Julio  Navarro

Research Interests
Galaxy Formation through Hierarchical Clustering
 
A collaborative effort that involves Matthias Steinmetz, Carlos Frenk, and Simon White.
 
 

The standard lore envisions galaxies forming as a result of the dissipative collapse of gas within an
evolving population of dark matter halos. In systems where the dynamical timescale exceeds the cooling timescale the gas component can radiate away efficiently its pressure support and collapse until it settles into a centrifugally supported disk, where it starts forming stars. The gas fraction that participates of this collapse and is eventually transformed into stars determines the total luminosity of a galaxy. The angular momentum of the gas determines the size and surface brightness of a disk galaxy. Combined with the total mass and spatial structure of the halo, it also sets the rotation velocity and the shape of the rotation curve of a galactic disk. Computing the amount of gas that can cool and form stars within a dark matter halo as well as its angular momentum is therefore of crucial importance for understanding the origin of disk galaxies and the significance of scaling laws that relate different properties of the galaxy population such as the Tully-Fisher relation.

Our numerical experiments have provided interesting clues to the process of hierarchical galaxy formation[4,5,9,10,15] . Dissipative effects are extremely efficient at high redshift (when all systems, and the universe as a whole, were in general denser than today) and, in the absence of a reheating mechanism, most of the gas in the universe collects at the center of protogalactic halos as so on as they collapse. There, it settles into tightly bound, rotationally supported disks which, as halos collide and merge, sink to the center of the new halos, eventually merging into a single object (Figure 1).
 

Face-on View
Edge-on View 
 
 

Figure 1:  Particle plots of a gaseous disk formed through dissipational collapse within a dark matter halo. Gas particles are shown in green and dark matter particles in red.

Radial Velocity vs radius 

Rotational Velocity vs radius 
 
 



 

Figure 2: Radial and tangential velocity profiles of the gas and dark matter. Note that the gaseous disk is centrifugally supported and that  the rotation curve is approximately flat out to about 20 kpc.

This efficient cooling and early collapse of baryons is very difficult to reconcile with the observation that galaxies contribute a very small fraction of the total baryonic density of the universe implied by primordial nucleosynthesis constraints and with the existence of massive amounts of X-ray emitting gas in clusters of galaxies.

A related problem that arises if disk galaxies assemble most of their mass through merging of the baryonic cores of protogalactic halos is that substantial angular momentum losses occur during the mergers of these cores. Although gaseous disks formed this way have several properties in common with spiral galaxies, such as flat rotation curves (Figure 2) and roughly exponential disks, excessive angular momentum transfer between gaseous disks and dark halos during mergers result in disks that are much more concentrated than real spirals. Figure 3 shows the specific angular momentum of dark matter halos (filled circles) and gaseous cores (open circles) as a function of mass for an ensemble of 30 simulations of galaxy formation in the standard CDM cosmogony. The specific angular momentum of gaseous disks is, on average, only a tenth of that of their surrounding halos, although tidal torques gave both components approximately the same spin during the expansion phase. Typical spiral galaxies have higher angular momenta than the gaseous disks formed in these simulations.

These are potentially serious problems for any hierarchical clustering scenario, and it is clearly important to investigate physical mechanisms that can reduce the efficiency of cooling, such as energy feedback from supernovae or the effect of a photoionizing background. It is possible that the same mechanism that prevents the baryons from condensing at the center of small halos at high redshift can change the dynamics of the collapse of the gas, allowing it to conserve a larger fraction of its initial angular momentum and leading perhaps to better agreement with observations.

We were able to show that the effect of a photoionizing background such as that inferred from the proximity effect on QSO absorption lines is insufficient to prevent the early collapse of large amounts of gas into low mass halos[23]. This implies that further progress towards understanding the formation of disk galaxies requires a realistic treatment of the effects of star formation and feedback from supernovae and evolving stars on the interstellar medium.
 

