Spatial properties of AGN in hierarchical models Federico Marulli Dipartimento di Astronomia, Università di Bologna In collaboration with: Silvia Bonoli (MPA), Enzo Branchini (Roma Tre), Lauro Moscardini (Bologna), Roberto Gilli (Bologna), Volker Springel (MPA), Simon D. M. White (MPA), Francesco Shankar (MPA)

Contents Goals: * test scenarios in which the AGN activity is triggered by galaxy mergers * constrain different models for the lightcurves associated with individual quasar events How: * using semi-analytic models on top of the Millennium simulation that follow the joint formation of galaxies and their embedded supermassive black holes * studying the spatial properties of simulated AGN Why: * to estimate the AGN lifetime: if AGN are (not) strongly clustered, their hosts must be rare (common) objects and therefore they must also be long (short) events in order to account for the total quasar luminosity density * information on clustering had not been considered in the construction of the model, and therefore must be regarded as genuine model predictions Papers: * Marulli, Bonoli, Branchini, Gilli, Moscardini, Springel, 2009, MNRAS, 396,1404M * Bonoli, Marulli, Springel, White, Branchini, Moscardini, 2009, MNRAS, 396,423B * Marulli, Bonoli, Branchini, Moscardini, Springel, 2008, MNRAS, 385,1846M

Method: run high-resolution, cosmological simulations of the DM component alone and apply semi- analytic prescriptions in post-processing to model the diffuse galactic gas and its accretion onto the central BH Millennium simulation: GADGET-2 code at the Computing Centre of the Max-Planck Society in Garching (MPA): * Number of DM particles: ≃ 1e10 DM * Mass of particles: 8.6e8 Msun/h (DM halo of 0.1L ⋆ galaxies with ~ 100 particles) * Box side: 500 Mpc/h * Spatial resolution: co-moving scale of 5 kpc/h * Cosmology: LCDM * FOF haloes + SUBFIND subhaloes + merger trees Hybrid models: DM + galaxies + BH Springel at al.2005

BH seeds every newly-formed galaxy hosts a central BH of 1000 Msun/h (a larger seed would only influence the BH at very high redshifts, because the large growth factor soon cancels any information about the seed mass) BH and AGN: recipes BH mass accretion: The BH mass accretion is triggered by two different phenomena: the merger between gas-rich galaxies – quasar mode the cooling flow at the centres of X-ray emitting atmospheres in galaxy groups and clusters – radio mode Light curve:

BH and AGN: details Quasar mode Radio mode AGN luminosity Hopkins et al Croton et al. 2006; De Lucia&Blaizot 2007

BH scaling relations BH mass function BH fundamental plane AGN luminosity function Model vs Observations

AGN clustering function Upper panels: Two-point correlation function of DM particles (dotted line) compared with the correlation of the AGN Central panels: bias between AGN and dark matter as a function of scale Lower panels: two-point correlation from the upper panels divided by a power-law fit * At all luminosities, the simulated AGN two-point correlation function is fit well by a single power-law in the range 0. 5<r < 20 Mpc/h, but its normalization is a strong function of redshift * The bias is approximately scale-independent and its average value increases with redshift * These results are independent of the lightcurve model

Faint AGN and lightcurves Correlation length as a function of redshift of the AGN sample divided in four bolometric luminosity bins, compared with the correlation length of two mass bins of the Millennium FOF haloes. f_Edd=1 f_Edd(t) faint AGN (Lbol<1e10Lsun) While at high redshifts there is hardly any difference between the two models, at low redshifts the faint objects obtained with f_Edd(t) are much more strongly clustered. This is because most of the population is composed of large BHs that are accreting at low f_Edd and that are hosted by large haloes

