I work on the physics of galaxy evolution. My main tools are numerical simulations, and I like to connect them with observations to get clues about the underlying physics. Many aspects of galaxy evolution interest me, but my work has focused on galactic inflows and outflows, the role of AGNs in galaxy evolution, and the quenching of star formation.
Inflows and Outflows
Galaxy evolution is largely regulated by complex inflows and outflows of gas that connect galaxies to the cosmic web. Inflows fuel galaxies over cosmic time, enabling them to maintain star formation for many gigayears. Difficult to detect directly, inflows are widely predicted by cosmological models, and they are all but required to explain the star-forming "main sequence" of galaxies with gas depletion times of only ~1 Gyr. But exactly how the gas gets into galaxies remains a difficult question.
Outflows are widely observed in strongly star-forming galaxies. They help control what fraction of available gas gets converted into stars. Outflows should be more efficient at removing gas from low-mass galaxies because the gravitational potential well is easier to escape -- this helps explain the extremely low stellar-to-dark matter ratios seen in small galaxies. Outflows also enrich the inter-galactic medium (IGM) and circum-galactic medium (CGM) with metals. Gas ejected from galaxies may "recycle" into galaxies at a later time. The launching of winds, their propagation through the IGM, and their interactions with inflows all are crucial to our understanding of galaxy evolution.
***
Delayed star formation due to turbulence-driving by gas inflows
Some recent papers examined how inflowing cold streams of gas could drive turbulence in high-redshift galaxies. Most calculations determined that cold inflows could not be driving high levels of turbulence in observed z~2 galaxies -- feedback from stars seems the likely culprit. But at very high redshifts, where gas inflow rates are higher, cold inflows could play an important role. We made some simple calculations to examine how this inflow-driven turbulence driving might affect star formation in high-redshift galaxies.
Our analytic model indicates that cold inflows could substantially delay star-formation in the galaxy. As gas falls into a galaxy, it carries gravitational energy with it. If the inflowing gas couples efficiently with the galaxy disk, it could drive turbulence whose dissipation timescale is longer than the normal cooling time. Thus the galaxy disk is "puffed up" due to the inflows, and the less-dense gas forms stars less rapidly.
An idealized simulation with cold stream inflows produces higher gas velocity dispersion, more stable gas, lower average density, and lower star formation efficiency than without inflows.
In order to better understand the efficiency of coupling between inflows and the galaxy disk, we ran some idealized simulations of gaseous infall into high-redshift galaxies. Unlike most simulations, resolution was maintained relatively high even outside the galaxy. These simulations demonstrated that the coupling efficiency can be quite high in small galaxies at high-redshift -- much of the gravitational energy of infall goes into driving turbulence. They also showed that the star-formation rates are small relative to the amount of gas available.
***
Propagation of galactic winds
The fate of galactic winds plays a crucial (and somewhat mysterious) role in galaxy evolution. Winds can return rapidly to their host galaxies as galactic fountains, return to their hosts or other galaxies only after a long sojourn, or escape to the IGM forever. In cosmological simulations wind gas makes up roughly a third of the gas in typical halos. The metals ejected by winds enhance cooling so that recycled wind gas provides a majority of the fuel for low-redshift star formation.
But winds are difficult to model. The processes that drive winds -- stellar radiation, AGN feedback, etc. -- are still debated. Winds are probably multi-phase, diffuse, fast-moving and out of equilibrium. The circum-galactic gas structures they may encounter as they leave the galaxy are not clear.
With these motivations (and others), we set out to construct an analytic model that could describe the propagation of galactic winds in cosmological simulations. I coded up the new model in Gadget, and it is currently undergoing testing.
Very preliminary results show that a more careful treatment of wind propagation can have a big impact on galaxy properties, such as the mass-metallicity relation.
The details of wind propagation can strongly affect the galaxy mass-metallicity and SFR-M* relations.
Star Formation Quenching
Galaxies can be divided into two broad categories: (a) those that are blue; contain cold gas, dust, and young stars; are disk-shaped; and live in relatively underdense regions of the cosmic web, and (b) those that are red; contain little gas and host mainly old stars; are elliptical in shape; and live in overdense regions such as galaxy clusters. The existence of the “red and dead” galaxies requires some process(es) to shut down, or “quench” star formation. This process, whatever it is, also limits the growth of massive galaxies so that they convert less than 20% of their cosmically available baryons into stars, and causes the exponential drop-off in the high-mass galaxy stellar mass function.
Some good candidate processes have been proposed over the last decade or so. Hot gas quenching relies on the fact that, in cosmological models, dark matter halos above about 1012 solar masses are pervaded by hot gas coronae. This hot gas mostly has very long cooling times, except perhaps at the center of the halo. If some heating engine such as a radio AGN can maintain a hot central region, then no gas will cool, and star formation will be quenched due to a lack of fuel. Another mechanism relies on galaxy mergers, which are thought to drive nuclear starbursts and powerful quasars. If the energetic output can blow the cold gas out of the galaxy, star formation will be quenched.
