The Host Galaxies of the Brightest Quasars: Gas-Rich Galaxies, Mergers, and Young Stars

Reports from Observers

1 Frédéric Courbin sars (MV < −23.5). A new observational of ‘quasar envelopes’, and demonstrated Géraldine Letawe 2 approach is implemented, based on that they were made of gas and stars, Georges Meylan1 spatial deconvolution and decomposi- and that they were at the same redshift Pierre Magain 2 tion of the spectra. Our clean, deep, as the quasars. These pioneering obser- Pascale Jablonka1,3 spectra of quasar host galaxies show vations unambiguously established the Knud Jahnke 4 that the brightest quasars reside in connection between quasars and galax- Lutz Wisotzki 5 all types of galaxies, that their stellar ies. Yannick Letawe 2 populations are compatible with those Pierre North1 of young discs, even when the host The huge quantities of luminous energy galaxy is an elliptical one. Most of the escaping from quasars immediately leads host galaxies show prominent gas to questions about the properties of their 1 Ecole Polytechnique Fédérale de emission lines, whose metallicity sug- host galaxies. Are quasar host galaxies Lausanne (EPFL), Observatoire, gests a rather inefficient star forma- normal galaxies? Are they more massive? Sauverny, Switzerland tion. The source of ionisation of the gas Where does the matter fuelling the cen- 2 Institut d’Astrophysique et de Géo- is the quasar in 50 % of the objects. tral black hole come from? What are the physique, Université de Liège, Belgium All these objects are undergoing inter- physical processes responsible for the ig- 3 Observatoire de Genève, Sauverny, actions, suggesting that interactions nition of a quasar and what is its life-time, Switzerland play a major role in the ignition and fuel- if any typical value exists at all? The an- 4 Max-Planck-Institut für Astronomie, ling of the brightest quasars. swers to these questions rely not only on Heidelberg, Germany accurate observations of quasars them- 5 Astrophysikalisches Institut, Potsdam, selves, but also on the characterisation of Germany Quasars are among the most luminous their host galaxies. Most quasar host objects in the Universe. Powered by studies carried out over the last two de- the accretion of matter falling onto a cades concentrate on imaging and have 7 10 Because they are faint and hidden in 10 –10 MA supermassive black hole, led to the global picture that the most the glare of a much brighter unresolved they radiate at almost every wave- luminous quasars reside in large elliptical source, quasar host galaxies still chal- length of the electromagnetic spectrum, galaxies. The VLT spectroscopic obser- lenge the most powerful telescopes, from g-rays and X-rays to the radio vations described in this article support a instrumentation and processing tech- domain. Although, at the time of their dis- different picture, where galaxies of all niques. Determining their basic morpho- covery, quasars were found to be point types can harbour bright quasars and logical parameters and their integrat- sources, it was soon realised that they where gravitational interactions between ed colours is feasible, but difficult, from were in fact located in the centre of much galaxies or even mergers, play a major imaging alone. However, detailed in- larger, but also much fainter, envelopes. role in the ignition and fuelling of quasars. formation on their stellar and gas con- The envelopes of the most nearby qua- This was suggested already by the tents and on their dynamics is achiev- sars extend over a few arcseconds, mak- discovery of binary quasars (e.g., able with deep spectroscopy. We have ing it possible to obtain spectra of their Djorgovski et al. 1987, Meylan et al. 1990) completed such a deep spectroscopic external parts, even in moderate seeing but has never been investigated in detail study, with the VLT and FORS1 in MOS, conditions. Boroson et al. (1984) obtained for single quasars. for the host galaxies of 20 bright qua- the very first spectra of a small sample

HE 1503+0228 (z = 0.135) Deconvolved image with the quasar removed HST+ACS/HRC (F606W)

Figure 1: Example of an HST/ACS observation of a low-redshift quasar with its spiral host galaxy, HE 1503+0228, at z = 0.135. Even with the excellent resolution of the High-Resolution Channel of the ACS, the bright central quasar makes it difficult to see the internal parts of the galaxy. The image on the right has been spatially deconvolved and the quasar has been subtracted using the MCS de- 1 arcsec convolution algorithm. The spiral arms are now obvious even in the centre of the galaxy, as well as the bulge of the galaxy (Letawe et al. 2006b).

