Astronomical Science

The Search for Intermediate-mass Black Holes in Globular Clusters

Nora Lützgendorf1 lution on all scales. Supermassive black cess to explain the rapid growth of Markus Kissler-Patig1 holes (SMBHs) have masses from a mil- SMBHs is through the merger of smaller Tim de Zeeuw1, 2 lion up to several billions of solar masses seed black holes of intermediate mass Holger Baumgardt 3 and are situated in the centres of massive (between 100 and 10 000 solar masses, Anja Feldmeier1 . When a SMBH accretes the e.g., Ebisuzaki et al., 2001). Karl Gebhardt 4 surrounding material it emits tremendous Behrang Jalali5 amounts of energy. This energy can be Numerical simulations have shown that Nadine Neumayer1 observed at all wavelengths and is detect- intermediate-mass black holes can form Eva Noyola 6 able at large distances. The most active in runaway mergers of in a dense galaxies are quasars and have been a environment, such as young clusters. puzzle to astronomers since their discov- Also, ultraluminous X-ray sources at off- 1 ESO ery in the early 1960s. Their presence at centre positions in galaxies provide strong 2 Sterrewacht Leiden, Leiden University, large redshifts indicates that they existed evidence of massive black holes in stel- the Netherlands at a time when the was only one lar clusters (e.g., Soria et al., 2011). The 3 School of Mathematics and Physics, billion years old or less, at a very early globular clusters observed today are as University of Queensland, Brisbane, stage of structure formation (Fan, 2006) old as their host : could they have Australia when the first galaxies were just forming. delivered seed black holes, the missing 4  Department, University of How could a black hole with a mass of a link, in the crucial early stage of galaxy Texas at Austin, USA billion or more solar masses have formed formation? 5 I. Physikalisches Institut, Universität zu so early in the history of the Universe, Köln, Germany when galaxies were not yet fully evolved? A further motivation for studying IMBHs 6 Instituto de Astronomia, Universidad is that observations have shown a cor­ Nacional Autonoma de Mexico (UNAM), Black holes can grow in two ways: relation between the masses of SMBHs Mexico through the accretion of surrounding and the of their host material and by merging with other black galaxies (see Figure 1 and e.g., Ferrarese holes. Even if a black hole accretes at & Merritt, 2000; and Gebhardt et al., Intermediate-mass black holes (IMBHs) the highest possible rate (the Eddington 2000). Extrapolating this relation to the fill the gap between supermassive rate) for a billion years, it would not reach mass ranges of IMBHs yields a velocity and stellar-mass black holes and they one million solar masses if it started with dispersion of between 10 and 20 km/s, could act as seeds for the rapid growth a one seed. Observations such as those observed from the stars of supermassive black holes in early have also shown that SHMBs do not con- in globular clusters. The origin and inter- galaxy formation. Runaway collisions of stantly accrete at the efficiency of the pretation of this relation is still under massive stars in young and dense stel- Eddington rate. Thus, the preferred pro- debate and it is not known whether it lar clusters could have formed IMBHs that are still present in the centres of 10 globular clusters. We have resolved the 10 central dynamics of a number of glob­ ular clusters in our Galaxy using the FLAMES integral-field spectrograph at the VLT. Combining these data with photometry from HST and comparing 10 8 them to analytic models, we can detect the rise in the velocity dispersion pro- file that indicates a central black hole.

Our homogeneous sample of globu- ) ी lar cluster integral-field spectroscopy (M 10 6 allows a direct comparison between BH clusters with and without an IMBH. M

Ω Cen

The missing link? NGC 6388 Figure 1. The MBH– σ 4 relation for supermas- 10 G1 Black holes have fascinated humanity for M 15 sive black holes in gal- axies is shown. Globular almost three centuries. Objects of pure 47 Tuc Globular Clusters clusters and dwarf gal- gravity forming singularities in space­time axies lie at the extrapo- were first postulated by John Michell lation of the relation in 1783 and still seem enigmatic to physi- towards lower black 102 hole masses. The points cists today, fuelling a strong drive to 10 100 for a few globular clus- understand their , growth and evo- σ (km/s) ters are plotted.

