Termination of star formation by BH feedback in equal- and unequal-mass mergers of disk and elliptical galaxies
aa r X i v : . [ a s t r o - ph ] S e p Astron. Nachr. / AN , No. 88, 789 – 792 (2006) /
DOI please set DOI!
Termination of star formation by BH feedback in equal- and unequal-mass mergers of disk and elliptical galaxies
Peter H. Johansson ,⋆ , Thorsten Naab , Andreas Burkert University Observatory Munich, Scheinerstr. 1, D-81679 Munich, GermanyReceived 22 August 2008Published online later
Key words galaxies: interaction– galaxies: active galaxies: evolution – galaxies: formation – methods: numericalWe present binary galaxy merger simulations of gas-rich disks (Sp-Sp), of early-type galaxies and disks (E-Sp, mixedmergers), and mergers of early-type galaxies (E-E, dry mergers) with varying mass ratios and different progenitor mor-phologies. The simulations include radiative cooling, star formation and black hole (BH) accretion and the associatedfeedback processes. We find for Sp-Sp mergers, that the peak star formation rate and BH accretion rate decrease and thegrowth timescales of the central black holes and newly formed stars increase with higher progenitor mass ratios. The ter-mination of star formation by BH feedback in disk mergers is significantly less important for higher progenitor mass ratios(e.g. 3:1 and higher). In addition, the inclusion of BH feedback suppresses efficiently star formation in dry E-E mergersand mixed E-Sp mergers. c (cid:13) Recent large-scale statistical galaxy surveys (e.g. the SloanDigital Sky Survey (SDSS) and the two-degree Field (2df)survey) have revealed a robust bimodality in the observedgalaxy population. Galaxies above a critical stellar mass of M crit ≃ × M ⊙ are typically non-star forming redspheroidal systems with old stellar populations that predom-inately live in dense environments, whereas galaxies belowthis critical mass are typically blue, star-forming disk galax-ies that lie in the field (e.g. Bell et al. 2003; Kauffmann etal. 2003; Baldry et al. 2004; Balogh et al. 2004).The observed bimodality can also be seen as a mani-festation of cosmic downsizing, in which the most massivegalaxies formed a significant proportion of their stars at highredshifts above z & . The star formation is efficientlyquenched in these systems by z = 1 as manifested in theobserved population of extremely red non-starforming mas-sive ellipticals (EROs, e.g. McCarthy 2004; V¨ais¨anen & Jo-hansson 2004). The lower mass systems, on the contrary,have typically been forming stars throughout the whole cos-mic epoch (e.g. Glazebrook et al. 2004; Juneau et al. 2005;Cimatti, Daddi, Renzini 2006).The quenching mechanism responsible for the observedtermination of star formation in massive galaxies at redshiftsbelow z . − needs to be both energetic enough to trig-ger the quenching and long-lasting enough to maintain thequenching over a Hubble time. There exists several theoret-ical explanations for the quenching, including the feedbackfrom AGNs (e.g. Bower et al. 2006; Croton et al. 2006),gaseous major mergers triggering star burst and/or quasar ⋆ Corresponding author: [email protected] activity (e.g Naab, Jesseit, Burkert, 2006), quenching byshockheated gas above a critical halo mass (Dekel & Birn-boim 2006; Birnboim, Dekel, Neistein 2007) and gravita-tional quenching by clumpy accretion (Naab et al. 2007;Dekel & Birnboim 2008).In this paper we study the termination of star forma-tion by black hole (BH) feedback in binary galaxy merg-ers. The study by Springel, Di Matteo, Hernquist (2005a)showed that BH feedback efficiently terminates star forma-tion in equal-mass disk mergers. Here we confirm this resultand expand the analysis to include unequal-mass disk merg-ers, mixed E-Sp mergers and dry E-E mergers. This paper isstructured as follows. In § § § We perform the simulations using the parallel TreeSPH-code GADGET-2 (Springel 2005) including star formationand the associated supernova feedback, where the multi-phase interstellar medium (Efstathiou 2000; Johansson &Efstathiou 2006) is modeled using the sub-resolution mo-del developed by Springel & Hernquist (2003). We imple-mented a black hole feedback description similar to the Sp-ringel et al. (2005b) model into our version of the code. Inthis model the BH accretion rate is parameterized using aBondi-Hoyle-Lyttleton description, with the maximum ac-cretion rate limited to the Eddington rate. Finally, some frac-tion of the accreted rest mass energy is available as thermal c (cid:13)
90 Johansson et al.: Termination of star formation by BH feedback in galaxy mergers
Fig. 1
