The Astrophysical Journal, 625:L71L74, 2005 June 1

2005. The American Astronomical Society. All rights reserved. Printed in U.S.A.

A PHYSICAL MODEL FOR THE ORIGIN OF QUASAR LIFETIMES Philip F. Hopkins,1 Lars Hernquist,1 Paul Martini,1 Thomas J. Cox,1 Brant Robertson,1 Tiziana Di Matteo,2 and Volker Springel2
Received 2005 January 28; accepted 2005 April 15; 2005 May 9

ABSTRACT We propose a model of quasar lifetimes in which longer periods of rapid black hole accretion are distinguished from observationally inferred lifetimes of quasars owing to gas and dust obscuration. In our picture, quasars are powered by gas funneled to galaxy centers, fueling starbursts and feeding black hole growth, but are buried until feedback from the accretion disperses the obscuring material, creating a window in which the black hole is observable as an optical quasar. Eventually, the activity ceases when the accretion rate drops below that required to maintain quasar luminosities. We study this process by simulating galaxy mergers, using the gas density to infer the bolometric luminosity of the black hole and the gas metallicity and column density to determine the B-band attenuation along arbitrary lines of sight. Dening the visible quasar lifetime as the total time with an observed B-band luminosity greater than some lower limit L B, min , we nd lifetimes 1020 Myr for L B, min p 10 11 L , (MB 23), in good agreement with observations, but signicantly shorter than the intrinsic lifetime 100 Myr obtained if attenuation is neglected. The ratio of observed to intrinsic lifetimes is also a strong function of both the limiting luminosity and observed frequency range. Subject headings: cosmology: theory galaxies: active galaxies: evolution galaxies: nuclei quasars: general
1. INTRODUCTION

The correlation between the velocity dispersions of galaxy spheroids and the masses of their central, supermassive black holes (SMBHs; Ferrarese & Merritt 2000; Gebhardt et al. 2000) links black hole growth to galaxy formation. An important quantity underlying the evolution of SMBHs is the duration of rapid black hole growth, i.e., the quasar lifetime, tQ. Observations constrain tQ to the range tQ 10 610 8 yr (for a review see Martini 2004). These estimates are primarily based on arguments that combine observations of the populations of local SMBHs and high-redshift quasars (e.g., Soltan 1982; Yu & Tremaine 2002) or incorporate quasars into models of galaxy evolution (e.g., Kauffmann & Haehnelt 2000; Di Matteo et al. 2003; Granato et al. 2004). Clustering measured in quasar surveys (e.g., Porciani et al. 2004; Grazian et al. 2004) also gives tQ 10 7 yr. These estimates constrain only the total time an SMBH radiates at quasar luminosities and do not address scenarios involving multiple short, episodic quasar phases. A minimum episodic lifetime tQ 10 5 yr is required to explain the proximity effect in the Lya forest (Bajtlik et al. 1988), but observations of the proximity effect at high redshifts (Haiman & Cen 2002; Yu & Lu 2005) and the transverse proximity effect in He ii (Jakobsen et al. 2003) also imply tQ 10 7 yr. If the net and episodic lifetimes are both 10 7 yr, quasars could be produced by relatively rare major mergers that produce rapid inows of gas (e.g., Barnes & Hernquist 1996; Mihos & Hernquist 1996). Semianalytical models suggest that beyond a certain threshold, feedback from SMBH growth expels gas and shuts down accretion (Silk & Rees 1998; Ciotti & Ostriker 2001; Fabian 1999; Wyithe & Loeb 2003). However, these models do not predict the characteristic lifetime of the accretion phase prior to its self1 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138. 2 Max-Planck-Institut fu r Astrophysik, Karl-Schwarzchild-Strasse 1, 85740 Garching bei Mu nchen, Germany.

