Gas accretion in Galaxies

Dušan KerešTAC, UC Berkeley

Hubble Fellow

Collaborators: Romeel Davé, Mark Fardal, C.-A. Faucher-Giguere,Lars Hernquist, Phil Hopkins, Neal Katz, Chung-Pei Ma, Ben

Oppenheimer, Volker Springel, Mark Vogelsberger, David Weinberg

Massive Galaxies Over Cosmic Time 3, Tucson 11/2010

Gas supply is needed• Galaxies are actively forming stars

at all epochs. What is the sourceof fuel for star formation?

• Stars form from molecular gas.• Molecular gas consumption

timescales are << t_H– > Additional gas reservoir is

needed to support long term starformation.

• Dense atomic phase (slowconsumption)?

• Amount of HI in DLAs at high-z ismuch less than the mass locked instars at z=0.

• Dense atomic phase also needs tobe constantly re-supplied.

• Galaxies contain <10% of baryons,huge reservoir available in the IGM

• Gas from the IGM suppliesgalaxies at all epochs (see alsoBauermeister et al 2010).

*ρ* z=0

Prochaska&Wolfe ‘08

Indirect evidence is widespread

• Star formation rate in the Solar neighborhood has beenrelatively constant for several Gyrs (Binney et al. 2000).– Implies not much change in gas density despite large gas

depletion• Deuterium in local ISM appears to be re-supplied (e.g. Linsky et

al. 2006)– Infall needed

• Gas depletion timescales are much shorter than the Hubbletime, but star formation sequence is evident at all times (Daddiet al. 2007, Noeske et al. 2007, Salim et al. 2007)

• Star forming galaxies are part of the “star forming sequence”and have long duty cycles of ongoing star formation. -> starformation in the bulk of galaxies is not bursty.

• Some continuous process keeps galaxies supplied with gasover Hubble time.

Global Accretion• We use cosmological SPH

simulations with Gadget (2,3)(Springel 2005).

• Gas cooling and star formation,no outflows.

• We count particles that joinedgalaxies between two outputs

• Galaxies grow through mergersand smooth gas accretion

• Smooth gas accretion dominatesglobal gas supply at all times

– dominates growth of centralgalaxies in < 1012M⊙ halos

• Star formation follows smoothgas accretion -> trivialconsequence of short starformation timescales.

• Mergers re-distribute material:globally important after z=1Kereš et al. 2009

Gas accretion in the standard model

● E.g. (Rees & Ostriker 1977, White & Rees 1978)

● Gas falling into a dark matter halo, shock heats tothe virial temperature Tvir at the Rvir, and continuouslyforms quasi-hydrostatic equilibrium halo.

Tvir=106(Vcirc / 167 km/s)2 K.

● Hot, virialized gas cools, starting from the centralparts, it loses its pressure support and settles intocentrifugally supported disk –> the (spiral) galaxy.

● The base for cooling prescriptions used in Semi-Analytic models – SAMs (e. g. White & Frenk 1991).

Tvir

Temperature history of accretion• We utilize the Lagrangian nature of our code

• We follow each accreted gas particle in time and determine itsmaximum temperature - Tmax before the accretion event.

• In the standard model one expects Tmax ~ Tvir

IGM

Shocked Tmax = ?

Galactic gas

Kereš et al. 2005Katz, Keres et al. 2002

250000K

Evolution of gas properties ofaccreted particles

• Empirical division at 2.e5K

• Gas that was not heated tohigh T -> COLD MODEACCRETION.

• Galaxies accrete fresh gasdirectly from cold denseintergalactic filaments.

• Gas heated to high T ->HOT MODE accretion;• Cooling of the HOTvirialized atmospheres.

Examples from SPH simulations

Kereš et al. 2009

- Gadget-3 (Springel 2005), m_p ~106 M⊙, resolution ~300pc (at z=3)

- Low mass halos contain mostly cold non-virialized gas, infaling infilaments. This cold filamentary accretion dominates at high redshift andin halos below 3e11Msun (Katz, Keres et al. 2003, Keres et al. 2005)

- 1D Shock stability analysis shows that virial shock does not reach Rvir Inhalos below ~1e11Msun (Birnboim&Dekel 2003)

-Massive halos contain mostly hot shock-heated gas, but coldmode accretion might still operate.

