Galaxy merging in the Millennium simulation

Serena Bertone - UC Santa Cruz

Chris Conselice - U. Nottingham

arXiv:0904.2365 MNRAS, in press

Cosmoclub, April 27th, 2009

Overview

• The Millennium simulation

• Techniques to identify mergers:

– in observations

– in simulations

• Results:

– merger fraction evolution

– merger rate evolution

– dependence on stellar mass

– dependence on time-scale for merging

– problems in models and observations

The Millennium simulation

Still the largest N-body simulation ever run(Springel et al 2005):

• 500 Mpc/h box• >10 billion DM particles• >20 million galaxies with MDM>1010 M

4 public galaxy catalogues:• Croton et al 2005• Bower et al 2006• De Lucia & Blaizot 2007• Bertone, De Lucia & Thomas 2007

Cosmological modelIC from CMB:

WMAP(1)

DM evolutionMillenniumSpringel et al.

2005

N-body simulation

FOF group finder

Halo merger trees

Evolution of galaxies

SA galaxy formation model

Galaxy catalogues: observed galaxy

propertiesSB, De Lucia & Thomas 2007

stellar pop synt. modelsdust extinction etc

Galaxy Formation in a Nutshell

Galaxy Formation Physics

Cold Gas(ISM)

Hot Gas(ICM)

Winds(SN

feedback)

Stars

recycling

star formation

coolin

g

ejection

re-incorporation

shock-heating

Black Hole accretion

Bertone et al 2007 SA model:

• same physical model as De Lucia & Blaizot 2007

• New: SN winds modelled as astrophysical blastwaves

in a cosmological context (Ostriker & McKee 1988)

• two-phase model for the long-term evolution of winds

– adiabatic, pressure-driven expansion

(Hoopes et al 2004, Strickland & Stevens 2000)

– momentum-driven snowplough

(Aguirre et al 2001, Theuns et al 2001)

Physics of Winds

hotbubble

coldbubble

thin cold shell

How do galaxies accrete mass?

• Merging: stars (gas)

• Cold accretion gas

• Hot accretion gas

A. Evans, HST

Dekel et al 2007

Ocvirk et al 2008

Merging vs. star formation

• Channels for galaxies to build up stellar mass:

– accretion by merging

– direct star formation

• Which one channel prevails depends on stellar mass

• Mergers are essential to build up the stellar mass in massive galaxies

Guo & White 2008star formation

minor mergersmajor mergers

more

massiv

e

Identifying mergersMethods to identify mergers in observations:(see Mark’s talk last week)

• CAS (concentration-asymmetry-clumpiness, Conselice 2003)

• Gini-M20 (Lotz, Primack & Madau 2004)

• close galaxy pairs (Patton et al 2000)

Each method may identify different populations of merging galaxies:• dry mergers• gas-rich mergers• mergers with different mass ratios…

The interpretation of results is not straightforward

Identifying mergers

CAS & Gini-M20:• merger has already

occurred• structural methods• CAS: A>0.35 & A>S

Pairs: • galaxies haven’t yet

merged• magnitudes within 1.5• physical separation

< 30 kpc/h

• merging timescale ≤ 400 Myr

• mass ratio ≥ 0.25

The observed sample

• Low redshift data: mergers identified by structural asymmetries (CAS)

– De Propris et al 2007 @ z=0– Conselice et al 2009 (COSMOS + EGS)

• High redshift data:

– structural asymmetry: Conselice et al 2008 (HDF+UDF)

– galaxy pairs: Bluck et al 2009 (GOODS)

Mergers in the Millennium

• Most previous works use galaxy pairs:– Kitzbichler & White 2008

– Patton & Atfield 2008

– Mateus 2008

– Genel et al 2008, 2009

• Bertone & Conselice 2009: direct counting of mergers in the simulation:– better consistency with CAS and Gini-M20 counts

– no ambiguity from merging time-scales, mass ratios and pair separation

Counting mergers

Procedure:

• stellar masses, mass ratios, time of merging are known from model

• set a time-scale: τ=0.4 Gyr and τ=1 Gyr:

