New video: Largest-scale cosmic test ever favors dark matter over MOND

Anytime there’s a chance to subject our Universe to a new cosmic test, it’s imperative that we do it. First off, it gives the Universe a chance to surprise us: to show us something that we never could have learned unless we looked. Second off, it gives us a chance to test our underlying assumptions about the Universe: what it’s made of, what laws and rules it obeys, and how it behaves in never-before-probed ways. And third — and perhaps most importantly — if there are rival ideas for making sense of the Universe, a new test has the potential to answer the question definitively: with data from the cosmos itself.

With unprecedented data about both:

• the large-scale structure of the Universe, arising from galaxy surveys like SDSS and DESI,
• and the cosmic microwave background radiation, from Planck, the Atacama Cosmology Telescope, and the South Pole Telescope,

we can finally conduct a long-awaited cosmic test: we can measure the kinetic Sunyaev-Zel’dovich effect. These measurements have the potential to test the idea of dark matter against the idea of MOND (MOdified Newtonian Dynamics), testing gravity on the largest scales and at the smallest accelerations ever. In a brand new video, I summarize the results and put them in their proper cosmic context. Have a watch below, and read the transcript beneath it to make sure you catch every detail!

Transcript

Is the Universe made up of an enormous amount of dark matter, and is that responsible for the gravitational effects that we see on galactic and cosmic scales that normal matter alone can’t explain?

Or is there no extra matter, and was Einstein wrong about how gravity works on the largest scales in the Universe?

The leading hypothesis is dark matter, but a rival idea – Modified Newtonian Dynamics, or MOND – paints a very different picture.

In a brand new study, the largest-ever direct test of dark matter vs. MOND has just been conducted, by looking at both the large-scale structure of the Universe and the temperature properties of the cosmic microwave background together, and inferring a very difficult-to-measure effect: the kinetic Sunyaev-Zel’dovich effect. These measurements enabled astronomers to determine, on average, how quickly two galaxies separated by a given distance accelerate towards one another: a measure of the acceleration due to gravity on the largest scales ever.

When the results are obtained, on enormous cosmic scales – between 100 million and up to 750 million light-years – they were able to show that the law of gravity behaved as Einstein’s General Relativity predicted, and not as the alternative, MOND, predicts on those same cosmic scales.

It’s a strong piece of evidence in our quest to uncover both how the Universe works and exactly what it’s made of, posing another colossal difficulty for alternatives to our standard picture of how reality truly is.

Introduction

Our main goal, always, in physics and astrophysics, is to understand the Universe as it actually is. Two of the biggest questions we know how to ask are:

• what are the laws that govern reality,
• and what are the ingredients that make up our Universe?

Over recent decades and centuries, we’ve built up a consensus picture that represents humanity’s best approximation to reality: with General Relativity and Quantum Field Theory providing the laws, and the Standard Model plus dark matter and dark energy making up the ingredients.

However, it’s important to recognize that there are alternative formulations and viewpoints that challenge this consensus picture, and that ability, to challenge the consensus, is central to the advancement of science.

Testing those alternatives against our present consensus model, side-by-side with actual data, is one of the most powerful ways to reveal how the Universe is and isn’t.

How we learn

Anytime you make a new observation – anytime you view or analyze the Universe in a fundamentally different or novel way – it’s an opportunity to test our best answers to these questions, and to test alternatives against the consensus wherever their predictions differ.

If the new observations and the prevailing theory are compatible, that’s good: the theory is validated.

If the new observations and the prevailing theory are incompatible, that’s also good: they expose a flaw or incompleteness in our current understanding, and may pave the way for a new, superior picture of reality.

And if the new observations can tell two fundamental ideas about what the Universe is made of or what laws govern it apart, that’s the best possible outcome of all. It points towards a major success of one idea while simultaneously identifying a major flaw in the other.

This brings us to an exciting new piece of research that puts the question of “dark matter vs. MOND” to its largest-scale ever direct test. First, a little background.

Dark matter vs. MOND

According to the dark matter hypothesis, we’ve got the laws governing gravity correct with General Relativity, but we need an extra ingredient to the Universe, some form of dark matter, to explain all that we see. This idea first came about all the way back in the 1930s when looking at the motions of individual galaxies within galaxy clusters, but it grew to prominence in the 1970s with observations revealing the too-rapid rotation at too great of a distance within spiral galaxies. It was those internal motions of spiral galaxies, when we could measure them well enough, that first brought the idea of dark matter to mainstream prominence.

Alternatively, according to the MOND hypothesis, there’s no need for a new ingredient, and instead we should just modify the law of gravity. This idea originated in the early 1980s by noting that individual galaxies would have their rotation curves match the data without a new ingredient if instead a new modification to gravity was introduced: one with a “fundamental acceleration” scale. At larger accelerations, you’d recover Newton’s 1 over r squared force law, but at small accelerations, you would transition to just 1 over r, as that “fundamental acceleration scale” would lead to a change in behavior. That MOND scale only appears, however, at very small accelerations. The acceleration of gravity here on Earth is 9.8 meters per second-squared, but for comparison, the MOND scale is 1.2 Angstroms per second-squared.