The total star formation rate (top), the total blackhole accretion rate (middle) and the evolution of the totalblack hole mass (bottom) as a function of time for three 3:1(solid lines) and one 1:1 (dashed line) merger with initialgas mass fractions of 20% (black), 40% (green) and 80%(red) (left panel) and for three 3:1 mergers with varying ini-tial gas mass fractions (black, green, red lines) and three3:1 mergers with varying orbital and initial geometries for afixed gas fraction (blue lines) (right panel) . The filled circlesindicate the time of merging of the BHs.feedback energy that couples to the surrounding gas. Fur-ther details about the code implementation and parameterchoices can be found in Johansson, Naab, Burkert (2008),J08 hereafter.The disk galaxies are setup as described in J08, withHernquist dark matter profiles populated by exponential dis-ks with initial gas fractions ranging from 20% to 80%. Allmodels contain initially a bulge with a third of the total diskmass. We also model mergers of early-type galaxies, wherethe early-type galaxies are setup initially using merger rem-nants of binary disk mergers (Naab & Burkert 2003). Weset the primary galaxy models up with 20,000 gas and stellardisk particles, 10,000 bulge particles and 30,000 dark matterparticles and scale the other models correspondingly. Thegravitational softening lengths of gas, newly formed starsand the black hole particles are set to ǫ = 0 . h − kpc andthe softening lengths of the disk, bulge and more massivedark matter particles are scaled with the square root of theparticle masses resulting in ǫ = 0 . h − kpc for the bulgeand disk particles and ǫ = 0 . h − kpc for the dark matterparticles, respectively.The galaxies are initially placed on parabolic orbits wh-ere the initial separation of the progenitors is set to the meanof the two galaxy virial radii and the pericentric distance tothe mean of the two disk scale radii. All simulations pre-sented in this paper were evolved for a total of t = 3 Gyr using the local Altix 3700 Bx2 machine hosted at the Uni-versity Observatory in Munich. In the left panel of Fig. 1 we plot the total star formationrates, BH accretion rates and the BH mass growth for three3:1 mergers (solid lines) with 20% (black), 40% (green),80% (red) initial gas mass fractions together with one 20%gas fraction 1:1 merger (dashed lines). The star formation isvery efficiently terminated by the BH feedback in 1:1 merg-ers, compared to a generally much shallower decline in starformation for 3:1 mergers. In addition, the final BH massare lower in 3:1 mergers typically by a factor of 2-5, butwith a relatively large scatter depending on the progenitormasses and initial gas fractions.In the right panel of Fig. 1 we show the results for three3:1 mergers with varying initial gas mass fractions for afixed orbit and initial disk geometry (black, green, red lines)and for three 3:1 mergers with varying initial orbits and ori-entations for a fixed gas mass fraction (blue lines). The ini-tial gas fraction has a large effect on the height of the starformation and BH accretion peaks, with larger initial gasfractions producing higher values, as expected. This resultsalso in relatively large differences in the final BH masses,with the f gas = 0 . simulations producing final BH massesthat are larger by an order of magnitude compared to the f gas = 0 . runs. The variation of the orbit and initial ge-ometry for a fixed gas mass fraction produces much smallerdifferences. The peaks of the star formation rates and BHaccretion rates only vary within a factor of two with chang-ing orbits and initial disk geometries.In Fig.2 we study the star formation and BH accretionhistories for unequal-mass mergers with varying mass ra-tios. To this end we ran five mergers with mass ratios of1:1, 2:1, 3:1, 4:1 and 6:1 on co-planar prograde orbits withinitial gas mass fractions of 20%. As can be seen in Fig. 2increasing the mass ratio of the merger systematically low-ers the peak star formation rate and increases the durationof star formation activity after the merger. For the highestmass ratio merger the star formation rate is virtually con-stant throughout the simulation with only a mild peak dur-ing the first passage (see also di Matteo et al. 2007). The fi-nal BH masses are systematically lower for increasing massra-tio of the merger. Furthermore, the slope of the M BH growth as a function of time becomes shallower with in-creasing merger mass ratio. There is also a systematic delayin the time of the BH merger with increasing mass ratio, asindicated by the filled circles in Fig. 2. For the lower massratio mergers the peak of the BH activity (the filled trianglesin Fig. 2) typically occurs shortly after the merging of theBHs, whereas for the higher mass ratio mergers the peak ofthe BH activity is not directly related to the merging time ofthe BHs.In Fig. 3 we study the duration of the BH accretion andstar formation activity as a function of merger mass ratio. c (cid:13)2006 WILEY-VCH Verlag GmbH&Co. KGaA, Weinheim
