This post refers to an old banner image, which you can find here.

What’s that picture up there at the top of the page?

The easy answer is that it’s a computer-generated image from the Millennium Simulation of what we think the universe looks like on the largest scales.1 That’s not quite true, though. It’s certainly an image from the Millennium Simulation, but it’s not what we think the universe “looks” like, at least not literally; those luminous dots indicate dark matter, not galaxies, meaning they’re not made of anything that actually gives off light of any kind.2 More importantly, it’s not what we think our universe looks like because it’s not a map of our universe. It’s a map of a fake universe, a wholly simulated construction living inside a computer, a universe that does not have a Milky Way, much less a Sun or an Earth. This simulation gives coordinates for massive clusters and superclusters of galaxies, and it’s detailed enough that we can make stunning fly-through movies, but none of those clusters directly correspond to the location of any known structures in our universe. In fact, if they did — if we found groups of galaxies in the sky in the same positions as the galaxies in this simulation — nobody would be more surprised than the physicist-programmers who created this mock universe. Directly mapping our universe was never the goal of this simulation — yet it was considered an unqualified success, and the resulting pictures are widely described as pictures of what our universe looks like on the largest scales. What’s going on here?

Our best theories of how the universe works on the largest scales don’t make specific, deterministic predictions about where we will find galaxies and clusters of galaxies in the sky. The standard model of cosmology,3 the ΛCDM model, gives predictions about the behavior of the universe that match our best data to better than 1% — but the ΛCDM model can’t tell us where to point Hubble if we want to see the first galaxy that ever formed. Our cosmological models are statistical in nature: they tell us about overall average properties of the universe, but not specific properties of tiny patches of the universe.4 So we can’t get the locations of galaxies or stars out of ΛCDM, but we can get the average mass of galaxy clusters at a particular period of time, or the probability that two galaxies are a certain distance apart from each other. And on these statistical measures, the Millennium Simulation looks like our universe’s twin, despite the fact that the actual positions of galaxy clusters and dark matter halos are very different between the two.

In short: that picture isn’t really a picture of our universe at all. If you had a huge camera (which was somehow sensitive to dark matter) and took a picture of our universe on the same scale, it wouldn’t look much like that. But if you looked at your picture and that picture, side by side, you’d probably think that they were two different pictures of the same kind of thing. In fact, if you take a look at a picture of the galaxies from the Millennium Simulation and a picture of one of our largest galaxy surveys, they do look similar — almost as if they’re pictures of two different regions of the same kind of forest.

Pretty, no?

1. On the scale of the image, the width of this sentence is about a billion light years — roughly 10,000 times the diameter of the Milky Way. [↩]
2. Galaxies are certainly associated with dark matter halos, but a map of galaxies from the same computer simulation doesn’t look exactly like the map of dark matter. [↩]
3. Not to be confused with the Standard Model of particle physics [↩]
4. Interestingly, while this comes as a surprise to most non-scientists, it’s something so basic to the practice of cosmology that most cosmologists don’t even think about it. [↩]