Not-So-Dark Matter? Cosmologists Discover Substructures Inconsistent With Current Theory

Art by Charlotte Leakey.

If you’ve ever read a schlocky sci-fi novel or tuned into an episode of Cosmos, you’ve probably heard of dark matter, the mysterious sister to ordinary baryonic matter that makes up some eighty-five percent of our universe. So-called because it doesn’t interact with EM radiation or normal matter, dark matter has only ever been observed indirectly via its gravitational influence. What is it made of? No one is quite sure—candidate explanations range from new elementary particles to primordial black holes. Nearly all of the major schools of thought in the astrophysics community subscribe to a “cold dark matter” model, in which the constituent particles move slowly and larger structures emerge hierarchically from the bottom up. But, while the consensus CDM model has been very successful, there are still some inconsistencies with observation—and another big one has just emerged.

The Findings

In a recent high-profile study published in Science, several cosmologists using the Hubble Space Telescope and the aptly named Very Large Telescope in Chile observed eleven different galaxy clusters, studying gravitational lensing effects—the warping of light from distant background galaxies under high gravity—to map out their dark matter distributions. What they found was consistent with expectations on the cluster scale. However, upon deeper inspection, the cosmologists uncovered a shocking phenomenon on smaller scales (five to ten kiloparsecs): within the individual cluster member galaxies the dark matter was so concentrated that it produced lensing effects stronger than predicted by a factor of ten! In other words, the dark matter in these galaxies was packed together much more tightly than it should have been.

This was an incredible discrepancy, but the researchers made sure their methodology was airtight. They arrived at their predictions by generating a probability metric for lensing effects using models of similar galaxy clusters in various numerical simulations. First, in the real universe, they determined the speeds of stars in several cluster galaxies using spectroscopy, which enabled them to constrain the dark matter distribution with high precision. They then analyzed simulated galaxy clusters with similar masses and distances from Earth and compared the resulting dark matter distributions with their observations. To top it all off, this procedure was repeated with several numerical simulations and methodologies developed by independent research groups. And the outcome? The discrepancy was virtually unchanged across all simulations: a full order of magnitude.

A Quest for Contradictions

This method for investigating dark matter distribution, which has recently become the standard in the field, was originally developed by team member Professor Priyamvada Natarajan of Yale University. Natarajan has made seminal contributions over the years to topics like black hole physics and dark matter physics, particularly the use of gravitational lensing distortions to put theoretical predictions to the test. “Lensing is as direct of an observation as you can make to derive the spatial distribution of dark matter,” Natarajan said. “In my Ph.D., my first paper was actually a conceptualization of the dark matter distribution in a cluster that would make it amenable to be tested with cosmological simulations of structure formation in the universe. I have always been interested, scientifically, in confronting observations with theory.”

Indeed, Natarajan’s publication history is speckled with similar studies of dark matter distribution conducted in intervals of roughly five years, each aiming to harness the best available simulations and compare them with high-power observations to probe for newly detectable differences. “In the past we found broad-based agreement,” Natarajan said, “but initially because of the data quality, we couldn’t push the theory too much.” For instance, previous studies that could only compare the number of dark matter clumps (the granularity of the dark matter) with predictions found no discrepancy—because there was none. Only in 2017, when Natarajan led a study with her collaborators that used these events to map the spatial distributions and internal structure of dark matter in individual galaxies, equipped with the best available Hubble Space Telescope data and state-of-the-art simulations, did small discrepancies begin to appear.

“In 2017, we saw a small hint… But finally now, the quality of the data and the resolution of the simulations have sort of converged and we detected this gap,” Natarajan said. By collecting high-quality observational data en masse and comparing it with simulations based on current theory, Natarajan has subjected the “cold dark matter” theory to ruthless scrutiny—and it seems the model may have finally cracked under the pressure. “We did this computation for many different, independent simulations. And they all agree with each other, but not with the observational finding – so it’s something fundamental, it’s a missing ingredient,” Natarajan said.

The Missing Ingredient

Hence, the situation: the simulations have been independently produced, the results verified, the process peer-reviewed, and the conclusions published. An undeniable gulf now yawns between observation and theory. What, then, could be missing? Natarajan excitedly identified two major possibilities: “When you find a gap, usually it means that the current model you have is missing something—you can often just add in [some new parameters] and get things to match. But very occasionally, you find a mismatch that absolutely cannot be explained with the current theory, but instead points the way to a future theory with more explanatory power.”

To drive home her meaning, Natarajan produced two poignant examples from the history of science. In the early 1800s when the orbit of Uranus was first properly mapped, it did not fit the projections of Newton’s and Kepler’s laws. Using only pencil and paper, the French mathematician Urbain Le Verrier explained the discrepancy by predicting the existence of Neptune. Le Verrier mailed a letter to the Berlin Observatory, and the planet was discovered the very night it arrived—no change to the laws of physics was required. Later in 1859, a similar problem would crop up with the unexplained precession of Mercury’s perihelion in its orbit about the Sun. Le Verrier again predicted an undiscovered planet—Vulcan—but this time, the proper explanation had to wait fifty years for the arrival of Einstein’s theory of general relativity and its complete upheaval of prevailing theory.

“So, you never know, when there’s a gap, whether you’re in the Uranus situation or the Mercury situation,” says Natarajan. It could be that there is some unknown process which occurs, for instance, as galaxies are pulled toward the center of a cluster, that strips or otherwise impacts the dark matter within in ways we cannot yet comprehend. But the unexplained density of dark matter in these galaxies could also indicate a misunderstanding of the esoteric substance’s fundamental nature. Things tend to condense because some force is pulling them together—perhaps there are unknown interacting or self-interacting properties of dark matter? If so, would it really be “dark” after all? Could this discrepancy even be a clue on the long and winding trail toward a quantum theory of gravity? One thing is clear—to one degree or another, it’s back to the drawing board with dark matter.


Meneghetti, M., Davoli, G., Bergamini, P., Rosati, P., Natarajan, P., Giocoli, C., . . . Metcalf, R. B. (2020). An excess of small-scale gravitational lenses observed in galaxy clusters. Science, 369(6509), 1347-1351. doi:10.1126/science.aax5164

Natarajan, P. (2020, September 25). Science Magazine Interview [Telephone interview].

About the Author:

Christopher Poston is a third-year Mathematics/Computer Science major in Pauli Murray College. In addition to writing for YSM, he works as a student software developer at the Peabody Museum and belongs to the Yale Undergraduate Math Society. In his free time he enjoys playing fingerstyle folk guitar, reading hard science fiction, and spending time with his two puppies.


The author would like to thank Dr. Priyamvada Natarajan for her time and enthusiasm in describing her research.

Extra Reading:

Natarajan, P., Chadayammuri, U., Jauzac, M., Richard, J., Kneib, J., Ebeling, H., . . . Vogelsberger, M. (2017). Mapping substructure in the HST Frontier Fields cluster lenses and in cosmological simulations. Monthly Notices of the Royal Astronomical Society, 468(2), 1962-1980. doi:10.1093/mnras/stw3385

Natarajan, P., De Lucia, G., & Springel, V. (2007). Substructure in lensing clusters and simulations. Monthly Notices of the Royal Astronomical Society, 376(17), 180-192. doi:10.1111/j.1365-2966.2007.11399.x

Natarajan, P., & Kneib, J. (1997). Lensing by galaxy haloes in clusters of galaxies. Monthly Notices of the Royal Astronomical Society, 287(4), 833-847. doi:10.1093/mnras/287.4.833