. u 2001), we used the standard tools of XMM-SAS software v. 5.4.1 and HEAsoft Xspec (11.2.0) Xronos (5.19) and followed the standard procedures described in Snowden et al. (2002). In some cases, the observations were a?ected by solar soft-proton ?ares, so that a preliminary cleaning was necessary.

2.1. Time analysis
To study the evolution of M33 X?8 and check for possible state transitions, we extracted from the observations reported in Table 1 EPIC-PN light curves with ? 73 ms time resolution. We extracted the data from a circle with 35′′ radius and centered in the position of M33 X-8 (RA = 01 : 33 : 50.89, Dec = +30 : 39 : 37.2, J2000, uncertainty < 4′′ ). The background was derived from a region 2′ wide near the source in the PN camera. Except for the third observation, during which the soft-proton ?ares limit the good time to about 5 ks, the others had a net exposure time of ? 9 ? 10 ks. From all these light curves, but one, we produced power density spectra (PDS) on interval of ? 300 s, and all the resultant PDS of a single observation were averaged and Leahy normalized (cf Leahy et al. 1983). The resultant PDS are thus ?tted between 3.3 mHz and 6.8 Hz (Fig. 1). Although the PDS above ? 20 mHz is ?at and compatible with white noise, evidence for a red-noise component

L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33


Table 1. XMM-Newton Observation Log. Columns: (1) Observation Identi?er; (2) Date of the observation; (3) Duration of the observation [s]; (4,5,6) Observing mode of MOS1, MOS2, and PN, respectively [FF: Full Frame; SW: Small Window]; (7) Position with respect to the centre of the ?eld of view.
ObsID (1) 0102640101 0102640301 0102640601 0102641001 0102642001 0102642101 0102642301 Date (2) 04 Aug 2000 07 Aug 2000 05 Jul 2001 08 Jul 2001 15 Aug 2001 25 Jan 2002 27 Jan 2002 Exposure (3) 18672 14862 12525 13275 12266 13001 12999 MOS1 (4) SW FF FF FF FF FF FF MOS2 (5) SW FF FF FF FF FF FF PN (6) FF FF FF FF FF FF FF Position (7) on axis o? axis o? axis o? axis o? axis o? axis o? axis

PHASE (Cycle)

Fig. 2. Global light curve of M33 X-8 (all the observations) folded with a period of 5000 s. Error bars are at 1σ. is found below that value: our best ?t with a power law gives an index ?1.5+0.3 (χ2 = 130.6, 137 d.o.f.). The red ?0.2 noise that appears to be present is apparently not due to the source. Given te present statistics, it is not possible to have su?cient frequency resolution below 3.3 mHz to study the signal at 0.2 mHz (5000 s) reported by La Parola et al. (2003), but the folding of the light curve on this timescale indicates the presence of signal (see Fig. 2). Although the exposure time is not su?cient to have a highly signi?cative detection, the χ2 test gives a probability of 4 % for the constancy of the source, with an excess variance < 2.1 % (3σ), thus con?rming the results obtained by La Parola et al. (2003). No state transitions were observed: we detect only a ?ux increase in the observation of 15 August 2001, without signi?cant spectral changes with respect to the best-?t model described in the Sect. 2.1, and the ?ux variation was consistent with the well-known modulation of ? 20% already observed by early satellites.

2.2. Spectral analysis
For the spectral analysis, we retrieved the on-axis observation of M33 performed on August 4th , 2000. The EPIC

