AST541 Lecture Notes: Galaxy Formation Dec, 2016

Transcription

1 AST541 Lecture Notes: Galaxy Formation Dec, 2016 GalaxyFormation 1 The final topic is galaxy evolution. This is where galaxy meets cosmology. I will argue that while galaxy formation have to be understood in the context of cosmology, we learn very little about cosmology through studying detailed galaxy properties and their statistical distributions. Rather. the observational tests of galaxy evolution underline our poor understanding of barynoic physics on small scales; I also question whether we can make any progress through this approach, or maybe we should focus on detailed physics, before using galaxies as cosmology tools. Let s put it all together: Can we use everything we ve learned in galaxies and cosmology to create a single model that explains how galaxies form from perturbations we see in the CMB? Together with a good understanding of linear theory and early universe physics, this would describe the universe from the Big Bang until today. Of course, we are far from this goal, but much progress is being made. There is no easy theoretical predictions for galaxy properties. Models must be developed to follow highly nonlinear evolution, plus the detailed baryonic processes that shape observable parts of galaxies. While we think we understand the basic cosmological models fairly well, to translate that into a successful galaxy formation theory faces two main challenges: Dynamic range. Basic problem is one of dynamic range = Ratio of largest to smallest length scales in the problem. Large scale Gpc, small pc. So that s To represent with cells of particles, need the cube of this: Largest current gravity-only models achieve 10 <11, and full galaxy formation models are 100 smaller. We might believe that Moore s law on computing (Figure) will continue for a while, but it is still very far off. Galaxy physics. Furthermore there is the issue of the input physics, which is not well understood or modeled on small scales: e.g. B fields, star formation, etc. Even worse, pc scale phenomena like AGN and SNe can have large-scale impact. 1 Numerical simulations Cosmological numerical simulations attempts to capture as much non-linear physics as possible, given the limitation of computing power. In numerical simulations, we use particles to represent mass. Typically M per particle within Mpc cubic volumes. Note

2 GalaxyFormation 2 that at this level, a galaxy can be represented by of order a few to a few million particles, therefore there is no way to resolve detailed ISM processes and even all dynamical processes within a galaxy. We use Zeldovich approximation to carry the linear evolution of perturbation to the semi-linear region, and generate initial conditions. Then, there are two general flavor of simulations. 1. N-body simulation: Gravity only. - Particle-mesh (PM) code: Uses FFT to solve 2 Φ = 4πGρ in Fourier space. Then F = ρ. Very fast, but not adaptive in space. - Tree code: Represents distant groups of particles as multipole. Forces then summed directly. Slower, but adaptive. Force softening: To avoid two-body scattering, force is softened within ɛ, e.g. a = GM/(r + ɛ) 2. ɛ then sets the minimum resolvable scale in the simulation. Dynamic range is then the size of volume divided by softening length: L/ɛ. Current computing power allows dynamic range of few Multiple Timestepping: Particles in dense regions have shorter dynamical times. Their forces need to be recalculated more often. Depends on softening, e.g. t < ɛ/a. 2. Hydrodynamics: We treat baryons here. Including heating, cooling, shocks, hydrodynamical instabilities, which result in the high density regions in galaxy and lead to star formation and chemical evolution. The code will need to incorporate gaseous processes and chemistry, and radiative transfer when appropriate. Many ways to do this; much less accurate and much slower. Most commonly used method is Smoothed Particle Hydrodynamics (SPH), which is Lagrangian and hence easily adaptive in space and time. Mesh codes are faster and more accurately handle shocks & instabilities. dynamic range need adaptive mesh, which makes it much slower. But to get high However, it is clear that we are not actually simulating the small scale, sub-grid, physical processes happening in the galaxy. Still need rules for star formation (Schmidt Law) and feedback (AGN, SN). Advantage over SAMs is that rules can be implemented within the dynamical evolution. But really, main reason to burn the CPU hours is to model the dynamics as similarly to nature as possible, without having to make assumptions/rules about it. This removes the vast majority of SAM parameters, and makes predictions commensurately more robust (within the dynamic range afforded by the simulation). The advantage of using simulation is to capture as much physics as possible. The disadvantage