termination but take it to be a free parameter. The lifetime is either matched to observations or timescales such as the galaxy dynamical time or the e-folding time for Eddington-limited black p 4 # 10 8el yr for accretion with rahole growth tS p MBH /M 2 0.1 and l p L /L Edd 1 (Saldiative efciency e p L /Mc peter 1964). Simulations of galaxy mergers that include black holes (Di Matteo et al. 2005; Springel et al. 2005) show that black hole growth is driven by gas inows and terminates abruptly when signicant gas is expelled in response to feedback from accretion. This determines a duration for the strong accretion phase of 10 8 yr, approximately equal to the gas inow timescale set when strong torques are present in the inner regions, and substantially shorter than the merger timescale of 2 # 10 9 yr. The accretion rate and obscuring column both increase with the gas density, so the period of most rapid black hole growth is likely to be heavily obscured. We propose that quasars originate in mergers but are heavily obscured for long periods by the high gas densities powering them. Eventually, feedback drives away the gas, creating a brief window in time in which the black hole is visible as a quasar, until accretion levels drop below quasar thresholds. We measure quasar lifetimes in simulations of galaxy mergers that incorporate black hole growth. We determine the intrinsic lifetime of the strong accretion phase, and, by accounting for the effects of obscuration, we trace column densities along different lines of sight to the SMBH and infer a lifetime for which the active galactic nucleus (AGN) would be seen as a quasar. We nd that the intrinsic lifetime is much longer than the period of visible quasar activity and that our predicted quasar lifetimes agree well with observations. This effect can account for the presence of an obscured population of quasars that are missed by optical, UV, or soft X-ray surveys but contribute signicantly to the cosmic X-ray background (Brandt & Hasinger 2005 and references therein), and reprocessing of quasar radiation by dust can explain observations of luminous and ultraluminous infrared galaxies that seem to contain obscured AGNs (e.g., Komossa et al. 2003). L71

L72

HOPKINS ET AL.

Vol. 625

galaxy is initially composed of 168,000 dark matter halo particles, 8000 bulge particles, 24,000 gaseous disk particles, and one BH particle, with a small initial seed mass of 10 5 M,. For these choices, the dark matter, gas, and star particles are all of roughly equal mass. The galaxies are then set to collide in a pure prograde encounter with zero orbital energy and a pericenter separation of 7.1 kpc.
3. COLUMN DENSITIES AND QUASAR ATTENUATION

Fig. 1.Projected gas density is shown in boxes 140 kpc on a side, colorcoded by temperature (blue to red, upper panels). The bolometric luminosity of the central black hole(s) is shown in the lower panel, with diamonds marking the times shown above. Bolometric luminosities prior to the merger are the sum of the two individual black hole luminosities.

2. THE SIMULATION

We employ a new version of the parallel TreeSPH code GADGET (Springe et al. 2001), which uses an entropy-conserving formulation of smoothed particle hydrodynamics (SPH; Springel & Hernquist 2002) and includes a subresolution, multiphase model of the dense interstellar medium (ISM) to describe star formation (Springel & Hernquist 2003). The multiphase gas is pressurized by feedback from supernovae, allowing us to stably evolve even pure gas disks (see, e.g., Robertson et al. 2004). SMBHs are represented by sink particles that accrete gas, with an accretion estimated using a Bondi-Hoyle-Lyttleton parameterization, rate M with an upper limit equal to the Eddington rate (Springel et al. 2005). The bolometric luminosity of the SMBH is then L bol p 2, where e p 0.1 is the radiative efciency. We further allow eMc a small fraction (5%) of L bol to couple dynamically to the gas as thermal energy. This fraction is a free parameter, determined in Di Matteo et al. (2005) by tting the MBH-j relation. We do not attempt to resolve the gas distribution immediately around the black hole but instead assume that the time-averaged accretion can be estimated from the gas on the scale of our spatial resolution. We generate two stable, isolated disk galaxies, each with an extended dark matter halo having a Hernquist (1990) prole, an exponential disk, and a bulge. Our simulation is one of the series described in detail in Springel et al. (2005), with virial velocity Vvir p 160 km s1, a ducial choice with a rotation curve and mass similar to the Milky Way. We begin our simulation with pure gas disks, which may correspond to the highredshift galaxies in which most quasars are observed. Each

