Statement of Research Interests:
Feedback Processes in Galaxy and IGM Evolution

Romeel Davé, Steward Observatory
May 2008

Overview: The evolution of our Universe is traced observationally using two primary probes: Galaxies and intergalactic gas. Understanding the evolution of these observable cosmic components from the Big Bang until today is the primary goal of extragalactic cosmology. My work focuses on using high-performance computing to understand critical questions at the frontiers of galaxy formation theory.

Solving galaxy formation subdivides into two fundamental aspects: Understanding the accretion of material from the diffuse intergalactic medium (IGM) into the dense regions where stars form, and understanding the feedback on gas from energetic processes of galaxy formation such as supernovae and black hole accretion. The theory of accretion has a long history that dates back to seminar works in the 1970's, but recent studies by our team (Keres et al. 2005) among others have yielded interesting new interpretations that have revised our core understanding of how stellar growth relates to dark matter growth. In particular, we identified ``cold mode" accretion (now a standard term), where gas cools via line emission in dense filamentary structures feeding galaxies, as the dominant accretion path for most galaxies across cosmic time.

In comparison to accretion, galactic feedback processes are not well understood. They are detected observationally but are difficult to quantify, and they are expected theoretically but are poorly constrained. A key theoretical issue known as the ``overcooling problem" identified in the 1970's shows that feedback must be important: Without it, the amount of mass ending up in stars is predicted to far exceed that observed. Hence feedback must suppress star formation by large factors. Feedback is hence clearly a critical element of galaxy formation theory, and lies at the heart of forefront work in this field.

Som candidate physical processes to suppress star formation are photoionization from young stars, which might heat cold gas both in and out of galaxies; supernovae that impart substantial heat and kinetic energy to ambient cold gas; and active galactic nuclei (AGN), where accretion onto the central black hole results in both jets and hard photons from accelerated electrons. These processes have all been observed to be operating in galaxies. More exotic processes such as magnetic conduction and cosmic rays may also have an impact, but its impact on star formation has not been directly seen.

My work mainly focuses on understanding these three major feedback processes in galaxy formation. I use cosmological hydrodynamic simulations of galaxy and structure formation. based on the smoothed particle hydrodynamics (SPH) technique, beginning with the world's first parallel TreeSPH code (Davé, Dubinski & Hernquist 1997) that I wrote and used for my thesis. Thanks to algorithmic, speed, and input physic improvements, I now mostly use Gadget-2 from V. Springel, with substantial additions made by our group.

Feedback from star formation: This has been my primary focus in the past several years. My graduate students and I have made fundamental contributions in this area, and our group is internationally recognized as leading current studies of cosmic-scale galactic outflows.

Feedback from star formation must rapidly distribute matter over enormous (megaparsec) scales, far beyond galaxies themselves. This is evident because the diffuse intergalactic medium (IGM) is seen to be enriched with heavy elements (>He, produced exclusively in stars) by redshift z~5, or ~10% of the Universe's age. The IGM can be traced by absorption lines seen against distant quasar light. The most commonly observed line is HI Ly-alpha (1216A), which falls in the optical band at z>2. The lines are so thick they are referred to as the Ly-alpha ``forest". High-quality data now available from 8-10m class telescopes has characterized both the Ly-alpha forest and associated metal lines (mostly CIV and OVI, as it turns out) with exquisite precision.

My thesis work primarily involved using simulations to study the Ly-alpha forest absorption lines. In Davé et al. (1997) I developed an automated tool to fit Voigt profiles to simulated and real quasar spectra, enabling the first true side-by-side comparison of IGM simulations and data. This work helped constrain models and establish a new paradigm for the Ly-alpha forest as arising from diffuse IGM gas tracing uncollapsed filamentary structures in the Cosmic Web. In order to understand the evolution of the Ly-alpha forest to the present epoch, I had to develop the world's first distributed-memory parallel cosmological hydrodynamics code, PTreeSPH~(Davé, Dubinski, & Hernquist 1997). This code harnessed the computing power of Beowulf clusters to evolve simulations over the latter 80% of cosmic time from z=2-0. My work on the low-redshift Ly-alpha forest (Davé et al. 1999) established the drivers of IGM evolution as cosmic expansion and a rapidly dropping metagalactic ionizing background. This settled a major controversy on the relationship between low and high redshift Ly-alpha forest absorbers, and provided a unified model for IGM evolution from z~6-0.

During this time, I also worked on understanding metal absorbers in the IGM. In Hellsten et al. (1997) we developed machinery to generate metal absorption spectra from simulations, using the photoionization code CLOUDY. We invented the LOX (line observability index) to characterize which metal ions are optimal for tracing IGM gas of a given overdensity. In Davé et al. (1998) we extended this work to study the best lowest-density IGM tracer, OVI. Because OVI (1032,1038A) falls within the dense Ly-alpha forest, we developed new pixel-based tools to assess HI ``contamination". We obtained the first constraints on the metallicity and ionization state of the z~3 IGM at densities near the cosmic mean.

