A YOUNG MASSIVE STELLAR POPULATION AROUND THE INTERMEDIATE-MASS BLACK HOLE ESO 243-49 HLX-1

The formation of stellar-mass black holes (BHs; ∼3–80 M_☉; Belczynski et al. 2010) through the collapse of massive stars is well accepted, but it is not yet completely clear how the supermassive BHs (∼10⁶–10⁹ M_☉) are formed. They may form through the merger of ∼10²–10⁵ M_☉ intermediate-mass black holes (IMBHs; Ebisuzaki et al. 2001). IMBHs are thus a crucial missing link between stellar-mass and supermassive BHs, with likely environments for their formation including globular clusters (Miller & Hamilton 2002) and the nuclei of dwarf galaxies (Ricotti & Ostriker 2004). The existence of IMBHs also has implications for other areas of astrophysics including the search for dark matter annihilation signals (Fornasa & Bertone 2008), the epoch of reionization of the universe (Ricotti & Ostriker 2004), and the detection of gravitational wave radiation (Abbott et al. 2009). The study of IMBHs and the environments in which they are found thus has important connotations for a wide range of important questions in modern astrophysics.

The brightest ultra-luminous X-ray source ESO 243-49 HLX-1 currently provides the strongest evidence for the existence of IMBHs (Farrell et al. 2009). HLX-1 is located in the halo of the edge-on S0a galaxy ESO 243-49, ∼0.8 kpc out of the plane and ∼3.3 kpc away from the nucleus. At the redshift of ESO 243-49 (z = 0.0223) the maximum 0.2–10 keV X-ray luminosity of HLX-1 is ∼1.3 × 10⁴² erg s⁻¹ (Godet et al. 2011a), a factor of ∼400 above the theoretical Eddington limit for a 20 M_☉ BH. Luminosities up to ∼10⁴¹ erg s⁻¹ can be explained by stellar-mass BHs undergoing super-Eddington accretion (Begelman 2002) and/or experiencing significant beaming, which makes them appear to exceed the Eddington limit for isotropic radiation (King 2008; Freeland et al. 2006). However, luminosities above ∼10⁴¹ erg s⁻¹ are difficult to explain without a more massive BH. Following the discovery of an optical counterpart by Soria et al. (2010), the distance to HLX-1 was confirmed through the detection of the Hα emission line at a redshift consistent with the host galaxy (Wiersema et al. 2010), confirming the extreme luminosity. Modeling the accretion disk emission using relativistic (Davis et al. 2011) and slim disk (Godet et al. 2011b) models implies a mass between ∼3000–100,000 M_☉, consistent with values obtained by scaling from stellar-mass BH binaries (Servillat et al. 2011).

Long-term monitoring with the Swift observatory has shown that HLX-1 varies in X-ray luminosity by a factor of ∼50 (Godet et al. 2009), with correlated spectral variability reminiscent of that seen in Galactic stellar-mass BHs (Servillat et al. 2011). Since Swift monitoring began in 2009, HLX-1 has been observed to undergo three outbursts, each spaced approximately one year apart (Godet et al. 2011b). The characteristic timescales of the outbursts are inconsistent with the thermal-viscous instability model, and the outburst mechanism could instead be tidal stripping of a companion star in an eccentric orbit (Lasota et al. 2011).

Excess UV emission was detected consistent with HLX-1 with the GALEX and Swift observatories, although this emission could not be resolved from the nucleus of ESO 243-49 (Webb et al. 2010). At least some of the UV excess is likely to be associated with a z ∼ 0.03 background galaxy (Wiersema et al. 2010; Farrell et al. 2011). Using preliminary UV fluxes obtained from the Hubble Space Telescope (HST) data presented in this Letter, Lasota et al. (2011) estimated the age of the stellar environment around HLX-1 to be ∼0.3–0.6 Gyr through stellar population synthesis modeling and the assumption of a 5 × 10⁶ M_☉ cluster mass and solar metallicities. They therefore concluded that the most likely progenitor of the companion star would be an asymptotic giant branch star with a mass ∼2.7–3.5 M_☉. However, the contribution from accretion disk emission (in particular, possible irradiation in the outer disk) was not taken into account, and so a more rigorous approach employing broadband modeling is required.

