The Impact of Nuclear Star Formation on Gas Inflow to AGN

Many studies of active galactic nuclei (AGN), particularly those concerned with understanding the co-evolution of black holes and their host galaxies through cosmic time, rely on observations of quasars. The reason is simply that these are the most luminous of such objects and can be studied at all redshifts. While they form an important component of the AGN population, the luminosity function ([Hao et al. (2005), Hao et al. 2005]) indicates that the number density of lower luminosity AGN is orders of magnitude greater. This is particularly true in the local universe because of the effects of downsizing, which means that at low redshift black hole growth occurs primarily in AGN with black hole masses $<10^{8}$ M_⊙ ([Best et al. (2005), Best et al. 2005]) and Seyfert-like luminosities ([Hasinger et al.(2005), Hasinger et al. 2005]). As a result, with only a few exceptions such as Mkn 231 ([Davies et al. (2004b), Davies et al. 2004]), detailed studies of nearby AGN have to focus on the lower luminosity AGN such as Seyferts.

This has important implications. Although most quasar activity may be fuelled via mergers of gas rich galaxies, Seyfert nuclei typically reside in (preferentially early type) spiral galaxies ([Ho (2008), Ho 2008]). As a result, Seyfert activity is likely to be fuelled via secular processes associated with disk evolution such as those outlined by [Shlosman et al. (1990)] and [Wada (2004)]. One obvious part of the overall scheme is the role played by bars, which are known to drive gas inwards from the large-scale disk. Yet despite the many attempts to find a correlation between the presence of a bar and an AGN, the statistical outcome is, at best, for only a marginal link between these phenomena. This, perhaps, should not be surprising. Bars drive large amounts of gas inwards (of order 0.1 M_⊙ yr^-1, [Regan & Teuben (2004), Regan & Teuben 2004]), working on spatial scales of kpcs and timescales of Gyrs. In contrast, a Seyfert nucleus requires typically $\lesssim 0.01$ M_⊙ yr^-1 of gas ([Jogee (2006), Jogee et al. 2006]), and the relevant temporal and spatial scales are Myr and pc. The two regimes are orders of magnitude different in every respect. Thus, while it is well understood how bars can create a gas reservoir and a starburst in the central kiloparsec (and there is plenty of observational evidence to support the link between bars and both of these phenomena: [Sakamoto et al. (1999), Sakomoto et al. 1999], [Laurikainen et al. (2004), Laurikainen et al. 2004], and [Jogee et al. (2005), Jogee et al. 2005] are a few exmaples among many), this is only a single possible step in the process that leads to AGN fuelling.

It is now understood that the properties of the central kpc – the circumnuclear region – can play an important role in driving gas further inwards. But although observations have yielded many tantalising results, reaching definitive conclusions is difficult. For example, [Martini et al. (2003)] looked at the detailed circumnuclear dust morphologies of matched samples of active and inactive galaxies. One of their results was that AGN are only found in galaxies that exhibit dusty spiral patterns. But they also found that spiral patterns of any sort were associated equally often with active and inactive galaxies. This, again, is likely a result of the differing spatial and temporal scales of the phenomena. The aspect of time variability is one of the issues addressed by NUGA ([García-Burillo et al. (2007), García-Burillo et al. 2007]), a study of the distribution and kinematics of the molecular gas in nearby low luminosity AGN. One of their key results was that gravitational torques can not only drive gas in to form a circumnuclear ring, but also tend to drive gas out from the nuclear region towards the ring. This led [García-Burillo et al. (2005)] to propose a scenario in which further inward flow was possible episodically if two conditions were fulfilled: the bar should be weakened by dynamical feedback, and the viscosity should increase due to a sufficiently steep density gradient in the circumnuclear ring.

These examples show that there is still much we do not understand about the conditions under which gas can be transported inward from kpc scales to the nuclear region, and it is one of the topics addressed here: we describe two highly contrasting case studies where gas is being driven inwards to the central tens of parsecs. We also show evidence that, once there, the gas is likely to trigger a starburst. And we discuss the impact of the starburst on the fate of the gas, and a possible link to the sub-parsec scales.