– τ is equivalent to the time-scale to which structural methods are sensitive to identify mergers in observations

– investigate dependence on τ: source of uncertainty in obs

• at given redshift, count how many galaxies have undergone a merger within τ

• calculate merger rates, fractions etc

Merger fractions

Definition:

• Fraction of galaxies that have undergone a merger within the last τ Gyr

• strongly dependent on τ

• more mergers at high redshift and in massive galaxies

€

fm =Nmergers (M*, z)

Ngalaxies (M*, z)

SB & Conselice 2009

Merger fraction vs. redshift

• good agreement at high stellar masses and z<2

• observations underestimated by a large factor when low mass galaxies are considered

Gamma vs. redshift• Γ= fgm / τ with fgm= 2fm/(1+fm)

• average time between mergers inverse of the merger rate per galaxy

too long!

Merger rates

Definition:

• in the simulation R is independent of the time-scale used to count mergers

• rate is highest for low mass galaxies

• visible evolution with redshift

€

R(z) =fgm (z) ⋅ngm (z)

τ

Shape of merger rate vs. M*

• the shape of the merger rate vs. stellar mass is defined by the shape of the stellar mass function ngm(z)

merger fraction vs stellar mass≈2 orders of magnitude

stellar mass function≈5 orders of magnitude

merger rate vs stellar mass≈3 orders of magnitude

Merger rates vs. redshift

• Good agreement at high stellar masses

• shape vs. redshift well reproduced

• But: how can there be good agreement for galaxies with M>1010 M when the merger fractions disagree?

Something wrong at M>1010 M?

• The merger rate agreement with obs for M*>1010 M is a coincidence

- stellar mass density: overestimated by factor ≈10

- merger fraction: underestimated by factor ≈10

• too many galaxies and not enough mergers in the Millennium at M<1011 M?

stellar mass density vs. redshift

Merging times

• median merging time of satellites in simulation increases with redshift

• longest merging times for low mass galaxies

• at z ≤ 1 it is comparable or larger than the Hubble time!

• problem in the semi-analytic model?

Median galaxy merging time

Comparison with other models

De Lucia & Blaizot 2007:

• same merger trees, galaxy formation prescriptions and parameters

• different SN feedback model

• predicted rates and fractions differ by factor of a few, similar redshift evolution

• similar agreement with observations, sometimes worse

Other models

Bower et al 2006:

• same DM evolution

• different SA model

• different merger trees

• better reproduces the high redshift data

• difference in results at high z: is it due to the SA modelling or to the merger trees?

Mateus 2008

Pair fractions dataKitzbichler & White 2008:

• calibrate the relationship between the fraction of galaxy pairs and the merger rate at high z

• close galaxy pairs are a reliable tool for extracting the merger history of galaxies

• the merging times used to convert to merger rates are overestimated by at least a factor of 2 in current observations

• does not solve the discrepancy we find at high z with pair fraction data

• problem: the position of type 2 satellites in the Millennium is uncertain

What do we learn?

• The simulated merger history is very sensitive to the semi-analytic prescriptions:

– differences in results between Bertone et al 2007, De Lucia & Blaizot 2007 and Bower et al 2006, even using same merger trees dependence on global star formation history

– merger time-scale too long for low mass galaxies? Can this help solve other problems of the models? Too many red satellite galaxies, too many low mass star galaxies…?

• Some quantities in observations not fully understood might also introduce uncertainties in the results:

– time-scale for merging sensitivity (CAS and pairs)

– mass ratios

Conclusions

We have recovered the merger history of galaxies in the Millennium simulation:

• merger rates and fractions vs. redshift and stellar mass

• massive galaxies experience on average more merger events than less massive ones, but have a lower merger rate

• model results agree with observations for massive galaxies, but disagree when galaxies with 1010 M < M* < 1011 M are considered

• too few mergers in the simulation between low mass galaxies

Galaxy Formation Physics

Cold Gas(ISM)

Hot Gas(ICM)

Winds(SN

feedback)

Stars

recycling

star formation

coolin

g

ejection

re-incorporation

shock-heating

Black Hole accretionBertone et al 2007