Model vs Observations luminosity dependence of AGN clustering Correlation length and bias for the optical selected AGN: b(z)= (1+z)+0.25(1+z) 2 * at low and intermediate redshifts the correlation length and the bias depend weakly on luminosity when a narrow range of luminosities is examined * bright AGN are powered by BHs accreting close to the Eddington limit  it is difficult to use quasar clustering observations to disentangle between different light-curve models, unless much larger luminosity ranges are probed * the present observations indicate however that, over the range of luminosities observed, quasars reside in haloes of similar masses

Redshift evolution of the median mass of dark matter haloes hosting AGN of different luminosities L bol <L * (red) and L bol ~L * (green) Model vs Observations AGN environment * The typical halo mass hosting L * AGN grow up to z ≈ 1. 5 − 2, and then it decreases at higher redshifts  M halo = a 0 +a 1 z+a 2 z 2 +a 3 z 3 with a i =[ ;0. 944; − ; ] * The similar behaviour of haloes hosting L * AGN in both models suggests that L * objects are mainly BHs accreting close to the Eddington limit f_Edd=1 f_Edd(t)

AGN clustering function (Roncarelli et al. 2007) Space-like projections of a past light cone and of a mock field of view with the same selection effects as the CDFS

AGN number counts The predicted AGN number counts in the mock CDFs compared to the one determinated by Bauer et al. (2004). The left-hand and right-hand panels display the number counts of the AGN selected in the soft and hard X-ray bands, respectively. The dark and light grey shaded areas show the observed AGN counts obtained with two different classification schemes used to separate AGN from star-forming galaxies. Model predictions: the dashed black curves represent the median of all 100 CDF mocks and the bands indicate the 5th and 95th percentiles. Different colours characterize the different lightcurve models described in Section 2.2, as indicated by the labels

AGN clustering function Spatial two-point correlation function measured in 100 mock Chandra fields, as a function of the different lightcurve models adopted. The grey shaded areas have been computed using the best-fit power-law model with fixed slope by Gilli et al. (2005) to the CDF AGN real space correlation function. The values of r0 as a function of the median X- ray luminosity for the AGN in the mock CDFN and CDFS. The black dots are the results of Plionis et al. (2008). Model predictions have been obtained by averaging over 100 different mock catalogues for each lightcurve model.

Conclusions: general * We used the spatial distribution of AGN as a further test of our model and extracted several independent AGN mock catalogues that closely resemble the CDFS and CDFN * Independent of the model adopted for the lightcurve, the two point correlation function of our simulated AGN can be approximated by a single power-law in the range 0.5< r < 20 Mpc/h. The bias between AGN and the dark matter is a strong function of redshift, but, at a given epoch, it is approximately constant * The correlation length of faint AGN obtained assuming a time-dependent f Edd is consistent with the correlation length of −10 13 Mpc/h haloes, whereas faint AGN obtained with f Edd =1 exhibit the same clustering as −10 12 Mpc/h haloes * The lack of a significant dependence of clustering on luminosity is not primarily a result of invoking lightcurve models with a wide distribution of Eddington ratios, but rather arises because in a merger-driven scenario there is a small scatter in the typical halo mass hosting quasars close to their peak luminosity

Conclusions: model vs data * The clustering of our L ∗ quasar is in very good agreement with the most recent observational data. Since quasars at these luminosities are objects accreting at high Eddington ratios, we cannot use these results as a sensitive test of our lightcurve models * If one set the slope γ = 1.4, as in Gilli et al. (2005), then the correlation length agrees with the one measured in the CDFN, once cosmic variance is accounted for. On the contrary, the mock AGN in the CDFS are much less correlated than the real ones. In this case, the discrepancy in the correlation lenght is of the order of σ, depending on the lightcurve model adopted * The agreement between correlation functions in the XMM-COSMOS field (Gilli et al. 2009) and in the CDFN which, as we have shown, is well reproduced by our AGN models, suggests that the AGN clustering in the CDFS is indeed unusually high * The models predict that the clustering amplitude depends little on the luminosity of AGN, in disagreement with the strong dependence found by Plionis et al. (2008) but in agreement with the measurements of the clustering of luminous AGN in the XMM-COSMOS catalogue