Modeling Star Formation and Feedback

Our poor grasp of the physics of star formation compromises numerical approaches that attempt to describe this process and forces us to adopt crude ``recipes'' for including star formation in numerical experiments.  In a particle-based treatment such as ours, star formation is most naturally modeled by the creation of collisionless ``star'' particles in regions where the gas is locally Jeans unstable and where the cooling timescale is shorter than the local dynamical timescale. The orbits of these newly formed stars are subsequently followed in a self-consistent fashion, assuming that they are only affected by gravitational forces.  At any given time, each star particle represents a stellar population of a few million stars and is labeled by an age (time since creation) and a metallicity equal to that of its gas progenitor at the time of formation.  We can then use population synthesis techniques to ``observe'' our galaxy models in arbitrary bandpasses. Our libraries contain spectral synthesis of stellar populations of arbitrary metallicity and the UBVRIJHK filters.

Young star particles can also devolve energy and mass to their surrounding gas, an effect that attempts to mimic the energy and mass input by supernovae and evolving stars into the ISM. The energy released by each star particle can be invested in heating up the neighboring gas or in perturbing its kinetic energy. Preliminary experiments[5] indicate that the results of a simulation are extremely sensitive to the fraction of energy invested in modifying the gas bulk motions, fv. This is the single most important free parameter in the whole procedure and can in principle be constrained by analytic models of SN explosions and by 1-D simulations of the interaction of SN remnants with the ISM.

Because our numerical treatment is fully Lagrangian, we can follow the process of metal enrichment in a spatially resolved manner.  Our code tracks ``metals'' dumped into the neighboring gas by young star particles. These metals are assumed to be released with two distinct timescales, corresponding roughly to the different evolutionary timescales of type I and type II supernovae[20]. This procedure allows us to study the global metal enrichment process of a galaxy, the establishment of metal gradients within galaxies, and the amount and composition of metals expelled from a galaxy.
 
 

The morphology, nature and appearance of galaxies at high and intermediate redshifts

The quality of HST images, combined with the spectroscopic capabilities of 10m-class telescopes, provide spatially resolved information about galaxies at redshifts ~3 or higher and allow us to measure their luminosities, sizes and potential depths. Only high-resolution simulations can address directly issues regarding the morphology and internal dynamics of individual galaxies at high redshift, and exploit fully this wealth of observational information. The power of our technique is illustrated in Figure 4, where we show the high-redshift appearance of the progenitor of a galaxy like the Milky Way, taken from a Cold Dark Matter simulation similar to the ones we intend to perform. This protogalaxy is shown at redshift of ~3, and has been imaged in six different broad-band colors, UBVRIK. The images include a point spread function similar to that of HST, but neglect noise and the effects of dust extinction or intervening absorption. This image illustrates how our simulations can be ``observed'' and compared directly high-resolution observations of distant galaxies.  Much work is needed before such image can be compared directly with those of HST, but the morphological and dynamical detail that we can reach in our numerical experiments demonstrates the vast potential of our technique. Another example is the computer-generated Hubble Deep Field (courtesy of Matthias Steinmetz).

By modeling galaxies selected to have different morphological types at z=0 we expect to gain insight into the interplay between the underlying cosmology and the structure, kinematic properties, and observability of the progenitor(s) of galaxies like our own. Also, by matching simulations to observational data at different redshifts we hope to constrain the star formation history of galaxies of various morphological types and to learn of possible signatures that may relate particular cosmological models with the appearance of galaxies at different redshifts.

The origin of correlations linking global galaxy parameters

The origin of global correlations linking the mass, velocity dispersion, luminosity, size and metal content of galaxies, such as the fundamental plane, the Tully-Fisher relation, or the color-magnitude relation has eluded astronomers for many years. Although there has always been strong suspicion that these correlations reflect cosmologically induced differences in the formation paths of systems of different mass, there is no consensus regarding which parameter actually controls the shape or normalization of these scaling laws.