MOS cameras were set in the small-window mode, and the net exposure was 12.5 ks long. The PN camera was in full-frame mode, and the exposure was 13.5 ks long. We extracted the data from the same regions described in the Sect. 2.1. The background was derived from a region 2′ wide near the source in the PN camera, but for the MOS cameras, since they operated in small-window mode, we used the background in the closest chip. The spectra were rebinned so that each energy bin contained a minimum of 30 photons, and we ?tted only in the 0.5 ? 10 keV energy range because of the uncertainties in the MOS cameras calibration at low energies (cf. Kirsch 2003). The photon redistribution matrix and the related ancillary ?le were created appropriately with the rmfgen and arfgen tasks of XMM-SAS. We tried to ?t the spectrum obtained with EPIC PN and MOS cameras with several models. Results of this are reported in Table 2. The best-?t model is composed of a multicolour accretion disc with temperature at the inner disc Tin = 1.16 keV plus a power law with Γ = 2.5 (Fig. 3). The ?ux in the band 0.5 ? 10 keV is of 1.7 × 10?11 erg cm?2 s?1 and is in agreement with earlier observations of other satellites. The power law accounts for about 57% of the total ?ux. The absorption column is higher than the Galactic value of NH = 5.6 × 1020 cm?2 along the direction of observation, the latter being evaluated according to Dickey & Lockman (1990). In past observations, intrinsic absorption was never mandatory: Schulman & Bregman (1995) with ROSAT and Dubus & Rutledge (2002) with Chandra found that no absorption was required in addition to the Galactic hydrogen column. On the other hand, Gottwald et al. (1987) with EXOSAT, Trinchieri et al. (1988) with Einstein, Takano et al. (1994) with ASCA, Parmar et al. (2001) with BeppoSAX, and La Parola et al. (2003) with Chandra found that it was necessary to include an additional absorption component. In the present case, the additional absorption is required with statistical signi?cance greater than 99.99% (see Fig. 4 for the 2-dimensional ?t-statistic contour plot of the power-law photon index and the absorption column). The absorption along the line of sight appears to be the same for both the multicolour disc and the power law model.

Normalized Rate


L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33

Table 2. Results from the ?t of the X-ray data. Columns: (1) Model: power law (PL), power law with high-energy cuto? (COPL; cutoffpl model in xspec), multicolour black body disc (MCD; discbb model in xspec), RaymondSmith (RAY), unsaturated Comptonization (CST; compst model in xspec); (2) Absorption column [1021 cm?2 ]; (3) free parameters of the model: (PL) photon index Γ; (COPL) photon index Γ and cuto? energy Ecut [keV]; (MCD) temperature [keV] at the inner disc (Tin ); (RAY) plasma temperature [keV] and metal abundances a; (CST) temperature [keV] and optical depth τ ; (4) χ2 and degrees of freedom of the spectral ?tting; the reduced χ2 is reported between brackets; (5) ?ux in the 0.5?10 keV band (10?11 erg cm?2 s?1 ); (6) X-ray luminosity in the 0.5?10 keV band (1039 erg s?1 ) calculated for d = 795 kpc and corrected for the absorption. The Galactic column is NH = 5.6 × 1020 cm?2 . The uncertainties in the parameters are at the 90% con?dence level.
Model (1) MCD+PL MCD+COPL NH (2) 1.8 ± 0.2
0.5 1.40.4


1.00.2 0.8


1.65 ± 0.07

Parameters (3) Γ = 2.50.2 0.1 Tin = 1.16 ± 0.04 keV Γ = 1.80.7 0.8 Tin = 1.20.2 keV 0.1 Ecut > 2.3 keV Γ < 0.8 a < 0.97 kT = 1.30.5 keV 0.3 Ecut = 1.80.3 keV 0.2 kT = 1.13 ± 0.03 keV τ = 20.9 ± 0.6

χ2 /d.o.f. (4) 1221.5/1175 (1.04) 1219.4/1174 (1.04) 1216.8/1173 (1.04)

FX (5) 1.7 1.7

LX (6) 1.7 1.5



1244.6/1176 (1.06)



It is most probable that the earlier negative detections were due to low statistics, rather than other e?ects: indeed, the Chandra spectrum with no absorption (Dubus & Rutledge 2002) was 10 ks long and had 23 degrees of freedom; the observation from which La Parola et al. (2003) found additional absorption was 92 ks long and had 333 degrees of freedom. It is worth noting too that the better statistics obtained thanks to the large collecting area of XMM-Newton results in a smaller error range with respect to the previous measurement of the additional absorption. The measured value of 1.24×1021 cm?2 (already subtracted for the Galactic value) corresponds to an optical reddening of E(B ? V ) = 0.21 mag, in agreement with the latest HST observations that found E(B ? V ) = 0.22 mag (Long et al. 2002). La Parola et al. (2003) found that they needed to add a thermal plasma component. However, substituting the multicolour disc model with a thermal plasma model (e.g., Raymond-Smith) leads to a worse result, with parameters not properly constrained. If the thermal plasma model is added to, instead of substituting, the MCD, the results (not reported in Table 2) are even worse. Furthermore, using a power law with an exponential cuto?, a model successfully used by some authors (e.g., Gottwald et al. 1987; Trinchieri et al. 1988) did not improve the ?t, and in this case some parameters are also not properly constrained. The only real alternative model to the reported best ?t appears to be the unsaturated Comptonization model of Sunyaev & Titarchuk (1980). The plasma temperature is compatible with that obtained from the multicolour accretion disc and the optical depth τ , which is known to vary according to the disc inclination (θ), is compatible to a high value θ ≈ 60? . This agreement is expected in

