Transcript 1_Moore

Obergurgl, December 2009
Galaxy formation: is the end in sight?
Ben Moore, Institute for Theoretical Physics, University of Zurich
+Oscar Agertz, Roman Teyssier+
GHALO: A billion particle simulation of the dark matter distribution surrounding a
galaxy. 3 million cpu hours with the parallel gravity code pkdgrav (Stadel et al 2008)
50 parsec, 1000Mo resolution, 100,000 substructures
GHALO: A billion particle simulation of the dark matter distribution surrounding a
galaxy. 3 million cpu hours with the parallel gravity code pkdgrav (Stadel et al 2008)
50 parsec, 1000Mo resolution, 100,000 substructures
What is the origin of morphology?
There is a large diversity in the galaxy population. Most stars are in disk
galaxies, most galaxies are dSph/dE.
What about S0, E, dIrr, LSB, Polar Rings, Bars……
Many of the observed galaxies start off as disks and undergo
transformation between classes
Malin1
200 kpc
Malin2
Moore & Parker 2007
Malin1 has hardly had time to rotate once! How can
stars form ~100kpc from the center?
The Cartwheel Ring Galaxy
Mapelli, Moore et al 2008
T=100Myr=Cartwheel
MNRAS, 383, 1223
Very low surface brightness
galaxies could be the evolved
state of ring galaxies
200kpc
T=1Gyr=Malin1
Here are some of the key baryonic components of galaxies that we need to simulate
correctly
Evrard, Summers & Davis 1994
A few hundred SPH particles per ‘glob’
Steinmetz & Navarro 1998
1000-10000 SPH particles
Abadi et al 2002
Most of stars in a massive spheroid
50,000 SPH particles
Baryons far too concentrated
Governato et al 2006
10^5 SPH particles in total, 500pc resolution
Better match with T-F.
Huge spheroid, disk is unresolved single phase cold gas
No thin disk dominated system, no morphological detail, no spiral patterns,
bulge/spheroid dominated systems.
What is resolved and what is put in by hand?
The dark matter distribution is resolved to about 0.5% of the virial
radius with 1 million particles – that’s about ~1kpc for the MW.
The shock heating of baryons and cooling processes are well
resolved.
Multiphase flows are not followed correctly with SPH (Agertz et al
2008) – almost all galaxy formation studies have used SPH.
SPH
AMR
T=0.2
GRID
SPH
T=0.5
T=0.8
Hydro/star formation simulations have reached 10^5Mo per gas
particle and gravitational force resolution of 0.5kpc, little else is
resolved, not even the disk scale height and certainly not the 10pc
molecular disk.
SPH disks are smooth featureless blobs of single phase gas – no
molecular clouds, no spiral patterns – but lets see the next talk?!
Star particles put in places where the gas density is high (sets a
maximum radius) – but stars form in molecular clouds and the
molecular cloud mass function depends on gravitational instability
which varies with radius.
Simulating star formation in disk galaxies:
the present status - 2009
•
•
•
The ISM is under-resolved
To radially resolve a galactic disk such as the Milky Way can be done
using ∆x∼0.5 kpc. At this resolution the whole scale height is within one
resolution element. The true disk stability is not captured as both the
density and velocity structure (gas and stellar dispersion) is numerically
affected. This still allows for the disk to have the correct global
characteristics such as gas and stellar mass compositions, thin and thick
disk, and at high resolution, realistic spiral structure. In this case a
statistical star formation recipe based on the local gas density and free
fall time is well motivated both theoretically and observationally
Satellites remain unresolved (the gas doesn’t reach the density threshold)
This is possible but the relevant variables (star formation threshold
and efficiency) must be understood robustly!
Simulating star formation in disk galaxies:
the future - 2015
•
•
The ISM is fully resolved
This means that the scale height of all ISM components are resolved
using at least 10 resolution elements (Romeo 1994). This translates into
at least ∆x∼1 pc. If this is not satisfied the true disk stability will not be
modelled accurately. At such a resolution, star formation occurs in their
natural sites i.e. massive clouds such as GMCs. This treatment is the
goal of most simulations but is due to their computational load beyond
the capabilities of modern simulations attempting to study the assembly
and evolution of large spiral galaxies to z = 0. In addition, as the star
formation sites become resolved the codes need to incorporate the
radiative feedback in order to accurately treat the life-times of the GMC
structures (Murray et al. 2009).
Pandora’s box (new small scale physics must be treated)!
Computationally impossible in a cosmological context today
Agertz et al 2008:
AMR simulations of isolated disks at
high resolution can resolve the
formation and evolution of molecular
clouds.
Gravitational instability and MC
interactions sets a minimum floor to the
disk velocity dispersion.
A 1 kpc region of the disk
Agertz et al ( 2008)
“The observed dispersion
velocity is generated
through molecular cloud
formation and subsequent
gravitational encounters”
What is the state of the art in cosmological simulations of galaxy formation?
We can resolve the ISM to ~50-200pc, enough to follow giant molecular cloud
complexes.
AMR simulations – Agertz, Teyssier & Moore
•
•
•
Feedback
SNII: 10^51 ergs dumped thermally after ~30 Myr. These are the
M>8 Msun stars. Given a typical IMF, this occurs for 10 % of the
mass of a given population and 10 % of that mass is returned as
enriched material. We keep track of the young stars that form and
turn off cooling in these regions allowing for the hot pockets of gas to
develop into a blastwave. If this is not done the under-resolved
medium unphysically radiates away that energy (McKee & Ostriker,
Thacker & Couchman)
SNI: As a star of mass 1<M<8 Msun exits the main sequence we
calculate the fraction of them being in a binary system and hence is
eligible for a type I event. We enriched the ISM with the outcoming
metals and Sn energy.
