How did this form? (Image: NASA, ESA, N. Smith (University of California, Berkeley), and The Hubble Heritage Team (STScI/AURA)/UCSD)
A SUPERNOVA has exploded in a galaxy not unlike our own. The photons unleashed will travel for billions of years through vacuum, dust and Earth’s atmosphere before being focused on to a telescope. From faint specks in blurred pictures, astronomers will use that supernova and many others to gauge the strength of dark energy, the mysterious force stretching our universe apart.
Advertisement
Only this isn’t our universe. It’s a giant fake – a simulation with a known amount of dark energy plugged in at the beginning. As if checking their answers against the ones in the back of the book, researchers compare the programmed strength of dark energy against that measured from the mocked-up images the virtual telescopes create. The difference between the two will let astronomers refine their real-world methods and help the hunt for dark energy in the real universe.
The telescope that will benefit from this cosmic charade is the Large Synoptic Survey Telescope (LSST), which will look for dark energy and more for real atop a Chilean mountain from 2022. But such experiments are finding uses in all areas of astronomy.
“I would describe them as test universes,” says Rachel Mandelbaum of Carnegie Mellon University in Pennsylvania. “We can test that our analysis algorithms are able to remove all the junk we don’t care about.”
That leaves what we do care about – and not just dark energy. Similar simulations are now being used to probe the equally elusive dark matter, which is thought to make up the bulk of the universe’s mass yet scarcely interacts with regular matter; star-forming nebulae; gravitational lenses; and other cosmological quandaries (see “Mini Multiverses“). It’s an entire fake multiverse that astronomers can spar with before going out to face the real thing.
“It’s a fake multiverse that astronomers can spar with before going out to face the real thing”
Test universes
Of course, the simulations didn’t get this good overnight. For decades, cosmologists have turned to simulations when the interplay between objects got too complicated – and tedious – to solve with pen and paper. One of the first was in the 1970s, when brothers Juri and Alar Toomre collided two galaxies represented by ASCII characters to try to explain the bizarre real-world Antennae Galaxies, which look like a contorted question mark. A video camera, pointed at the monitor as the simulation crunched through frame after frame, to show that a galactic merger yielded a crude version of the galaxies’ shape.
The intervening years have seen ever-bigger and more accurate simulations on supercomputers. “What you have had is progress on three fronts: the speed of computers, our algorithms and our understanding of physics,” says at the Massachusetts Institute of Technology.
In 2005, for example, the Millennium simulation generated about 20 million galaxies in a virtual cube 2 billion light years to a side. Torrey is a member of the ongoing Illustris simulation, which also aims to make a universe from scratch in a jar. Illustris includes not just dark matter, but the devilishly complex motions of gases.
“Experiments for the most part are just not possible to us,” Torrey says. “How do you understand not just the current state of objects, but where they came from and where they’re going? Simulations allow us to fast forward and rewind.”
While these efforts to build a cosmos continue, in the past few years we have seen another use for such simulations: donating fake galaxies to be used as predictive tools for other projects.
The LSST is a prime example. As the actual telescope slouches to completion, simulations are already lighting the way forward.
When it’s up and running, the telescope will map the entire sky every few nights for 10 years, transforming the cosmos into a huge database. It will probe by looking at distant supernovae, weigh dark matter from the distorted shapes of background galaxies, and track potentially dangerous asteroids.
But , who runs LSST’s 10-person simulation team at the University of Washington, doesn’t want to wait to see the universe through the LSST’s eyes.
The simulations Connolly has planned will take tens of millions of hours of processing time on supercomputers. They all start with galaxies from the Millennium simulation. Every step the photons take from those galaxies to a telescope on a virtual Earth, and the creation of simulated maps of dark matter and dark energy, is meticulously traced. All that effort is needed because each intervening step distorts the original signal.
“It’s a simulated view of a perfect representation of the universe,” Connolly says.
Simulations have already helped sharpen our search for dark matter. In 2013 and 2014, Mandelbaum led a – the third so far – in which algorithm was pitted against algorithm to see which made the best map of how gravity distorted images of mocked-up galaxies compared to the “true” answer. And modelling efforts are also planned to help NASA’s upcoming Wide-Field Infrared Survey Telescope, which will look for dark matter and dark energy.
But even as this kind of test comes into wide use, the idea of learning about reality from a fake universe still feels risky. “It’s dangerous to say: I generated a simulated observation that looks like what I see through a telescope, and this looks like reality,” says at the National Center for Supercomputing Applications in Illinois. Since we are counting on simulations to teach us about our scientific methods, trusting the wrong one just because it can make realistic-looking galaxies could lead us to misleading measurements.
Turk is part of the , an open software collaboration that he hopes will help scientists from all over astronomy weigh up their models. Currently, simulations are often run with incompatible software, and researchers don’t always think about how their mocked-up observations can be compared with real data.
Setting a standard is now more crucial than ever, says Turk, especially as astronomy comes to depend on them more and more. “Just doing an awesome simulation is so far removed from contributing to scholarship,” he says. “We need to be very careful about which wheels we decide to reinvent.”
Mini Multiverses
Making observations in simulated universes is helping astronomers perfect techniques before trying them out in the real world, by letting them compare what is measured with the answer that was programmed in (see main story). That approach is shedding light on some of the most pressing problems in the cosmos:
Dark matter
Very slight warps in the shapes of galaxies we see across the sky can show us where dark matter is hiding – but only if we interpret those distortions correctly. Simulations are helping us figure out the best way to pick up and trust in this subtle effect.
Star-forming nebulae
The gas and dust in stellar nurseries, where stars blast their surroundings with ultraviolet light, is too diffuse to recreate in a physical experiment. But simulations are possible, helping us translate between what we observe and the chemistry at play. This can show us the gas that stars coalesce from and the grains of interstellar dust that can grow into planets.
Exoplanets
Models of what possible multi-planet systems would appear like through NASA’s Kepler telescope are being used to look back through Kepler’s catalogue to help find previously undiscovered worlds that may resemble Earth.
Refining simulations
In a weird twist, citizen scientists volunteering with the Galaxy Zoo project are classifying both real galaxies and ones from the Illustris simulation by shape to see how they differ. The results will feed into future simulations, making their galaxies resemble real ones even more closely.
This article appeared in print under the headline “Mocking up the universe”