the case of steady accretion discs around black hole candidates, as shown by Ebisawa et al. (1991). There is, however, no evidence of any anomalous Comptonization as found by Kubota et al. (2001) in GRO J1655?40. By adding to the best-?t model a Comptonized blackbody component (compbb model in xspec) and linking the blackbody temperature to the temperature of the inner disc (Tin ), the new three-component model does not converge. We tried also the bulk motion Comptonization model (bmc model in xspec, Laurent & Titarchuk 1999), which has been used successfully to ?t the soft state of several Galactic black hole candidates (Borozdin et al. 1999), but also some ULX, with M33 X–8 among them (Schrader & Titarchuk 2003); however, in the present case, the ?t gives unphysical results, with pegged parameters. We therefore do not mention this in Table 2. Having detected ?ux variations in the observation of 15 August 2001, we extracted the source spectrum to investigate the possibility of state transitions. The data were a?ected by pile-up, mildly for the PN and strongly for the MOS cameras. Therefore, we analyzed the data from PN only, and we extracted the source spectrum from an annulus centered on the nucleus coordinate and with radii 10′′ and 40′′ , thus excluding the central region a?ected by pile-up. The background was extracted from a nearby region of 2′ radius. The data were ?tted to the best-?t model, i.e. multicolour accretion disc and power law. The present ?t gave χ2 = 182 for d.o.f. = 191, with NH = (1.9 ± 0.5) × 1021 cm?2 , Γ = 2.3+0.8 , and Tin = 1.4+0.3 keV. Their re?0.4 ?0.4 sults are consistent with the reference spectrum within the measured errors. No state transition was observed, but the measured unabsorbed ?ux in the energy band 0.5?10 keV

L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33


3. Observations of the nuclear region at other wavelengths
Observations of the nuclear region of M33 at wavelengths other than X-rays are very di?cult because the stellar density in the nuclear region of M33 is so high. Even with the highest available resolution, the innermost region of the nucleus remains unresolved (Lauer et al. 1998). Chandra observations by Dubus & Rutledge (2002) placed the X-ray position within 0.6′′ (2.3 pc at the distance of 795 kpc) of the near-IR position from 2MASS and the radio position from the VLA observations. The Chandra coordinates are compatible with what has been found by ROSAT and the present observation with XMMNewton. The reference radio observation has been performed by Gordon et al. (1999). They observed M33 with the VLA at 6 and 20 cm with 7′′ resolution and detected the centre of M33 at RA = 01 : 33 : 50.89 and Dec = +30 : 39 : 37.33 (J2000). The ?ux density at 20 cm was S20 = 0.6 ± 0.1 mJy, while it was 0.2 ± 0.1 mJy at 6 cm (signal-to-noise ratio > 3). The spectral index was α = ?0.8 ± 0.2, where Sν ∝ ν α . By comparing the radio data with optical observations performed with the 4 m telescope at the Kitt Peak Observatory, Gordon et al. (1999) exclude the possibility that the M33 centre is related to supernova remnants or H II regions. Intermediate-age stars, common in the nuclear region (Lauer et al. 1998), should not generate strong radio emission; the only other type of source that can display an α = ?0.8 spectrum would be a background AGN. But the periodicity of 106 days in X-rays excludes the possibility of an AGN. Therefore, it is reasonable to assume that the VLA detection is genuinely linked with M33 X-8, although the current spatial resolution of either the X-ray or radio data cannot yet de?nitively prove this. However, there are problems to accept the 2MASS detection as simply the counterpart of X-8. In the 2MASS All Sky Data Release Catalog1 (released on March 2003) the centre of M33 is located at RA = 01 : 33 : 50.9 and Dec = +30 : 39 : 36.6 (J2000, spatial resolution 3′′ ). The apparent magnitudes are J = 12.06 ± 0.03, H = 11.44 ± 0.03, and K = 11.22 ± 0.03. The detection ?ags of the catalog indicate a good quality processing of a pointlike source, although 3′′ at 795 kpc is equivalent to about 11 pc. The total absorption column measured by the present work (NH = 1.8 × 1021 cm?2 ) allows us to calculate a visual extinction of AV = NH · 5.3 × 10?22 = 0.954 mag (by using RV = 3.1; cf. Cox 2000). Then, it is possible to calculate the extinction factors at the 2MASS wavelengths according to Cardelli et al. (1989): AJ = 0.27, AH = 0.18, and AK = 0.11 mag. The dereddened magnitudes are J =
http://www.ipac.caltech.edu/2mass/releases/allsky/. The Two Micron All Sky Survey (2MASS) is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.