We calculate the column density between a black hole and a hypothetical observer as follows. We rst generate 1000 radial lines of sight (rays), each with its origin at the black hole and directions uniformly spaced in solid angle. For each ray, we begin at the origin, calculate and record the local gas properties using GADGET, and then move a distance along the ray Dr p hhsml, where h 1 and hsml is the local SPH smoothing length. This process is repeated until a ray is sufciently far from its origin (100 kpc). The gas properties along a given ray can then be integrated to give the line-of-sight column density and metallicity. We test different values of h and nd that gas properties along a ray converge rapidly and change smoothly for h 0.5. We similarly test different numbers of rays and nd that the distribution of line-of-sight properties converges for 100 rays. Given the local gas properties, we use the multiphase structure of the ISM described in Springel & Hernquist (2003) to calculate the mass fraction in hot (diffuse) and cold (molecular and H i cloud) phases of dense gas, and, assuming pressure equilibrium, we obtain the densities of the two phases and their corresponding volume lling factors. Using only the hot-phase gas allows us to place a lower limit on the column density along a particular line of sight, as it assumes a ray passes only through the diffuse ISM, with 90% of the mass of the dense ISM concentrated in cold clumps. Given the small volume lling factor (!0.01) and cross section of cold clouds, we expect the majority of sight lines to pass only through the hot phase, with rare outliers dominated by single clouds along the line of sight (covering fraction 1%). 2, we model the intrinsic quasar continuum Using L bol p eMc spectral energy distribution following Marconi et al. (2004), based on optical through hard X-ray observations (e.g., Elvis et al. 1994; George et al. 1998; Vanden Berk et al. 2001; Perola et al. 2002; Telfer et al. 2002; Ueda et al. 2003; Vignali et al. 2003). This gives a B-band luminosity log ( L B ) p 0.80 0.067L 0.017L2 0.0023L3, where L p log ( L bol /L , ) . We then adopt a gas-to-dust 12, and we take lB p 4400 A ratio to determine the extinction wavelength. Observations suggest that the majority of quasars have reddening curves similar to that of the Small Magellenic Cloud (SMC; Hopkins et al. 2004), which has a gas-to-dust ratio lower than the Milky Way by approximately the same factor as its metallicity (Bouchet et al. 1985). We consider both gas-to-dust ratios equal to that of the Milky Way, ( AB /NH ) MW p 8.47 # 1022 cm2, and scaled by metallicity, AB /NH p ( Z /0.02)( AB /NH ) MW. For both cases we use the SMC-like reddening curve of Pei (1992).
4. RESULTS

Figure 1 shows the gas at four representative times during the run, with the bolometric luminosity of the SMBH as a function of time below. After the rst passage (upper left), there is an extended period of strong accretion, but central gas densities are very large and the intrinsic SMBH luminosity is attenuated well below our quasar threshold of L B, obs 10 11 L , (MB 23).

No. 2, 2005

MODEL FOR QUASAR LIFETIMES

L73

Fig. 2.Top panel: Thick contours plot the median column density NH as a function of simulation time, with thin contours at 25% and 75% inclusion levels. Solid contours represent the density of the hot-phase ISM, and dashed contours represent the total simulation density. Middle panel: Bolometric lu 2 (thick line), and ratio of bolometric minosity of the black hole, Lbol p eMc to Eddington luminosity, l { Lbol/LEdd (thin line). Bottom panel: Observed Bband luminosity calculated given the median hot-phase ISM density. Diamonds mark times shown in Figure 1.

Fig. 3.Upper panel: Histograms of the observed quasar lifetime tQ, dened as an observed B-band luminosity LB, obs 1011 L,, for the simulation along different lines of sight to the black hole, given the hot-phase ISM density. The thick line shows the lifetime if attenuation is ignored, tQ 113 Myr. Lower panel: Same as above, but with lifetime tQ dened as an inferred bolometric luminosity (estimated from LB, obs) Lbol 1 0.1LEdd.

Once the merger begins (upper right), the intrinsic luminosity peaks as gas is channeled to the center. However, this also produces very large columns that heavily obscure the quasar. After a short time (lower left), feedback disperses the central gas and the object can be seen as a quasar. In this stage, rapid black hole growth takes place and most of the nal mass is accumulated. Shortly after the merger (lower right), gas has been expelled or consumed by star formation and accretion, and densities have dropped well below the levels needed to maintain quasar activity. Figure 2 shows the bolometric luminosity of and column densities to the SMBH as a function of time, shortly before and after the merger (1.27 Gyr). Median column densities and 25%75% contours are shown for the total and hot-phase densities, as described in 3. The metallicity-weighted and nonweighted column densities are similar because the column is dominated by gas in the central, most dense regions, where star formation rates (and thus metallicities) are large. The bolometric luminosity rises rapidly above 10 12 L , during the merger, but column densities simultaneously become very large, up to 10 25 cm2 (total) or 10 23 cm2 (hot phase). These correspond to the large columns seen toward obscured quasars and suggest that the quasar will not be observable at visible wavelengths for a signicant fraction of this period. Quasar lifetimes are shown in Figure 3. For each sight line, the observed lifetime is the integrated time during which an observer sees a B-band luminosity above a given threshold, L B, obs 10 11 L ,. The histogram plots the fraction of sight lines with a particular observed quasar lifetime, tQ , spaced in logarithmic bins. Lifetimes are calculated using the metallicityweighted hot-phase density. The total column density would attenuate the quasar below this limit along all sight lines at all times. The line at tQ 1.1 # 10 8 yr shows the intrinsic lifetime along all sight lines if attenuation is ignored. We note that below 5 Myr (the simulation output frequency) our estimates of tQ become uncertain owing to the effects of variability and our inability to resolve the small-scale physics of the ISM. However, the more reliable estimates of a typical observation, from

