Research Spotlight: N-body Shop Simulates the Universe

They say that the best simulation is the one you never have to run. We respectfully disagree. We prefer the simulation you only have to run once*. This is how George Lake did it, it’s how the N-Body Shop does it, and it has worked out pretty well so far. We present to you the newest in the N-Body Shop large scale cosmological simulations. Ladies and gentlemen, for your consideration ROMULUS25!

Why we need cosmological simulations

Running an astrophysical “experiment” is hard. We cannot travel a hundred million light years to a distant galaxy and study its evolution over the course of billion of years. However, N-Body simulations are a way to perform experiments to understand how galaxies form and evolve using a supercomputer. Specialized parallel software models a large region of the universe as a collection of billions of mass elements, integrates the relevant equations that model… the expansion of the Universe, gravity, electromagnetic and hydrodynamic forces, dark matter, cosmology, star formation, supernovae explosions, and the formation and growth of supermassive black holes (rocket science? That is easy). The outcome is a series of detailed snapshots of the state of that region of the Universe, described by the position, velocity and thermodynamical properties of each mass element. Scientists can then identify “galaxies” within the simulations as a specific collection of mass elements and study their evolution in time.

romulus25_3panel
A color density slice of Romulus25 at z=0.4, centered on a group of galaxies with total mass 10 times the size of our own Milky Way. Left: The underlying DM distribution. Center: the distribution of stars, color coded with their metallicity (red: metal poor, red: metal rich). Right: the distribution of stars, color coded with their age. Red: old, Blue: young. Most stars are older than 1 Gyr as star formation is “quenched” in the dense group environment. The White dots mark Black Holes.

What is ROMULUS25?

A new, state of the art cosmological simulation encompassing over 15,000 cubic Megaparsecs (Mpc) that is able to resolve the internal structure of galaxies down to dwarf galaxies and over the entire Hubble time. The simulation includes a novel treatment of Super Massive Black Holes physics and it is being run on the Blue Waters supercomputer, using our Tree+SPH code ChaNGa.

Science Goals

With ROMULUS25, we plan to answer the following questions:

  • How many small galaxies are there? We are quantifying the number evolution of the faint end of the galaxy Luminosity Function at z >5 and its contribution to the reionization of the Universe (Lauren Anderson).
  • Understand the genesis and evolution of empirical scaling relations for galaxy properties (Michael Tremmel in collaboration with Dan Taranu at UWA and the SAMI Galaxy Survey team)
  • Study the co-eval assembly of supermassive black holes and galaxies (Michael Tremmel, Zoe Deford, and Josh Smith in collaboration with Marta Volonteri, IAP Paris)
  • Examine when and in what galaxies do SMBH mergers and dual AGN occur and where these events fit in the bigger picture of galaxy-BH coevolution (Michael Tremmel, Daven Cocroft, and Daniel Simons in collaboration with Marta Volonteri)
  • Where do Supernovae and SMBH winds go? We will measure the amount of elements pushed by SNae and SMBH winds into the CGM as a function of galaxy mass, SMBH activity, and redshift. (Michael Tremmel in collaboration with Prof. Jessica Werk)
  • Which galaxies host Super Massive Black Holes? Predicting the occupation fraction of SMBHs in dwarf galaxies (Michael Tremmel in collaboration with Prof. Marta Volonteri)
  • Disentangle the effects of environment and Black Holes physics on the observed shutting off of star formation in massive galaxies (Michael Tremmel in collaboration with Andrew Pontzen at UCL)
  • Dust attenuation in high redshift galaxies. In particular, Understanding how the relationship between UV and IR emission is different in the Early Universe (Danielle Skinner and Lauren Anderson)

Why another large scale simulation?

Excellent cosmological simulations already exists, like EAGLE and MILLENIUM and they have been instrumental in shaping our knowledge of galaxy formation. While all recent simulations adopt a cosmological model based on a Cold Dark Matter model and an accelerating expansion driven by dark energy, the original advantages that ROMULUS25 brings are multiple:

  • Sub-kpc spatial resolution that allows us to simulate at the same time the evolution of the bulge and disk structure of systems time size of Andromeda and above and the properties of dwarf galaxies as small as the Magellanic Clouds.
  • Two databases (one halo based, in collaboration with Andrew Pontzen, and the other particle based) that allow us to follow a) the evolution and merging history of various galaxies and the b) thermodynamical history of each simulated mass element across cosmic time. This is a unique strength of particle based, SPH tree-codes such as ChaNGa.
  • A novel physical description of SMBH formation, dynamics and accretion that will allows us to, for the first time, realistically capture the merging rate of SMBHs over cosmic time and their frequency as a function of host galaxy properties.
  • Having a detailed knowledge of the spatial distribution, age, and metal content of the stars and neutral hydrogen in each simulated galaxy allows us to create “artificial observations,” which can be used to compare the outcome of the model with the properties of galaxy samples in the real Universe
Interacting z=1 galaxies color map
A color map of interacting galaxies at z=1. Gray maps the dark matter distribution, red and blue old and young stars respectively, while the large green dots mark the position of the active Super Massive Black Holes. One galaxy has had its star formation “quenched,” possibly by a combination of intense feedback from SMBHs and having gas stripped away by its companion, while the one on the right is a disk galaxy that is actively forming stars. The frame is about 100 kpc across. This system eventually results in a BH merger and a quenched galaxy by z = 0.5.

The UW Team

Several people are involved in this project: Michael Tremmel and Lauren Anderson are the grad students currently involved in the project. Prof. Tom Quinn develops and maintains ChaNGa as part of a collaboration with a computational group group at NCSA. Prof. F. Governato is involved in the science planning and analysis. We also have a group of undergraduate students working on the project: Danielle Skinner, Zoe Deford, Josh Smith, Daven Cocroft, and Daniel Simons.

Powers of ten fun facts

  • The database will contain 10,000+ galaxies
  • Romulus25 is being run using up to 100,000+ cores
  • The amount of data created (about 100 time snapshots) will reach 10+ Terabytes and requires a database and an efficient analysis toolset to properly analyze it.
  • To reach redshift zero the simulation will require 100 million CPU hours

The main physics modules in the ROMULUS25 simulation

  • Gravity duh!
  • Gas heating, cooling, metal enrichment and diffusion
  • SN feedback and cosmic UV radiation
  • BH formation, dynamics, accretion, and feedback
  • Gas dynamics resolved down to 30-60pc

Collaborations and Future Work

We have ongoing collaborations with the IAP in Paris (Volonteri), UCL in London UK (Pontzen) and with the SAMI survey in AU (Taranu). But with so much data and a rich science agenda we welcome scientists near and far and especially students interested in applying for grad school at UW to get in touch with us. We work hard to make our group a most effective and welcoming work environment for everybody interested.


A color density map of the DM distribution centered on a group of galaxies in ROMULUS25. The movie shows a large number of subhalos orbiting in the potential of the group, which weights about ten times the one of our own Milky Way. Each of the most massive subhalos hosts a luminous galaxy. The frame is about 2Mpc per side.

A color density map (red: hot, blue: cold) of the gas distribution. Note the galaxy with a bipolar outflow (possibly driven by the central SMBHs and the interaction with the cluster’s hot diffuse gas).

This research has been made possible by two NSF awards, NSF PRAC OCI-1144357 and NSF AST-1311956, with a total time allocation of 200 million CPU hours. NSF

*for now!