Collecting light from distant stars and galaxies is pretty straightforward. You just need a really big light bucket (also known as a telescope). But turning that light into a story about how the Universe was born, grew up, and will die is much harder.
At the huge distances between galaxies, gravity reigns*. To see how structures in the Universe change over time, researchers use computer simulations: they throw a bunch of virtual particles in a box, turn on gravity, and see what happens.
Of course, the devil is in the details. Starlight, supernova explosions, piles of gas and dust flying between galaxies… all of this is a complex interaction of much more than just gravity. The fun really begins when scientists start adding more detailed physics to their simulations. There are two main approaches: smoothed-particle hydrodynamics and adaptive mesh refinement.
The first case, smoothed-particle hydrodynamics, does pretty much what it sounds like: individual particles, each representing thousands of stars or large chunks of empty space, are free to flow through a virtual Universe over time. The particles’ motions are governed by whatever laws of physics the box is told to use. The other case, adaptive mesh refinement, is sneakier: instead of tracking every virtual particle, the Universe-box is divided into chunks of different sizes depending on how much is happening in that area. A relatively empty region can be large and have low-resolution, because it doesn’t need much computing power, while a relatively full region that is busy forming galaxies and stars is small and has high-resolution.
But why all the caveats? Why not simulate every portion of a Universe-in-a-box down to the size scale of, say, a star or solar system? Wouldn’t that be more accurate than “smoothing” elements of the Universe into unwieldy virtual particles or “adapting” your box to different resolutions in different regions?
The problem is dynamic range. Important physical processes govern the Universe over an immense range of size scales, and even the most powerful supercomputers can’t handle that level of detail in a simulation. For instance, hydrogen fusion is our Sun’s source of power. Fusion takes place on size scales of about 10-10 meters. On the other extreme, galaxies tug on one another to shape the Universe on size scales around 1022 meters. That spans more than 30 orders of magnitude!
You can experience a miniature version of this problem for yourself. Recently, NASA organized a “Global Selfie” for Earth Day. The idea was for people all around the world to take a picture of themselves and post it using social media. The result is a massive 3.2 gigapixel mosaic that you can explore in a web browser.
Notice how your computer has to work really hard to load that mosaic, and zoom around it. It’s hard work to go from fully zoomed-out to a clear view of over 35,000 individual images. And this is a mere 5 orders of magnitude.
Despite this hurdle, astronomers have made great leaps in simulating our Universe. For instance, the Illustris project uses a technique similar to adaptive mesh refinement to simulate an impressively realistic chunk of the Universe. Even so, the fact remains that the real Universe is far more intricate than our most powerful computers can imagine.
*At the largest scales, the mysterious force known as dark energy actually reigns.