Rewinding the universe

Heitmann runs one simulation project, “MockBOSS: Calibrating BOSS Dark Energy Science with HACC.” A second project, “Simulating Cosmological Lyman-α Forest,” is headed by Zarija Lukić, a physics project scientist at Berkeley Lab.

MockBOSS will deploy a new N-body code called HACC (Hybrid/Hardware Accelerated Cosmology Code). N-body codes simulate the complicated movements of objects as they interact through their mutual gravitational pulls.

“HACC will model the evolution of matter over time into the cosmic web,” Heitmann explains. “In post-processing, galaxies are added using a halo-occupation distribution model. This model is calibrated against observations.”

For two years in a row, the code has been a finalist for the Association for Computing Machinery’s prestigious Gordon Bell Prize recognizing outstanding achievement in high-performance computing.

The largest cosmological simulation runs on Mira, a Blue Gene/Q at the Argonne Leadership Computing Facility. The cube-shaped volume of space in the simulation measured 15 billion light years on a side. If that block of space were shrunk to fit into Arizona’s kilometer-wide Meteor Crater, the disk of our Milky Way galaxy – 100,000 light years from edge to edge – would be only 2 millimeters wide, the diameter of a No. 2 pencil lead.

The simulation can reveal details with a resolution of about 23,000 light years. That’s roughly the distance from Earth to the center of the Milky Way, or one-half millimeter on the Meteor Crater scale. Working with a trillion particles, each representing a mass equivalent to a billion suns, the HACC simulation was more than three times larger than any prior model.

With HACC, “you can investigate the effects of different cosmological scenarios,” Heitmann says. “You can change properties of the dark matter and dark energy for example and investigate how best the resulting differences can be observed by the surveys.”

BOSS and MS-DESI – and therefore HACC simulations – target luminous red galaxies, bodies bright enough to be seen from about 6 billion light years away and bearing evidence of their distance and velocity in their light.

Lukić’s project will study light that comes from even farther away – from quasars, the brightest objects in the universe. A quasar is thought to be the core of a so-called active galaxy, where a supermassive black hole drags stars, dust and any other material within reach to an implosive death. Quasars shine from as far as 11.8 billion light years away.

The electromagnetic energy a quasar emits spans the spectrum, from radio waves to gamma waves. As the energy travels through space, the waves are continuously stretched to longer wavelengths, or redshifted, by the universe’s expansion. At the same time, the energy passes through clouds of hydrogen gas. Each cloud robs the emission of a specific wavelength of ultraviolet light. When astronomers examine the spectrum, the missing wavelength shows up as a dark strip called a Lyman-alpha line. If the energy passed through several hydrogen-containing clouds and was redshifted in between, then Lyman-alpha lines are spaced out like a stand of trees, a so-called a Lyman-alpha forest. Each line records the distance of a different cloud.

“Each quasar we observe gives us about 1 billion light years of skewer – a one-dimensional map (of the clouds) in that direction on the sky,” Lukić says.

BOSS will take spectra from about 160,000 quasars. MS-DESI will hunt up about a million. “Having so many one-dimensional skewers will enable us to reconstruct a three-dimensional distribution of neutral hydrogen in the universe, about 10 billion to 12 billion light years away from us,” he says.

For his simulations, Lukić will use a code named Nyx on Hopper, a Cray XE6, and Edison, a Cray XC30, both at the National Energy Research Scientific Computing Center. The code’s lead developer is Ann Almgren in Berkeley Lab’s Center for Computational Sciences and Engineering. Like HACC, Nyx is an N-body code, but also incorporates hydrodynamics. It treats dark matter as a pressureless fluid and visible, baryonic matter as an ideal gas. Nyx can follow gas distribution in the simulation, while HACC focuses on the dark-matter distribution. Nyx incorporates richer physics that are important on smaller scales but also simulates smaller volumes because it demands more computing power.

Lukić says the various studies may finally converge on dark energy’s identity. “Having a good map of what primordial fluctuations are via the CMB, what the cosmic web is now via galaxies and clusters of galaxies, and also intermediate times via the Lyman-alpha forest,” he says, “will enable us to understand better what dark energy can be, and what models can reliably be ruled out.”