14
How well do we understand outflows and accretion on cosmic scales? Romeel Davé

How well do we understand outflows and accretion on cosmic scales? Romeel Davé

  • View
    217

  • Download
    0

Embed Size (px)

Citation preview

How well do we understand outflows and accretion on

cosmic scales?Romeel Davé

Galactic Galactic OutflowsOutflows• Cold mode accretion is dense & filamentary Need

bouncer feedback to prevent overcooling: Outflows• Test outflow scaling relations by comparing hydro

simulations to outflow-related observables, e.g.:– IGM enrichment [Oppenheimer & RD 06]– Early galaxies & overcooling [RD, Finlator, Oppenheimer 07]– Mass-metallicity relation [Finlator & RD 07]– DLA kinematics [S. Hong, Katz etal in prep]– ICM metals & energy [RD etal in prep]

• A single wind scaling relation matches all these!

Quantifying Quantifying OutflowsOutflows

• Outflows rare locally, but probably the norm at z>~2.

• Two basic parameters:

– Outflow velocity: vw

– Mass loading factor: • Martin 05, Rupke etal 05:

Starbursts show vwvcirc. • Murray etal 05: Such a

scaling arises in momentum-driven winds: vwvc, 1/vc

• Implement into Gadget-2, Monte Carlo ejection of particles, vc computed from Mgal using on-the-fly finder.

Martin 2005log

Erb etal 06z~2 SFG’s

M82 MIPSEngelbracht etal

How unique is this outflow How unique is this outflow model?model?• Short answer: Not terribly.

• Key features that seem necessary to match data:– Winds eject mass & metals from ALL galaxies, not just

dwarfs.– Small galaxies expel a higher fraction of their accreted gas.– Outflow rate » star formation rate @ early epochs.

• M-D scalings work, but feel free to invent your own…

Erb et al 2006

Log W

ind k

ineti

c/Pote

nti

al

How do outflows set galaxy How do outflows set galaxy properties?properties?

• Key insight: Accreted gas is processed quickly Inflow ≈ Outflow + SFR.

SFR = Inflow/(1+). This inflow equilibrium relation broadly governs galaxy properties (e.g. SFH, Z, fgas).

• e.g. fgas is set by (M*). If this doesn’t vary with z, then fgas(M*) doesn’t vary with z [as observed].

• Energetics arguments for outflows are not relevant. Outflows don’t share energy, they blow holes and leave.

• Bottom line: is key! vwind irrelevant, beyond >vesc.

Momentum-drivenscalings

Constant vw,

Mass

outfl

ow

rate

out

of

halo

Red: input scalingsBlack: actual

z=2

What does this mean for outflow What does this mean for outflow physics?physics?• Unclear; approx scalings could in principle be generated

from momentum or energy driven winds.• SN-driven sims usually fail to remove much gas mass

from the ISM (Mac Low & Ferrara; Teyssier’s talk).• In principle, lots of momentum available from light and

stellar winds to drive gas out, but coupling unclear.• Need ISM sims of momentum-driven winds!• Much to be done on feedback… What about accretion?

log

The M*-SFR The M*-SFR RelationRelation

• Gas accretion star formation• M*-SFR constrains SFH form:

• Observations of SFGs (z~0-2):– M*SFR0.7-0.9 at all z.– Small scatter (<0.3 dex) around

“main sequence” of SFGs.– Evolution is M*-independent.

Daddi etal 07z~1.4-2.5

Elbaz etal 07z~0.8-1.2

Noeske etal 07z~0.2-1.1

M*-SFR vs. ModelsM*-SFR vs. Models• Green: Millenium

SAM • Red, magenta: SPH • Blue: Data (=0.3)

• Slope <~unity? • Scatter small? • Evolves

independent of M*?

• Evolves at observed rate? ×

Star Formation Activity Star Formation Activity ParameterParameter

(i.e. fraction of Hubble time required to form M* at current SFR).

• Models: sf~1 at all z. • Cold accretion similar

forms of SFH at all M*.• Observed: sf(z) evolves

strongly. Oops!• Possibilities:

– Simulated SFH wrong?– Measurements wrong?– Or…

Data

Models

log S

FR (

Mo/y

r)

1011M1010.5

1010

109.5

IMF wrong?IMF wrong?[insert Stacy McGaugh MOND dance][insert Stacy McGaugh MOND dance]

• Need less M* formed per unit high-mass SF

• Conservatively, SFR/M* should be reduced by ~x3 at z=2, and ~x2 at z=1: This would yield unevolving sf.

• Larson (98,05): IMF today has Mchar≈0.5 M. High-z ISM hotter Mchar higher.

• “Evolving Kroupa” IMF (0.1-100 M):

dN/dlogMM-0.3 for M<Mchar.

dN/dlogMM-1.3 for M>Mchar.

Mchar=0.5(1+z)2 M from PEGASE modeling

Evolving IMFEvolving IMF• No effect on high-

mass SF/feedback/ metals; only detectable in M* accumulation rate.

• SFR down by ~×(1+z)

• Fardal etal: Reconciling fossil light (K, EBL) and integrated cosmic SFH“Paunchy” IMF.

• Perez-Gonzalez etal (IRAC): M* to z~4. dM*/dt < SFR @z>2.

• Not crazy…

SummarySummary• It is possible to constrain basic outflow

parameters across cosmic time by comparing hydro sims to galaxy SFR and ZIGM data.

• Best matches are for scalings reminiscent of momentum-driven winds, but actual physics of wind propagation unknown.

• Mass loading factor is key: SFRZ(1+)-1.• Accretion appears to be reasonably well

understood, but at face value the evolution of SFR-M* doesn’t agree.

• An IMF that is more bottom-light at high-z is an explanation that seems equally as (un)likely as any of the alternatives, and may be favored from fossil light considerations.

Simulated SFH wrong?Simulated SFH wrong?• At z~2, observed sf~0.2.• Problem: Can’t reconcile sf~0.2 with

other data, let alone models.• Bursts? (tip of iceberg)

– M*-SFR tight; Lower SFRs would’ve been seen.

• Delayed SF? (strong early feedback)– sf~0.2 implies z~2 systems began SF at

z~2.3! Plus, low scatter for 1.4<z<2.5.

• Unseen passive galaxies? (downsizing)– Mass-selected samples do not see enough

passive galaxies; sBzK selects dominant population at z~2.

• All seem dubious (besides being inconsistent with models).

Measurements wrong? Measurements wrong? (Systematics)(Systematics)

• Need to lower SF / raise M* by ~x3-5.• Raising M* generically hard: Unless stars

put out a LOT less red light than locally. [note: Maraston vs. BC03 goes wrong way]

• Something else mimicking SF?– AGN: Possible, but would have to be strange

to exactly mimic tight M*SFR.– PAH emission: Rest-8m dominated by PAHs,

so perhaps PAH emission per unit SF much stronger at high z.

• Can’t be ruled out, but would require dramatic differences vs locally calibrated relations. Such differences not seen locally even in extreme systems.