Sign up for the Starts With a Bang newsletter
Travel the universe with Dr. Ethan Siegel as he answers the biggest questions of all.
In most scientific fields, one of the most exciting things we can encounter is data — high-quality, robust data — that doesn’t align neatly with the expectations of our currently leading theories. Since the late 1990s, our leading theory of the Universe has been known as either the “ΛCDM” or “concordance” cosmology, where our Universe:
began with a period of cosmic inflation that preceded and set up the hot Big Bang,
then the hot Big Bang occurred, creating a dense, hot, mostly uniform Universe,
containing normal matter and radiation, but dominated by dark matter and dark energy,
which gravitated, expanded, and cooled,
forming the light elements, neutral atoms, stars, galaxies, and black holes,
and giving rise to the Universe as we observe it today.
Today, that concordance picture looks like a Universe presently expanding at a rate of ~70 km/s/Mpc, made of around 70% dark energy as a cosmological constant and 25% dark matter as non-interacting cold dark matter (with the remaining 5% being “normal” stuff), and forming black holes, stars, and galaxies from the normal matter in our Universe, under the gradual but relentless influence of gravity.
However, many observations now compel us to question one or more of these facets of the Universe. When we combine cosmic microwave background (CMB), large-scale structure, supernova, and ultra-distant JWST data together, all the pieces don’t necessarily fit together so nicely. It’s possible that dark energy is evolving, sure, as many have hypothesized. But a new theory by Prof. Jonathan Tan, hypothesizing a novel, early population of supermassive stars, offers a fascinating alternative possibility. Here’s what it’s all about.
This animation of DESI’s 3D map of the large-scale structure in the Universe, the largest such map to date, was created with the intention of studying dark energy and its possible evolution. However, although they found evidence for dark energy evolving, that’s likely due to the assumption that it’s dark energy’s evolution that’s causing the discrepancies in the data compared to our standard cosmological model. This is not necessarily the case.
Credit: DESI Collaboration/DOE/KPNO/NOIRLab/NSF/AURA/R. Proctor
To start off, we have to take stock of the observations that we have, and the fact that they don’t all point to a single, consistent picture of the Universe that matches up with our standard ΛCDM concordance cosmology. We have:
CMB data, such as from the Planck satellite, which indicates a low expansion rate (H0 = 67 km/s/Mpc), a Universe that’s 68% dark energy and 27% dark matter, and a single, late period of gradual reionization of the neutral atoms that formed (where the optical depth, τ, is constrained to be 0.054 ± 0.007),
high-redshift distance indicators, such as Type Ia supernovae (but also including many others) that favor a greater expansion rate (H0 = 73 km/s/Mpc), with less dark energy but more dark matter at 66% and 29%, respectively,
large-scale structure data, such as from DESI, that can be consistent with either expansion rate but that points to more dark energy (71%) and less dark matter (24%) than either of the previous options, and that furthermore indicates that dark energy is evolving by weakening at late cosmic times (and puzzlingly favors negative neutrino masses, which is unphysical),
and data about the ultra-distant Universe from JWST, which points to large, massive, early black holes — black holes that were more massive than could have been formed by normal stars alone — and bright, early galaxies from as early as just ~280 million years after the Big Bang.
Putting all of these pieces of data together, rather than pointing to one consistent picture of the Universe and its parameters, instead indicates a tension where they, plus our view of standard ΛCDM, can’t all be right together at once.
This image shows 15 of the 341 hitherto identified “little red dot” galaxies discovered in the distant Universe by JWST. These galaxies all exhibit similar features, but only exist very early on in cosmic history; there are no known examples of such galaxies close by or at late times. All of them are quite massive, but some are compact while others are extended, and some show evidence for AGN activity while others do not. The ones exhibiting AGN activity all have heavy inferred supermassive black hole masses.
Credit: D. Kocevski et al., Astrophysical Journal Letters accepted/arXiv:2404.03576, 2025
Sure, it’s possible that dark energy is evolving, as many have suggested, and that could help resolve the tensions that arise from different lines of observation supporting different expansion rates and different dark matter and dark energy ratios. However, many puzzles would still remain, such as how the supermassive black holes — found in many early galaxies, including the “little red dot” galaxies shown above — get so massive so early on, or why our current “best fits” to the data indicate negative neutrino masses, which cannot physically be the case.
It’s also possible that there are other non-standard (i.e., exotic, with novel or more complex ingredients) cosmic scenarios that could explain these observations, each with their own pros and cons. It’s possible that something happens early on, between when the fluctuations in the CMB are seen (at a cosmic age of 380,000 years) and when the first luminous objects in the Universe are seen (at a cosmic age of 280 million years), to enhance the formation of structure while changing the energy contents of the Universe. This could include some sort of novel, exotic form of mass or energy, a non-standard form of dark matter or dark energy that decays or otherwise interacts, or a modification to one or more laws of physics.
