Paul Martini
Assistant Professor
Department of Astronomy
The Ohio State University

The figure to the left illustrates the basic approach we have used to identify AGN in groups and clusters: We obtain X-ray (black contours) and visible-wavelength images (grayscale) of a group or cluster and then obtain spectra of all galaxies with sufficiently bright X-ray counterparts that they would be AGN if the galaxies are cluster members. This figure is from Martini et al. (2006).

With a sample of 32 clusters of galaxies from the local universe out to z~1.3 we have found that there is a factor of eight increase in the fraction of cluster galaxies that host AGN. In the figure to the left (from Martini et al. 2009) the filled circles represent the average AGN fraction in 17 z<0.4 clusters (2 AGN in ~1500 cluster galaxies) and 15 z>4 clusters (18 AGN in ~2400 cluster galaxies). (The open circles represent various estimates of the field AGN fraction.) A simple power-law fit to this evolution indicates that the AGN fraction increases as (1+z)^5.3. It is very intriguing that the evolution of the fraction of star-forming cluster galaxies is essentially identical (Haines et al. 2009 find (1+z)^5.7). My Ph.D. student David Atlee and I are actively working on the connection between star formation and black hole accretion in cluster galaxies.

The distribution of AGN in clusters as a function of radius provides important information about their origin. If the AGN are primarily at large radii, then that suggests most are hosted by galaxies that have been recently accreted from the field, while a more centrally concentrated distribution would suggest that the AGN are being triggered in situ. Surprisingly, we have found evidence for both distributions. At high-redshift we find the distribution (in AGN number) as a function of radius is flat. As the galaxy population decreases dramatically with clustercentric distance, this indicates that the AGN fraction is increasing toward the cluster outskirts. However, at low redshift we find the opposite and here the central concentration suggests these (lower-luminosity) AGN are mostly hosted by the massive galaxies that dominate the cluster core and have been in the cluster for many Gyr. This figure is adapted from Martini et al. (2007) and (2009).

Another long-standing question in AGN research is how they are fueled. Observations of the AGN population in the field, groups, and clusters can help answer this question by identifying the preferred environment for AGN to form. In the standard merger-driven scenario (e.g. Hopkins et al. 2008), low-mass groups have the ideal combination of high space density, low relative velocity, and gas-rich galaxies. We have searched for AGN as a function of local overdensity (parameterized by the velocity dispersion) and indeed find some evidence for a higher AGN fraction in groups relative to both clusters and the field (the latter measurement is from Lehmer et al. 2006). We are currently working to secure a large sample of nearby groups and clusters to obtain a more statistically significant sample. The figure at left is specifically the AGN fraction in early-type galaxies and was obtained from Arnold et al. (2009).

While I was a Carnegie Fellow at the Carnegie Observatories in Pasadena, I helped Eric Persson build the PANIC NIR camera for Magellan. PANIC (Persson's Auxilliary Nasmyth Near-Infrared Camera) is a relatively straightforward NIR imaging system that has a 2x2 arcminute field of view and 0.125 arcsecond pixels (on a 1024x1024 array). PANIC was commissioned in April 2003 and has obtained some of the best image quality at Magellan, often producing images better than 0.3'' FWHM. My contributions to PANIC included thermal modeling, mechanical design (including the optics mounts), system integration, construction of the data acquisition system, data reduction software, and documentation. More information is available in the PANIC Online Documentation.

During my two years as a Clay Fellow at the Harvard-Smithsonian Center for Astrophysics I was the Instrument Scientist for MMIRS (MMT/Magellan Infrared Spectrograph), working with the PI Brian McLeod up through the Critical Design Review in June 2005. MMIRS is an imager and multi-object spectrograph and has a Hawaii-2 detector with 0.2'' pixels. The imaging field of view is 6.8' x 6.8', while the multi-slit spectroscopy can be obtained over approximately a 4' x 6.8' field (although only with complete spectral coverage over 2'). MMIRS was commissioned at the 6.5m MMT in May 2009. The MMIRS Observing Information contains further details.