Kereš et al. 2009

What have we learned from theory?• Based on the cosmological simulations and analytic arguments (Kereš,

Katz, Birnboim, Dekel, Brooks, Ocvirk, Teyssier, Agertz, Ceverino andothers), we know that:

– Filamentary cold mode accretion of non-virialized gas is the dominant way ofgas supply into high redshift galaxies.

– Accretion from the IGM gas provides continuous fuel source and is drivinghigh star formation rates of these galaxies (e.g. Genzel et al.)

– Filamentary gas is dense; virial shocks cannot propagate trough the filamentseven in halos that start developing hot atmospheres, at least at high redshift.

– Cold mode provides gas supply to galaxies in < few 10^11Msun halos atlower redshifts.

– Properties and geometry of filaments at fixed mass are changing with redshift.

– Drop in cosmic density -> decreases of gas supply with time -> lower starformation rates at late times.

Robust global picture butmany open questions

• Do different simulation techniques agree in the study of gas infall?

• What happens at much higher resolution?

• What happens in massive halos, do filaments survive ?

• How to *directly* detect cold mode accretion?

• With all this theoretical evidence for infalling gas, why we only seeoutflows?

• Do outflows affect infall?

• How does the infall stop? Do we need feedback at the massive endand what causes galaxies to be red-and-dead?

Different simulationtechniques

PRELIMINARY (work in progress)• So far good agreement between AMR and SPH codes in cold mode studies.• SPH Gadget-3 (Springel ‘05) and moving mesh code Arepo (Springel ‘10)

– Same gravity, same ICs, better instabilities, naturally adaptive, Galilean invariant code– 10/h Mpc box 2x128^3 particles.

• Global properties of cold halo gas are largely insensitive to simulation technique– However, Arepo simulations show higher “hot mode” cooling rates at late times and

more extended gas rich disks whose stripping contributes to cold gas.

Gadget-3

Arepo

Kereš, Vogelsberger, Springel, Hernquist; in preparation

Higher Resolution

Halo mass at z=0: ~7x1011M_sunGas particle mass: ~4x104M_sunForce resolution ~70/h pc at z=3, ~30pc peak hydro resolution~8 million particles within Rvir at z~0.

Halo clouds at z=0

200/h kpc box

n > 4x10-4 cm-3

Kereš & Hernquist ‘09

- At high resolution infalling gas forms clouds at z <1-2(depending on mass)- Clouds infall from a flattened distribution in the disk plane.- Large fraction of the infall is co-rotating with the galaxy

Origin of halo clouds• Most clouds form from 1-1.5e5K infalling gas

– Leftovers of cold mode accretion– Stripping of the satellite gas– A fraction forms from hot halo (likely sensitive

to halo gas metallicity)• Penetrating filaments are cooling unstable

and create density inversions in agravitational field, susceptible to R-Tinstabilities.– Infalling gas is also compressed by the

surrounding hot medium and shocks fromstructure formation.

• Cloud masses are <1e6-1e7Msun: highresolution is needed to resolve them.

• Properties do depend on resolution: not yetclear what determined the mass spectrum?

• Clouds form earlier in massive halos.• In early type remnants cloud formation or

their infall has to prevented in order toproduce passive systems!

Observational Signatures

Nearby galaxies• Only direct detection of the

accreting gas comes from“local” HI observations.– infalling high-velocity HI clouds

in the Milky Way– Similar clouds around nearby

galaxies.– Simulated clouds broadly

consistent with observations, interms of mass and columndensity

• Estimates of infall rates basedpurely on HVCs are biased low:– Large fraction of accretion is

co-rotating with the disk, hardto separate in velocity space.

– HI observations probe onlydense, innermost clouds, therest is ionized.

– Work in progress M31, Thilker et al. 2004

Higher redshift• Absorption studies, both directly from

galaxies and using backgroundsources are used as probes of thehalo gas (e.g. Weiner et al. 2009,Steidel et al. 2010)– Spectra show signature of outflows.– Metal lines, but very sensitive (would

detect dense cold gas with even smalltraces of metals)

– Models that try to predict thisabsorption can also explain absorptionat higher impact parameters, withoutneed for gas infall (Steidel et al. 2010)

– Without ionizing radiation transfer, lowresolution simulations suggested largecovering fractions of infall (e.g. Dekelet al. 2009).

– However, detailed predictions aremissing-> time to improve this (checkthe astro-ph tonight!)

Weiner et al. 2009

Steidel et al. 2010

Why we see only outflows?