***
Hot gas quenching is both necessary and sufficient to produce a realistic red sequence; quenching by merger-driven outflows is neither.
I implemented these quenching mechanisms in cosmological hydrodynamic simulations. In hot gas quenching, the code identifies hot halos on-the-fly, and deposits thermal energy continually around the galaxies in such halos. In merger quenching, the code identifies galaxy mergers on-the-fly, and ejects the gas from merger remnants.
Hot gas quenching produces a realistic red sequence that matches the observed luminosity function, while merger quenching produces far too few quenched galaxies.
Merger quenching does not produce much of a red sequence. This is not because mergers are insufficiently frequent. Rather, it’s because galaxies in a cosmological context are continuously accreting gas from the intergalactic medium. Thus, even if you get rid of all of a galaxy’s gas at one time, it will resume forming stars within a Gyr or two just because of new gas accretion.
Hot gas quenching, on the other hand, produces a red sequence with the right numbers of galaxies (e.g. the luminosity function matches observations). The hot halos effectively starve both central and satellite galaxies of fuel for star-formation. As long as the halo remains hot, these galaxies will remain quenched.
***
Red sequence build-up -- first massive centrals, then low-mass satellites
Using our model of hot gas quenching in cosmological simulations, we can follow the build-up of the red sequence over cosmic time. Hot gas coronae form in halos above 1012 solar masses (independent of redshift), corresponding roughly to galaxies with stellar masses of 1010.5 solar masses. Once a galaxy achieves roughly this mass, starvation begins, and it inevitably exhausts its fuel for star formation. Stellar mass growth of galaxies slows after reaching the threshold, and this mass forms the “knee” in the stellar mass function. This mass threshold is reached first by the most massive galaxies at z~4. These galaxies continue to grow in both mass and size via mergers and accretion of their satellite galaxies, but the mass growth is small compared to the growth of their host halos.
In our hot gas quenching model, massive galaxies are the first to be quenched at redshifts above 2, after which they grow through mergers. At lower redshifts, lower mass galaxies fill in the red sequence.
At later times, additional massive galaxies reach the threshold mass, and they also quench, adding to the red sequence. Concurrently, many low-mass galaxies become satellites of groups and clusters (really, any hot halo), and these satellites are also quenched due to the presence of hot gas. These satellites “fill in” the low-mass end of the red sequence.
***
Hot gas quenching drives both “mass quenching” and “environment quenching”.
"Mass quenching" and "environment quenching," described in observations by Peng et al. 2010, emerge naturally from hot gas quenching (top left). Galaxies with large masses or living in dense environments are mostly quenched. We also show relationships with halo mass.
Hot gas quenching naturally leads to trends with mass and environment. These trends ultimately stem from the sharp transition mass -- halos with masses above ~1012 Msun form hot coronae, while lower mass halos do not.
Galaxies with stellar masses above about 1010.5 Msun almost always live in a hot halo, and the quenched fraction of galaxies increases rapidly above this mass. This mass quenching is roughly independent of the large scale environment of the cosmic web. Even though we sometimes think of massive halos as living in dense environments (e.g. groups or clusters), 1012 Msun halos actually live in a wide range of environments.
On the other hand, galaxies in very dense cosmic environments almost always live in groups or clusters. The vast majority of these galaxies are satellites (e.g. a cluster of 100 galaxies has 99 satellites!). These group and cluster halos are permeated by hot gas that quenches the resident galaxies, so the quenched fraction increases sharply with environment. This is called environment quenching.
Along with these trends, we found some other interesting effects. Roughly a third of quenched satellite galaxies were actually "pre-processed": they were quenched in smaller halos before joining their z=0 group or cluster (shown at right). Almost all massive quenched satellites were pre-processed, and most of these were quenched as the central galaxies in their own hot halos.
A nontrivial number of small galaxies appear as though they are quenched central galaxies, but in fact they are ejected former satellites from massive groups or clusters. A small number of galaxies are “neighborhood quenched”: they live in hot gas that extends several virial radii beyond the edge of a cluster.
AGNs in Galaxy Evolution
Active galactic nuclei (AGNs) could play an important role in galaxy evolution. Most massive galaxies in the local universe host a supermassive black hole, and those black holes’ masses correlate with host galaxy properties such as the bulge mass. This means that AGNs (which result from the growth of black holes) must have occurred at some time in most every galaxy’s past, and raises the possibility that the AGNs had important effects on the star formation in the host galaxy.
Many open questions remain: what triggers an AGN episode? Which kinds of galaxies host AGNs? How does an AGN affect the gas in the host galaxy? Could an AGN drive powerful gas outflows? How do AGNs affect their inter-galactic (and intra-cluster) environments?