Fig. 3. (top) Best-?t spectrum of M33–X8 with EPIC MOS1, MOS2, and PN data. The model used is the multicolour disc plus a power law. See Table 2 and the text for details. The ratio data/model is shown in the bottom panel. (bottom) Corresponding unfolded spectrum.

Fig. 4. 2-dimensional ?t-statistic contour plot of Γ versus NH . The contours are with ?χ2 = 2.3, 4.61, 9.21, corresponding to the con?dence levels of 68%, 90%, and 99%.

is 2.8 × 10?11 erg cm?2 s?1 , which is 27% higher than the ?ux measured during observation 0102640101, consistent with the well-known modulation of ? 20%.


L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33

11.79, H = 11.26, and K = 11.11, which yield colours of J ? H = 0.53 mag and H ? K = 0.15 mag, indicating an infrared excess that cannot be due to a single star. The near-IR emission is consistent with the light expected from the known nuclear star cluster of M33. According to Kormendy & McClure (1993), the central cluster has a B-band magnitude of ? 14.6. Long et al. (2002) recently concluded that M33’s nucleus has an age of ? 107 ? 109 yrs. From the population synthesis models of Bruzual & Charlot (2003), we anticipate that a cluster in this age range should have B ? V ≈ 0 ? 0.6 mag and V ? K ≈ 2 ? 3 mag. Hence, the nuclear star cluster of M33 is expected to have a K-band magnitude of ? 11?12, consistent with the value seen by 2MASS. Therefore, the 2MASS detection is likely to be dominated by the integrated stellar emission of the nuclear star cluster, not by near-IR emission intrinsic to X-8.

4. Discussion
The source shows interesting similarities with other well known Galactic BHC (see, e.g., McClintock & Remillard 2003). The radio spectral index of M33 X–8 is very similar to the ones of the famous microquasars 1E1740.7 ? 2942 (cf Mirabel et al. 1992) or SS433 (cf Dubner et al. 1998). Moreover we con?rm the 5000 s modulation of the emission of the source. If this period refers to a signal propagating at the speed of sound (with a typical value of cs = 10 km s?1 ) or at the speed of light, the corresponding physical dimensions are between 5 × 109 and 1.5 × 1014 cm. A similar variability has been observed also in SS433 (jet 1000 s) and for this source the typical dimension has been calculated to be about 1013 cm (Kotani et al. 2002). Moreover the source X–ray ?ux recorded by XMM– Newton is almost equal to that previously measured. The only change observed is in good agreement with the 106 days modulation previously measured (Dubus et al. 1997). By this evidence, we may assess that the source has been observed to be almost stable during the last 20 years. No particularly strong variation in the spectral and/or ?ux state of the source has ever been observed. These characteristics are very similar to what is observed in the BHC LMC X–1 (cf Nowak et al. 2001, Wilms et al. 2001). From these indications we are led to classify the ULX M33 X–8 as a black hole with stellar mass. This picture is supported also by the X–ray spectra we obtained with XMM–Newton. In particular, the temperature of the thermal component seems to be too high to be referred to an intermediate mass BH unless one assumes relativistic effects (see the next subsection).

4.1. Evaluation of the mass of M33 X-8
To calculate the mass of M33 X-8, we have to take into account that it cannot be greater than the upper limit of the non-luminous mass of the nucleus, 1500 M⊙ (Gebhardt et al. 2001).