the hot-phase column density, lie well above this limit. The majority of such sight lines see lifetimes in the range 10 20 Myr, in good agreement with observations. The longest observable lifetimes in this case are approximately 20 Myr, 1 6 of the duration of the intrinsic strong accretion phase. We also show the observed lifetime with an inferred bolometric luminosity (calculated from the observed L B, obs using the bolometric corrections dened in 3) greater than 0.1L Edd, where L Edd is the Eddington luminosity for the black hole mass at a given time (L Edd p 6 # 10 12 L , for the nal black hole mass 1.9 # 10 8 M,), and nd similar lifetimes. In Figure 4 we plot tQ as a function of the limiting B-band luminosity L B, min. The uppermost curve shows the intrinsic i lifetime tQ , the total time the intrinsic L B L B, min . The bottom curves show the integrated time that the observed B-band luminosity meets this criterion, using the total metallicityweighted column density. The solid curve shows the integrated time, weighted by the fraction of sight lines f along which the above denition is met [tQ p f ( L B, obs L B, min ) dt]. The dotted and dashed curves show the integrated time during which 50% and 10% of sight lines meet the criterion, respectively. The middle curves show the same, using the hot-phase density. In all cases, changing the denition of tQ based on the fraction of sight lines with a given column does not signicantly change tQ. However, tQ is, unsurprisingly, a strong function of L B, min. The observed tQ is signicantly smaller than the intrinsic lifei time for all L B, min 1 10 9 L ,, and the ratio tQ /tQ decreases with increasing L B, min. We also plot the lifetime as a function of l, the ratio of the inferred bolometric luminosity to the Eddington luminosity, and nd a similar trend.
5. CONCLUSIONS

Incorporating black hole feedback and obscuration in galaxy mergers gives observed quasar lifetimes of 10 7 yr, in good agreement with observations. For a signicant fraction of the intrinsic timescale for black hole growth, the quasar is heavily obscured, and attenuated well below observable limits in the B band. The signicant difference between the observed and

L74

HOPKINS ET AL.

Vol. 625

Fig. 4.Upper panel: Quasar lifetimes as a function of LB, min. The thick line shows the lifetime if attenuation is ignored (time with observed B-band luminosity LB, obs LB, min). The lower three lines use the total ISM density, while the middle three use the hot-phase density. Solid lines represent the integrated lifetime weighted by the fraction of sight lines along which the denition is satised. Dashed lines represent the lifetime for which 10% of sight lines meet the denition, dotted lines the limetime for which 50% meet the denition. Lower panel: Same as above, but with tQ dened as an inferred bolometric luminosity (estimated from LB, obs ) Lbol 1 lLEdd. The total ISM density attenuates the quasar below the limits shown, so only the hot-phase results are shown.

intrinsic lifetimes is critical for modeling the quasar luminosity function, the present-day population of SMBHs, or the cosmic X-ray background. Moreover, we predict that the observed lifetime is itself a function of the observed frequency, an effect that must be considered in extending an observed quasar luminosity function to other wavelengths or using observational constraints of quasar lifetimes to study the infrared and X-ray backgrounds. Estimates of the fraction of starbursting galaxies with nuclear activity must similarly combine multiwavelength observations as optical or IR observations alone may miss a large fraction of AGNs. Provided that self-regulating feedback mechanisms terminate rapid black hole growth at a certain mass, the relationship between observed and intrinsic quasar lifetimes should remain qualitatively similar under a wide range of conditions, even if collimation or other effects alter the absolute time when growth or obscuration terminates. However, we defer a full treatment of the statistics of quasar lifetime distributions as a function of galaxy sizes, merger conditions, and feedback mechanisms to a later paper. Using this model, a wider examination of simulations will allow for a consistent prediction of quasar lifetimes as a function of observed wavelength, redshift, and host galaxy properties.