In this Letter we report the results of detailed broadband spectral modeling of HLX-1 using UV, optical, and near-IR observations performed with the HST in conjunction with simultaneous observations performed in X-ray wavelengths with the Swift X-ray Telescope (XRT). The Letter is organized as follows: Section 2 lists the data reduction steps, while Section 3 describes the analysis methods employed and the results obtained. Section 4 discusses the implication of these results and the conclusions that we have drawn.

Following the peak of the second outburst we obtained three orbits of observations with the HST on 2010 September 13 and 23 under program 12256 in order to constrain the nature of the environment around HLX-1. During these observations HLX-1 was in the thermally dominated high/soft spectral state. Observations were performed in the far-UV (FUV) band using the Advanced Camera for Surveys (ACS) Solar Blind Camera (SBC), and in the near-UV (NUV), Washington C, V, I, and H bands with the Wide Field Camera 3 (WFC3) UVIS and IR cameras. Table 1 presents the log of the HST observations. We analyzed the final HST images generated by the pipeline (drz files) using the latest calibration data (CALWF3 = 2.1 as of 2010 May 15, and CALACS = 5.1.1 as of 2010 April 27). As part of the pipeline these images were flat-fielded, combined (with cosmic ray rejection) and geometrically corrected using PyDrizzle v6.3.5 (2010 May 19).

Table 1. Log of the HST Observations and Details of the Astrometric Corrections Applied to the Images

Instrument	Filter	Date	texp	NISPIa	Fit rms	Astrometric Errorb
			(s)			HST	Total
ACS-SBC	F140LP	2010 Sep 13	2480	4c	001	035	047
WFC3-UVIS	F300X	2010 Sep 23	1710	6	005	035	047
WFC3-UVIS	F390W	2010 Sep 23	712	8	005	035	047
WFC3-UVIS	F555W	2010 Sep 23	742	18	009	040	050
WFC3-UVIS	F775W	2010 Sep 23	740	25	009	040	050
WFC3-IR	F160W	2010 Sep 23	806	22	010	041	050

Notes. ^aNumber of ISPI sources used in the correction. ^bAll errors quoted are at the 95% significance level. ^cThe F140LP image was aligned on the F300X image, and not on ISPI directly.

Download table as:  ASCIITypeset image

We corrected the astrometry of each HST image following a two-step procedure with an intermediate image and using the US Naval Observatory CCD Astrograph Catalog (UCAC3; Zacharias et al. 2010) as an astrometric reference (typical rms error of 01). As the HST WFC3 and ACS fields of view are relatively small and the target is located in a particularly empty field that only contains one or two UCAC3 sources (including the extended source ESO 243-49 itself), we used an intermediate image obtained with the Infrared Side Port Imager (ISPI) at the Cerro Tololo inter-American observatory (CTIO; van der Bliek et al. 2004) on 2009 August 2. This image covers a 10' × 10' region around ESO 243-49 in the J band. We aligned the ISPI image on the UCAC3 catalog using the Starlink/GAIA software and obtained a precision of 013 (rms) using nine reference stars. We then used the ISPI image as a reference to align the WFC3 images. The number of stars used in the process and the resulting absolute position error are given in Table 1 for each band. For the ACS-SBC image (F140LP filter), this process could not be applied because of the lack of detection of ISPI stars in this band. We thus used the WFC3-UVIS near-UV image as a reference, finding four common objects. The precision is also reported in Table 1. The HLX-1 Chandra X-ray position is 03 at 95% confidence (Webb et al. 2010). The HST errors include the UCAC3 absolute error of 005 and the ISPI relative error of 013 (1σ). The final position error, adding in quadrature all errors, is thus 05 at 95% confidence level for each HST image. A single counterpart is clearly detected in each band within this error circle (Figure 1), although in the F160W H-band filter image it is difficult by eye to see the counterpart against the strong diffuse emission from the galaxy.