The Messenger 147 – March 2012 21 Astronomical Science Lützgendorf N. et al., The Search for Intermediate-mass Black Holes

UT2INTEGRATION IN OZPOZ ARGUS ARRANGEMENT Figure 2. A schematic of the VLT FLAMES ARGUS integral field unit 300 μm in the Ozpoz fibre posi- Sky

Spectr tioner is shown (centre).

button 15 The ARGUS fibre geom-

Tumbler subslits etry and the allocation Argus ograph

2 mm of fibres into sub-slits head 4. feeding the GIRAFFE spectrometer is shown at right. 6.6 mm Array of 22 × 14 channels truly holds at the lower mass end. The hole. Obtaining velocities and velocity and Planetary Camera 2 (WFPC2) in the study of IMBHs and their environments dispersions in the centres of globular clus- I-band (F814W) filter in May 1998 (GO- might shed light on many astrophysical ters, however, is very difficult due to 6804, PI: F. Fusi Pecci) and retrieved from problems, helping us to develop a better the high stellar densities in these regions. the European HST archive. The choice understanding of black hole formation Resolving individual stars in the centre of filter is related to our spectral observa- and galaxy evolution. and measuring spectra with ground- tions, which are obtained in the wave- based telescopes requires adaptive op­­ length range of the calcium triplet, a prom- The idea that IMBHs might exist in globu- tics, extraordinary weather conditions inent absorption feature around 860 nm lar clusters was first proposed decades and time-consuming observations. For and ideally suited to the measurement ago. Many observational attempts to find these reasons we used a different method of radial velocities. By measuring the rela- direct evidence of a black hole in globu- — integrated light measurements (Noyola, tive shift of the calcium triplet through lar clusters have followed. However, most 2010; Lützgendorf, 2011). absorption line fitting, we obtained a ve­ studies have only observed the light locity for each spectrum. The final veloc- ­profiles in the centre of the clusters and We used the large integral field unit ity map for NGC 2808 is also shown in compared these with models. A central ARGUS, a mode of the FLAMES facility Figure 3. Velocity maps are instructive black hole does produce a prominent mounted on Unit Telescope 2 of the Very to look at and good for identifying sus­ cusp in the surface brightness profile of Large Telescope (VLT). Figure 2 shows picious features, but in order to compare a (Bahcall & Wolf, 1976; schematically the properties of the instru- our data to model predictions we need Baumgardt et al., 2005), but this signature ment. ARGUS consists of a 22 × 14 array the velocity dispersion profile. It is can be confused with the one expected of microlenses (called spaxels for spatial the velocity dispersion s, or the second from a core-collapse process. For this pixels) with a sampling of 0.52 arcsec- moment of the velocity (the quadratic reason, it is crucial to study the kinematic onds per microlens and a total aperture sum of the expectation value of the veloc- profile in addition to the light profile. of 11.5 by 7.3 arcseconds on the sky; ity and the velocity dispersion), which this field allows the core of a globular holds information about the mass distri- Our group set out to search for the kine- cluster to be covered in a few pointings. bution. matic signatures of intermediate-mass A more detailed description of FLAMES black holes in Galactic globular clusters and the ARGUS mode can be found in In order to derive a velocity dispersion by trying to understand which of our Pasquini (2002) and Kaufer (2003). For profile, we used radial bins around the observed globular clusters host an inter- the observations the GIRAFFE spectro- centre of the cluster and combined all mediate-mass black hole at their centre graph was set to the low spectral resolu- the spectra included in each bin. The and which do not. With a sample of tion mode LR8 (coverage 820–940 nm absorption lines of the resulting spectra ten clusters, our goal is to understand at a spectral resolution of 10 400) and the are broadened due to the different veloci- whether the presence of an IMBH corre- ARGUS unit was pointed to different ties of the individual stars contributing lates with any of the cluster properties. positions, each of them containing three to the spaxels. Measuring the width of exposures of 600 s with 0.5 arcsecond these lines yields the velocity dispersion dithering, to cover the entire core radius. for each bin, thus a velocity dispersion Mapping the integrated light The position angle of the long axis of profile as a function of radius. Great care the integral field unit was varied from 0 to is taken to measure the true velocity A central black hole in a dense stellar 135 degrees in order to achieve maximum ­dispersion and not artefacts of the instru- system, such as a globular cluster, coverage of the cluster centre. ment, the data reduction or the data affects the velocities of its surrounding combination process. stars. As an additional gravitational Figure 3 shows the combined pointings potential it causes the stars in its vicinity from the ARGUS observations of the to move faster than expected from the cluster NGC 2808 together with the re­­ Black hole hunting gravitational potential of the cluster itself. constructed field of view on the Hubble Measuring this difference provides an Space Telescope (HST) image. The HST A black hole is suspected when we see estimate of the mass of the possible black images were obtained with the Wide Field a difference between the measured ve­-