32 The Messenger 124 – June 2006 Figure 2: Example of spectra decomposition for the the quasar spectrum. The continuum and emis- z = 0.23 quasar HE 2345-2906, using the MCS sion lines of the host can be followed even ‘under- deconvolution algorithm (Magain et al. 1998, Courbin neath’ the quasar PSF. The height of the spec- et al. 2000). The top panel shows a portion of the trum is 19 arcseconds. The seeing during the obser- original VLT/FORS1 spectrum, after reduction and vations was 0.7 arcsecond and the spatial resolu- sky subtraction. The bottom panel shows the tion after deconvolution is 0.2 arcsecond (Letawe et spectrum of the host galaxy alone, after removal of al. 2006a). A VLT/FORS1 MOS study of quasar host galaxies

The high contrast between the central quasar and its host, and the small angular size of the host are the main limita- tions to a spectroscopic study. Quasar host galaxies are typically 2–3 mag- nitudes fainter than their central quasar. Their angular size is only a few arcsec- Hβ [O III][O III] onds at low redshift (z ~ 0.1–0.3). Fig- ure 1 gives an example of a spiral galaxy hosting a quasar. Even with excellent seeing or HST data, the observations re- main challenging, because the wings of the observed PSF, even a few arcseconds away from the quasar, still strongly contaminate the data. Two observational strategies are then possible to minimise the mutual contamination of the quasar and host spectra: (1) to carry out ‘off-axis the host spectrum by that of the quasar. about 50 % are undergoing interactions spectroscopy’, by placing the slit of With a mean signal-to-noise of 10–20 per to the point that it is hard to even infer a the spectrograph a few arcseconds away spectral resolution element, this spec- morphological type. While the fraction from the quasar in the hope that less trum allows us to measure several of the of quasar hosts with gravitational interac- quasar light will contaminate the host important Lick indices describing the tions is about the same as for normal galaxy, (2) to develop reliable decomposi- stellar content of the galaxy. The promi- galaxies, we show with our data (see be- tion techniques in order to separate the nent emission lines are used to infer the low) that the most powerful quasars individual spectrum of the quasar from ISM gas metallicity as well as the ionisa- are systematically in hosts with signs of that of its host galaxy. We make the latter tion source (stars versus central AGN) interactions. In one extreme case, the choice for our study, which considers responsible for the measured emission host galaxy may have been even com- ‘on-axis spectra’ of 20 bright quasars line ratios. Curved emission lines are pletely disrupted in a strong collision with with the VLT and FORS1. seen in the 2-D spectra of the host galaxy a galaxy (HE 0450-2958; Magain et al. (Figure 2), making it possible to measure 2005; see also ESO Press release 23/05). We adopt the Multi-Object-Spectroscopy its velocity field. mode (MOS) of FORS1 and devise a Aside from the extreme cases, our spec- new observational strategy in which one The variety in quasar and quasar-host troscopic sample allows us to infer gen- of the 19-arcsecond-long slitlets is used properties is such that many objects in eral characteristics of quasar hosts, as to observe the quasar while the other slit- our sample require slight modifications compared with their quiescent, non-AGN, lets are used to simultaneously observe of the general techniques used to analyse counterpart. While the full analysis and field stars used as PSFs. The simultane- the sample as a whole. While a few qua- interpretation are described in Letawe et ous observation of the object of scien- sar hosts are in isolated spiral galaxies al. (2006a), we summarise here the main tific interest and of several PSF stars al- (e.g. HE 1503+0228; Courbin et al. 2002), conclusions. lows us to use the spectroscopic version of the ‘MCS deconvolution algorithm’ 10 (Magain et al. 1998, Courbin et al. 2000), HE 2345-2906 which takes advantage of the spatial information contained in the spectrum of PSF stars in order to carry out a spatial 5 ) /Å deconvolution and decomposition of the 2 cm

spectra. This strategy was already pro- s/ QSO Figure 3: Integrated spectrum of HE 2345-2906. The g/

posed by Courbin et al. (1999, 2002). It er 0 top and bottom panels display respectively the 6

–1 HOST has now been used to decompose nu- 0 spatially deconvolved quasar and host spectra, inte- (1 merous VLT spectra of quasar hosts and grated in the spatial direction. Each spectrum is ob- Flux tained by using the three FORS1 grisms G600B,R,I, of galaxies lensing distant quasars (e.g., 0.5 (R ~ 400) in order to cover the full optical range. Eigenbrod et al. 2006). Figures 2–3 give The spectra are displayed in the rest frame. Note the an example of our spectra decomposi- quasar broad emission lines, that are seen narrow tion, applied to a FORS1 spectrum of the in the host galaxy. The Ha + N ii (6565 Å) emission 0 line is even resolved in the spectrum of the host, be- z = 0.23 quasar HE 2345-2906. Both fig- 4 000 5 000 6 000 7000 cause of the lower circular velocity than in the qua- ures indicate no residual contamination of Restframe Wavelength (Å) sar (Letawe et al. 2006a).

