The Eris Simulation

The soldiers fought like wolves while Eris, the Lady of Sorrow, watched with Pleasure. -- The Illiad.

ADS | astro-ph

Eris is the highest-resolution cosmological simulation of the formation of a late-type spiral galaxy to date. It was sampled using the "zoom-in" technique to archieve very high resolution at the center while keeping a coarse dark matter distribution in the outer regions. The run was performed using Gasoline, an N-body + smoothed particle hydrodynamics (SPH) code that allows for a large dynamic range due to its Lagrangian nature and includes the physics of star formation, heating by an UV background, SNe type Ia and type II feedback (blast-wave) and enrichment, and cooling. No black holes or AGN feedback are included in this run. Eris was run on the NASA Pleiades supercomputer consuming roughly 1.6 million cpu-hours (9 months including testing overhead) from z=90 to z=0.

Here are some basic facts of Eris at z=0 (and a rough comparison to the MilkyWay):

The baryon fraction within Eris' virial radius is 30% below the universal value due to the loss of baryons via outflows triggered by supernovae explosions (there is no AGN in the simulation). The extent of these outflows and their effect on the enrichment of the IGM is currently being studied (Shen at al. 2011, in prep). Eris' disk is extended and its bulge is modest, with a B/D ratio of 0.35, although this is a very conservative value since our photometric decomposition does not detangle the light from the bulge and the bar. Furthermore, the bulge's Seric index is low (~1) and a large fraction of the "bulge" particles have a high specific angular momentum, so Eris' bulge could be classified as a pseudo-bulge.


Images: Gas (gray), young stars(blue) and old stars (red) or Eris at z=0. The box sizes are 38 kpc (left) and 25 kpc (right).

Movies: Click on the images to view/download the movies.


Eris' rotation curve at z=0 rises and falls off slowly, having a peak circular velocity Vpeak=238 km/s at r=1.34 kpc. In order to assess the important of the star formation threshold on the resulting galaxy, we ran a twin simulation to Eris, called ErisLT (low threshold). ErisLT produced a much more centrally concentrated galaxy (see the comparison at z=1 below), with a larger bulge and a smaller disk. In turn, Eris' rotation curve is in good agreement with observations of blue HB stars in the Milky Way. Eris' HI rotation curve fits quite nicely with THINGS data (Oh et al. in prep, see Lucio Mayer's talk at Durham, page 14).

They key for the success of Eris and simulations of bulgeless dwarf galaxies is in the high star-formation threshold: we allow stars to form only in the highest density peaks, creating an inhomogeneus ISM. Feedback is much more efficient in this case, because it removes preferably high-density / low angular momentum gas from the galaxy. Below is a map of the cold gas and the star formation rate density. The HI gas is much more extended than the stellar component, and the regions of star formation are confined to very high-density HI regions (>20 cc) along the spiral arms. Since we don't have H2 physics in the simulation, we don't reach the true densities at which star formation occurs (>100 cc), but our HI data matches quite well the Kennicutt-Smith relation observed in nearby spirals.

The resulting outflows have two important observational consequences 1) lower the baryon fraction inside the virial radius and 2) enrich the IGM with metals (Shen et al. 2011, in prep, see Sijing Shen's talk at Joelfest).

A nice way to see why the star formation threshold is so important is by looking at the PDF of the gas phase. Below are the PDFs of our high-threshold (left) and low-threshold (right) simulations, color-coded by gas temperature. The star formation threshold is shown in green. In Eris, the ISM is quite inhomogenerous, and there is a higher fraction of cold gas.

In ErisLT, supernova feedback affects mostly low-density gas and therefore is very inneficient at quenching star formation and at removing low-angular momentum material from the inner regions. This is why previous simulations that used low SF thresholds yielded disk galaxies that formed too many stars for their halo mass and were too centrally concentrated.

Lastly, here is a (very cheesy) I-band image of Eris compared to the 2MASS image of the Milky Way.

Image credit: 2MASS, Simone Callegari, Javiera Guedes.

For our latest work on Eris go to my research page.