Over the last few years I have worked to build the OSMOS instrument for the 2.4m Hilter telescope at the MDM Observatory. OSMOS is an imager and multi-object spectrograph with 0.3'' pixels, a circular 20' diameter field of view, and produces good images from about 370nm to 1 micron with an all-refractive design. OSMOS will have a variety of spectroscopic resolutions from a few hundred (via a triple prism) to several thousand (via VPH grisms). Commissioning is scheduled for April 2010, although at first only imaging and long-slit spectroscopy will be available. Once the multi-object capability is complete, it should be possible to observe up to 50 targets per mask.

Clustering provides a very important tool to estimate the lifetime of QSOs with only one very simple assumption about how QSOs inhabit dark matter halos. If we assume that QSO luminosity is a monotonic function of the mass of the host dark matter halo, then the measurements of the QSO space density and clustering provide an estimate of the QSO lifetime because more massive dark matter halos are more strongly clustered. For example, if the QSO lifetime is long, then only the most massive halos are required to match the observed space density and the clustering will be large, while if the QSO lifetime is short then many more (and therefore less massive and less clustered) halos are required to match the observed space density and the QSO clustering will be lower. The figure at left from Martini and Weinberg (2001) illustrates the relationship between clustering and lifetime for various space densities and models for QSO luminosity evolution.

An alternative way to estimate the lifetime of QSOs is to assume they are the product of major mergers between gas-rich galaxies and measure the lifetime of the accretion onto the supermassive black hole in a numerical simulation. Here one requires a simple prescription for accretion onto the black hole because the simulation does not have sufficient resolution to probe the immediate environment of the AGN. The figure at left shows the results of a calculation that I performed with Phil Hopkins, Lars Hernquist, and others using the GADGET code. We examined a large number of sightlines toward the central AGN and calculated the fraction of these that were not obscured as a function of time to determine the lifetime of the QSO that is produced in this merger and found that the lifetime is on order 10 Myr.

Since my earlier work on QSO clustering with David Weinberg, observations of large numbers of QSOs from the SDSS have finally made it possible to measure the clustering of QSOs at high redshifts (z>3). The results of this measurement were incredibly surprising: while low-redshift QSOs are clustered in a similar manner to galaxies, measurements by Shen et al. (2007) found that the highest redshift QSOs have enormous correlation lengths, so large that the best match to my earlier theoretical models require essentially a perfect correlation between QSO luminosity and dark matter halo mass as well as QSO lifetimes comparable to the age of the universe (at the observed redshift). With Martin White and Joane Cohn, I explored how much scatter between QSO luminosity and halo mass could be allowed and yet still be consistent with the observations. The figure at left demonstrates that the observed bias (clustering strength) of the z>4 QSOs (dashed lines, upper gray band) is so large that only a small amount of scatter is allowed. Furthermore, the allowed scatter is comparable to the estimated scatter between black hole mass and bulge velocity dispersion in the local universe, even though this relation is unlikely to constitute the entire scatter between QSO luminosity and halo mass.

A key question in secular evolution is how gas is transported to the circumnuclear region of galaxies, typically defined to be the central 100pc to kpc. One of the most useful pieces of information on the origin of the gas is its morphology, and to better quantify the morphology I developed a classification system for the dust lanes found in HST images of the centers of nearly all spiral galaxies.Examples of these morphological classes are shown in the figure to the right and largely reflect the many types of spiral structure seen in the centers of galaxies. (Note that these dust spirals are generally not associated with the stellar spiral arms seen on larger scales.)

With Mike Regan, John Mulchaey, and Rick Pogge, I investigated the incidence of different forms of nuclear dust spiral structure for galaxies with and without large-scale bars. The most striking correlation to come out of this is that so-called 'grand design' nuclear dust spirals are exclusively found in galaxies with large-scale bars. Furthermore, in essentially all cases these two-arm dust spirals extend from the unresolved nucleus to the dust lanes that trace the leading edges of the large-scale bar. These dust lanes are highly suggestive of gas transport into the nuclear region.