• We use zoom-in simulations with peak hydroresolution of ~30pc, 4e4Msun gas particles

• Covering fraction of DLAs, neutral hydrogenwith N_HI > 2e20/cm^2, is very small at z~2,only few few percent of the Rvir.

• Relevant for metals, several low-ion speciesoriginate from this self shielded gas.

• Given very large covering factor of outflows,signatures of gas accretion, especially instacked spectra will be washed out.

Faucher-Giguere & Kereš 2010

3e11Msun

9e11Msun

Halo gas in emission?• Direct Ly_alpha emission from the infalling gas in z~3 halos:

– Cooling radiation form the infalling gas is a strong source ofLy_alpha emission (e.g. Katz &Gunn 1991, Fardal et al. 2001,Dijkstra&Loeb 2009, Goerdt et al. 2010)

– Other possibilities: star formation, outflows, AGN.– Complex: Need to precisely identify self-shielded regions and the

gas temperature– To get line profiles and surface brightness distributions need to do

Ly-alpha line radiative transfer• Regardless of the origin of such emission blobs contain

information about massive galaxy formation at high-z!

Luminosity of Lyα blobs

• Predictions are extremely sensitive on the correct treatment of star forming andself-shielded regions (strong temperature dependence at 1-2e4K):– Orders of magnitude difference if not careful!– Need to be careful with treatment of self-shielding and star forming gas!– We apply proper ionizing radiation transfer, both in post-processing and on the fly.– With radiation coming out of star forming regions excluded, cooling radiation is

insufficient to power brightest blobs, need additional sources to get to ~few 1e44ergs/s.– Lower luminosity extended sources can be powered by the cooling radiation.

Faucher-Giguere, Kereš et al. 2010

Do we need feedback?

Mass function without feedback

-No feedbacksimulations, produce toomany stars in galaxies

-Problems at the highmass and low mass ends

-Removal of hot modedoes not make much of adifference, in no-feedback simulations

-All of the galaxyformation happens early.

Kereš et al. 2009

Which mode is responsible?• Majority of galaxies were built mostly

by cold mode• M > 5x1010M⊙ (0.5-1x1012M⊙ halo)

the contribution of cold modeincreases

– growing merger contribution– Mergers of smaller galaxies built the

more massive ones, not the recentaccretion

– These small galaxies were builtthrough cold mode

– Without metal cooling and feedbackhot mode not important for massaccumulation: enables mergers(minor and major) to dominate massgrowth

– Early feedback in low mass galaxieswill affect massive halos at late times

Kereš et al. 2009

Low mass end solution?

• “Momentum driven” galacticwinds and gas re-accretionhelp with the low mass end

• Re-accretion of ejectedmaterial increases gasaccretion in massive halosand ruins the high mass end

• Solution to the high massend depends on the lowmass end feedback.

Oppenheimer, Dave, DK et al. 2009

Possible high-mass end solution?• Problems with simulated massive

galaxies:– Current star formation (colors)– Mass of galaxies (and morphology)

• Can gas rich major mergers stop starformation?– Trigger quasar mode of AGN– Shock heat the halo gas– Produce burst of star formation– Grow massive black holes, condition

necessary to have long termmaintenance in “radio mode”

– In hot halos, all of these are efficient.– Right scaling with mass to produce

proper mix of morphological types.• However, merger driven feedback

needs to be carefully implemented incosmological environment to see if itcan solve the high mass problem.Hopkins et al. 2008

Kereš et al. 2009

Salim et al. 2007

simulations

X Bell et al. 2003

Summary• Cold mode accretion is now a theoretically robust scenario.• Star formation in the Universe is driven by continuous accretion of gas• Predictions for the direct observability of gas infall are coming out,

– We need more careful work, cannot just observe simulations!• Future observations with the next generation of 25-30 meter

telescopes will help avoid stacking and increase sample ofbackground sources probing galactic vicinity.

• This will help constrain different models of gas accretion and feedbackin a very direct way.

• Feedback is needed to regulate galactic growth!• Infall and outflows co-exists and likely interact -> interesting complex

physics to model and observe.• Merger driven bursts are not important for cosmic star formation but

major merger events are important in regulating galaxy evolution inmassive systems.

Lyα emission from simulatedhalos

• We incorporated, ionizing radiation R-T, self-shielding & Ly_alpha line R-T.• Example: 2.5e11Msun halo at z=3, from zoom-in simulation.

– Cooling radiation from the halo gas emission is detectable and contributes toLy_alpha emission from extended blobs.

Faucher-Giguere, Keres et al. 2010