Cold mode of gas accretion

1. Missing Baryons, University of Sydney 12/01/2009 Cold mode of gas accretion Dušan Kereš Institute for Theory and Computation - Harvard Collaborators: Neal Katz, Lars Hernquist, Romeel Davé, David Weinberg, Mark Fardal

2. Stellar mass locked in galaxies increases with time Marchesini et al. 2008

3. Cosmic star formation drives the stellar mass growth Hopkins & Beacom 2006 What drives the star formation history?

4. Stars form from molecular gas • Galactic disks were (molecular) gas rich at high redshift. • High SFRs at early times. • SFRs drop with time. • Is this simple model in agreement with observations? • Gas consumption timescales for molecular gas are short, ~2Gyrs at z=0 (Kennicutt 1998), ~1 Gyr at z~1.5 (Daddi et al. 2008) • Kennicutt law - > short SFR timescales. • Molecular phase needs to be re-supplied over time.

5. ** z=0 Dense, neutral atomic gas • Molecular phase gets replenished from atomic phase • DLAs contain most of the neutral gas • Is there enough gas in the dense atomic reservoir to explain the evolution of SFR density? • Amount of gas in DLAs at high-z is less than the mass locked in stars at z=0. • Most stars form after z=2: HI phase is constant at 0< z <2. • DLA phase also needs to be constantly re-supplied. • Most of the baryons are outside of galaxies in ionized form (galaxies contain only 6-10% of baryons)-> huge reservoir Prochaska&Wolfe ‘08

6. Milky Way • SFR 1-3 M⊙/yr (Kennicutt 2001) • Gas reservoir < 5e9 M⊙. • This can be quickly consumed • Star formation rate in the Solar neighborhood has been relatively constant for several Gyrs (Binney et al. 2000) • not much change in gas density despite large gas depletion • Infall needed • Deuterium in local ISM is close to primordial value even though is should be depleted in stars (e.g. Linsky et al. 2006) • Infall l needed

7. The observed evolution of the cosmic and “local” star formation requires gas accretion from the intergalactic medium at all redshift

8. Standard Model • E.g. White & Rees 1978 • Gas falling into a dark matter halo, shocks to the virial temperature Tvir at the Rvir, and continuously forms quasi-hydrostatic equilibrium halo. Tvir=106(Vcirc / 167 km/s)2 K. • Hot, virialized gas cools, starting from the central parts, it loses its pressure support and settles into centrifugally supported disk –> the (spiral) galaxy. • Mergers of disks can re-distribute the gas and produce spheroids. • The base for simplified prescriptions used in Semi-Analytic models – SAMs (e. g. White & Frenk 1991). Tvir

9. SPH simulations of the CDM universe • Gadget-2, 3 (Springel 2005) Smoothed Particle Hydrodynamics code: • Cooling (no metals), UV background that heats and ionizes the gas • Schmidt law + star formation timescale is selected to match the normalization of the Kennicutt law. • Two-phase sub-resolution model: SN pressurize the gas, but does not drive outflows • Usually no winds from star formation (but see Dave and Oppenheimer’s talk) • Large volume 50/h Mpc on a side, with gas particle mass ~8x107 M⊙, but most of the findings confirmed with resolution study down to 4x104 M⊙ • Lagrangian simulations -> naturally adaptive and we can follow fluid (i. e. particles) in time and space.

10. Global Accretion • We count particles that joined galaxies between two outputs • Galaxies grow through mergers and accretion of gas from the IGM • Smooth gas accretion dominates global gas supply at all times • dominates growth of central galaxies in < 1012M⊙ halos • Star formation follows smooth gas accretion • It is within factor of ~2 of the “Madau diagram” (e.g. Hopkins & Beacom 2006) • Mergers only re-distribute material: globally important after z=1 Kereš et al. 2009

11. Tmax = ? Shocked IGM Temperature history of accretion - We follow each accreted gas particle in time and determine its maximum temperature - Tmax before the accretion event. - In the standard model Tmax ~ Tvir Galactic gas

12. 250000K -T evolution of accreted particles - The gas that was not heated to high temperatures -> COLD MODE ACCRETION - The gas that follows the standard model -> HOT MODE ACCRETION - Empirical division at 2.e5K, but results are robust for 1.5-3e5K Katz, Keres et al. 2002 Kereš et al. 2005

13. How important are these two accretion modes? • Cold mode dominates the gas accretion at all times. • Hot mode starting to be globally important only at late times Kereš et al. 2009

14. Low mass halos are not virialized • Transition at Mh=1011.4 M⊙ • Early hints (e.g. Binney 1977, Katz & Gunn 1991) • 1D description by Birnboim&Dekel 2003 • Shock cannot propagate if post shock cooling times are shorter than the gas compression time (multiple of free fall time) • Condition valid at Rvir when Mh~1x1011 M⊙ • More realistic transition is gradual: dense cold filaments survive within hot halos (Keres et al. 2005) • Transition mass increases in sims. with feedback and metal cooling Kereš et al. 2009

15. Smooth accretion at high-z Cold mode Hot mode • Gas accretion rate is a strong function of mass (even steeper when mergers are included) • Cold mode dominates at all halo masses. • Even in massive halos, full of hot gas, cold mode filaments supply galaxies with cold gas. • Cooling from the hot atmosphere is not efficient • Cold mode accretion follows infall into halos (within a factor of 2-3). Accretion rates of central galaxies

16. m_p ~106 M⊙, resolution ~500pc (at z=2) Kereš et al. 2009

17. 5kpc/h physical New zoom-in simulations Z~2.6 110/h pc physical res. M_p~1.e5M_sun M_h~1.3e11M_sun Yellow Tvir Blue ~ 1e4K 300/h kpc ~ 2Rvir Simulation of a single halo takes a month on our local cluster! 1.2/h Mpc cmv.

