A law of physics written by Isaac Newton in 1687 has just been tested across a distance of 750 million light-years, and it still works. Not approximately. Not close enough. Within the margin of error the instruments can detect, the universe is pulling galaxy clusters together at exactly the rate Newton calculated three and a half centuries ago, using observations of falling apples and orbiting moons.
Findings published in Physical Review Letters in April 2026 quantify the gravitational pull between galaxy clusters separated by distances ranging from 98 million to 750 million light-years, the largest direct test of gravity’s basic rules ever carried out.
To understand why this matters, start with the rule itself. Newton’s insight was that gravity between any two objects follows a precise pattern: every time you double the distance between them, the gravitational pull drops to one quarter of what it was. Triple the distance and it falls to one ninth. The force does not just get weaker; it gets weaker in a very specific, mathematically predictable way. This relationship is called the inverse square law, because the force weakens in proportion to the square of the distance. Newton worked it out from observations of the moon and the planets in our solar system, distances measured in millions of kilometres. The new measurement tests whether that same relationship still holds at distances measured in hundreds of millions of light-years, a scale so vast that a single light-year, roughly 9.5 trillion kilometres, is not even a meaningful unit to think about. At those scales, the building blocks are galaxy clusters: collections of hundreds or thousands of entire galaxies, each containing hundreds of billions of stars, all bound together by gravity into structures tens of millions of light-years across.
The question the measurement set out to answer is one that has quietly unsettled physics for decades. When astronomers measure how fast galaxies and galaxy clusters move, the numbers do not add up. Stars at the outer edges of galaxies orbit their galactic centres far faster than they should, given the amount of visible matter inside those galaxies. Entire galaxies within clusters travel at speeds that the combined mass of every star and gas cloud in the cluster cannot gravitationally justify. Something is generating extra gravitational pull that visible matter alone cannot account for. Two competing explanations have sat unresolved for years. Either the universe contains enormous quantities of matter that is genuinely invisible, emitting no light at any wavelength any instrument can detect, a substance called dark matter. Or the gravitational equations themselves are wrong, and gravity does not actually follow Newton’s inverse square law at very large scales or very weak strengths.
The second of those explanations has a formal name: Modified Newtonian Dynamics, or MOND. A physicist named Mordehai Milgrom proposed it in 1983. His idea was that Newton’s law works perfectly well in strong gravitational environments, such as inside solar systems, but breaks down at extremely weak gravitational strengths, the kind found in the outermost regions of galaxies or in the vast stretches between galaxy clusters. Under MOND, gravity at those weak strengths weakens more slowly with distance than Newton’s law says it should. The practical effect is that galaxies and clusters feel a stronger gravitational pull than Newton predicts, which would explain why they move faster than their visible mass justifies, without requiring any invisible matter at all. MOND has been a serious and persistently debated alternative for more than 40 years, and testing it directly at the scales where it would need to operate has been technically out of reach until now.
The instrument that made the test possible is the Atacama Cosmology Telescope, a six-metre dish built into the high desert of northern Chile at an elevation of 5,190 metres, roughly two thirds of the way to the cruising altitude of a commercial aircraft. Its target is not galaxies directly. It reads the oldest light in the universe, a faint glow of microwave radiation called the cosmic microwave background, or CMB. This light was released approximately 380,000 years after the Big Bang, when the universe had cooled enough for atoms to form and light to travel freely for the first time. That glow has been travelling ever since, filling all of space at a temperature just 2.7 degrees above absolute zero. It is extraordinarily uniform, but not perfectly so, and those tiny variations carry detailed information about what the light has passed through on its journey. When the CMB passes through the hot gas surrounding a moving galaxy cluster, the cluster’s motion leaves a faint imprint on the light: a tiny temperature shift, detectable with precise enough instruments. By reading those imprints across hundreds of thousands of galaxy clusters, physicists can measure how fast the clusters are moving toward or away from Earth without ever visiting them. This technique is called the kinematic Sunyaev-Zeldovich effect, named after the two Soviet physicists who predicted it in 1980.
The team combined those velocity measurements with a separate catalogue of 343,647 galaxies from the Sloan Digital Sky Survey, a ground-based telescope programme that has mapped the positions and distances of millions of galaxies across large portions of the sky. From that catalogue, they calculated how galaxies cluster together at different separations, which tells them how matter is distributed across space. Matter distribution is what generates gravitational pull, so knowing the distribution allows the team to calculate the acceleration that should be driving cluster pairs toward each other. Comparing that predicted acceleration to the actual velocities measured from the CMB imprints gives a direct read of how gravity is behaving at each separation. The test covered 15 distance bins ranging from 30 to 230 megaparsecs, with one megaparsec equal to 3.26 million light-years.
The result is expressed as an exponent. Under Newton’s inverse square law, the exponent governing how gravity weakens with distance is 2. Under MOND, the exponent at weak gravitational strengths drops to 1, meaning gravity weakens more slowly. The measurement returned an exponent of 2.1, with an uncertainty of plus or minus 0.3. The standard gravity prediction of 2 sits comfortably inside that uncertainty range. The MOND prediction of 1 sits more than three standard deviations outside it, a gap large enough that statisticians consider it a serious rejection. The probability of getting data this inconsistent with MOND purely by chance is 0.365 percent.
What this result does to MOND is close a door that had remained technically open. Previous tests of MOND had relied on indirect measurements, or on scales where MOND’s predictions were not dramatically different from standard gravity. This test operates directly in the regime where MOND was designed to deviate from Newton, across the vast low-gravity separations between galaxy clusters, and finds no deviation. The gravitational pull weakens with distance at almost exactly the rate Newton calculated, not the slower rate MOND requires to explain fast-moving galaxies and clusters. If gravity is not weaker over large distances than Newton and Einstein say it is, then the extra pull driving those anomalous velocities must come from somewhere else. The only candidate remaining with a working theoretical structure is dark matter, invisible mass distributed through and around galaxies and clusters, pulling on visible matter through gravity alone without ever interacting with light.
Dark matter has never been directly caught. No particle detector on Earth, including the large underground experiments designed specifically to capture dark matter interactions, has confirmed a single collision between a dark matter particle and ordinary matter. What is documented is the gravitational effect: something non-luminous is pulling on visible matter at every scale from individual galaxies to the largest structures in the observable universe, and no modification to Newton’s equations survives this scale of direct test. The Atacama Cosmology Telescope’s observing programme concluded in 2022, with the dataset underpinning this result accumulated during years of operation. The next instrument at the same Chilean site, the Simons Observatory, is currently under construction and will map the CMB with roughly ten times the sensitivity, covering a larger area of sky and reaching fainter cluster signals across even greater separations than the 750 million light-years probed here.
SOURCES
Gallardo, P. A., Pardo, K., Philcox, O. H. E., et al. (2026). The Atacama Cosmology Telescope: A Test of the Gravitational Force Law on Cosmological Scales Using the Kinematic Sunyaev-Zeldovich Effect. Physical Review Letters. https://doi.org/10.1103/rk8v-rcm3
arXiv preprint: https://arxiv.org/abs/2604.14327
