How has cosmology changed from 2000 to 2025?

Everywhere you look these days, you’re likely to hear declarations that cosmology — the science of what the Universe is made of, where it came from, how it got to be the way it is today, and what its ultimate fate will be — is in crisis. The Hubble tension, or the controversial fact that different methods of measuring how fast the Universe is expanding give different, mutually inconsistent answers, has gotten so significant that some now call it the Hubble crisis. The DESI survey, designed to measure the evolution of large-scale structure in the Universe, indicates a preference for evolving, rather than constant, dark energy, as well as a preference for (unphysically) negative neutrino masses. And the CMB, often considered the “rock” of 21st century cosmology, stands in disagreement with both supernova and large-scale structure data.

Is the data now compelling enough that we have to abandon our notion of a standard dark energy and dark matter-dominated Universe? Is this truly a crisis for ΛCDM cosmology: something that was only pieced together just before the turn of the century? It’s worth taking a long-term perspective here, and to look at just what we’ve learned over the last 25 years, including what supports and what threatens to upend our current way of viewing the Universe. Cosmology has come a long way so far in the 21st century, but there’s a lot farther to go as far as figuring out what’s really going on in the Universe.

Measuring back in time and distance (to the left of “today”) can inform how the Universe will evolve and accelerate/decelerate far into the future. By linking the expansion rate to the matter-and-energy contents of the Universe and measuring the expansion rate, we can come up with an estimate for the amount of time that’s passed since the start of the hot Big Bang. The supernova data in the late 1990s was the first set of data to indicate that we lived in a dark energy-rich Universe, rather than a matter-and-radiation dominated one; the data points, to the left of “today,” clearly drift from the standard “decelerating” scenario that had held sway through most of the 20th century.

25 years ago, we were just assembling our modern picture of the Universe. We had:

• CMB measurements from COBE, BOOMERanG and Maxima, which indicated that the Universe was flat, or that the total sum of all the different types of energy within it equaled ~100%, with no spatial curvature.
• Large-scale structure data from surveys like PSCz and the 2dF galaxy redshift survey, which taught us that there is a large amount of cold mass/matter in the Universe, but only around ~30% of the total energy present.
• And supernova data from the High-z supernova search team and the supernova cosmology project, which — unlike matter — would cause the Universe’s expansion to speed up, rather than slow down, over time.

We also had data about neutrinos, showing there were three species and that those species oscillated into one another, indicating their massive nature. However, they couldn’t be the dark matter, as it would be hot, not cold. We knew how much total normal matter was present in the Universe from Big Bang nucleosynthesis and the abundance of the light elements: around 5% of the total, maybe a little more or less, but nowhere near the 30% that large-scale structure showed.

And then, right at the start of the 21st century, we got a key measurement at long last: the results of the Hubble Space Telescope’s key project, the one that finally measured the expansion rate of the Universe.

Before converging on a value of ~71 km/s/Mpc, values for the modern-day Hubble expansion rate underwent an enormous number of changes, as big discoveries such as the existence of two types of Cepheids, an understanding of peculiar velocities, calibration issues and assumptions over the properties of distance indicators represented real, physical issues whose resolution resulted in a better understanding of the astrophysics governing the Universe. In more recent times, two separate methods appear to yield incompatible results, with one group preferring ~67 km/s/Mpc and the other preferring ~73 km/s/Mpc for the expansion rate.

For decades prior to the HST key project’s results, cosmologists were split into two camps:

• a camp arguing for a fast Hubble constant, like ~90-100 km/s/Mpc, giving a young, critical density Universe,
• and a camp arguing for a low Hubble constant, like ~50-55 km/s/Mpc, giving an old, underdense Universe.

The results of the HST key project showed that both groups were wrong: the Universe was expanding at a rate of ~72 km/s/Mpc, with an uncertainty of just ~10%. Combining that result with CMB, large-scale structure, and supernova data led to an astonishing picture for our Universe: one that would have shocked astrophysicists at any point throughout the 20th century. That picture was that our Universe:

• was dominated by dark energy, some form of energy that caused the Universe’s expansion to accelerate and was consistent with a cosmological constant, to the tune of about ~70% of the cosmic energy budget,
• was sub-dominated by dark matter, which clumps and clusters like normal matter but doesn’t collide with or interact with itself, with normal matter, or with light, contributing ~25% to the cosmic energy budget,
• and where normal matter — including stuff made of protons, neutrons, and electrons, but also including a small amount of massive neutrinos (and antineutrinos) plus photons — makes up just ~5% of our Universe.