From the X-ray analysis, it is obvious that X-8 is in a very high state, with a thermal component with kT ? 1 keV and a power law with photon index ? 2.5. According to some authors (e.g. Done & Gierlinski 2003), these spectral characteristics are the typical signature of the accretion disc around a black hole. In the present work, the best model to ?t the ultrasoft component is the multicolour accretion disc (MCD) by Mitsuda et al. (1984). The MCD model require, to be correctly used, some additional parameters (cf Merloni et al. 2000, Ebisawa et al. 2003). Therefore, we recall some basic de?nitions to explain the values of the parameters we used to calculate the mass of the compact object in the present case. We refer to the works of Makishima et al. (2000) and Ebisawa et al (2003) for further details and deeper analysis on the MCD model applied to ULX. The normalization of the MCD model AMCD allows a direct estimate of the inner disc radius Rin , by means 2 of AMCD = Rin cos θ/D2 , where D is the distance of the source in units of 10 kpc, θ is the inclination of the disc (θ = 0? means face-on; θ = 90? refers to the disc edgeon). Rin is expressed in km and depends on the spin of the black hole. In the case of a Schwarzschild black hole (spin 0), Rin is equal to three times the event horizon radius RS = 2GM/c2 (that is twice the gravitational radius), while for a Kerr black hole, the radius of the inner disc can be down to 1.24RS in the most extreme case of spin +1. Two more correction parameters should be taken into account: the ?rst correction, indicated with the parameter ξ, is to represent the fact that Tin , the temperature at the innermost disk boundary, is related to a radius a bit larger than Rin (ξ ≈ 0.41, Kubota et al. 1998). The second parameter is the spectral hardening factor f of Shimura & Takahara (1995), which takes into account the fact that in the MCD model Tin is the maximum disc colour temperature, and therefore it has to be converted into the e?ective temperature. The hardening factor weakly depends on the accretion rate and the viscous parameter α of the standard model (Shakura & Sunyaev 1973). For Galactic black holes, f is generally constant and within the range 1.7 ? 1.9 (see Ebisawa et al. 2003 for a discussion on these values). We assume f = 1.7. The mass of the compact object is therefore given by M = (Rin ξf 2 /8.86s) M⊙, where s is a coe?cient depending on the spin of the black hole. In the case of a Schwarzschild black hole s = 1. A further uncertainty in the evaluation of the mass is given by the inclination of the accretion disc, which is generally unknown. By assuming cos θ = 1, we obtain a lower limit of the mass. The MCD normalization in the best-?t model of M33 X-8 gives Rin = 46 ± 3 km, which corresponds to a mass of (6.2 ± 0.4) M⊙ . It is worth noting that the inclination of the accretion disc and the corresponding relativistic corrections can increase the value of the mass. To have an estimate of the possible inclination of the accretion disc, we note in Table 2 that the X–ray spectrum of M33 X–8 is also well ?t-

L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33


ted by the unsaturated Comptonization model by Sunyaev & Titarchuk (1980), with a temperature compatible with that of the MCD model. Therefore, it is reasonable to think that the power law component of the MCD+PL model could be due to the Comptonization. From the data of ?t with unsaturated Comptonization and the studies performed by Ebisawa et al. (1991), we can infer that a possible value for θ = 60? . If true, the mass would be M = (8.7 ± 0.5) M⊙ . If we include also relativistic e?ects of the disc inclination, including Doppler boosting and gravitational focusing, by assuming that the accretion disc shares the same behavior as that seen in AGNs (see Sun & Malkan 1989), the mass might rise to M = (12 ± 1) M⊙ . De?ning LEdd = 1.3(M/M⊙) × 1038 erg/s (e.g. Frank et al. 2002), and assuming L0.01?100keV = 2.7 × 1039 erg/s (extrapolated from the X–ray data) as the best approximation of the bolometric luminosity, it results that we are observing the source at about 1.7 LEdd for a 12 M⊙ black hole. To reach such a super–Eddington luminosity there are two possibilities: to have genuine super–Eddington accretion rate or sub–Eddington rate with some type of anisotropies or collimation. The bolometric luminosity of the accretion disc only can be calculated by using the Stefan–Boltzmann law with the dimension and temperature of of the innermost stable region. In the case of the inclination angle θ = 60? , the disc can account for Ldisc = 1.0 × 1039 erg/s. The remaining part of the accretion, related to the power law component, is more di?cult to explain, since it depends on the physical interpretation. In the case of sub–Eddington rate, a little anisotropy (solid angle of the emission < 1.6π) is su?cient to account for the observed ?ux. These values are similar to those of the Galactic microquasar GRS1915 + 105, where the inclination angle is 70? and the solid angle of the emission is about 2.4π (see King et al. 2001). We cannot exclude at all that the M33 X–8 could accrete at super–Eddington rates, although, in this case, we would expect values of the excess variance greater than those observed. We caution that the mass evaluation is only as good as the MCD model, the present best ?t, and the parameters adopted/inferred for the correction of the MCD. The above value is the best estimate, according to the XMM–Newton observation and in the hypothesis of a Schwarzschild black hole. It is worth noting in case of intermediate mass black hole, the temperature of the inner disc is expected to be much lower than the 1 keV measured in the present case. Miller et al. (2003) found a temperature of kTin = 0.15 keV for the two ULX in NGC1313 from which they inferred the presence of a black hole of ≈ 103 M⊙ . It is possible to reach a high value of the mass only if the black hole is maximally rotating (s = 1/6, Kerr black hole). In this case, taking into account all the e?ects and parameters described above, the mass of M33 X–8 could be as high as M = (147 ± 8) M⊙ .