While many investigators have relied on the simple and convenient classification of galaxies as either barred or unbarred, in practice there is a continuous distribution of bar strengths (bar strength is a measure of the magnitude of the gravitational torque that the bar can exert on the underlying disk material). With graduate student Molly Peeples, we looked at circumnuclear dust structure as a function of bar strength and found that the incidence of grand design dust lanes increases in stronger bars. The figure to the left is from our 2006 paper and shows the circumnuclear regions of twelve of the most strongly barred galaxies we studied.

I have more recently become interested in secular evolution and star formation in the very latest-type disk galaxies (that have no bulge at all). These disks are the best candidates to study secular processes because they do not have any evidence at all of a 'hierarchical' bulge component. To study the gas distribution, star formation, and morphologies of the latest-type disks, I have collected HST, VLA HI, IRAM CO, IRAC, and Halpha images of 20 nearby galaxies. The efficiency of star formation and secular evolution in this sample constitutes the dissertation project of my Ph.D. student Linda Watson and is also in collaboration with Eva Schinnerer, Torsten Boeker, and Ute Lisenfeld.

Many mechanisms could dissipate angular momentum and fuel AGN, although they vary in their effectiveness. For example, major mergers between gas-rich galaxies are thought to be the only mechanism that could fuel the very high mass accretion rates that power QSOs, while very low accretion rates could be produced by processes as benign as stellar mass loss. The figure at left lists mechanisms that have been proposed to fuel AGN and are approximately ordered by the relative amount of accretion they could produce. I created this figure for my 2004 review article on AGN fueling, which I entitled "Why does Low-Luminosity AGN Fueling Remain an Unsolved Problem?"

The main conclusion from my work on fueling low-luminosity AGN is that all of the mechanisms in the previous figure must be important at some level, and in fact the large number of these mechanisms present in all galaxies (not just those that presently host AGN) strongly suggests that low-luminosity AGN are an episodic phenomenon common to all galaxies. This explains why previous studies, including my own, have not found a substantial difference in the presence of, e.g. bars or shocks traced nuclear spiral structure, in AGN relative to well-match control samples of inactive galaxies: more than just bars or shocks are important. Furthermore, the dynamical timescale of these phenomena (1-10 Myr) must be on order the AGN timescale. For example, an inactive galaxy may have substantial mass inflow at large radii due to a large-scale deviation from axisymmetry, yet by the time the that material reaches the nucleus, the nonaxisymmetric component of the potential may have been erased. This is qualitatively illustrated with the figure to the left, which is also from my 2004 review article on AGN fueling.

The one difference between AGN hosts and inactive galaxies that I did find in my large HST program is that while only a small fraction of galaxies have no evidence for circumnuclear dust, these 'dustfree' galaxies are exclusively inactive. This implies that the presence of circumnuclear dust is a requirement, although not a guarantee, of AGN fueling. The figure at left is from my 2003 papers with Mike Regan, John Mulchaey, and Rick Pogge.

My previous work on AGN fueling focused on spiral galaxies because they are known to have dust, yet early-type galaxies are interesting too because they also host AGN and are simpler dynamical systems. With Ramiro Simões Lopes and Thaisa Storchi-Bergmann, I investigated the circumnuclear dust structure of early-type galaxies with HST WFPC2 observations and the structure map technique I developed with Rick Pogge. The figure on the left shows structure maps of four early-type galaxies (from over 60 in our sample). The left column displays two that host low-luminosity AGN, while the right two display inactive galaxies. Note that both AGN hosts have substantial circumnuclear dust, while both inactive galaxies do not. The main result of this work was that all of the AGN hosts have circumnuclear dust, while circumnuclear dust is present in only 25 percent of the inactive early-type galaxies. This suggests that dust is required to produce an AGN in an early-type galaxy, and almost always does so. Our main area of investigation now is to understand why some early-type galaxies have dust and some do not.