Three different types of measurements, distant stars and galaxies (from supernovae), the large-scale structure of the Universe (from BAO), and the fluctuations in the CMB, tell us the expansion history of the Universe and its composition. Constraints on the total matter content (normal+dark, x-axis) and dark energy density (y-axis) from three independent sources: supernovae, the CMB (cosmic microwave background) and BAO (which is a wiggly feature seen in the correlations of large-scale structure).

This picture became known as our “vanilla” ΛCDM cosmology, or our concordance cosmology, and the 2000s and early 2010s cemented this as the “standard model” of cosmology that seemed to fit the full suite of data. First the CMB data from WMAP came in, showing that not only was the Universe spatially flat, but that the acoustic peaks seen on smaller angular scales allowed us to measure the dark matter to normal matter ratio spectacularly: it was about 5-to-1. Large-scale structure data from the Sloan Digital Sky Survey (SDSS) began pouring in, showing features like baryon acoustic oscillations and again revealing that 5-to-1 dark matter to normal matter ratio.

Supernova data improved, allowing us to rule out versions of dark energy that were significantly different from a cosmological constant: domain walls and accelerating “Big Rip” dark energy were ruled out. And, perhaps most spectacularly, we acquired data about colliding galaxy clusters in various stages of collision that showed:

• where the starlight was coming from,
• where the X-rays (arising from heated gas/plasma) was coming from,
• and where the gravitational signal, from gravitational lensing, was coming from.

Critically, this data showed that galaxy clusters, when they collided, saw the majority of the normal matter within them slowed down and emitted X-rays, but that the starlight and the gravitational signal both passed right through, showcasing a separation between the gravitational signal (i.e., dark matter) and the location of the normal matter (i.e., X-rays). Dark matter was real, outmassed normal matter by 5-to-1, and couldn’t be made of any normal “stuff.”

The X-ray (pink) and overall matter (blue) maps of various colliding galaxy clusters show a clear separation between normal matter and gravitational effects, some of the strongest evidence for dark matter. The X-rays come in two varieties, soft (lower-energy) and hard (higher-energy), where galaxy collisions can create temperatures ranging from several hundreds of thousands of degrees up to ~100 million K. Meanwhile, the fact that the gravitational effects (in blue) are displaced from the location of the mass from the normal matter (pink) shows that dark matter must be present. Without dark matter, these observations (along with many others) cannot be sufficiently explained.

It seemed like our concordance picture of the Universe, our “standard model” of cosmology with a cosmological constant for dark energy (Λ) and with cold dark matter (CDM) dominating the rest, with our “normal stuff” making up just ~5% of the cosmic energy budget, was succeeding in every possible avenue. Our measurements from the next CMB satellite — Planck, the successor to WMAP — seemed to only refine this picture. With a superior map of galactic dust and other foregrounds, the best ever, and with better angular resolution and more frequency bands to measure the Big Bang’s leftover glow in than ever before, we started pinning down these figures to even greater precision.

• The total baryon (proton+neutron) density in the Universe was 4.9% of the critical density, with an uncertainty of just ±0.1%.
• The dark energy density was 68% and the dark matter density was 27% of critical, with uncertainties at the 2-3% level only.
• The Universe has an age of 13.81 billion years as marked from the start of the hot Big Bang, with an uncertainty of only about ±100 million years on it.
• And the expansion rate of the Universe, the long-sought Hubble constant (or the Hubble parameter as measured today) was determined exquisitely: to be 67 km/s/Mpc, with an uncertainty of just ±1 km/s/Mpc on that figure.

The fluctuations in the cosmic microwave background were first measured accurately by COBE in the 1990s, then more accurately by WMAP in the 2000s and Planck (above) in the 2010s. This image encodes a huge amount of information about the early Universe, including its composition, age, and history. The fluctuations are only tens to hundreds of microkelvin in magnitude. On large cosmic scales, the error bars are very large, as only a few data points exist, highlighting a large inherent uncertainty.