Zoom In Zoom Out Reset image size

The field of HLX-1 was also observed with the Swift XRT (ObsIDs: 00031287055, 00031287056, 00031287057, 00031287058, and 00031287059) on 2010 September 13, 14, and 23 for a total of 17.5 ks. The data were reduced, and spectra were extracted in the same manner as described in Servillat et al. (2011). The spectra were binned to a minimum of 20 counts per bin in order to use χ² statistics for the fitting.

A single point source was significantly detected in all the HST images within the 95% confidence levels of the combined Chandra and HST astrometry of HLX-1 (see Figure 1). The source is unresolved in all the images.¹⁰ The 008 FWHM of the point-spread function (PSF) in the raw WFC3-UVIS images converts to a diameter upper limit of 40 pc at the distance to HLX-1 (95 Mpc; Wiersema et al. 2010).

We extracted the flux of the HLX-1 counterpart using aperture photometry with the astrolib IDL procedure¹¹ aper, after centering with the cntrd procedure. We used extraction radii that encircle at least 90% of the energy from the source (see Table 2). The encircled energy curve was calculated from the Tiny Tim PSF models¹² (Krist 1995) which consistently match the identified stars in the raw images, degraded with a Gaussian filter to reproduce the effect of image combination with drizzle. We then subtracted the background, estimated in an annulus around the source, and applied aperture corrections (see Table 2) to the net flux to obtain the total flux from the target.

Table 2. HST Photometry of the HLX-1 Counterpart

Band	Filter	λpivot	FWHM	Aperture Size		Encircled Energy		Magnitude
		(Å)	(Å)	Src	Bkg	Srca	Bkgb	(AB mag)
FUV	F140LP	1527	294.08	05	05–06	90.5%	1.8%	24.11 ± 0.05
NUV	F300X	2829.8	753	04	04–05	90.4%	1.2%	23.96 ± 0.04
C	F390W	3904.6	953	04	04–05	91.6%	1.2%	23.92 ± 0.06
V	F555W	5309.8	1595.1	04	04–05	92.1%	1.1%	23.83 ± 0.08
I	F775W	7733.6	1486	03	02–03	90.2%	2.1%	23.91 ± 0.08
H	F160W	15405.2	2878.8	05	05–06	90.6%	1.2%	24.4 ± 0.3

Notes. ^aFraction of encircled energy in the aperture. ^bFraction of the source energy included in the background annulus.

Download table as:  ASCIITypeset image

For the WFC3-IR F160W image, the galaxy emission is about 10 times higher than the flux of the counterpart to HLX-1 and the background is not symmetric. We therefore interpolated the extended emission of the galaxy in a 10 pixel radius region around HLX-1 using the procedure grid_tps in IDL, and subtracted it from the image in order to obtain a flat and symmetric background around the source. We then performed similar aperture photometry as for the other bands.

The fluxes were converted into AB magnitudes using the magnitude zero points given in the HST WFC3 and ACS documentation and Web sites.¹³,¹⁴ Errors on the photometry were computed with the aper procedure as the quadratic sum of the scatter in background values, the random photon noise, and the uncertainty in mean sky brightness. The magnitudes obtained for the HLX-1 counterpart in each filter are given in Table 2. The fluxes and errors were then converted into units of photons cm⁻² s⁻¹ and XSPEC-readable spectral files (with diagonalized response files) were generated using the ftools task flx2xspec.