3 GalaxyFormation 3 is that it takes time, and it is difficult to explore the parameter space and see the impact of different cosmological models and physics assumptions on small scales. An alternative way to do this is the SAM approach. 2 Semi-Analytic Models: Idea is to connect (relatively) well-understood dark matter evolution to the evolution of baryons using observationally-constrained parameterizations. There are a few groups in the world doing this. Their approaches differ in detail, but with general agreements. A typical SAM model consists of the following steps: 1. Choose a cosmological model, and the initial fluctuation power spectrum. 2. Use either N-body simulations, or EPS, to trace the merger histories for a series of DM halos. 3. Adopt prescriptions to specify the dynamical evolution of the various components when halos merge. Including: hot gas components in the progenitors are heated to a new virial T as they merge. The least massive progenitor becomes a satellite galaxy in the new system and ceases to accrete gas. The most massive progenitor becomes the new central galaxy and keeps on accreting as far as cooling is efficient. 4. Assume that satellites can merge with central objects on a timescale set by dynamical fraction and make simple assumption of the output of the merger. Presumably, if the mass of the satellite is small, it is cannibalized without major impact on the morphology of the central galaxy. If they are comparable, the disks of both progenitors are likely to be destroyed, resulting in the formation of an elliptical. If new gas is able to subsequently cool, a new disk may grow around the spheroid, giving rise to a galaxy consisting of disk and bulge. 5. Within each DM halo, follow three distinct baryonic components hot gas, cold gas (disk) and stars. 6. Use simple prescriptions to specify the conversion rates between these baryonic components: cooling converts hot gas into cold gas, star formation converts cold gas into stars and feedback from massive stars and AGNs either converts cold gas into hot gas, or reheat hot gas directly. 7. At the same time keep track of the metallicity of each of these components using chemical evolution models.

4 GalaxyFormation 4 8. Use a stellar population model to convert the star formation history and metallicity of the stellar pop into luminosity and color, including dust. 9. Repeat this process for a large number of halos. Thus, in SAM, the complicated, tightly intertwined astrophysical processes are modeled with a set of recipes and prescriptions, which carry a number of free model parameters. Some typical rules: 1. Stars form out of disk based on Kennicutt-Schmidt Law. 2. When halos merge, the galaxies merge on a dynamical friction time. 3. Hot halo gas is redistributed into new (larger) halo. 4. Major mergers (e.g. > 1 : 3) result in elliptical, else grow spiral. 5. If spiral, cold gas is redistributed into a new disk. 6. If elliptical, remaining cold gas is heated to virial temperature. 7. Any galaxy that falls into a larger halo but hasn t yet merged has gas supply shut off. 8. Ellipticals can regrow disks by subsequent small gas-rich mergers. Can add many more rules: e.g. for AGN/BH growth and feedback, detailed movement along Hubble sequence, metal enrichment, supernova feedback, secular evolution (bars), etc. Each rule typically has parameters associated with it. These parameters govern not only microphysics but also gas and stellar dynamics. Typical SAM has parameters. Parameters are constrained by comparison with a wide suite of data, typically at z 0 since that s where most detailed constraints are available. Hence SAMs usually represent local Universe accurately by construction. Then predictions can be made for higher-z galaxies and galaxy evolution. The observations SAM tries to fit include: Primary constraints: morphological mix, metallicity, Tully-Fisher, LF, sizes and gas fractions. Further constraints: M/L, color, color-morphology, star formation history, clustering, luminosity-metallicty, AGN etc. Advantages: Calculations are fast and flexible can explore wide range of model variations. Can be tuned to match desired data very closely (good for e.g. mock catalogs/cosmic variance). Disadvantages: Big uniqueness problem many SAMs exist today, all match lots of data, but do so with very different parameterizations. This limits physical insight gained.

5 GalaxyFormation 5 Opinion: SAMs are good for lots of things, but learning the physics by which galaxies form isn t one of them. 3 Successes of Galaxy Formation Models - Models are very successful at predicting the matter distribution on scales larger than galaxy halos: Correlation functions, large-scale mass distribution, etc. - Models generally predict that cosmic star formation and galaxy assembly was much more rapid in the past (z 2 4) than today. - Models are able to predict merger-driven morphological evolution as observed (though not in detail fractions of various Hubble types). - Models get the general structure of disk galaxies correct: Disk w/spiral arms, bulge, halo. Finer details, like globular clusters and metallicity gradients, remain poorly understood. - Models qualitatively reproduce properties versus environment: Older, earlier type, more massive, more enriched galaxies in denser environments. 4 Current Challenges to Galaxy Formation Models 1. Dwarf galaxy problem: Why are there so few Local Group dwarf galaxies, when an LG-sized halo is supposed to have hundreds to thousands of subhalos? - Photoionization retards gas s ability to cool. T IGM = 10 4 K means that halos with T v less than this can t cool gas to form stars. Corresponds to M vir = 10 8 M, or v c 20 km/s. - BUT... v c function is shallow all the way to 100 km/s. - If photoionization retards galaxy growth early enough, say only 10 8 M halos that form before z = 7 can form stars, then can affect much larger halos through hierarchical growth. - BUT... Dwarf galaxies in LG show bursty SF histories all the way until today. - Seems like some other feedback required- supernovae? 2. Cusp problem: Why do density profiles of at least some LSB dwarf galaxies and clusters seem to be shallower than NFW predicts? - LSB dwarfs and clusters are dominated by dark matter very far into the center, and hence can probe non-luminous contribution to potential down to 1% of virial radius.