This work was supported in part by NSF grants AST 0206299 and AST 03-07690.

REFERENCES Barnes, J. E., & Hernquist, L. 1996, ApJ, 471, 115 Bajtlik, S., Duncan, R. C., & Ostriker, J. P. 1988, ApJ, 327, 570 Bouchet, P., Lequeux, J., Maurice, E., Prevot, L., & Prevot-Burnichon, M. L. 1985, A&A, 149, 330 Brandt, W. N., & Hasinger, G. 2005, ARA&A, in press (astro-ph/0501058) Ciotti, L., & Ostriker, J. P. 2001, ApJ, 551, 131 Di Matteo, T., Croft, R. C, Springel, V., & Hernquist, L. 2003, ApJ, 593, 56 Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604 Elvis, M., et al. 1994, ApJS, 95, 1 Fabian, A. C. 1999, MNRAS, 308, L39 Ferrarese, L., & Merritt, D. 2000, ApJ, 539, L9 Gebhardt, K., et al. 2000, ApJ, 539, L13 George, I. M., Turner, T. J., Netzer, H., Nandra, K., Mushotzky, R. F., & Yaqoob, T. 1998, ApJS, 114, 73 Granato, G. L., De Zotti, G., Silva, L., Bressan, A., & Danese, L. 2004, ApJ, 600, 580 Grazian, A., Negrello, M., Moscardini, L., Cristiani, S., Haehnelt, M. G., Matarrese, S., Omizzolo, A., & Vanzella, E. 2004, AJ, 127, 592 Haiman, Z., & Cen, R. 2002, ApJ, 578, 702 Hernquist, L. 1990, ApJ, 356, 359 Hopkins, P. F., et al. 2004, AJ, 128, 1112 Jakobsen, P., Jansen, R. A., Wagner, S., & Reimers, D. 2003, A&A, 397, 891 Kauffmann, G., & Haehnelt, M. 2000, MNRAS, 311, 576 Komossa, S., Burwitz, V., Hasinger, G., Predehl, P., Kaastra, J. S., & Ikebe, Y. 2003, ApJ, 582, L15 Marconi, A., Risaliti, G., Gilli, R., Hunt, L. K., Maiolino, R., & Salvati, M. 2004, MNRAS, 351, 169 Martini, P. 2004, in Coevolution of Black Holes and Galaxies, ed. L. C. Ho (Cambridge: Cambridge Univ. Press), 170 Mihos, J. C., & Hernquist, L. 1996, ApJ, 464, 641 Pei, Y. C. 1992, ApJ, 395, 130 Perola, G. C., Matt, G., Cappi, M., Fiore, F., Guainazzi, M., Maraschi, L., Petrucci, P. O., & Piro, L. 2002, A&A, 389, 802 Porciani, C., Magliocchetti, M., & Norberg, P. 2004, MNRAS, 355, 1010 Robertson, B., Yoshida, N., Springel, V., & Hernquist, L. 2004, ApJ, 606, 32 Salpeter, E. E. 1964, ApJ, 140, 796 Silk, J., & Rees, M. J. 1998, A&A, 331, L1 Soltan, A. 1982, MNRAS, 200, 115 Springel, V., Di Matteo, T., & Hernquist, L. 2005, MNRAS, in press (astroph/0411108) . 2005, ApJ, 620, L79 Springel, V., & Hernquist, L. 2002, MNRAS, 333, 649 . 2003, MNRAS, 339, 289 Springel, V., Yoshida, N., & White, S. D. M. 2001, NewA, 6, 79 Telfer, R. C., Zheng, W., Kriss, G. A., & Davidsen, A. F. 2002, ApJ, 565, 773 Ueda, Y., Akiyama, M., Ohta, K., & Miyaji, T. 2003, ApJ, 598, 886 Vanden Berk, D. E., et al. 2001, AJ, 122, 549 Vignali, C., Brandt, W. N., & Schneider, D. P. 2003, AJ, 125, 433 Wyithe, J. S. B., & Loeb, A. 2003, ApJ, 595, 614 Yu, Q., & Lu, Y. 2005, ApJ, 620, 31 Yu, Q., & Tremaine, S. 2002, MNRAS, 335, 965