The fitting of the broadband spectral energy distribution (SED) constructed from the HST and Swift data was performed using XSPEC v12.6.0q (Arnaud 1996). The Swift XRT data were consistent with both multi-color disk blackbody and power-law models; however, the power-law photon index of Γ = 4.9 is unphysically steep. We thus assumed a thermal model for the X-ray emission and fitted the SED with an irradiated disk model (diskir; Gierliński et al. 2008, 2009), which includes thermal emission from the inner disk, non-thermal contribution from the Compton tail, and reprocessing in the outer disk. Components representing absorption by the neutral hydrogen column (using the tbabs model and the elemental abundances prescribed in Lodders 2003) and dust extinction (using the redden model and the extinction curves in Cardelli et al. 1989) were included. An acceptable fit was not obtained (χ²/dof = 48.26/30), with a clear excess of UV/optical/near-IR emission evident in the residuals.

HLX-1's position ∼1 kpc out of ESO 243-49's plane is naturally explained if the BH was embedded in a cluster of stars. We therefore attempted to fit the red excess in the SED using stellar population models representing emission from such a stellar cluster. We generated XSPEC additive table models¹⁵ for the Maraston (2005) stellar population models (based on theoretical atmospheres with the Salpeter initial mass function) using the ftools routine wftbmd. We added this component to the absorbed, reddened, irradiated disk model with no priors imposed on the age or metallicity of the stellar component. The redshift was frozen at the value of z = 0.0223 obtained from Very Large Telescope (VLT) spectroscopy (Wiersema et al. 2010) but the remaining parameters were free to vary.¹⁶

Two distinct fits were obtained with the irradiated disk plus stellar population model (Figure 2). One solution had a young stellar population (13 Myr) with a low fraction of bolometric flux reprocessed in the outer disk. The second solution was with an older stellar age (13 Gyr) and high levels of disk irradiation. Ages up to ∼200 Myr are also allowed (giving f_out ∼ 1 × 10⁻³), but solutions between ∼200 Myr to 10 Gyr required unphysically high fractions of reprocessing in the outer disk (≫10% of the bolometric flux, implying that the solid angle of the disk seen by the central BH is >4π sr; see Gierliński et al. 2009). The best-fit outer disk radii in both fits were ∼10³ times the inner disk radius. Such a small disk would be stable in the context of the thermal-viscous disk instability model (Lasota et al. 2011), adding further weight to the arguments against this mechanism driving the observed outbursts. The stellar metallicity, irradiated disk fraction, and outer disk radius were poorly constrained, preventing us from drawing any strong conclusions regarding the nature of the stellar population or the size and level of irradiation of the disk. In addition, the fact that we can fit the data with two models with such disparate parameter values indicates degeneracies in the fit, particularly between the stellar age and fraction of outer disk irradiation. Nonetheless, using the age, metallicity, and luminosity of the stellar components in each fit, we are able to constrain the stellar-mass to be ∼(4–6) × 10⁶ M_☉. Table 3 lists the parameter values obtained with this fit.

Zoom In Zoom Out Reset image size

To test for model dependency in the fit, we also generated additive table models using stellar population models based on empirical stellar libraries by Maraston & Strömbäck (2011). In particular, we used the models based on the Pickles (1998) solar metallicity library because of their extended wavelength range. Replacing the theoretical atmosphere models with these empirical stellar library models also obtained fits with both young (χ²/dof = 23.81/28) and old (χ²/dof = 24.52/28) stellar populations.

The detection of a stellar population around HLX-1 provides insights into the origin of and environment around an IMBH. Our SED fitting could not constrain the age of the stellar cluster around HLX-1 other than to show that only young or old populations are allowed. However, the reprocessing fraction in the outer disk required for old solution is ∼10%, which borders on being non-physical (reprocessing fractions >10% require the disk to subtend an unfeasibly large solid angle; Gierliński et al. 2009). In addition, the size of the accretion disk is constrained to be extremely small with an outer radius ∼1000 times the inner disk radius.