4.2. A mildly relativistic jet or a Compton heated wind?
We propose two possible hypotheses to explain the power law component of M33 X-8. The ?rst is the presence of collimated emission (mildly relativistic jets). The dimensions of the variable region appear to be compatible with the X-ray emitting region from jets in microquasars: e.g. for SS433 (jet speed 0.26c, time variability 1000 s), this region has dimension ? 1013 cm (Kotani et al. 2002). The radio spectral index is compatible with the values found in the hot spots of some Galactic microquasars, like 1E1740.7?2942 (Mirabel et al. 1992) or SS433 (Dubner et al. 1998). Although it is not possible to resolve the radio emission from the nucleus of M33, the steep spectral index is consistent with the synchrotron emission from charged particles accelerated in shocks generated by the propagation of a jet in a di?use region. But a cautionary note should be stressed, which is the great uncertainty in the determination of the counterpart of M33 X-8. The second hypothesis takes into account the presence of Compton-heated winds, like, for example, the Galactic black hole GRO J0422+32 (van Paradijs et al. 1994). This model was never proposed to explain the high luminosity of ULX. In this case, the hard X-ray emission represents the signature of a speci?c physical process, generally taken to be inverse-Compton scattering of photons of thermal origin on a population of hot electrons. The geometry of the region where this process occurs is rather di?cult to understand. The two-phase model by Haardt & Maraschi (1991) of a corona in hydrostatic equilibrium around the accretion disc is one of the standard solutions. In this case, thermal radiation emitted by the accretion disc enters the hot corona and is Comptonized into hard X-rays. Part of this radiation is then reprocessed by the accretion disc, and a small fraction is re?ected. The fact that the spectrum of M33 X-8 is well ?tted also by the unsaturated Comptonization model strengthens the importance of the corona for this source (cf. Table 2). We consider, as a reference, the model developed by Begelman et al. (1983). A nearly hydrostatic corona exists at a distance r from the centre of the accretion disc system if the Compton temperature TIC is less than the escape temperature. This occurs inside the radius RIC = (1.0 × 1010 /TIC8 )M/M⊙ , where RIC is in cm and TIC8 is the inverse-Compton temperature expressed in units of 108 K (Begelman et al. 1983). If r > RIC , Compton heating can cause a strong wind (see Begelman et al. 1983, Begelman & McKee 1983, Shields et al. 1986). In this case, a winddriven relaxation cycle is set up, causing oscillations in the interplay of the disc accretion rate and the wind ejection rate. In the standard disc model, these oscillations have a 4/3 ˙ 1/3 ˙ period P = (3400 s) · M 14/9 /(α7/9 TIC8 Ma17 ), where Ma17 17 ?1 is the disc accretion rate in units of 10 g s . In the hypothesis that M33 X–8 is accreting at about 60% the Eddington limit, from the above equation, it is possible to calculate TIC8 , which in turn gives us RIC . For the values of mass of the compact object M = (6 ? 12) M⊙ , we infer


L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33 NASA grants from the Space Telescope Science Institute (operated by AURA, Inc., under NASA contract NAS5-26555). LF wishes to thank Giorgio Palumbo, Paola Grandi, and Massimo Cappi for useful discussions. This publication has made use of public data obtained from the High Energy Astrophysics Science Archive Research Centre (HEASARC), provided by NASA Goddard Space Flight Centre.

a value of RIC ≈ 7 × 109 cm. Therefore, if the oscillation occurs at this distance, the perturbation speed is about 14 km s?1 , compatible with the sound speed in a isothermal plasma at a temperature of about 2 × 104 K. This value can be compared with what has been found in the case of GRO J0422 + 32, where the temperature of the plasma is ? 3 × 104 K (van Paradijs et al. 1994). It is worth mentioning that out?ows have been invoked to account for the high luminosity in ULXs (e.g., Begelman 2002; King 2002; King & Pounds 2003) and the Compton-heated winds solution proposed here can be considered a variant of these models.