The total star formation rate (top), the total blackhole accretion rate (middle) and the evolution of the totalblack hole mass (bottom) as a function of time for three 3:1(solid lines) and one 1:1 (dashed line) merger with initialgas mass fractions of 20% (black), 40% (green) and 80%(red) (left panel) and for three 3:1 mergers with varying ini-tial gas mass fractions (black, green, red lines) and three3:1 mergers with varying orbital and initial geometries for afixed gas fraction (blue lines) (right panel) . The filled circlesindicate the time of merging of the BHs.feedback energy that couples to the surrounding gas. Fur-ther details about the code implementation and parameterchoices can be found in Johansson, Naab, Burkert (2008),J08 hereafter.The disk galaxies are setup as described in J08, withHernquist dark matter profiles populated by exponential dis-ks with initial gas fractions ranging from 20% to 80%. Allmodels contain initially a bulge with a third of the total diskmass. We also model mergers of early-type galaxies, wherethe early-type galaxies are setup initially using merger rem-nants of binary disk mergers (Naab & Burkert 2003). Weset the primary galaxy models up with 20,000 gas and stellardisk particles, 10,000 bulge particles and 30,000 dark matterparticles and scale the other models correspondingly. Thegravitational softening lengths of gas, newly formed starsand the black hole particles are set to ǫ = 0 . h − kpc andthe softening lengths of the disk, bulge and more massivedark matter particles are scaled with the square root of theparticle masses resulting in ǫ = 0 . h − kpc for the bulgeand disk particles and ǫ = 0 . h − kpc for the dark matterparticles, respectively.The galaxies are initially placed on parabolic orbits wh-ere the initial separation of the progenitors is set to the meanof the two galaxy virial radii and the pericentric distance tothe mean of the two disk scale radii. All simulations pre-sented in this paper were evolved for a total of t = 3 Gyr using the local Altix 3700 Bx2 machine hosted at the Uni-versity Observatory in Munich. In the left panel of Fig. 1 we plot the total star formationrates, BH accretion rates and the BH mass growth for three3:1 mergers (solid lines) with 20% (black), 40% (green),80% (red) initial gas mass fractions together with one 20%gas fraction 1:1 merger (dashed lines). The star formation isvery efficiently terminated by the BH feedback in 1:1 merg-ers, compared to a generally much shallower decline in starformation for 3:1 mergers. In addition, the final BH massare lower in 3:1 mergers typically by a factor of 2-5, butwith a relatively large scatter depending on the progenitormasses and initial gas fractions.In the right panel of Fig. 1 we show the results for three3:1 mergers with varying initial gas mass fractions for afixed orbit and initial disk geometry (black, green, red lines)and for three 3:1 mergers with varying initial orbits and ori-entations for a fixed gas mass fraction (blue lines). The ini-tial gas fraction has a large effect on the height of the starformation and BH accretion peaks, with larger initial gasfractions producing higher values, as expected. This resultsalso in relatively large differences in the final BH masses,with the f gas = 0 . simulations producing final BH massesthat are larger by an order of magnitude compared to the f gas = 0 . runs. The variation of the orbit and initial ge-ometry for a fixed gas mass fraction produces much smallerdifferences. The peaks of the star formation rates and BHaccretion rates only vary within a factor of two with chang-ing orbits and initial disk geometries.In Fig.2 we study the star formation and BH accretionhistories for unequal-mass mergers with varying mass ra-tios. To this end we ran five mergers with mass ratios of1:1, 2:1, 3:1, 4:1 and 6:1 on co-planar prograde orbits withinitial gas mass fractions of 20%. As can be seen in Fig. 2increasing the mass ratio of the merger systematically low-ers the peak star formation rate and increases the durationof star formation activity after the merger. For the highestmass ratio merger the star formation rate is virtually con-stant throughout the simulation with only a