22 The Messenger 147 – March 2012   object spectrographs such as FLAMES/     Medusa. A Fabry Perot instrument can

W   also operate as an integral field unit: nar-

%KT     row wavelength band, wide-field images   at a series of different wavelength steps         are recorded. This results in a set of λ ML images with high spatial resolution from

'23(, &$ 1&421$".-2314"3$# 5$+."(38, / which the stellar absorption line fea- tures can be extracted by combining the ­photometry of the objects from each wavelength bin (Gebhardt et al., 1992).

After deriving the full kinematic and ­photometric profiles, the information is fed into models. We use different kinds of models. The first and the simplest kind are analytical Jeans models. The models take the surface brightness profile, de­ ll ll l l  JLR project it into a three-dimensional density profile and predict a velocity dispersion Figure 3. The process of deriving a velocity disper- since the density of stars is lower than in profile. The latter is then compared to the sion profile with integral-field spectroscopy is illus- the centre. The instruments we used actual measured velocity dispersion pro- trated. Shown are the field of view on the HST image (left panel), the ARGUS reconstructed pointing to obtain the outer kinematics are the file. We study the variations of the model (middle panel), a datacube with each spaxel (spatial Rutgers Fabry Perot on the Blanco when including the additional gravita- element) being a spectrum in the calcium triplet 4-metre telescope at Cerro Tololo Inter- tional potential of a black hole into the region (shown above), and the resulting velocity map American Observatory (CTIO) and multi- equations. By trying different black hole (right panel). The circles indicate the bins applied to derive the velocity dispersion profile. loc­ities and the expected velocities from the gravitational potential of the cluster derived from its light distribution only. In order to determine these expected ve­­loc­ ities, we needed an estimate of the stellar density in the cluster. This is done by tak- ing images from HST and measuring the surface brightness profile. We determined the photometric centre of the cluster by measuring the symmetry point of the NGC 1851 NGC 1904 NGC 2808 spatial distribution of the stars. Using col- our– diagrams, we also identi- fied the stars which do not belong to the cluster.

In addition to the VLT and HST observa- tions, data from other instruments are required for kinematic measurements of outer regions that are too faint to be studied using integrated light. In order to NGC 5286 NGC 5694 NGC 5824 detect a black hole, only the innermost points are critical, but in order to under- stand the global dynamics of the cluster and to determine its mass, the entire velocity dispersion profile is needed. The outer kinematics can be obtained through the measurement of individual stars,

Figure 4. Velocity maps of all the Galactic globular clusters of our sample (except ) are shown superposed on HST images. NGC 6093 NGC 6266 NGC 6388

The Messenger 147 – March 2012 23 Astronomical Science Lützgendorf N. et al., The Search for Intermediate-mass Black Holes

NGC 6388 NGC 2808 30 10 16 8 3 15 MBH = 45 × 10 Mी 6 2 χ 2 10 χ 4 14 5 25 2 0 0 0 10 20 30 40 0 1000 2000 3000 4000

MBH (Mी) MBH (Mी) 12 s) s)

MBH = 0 km/ km/ ( 20 ( S S 10 RM RM V V

8 15

6 10 0.0 0.51.0 1.5 0.0 0.51.0 1.52.0 2.5 log r (arcsec) log r (arcsec)