Within the Solar System, there’s no way to test MOND, because all of our accelerations are far too large. Even Pluto’s acceleration around the Sun is many orders of magnitude too large; you’d need for the separation distances to be far greater than what we can actually observe.

Astrophysical tests

People have, in recent years, attempted to look at what we call “wide binary” systems of stars, because at very large separation distances, the accelerations actually drop to low enough values that you could potentially test for the effects of MOND. Indeed, several different groups have identified large numbers of systems and looked at their motions statistically, reaching wildly different and mutually incompatible conclusions in the process. That’s because it would take centuries-to-millennia of observations to actually determine the orbits of these systems individually, and confirm that they are indeed gravitationally bound, rather than just coincidentally close objects, while ruling out the presence of a higher-multiplicity (like a trinary or quaternary) system. Since we can’t do that, we have to make guesses as to which points we keep in and which ones we throw

out. And – surprise, surprise – the conclusion you draw depends on which guesses and assumptions you make. So we can’t draw any surefire conclusions from that.

But if we could go to larger scales, we’d be able to test these two hypotheses against one another directly: where accelerations are small, at large cosmic distances, and where the two competing ideas make vastly different predictions for how quickly any two pairs of massive objects should be seen accelerating towards one another on average. There’s a chance to do this not just on solar system, stellar, or galactic scales, but on cosmic ones. The key is through an effect known as the kinetic Sunyaev-Zel’dovich effect, which has taken us decades to actually be able to measure.

Astrophysics background

In our Universe, we have a background source of light that comes to us from all directions in space from our perspective: the cosmic microwave background. When it passes through empty space, the light we observe is identical to the light that was emitted, save for the redshifting effects of cosmic expansion. But if that light passes through a cloud of ionized matter – particularly if it’s an electron-rich cloud – it’s going to have two different ways of altering that light.

First, depending on the temperature of those electrons, the photons that pass through them will get boosted to higher energies. This creates what looks like a “cold spot” in the background radiation, because there’s a smaller amount of energy in the wavelengths where you’ve been seeing the rest of that background. But in reality, that radiation is hotter, because it’s been altered, and shifted up to hotter temperatures and higher energies. Because it’s dependent on electron temperature, this is called the thermal Sunyaev-Zel’dovich effect.

But remember, we were interested in the large-scale motions of objects, relative to one another, caused by their gravitational accelerations. A thermal effect doesn’t shed light on that. If the galaxies that have these ionized electrons in them are also in motion, however, that motion will also imprint itself onto the background light that passes through it, either boosting or de-boosting the energy of that light in the process. In other words, there isn’t just a thermal Sunyaev-Zel’dovich effect, but also a kinetic (or, sometimes, kinematic) Sunyaev-Zel’dovich effect as well.

The new test

The way the researchers approached this problem was to take regions of sky that had both really excellent cosmic microwave background data, using the Atacama Cosmology Telescope and really excellent large-scale structure data, using spectroscopic galaxy maps from the Sloan Digital Sky Survey. Although isolated galaxies don’t generally affect the CMB, or cosmic microwave background, in meaningfully observable ways, galaxies within galaxy clusters – which often do have large quantities of ionized electrons inside – makes a great laboratory for finding galaxies that can have their Sunyaev-Zel’dovich effects measured.

So that’s the procedure:

• Take large numbers of galaxies,
• Where the kinetic Sunyaev-Zel’dovich effect can be measured in that same region.
• You measure their properties, spectroscopically, to determine their 3D positions and also their velocities relative to the Hubble flow.
• Then, you correlate all pairs of galaxies together, where you can estimate both the relative velocities between different pairs, which gives you a window into their mutual gravitational accelerations, and the separation distance between each pair.
• And then, you see what the relationship is between those three things at once. You want to see the relationship between:

1. The kinetic Sunyaev-Zel’dovich effect, which you measured from the cosmic microwave background. You want to see the relationship between that and
2. the pairwise velocity of each set of galaxies as a function of your chosen theory of gravity, relative to the MOND theory, or relative to Einstein’s General Relativity with dark matter,
3. and then you want to see how that evolves, Sunyaev-Zel’dovich effect versus pairwise velocity, as a function of separation distance between those pairs of galaxies. You’ll divide large numbers of galaxies at all different distances from each other into “bins” for analysis.

On the scales that the research teams looked at, which range from 30 megaparsecs to 230 megaparsecs – or about 100 million up to 750 million light-years – the accelerations due to gravity are extremely low, so this should be a really good test of that low-acceleration MOND regime. If the acceleration doesn’t follow a 1 over r squared law, but rather follows a 1 over r law, it would provide evidence for MOND and not only against dark matter, but against Einstein’s General Relativity itself being valid on cosmological scales. However, if it does follow a 1 over r squared law, and not a 1 over r law, it would directly show that MOND’s predictions are invalid on cosmic scales, and that Einstein’s General Relativity and dark matter remain consistent with all that we see.