At last, we had a full, complete, self-consistent picture of how the Universe worked: what it was made of, how it evolved from the distant past until today, what it’s ultimate fate would be (a heat death, dominated by dark energy), and even where it came from if you folded in all the observational evidence that had come in in support of cosmic inflation preceding and setting up the hot Big Bang. With the exquisite measurements we had of the CMB, it seemed like the next stages — getting comparably exquisite data concerning the Universe’s large-scale structure and expansion rate directly (from high-redshift supernovae) — would pin down these cosmic parameters further, and was an inevitability.

But there’s an important caveat here: as scientists, we cannot assume we know what the end result, or the conclusion, is going to be before we make the critical measurements. We have to perform our analyses honestly and in an unbiased fashion, irrespective of our preconceived prejudices about whatever outcome(s) we might expect or anticipate. Put more generally, if you want to know something about the Universe, you have to ask the Universe questions about itself in a way that will compel it to give up its answers, and then you, the scientist, have to listen to what those answers are and interpret it, as best you can, in the context of everything else we’ve already learned.

And that is precisely where modern cosmology starts to get into trouble.

The construction of the cosmic distance ladder involves going from our Solar System to the stars to nearby galaxies to distant ones. Each “step” carries along its own uncertainties, especially the steps where the different “rungs” of the ladder connect. However, recent improvements in the SH0ES distance ladder (parallax + Cepheids + type Ia supernovae) have demonstrated how robust its results are.

Starting in the mid-2010s, the first hint of a crack in this picture began to show. Sure, the CMB had pinned down the expansion rate, at present, to be ~67 km/s/Mpc, with an uncertainty of just ±1 km/s/Mpc on that figure, which seemed to agree with the original HST key project’s results of ~72 km/s/Mpc, even though that had a larger uncertainty of ±7-8 km/s/Mpc on it. With more data, better calibration, and reduced errors, however, people using the key project’s original method — the distance ladder method — were now coming up with superior data, and an independent way of measuring the expansion rate of the Universe.

Instead of starting at the start of the hot Big Bang and evolving the Universe forward until the present day, which is what CMB measurements (and later, baryon acoustic oscillation studies from large-scale structure surveys) did, yielding ~67 km/s/Mpc, supernova astronomers using the distance ladder method were starting here, locally, and by:

• measuring variable stars in our Milky Way with parallax,
• measuring those same types of variable stars in nearby galaxies,
• then measuring type Ia supernovae in those same galaxies,
• and finally by “jumping” to type Ia supernovae all throughout the Universe,

they could also, independently, measure the expansion rate of the Universe. What they were finding, consistently, were higher values than the CMB method: of between 72 and 74 km/s/Mpc, at the same time their errors were dropping, to just ±1-2 km/s/Mpc in that very method.

A series of different groups seeking to measure the expansion rate of the Universe, along with their color-coded results. Note how there’s a large discrepancy between early-time (top two) and late-time (other) results, with the error bars being much larger on each of the late-time options. Although these two classes of measurements give incompatible results, no one knows the resolution to why the Universe appears to expand differently dependent on the method used to measure the expansion.

This led to what became known as the Hubble tension, which is now so significant — at greater than 5-sigma statistical significance — that it’s been upgraded by many to the “Hubble crisis.” It’s not really a crisis or a tension in the expansion rate of the Universe, however, as it is a crisis or tension in how these two very good data sets, which should be pointing to the same conclusion, are instead pointing to mutually incompatible results as one another. If we only had one of these individual results, on its own, we wouldn’t be particularly worried. But to have them both in hand, and to have the “distance ladder” method give consistent results even when we use different indicators for our method of assembling the distance ladder, indicates that all the puzzle pieces we have don’t fit together. Recent superior data about our local Universe, from the ESA’s Gaia mission, has only reinforced and strengthened this tension.