Figure 5. The velocity dispersion profiles of the ities. Then, the system evolves by letting is rather flat and an IMBH larger than globular clusters NGC 6388 and NGC 2808 are the particles interact gravitationally. By ~ 1000 M is excluded. Among all our shown. Overplotted are a set of Jeans models with A different black hole masses (grey lines) and the following the evolution of this cluster over candidates we find clusters like NGC 2808 ­best-fit model (black line). In the corner of each plot a lifetime (~ 12 Gyr), we studied its evolu- as well as clusters where the central the variation of χ2 of the fit as a function of black- tion in detail and compared the final state points of the velocity dispersion profile are hole mass is shown. The shaded area marks the with the observations. By changing the rising just as in NGC 6388 and Omega region for Δχ2 < 1, and thus the 1σ error on the black hole mass. While for NGC 2808 the profile is rather initial conditions, we found the simula- Centauri. flat and does not require a black hole, NGC 6388 tions which fitted the observed data best has a steeper profile and a best-fit model with and learnt about the intrinsic properties After finalising the analysis and the mod- a ~ 17 000 M black hole (Lützgendorf et al., 2011, A of the globular clusters and their black elling of all the clusters in our sample, Lützgendorf et al., 2012, in prep.). holes, such as mass-to-light ratio varia- we will be able to start correlating the tions and the anisotropy in the cluster. presence and mass of an IMBH with the masses, we find which model fits our cluster properties. We will also be able data best. Figure 5 shows the entire to set the selection criteria for further velocity dispersion profile for two exam- Black hole or no black hole? studies and use our experience and ples: NGC 2808 and NGC 6388 with expertise to extend the search for IMBHs a set of Jeans models overplotted. A Our sample, so far, includes ten Galactic to more distant objects and larger sam- central black hole causes a rise in the globular clusters observed in 2010 and ples. This strategy will advance our velocity dispersion profile. In the case of 2011, including the pilot project target knowledge of globular cluster formation NGC 2808 no black hole is needed in Omega Centauri. All of them have been and evolution and bring us closer to an order to explain the observed data but in analysed and velocity maps have been understanding of black hole growth and the case of NGC 6388 the profile shows computed (see Figure 4 for all the clus- galaxy formation. a clear rise, making this cluster an ex­­ ters except Omega Centauri). Modelling cellent candidate for hosting an interme- all the clusters, however, is not straight- diate-mass black hole. forward as each cluster shows peculi­ References arities: some show very shallow kinematic Bahcall, J. N. & Wolf, R. A. 1976, ApJ, 209, 214 In addition to these analytical models, profiles, while others seem to have multi- Baumgardt, H., Makino, J. & Hut, P. 2005, N-body simulations are employed in order ple centres — one photometric and one ApJ, 620, 238 to reproduce the observations (Jalali kinematic. Ebisuzaki, T. et al. 2001, ApJ, 562, L19 et al., 2011). This approach is crucial in Fan, X. 2006, New A Rev., 50, 665 Ferrarese, L. & Merritt, D. 2000, ApJ, 539, L9 order to exclude alternative scenarios However, for most of the clusters we are Gebhardt, K. et al. 2000, AJ, 119, 1268 which could cause a kinematic signature confident that we can determine from Gebhardt, K. et al. 1992, in Bull. AAS, 24, 1188 similar to a black hole, such as a cluster the central kinematics whether they are Jalali, B. et al. 2011, MNRAS, 400, 1665 of dark remnants (e.g., neutron stars). likely to host an intermediate-mass black Kaufer, A. et al. 2003, The Messenger, 113, 15 Lützgendorf, N. et al. 2011, A&A, 533, A36 N-body simulations are ideally suited for hole or not and that good constraints on Noyola, E. et al. 2010, ApJ, 719, L60 globular clusters since their kinematics their masses can be provided. The two Pasquini, L. et al. 2002, The Messenger, 110, 1 are only determined by the motions of examples in Figure 5 show that in the Soria, R. et al. 2011, MNRAS, in press, arXiv: IIII.6783 their stars/particles. Assuming certain ini- case of NGC 6388 a model with a black tial conditions, the stars are distributed hole of ~ 17 000 MA is needed to repro- in space according to a mass function duce the data, while the velocity dis­ (i.e., luminosities) and are assigned veloc- persion profile in the core of NGC 2808

24 The Messenger 147 – March 2012