We expect that the numerical study we intend to carry out will help identify the causes of these observed regularities between galaxies of a given morphology. We plan to simulate a large number of systems of different mass, and to analyze the sensitivity of the resulting scaling laws to the particular cosmological model adopted and to our assumptions about the star formation and feedback processes. Since our simulations contain spatially resolved information, we will be able to assess the importance of possible biases introduced by observational procedures. For example, we will be able to compute Tully-Fisher relations in different passbands, constructed from simulated HI velocity widths and from rotation curves (optical or HI) at different redshifts. A similar analysis will be applied to the progenitors of present-day elliptical galaxies. These results can be directly compared with ongoing surveys that address the shape and normalization of these scaling laws at different redshifts.

Early results are encouraging[7,11,14] We have included the effects of star formation in a subsample of simulated halos which have suffered recent major mergers and are therefore unlikely to harbour stellar disks.  Stellar systems formed this way are found to be slowly rotating, mildly triaxial ellipsoids supported by an anisotropic velocity dispersion tensor. They also appear to obey similar correlations to those of real ellipticals, giving some insight into the origin of the ``fundamental plane'' and the Faber-Jackson relation. This is an example of the way in which outstanding problems regarding the origin and evolution of elliptical and spiral galaxies can be approached in the numerical study we intend to carry out.
 
 
 

Ly-alpha absorbers and present-day galaxies

Over the past few years, cosmological gasdynamical simulations of hierarchical structure formation have revolutionized theoretical studies of Ly-alpha absorption systems in QSO spectra. The large scale distribution of gas seen in these simulations provides a natural and elegant explanation for the number density, the redshift evolution, and the relative abundance of intervening absorbers of different column density. In these simulations the Ly-alpha forest arises as lines of sight to QSOs cross filaments or the outer regions of galaxy halos, and even as the result of caustics in velocity space of only slightly overdense structures. Lines of sight which penetrate galaxy halos or that cross the central regions of gas rich galaxies account for Ly-limit and damped Ly-alpha systems. These numerical simulations have solved a number of apparent contradictions regarding size estimates of these systems and pointed out large systematic uncertainties in determinations of metal abundances that assume photoionization equilibrium. These examples demonstrate clearly the potential of these simulations for interpreting high-resolution observations of QSO spectra.

None of the simulations published so far include star formation, but future investigations are likely to benefit dramatically from its inclusion. The simulations we propose can track the metal enrichment and abundance patterns in gas that inhabits galaxy halos and that has been expelled from galaxies by supernova-driven winds. These simulations would thus provide clear interpretation for observations of the metal abundances in low column density systems.  One issue of particular interest that could be addressed with our techniques concerns recent observations that indicate that all Ly-alpha alpha clouds with measured metal abundances seem to exhibit a minimum metallicity of about 0.01 Zsolar. This observation constrains very strongly the star formation process and the efficiency of feedback. Our simulations can help sort out whether this indicates that essentially all the intergalactic medium (IGM) has been polluted by Pop III stars that have since faded out of view or whether feedback is so strong that metal enriched gas can be expelled from galaxies and pollute the IGM far beyond the sites of galaxy formation. We are confident that our numerical simulations will show under which conditions (or for which cosmological models) such strong mass losses are expected or even allowed. Finally, full knowledge of the thermodynamic state of the gas provided by our simulations can be used to simulate spectra and to find reliable ways to limit uncertainties in estimates of metal abundances from analysis of absorption lines.

Another planned application of our simulations concerns the relationship between damped Ly-alpha systems and evolving galaxies. Simulations which neglect star formation are not ideal for this task because gas densities tend to be severely overestimated unless stars are allowed to form. We intend to use our simulations to characterize the optical and infrared properties of galaxies responsible for damped systems. This can be used to interpret the observations of the few cases in which galaxies responsible for the damped system have been identified and to improve the observing strategy of future search. We also expect to address issues such as the typical distance from a galaxy at which damped systems arise, and the extent to which actual kinematic and spatial information can be extracted from studies of unsaturated lines in these systems. We expect these numerical studies to have a strong impact in this very active area of research.