One megaparsec is about three million light years, a distance that sounds enormous until you set it against the whole observable universe. On cosmic scales it is small. And it is exactly where three astrophysicists, Ethan O. Nadler, Keir K. Rogers and Alex Drlica-Wagner, say the most interesting arguments about dark matter are now happening. Their review, posted to the arXiv on 30 July 2026 and submitted to Reviews of Modern Physics, runs to 73 pages with 7 figures and 2 tables, and it takes stock of what the smallest structures in the universe can tell us about the substance that makes up most of its matter.
The logic behind this is worth spelling out, because it is one of the more elegant arguments in cosmology. Dark matter does not shine, so nobody can point a telescope at it directly. But whatever it is made of, its microscopic physics leaves fingerprints on the size and number of the clumps it forms. The authors describe three families of models and the mechanisms that distinguish them. Warm dark matter particles move fast enough in the early universe to stream out of small clumps before those clumps can collapse, a process called free-streaming. Fuzzy dark matter is light enough to behave like a wave, and wave interference smooths out structure below a characteristic scale. Self-interacting dark matter, and dark matter that interacts with ordinary Standard Model particles, changes both how many halos form and how matter is distributed inside them.
Each of these mechanisms suppresses or reshapes structure at small scales in its own way. So counting and characterising the smallest cosmic structures becomes a test of particle physics conducted with telescopes.
Five ways to weigh the invisible
The review surveys five observational probes, and the variety is striking. Dwarf galaxies are the faintest, smallest galaxies known, and how many of them exist around the Milky Way depends on how easily small halos can form. Strong gravitational lensing uses a massive foreground object to bend and distort the light of something behind it, and small clumps of dark matter perturb those distorted images in detectable ways. The Lyman-alpha forest is the thicket of absorption lines that intergalactic hydrogen imprints on the spectrum of a distant quasar, and it traces how gas is clustered along the line of sight. Stellar streams are the ribbons of stars pulled out of disrupted star clusters and satellite galaxies, and passing dark matter clumps can gouge gaps in them. High-redshift galaxies, seen as they were when the universe was young, constrain how early structure could get going.
No single one of these is decisive, and the authors are candid about why. They devote substantial space to what they call key modeling uncertainties and observational systematics: the theoretical machinery that converts a dark matter model into a prediction is imperfect, and every observation carries its own biases. A constraint is only as good as the simulation and the instrument behind it.
That honesty shapes what the review recommends. Nadler, Rogers and Drlica-Wagner point to two developments they consider increasingly important. The first is probe combination, meaning that the five methods above are stronger together than apart, partly because their weaknesses do not overlap. The second is simulation-based inference, a statistical approach suited to problems where you can simulate the data but cannot write down a clean likelihood for it. Both are signs of a field that has moved past the stage of any one measurement settling the question.
Why it matters
Dark matter is the largest unsolved problem in physics that we can be confident is a real problem. Something holds galaxies together and shapes the cosmic web, it outweighs ordinary matter several times over, and decades of laboratory experiments designed to catch a dark matter particle have not caught one. The small-scale structure approach sidesteps the laboratory entirely. It asks the universe to run the experiment and reads the answer in the distribution of galaxies.
What this review offers is not a new result but a map of where the boundaries currently sit, and how they were drawn. That is a genuinely useful thing when a field is generating constraints faster than anyone can keep track of them, and when those constraints depend on modelling choices that a reader outside the field would never see. Reviews like this one are how a community agrees on what it actually knows.
The authors close by looking forward, surveying upcoming observational facilities that will sharpen these tests. They do not promise a discovery. The realistic prospect is narrower and, in its way, more interesting: the space of viable dark matter models keeps shrinking, and small structures are doing much of the shrinking.
This is a preprint. It has been submitted to Reviews of Modern Physics but has not yet completed peer review, and the authors have explicitly invited comments.