Begelman M.C., 2002, ApJ 568, L97 Begelman M.C., McKee C.F., 1983, ApJ 271, 89 Begelman M.C., McKee C.F., Shields G.A., 1983, ApJ 271, 70 Borozdin K., Revnivstev M., Trudolyubov S., Shrader C., Titarchuk L., 1999, ApJ 517, 367 Bruzual G., Charlot S., 2003, MNRAS, 344, 1000 Cardelli J.A, Clayton G.C., Mathis J.S., 1989, ApJ 345, 245 Colbert E.J.M., Ptak A.F., 2002, ApJS 143, 25 Colbert E.J.M., Heckman T., Ptak A.F., Strickland D., 2003, ApJ, accepted for publication (astro-ph/0305476) Cox A.N. (editor), 2000, Allen’s Astrophysical Quantities. IV Edition, Springer, New York Dickey J.M. & Lockman F.J., 1990, ARAA 28, 215 Done C., Gierlinski M., 2003, MNRAS 342, 1041 Dubner G.M., Holdaway M., Goss W.M., Mirabel I.F., 1998, AJ 116, 1842 Dubus G., Charles P., Long K.S., Hakala P.J., 1997, ApJ 490, L47 Dubus G. & Rutledge R.E., 2002, MNRAS 336, 901 Ebisawa K., Mitsuda K., Hanawa T., 1991, ApJ 367, 213 ˙ Ebisawa K., Zycki P., Kubota A., Mizuno T., Watarai K., 2003, ApJ 597, 780 Fabbiano G., 1989, ARAA 27, 87 Foschini L., Di Cocco G., Ho L.C., et al., 2002, A&A 392, 817 Frank J., King A.R., Raine D.J., 2002. Accretion power in astrophysics, Cambridge University Press, Cambridge. Gebhardt K., Lauer T.R., Kormendy J., et al., 2001, AJ 122, 2469 Georganopoulos M., Aharonian F.A., Kirk J.G., 2002, A&A 388, L25 Gordon S.M., Duric N., Kirshner R.P., Goss W.M., Viallefond F., 1999, ApJS 120, 247 Gottwald M., Pietsch W., Hasinger G., 1987, A&A 175, 45 Haardt F., Maraschi L., 1991, ApJ 380, L51 Ho L.C., Filippenko A.V., Sargent W.L.W., 1997, ApJS 112, 315 King A.R., 2002, MNRAS 335, L13 King A.R., Davies M.B., Ward M.J., Fabbiano G., Elvis M., 2001, ApJ 552, 109 King A.R., Pounds K.A., 2003, MNRAS, 345, 657 Kirsch M., 2003. EPIC status of calibration and data analysis. XMM-SOC-CAL-TN-0018, v. 2.1, 4 April 2003. K¨rding E., Falcke H., Marko? S., 2002, A&A 382, L13 o Kormendy J., McClure R.D., 1993, AJ 105, 1793 Kotani T., Trushkin S., Dennisyuck E.A., et al., 2003, in “New Views on Microquasars”, Ph. Durouchoux, Y. Fuchs and J. Rodriguez Eds. Center for Space Physics: Kolkata, p.265 (astro-ph/0208250) Kubota A., Tanaka Y., Makishima K., et al., 1998, PASJ 50, 667 La Parola V., Damiani F., Fabbiano G., Peres G., 2003, ApJ 583, 758

