Cosmic variance and ergodicity in finite systems with correlations

To understand the universe, cosmologists often rely on a powerful mental shortcut. They imagine that if they could step outside our single, unique universe and look at a vast collection of possible universes, the average properties of that collection would match what we see in our own backyard. This idea, known as ergodicity, suggests that the randomness of the cosmos is distributed evenly enough that measuring a large enough patch of space gives you the same answer as averaging over all possible realities. It is a comforting assumption that allows scientists to compare their theories with the single map of the sky we have. However, this shortcut works best only if the universe is chaotic enough that distant regions have no memory of each other, a condition that breaks down when correlations stretch across the entire observable cosmos.

A recent study by Dipayan Mukherjee and Syksy Räsänen investigates what happens when this shortcut fails. They examined the difference between the theoretical average of all possible universes and the actual average of the one universe we can observe. While it has long been known that this difference shrinks as the size of the observed volume grows, the authors calculated exactly how fast it shrinks and whether it shrinks enough to matter for the scales we can actually measure. They focused on three specific types of cosmic ripples: the curvature of space itself, the clumping of matter, and the flow of cosmic fluids. Their work reveals that for the vast scales we can observe, the difference between the theoretical average and our local reality is not a tiny, negligible error, but a massive distortion that can completely alter how we interpret the data.

The researchers began by asking a simple question: how large must a volume of space be before the average of that volume becomes indistinguishable from the average of all possible universes? In a system where correlations fade quickly, this happens relatively fast. But the universe is different. The patterns of density and curvature that formed in the early moments of the cosmos are linked across enormous distances, creating a long-range order that does not fade away. Because of this, the mathematical rule that usually guarantees the two averages will match up does not apply in the way scientists previously assumed. The authors calculated the specific conditions required for the averages to converge and found that the power spectrum of the universe—the map of how much structure exists at different sizes—must fall off very rapidly for the shortcut to work. In our universe, it does not fall off fast enough.

When the team applied these calculations to the density of matter, which is what we see when we map galaxies, the results were striking. They found that for separations smaller than about 177 megaparsecs, the error introduced by assuming the two averages are the same is manageable. However, once the distance between two points in space exceeds 177 megaparsecs, the error becomes larger than the measurement itself. For any distance greater than 560 megaparsecs, the difference between the theoretical average and the volume average is more than double the size of the signal being measured. This means that for the largest structures in the universe, the "average" we calculate from our single observation is wildly different from the true statistical average of all possibilities. The error is not a small correction; it is a dominant factor that must be accounted for to avoid drawing false conclusions about the nature of the cosmos.

The study also looked at the curvature of space and the velocity of cosmic flows, finding that the bias is significant for these as well. In the case of the cosmic microwave background, the temperature fluctuations we see in the sky, the error is also large on the biggest angular scales. However, the authors noted a crucial distinction here: the analysis of the cosmic microwave background does not involve averaging over a volume of space in the same way as galaxy surveys. Instead, it compares the sky to a theoretical ensemble directly. Therefore, while the mathematical bias exists, it does not complicate the interpretation of the cosmic microwave background data in the same way it does for the large-scale structure of the universe.

The core finding of this work is that the standard assumption of ergodicity, which treats our observable universe as a fair sample of all possibilities, breaks down at the very scales where we are most interested in testing our theories. The authors did not suggest that our theories are wrong, but rather that the way we compare those theories to our observations needs to change. When looking at the largest structures in the universe, the difference between what we see in our specific volume and what the math predicts for the whole ensemble is not a minor detail to be ignored. It is a fundamental feature of a correlated universe that must be included in the error bars of any measurement. By quantifying this bias, the researchers have provided a necessary correction for cosmologists, ensuring that when they look at the vast, connected web of the universe, they are not mistaking a local fluctuation for a universal truth.