18. Halo mass at z=0, 7x1011M_sun Gas particle mass ~105M_sun

19. Halo clouds at z=0 n > 4x10-4 cm-3 200/h kpc box Kereš & Hernquist 2009

20. Thermal history • We determine the maximum temperature ever reached by particles currently in clouds • Clouds form out of 1-1.5 x 105 K gas, which is the temperature of low-z filaments infalling into MW size galaxies. • A small fraction forms from hot component (Mo&Miralda-Escude ‘96, Maller&Bullock ‘04) Kereš & Hernquist. 2009

21. Cloud Formation • Penetrating 1.e5K filaments create density inversions susceptible to Rayleigh-Taylor instabilities. • These regions are further compressed by shocks, pressure from hot medium and ram pressure, triggering faster gas cooling and cloud formation. • Cooling instabilities and R-T instabilities have similar timescales in the regions where clouds form and act together to form them. • Both are comparable to the free fall time. • Galactic winds increase the number of formed clouds: • More density inversion, more gas compression, cold gas kicked out (interpretation is more complex).

22. Cloud properties • T~104K clouds (cooling curve limit) • Pressure confined by ~106K gas. • No associated dark matter • Masses few 106 - 5 x 107 Msun. • Total mass ~1-2 x 109 Msun • Hydrogen column densities ~1018-1021 cm-2 • 40/h x 60/h kpc region in halo center. • Prominent rotational component, clouds are building the outer disk. • Cloud formation process similar after z ~1, but accretion increased around merger events and decreases afterward. • These might be analogs of: • Clouds and extra-planar gas observed around nearby galaxies • HVCs around the Milky Way • Halo absorbers

23. Survivability of Clouds • SPH approach has difficulties to reliably follow the cloud evolution (Agertz et al. 2008) • Details of cloud disruption are complex but simple criterion exists: • Clouds disrupt after they encompass mass in the hot medium comparable to their own mass (Murray & Lin ‘04) • For a simple orbit calculation, gas profile of simulated halo and clouds launched at 50 kpc: Mc > 106 Msun survive • Clouds < few 105 Msun do not survive intact. • More detailed calculations are needed (e.g. Heitsch& Putman 09) Murray and Lin 2004

24. Infalling gas around local galaxies • HVCs around MW, properties similar to simulated clouds • Clouds often show infalling velocities, net infall 0.25M/yr (Wakker ‘99) • Extra-planar clouds at distances up to 50kpc are detected around M31 and several other galaxies (e.g. Wakker, van der Hulst ) • Similar net infall around other galaxies • Low metallicities: unlikely a result of Galactic fountain Van Woerden et al. 2004

25. Consequences • Cold, pressure supported clouds are supplying galaxies with gas: 0.6-1 Msun/yr for 7e11Msun halo. • Observed infall rates are factor of 5-10 lower than SFRs. • Small correction: • clouds at large distances and edges of clouds have lower densities and can be partially or completely ionized • Large corrections: • HVCs are probing only a part of the infalling gas, with the largest deviation from orderly rotation. • A large fraction of infalling gas is co-rotating • Hard to separate galactic fountain gas from the IGM infall (e.g. NGC 891) • => rates of gas accretion around nearby galaxies are underestimated, possibly by large factors

26. Signatures of cloud infall in external galaxies NGC 6946 Boomsma 2007 Holes punctured In the HI disk (out of the SF region) Velocity wiggles - Simulated galaxies: outer disks often show warps, wiggles and slight change in AM direction - Simulated clouds occasionally make “holes” in the outer disk

27. How to detect accreting gas? • In the local universe we can use HI but what about higher redshift? • Halo absorbers: Ly-limit and MgII, especially around z~1 • Time when filaments turn into clouds and increase cross section • Important in few 1.e11- few 1.e12 Msun halos, where cold gas and hot medium co-exist • Empirical models based on Mg II absorbers imply largest covering fractions in 0.5-1.e12 halos (Tinker & Chen 2008). • Direct Ly_alpha emission from cooling radiation in z~3 halos. Work in progress, with Faucher-Giguere et al. • Line radiative transfer • Ionizing radiation radiative transfer • Problem: larger sample of simulations for a range of halo masses is needed for detailed predictions: • Simulation of a single halo takes more than a month of computing time; sample will dramatically increase in ~year timescale • Resolution, instabilities, metallicity issues

28. Conclusions • Cold mode accretion is the dominant way of galactic gas supply • Filamentary nature of infall is very important, denser gas prevents shock propagation in massive halos • Cosmological simulations can follow the formation of halo clouds form the “first principles”. • In the MW size halos, infalling filaments form cold gaseous clouds that supply galaxies with gas • This is likely the formation mechanism of a large fraction of halo absorbers and HVCs. • We are starting to make quantitative predictions for future observations of gas accretion onto galaxies: • 21 cm emission of HI, halo absorbers, Ly_alpha emission • This will take some time, but will enable more direct tests of models and help with detection of ongoing gas accretion.

29. Low mass halos Kereš et al. 2009