Results

What you see here is the main result from the paper: Gallardo et al., Physical Review Letters, 2026. It shows a series of binned data points of galaxies at a variety of separation distances from one another, with the measured effects of the pairwise kinetic Sunyaev-Zel’dovich effect shown on the y-axis. Note how small this effect is: they are looking at ten-to-a-hundred billionths of a Kelvin in temperature fluctuation differences here! What you’ll notice is that at very large distances, the effect is extremely small, but that at small separation distances, the effect gets stronger.

The key question is: by how much?

According to MOND, the increase should be slow: corresponding to a one over r force law.

According to Einstein with dark matter, the increase should be faster: corresponding to a one over r squared force law.

Visually, the data looks like it agrees with the consensus picture extremely well, and looks like it disagrees with the MOND predictions. In more detail, what they’re plotting – known as the pairwise velocity – scales as the integral of the correlation function multiplied by the force law, and that’s what is represented by the blue and yellow lines (with uncertainties), as shown for dark matter and MOND, respectively.

But how do things stack up when we perform a more robust statistical analysis?

As you can see, the standard picture, where we get a one over r squared force law at large distances, looks exactly right: completely consistent. However, the MOND picture, with a one over r force law, doesn’t match the data, and is ruled out, already, at around the 99.9% confidence level.

In private communication with the lead author of the study, Patricio Gallardo, I asked him what would happen if they used newer, superior DESI data instead of older Sloan Digital Sky Survey data, and he pointed me to a paper from earlier this year,

where they showed that DESI data detects the existence of the kinetic Sunyaev-Zel’dovich effect at a whopping 9-sigma statistical significance, compared to just 6-sigma for the Sloan Digital Sky Survey data that was used in the present paper. This means that a future analysis using already existing DESI data could rule out MOND at the 99.999% confidence level: tantalizingly close to the 5-sigma “gold standard” in detection significance for astrophysics and cosmology.

The results in context

They’ll point to the external field effect and the baryonic tully-fisher relation, saying that “all galaxies follow this, so some analogous effect should also exist on cosmic scales.” Maybe it should, but when they say that, they’re surreptitiously ignoring the ultra-diffuse galaxies that don’t follow that relationship at all, proving that it’s not universal.

They’ll look at the large-scale structure of the Universe or the fluctuations in the cosmic microwave background’s temperature, and say, “well you need some other modification to make MOND work on large cosmic scales anyway,” and then they’ll put forth modifications that – surprise surprise – behave identically to how dark matter behaves under Einstein’s General Relativity.

Or they’ll look at strong independent evidence for dark matter, such as from colliding galaxy clusters, and resort to the ancient argument that “these collision speeds are too fast for General Relativity and a dark matter-rich cosmology to explain,” while completely ignoring the revised collision speeds that we’ve been able to infer since we spotted colliding galaxy cluster MACS J0018.5, which has the collision occurring along the line-of-sight, enabling us to measure cluster collision velocities and the emergence of shocks directly. This system showed that the actual collision speeds between clusters was much lower than previously inferred, because shocks depend on the circumcluster medium in a way that was not understood or appreciated until we made those line-of-sight observations in 2024. They reduce the estimated speeds of cluster collisions, including from the Bullet cluster, to speeds that fall exactly in line with the predictions of a standard dark matter cosmology, while still allowing for the emergence of the shocks that we see.

Concluding thoughts

It’s always fashionable to challenge the scientific consensus, and to favor ideas that would lead to a scientific revolution. But to get there, you have to have data on your side: data that refutes the consensus position and that supports your alternative. The way you do good science is you extract predictions from a theory that lead to observables, and then you use those observables to test the validity of your theory. When your theory passes the test, it remains in the realm of validity; when your theory fails the test, you need to either revise or discard it.

What you mustn’t do – and this is a tantalizing temptation to have to resist – is move the goalposts, or as we say in cosmology, to “invoke the tooth fairy a second time,” to save a fundamentally broken idea. The new test is so powerful because it’s so simple and straightforward: in such a clean cosmic laboratory. It shows how gravity works at low accelerations on large cosmic scales: where the density in any location departs only slightly (by much less than 100%) from the average cosmic density. It shows that gravity still obeys a 1 over r squared force law, not the alternative 1 over r force law that’s at the core of modified Newtonian dynamics. And it shows that we can use easily accessible data, from the CMB and large-scale structure, to conduct this test, implying that as better data comes in, from ACT, from the South Pole Telescope, and from the Simons Observatory on the CMB side, and from SPHEREx, Euclid, Nancy Roman Telescope, and Vera Rubin Telescope on the large-scale structure side, we’ll be able to conduct an even more stringent version of this test. It’s a wonderful illustration of how science works and progresses, if only we can bring ourselves to accept the conclusions that the data is unambiguously telling us.

This article is featured on Big Think.