In this section, we describe contrasting cases where gas can be observed as it flows in towards an AGN. Although we present only two examples here, analyses of non-circular motions and gas inflow mechanisms on similar scales have been performed for other galaxies as well: NGC 7469 ([Davies et al. (2004a), Davies et al. 2004]), NGC 6951 ([Storchi-Bergmann et al. (2007), Storchi-Bergmann et al. 2007]), NGC 4051 ([Riffel et al. (2008), Riffel et al. 2008]), NGC 7582 ([Riffel et al. (2009), Riffel et al. 2009]), NGC 4151 ([Storchi-Bergmann et al. (2009), Storchi-Bergmann et al. 2009]).

In one case, NGC 1068 ([Mueller Sánchez et al. (2009), Mueller Sánchez et al. 2009]), the inflow rate is prodigious, due to chance events, and probably cannot be sustained. In the other, NGC 1097 ([Davies et al. (2009), Davies et al. 2009]), the inflow is much more moderate and appears to be regulated in a way that is sustainable for timescales of a Gyr. Intriguingly, in neither case do we see gas flowing all the way down to the AGN: the inflow is terminated at spatial scales of a few to tens of parsecs, where it is instead expected to fuel a starburst.

We have seen in the previous section how, at least for these 2 cases, gas flows into the central few to tens of parsecs. If it is stalled here, one would naturally expect it to trigger a starburst. Evidence for starbursts on these scales was presented by [Davies et al. (2007)], who used near infrared integral field spectroscopy combined with adaptive optics to achieve high spatial resolution. Using the spectral information, these authors were able to separate out the non-stellar continuum (hot dust emission) associated with the AGN, and show that in every case the stellar continuum itself was spatially resolved with a typical size scale $<$50 pc. Furthermore, they showed that this was a distinct stellar population, younger and dynamically cooler than the bulge: photometrically, from an excess in the stellar continuum above an $r^{1/4}$ law fitted to 1–2^′′ scales and extrapolated inwards; and kinematically, from a drop in the stellar dispersion on the same spatial scales. Constraints on stellar population synthesis models provided by several independent diagnostics indicated that the starburst ages were young, lying in the range 10–300 Myr. One important piece of evidence was the remarkably low equivalent width of the Br$\gamma$ line, which shows unambiguously that the star formation has ceased: i.e. although still young, these are post-starbursts. Intriguingly, starburst models imply that during the short (perhaps only 10 Myr) actively star forming phase, the luminosity of the nuclear starburst would have been an order of magnitude greater than it is now – with a star formation rate per unit area comparable to that in ULIRGs. This high intensity appears to make sense in terms of the high gas mass surface density which, from the Kennicutt-Schmidt law, is expected to lead to high star formation rates.

More pertinent to the issue of gas inflow, the authors showed that there appears to be a relation between the characteristic age of the most recent episode of star formation and the luminosity of, or accretion rate onto, the AGN (Fig. 2). This figure suggests that there is a delay of 50–100 Myr between the short duration starburst and the onset of accretion onto the AGN. This timescale is interesting because it corresponds to the age at which, for a short burst, one expects the type ii supernovae phase to be finishing. This topic is discussed in more detail in Section 4, which addresses the impact of stellar winds and supernovae on gas inflow and outflow, and how these might provide a link between phenomena observed with adaptive optics on $\sim$10 pc scales and those seen with infrared interferometric techniques on 0.1–1 pc scales.

The data presented in [Davies et al. (2007)] and [Hicks et al. (2009)] show that nuclear star formation is occuring on the same spatial scales, up to a few tens of parsecs, as the gas distribution; and that on these scales the two components have similar kinematics. This means the stars and gas must be mixed. It is inevitable that the star formation will have some impact on the gas, and this was argued from a phenomenological approach by [Davies et al. (2007)]. Hydrodynamical simulations of the impact of stellar evolution, such as those presented in [Schartmann et al. (2009)] give a clearer picture. To understand this in the context of Seyfert nuclei, these simulations have been scaled to match a typical starburst there, as described by Schartmann et al. (these proceedings): M${}_{\rm BH}\sim 10^{7}$ M_⊙, M${}_{\rm stars}\sim 2\times 10^{8}$ M_⊙ out to 50 pc, a dispersion of $\sigma*\sim 100$ km s^-1 as well as rotation, and a mass loss rate $\sim 0.1$ M_⊙ yr^-1 corresponding to an age of 50 Myr but which of course would vary with time. The simulations show that there are two, fairly well defined, regimes separated by a stellar outflow speed of about 250 km s^-1. Fast outflows (which could represent winds from OB stars, as well as type ii and type i supernovae) lead to a starburst wind which efficiently removes the diffuse ISM. Slower outflows (which represent, for example, planetary nebulae or winds from AGB stars) lead to a large number of clumps that can accrete efficiently to smaller scales, giving rise to a dense and turbulent disk with a radius of 0.5–1 pc. Taken together, these imply that there is a phase in a starburst when accretion to smaller scales is possible and efficient. For a short starburst, this phase would begin after about 50 Myr, once there are no more OB stars or type ii supernovae; and it would end after about 300 Myr once type i supernovae start to appear.