6 GalaxyFormation 6 - BUT... cleanest cases generally show a rise in v much slower than predicted from NFW. - Non-circular motions or misidentification of center can mimic shallower slope. - BUT... 2D IFU maps also show less central mass concentration than NFW. Also cluster mass maps from lensing & dynamics indicates shallower profile. - Wide range seen, so perhaps some cases are not relaxed. - Also, simulations have yet to completely converge on prediction, and no inherent understanding of why NFW arises. 3. Cooling flow problem: Why are large galaxies not forming more stars? - Need a constant (or recurring) source of non-gravitational energy input. - BUT... no star formation. no source of photoionization. where does energy come from? - Black holes contain large energy reservoir; small amount of accretion can in principle yield large amount of energy. - BUT... AGN activity in large ellipticals is rare: Only 1% show activity. - Perhaps a recurrent, self-regulating process? A little bit of cooling results in lot of feedback energy, which then stops cooling, stops feedback, and the system re-established CF which starts process all over again. - BUT... feedback energy from AGN comes in form of jets, how to stop spherical cooling? - Sound waves? Hand waves? 4. Downsizing problem: Why does the typical SF galaxy have lower mass with time? Why did large galaxies form their stars a long time ago and none recently? (tight red sequence) - Virial shock forms only in large mass halos, because in small halos the cooling radius is outside the virial radius. - BUT... central regions are denser, and will still cool if undisturbed. - Major mergers that form early types are somehow responsible for suppressing future SF? Would also evolve with z because major mergers more common in past. - BUT... mergers throw out very small amount of gas dynamically. - Need mergers to produce some sort of feedback. Do they generate quasars? (Hernquist group s models) Seem to agree with many properties of AGN (Hopkins papers). - BUT... BH feedback likely highly anisotropic, whereas current models assume isotropic energy injection. 5. Dry merger problem: Why do the largest galaxies not grow from z = 1 0, although

7 GalaxyFormation 7 their halos should have grown by 2? - Halos merge but galaxies don t merge? - BUT... see only 1 large cd galaxy in big clusters at all redshifts (to 1.5). - CDM predicts higher merger rates in the past. - BUT... testing this very hard; can find close galaxies with low v but still don t know orbits. - Does hierarchical structure formation (and hence Press-Schechter/ Sheth-Tormen/SAM models) somehow break down at high masses? 6. Angular momentum problem: Disk galaxies are extremely hard to form in models; too much early merging, never get Sd/bulgeless galaxies. In general, unsolved galaxy evolution issues point to two directions: small scale, dwarf galaxies, solution is to understand dark matter. massive galaxies, solution is likely feedback. Having spent most of the semester covering subjects that we claim to know something about, I m going to spend this lecture talking about things we know little about! As I emphasized a number of times in our class, we now have a very good ΛCDM model which describes our cosmological observations from BBN, to CMB, the the expansion history and growth of the universe. However, it is clear that we are still missing some fundamental physical understandings of why this model works. When I think about how the field of cosmology is going, I see three basic areas of research: Early Universe. What s the origin of the expansion of the universe and fluctuations in the universe? Dark matter and dark energy. Both astrophysical measurement and physical understanding, the latter of which is very closely related to the first. First light and reionization. This is where cosmology first met galaxy formation. Unlikely low-z galaxy evolution, this is still a strong test to cosmology, as cosmological environment likely play a dominant role. Note that I leave the subject of galaxy evolution out. I think many people view that galaxy evolution is by and large a problem of galaxy physics while cosmology provides the necessary background. However, because of the highly non-linear and complex physics involved,

8 GalaxyFormation 8 opportunities for direct cosmological tests using observations of galaxy properties and their evolution, other than using them as test particles or distant indicators, are limited. On a number of occasions in this class, I also talked about the somewhat contradictory nature of this field, the glass-half-full vs. glass-half-empty views. On one hand, we have the beautiful results from both CMB and low-z redshift survey, that allow all cosmological observations to be united into a simple 6/7 parameter model, on the other hand, 95% of the universe remains unknown to us. This very much reminds us the situation a bit more than 100 years ago, when physicists enjoyed the great success of thermal dynamics and E&M, while great puzzles, such as the UV catastrophe of black body and the constant speed of light pointed to deep unknown components of physics. I personally believe that within our lifetime, we will see great advance in physics that finally unveil the true nature of the observations that we now call dark energy and dark matter, and I strongly feel that they are likely not what we think they are, namely, vacuum energy and massive weakly interacting particles. And the only way to find out is to continue conduct cosmological observations, by testing current models in the parameter space that we have yet to reach.