These analyses inserted metals ``by hand" into simulations, as was commonly done at the time. This begged the question, how exactly did metals get transported from galaxies into the diffuse IGM? To answer this, student Ben Oppenheimer and I implemented dynamical models for galactic outflows into Gadget-2. We initially followed V. Springel's outflow implementation, but modified it to account for recent observations of local starburst outflows that showed the outflow velocity scaled with galaxy circular velocity, as expected for momentum- or radiation-driven winds. These observations provided, for the first time, a concrete physical model for how outflow properties should be tied to galaxy properties.

Our initial results (Oppenheimer & Davé 2006) showed that IGM enrichment data, most commonly traced by CIV absorbers, were very constraining for outflow properties. If outflow speeds are too low, the metals do not reach the diffuse IGM as observed, whereas if they are too high, outflows overheat the IGM and produce overly large CIV linewidths. If the mass in outflows is too low the diffuse IGM contains too few metals, whereas if mass loss is too efficient, star formation becomes so suppressed that not enough metals are produced overall. Remarkably, the momenentum-driven wind scalings calibrated by local starburst data turned out to be the best-fitting model amongst the dozen or so we tried. This suggested that ubiquitous winds at high-z may be analogous to the rare winds seen locally, with the primary difference being that star formation is more vigorous in early galaxies.

The success of our momentum-driven wind model was impressive and the connection to local outflows satisfying. We then explored a number of implications of such outflows. In Davé & Oppenheimer (2007) we we presented a novel solution to the ``missing metals problem", in which the amount of metals seen in z~3 galaxies falls an order of magnitude short of the metals produced by the stars in those galaxies. Our simulations suggested that the IGM actually hides metals in a warm phase, since the winds ejecting the metals concurrently heats the IGM and alters the ionization state from pure photoionization. In Davé, Finlator, & Oppenheimer (2007) we compared outflow models to observed z~6 galaxy luminosity function, and found that an order of magnitude suppression in star formation was required to solve overcooling, which is naturally achieved in our momentum-driven wind scenario. In Davé, Oppenheimer, & Sivanandam (2008) we showed that momentum-driven winds can successfully reproduce the observed enrichment and entropy excess seen in intragroup gas, something that no previous model had been able to do.

In Davé, Finlator, & Oppenheimer (2007) we found that momentum-driven winds also best matched the observed galaxy mass-metallicity relation (MZR). In Finlator & Davé (2008) we developed a simple analytic model, guided by simulations, for the origin of the MZR. Our model is completely different from the canonical view of the MZR being governed by the outflow speed relative to the galaxy's potential depth (around since the 1980's), and instead suggests that the MZR is an equilibrium relation reflecting the balance of recent accretion relative to star formation. If true, the MZR is providing tight constraints on mass outflow rates. This model has attracted considerable attention, and is now standardly presented as an alternative to the conventional scenario.

In the past year, Ben Oppenheimer and I have made major improvements to our outflow and chemical enrichment modeling, including independently accounting for Type~Ia and Type~II supernovae and AGB mass loss. In Oppenheimer & Davé (2008) we studied the dynamics of outflows across cosmic time, using our momentum-driven wind model, and found (contrary to conventional wisdom) that wind material ejected from a galaxy is far more likely to return to that galaxy than stay in the IGM, despite being ejected at speeds far in excess of the escape velocity. Wind recycling happens on typical timescales of 1~Gyr, and the median wind particle goes out to roughly 100~kpc physical at all epochs, for all galaxy masses (though there are significant trends with each). At late times, wind recycling may produce what we call ``halo fountain" in galaxies such as the Milky Way, and may dominate over pristine accretion from the IGM. Our work closely ties galaxies and IGM in a sort of cosmic ecosystem of inflows and outflows in a manner never before appreciated.

The momentum-driven wind scenario has been remarkably successful, but much work remains to be done to investigate new physical consequences of such outflows, and to develop a less heuristic description of galactic winds. Ben and I are actively working on a unifying scenario for IGM metal evolution from high to low redshifts as traced by various ions, studying how wind recycling might be observed in absorption and emission signatures around nearby galactic halos (including my work on the IGEM SMEX mission proposal), and characterizing the Ly-alpha and metal line forest for upcoming observations with HST's Cosmic Origins Spectrograph (I was involved with 5 Cycle 17 COS proposals). I am working with individual galaxy simulators to understand the physics of wind propagation through ISM and halo gas at very high resolution. There is a vast uncharted territory of observational comparisons to be done, notably in the evolution of galaxy gas and stellar contents.

Feedback from stellar photoionization: The epoch of reionization (z>~6) is a key frontier in modern cosmology, when the first stars slowly ionized away ambient neutral hydrogen of the Dark Ages. Led by student Kristian Finlator and with Feryal Ozel (Physics), we are significantly advancing the theoretical study of this epoch by developing a fast, robust, and accurate radiative transfer hydro code for studying reionization-epoch galaxies.