Winds: A population of stars loose 30-40% of its mass in winds. We
return this mass to the gas component at the end of a stars life.
SN Feedback Thermal Dump
z=10
SN Feedback Thermal Dump and Delayed Cooling z=10
The complex gas flows into a dark matter halo with a forming disk galaxy at a redshift
z=3. R=temperature, G=metals and B=density. (Agertz, Teyssier & Moore 2009). One
can clearly distinguish the cold pristine gas streams in blue connecting directly onto the
edge of the disk, the shock heated gas in red surrounding the disk and metal rich gas in
green being stripped from smaller galaxies interacting with the hot halo and cold streams
of gas. The disk and the interacting satellites stand out since they are cold, dense and
metal rich.
The complex gas flows into a dark matter halo forming a disk galaxy
R=temperature, G=metals and B=density
Agertz et al (2009)
Elmegreen et al 2009: Clumpy high redshift galaxies – chains, clusters etc.
Clumpy (10^8Mo), high star-formation rates, extended over ~10kpc radii
ACS images (Elmegreen et al)
No direct evidence for cold infalling
gas…very hard to detect.
NGC 4650A
Maccio’, Moore, Stadel
(2005)
serendipitous detection of a
polar ring galaxy in a
cosmological hydrosimulation.
Evidence for cold accretion
on sub-L* scales.
We managed to make a nice looking disk at z=3, but what about the evolution to z=0?
•
•
•
•
Parameter study
Resolution
Star formation threshold
Star formation efficiency
Inclusion of SN 1 events and wind massloss
For realistic choices we always get
nice disk galaxies.
•
•
•
•
Due to regulation (balance of star
formation and destructive
feedback), a resolved disk is less
sensitive to the actual choice of
parameters.
If the disk is under-resolved, the
efficiency sets the bulge to disk
ration. The threshold for star
formation must be set low enough
to avoid “missed star formation
events” in the outer disk.
Analytical derivations of required
resolution given a star formation
recipe will be provided!
The disk, SR5n1e1ML
Gas density
Gas temperature
Gas density
Stars
Gas temperature
Gas metallicity
• Thin extended disk of stars and gas
• Thick stellar disk and bulge
• Hot pockets of gas from supernovae
• Massive hot gaseous halo
• Gravitational instability: we resolve spiral structure
• Stars preferentially form in the arms
• Galactic fountain
• Realistic metallicity gradient
Stars
Gas metallicity
Regardless of parameter choice, all
disks get roughly the same global
properties.
Disks with low B/D ratios are obtained
when star formation is resolved at large
disk radii and when the efficiency per
free fall time is kept below 2%.
Detailed properties are unconstrained.
The Kennicutt-Schmidt law
Bigiel (2009), fig. 15. Compilation
using THINGS data + other
Our AMR simulations – different
colours are different star formation
efficiencies
The disk, SR6n01e1ML
Gas density
Stars
Gas temperature
Gas metallicity
Gas density
Stars
No dominating bulge!
Sb galaxy where the bulge
forms from a buckled bar.
Gas temperature
Gas metallicity
The rotation curve of the simulated galaxy with low star formation efficiency matches
well that of the Milky Way.
Compare SDSS stellar mass function
with DM halo mass function
M MW  2.5  1012 M⊙
Forero et al 2009
Gao et al 2009
Halo gas, SR5n1e1ML
Temperature
Metallicity
• Complex large scale structure in temperature and metallicity
• Satellite stripping a la Magellanic Stream
A diffuse gaseous halo is robustly predicted with a density and temperature profile in good
agreement with the Milky Way non-detection. It can explain the observed HVC head-tail features as
well as a ram-pressure origin of the Magellanic Stream.
Using the obtained values (n~10^-4 cm-3 and T~ few 10^6 K) we can explain e.g. Smith’s Cloud:
The cloud of 10^6 Msun of HI gas is moving towards the disk of the Milky Way at 73 km/s. Smith's
Cloud is expected to merge with the Milky Way in 27 million years at a point in the Perseus arm.
Observation
Simulation
The interacting Magellanic Clouds – MW system
Our goal is to resolve molecular cloud formation within a cosmological context:
i.e. <10 parsec resolution.
Aim is to study formation and evolution of galaxies across the Hubble sequence,
in different environments and within the current cosmological context
Star-formation will still be sub-grid, but will be spatially, kinematically and
physically more believable
This is just a few years away with planned 10-50 x code improvements.
Agertz et al (2008)
ISM kinematics using
the RAMSES AMR
code +/- star
formation +/- SN
feedback
Lots of predictions, no evidence for CDM yet, no
strong counter evidence. Complicated by galaxy
formation
Direct or indirect detection of dark matter may happen
in the next 5 years, but it may also never happen
since the cross section may be so small
A new complex model of galaxy formation is emerging
in which baryons accrete to the halo center in three
ways – cold streams, cooling flow, metal enriched
stripped debris.
At z=2 we can match the clumpy morphology of HST
galaxies.
Gas
Stars
At z=0 we can make thin disks with small bulges and
nice spiral patterns. The parameters have to be tuned
since molecular cloud (star) formation is unresolved.
The global properties are correct but details such as
disk scale lengths are uncertain.
Is the end in sight? Yes, but still not very predictive.
By 2015 we will reach the 1 parsec resolution required
to resolve the molecular disks and spatially resolved
star formation.