The Messenger 124 – June 2006 33 Reports from Observers Courbin F. et al., The Host Galaxies of the Brightest Quasars

Stellar populations 0.4 Figure 4: Lick Mg2 and Hb indices for 14 of the quasar host galaxies (red symbols), compared with the non-active galaxies of Trager et al. (1998) and Studies based on imaging only conclude Kennicutt (1992). Circles stand for elliptical galaxies, that almost all of the most luminous qua- 0.3 while triangles indicate later-type galaxies. Red open sars reside in massive elliptical galax- squares indicate the objects that contain gas ion- ies (e.g., Floyd et al. 2004, based on HST ised by the central AGN itself. Our values of the Lick indices match well those measured for late-type imaging). We measure in our spectra the 0.2 galaxies. The only two objects that have metallicities 2 Mg2, H and G4300 Lick indices, indic- compatible with elliptical galaxies are undergoing

b Mg ative of the stellar content of the galaxies. interactions (Letawe et al. 2006a).

The results are plotted in Figure 4 along 0.1 with the same measurements for two control samples of non-AGN galaxies.

Most objects have stellar populations that 0 compare well with late-type rather than early-type galaxies. Only two objects –20 –15 –10 – 5 0 5 have stellar populations consistent with Hβ their early-type morphology. Quasar host galaxies show on average bluer colours than those of non-AGN galaxies, in- dicative of an additional underlying young stellar population (e.g., Jahnke et al. 2004) that might be explained by recent gravitational interactions. The majority Figure 5: Emission-line ratios for 14 of the quasar Ionised by nucleus of the objects we study here shows stel- host galaxies. The solid line indicates the limit between objects with an ISM ionised by star formation 1 lar populations inconsistent with nor- and the objects with an ISM ionised by the central mal inactive isolated early-type galaxies quasar. Filled symbols represent objects with ob- or massive bulges, but rather shows vious signs of interactions, while open symbols are isolated objects. All objects with ionisation by the

spectral features more typical for young ] β 0.5 quasar are undergoing interactions. They also cor- /H discs. I II respond to the brightest quasars. Measurements

g [O Ionised by stars connected by a dashed line are for the same object lo but correspond to measurements in the central or Gas ionisation and metallicity external parts of the galaxy, where the ionisation 0 is found to be respectively by stars or by the AGN Disc (Letawe et al. 2006a). Emission-line ratios are well-known indi- Elliptical Undefined cators of the source of ionisation of the Undefined + YP

Interstellar Medium (ISM). We measure in –0.5 Figure 5 the [O iii]/H ratio as a function –0.8 –0.6 –0.4 –0.2 0 0.2 b log [N II/Hα] of the [N ii]/Ha ratio for 14 quasar host galaxies that have significant narrow emission lines. These two ratios compare high and low ionisation lines, hence allowing us to discriminate between high- energy ionisation sources and sources of lower energy. Galaxies where the ISM 45 Figure 6: Luminosity of the quasar in the Hb emission line (erg s−1) as a function of its [O iii] (5007 Å) is ionised by star formation are there- Disc Elliptical luminosity. The typical 1-s error bar is given in fore located, in such a diagram, at a dif- Undefined the lower right corner. The characteristics of the host ferent place than the galaxies where Undefined + YP galaxies are as stated in the legend, where open the ISM is ionised by the central quasar. symbols represent isolated galaxies, and filled sym- The result is striking: half of the galax- 44 bols are for galaxies that display signs of interactions. It is immediately seen that the quasars with the ) ies show the signature of an ISM ionised β

H brightest emission lines tend to reside in interacting by the quasar itself. All these are under- L( systems (Letawe et al. 2006a). g going interactions. In addition, the line ra- lo tios in these objects indicate that even 43 the gas located several kpc away from the centre of the quasar is ionised by the hard radiation field of the quasar. The objects are also the brightest of the sam- 42 ple, making it clear that interactions play 41 42 43 44 45 a major role in the ignition and fuelling of log L([O III])

