Transcript Document

“Making Normal Galaxies in a Cosmological Setting”
“Making Galaxies Red and Dead Without Feedback”
T.Naab, P. Johansson, K. Nagamine,
G.Efstathiou,
RYFeb
Cen2009:
and J.P.O.
Princeton, 27
jpo
Cambridge, 8 May 2008
Cosmological Simulation:
Start with WMAP CBR Sky
QuickTime™ and a
decompressor
are needed to see this picture.
Hinshaw et al; 2008
QuickTime™ and a
decompressor
are needed to see this picture.
WMAP Spectrum of Cosmic
Perturbations
(Amplitude)2
(Aplitude)2
Spherical Harmonic
I believe, with a
perfect faith…
Cosmic Structure Formation
Recombination
Nucleosynthesis
(cartoon version)
linear perturbation theory
nonlinear simulations
Computing the Universe: locally, growth of
perturbations computed classically;
numerical hydro required to reach the
current epoch
• Transformation to
co-moving
coordinates x=r/a(t)
• Co-moving cube,
periodic boundary
conditions
• Lbox >>lnl
Lbox
Physics Input Included
• Newtonian gravity.
• Standard equations of hydrodynamics.
• Atomic physics:adiabatic, + cooling, +heating,
+ non-equilibrium ionization.
• Radiative transfer: global average, +shielding
of sinks, +distribution of sources.
• Heuristic treatment of star-formation.
• -------------------------------------------------• Maxwell’s equations in MHD form.
Physics Input Missing
(important on galactic scales)
• Cosmic ray pressure and heating.
• Dust grain physics (depletion,
absorption and catalyzation).
• Magnetic field generation.
• Multiphase media.
Star Formation Algorithm
• Consider gas that is dense, cooling and collapsing.
• Make stellar particle:
DM* = const* x DMgas x dt/Max(Tcool,Tdyn).
(cf R. Kennicutt and M. Kuchner)
• Label particle with position, mass, metallicity and
epoch.
• Give particle velocity of gas and follow dynamics as if
dark matter particle.
• Allow output of mass, energy and radiation from each
particle consistent with a star-cluster of same mass
and age: “feedback”.
C* = 0.05
Global Simulations
(better mass resolution)
Naab et al (2007)
(better spatial resolution)
Choice of parameters (“free” or
otherwise)
• Cosmological parameters:
– Determined by observational constraints such as WMAP, SDSS
etc
• Star-Formation algorithm:
– Results nearly independent of algorithm so long as cooling gas
made into stars (in agreement w observation)
• Metallicity:
– Yield determined by match to cluster IGM
– ++++++++++++++++++++++++++++++++++++
THAT IS ALL THERE IS
But what about “feedback” ?
(but feedback is necessary and does cause some moderate variance: NB digression ->)
Bubbles blown by super-winds from forming galaxies heat the ambient medium
and retard subsequent gas infall: Cen et al 2004, Dave …
(feedback important for IGM, but relatively unimportant for galaxy properties)
Feedback Increases Number of Small Mass Galaxies
and Reduces Number of High Mass Galaxies.
(effects largely compensate and produce little net change in SF rate)
high feedback
no feedback
Star Formation history: Nagamine et al (2005)
TVD Hydro vs Data
SPH Hydro & SAM vs Data
Butcher-Oemler or
Gunn-Dressler
effect
Blanton, M.; Cen, R.;
Ostriker, J. P.; Strauss,
M. A.; Tegmark, M.
ApJ.531, 1 (2000)
TVD Hydro Simulation
In clusters, the fall off in
star formation since z=1
is much more rapid than
in the field.
Cause is simply
C2x > V2 gal,esc
Effect of hot gas in suppressing GF
Results of Global Simulations
• Reasonably good results on epoch of
galaxy formation & mass distribution of
galaxies.
• Reasonably good on spatial distribution
of galaxies and environmental effects
(eg early “red and dead” in clusters).
• Good treatment of the IGM.
• Essentially no information re internal
structure of galaxies.
Individual Galaxy Simulations
1) Find and isolate objects of interest in large box.
2) Add small scale power to region(s) of interest.
3) Nest within bigger and bigger boxes (but smaller
than the total volume) and add intermediate
scale power (for tidal forces).
4) Repeat simulation at higher spatial, temporal
and mass resolution in smaller regions.
5) Go back to step # (2) with still higher resolution
and repeat steps # (3) and # (4) to convergence.
Hydro Codes
• SPH (eg Springel & Hernquist, Weinberg &
Katz etc)
– Advantage: good spatial resolution, community
standard.
– Disadvantages: poor mass resolution, too much
viscosity, and cooling instabilities.
• AMR (eg Norman & Bryan, Klypin,Tessier etc )
– Advantage: more accurate hydro.
