32 AUSTRALIAN SKY & TELESCOPE July 2017
~10^10 parsecs
Observable
universe~10^4 -10^5 parsecs
Galaxy~10^7 -10^8 parsecs
Clusters, large-scale
structure0 h 23 h(^22) h
1 h
2 h
Distance (billion light-years)
2.5
2.0
1.5
1.0
0.5 0.5
1.0
1.5
Distance (billion light-years)2.0
2.5
BOLSHOI SLICES: N. MCCURDY (UC-HIPACC) / R. KAEHLER & R. WECHSLER (STANFORD) / M. BUSHA (UNIV. ZURICH) / SDSS
As a result, the cosmological simulations community has
long been waiting for a time when computers and algorithms
would be powerful enough to handle the complexity of
baryonic physics. That time is now.
Small beginnings
While the dark-matter-only simulation experts offer insights
into how the universe evolves as a whole — albeit without
directly including anything in the simulations that is observable
— those running cosmological simulations that include
ordinary matter have faced a dilemma akin to trying to make a
big cake with a delicate, complex taste: how to simultaneously
bind together large-scale dark matter structure growth with
smaller-scale baryonic objects and processes — things like black
hole feedback, gas dynamics and star formation. It all boils
down to computational cost: the effort, complexity and time
required to complete a realistic simulation.
“The community has bifurcated a bit now,” says Wetzel.
“Groups doing large-volume simulations that encompass
thousands of galaxies at a time gain excellent statistics to
compare against observational surveys. But the trade-off is
that these simulations are not able to resolve the relevant
physical processes within a galaxy.”
In contrast, groups conducting ‘zoom-in’ simulations start
with a large-volume box but only have high resolution in the
region immediately around a single galaxy. “This means that
such simulations only model one galaxy at a time, but they
resolve it exquisitely well,” he says.
Zoom-in simulations that focus on a smaller cosmic box can
be likened to home baking. Bakers can pick as many ingredients
as they want and take their time over the bake, because the
simulation is covering a smaller spatial area and thus can
incorporate more physical processes at higher resolution.
FIRE (Feedback In Realistic Environments) and CLUES
(Constrained Local Universe Simulations) are two projects
running such simulations, with vastly different approaches.
What sets FIRE — and its more recent higher-resolution
offspring, ‘Latte’ — apart is the researchers’ attempt to
build the simulations from the bottom up. “We start by
asking ourselves, for example, how stars behave on small
COSMIC PIE
The Bolshoi simulation’s
predictions for what large-scale
cosmic structure looks like (left)
closely match the real distribution of galaxies
observed by the Sloan Digital Sky Survey (right) —
in fact, statistically they’re virtually identical.
Each wedge spans a quarter of the way around the sky.
TCOSMIC SCALES Cosmological simulations try to replicate the universe and its evolution, and early ones produced remarkable matches
to large-scale structures (left-hand boxes in this illustration). But for many years astronomers were forced to fudge what was happening at
the smallest scales (other illustrations) — processes that, it turns out, heavily influence the larger structures. (A parsec is 3.26 light-years.)
UNIVERSE 2.0