***
Powerful AGNs in isolated disk galaxies (at high-redshift)
While theorists have long suspected that galaxy mergers trigger AGNs, observations suggest that most AGNs, at least at moderate redshifts (0.3-2), are hosted by normal disk galaxies. This led us to explore the problem with idealized, high-resolution simulations of isolated, star-forming disk galaxies.
We found that disk galaxies with high gas fractions (like those observed at redshift 2) can fuel powerful AGNs. In such galaxies, gravitational instability induces the formation of massive, dense clouds of gas. These clouds (or really, any density perturbations) will interact gravitationally and “scatter” with each other, exchanging angular momentum. The scattering enables some clouds to migrate to the galactic center until they collide with the black hole (see video below). As shown at right (top panel), the structure of the ISM produces rapid variations in the black hole accretion rate. When a particularly dense cloud collides with the black hole, it then undergoes a strong but brief (~10 Myr) episode of accretion. Owing to the high gas densities in the cloud, the accretion can reach the Eddington limit, and luminosities in the quasar regime (>1046 erg/s).
Based on our simulations, this mode of black hole accretion could power most of the AGNs detected in deep galaxy surveys at redshift 1 and 2. It could therefore drive a significant fraction of the black hole growth in the universe.
(full-resolution version here)
***
Powerful AGN-driven outflows in isolated disks, without quenching
Once we discovered that isolated gas-rich galaxies could fuel an AGN, it was natural to ask what effect the AGN could have on the gas and star formation in the galaxy.
We found that our simulated AGNs drive powerful gas outflows. The winds, driven by a strong shock, emerge with peak mass outflow rates around the star formation rate (tens of Msun/yr). In our simulations most of the outflowing gas is very hot (>106 K), and has velocities >1000 km/s; some cooler, slower-moving gas is entrained in the flow.
(full-resolution version here)
Despite the strong outflows, the AGN has only a minor effect on the wider host galaxy. Dense gas in the disk blocks the propagation of the outflowing shock, so only the central kpc is affected. Dense clouds above or below the black hole in the plane of the disk can even block the outflow from emerging from one side -- leading to asymmetric, unipolar outflows. The quantity of dense gas in the galaxy is essentially unchanged by the AGN, so star formation continues as normal.
AGN feedback may affect the gas distribution in a galactic disk, but it does not have a major impact on the quantity of dense gas on scales > 1 kpc.
***
AGN Host Galaxy Morphologies
My first approach to the issue of AGNs in galaxy evolution was observational. I analyzed Hubble Space Telescope data from the COSMOS survey to measure quantitative morphologies of Active Galactic Nuclei (AGN). We used about 300 X-ray and radio-selected AGN with optical spectroscopic redshifts between 0.3 and 1.0. Among our findings for X-ray AGN: their morphologies form a broad distribution between bulge-dominated early types and disk-dominated late types, and evidence for mergers or interactions are weak. The X-ray AGN host galaxies were not bulge-dominated, and appeared relatively isolated, leading to the suggestion that normal star-forming disk galaxies could play host to many or even most AGNs. This suggestion was largely confirmed by subsequent work, though this remains a complex issue.
Older Research
Back in the day I was lucky enough to try out lots of different research projects. I learned that there are many interesting problems out there, and many cool techniques to try to solve them. In the end I was most inspired by black holes and galaxy evolution, but here are some of the other things I’ve worked on:
Spectral Diagnostics of Age for (sort of) Young Stars
As an undergrad, I worked with Lynne Hillenbrand and Russel White analyzing spectra of stars with suspected ages between 3 million and 3 billion years. This work helped constrain ages for the FEPS Spitzer Legacy program, which studied circumstellar debris disks thought to precede and/or contain newly formed planets. This work resulted in a refereed journal article (astro-ph version).
Neutron Star Hot Spots
For an REU summer project in 2004, I worked with Dong Lai to model light curves and spectra of neutron stars with hot spots. Work of this kind can help constrain the physical sizes of neutron stars, which with mass estimates can constrain the equation of state of extremely dense matter.
Mid-IR Spectroscopy of z~2 Galaxies
I’ve worked with Alex Pope on analyzing spectra from Spitzer Space Telescope’s Infra Red Spectrograph. Our goal was to create a database of deep mid-IR spectra of high-redshift galaxies in the deep multi-wavelength surveys (mainly GOODS-N and GOODS-S).
Solar Energy Generation
During summer of 2008, I worked with Roger Angel as he developed new technologies for efficient concentrated solar energy generation. I primarily developed new diagnostic tools to characterize performance of solar cells and their cooling system. I conceived an automated way to measure solar cell I-V curves about 100 times faster and 100 times cheaper than alternativ