Fast and energetic AGN-driven outflows in simulated dwarf galaxies

Dwarf galaxies, the least massive galaxies in the Universe, are the smallest building blocks of the cosmic structure formation process. They are interesting cosmological probes, as there are a number of apparent discrepancies when comparing observations of dwarf galaxies with the predictions of structure formation models. For example, the large disparity in the number of observed dwarfs with respect to the predicted number of dark matter haloes that may be hosting these systems (the missing satellite problem) or the lack of observed massive dwarf galaxies (the too-big-to-fail problem). Some have suggested that these issues could be resolved by changing our models of dark matter, which provides the skeleton for cosmic structure formation. Others argue that the focus should be on improving our models of the luminous, ‘baryonic’ matter that we can observe directly.

One of the most important baryonic processes, reionization, stems from the radiation produced by the first galaxies in the Universe. This radiation shuts down star formation in the low-mass dwarf galaxies rendering them undetectable. For the more massive dwarf galaxies it has been suggested that violent stellar explosions – supernovae (SNe) – could shut down star formation, however this remains controversial. Most simulations indicate that another ingredient would be needed for the theoretical models to explain the sparsity of massive dwarf galaxies in our Universe.

Figure 1: Edge-on projections of all the simulation runs at t = 300 Myr. The first row shows surface densities, the second row shows temperatures, the third row shows vertical velocities, and the fourth row shows metallicities. The runs with low star formation efficiency are depicted on the left hand side, and the high star formation efficiency runs on the right hand side. With only SN feedback, the outflows are slow, warm, and high-density. With additional AGN feedback, the outflows have an additional hot, low-density component moving at high velocities.

Recently, it has been proposed that active galactic nuclei (AGN) – actively-accreting massive black holes – might be this missing ingredient. Previously, theorists did not include AGN in their simulations of dwarf galaxies as these were only observed in more massive galaxies. However, the systematic analysis of optical surveys has revealed a population of dwarf galaxies hosting AGN, which have been confirmed by X-ray follow-up observations. All of these AGN are extremely luminous relative to their black hole mass and are likely only the tip of the iceberg. There has also been some first evidence of direct AGN feedback from the MaNGA survey, which identified six dwarf galaxies that appear to have an AGN that is preventing on-going star formation. These observational results are very exciting as they suggest that AGN may be able to solve the too-big-to-fail problem.

It is therefore timely to study the physical properties of dwarf galaxies, in particular whether the presence of an AGN can affect their evolution. Using the moving mesh code arepo, we have investigated different models of AGN activity, ranging from simple energy-driven spherical winds to collimated, mass-loaded, bipolar outflows in high resolution simulations of isolated dwarf galaxies hosting an active black hole. Our simulations also include a novel implementation of star formation and mechanical SN feedback.

Figure 2: Comparison between the line-of-sight velocity maps for stars and ionized gas for a galaxy without observed AGN activity (left panel) and a galaxy with AGN activity (right panel). With AGN activity, the ionized gas is visibly offset from the stellar component.

Figure 1 shows the visuals of our simulation suite. Here, edge-on projections of surface density, temperature, vertical velocity and metallicity at t = 300 Myr are plotted. The control runs with neither SN nor AGN feedback (’NoFeedback’ runs) do not produce any outflows. With added SN feedback, we obtain relatively cold and slow-moving outflows. When we also include AGN activity in our simulations, the outflows are much faster as well as much more energetic, reaching temperatures of up to 10 9 K.

These results provide tantalizing evidence that the observed correlation between large kinematic offsets and AGN activity may be due to high-velocity AGN-driven outflows (see Figure 2 for an example of the different line-of-sight velocity maps observed with MaNGA). In our simulations, the same correlation arises because only the AGN-driven gas outflows are fast enough to alter the velocity patterns set by the rotational motion of the galaxy. The relatively slow-moving supernova-driven outflows on the other hand do not offset the rotational motion of the gas so that it stays aligned with the stars (see Figure 3, left panel). This then results in a significant difference between the kinematic position angles (PAs) of stars and gas for dwarf galaxies hosting AGN (see Figure 3, right panel).

Figure 3: Left panel: Comparison between the 2D velocity maps of the stars and gas in the simulations at t = 150 Myr. The colour coding of the pixels indicates the surface density. The spatial coverage of the primary MaNGA sample of 1.5R eff is indicated by a white circle. With added AGN feedback, the central circular motion is significantly altered by the outflows so that the gas would be seen as kinematically offset. With SN feedback alone, the circular motion remains dominant and we do not observe this feature. Right panel: Comparison between the line-of-sight velocity maps for stars and ionized gas at t = 150 Myr. Only gas and stars within 1.5R eff are included (in analogy to MaNGA), and the spatial resolution is matched to the MaNGA resolution at the mean redshift of the primary sample. For the AGN runs, the ionized gas is visibly offset from the stellar component, just like in the observed MaNGA galaxies.

We also investigated the effects of these AGN-driven outflows on star formation and found that AGN activity has a small but systematic effect on the central star formation rates (SFRs) for all set-ups explored, while substantial effects on the global SFR are only obtained with strong SNe and a sustained high-luminosity AGN with an isotropic wind. Our findings from the isolated galaxy set-up therefore indicate that AGN are unlikely to directly affect global dwarf SFRs. However, in realistic cosmological environments inflows are known to be important especially for
high-redshift dwarfs. It is hence possible that AGN-boosted outflows may prevent some of this cosmic ‘pristine’ gas reaching the dwarfs in the first place, providing a mechanism for indirect star formation regulation. This possibility will be addressed in a follow-up study using cosmological zoom-in simulations.