SemiAnalytic Galaxy Formation are we kidding ourselves Health

Semi-Analytic Galaxy Formation - are we kidding ourselves? Health warning - not a proper review; not a complete bibliography Thanks to the Galform team: Carlton Baugh, Andrew Benson, Shaun Cole, Andreea Font, Carlos Frenk, Juan Gonzalez, John Helly, Cedric Lacey, Rowena Malbon, Ian Mc. Carthy, (this talk in no way reflects the views of the group!)

What this conference is all about… Ø What are semi-analytic models for? ØA means of predicting the properties of the universe? ØThe ultimate multiscale simulation technique? ØA tool for interpreting observational data? ØA tool for understanding numerical simulations? ØA tool for assessing telescope proposals?

Something to think about… Ø If you ran the perfect simulation: Øreal Hydrodynamics Ø 1 Mo resolution, 1 pc smoothing ØMagneto-hydrodynamics ØBlack holes (relativistic magnetohydrodynamics) Ø …and matched every piece of observational data Ø Would you have learned anything?

What are Semi-Analytic Simulations? …take a few steps backwards…

Structure formation is hierarchical Ø Small things form first Ø Big things form later

Cosmological model ( m, , h); dark matter Primordial fluctuations Well established / (M, t) Dark matter halos (N-body simulations) Well understood Gas processes (cooling, star formation, feedback) Gasdynamic simulations Semi-analytics Formation and evolution of galaxies

Two approaches for populating the Dark Universe with galaxies Ø Semi-analytics Ø Direct simulation Ø Start from fundamental Ø Encapsulate physics in physical laws simple equations. Link Ø Gives the “correct” solution (for them in a network. the input physics, resolution, Ø Fast! numerical accuracy etc) Ø Easy to explore different Ø Need to add “subgrid” physics parameters and new to stabilise the solution. physical effects Ø Populate a vast volume with galaxies Complementary not Adversary!!! (the boundaries are blurring)

What’s the problem? Ø So few stars Ø Only 10% of the baryons form into stars (Balogh et al 2001, Cole et al 2001, Lin et al 2003) Ø “Down sizing” Ø “As the universe ages star formation moves fromthe larger to Is it just But is this really what data show? smaller objects” (Cowie et al 1996) the maximum star forming mass that increase with redshift? – or is it just the mass threshold for star formation? Ø “Anti-hierarchical” Ø “the big galaxies form first, while in CDM the large dark matter haloes form last” But the first haloes to form are now incorporated into the largest haloes today! Ø “The Broken Hierarchy” Ø“baryon physics introduces extra scales” (Rees & Ostriker 1978, Binney 1977, Silk 1977, White & Rees 1978, White & Frenk 1991)

Other problems for galaxy formation Ø Related problems: ØThe shape of the luminosity function ØThe “cooling flow” problem Ø Unrelated problems (? ): ØThe density-morphology relation Ø“(pre-)heating” the intra-cluster medium

Recent progress in semianalytics Feedback - regulating the formation of galaxies is the key issue

The galaxy luminosity function The halo mass function and the galaxy luminosity function have different shapes SNe winds Dark halos (const M/L) Complicated variation of M/L with halo mass galaxies Data: Cole et al 01: Kochanek etal: 01; Huang et al 03 Benson et al ‘ 03

What cooling+feedback need to do! feedback has sucessfully depressed galaxy formation in small haloes The same problem is seen in simulations: Balogh et al. , 2001; Springel & Hernquist 2003 dark matter mass function (fixed M/L) NB: exacerbated by the high value of WMAP Ωb but cooling is now too effective in high mass haloes (there's more gas left over) Benson et al 2003

A solution: AGN The two modes of AGN accretion

The GALFORM family Superwinds Radio Mode AGN Many recent papers have their own implementation of AGN “radio mode” feedback, eg. Crotton et al 2006, Cattaneo et al 2006, Kang et al 2007; Sommerville 2008

The Power of AGN Comparison of energies: Thermal energy of a 1013 Mo halo … 1061 erg Accretion energy of a 109 Mo black hole … 2 x 1062 erg It seems unlikely that AGN are unimportant!

The two forms of AGN feedback “Quasar” mode “Radio” mode feedback (eg. Granato et al. , 2004, Springel et al 2005) (eg. Croton et al 2006, Bower et al 2006 Okamoto et al 2007) Radio X-rays Temperature Shock heating Uplifting matterial? Mixed plasma and ICM? M 87: Forman et al 2006; Perseus: Fabian et al 2000, 2006 Springel et al 2005

Why does the “radio mode” work?

The AGN feedback loop AGN fuelling Cooling “radio” mode Hydrostatic ? tcool>tfree-fall Keres et al 2003; Dekel & Birnboim 2003; Binney 2004

The impact of AGN Feedback: An Example With AGN Without AGN bulge stars disk stars

Example from Cattaneo et al 06 Similar plots in Croton et al 06

Different implementations same aim Ø RGB Ø AGN “radio mode” offsets hydrostatic cooling if BH is sufficiently massive Ø Croton/De Lucia Ø Compute “radio mode” feedback energy from mass of halo and black hole (loosely based on bondi accretion of multiphase gas) Ø Cattaneo et al Ø Separate hot and cold accretion above a (redshift dependent) threshold mass. Ø Kang/Summerville Ø Radio mode driven by multiphase bondi accretion model Ø Menci/Monaco(/Baugh 05) Ø BH(SN) driven superwinds Ø Etc…

Are the semi-analytic recipies justified? Ø “Gastrophysics” is still a difficult problem Ø How does thermal energy from Sne couple with the ISM? Ø If resolution is low, this energy is just radiated away Ø How does the AGN interact; how is it triggered? Ø Still hard (impossible) to simulate a significant population of galaxies with adequate resolution Ø The prospects for “ab initio” simulation of galaxies Ø Learn and embed in semi-analytics Ø Embed sub-grid semi-analytics in simulation

How well does it work?