mild peak dur-ing the first passage (see also di Matteo et al. 2007). The fi-nal BH masses are systematically lower for increasing massra-tio of the merger. Furthermore, the slope of the M BH growth as a function of time becomes shallower with in-creasing merger mass ratio. There is also a systematic delayin the time of the BH merger with increasing mass ratio, asindicated by the filled circles in Fig. 2. For the lower massratio mergers the peak of the BH activity (the filled trianglesin Fig. 2) typically occurs shortly after the merging of theBHs, whereas for the higher mass ratio mergers the peak ofthe BH activity is not directly related to the merging time ofthe BHs.In Fig. 3 we study the duration of the BH accretion andstar formation activity as a function of merger mass ratio. c (cid:13)2006 WILEY-VCH Verlag GmbH&Co. KGaA, Weinheim stron. Nachr. / AN (2006) 791 Fig. 2
The evolution of the star formation rates, blackhole accretion rates and BH mass as a function of time forco-planar prograde 1:1 (black), 2:1 (green), 3:1 (red), 4:1(blue) and 6:1 (orange) mergers. The filled circles indicatethe time of the BH merger, the filled diamonds in the toppanel and the filled triangles in the bottom two panels showthe location of the maximum star formation and BH accre-tion peaks, respectively.We define a half-mass growth time ∆ T / during which halfof the final BH mass and half of the total stellar mass isformed respectively. In both cases ∆ T / is centered on thepeak of the corresponding activity, the BH accretion on themaximum of the BH accretion rate (the triangles in Fig. 2)and the star formation rate on the peak of the SFR markedwith diamonds in Fig. 2. The resulting half-mass timescalesare shown in the bottom panel of Fig. 3. The ∆ T / isstrongly correlated with the mass ratio of the merger. For1:1 and 2:1 mergers the growth of the M BH is very con-centrated in time, with half of the final BH mass growthoccurring in less than 100 Myr. For the 3:1 and 4:1 merg-ers ∆ T / ∼ . , with the 6:1 merger resulting in ∆ T / ∼ .The corresponding stellar half-mass timescales ∆ T star (triangles in the bottom panel Fig. 3) also show a clear cor-relation with the mass ratio of the merger. In the 1:1 mergerhalf of the final stellar mass is formed in a short major burstlasting about ∆ T star ∼ .
15 Gyr . This value is compa-rable to the star formation timescale derived by Cox et al.(2008), who calculated a full width at half maximum of
FWHM ∼ . for the star formation peak of a typical Fig. 3
The maximum BH accretion and star formationrates and timescales as a function of merger mass ratio.The bottom panel shows the half mass growth time of theBHs (circles) and the stellar mass (triangles) centered onthe maximum BH accretion and star formation rates, respec-tively. The top panel shows the corresponding peak BH ac-cretion (circles) and star formation rates (triangles) duringthe corresponding half mass growth time ∆ T / .1:1 merger. For higher mass-ratio mergers the resulting starformation timescales are ∆ T star ∼ . − . , withthe highest mass-ratio merger having the longest timescaleof ∆ T star ∼ . .In the top panel of Fig. 3 we plot the correspondingmaximum black hole accretion and star formation rates. Bydefining the variable q as the mass ratio between the pri-mary and secondary component we can fit the logarithmsof the peak BH accretion and star formation rates with thefollowing linear relation log Maximum [M ⊙ yr − ] = a + a × q , (1)where a and a are the inferred normalization and slope re-spectively. Both the maximum BH accretion rates and starformation rates are well fitted by Eq. 1 (dotted lines in Fig.3) resulting in the following best fitting parameters respec-tively: ( a ,BHAR = 0 . , a ,BHAR = − .
47) ( a ,SF R =1 . , a ,SF R = − . . The ratio of the peak BH accre-tion rate to the peak star formation rate is thus of the or-der of ˙ M BH , peak / ˙ M SF , peak ∼ − . On the other hand,the ratio of the mean BH accretion rate to the mean starformation averaged over the whole simulation is closer to ˙ M BH , mean / ˙ M SF , mean ∼ − , with this ratio varying sys-tematically between ˙ M BH , mean / ˙ M SF , mean = 2 · − (1:1merger) and ˙ M BH , mean / ˙ M SF , mean = 0 . · − (6:1 mer-ger). Interestingly, recent observations of z ∼ galaxies byDaddi et al. 2007 also find indications of an universal ratioof ˙ M BH / ˙ M SF ∼ − and this is also expected from theobserved M BH − M ∗ relation. c (cid:13)
92 Johansson et al.: Termination of star formation by BH feedback in galaxy mergers
