Dwarf Galaxies at the cutting EDGE of galaxy formation

This guest post was written by Katy Proctor for an assignment in the “Advanced Topics in Physics: Galaxies” class taught by Dr. Claudia Lagos. As part of the course, students were tasked with writing Astrobite-style summaries of recent and seminal extragalactic research papers.

Katy is a final year Masters student at the International Centre for Radio Astronomy Research (ICRAR) at the University of Western Australia. Katy is working under the supervision of Dr. Claudia Lagos and Dr. Aaron Robotham on constraining the free parameters of SHARK, a semi-analytic model of galaxy formation. When not thinking about galaxies, Katy enjoys getting outdoors, playing guitar, and reading.


Title: EDGE: the mass-metallicity relation as a critical test of galaxy formation physics

Authors: Oscar Agertz, Andrew Pontzen, Justin I. Read, Martin P. Rey, Matthew Orkney, Joakim Rosdahl, Romain Teyssier, Robbert Verbeke, Michael Kretschmer and Sarah Nickerson

First Author’s Institution: Lund Observatory

Status: Accepted in MNRAS [closed access], Open access on Arxiv.

Despite their small size, dwarf galaxies have proven to be extremely useful probes of galaxy formation. This can be attributed to the fact that dwarf galaxies are sensitive souls: the effect of feedback from either supernovae, re-ionisation, stellar winds or radiative feedback has been shown to be enough to completely extinguish their star formation.

Figure 1. An image of the dwarf galaxy NGC 5477 taken from the NASA/ESA Hubble Space Telescope.

This helps dwarf galaxies become prime candidates for constraining models of galaxy formation. Simulators have had success in reproducing a large number of the scaling relations obsreved in dwarf galaxies, often through different choices in physical models. However, there are some properties such as the stellar mass-halo mass relation and the stellar mass-metallicity relation that simulations still cannot convincingly reproduce – particularly for the faintest dwarfs. Another intriguing problem that eludes astronomers is coming up with a convincing explanation for the existence of star-forming ultra faint dwarf galaxies.

With these issues in mind, today’s authors introduce their new dwarf galaxy campaign: Engineering Dwarfs at Galaxy Formations Edge (EDGE). The goal of today’s paper is to investigate:

  1. What drives the regulation of star formation in the smallest dwarf galaxies; and
  2. Which dwarf galaxy observables are affected the most by changes to the physical models of galaxy formation.

Finding a home for the dwarf galaxy

The authors make use of RAMSES-RT, which is a radiation hydrodynamics code that allows them to investigate the physics of hydrodynamics, radiative transfer, and non-equilibrium radiative cooling/heating on the properties of dwarf galaxies. The authors choose to study one dwarf galaxy that forms in isolation – this allows them to probe down to high resolution limits while avoiding the additional complexities (both theoretical and computational) of simulating a host galaxy. Simulating at high resolutions is necessary if the authors want to investigate in detail physical processes that occur on the parsec scale, such as stellar winds.

To simulate their dwarf galaxy, the authors first had to find somewhere to put it. Given that the authors are interested in investigating a single dwarf galaxy that ends up isolated at the present day (redshift of z=0), the authors began by running a dark matter-only simulation and identifying suitable dark matter halos that were in cosmic voids at z=0. From this filtered selection of halos, the authors chose a halo of mass 109 M⊙ that was isolated from dark matter filaments. They then traced the history of all particles in the halo at z=0 back to z=99, and from there were able to perform a higher resolution zoom simulation of the halo that incorporated a description of baryonic physics.

The authors then probed the effects of different galaxy formation physics on the properties of the resulting isolated dwarf galaxy at two numerical resolutions. Their fiducial simulation probes down to baryonic masses of mbar = 161M⊙ and the high resolution simulation (referred to as Hires in the paper) probes down to mbar = 20M⊙. The physics they investigate includes (but is not limited to – see paper for comprehensive list):

  • The effect of including radiative transfer in their feedback model; and
  • Varying supernovae (SN) feedback strengths.