It could have meant a lot of things. In cosmology, there’s a very tight, corresponding relationship between how fast the Universe expands (and how the expansion rate changes with time) and what the various ratios are of the different forms of energy present within it. These two measurements of the expansion rate being incompatible could mean many things:

• perhaps the Universe isn’t made of what we think it’s made of,
• perhaps its energy contents change over time, such as through decays or phase transitions,
• perhaps dark energy, dark matter, or something else is evolving,
• or, perhaps, everyone is just a “little bit wrong” and the true answer lies somewhere in between the results of these differing groups/methods.

Yet, as much as you’d want to bet on that last option, that’s not where the data is pointing. And when we fold in the latest modern results about large-scale structure in the Universe, the situation worsens considerably.

This slice of the DESI data maps celestial objects from Earth (center) to billions of light years away. Among the objects are nearby bright galaxies (yellow), luminous red galaxies (orange), emission-line galaxies (blue), and quasars (green). The large-scale structure of the universe is visible in the inset image, which shows the densest survey region and represents less than 0.1% of the DESI survey’s total volume.

With the latest results from DESI, our best, most comprehensive view of the Universe’s large-scale structure to date (with over 15 million galaxies mapped in three-dimensional space), we now have a third high-quality suite of data: that measures the baryon acoustic oscillation feature more precisely than ever before, including how it evolves over time. DESI’s results, on their own, don’t indicate anything particularly weird: they measure the BAO signal and get something that’s generally consistent with our concordance picture.

That is, until you try combining it with other data sets to make one consistent picture. As I wrote just recently (in March) about these latest results:

• When you look at the DESI data alone, it only favors evolving dark energy over the standard ΛCDM model by less than 2-sigma significance: what we call “negligible evidence.”
• When you look at the DESI data combined with Planck CMB data, there’s a 3.1-sigma preference for evolving dark energy over ΛCDM, which is suggestive, but not conclusive.
• But then, if you add in the supernova data set, your significance for evolving dark energy as compared with a cosmological constant can either increase or decrease depending on which set you use. If you add in Union or DESY data, the significance increases to 3.8-sigma or 4.2-sigma, respectively, but if you add in Pantheon+ data instead, the significance decreases to just 2.8-sigma.

In other words, when you add in the third high-quality type of data set, you’ll find that any two of them, in combination, point away from the entire picture adding up. Add all three data sets together, and you simply can’t make one picture that’s fully consistent with our concordance model.

This figure, from the DESI collaboration’s second data release’s results paper, shows the different values of the matter density that are preferred by six different data sets: DESI’s first and second releases, the CMB, and the supernova samples of Pantheon+, Union, and DESY5. Note that BAO and supernova data sets are not really compatible with one another.

After 25 years of our concordance model holding up to scrutiny and successfully explaining all of the data we’ve acquired, we’ve now come to a point where that doesn’t appear to be true any longer. Three different high-quality data sets — from the CMB, from large-scale structure, and from supernova data — are mutually incompatible. Measuring the expansion of the Universe, if we do it by starting here and looking out or if we do it by starting at the Big Bang and coming forward, gives mutually incompatible results between these two methods. And if we insist that “this is what the Universe is made of” and just follow the equations, we see that the full suite of data simply cannot be consistent with the simple ΛCDM picture that’s held sway for so long.

So what does it all mean? Is dark energy evolving or did an early form of it exist and then decay early on? Is dark matter decaying, or is something decaying into dark matter? Was there a phase transition early on that we haven’t identified or reckoned with? Is there an extra type of neutrino, and is the fact that DESI data seems to favor negative neutrino masses telling us anything meaningful?

The most exciting aspect of the story is that these are all possibilities, and that there are many others that haven’t even been fully explored at present. We have a hint that something is wrong with our standard picture, and the next step will be acquiring more and better data to help us pin down precisely where our standard picture fails and by how much. With ESA’s Euclid, the NSF’s Vera Rubin Observatory, and NASA’s Nancy Roman Telescope (plus Caltech’s SphereX mission), we’re going to measure baryon acoustic oscillations, and acquire new type Ia supernovae, as never before. Perhaps we’ll gain new insights into refining our standard ΛCDM picture, or perhaps we’ve find out exactly how and where it fails. Either way, noticing these cracks in our consensus picture is potentially a harbinger of a new scientific revolution. The only way we’ll find out is to look, with better tools, techniques, and data than ever before.