34 The Messenger 124 – June 2006 Figure 7: Gas metallicities of the qua- the brightest quasars. Figure 6 adds 6 more support to this picture, by estab- Our sample of H II galaxies sar host galaxies (top), and of two 4 samples of metal-poor blue compact lishing a clear correlation between galaxies (middle panel), and metal- the luminosity of the central quasar in the 2 rich galaxies (bottom). Although our emission lines, and the fact that the sample is dominated by discs/spirals, quasar resides in an interacting system. 0 their metallicity remains low com- 7 7.5 8 8.5 9 9.5 pared with normal spirals, suggestive In Figure 6, all the bright quasars are of a low efficiency of star formation 15 undergoing interactions, while the faintest (Letawe et al. 2006a). Blue compact galaxies ones rather reside in disc-dominated 10 galaxies. Isolated elliptical galaxies host

only moderate-luminosity quasars. N 5

0 When the source of ionisation of the ISM 7 7.5 8 8.5 9 9.5 is stellar, it is possible to estimate the Star-Formation Rate (SFR) using the [O ii] 25 equivalent width. Although more objects 20 Spirals would be needed to draw a definite 15 conclusion, we note that three hosts out 10 of four with confirmed elliptical morphol- 5 0 ogy have significant [O ii] emission, in 7 7.5 8 8.5 9 9.5 contrast with non-AGN elliptical galaxies. 12 + log (O/H) In all the cases where it can be measured, the SFR is comparable to what is ing the presence of interactions. When with a neighbour. Indeed, the companion generally found in young discs. the host is a spiral galaxy, we fit a three- galaxy seen in the HST image is at the component velocity model including a same redshift as the quasar. It is however The metallicity of the gas traces how central point mass, a rotating disc, and already apparent from the FORS1 spectra galaxies have evolved. The hosts of our a dark matter halo. The masses we com- and from the HST images that interac- sample have a mean log (O/H)+12 of pute in that way are in the range 1011− tions are or have been present during the 12 about 8.4, i.e., in the low part of the met- 3 10 MA, integrated over a 10-kpc aper- history of this elliptical galaxy. allicity distribution of spiral galaxies ture, i.e., in good agreement with the (see Figure 7). Only two galaxies have a masses of normal, non-AGN galaxies. gas metallicity well compatible with the On the role of interactions in quasar mean value of 8.8, typical for spirals. This Not all symmetrical velocity curves can formation suggests that star formation has been be adequately fitted by three-component rather inefficient, and that the ignition of models, and other structures in the qua- Accurate image-processing techniques the quasar activity might therefore be re- sar host galaxies are found to mimic ro- combined with deep slit spectroscopy lated to an early stage of evolution of their tating discs when observed at such small yields a wealth of information that makes host. Note, however, that a statistical- angular sizes. The case of HE 1434-1600 it possible to unveil connections between ly more significant sample is necessary is a good example, where highly ionised the physical properties of quasars and to confirm this trend, as we have selected gas displaying a velocity field that can be of their host galaxies. the objects where we are sure that the mistakenly interpreted as circular motion, gas is ionised by star formation, i.e., a is in fact distributed in shell-like structures We find that bright quasars do not exclu- subsample dominated by a young stellar along the slit. However, accurate fit of the sively reside in large elliptical galaxies. population. velocity curve allows us to rule out cir- Half of the galaxies in our sample show cular rotation (Letawe et al. 2004), as con- stellar populations typical of young discs. firmed by HST imaging that unveils the Dynamics details of the gas distribution in the host. A second striking result is that most of Figure 8 shows an HST/ACS image that the host galaxies, including the ellipticals, The spectral resolution of the data is suf- we have obtained for HE 1434-1600. The contain large amounts of gas. This gas ficient to derive rough velocity information host galaxy of this quasar is classified has probably been brought into the ISM on the host galaxies, as can be noticed as an elliptical galaxy using ground-based through gravitational interactions. Indeed, already from Figure 2. This information is imaging. The HST image immediately the brightest quasars are systematical- valuable, not only in estimating the total confirms shell structures. By combining ly in interacting systems, independent of mass of the galaxies when unperturbed these data with 3D spectra, as can be their morphological type, when a mor- rotation is seen, but also to unveil signs of obtained with ARGUS (Figure 9), it will be phological type can be determined at all. interactions. Indeed, all objects that show possible to decide whether the shells irregular or even distorted shapes on are expanding away from the central qua- In all galaxies with interactions, the gas is the VLT acquisition images, also have dis- sar or whether they are residual materi- ionised by the central AGN rather than turbed velocity curves, hence confirm- al left in the host’s ISM after an encounter by star formation, even several kpc away