Changing the physics

The first change in physics the authors investigate is the effect of including radiative transfer (RT)  in their model. Figure 1 shows the effects on gas surface density, temperature and metallicity for models where feedback is dominated by supernovae (Hires) and where radiative transfer modelling is included (Hires+RT) at different redshift snapshots. At redshift z > 5 we are in the period prior to cosmic reionisation becoming the dominant ionisation source. We can see from Figure 1 that during this epoch the inclusion of RT has a big effect on how the galaxy self-regulates and evolves. The Hires model is dominated by large scale supernova driven outflow, even at early times. This expulsion of gas results in a hot and high metallicity region surrounding the galaxy (seen in rows 3 and 5 of Figure 1). The inclusion of RT has the effect of keeping the gas in a warm state, which also regulates the formation of stars. This suppression of star-formation is less explosive than the SN dominated model, which operates through highly energetic outflows, whereas large scale outflows are suppressed in the RT model until z=6.5. The differences between the two models becomes less apparent at z<5 since reionization becomes the dominant ionising source.

Figure 2. Gas surface density, temperature, metallicity in a 5×5 kpc2 region centred on the galaxy at various redshifts. Including radiative feedback changes the mode of galaxy formation.

The biggest takeaway here is that the inclusion of RT has a massive impact on the early behaviour and regulation mechanism of the dwarf galaxy. The authors also consider the resulting stellar mass at z=0 from both models, and find that the Hires+RT model results in a lower stellar mass than the Hires model, by a factor of ~10. While both lie within the observation scatter of the M* − M200 relation, the Hires+RT model is consistent with the faintest observed dwarfs at a halo mass of 109 M⊙. These results imply that radiative transfer contributes significantly to the suppression of star formation in low mass dwarf galaxies.

Constraining models

Ideally, the authors should have been able to rule out some of their physical models based on tensions with observed quantities. However, upon comparison to observed present-day dwarf galaxy structural properties such as V-band magnitude, half-mass radii, stellar velocity dispersions and dynamical mass- to-light ratios, the authors found that all of their physical models (with the exception of the no feedback model) can reproduce the observations. This includes their un-physically weak/strong feedback models, which has some big implications for modellers: it indicates that these scaling relations have no constraining power over the models of dwarf galaxy formation and that other properties of dwarf galaxies must be considered in order to rule out some of the theoretical formation scenarios. The authors suggest that the insensitivity in these scaling relations is likely due to these scaling relations being set by the host dark matter halo in galaxies with high mass-to-light ratios.

This leads the authors to consider a comparison with the mass-metallicity relation (MZR) at z=0. Figure 2 below shows one of the main findings of their paper: the MZR can be used to identify plausible models of dwarf galaxy formation, with the Hires and Hires+RT being the only models that can reproduce observations. The MZR seems to be highly sensitive to the physics involved in star formation and outflows. This can be seen in the consideration of the artificially weak and strong feedback models, shown in yellow, red and maroon in Figure 2. The weak feedback model quickly evolves to one dex above the observed average metallicity and likewise, the stronger feedback models rapidly suppress metallicities. Physically, this is due to SN feedback quenching star formation via the expulsion of metal-enriched gas from the halo.

Figure 3. Metallicity by V-band magnitude (left) and stellar mass (right). The left panel shows a comparison against observational data while the right panel compares with recent zoom simulations. Lines indicate how the metallicity of the galaxy evolved over time. Only the Hires and Hires+RT models appear to be in agreement with the data.

In summary, the authors present a new zoom simulation of dwarf galaxy formation and have undertaken a comprehensive study of various star-formation scenarios by exploring how different feedback mechanisms affect key observables. While the results are subject to further robustness testing with a larger sample of dwarf galaxies (allowing the authors to test the effect of different mass accretion histories), the key message from this paper for simulators to keep in mind is that the success of a galaxy formation model may not be in how many observations it can reproduce, but rather how well it can reproduce a specific few. In the case of the smallest known galaxies, the work of Agertz et al. suggests that the present day mass-metallicity relation seems to be the most promising discriminator of galaxy formation feedback models.

About Guest

This post was written by a guest author. If you're interested in writing a guest post for Astrobites, please contact us.

Leave a Reply