– Disadvantage: technically very costly to resolve
many regions simultaneously (communication).
High Resolution Simulation of
Massive Galaxy Formation
Naab, Johannson, Ostriker and Eftsatiou
Input
• Dark Matter Simulation (AP^3M: 50 h-1Mpc)
• Pick isolated halos (~ 10^12 Msolar)
• Re-simulate at higher resolution with gas
(SPH and GADGET-2)
• Standard star-formation algorithm
• Standard cooling
• No feedback
Questions
• Convergence: how do results change with
resolution improvement?
• Is feedback necessary to make an early formed
“red and dead” galaxy?
• Are the paradigms “monolithic collapse”, “merger”,
“dry” or “wet” accretion useful, relevant?
• What is the physics that matters?
• What is the expected evolution of an elliptical?
• How to test the picture presented?
Detailed Hydro Simulations (N,J,O&E : 2007ApJ, 658,710)
Convergence to
low and to a flat
rotation curve at
high resolution:
In Situ Star Formation
Convergence to
stellar system
formed very early
which quickly
becomes “red and
dead”.
Gas Properties
Gas, at all radii,
becomes hotter with
time despite fact that
the “cooling time”< the
Hubble time! Why?
Accreted Stellar Mass
Accreted stellar
mass, 45% of
total is added
late ( z < 1.5),
and at larger
radii.
Half-Light Radii of In-situ and Accreted Stars
A Normal Elliptical: fits Sersic Profile
(detailed kinematics ok as well)
Size Evolution
QuickTime™ and a
decompressor
are needed to see this picture.
Profile Evolution
QuickTime™ and a
decompressor
are needed to see this picture.
Dark Matter Evolution
QuickTime™ and a
decompressor
are needed to see this picture.
Physics
• Gas is steadily being heated by in-falling new gas
( -PdV and Tds) and by dynamical friction from
infalling lumps of DM and stars (dynamical
friction) via viscosity.
• These effects not included in SAM, but equal quantitatively - feedback used by the SAM
schemes.
• Of course feedback really exists and must be
complementary to effects listed above.
• Dynamical Friction due to in-falling stellar lumps is
very important for evolution of the stellar and DM
components.
Astronomy
• Two phase growth. First in situ star-formation from
in-falling cold gas, and then accretion of stellar
lumps.
• DM initially increases in density (adiabatic
contraction) and then decreases (dynamical friction)
• Metal rich component in center from in-situ starformation and metal poor component in outskirts
due to stellar infall of old and small systems.
• Stellar system grows in size with time and central
velocity dispersion actually declines with time
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
What have we learned?
• High resolution needed ~ 107 particles.
• For massive systems the three papers of
1977 (Binney, Silk and Rees & Ostriker)
appear to point to the correct physics:
Cooling time of gas becomes longer than
the dynamical time and star formation
ceases. Systems live in hot bubbles
and then grow by accretion of smaller
stellar systems.
Intermediate Mass Systems:
ie Normal Spirals
• Where ?: Filaments - not clusters or voids.
• When?: Slowly - bulge, then disc and halo.
• How?: Not understood and complicated (ie
cannot simulate), but at least four
phases/components:
– Bulges formed like early ellipticals, smaller scale
models of same: “dissipational collapse;
– Discs form slowly from inflowing cold gas streams;
– Accretion of satellites adds metal poor stars to halo and
gas to disc;
– Dynamical evolution can produce bars, thick discs,
globular cluster in-fall and destruction etc.
Technical Issues that make
this a VERY hard problem
• Very high mass space and time resolution is
needed, since discs are so thin and relaxation in
them is so easy to produce (spuriously).
• Since things happen slowly, one must have a
reasonably good model of star formation (ie if it is
short compared to T0 , one can get it wrong and it
matters little).
• Since thin discs are fragile, both feedback and
dynamical effects can strongly alter the evolution
• Etc, etc, etc - a difficult problem.
Many papers with beautiful
work: Examples include
“Simulations of Galaxy Formation in a Λ Cold Dark
Matter Universe. I. Dynamical and Photometric Properties
of a Simulated Disk Galaxy” & “II. The Fine Structure of
Simulated Galactic Disks”
by Abadi, Mario G.; Navarro, Julio F.; Steinmetz, Matthias;
Eke, Vincent R. Ap.J;591,499 (2003) & 597, 21
QuickTime™ and a
decompressor
are needed to see this picture.