5. Final remarks
We presented the spectral and temporal analysis of XMMNewton observations of M33 X-8. The present analysis of X-ray data suggest that M33 X-8 is a stellar mass black hole, whose luminosity is only apparently super– Eddington for geometrical reasons. The lower limit for the mass of M > 6 M⊙ , and a best estimate of M = 12 ± 1 M⊙ , although we cannot completely exclude a mass of ≈ 150 M⊙ if X–8 is a maximally rotating BH. These conclusions are in agreement with the X–ray binary interpretation already found by several other authors (Makishima et al. 2000, Dubus & Rutledge 2002, King 2002, La Parola et al. 2003, just to mention the latest). We con?rm the oscillation with a period of 5000 s discovered by La Parola et al. (2003), and we suggest that this oscillation is associated with the interplay between the mass loss from a Compton-heated wind and the accretion rate. It is worth noting that also a mildly relativistic jet could explain as well most of the observed data. In this case, the 5000 s variability is due to oscillations at the basis of the jet. The case of M33 X-8 sheds new light on ULX studies. The interpretation proposed here for M33 X-8 — a stellarmass black hole whose luminosity is boosted by orientation e?ects of the accretion disc and Compton-heated winds (or even a mildly relativistic jet) — may serve as a useful template for understanding other ULXs. To date, ULXs with little or no variability have generally been associated with young supernovae remnants. M33 X-8 illustrates that steady X-ray sources, with weak short term variability, can be stellar-mass black holes. It is also of interest to note that environments rich in hot plasma, coming from hot winds of young stars or from stellar collisions, as might occur in compact young star clusters or the nuclei of galaxies, may be particularly conducive to fueling and sustaining ULX sources. Finally, we would like to emphasize that very high resolution, multiwavelength, simultaneous observations of M33 X-8 are required to draw de?nitive conclusions on the nature of this enigmatic source.
Acknowledgements. LF and MD acknowledge partial ?nancial support by the Italian Space Agency (ASI). JR and YF acknowledge ?nancial support from the CNES. LCH is is supported by the Carnegie Institution of Washington and by

L. Foschini et al.: XMM–Newton Observations of the Ultraluminous Nuclear X-ray Source in M33 Lauer T.R., Faber S.M., Ajhar E.A., Grillmair C.J., Scowen P.A., 1998, AJ 116, 2263 Laurent P., Titarchuk L., 1999, ApJ 511, 289 Leahy D.A., Darbro W., Elsner R.F., et al., 1983, ApJ 266, 160 Long K.S., Charles P.A., Dubus G., 2002, ApJ 569, 204 Long K.S., D’Odorico S., Charles P.A., Dopita M.A., 1981, ApJ 246, L61 Makishima K., Kubota A., Mizuno T., et al., 2000, ApJ 535, 632 Maloney P.R., Begelman M.C., Pringle J.E., 1996, ApJ 472, 582 McClintock J.E., Remillard R.A., 2003, to appear in “Compact Stellar X-Ray Sources”, eds. W.H.G. Lewin and M. van der Klis, Cambridge University Press (astro-ph/0306213) Merloni A., Fabian A.C., Ross R.R., 2000, MNRAS 313, 193 Miller M.C., Colbert E.J.M., 2003, Int. J. Mod. Phys. D, submitted (astro-ph/0308402) Miller J.M., Fabbiano G., Miller M.C., Fabian A.C. 2003, ApJ 585, L37 Mirabel I.F., Rodriguez L.F., Cordier B., Paul J., Lebrun F., 1992, Nature 358, 215 Mitsuda K., et al., 1984, PASJ 36, 741 Nowak M.A., Wilms J., Heindl W.A., Pottschmidt K., Dove J.B., Begelman M.C., 2001, MNRAS 320, 316 Parmar A., Sidoli L., Oosterbroek T., et al., 2001, A&A 368, 420 Schulman E. & Bregman J.N., 1995, ApJ 441, 568 Shakura N.I., Sunyaev R.A., 1973, A&A 24, 337 Shields G.A., McKee C.F., Lin D.N.C., Begelman M.C., 1986, ApJ 306, 90 Shimura T., Takahara F., 1995, ApJ 445, 780 Shrader C.S., Titarchuk L., 2003, ApJ 598, 168 Snowden S., Still M., Harrus I. et al., 2002. An introduction to XMM-Newton data analysis. Version 1.3, 26 September 2002. Str¨der L., Briel U., Dennerl K., et al., 2001, A&A 365, L18 u Sun W.-H., Malkan M.A., 1989, ApJ 346, 68 Sunyaev R.A., Titarchuk L.G., 1980, A&A 86, 121 Takano M., Mitsuda K., Fukazawa Y., Nagase F., 1994, ApJ 436, L47 Trinchieri G., Fabbiano G. & Peres G., 1988, ApJ 325, 531 Turner M.J., Abbey A., Arnaud M., et al., 2001, A&A 365, L27 van Paradijs J., Telesco C.M., Kouveliotou C., Fishman G.J., 1994, ApJ 429, L19 Watarai K., Mizuno T., Mineshinge S., 2001, ApJ 549, L77 Wilms J., Nowak M.A., Pottschimdt K., Heindl W.A., Dove J.B., Begelman M.C., 2001, MNRAS 320, 327




All rights reserved Powered by 甜梦文库 9512.net

copyright ©right 2010-2021。