The Messenger 124 – June 2006 35 Reports from Observers Courbin F. et al., The Host Galaxies of the Brightest Quasars

Figure 8: HST image of the z = 0.144 from the quasar. This remains true for HE 1434-1600 (z = 0.144) quasars that have large reddening values, HST+ACS/HRC (F606W) quasar HE 1434-1600. The image has been taken through the F606W filter, hence ruling out the possibility that bright using the High-Resolution Channel quasars are seen brighter because of the Advanced Camera for Surveys the dust in their host has been destroyed onboard the HST. The slit of our during mergers or encounters. Companion galaxy FORS1 spectrum is about vertical in this image. The FLAMES/ARGUS field of view is also indicated. Note the Finally, we find that quasars that reside in obvious shell-like structures that may non-interacting galaxies appear to be expand away from the quasar. The in metal-poor but gas-rich spiral galaxies. companion galaxy to HE 1434-1600 is at the same redshift and is in gravi- These galaxies harbour the less luminous tational interaction with the quasar and quasars. its host galaxy (Letawe et al. 2004).

The general picture drawn from our study FLAMES/ARGUS field of view is that the brightest quasars are clear- (7 × 11 arcsec) ly related to gravitational interactions or even mergers between several galaxies. Aside from the clear case of the brightest HE 1434-1600 (z = 0.144) [O III] (5007 Å) Continuum quasars, the fainter quasars might form in the early stage of formation of their host galaxy, which are all metal-poor, i.e., galaxies where star formation has been rather inefficient so far. While we find very large amounts of gas in the latter galaxies, it is not clear what mechanism brings the gas to their central AGN. The mech- anisms involved are probably more ‘local’ than global mergers or interacting systems, e.g., the explosion of supernovae, gravitational instabilities or bar-driven fuelling. It will be interesting in the future to investigate the correlation between quasar luminosity and the presence of bars in the host galaxy.

On the technical side, our study shows that clean, deep, multi-slit spectroscopy is a powerful way to characterise high- contrast objects such as quasar host gal- 3 arcsec (7 kpc) axies. While integral-field spectroscopy is useful in order to map the ionisation state of the gas across the galaxy and to References Figure 9: ARGUS quasi-monochromatic images of measure the velocity field of the gas, ob- the z = 0.144 quasar HE 1434-1600. The fibre size Bahcall J. N. et al. 1997, ApJ 479, 642 in this image is 0.25 arcsecond, after subsampling of taining deep continuum spectra remains Boroson T., Oke J. B. 1984, ApJ 281, 535 the original data (which have 0.5 arcsecond fibres). only accessible to slit spectra in MOS. In Courbin F. et al. 1999, The Messenger 97, 26 The seeing was 0.6 arcsecond during the obser- addition to be much more sensitive, slit Courbin F. et al. 2002, A&A 394, 863 vations. The images display the two spectral regions spectrographs and their MOS capability Courbin F. et al. 2002, The Messenger 107, 28 centred on the [O iii] (5007 Å) emission line (right) Courbin F. et al. 2000, ApJ 529, 1136 and on a continuum region (left) with the same wave- provide easy access to the PSF informa- Djorgovski S. G. et al. 1987, ApJL 321, L17 length coverage. These data allow us to trace the tion mandatory to the removal of the Eigenbrod A. et al. 2006, A&A 451, 759 gas across the galaxy and to determine its ionisation central quasar. There is therefore a strong Floyd D. et al. 2004, MNRAS 355, 196 state from the central kpc to the outer parts. How- complementarity between 3D and slit Jahnke K., Kuhlbrodt B., Wisotzki L. 2004, ever, the stellar continuum remains only accessible MNRAS 352, 399 to slit spectra, hence pointing to the complemen- spectroscopy as far as quasar host gal- Kennicutt R. 1992, ApJS 79, 255 tarity between the two techniques. axies are concerned. The general trends Letawe G. et al. 2004, A&A 424, 455 unveiled by the present 20-object sam- Letawe G. et al. 2006a, submitted to MNRAS ple call for a more systematic study in- (astro-ph/0605288) Letawe Y. et al. 2006b, in prep cluding, e.g., all known bright low-redshift Magain P. et al. 2005, Nature 437, 381 quasars and Seyfert galaxies. Magain P., Courbin F., Sohy S. 1998, ApJ 494, 452 Meylan G. et al. 1990, The Messenger 59, 47 Trager S. C. et al. 1998, ApJS 116, 1

36 The Messenger 124 – June 2006