The size scale of the central disk is similar to that of the compact 8–12 $\mu$m structure in NGC 1086 (the galaxy to which the simulation is scaled) by [Raban et al. (2009)] using MIDI on the VLTI; and within a factor of a few of the compact structure seen in Circinus by [Tristram et al. (2007)] using the same instrument. As already suggested by [Davies et al. (2008)], these simulations imply that stellar outflows are a link between the gas structure seen on scales of 10–50 pc using adaptive optics, and those seen on 0.1–1 pc scales using infrared interferometry. This scenario has a number of similarities to the simulations of [Wada et al. (2009)]. Although the initial conditions and the distribution of energy input are rather different, the result is a vertically thick (up to 10 pc from the disk plane) distribution at large radii, and in the inner few parsec a compact turbulent disk.

If one considers either of these structures to be associated with the molecular obscuring torus, then one must also conclude that they both are. As illustrated in Fig. 3 and stated by [Hicks et al. (2009)], this implies that the torus comprises multiple components each of which fulfils different roles; that it must be a star forming torus; and, because the starbursts are episodic, that it must be a dynamical entity. What we see on 10–50 pc scales is the global structure, a diffuse and clumpy envelope where there are short repeated bursts of star formation. The slow stellar winds accrete down to smaller $\sim$1 pc scales where they form the more compact structure, a dense and turbulent disk. The evolution of this disk – in terms of further star formation and gas inflow – is addressed by Schartmann et al. (these proceedings).

One final point worth noting is the similarity between these simulations and those performed for the Galactic Center, on scales about 100 times smaller, by [Cuadra et al. (2006)] and [Cuadra et al. (2008)]. Both simulations were performed within a 0.5 pc volume in which the stars, their orbits, and their outflows, matched as precisely as possible the known configuration at that time. In the earlier version, where there were both slow $\sim$200 km s^-1 and fast $\sim$700 km s^-1 stellar winds, the result was that much of the ISM was blown out but that clumps of cooler dense gas accreted to form a compact disk that extended out to $\sim$0.04 pc. In the later version, the updated stellar wind speeds were all fast, and nearly the entire ISM was blown out. It is reassuring that qualitatively, these simulations yield the same result as those of [Schartmann et al. (2009)] even though the scales are drastically different, because in both cases it is the same processes (i.e. stellar outflows) that govern the evolution of the region.

We have presented a brief review of the current status of our work on the impact that star formation has on gas inflow to Seyfert nuclei. We are not yet able to follow gas all the way down to the AGN itself, and instead we have focussed on spatial scales from a few hundred parsecs down to the central parsec. Our main conclusions are:

• •

  We have seen 2 contrasting methods of gas inflow that feed a nuclear starburst. While showing how gas might reach the central few to tens of parsecs, this also raises questions about the sustainability of inflows, how far in they reach, and whether it is possible to define what a typical inflow mechanism or rate is.

• •

  There are short intense starbursts in the central tens of parsecs around Seyfert nuclei. In general, starbursts will be an inevitable consequence of gas inflow.

• •

  Stellar outflows play a key role in the evolution of the ISM on these scales. Specifically, fast outflows tend to blow out the diffuse ISM, while slow outflows can lead to efficient accretion of gas onto smaller scales. This may be a link between what is seen on 10–50 pc scales with adaptive optics, and what is observed on 0.1–1 pc scales with infrared interferometry.

• •

  The molecular obscuring torus most likely consists of multiple structures on different scales that fulfil different roles. Since the state of the torus is strongly affected by the episodic star formation that occurs within it, we need to think of it as a dynamic entity.