He is currently essentially done developing the radiative transfer code. In his first paper (this summer) he will use it to study the effects of radiation on the environments of the first forming galaxies. Then he will work on merging the radiative transfer with Gadget-2, and conduct the first fully self-consistent radiation hydro simulations of early galaxy formation. In our earlier work (Finlator et al. 2005), Kristian developed sophisticated tools to produce, for example, mock JWST observations that can be analyzed alongside real data. Already, some reionization-epoch galaxies have been observed with Spitzer, and in Finlator, Davé, & Oppenheimer (2007) we compared their photometric properties to simulations using a Bayesian SED fitting program Kristian developed called SPOC. Using these tools, Kristian will be well-positioned to be at the forefront of predicting JWST observations with e.g. NIRCAM (M. Rieke, PI). Kristian and I are also involved with a large Hubble/Spitzer program to study lensed reionization-epoch systems (E. Egami, PI). The code Kristian is developing will be a unique, forefront tool for studying a wide range of topics, from 21cm absorption/emission to be seen with upcoming radio arrays, to metal absorbers as tracers of reionization topology, to understanding Helium reionization, to studying damped Ly-alpha systems.

Feedback from AGN: Student Jared Gabor will begin working on this topic in the Fall. Our approach will be heuristic, much like with outflows, in trying to find a feedback model that quenches star formation in a manner consistent with a wide range of observations. This will then be used to study massive galaxy evolution, the development of the red sequence, the post-starburst galaxy phenomenon, and the energetic impact of quenching on intergalactic and intracluster gas.

IMF evolution: Recently, I embarked on a tangent investigation of whether the stellar initial mass function (IMF) might vary with cosmic epoch. Once a taboo suggestion, several papers have forwarded this idea in the past year owing to a host of puzzling observations. In Davé (2008) I presented arguments for this centered around the evolution of the stellar mass-star formation rate relation in simulations versus data. I was the first to present a concrete testable model of IMF evolution that reconciles observed and predicted galaxy evolution from z~2-0.

Research with Arizona students: As one of the few theorists in a mostly observational department, I have attempted to make our models available to students who want a theoretical component to their thesis. I have worked Chris Impey's students Kris Eriksen, Andy Marble, Lei Bai, and Caitlin Casey, on various aspects of Ly-alpha forest observations. Marble et al. (2008) examined how Ly-alpha flux statistics could be employed for an Alcock-Paczynski test, and Casey et al. (2008) probed Ly-alpha flux coherence using a lines of sight pair.

I worked Ann Zabludoff's students Yujin Yang and Suresh Sivanandam. For Yujin, I provided simulations to investigate Ly-alpha blobs (Yang et al. 2005), to complement his Magellan search for these enigmatic objects. Suresh followed up his work on intracluser light by developing code to compute X-ray flux of from gas particles' [rho,T,Z] (Davé, Oppenheimer & Sivanandam 2008).

I worked with former students Desika Narayanan (advisor C. Walker) and Hu Zhan (Physics). I helped with Desika's CO radiative transfer simulations, and was involved with four papers that comprised his thesis. His most recent and important work showed that the Schmidt Law fundamentally governs how gas turns into stars, and refutes other recent suggestions tied to molecular gas mass. In Zhan et al. (2005) we studied the Ly-alpha flux power spectra from high to low redshift, to assess its value for tracing cosmological evolution.

Other scientific interests: I am interested in constraining the nature of dark matter from astrophysical observations, and have worked on simulations of self-interacting dark matter (Davé et al. 2001), constraining the mass of warm dark matter (Narayanan et al. 2001), and the neutrino mass (Croft, Hu & Davé 2000). I am involved with an effort to bring a new NSF center on dark matter to Arizona (led by I. Sarcevic, Physics). I am also interested in the connection between galaxies and large-scale structure; I developed Filament Statistics to quantify topology (Davé et al. 1997), measured Omega using the mass-to-light ratio as a function of scale (Bahcall et al. 2000), studied halo occupation distributions (Berlind et al 2003, Zheng et al 2004), compared simulations with SDSS galaxies (Berlind et al 2004), used clustering to infer the halo masses of Damped Ly-alpha systems (Bouché et al. 2005), and investigated the relationship between satellites and central galaxies (Weinberg et al. 2008). I am involved with observational projects on low-redshift Ly-alpha and OVI absorbers and their relationship to galaxies (T. Tripp, UMass & J. Prochaska, UCSC), a multiwavelength absorption/emission study of the Clowes/Campusano Large Quasar Group (G. Williger, Louisville), and NASA mission concepts Inter-Galactic Emission Mapper (PI: S. Chakrabarti, Boston Univ.) and Baryonic Structure Probe (PI: K. Sembach, STScI).