Galactic low-mass X-ray binaries in disk-dominated states typically have fractions of reprocessing in the outer disk of f_out ∼ 10⁻³ (Gierliński et al. 2009), far lower than the value obtained in the old stellar population solution. The outer disk radius also has a strong effect on the reprocessing fraction. Freezing the outer disk radius at a more likely value of 10⁵R_in (C. Done 2011, private communication), we could not find an acceptable fit solution with an old population (χ²/dof = 225.42/29), indicating that an extremely high reprocessed fraction and an extremely small disk are necessary outcomes with an old star cluster.¹⁷ We therefore favor the young stellar population solution, although without additional data we cannot conclusively rule out an extremely old star cluster. The upper limit of 40 pc derived for the cluster diameter is consistent with either a globular cluster or young massive star cluster, which typically have half-mass radii of ∼10 pc (Harris 1996) and <50 pc (Portegies Zwart et al. 2010), respectively.

The presence of a young stellar population is difficult to reconcile with HLX-1 residing in a classical globular cluster (such as those observed in our own Galaxy), which are typically dominated by old stars (e.g., Forbes 2003). However, globular clusters with young stellar populations have been observed around disrupted galaxies such as the Antennae (Bastian et al. 2006) and the Magellanic Clouds (Elson & Fall 1985). The mass of the cluster around HLX-1 (calculated using the derived age and metallicity, the observed luminosity and the model mass-to-light ratio¹⁸) is ∼4 × 10⁶ M_☉, which is at the upper end of the standard classical globular cluster mass range (Maraston et al. 2004).

A young star cluster surrounding HLX-1 could also be explained in the accreted dwarf galaxy scenario (Knierman 2010). Tidally stripping a dwarf galaxy during a merger event could remove a large fraction of the mass from the dwarf galaxy, with star formation triggered as a result of the tidal interactions. This could result in the observed IMBH embedded in the remnant of the nuclear bulge and surrounded by a young, high-metallicity, stellar population. It has been proposed that such accreted dwarf galaxies may explain the origin of some globular clusters, with the remnant cluster appearing more like a classical globular cluster as its stellar population ages (Forbes & Bridges 2010).

A link has been drawn between the presence of prominent dust lanes in early-type galaxies with frequent gas-rich minor mergers, with the host galaxy nuclear BH becoming active within <200 Myr following the merger event (Shabala et al. 2011). The HST images of ESO 243-49 (see Figure 1) reveal pronounced dust lanes and yet no evidence of nuclear activity was detected in Chandra X-ray images (Servillat et al. 2011), implying that the merger events that contributed to the formation of the dust lanes took place in the recent past. This is thus consistent with the presence of a young stellar population surrounding HLX-1, where localized star formation would have been triggered as a result of tidal interactions with ESO 243-49 following a recent merger event.

We note that Soria et al. (2011) argue against a young massive stellar population using lower signal-to-noise UBVR data from the VLT. However, they do not fit the X-ray and optical data simultaneously and use alternative stellar population models. Additional HST observations obtained at different X-ray luminosities are thus required in order to test the competing theories and therefore constrain the disk irradiation and stellar population contributions.

We thank the referee for useful comments and J.-P. Lasota, T. Alexander, C. Done, S. Shabala, F. Grisé, K. Arnaud, and R. Starling for helpful discussions. S.A.F. is the recipient of an ARC Postdoctoral Fellowship, funded by grant DP110102889. S.A.F., T.J.M., A.J.G., and K.W. acknowledge funding from the UK STFC. M.S. acknowledges support from NASA grants DD0-11050X and GO 12256. J.P. and C.M. acknowledge the Marie-Curie Excellence Team grant Unimass, ref. MEXT-CT-2006-042754 of the TMR program financed by the European Community. N.A.W., D.B., and O.G. thank the CNES. Based on observations made with the NASA/ESA HST, obtained at the STScI, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. These observations are associated with program 12256. Also based on observations with Swift.

Facilities: HST (ACS, WFC3) - Hubble Space Telescope satellite, Swift (XRT) - Swift Gamma-Ray Burst Mission, Blanco - Cerro Tololo Inter-American Observatory's 4 meter Blanco Telescope