Comparison with data Ø Cirasulo et al. Bower 06 De Lucia 07 Faint end overshoots - but see Khochfar 08 et al

Evolution of the Stellar Mass Function Ø The evolution of the stellar mass function from Drory et al 2005. z=0 AGN model Mc. Clure et al 2006 Integrated SMD agrees with Stark et al 2006

Evolution of colours Ø Evolution of red sequence tracks passive evolution Ø …but the blue sequence also get bluer – matches the increase in SFR density Bower et al 2006 & De Lucia et al 2006 galaxy catalogues are public! www. mpa-garching. mpg. de and www. icc. dur. ac. uk

Problem solved? No Way! Challenges for galaxy formation models

Environment Models need more sophisticated treatment of environmental effects: Kang et al Font et al

Environmental Physics is not correctly handled All Galaxies Satellite Galaxies All satellites are red! No blue satellites!

Environmental Physics is not correctly handled Old Strangulation model • Remove gas reservoir as galaxy orbits larger halo Larson, Tinsley & Caldwell 1980 Mc. Carthy et al – an improved model for halo stripping – depends on the orbit of the satellite and the gas content of the satellite and main halo. (Actually, Gunn & Gott’s formulae re-calibrated for halo gas using numerical simulations) Hot gas reservoir Is this realistic? Strangulation = suffocation • Mass ratio of haloes = starvation • Gas atmosphere of the main halo SNe winds quickly exhaust disk gas

Blue galaxy fraction with an improved treatment of environment Weinmann et al 2006; Font et al, 2008

X-rays emission from groups and clusters The Achilles' heal of these models? ? ?

X-ray Emission from Groups and Clusters Ø L-T relation : well known that the self-similar relation fails Ø AGN: standard model just prevents cooling… it doesn’t affect the X-ray luminosity B 06 Model Data from Horner et al.

The AGN feedback loop (new version) AGN fuelling Cooling Hydrostatic ? Heating redistribute halo gas Based on the “excess energy” method (Wu et al 1999), plus the hydrostatic criterion

X-ray Emission from Groups and Clusters Ø L-T relation : well known that the selfsimilar relation fails AGN redistributes halo gas Ø AGN: standard model just prevents cooling Ø Revised model, AGN feedback redistributes halo gas until the cooling rate drops and AGN power is cut off Voit & Bryan 2001; Bower et al 2008, submitted Scatter driven by diverse assembly history A huge step forward - I’ve been trying to achieve this for ten years!

The baryon content of haloes where all those baryons? “Ejected” gas Stars and cold gas Hot X-ray emitting gas Halo mass

What about the galaxies? But note! The parameters have all changed!

Where are we? Ø Semi-analytics working well in many respects Ø Many aspects are coming out well! Ø Almost justified by numerical simulations (…discuss…) Ø But there are plenty of problems… Ø SCUBA galaxies Ø Morphology/Sizes (both in SA and numerical models) Ø Narrowness of the CMR Ø understanding BH accretion (Bondi can’t be correct!) Ø All the other problems…

I don’t believe any of this… with so many parameters you can fit anything!

Just how many parameters are there? Ø Not all parameters are equal Ø Some are set by external simulations Ø Some have a very weak effect Ø Some are physically degenerate Ø Just how many are there? Ø Input file contains 50 numbers (but many are legancy for older versions) Ø It makes sense to vary 20 parameters Ø 8 parameters dominate the variance Ø But acceptable models occupy less than 1% of the parameters space

Ø The methods… The space of acceptable models Ø Use model runs to sample the surface. Ø Latin hypercube provide maximum information on parameter dependencies Ø Construct “emulator” to interpolate between runs Ø Use low-order polynomial plus “Gaussian process”. Ø Rule out “implausible” regions of parameter space Ø Allow for emulator uncertainty make conservative choice Ø “What’s the answer” Ø How unique is the Bower 06 model? Ø How much do other properties vary within acceptable models? Ø Do parameter degeneracies have a physical interpretation? Ø Limit region of interest and generate a new wave of runs Ø surface is smoother and so emulator is more accurate With Ian Vernon & Michael Goldstein, Maths.

The Galform Parameter Space Ø 2 -sigma discrepant models occupy 1% of the volume. Ø Difficult to visualise an 11 -d space! Ø Project to 3 -d using the least discrepant point (still hard to fully sample!) Ø x, y, z = vhot, reheat, hot

The Galform Parameter Space Ø 2 -sigma discrepant models occupy 1% of the volume. Ø Difficult to visualise an 11 -d space! Ø Project to 3 -d using the least discrepant point (still hard to fully sample!) Ø x, y, z = vhot, reheat, hot

The Galform Parameter Space Ø 2 -sigma discrepant models occupy 1% of the volume. Ø Difficult to visualise an 11 -d space! Ø Project to 3 -d using the least discrepant point (still hard to fully sample!) Ø x, y, z = vhot, reheat, hot

Conclusions Semi-Analytic models - are we kidding?

Semi-Analytic Models: “are we kidding? ” Ø What I’ve told you: Ø Gas physics is difficult Ø Semi-analytics vs direct simulation Ø The challenges for galaxy formation models Ø Where we stand & future challenges Ø Environment Ø X-ray emission Ø Systematically exploring the parameter space… Ø Why you should listen! Ø Semi-analytic models are: Ø A fact of life Ø We need them! Ø Where do we draw the boundaries? Ø A method for multi-scale simulation Ø A tool for understanding physics Ø A tool for understanding observations �

Thank you!