But
Many other good papers, but typically suffer from same
problem, of too bulge dominated and too high a rotation
curve, with a recent excellent summary of the issues:
“The formation of disk galaxies in computer simulations”: Mayer,
L.; Governato, F.; Kaufmann, T.( astroph0801.3845v )
We review the progress made by numerical simulations carried out on large parallel
supercomputers. Recent progress stems from a combination of increased resolution
and improved treatment of the astrophysical processes modeled in the simulations,
such as the phenomenological description of the interstellar medium and of the
process of star formation. High mass and spatial resolution is a necessary condition in
order to obtain large disks comparable with observed spiral galaxies avoiding spurious
dissipation of angular momentum. A realistic model of the star formation history. gasto-stars ratio and the morphology of the stellar and gaseous component is instead
controlled by the phenomenological description of the non-gravitational energy budget
in the galaxy. We show that simulations of gas collapse within cold dark matter halos
including a phenomenological description of supernovae blast-waves allow to obtain
stellar disks with nearly exponential surface density profiles as those observed in real
disk galaxies, counteracting the tendency of gas collapsing in such halos to form
cuspy baryonic profiles. However, the ab-initio formation of a realistic rotationally
supported disk galaxy with a pure exponential disk in a fully cosmological
simulation is still an open problem.
However, see solution…
• Increase resolution
to ~ 107 mass
elements.
• Add supernova
feedback of types I
and II
• Repeat (Quinn et
al)…
Alternate approach: put in cosmological context, but model
the components separately without attempting hydro
“A simple model for the evolution of disc galaxies: the Milky Way”
Naab, Thorsten; Ostriker, Jeremiah P. (MNRAS; 366,899;2006)
A simple model for the evolution of disc galaxies is presented. We adopt three
numbers from observations of the Milky Way disc, Σd the local surface mass density, rd
the stellar scalelength, Vc, the amplitude of the rotation curve, and physically, the local
Kennicutt star formation prescription, standard chemical evolution equations assuming
a Salpeter initial mass function and a model for spectral evolution of stellar
populations. We can determine the detailed evolution of the model with only the
addition of standard cosmological scalings with the time of the dimensional
parameters. A surprising wealth of detailed specifications follows from this prescription
including the gaseous infall rate as a function of radius and time, the distribution of
stellar ages and metallicities with time and radius, surface brightness profiles at
different wavelengths, colors, etc.
Same but better: Schoenrich and Binney (astroph 0809.3006S):
Models of the chemical evolution of our Galaxy are extended to include radial
migration of stars and flow of gas through the disc. The models track the production
of both iron and alpha elements. A model is chosen that provides an excellent fit to
the metallicity distribution of stars in the Geneva-Copenhagen survey (GCS) of the
solar neighbourhood, and an acceptable fit to the local Hess diagram. The model
provides a good fit to the distribution of GCS stars in the age-metallicity plane
although this plane was not used in the fitting process. Although this model's starformation rate is monotonic declining, its disc naturally splits into an alphaenhanced thick disc and a normal thin disc. In particular the model's distribution of
stars in the ([O/Fe],[Fe/H]) plane resembles that of Galactic stars in displaying a
ridge line for each disc. The thin-disc's ridge line is entirely due to stellar migration
and there is the characteristic variation of stellar angular momentum along it that
has been noted by Haywood in survey data. Radial mixing of stellar populations
with high sigma_z from inner regions of the disc to the solar neighbourhood
provides a natural explanation of why measurements yield a steeper increase of
sigma_z with age than predicted by theory. The metallicity gradient in the ISM is
predicted to be steeper than in earlier models, but appears to be in good agreement
with data for both our Galaxy and external galaxies. The absolute magnitude of the
disc is given as a function of time in several photometric bands, and radial colour
profiles are plotted for representative times.
What have we learned?
• Very high resolution (N ~ 107) and SN
feedback are both necessary.
• For lower mass systems cool gas accretion is
important at late times (Weinberg, Katz &
Dekel, Birnboim).
Cooling time of gas is shorter than the
dynamical time and star formation
continues via accretion of gas to discs
which become the familiar spiral systems.
Conclusions: High Mass Systems
• High resolution SPH simulations without
feedback produce normal, massive but small
elliptical galaxies at early epochs.
• Accreted smaller systems add, over long
times a lower metallicity stellar envelope.
• Physical basis for cutoff of star-formation is
gravitational energy release of infalling matter
acting through -PdV and +Tds energy input to
the gas (R&O, 1973)
• Feedback from SN and AGN is a real
phenomenon - but secondary and mainly
important for clearing out gas at late times.
Conclusions: Lower Mass
(predominantly spiral) Systems
• Preliminary hydro simulations indicate cool gas
accretes onto disks (around old bulges) and
produces familiar spiral late forming galaxies.
• Physical basis for transition is cooling time vs
in-fall time of gas. input to the gas.
• Dynamical evolution at late times is important.
• Major mergers at late times are relatively
unimportant (would overly thicken discs if they
occurred), but satellite accretion is significant.