Finally, we analyse in Fig. 4 the star formation and BHaccretion rates for a sample of four mixed E-Sp mergers(left panel) and four dry E-E mergers (right panel). The starformation rate of the mixed E-Sp mergers is lower than inSp-Sp (Fig. 1) mergers, due to the lower amount of coldgas available for star formation. The disk progenitor con-tains f gas = 20% of gas initially, whereas the early-typeprogenitors typically have an initial gas mass fraction of f gas ∼ , with typically only ∼ of this gas beingcold and dense. After the merger of the E-Sp galaxies thestar formation rate declines rapidly in all the merger rem-nants. We define a BH mass growth factor as f BH , insitu =∆ M BH , insitu /M BH , final , which gives the ratio of BH massgrowth due to gas accretion during the simulation with re-spect to the final BH mass. Quantitatively, the fraction ofthe BH mass that accumulates by gas accretion during themixed mergers is in the range of f BH , insitu ∼ − ,with typical mean values of f BH , insitu ∼ . For the E-E mergers the initial star formation rates are generally verylow due to the low gas fractions of f gas ∼ − . The ini-tial gas fraction directly depends on the strength of the ini-tial interaction that gave rise to the merger remnants used asprogenitors for the E-E remergers. Increasing the masses ofthe progenitors and using more direct planar orbits producesmore violent initial encounters, thus decreasing f gas . Thestar formation is very effectively terminated shortly after themerging of the E-E progenitors on comparable timescales to1:1 Sp-Sp mergers and thus more efficiently than in 3:1 Sp-Sp and mixed E-Sp mergers. In this paper we have studied the termination of star for-mation in merger simulations including BH feedback. Wefind that the termination of star formation by BH feedbackis significantly less important for unequal-mass disk merg-ers compared to equal-mass disk mergers. The timescalefor star formation termination systematically increases withincreasing progenitor mass ratios. Similarly, a systematicincrease is seen in the half-mass growth timescales of theBHs, with this timescale varying from ∼ . Gyr for equal-mass mergers to ∼ Gyr for 6:1 mergers. This systematictrend can be used as input in modeling BH accretion morerealistically in semi-analytic galaxy formation models (e.g.Croton et al. 2006). For mass-ratios of 3:1 and higher merg-ers with BH feedback are unable to completely quench thestar formation, with the merger remnants showing star for-mation rates roughly on the pre-merger level even 1 Gyr af-ter completion of the merger. In addition, the star formationis efficiently terminated in mixed E-Sp and dry E-E mergersdue to the presence of the fully grown super-massive BHsin the early-type progenitors.
Acknowledgements.
The numerical simulations were performedon the local SGI-Altix 3700 Bx2, which was partly funded by theCluster of Excellence: ”Origin and Structure of the Universe”.
Fig. 4
The total star formation rate (top), the total blackhole accretion rate (middle) and the evolution of the totalblack hole mass (bottom) as a function of time for four E-Sp mixed mergers (left panel) and four E-E remergers (rightpanel) . The gas disks have initially f gas = 20% and thevirial velocities as indicated in the Figure. The filled circlesindicate the time of merging of the BHs. References
Baldry, I. K. et al.: 2004, ApJ, 600, 681Balogh, M. L et al.: 2004, ApJ, 615, 101Bell, E. F., et al.: 2003, ApJS, 149, 289Birnboim, Y., Dekel, A., Neistein, E., 2007, MNRAS, 380, 339Bower, R. G et al.: 2006, MNRAS, 370, 645Cimatti, A., Daddi, E., Renzini, A.: 2006, A&A, 453, 29Croton, D. J., et al.: 2006, MNRAS, 365, 11Cox, T.J. et al., 2008, 384, 386Daddi, E. et al.: 2007, ApJ, 670, 156Dekel, A. & Birnboim, Y.: 2006, MNRAS, 368, 2Dekel, A. & Birnboim, Y.: 2008, MNRAS, 383, 119di Matteo, P. et al.: 2007, A&A, 468, 61Efstathiou, G: 2000, MNRAS, 317, 697Glazebrook, K. et al.: 2004, Nature, 430, 181Johansson, P. H. & Efstathiou, G., 2006, MNRAS, 371, 1519Johansson, P. H., Naab, T., Burkert, A., 2008, ApJ in press,arXiv:0802.0210Juneau, S. et al: 2005, ApJ, 619, 135Kauffmann, G. et al.: 2003, MNRAS, 341, 33McCarthy, P. J.: 2004, ARA&A, 42, 427Naab, T. & Burkert, A.: 2003, ApJ, 597, 893Naab, T., Jesseit, R., Burkert, A.: 2006, MNRAS, 327, 839Naab, T., Johansson, P. H., Ostriker, J. P, Efstathiou, G.: 2007, ApJ,658, 710Springel, V.: 2005, MNRAS, 364, 1105Springel, V. & Hernquist, L.: 2003, 339, 289Springel, V., Di Matteo, T., Hernquist, L.: 2005a, ApJ, 620, 79Springel, V., Di Matteo, T., Hernquist, L.: 2005b, MNRAS, 361,776V¨ais¨anen, P. & Johansson, P. H.: 2004, A&A, 421, 821 c (cid:13)2006 WILEY-VCH Verlag GmbH&Co. KGaA, Weinheim