However, if we adhere to the standard scenario of ΛCDM, the only non-exotic way to form these early, very massive black holes, if not from normal, early stars, is through the direct collapse of cold gas.
This snippet from a supercomputer simulation shows just over 1 million years of cosmic evolution between two converging cold streams of gas. In this short interval, just a little over 100 million years after the Big Bang, clumps of matter grow to possess individual stars containing tens of thousands of solar masses each in the densest regions, and could lead to direct collapse black holes of an estimated ~40,000 solar masses. This could provide the needed seeds for the Universe’s earliest, most massive black holes, as well as the earliest seeds for the formation of stars and the growth of galactic structures.
Credit: M.A. Latif et al., Nature, 2022
The idea behind this direct collapse isn’t necessarily simple, but it is straightforward. It basically states that:
the Universe is born with slight overdensities seeded (by inflation) across it on all scales,
those overdense regions gravitate, resulting in the “cold spots” in the CMB because there’s more matter than average in them,
with the coldest cold spots corresponding to the greatest-magnitude overdensities,
which will gravitationally collapse first, pulling normal and dark matter into them and growing in a runaway fashion,
where the normal matter that efficiently cools creates cold streams of gas,
and where these cold streams of gas converge the fastest and earliest, as shown in the simulation above,
generating supermassive black hole seeds of between 10,000 and 40,000 solar masses that arise early on in cosmic history, well before the earliest JWST galaxies have ever been spotted.
However, when considering this scenario, we have to confront the big question that separates the hard sciences from all other endeavors: can this idea explain the black holes that we actually observe in a quantitative fashion? In other words, can it predict the right number of early, very massive black holes of the right masses to be consistent with what JWST and other telescopes actually observe are present in our Universe?
If you begin with an initial, seed black hole when the Universe was only 100 million years old, there’s a limit to the rate at which it can grow: the Eddington limit. To be consistent with what’s observed, seeds of several tens-of-thousands of solar masses, at least, must arise early on and grow rapidly thereafter, or otherwise, a slightly larger set of seeds at slightly later times could do it. The lack of lower-mass black holes at early epochs suggest a large set of similar-mass seeds at some early time.
Credit: F. Wang, image taken at AAS237
According to several recent studies, the answer seems to be “perhaps not.” There are two major problems. First, the black holes that are predicted do arise sufficiently early — at between 100-150 million years of cosmic age (or a redshift of between 23 and 30) — but tend to have masses of around 10,000 or maybe a few ten-thousand solar masses. For comparison, the observed black holes, although they only appear at late times, seem to require seeds that are a bit larger, of around 100,000 solar masses or so. And second, the black holes formed by direct collapse should also come in lower-mass varieties, of a few thousands or a few hundreds of solar masses at that time, which should lead to a population of smaller black holes found alongside the largest ones.
However, it seems like the galaxies that we observe from the first ~1 billion years of cosmic history, when we focus on the ones that do contain these early, massive black holes, all require the same large, 100,000-ish solar mass seeds in order to form the black holes that we observe at these later cosmic times. There is no spectrum of black holes that includes lower-mass seeds, and even under the accepted direct collapse scenarios, the large, massive seeds that are produced are still a little bit too small to be fully consistent with the observed black holes that appear in JWST, Hubble, and Chandra data, among others.
When the data from a variety of “little red dot” galaxies are broken up into their stellar mass component versus the component arising from an active supermassive black hole, the mass ratios of the galaxy’s total stellar mass compared with the supermassive black hole’s mass can be determined. Many, and perhaps even most, of these black holes are found to be significantly overmassive: at much more than 0.1% of the mass of the stellar component.
Credit: D. Kocevski et al., Astrophysical Journal Letters accepted/arXiv:2404.03576, 2025
There have been plenty of other scenarios proposed for how these supermassive black holes could have arisen, and all of them are even more problematic than the direct collapse scenario.
There’s the idea that the seed black holes could have arisen from the very first stars that formed, lived, and died in the Universe. Unfortunately, these stars, even taking into account that they form early and are made from hydrogen and helium alone (and hence will be more massive than today’s stars), cap out at about ~1000 solar masses only, meaning that the seed black holes that will result from them are even less massive than the direct collapse black holes formed from streams of gas.
There’s also the idea that there could have been a population of primordial black holes, or black holes that formed shortly after the Big Bang itself. Although this idea is impossible to rule out at present, it’s an ad hoc idea that requires a new mechanism (that won’t mess up either structure formation or the CMB) that leads to a remarkably powerful spike in the matter power spectrum, enhancing structure on one particular scale — the scale of supermassive black hole seeds — by factors of 100,000+, without affecting either larger or smaller scales.
Moreover, neither of these scenarios, nor the now-default direct collapse scenario, can do anything to either alleviate the Hubble tension or to explain the puzzles associated with large-scale structure observation and its implications for dark energy or for neutrino masses.
The overdense regions that the Universe was born with grow and grow over time, but are limited in their growth by the initial small magnitudes of the overdensities, the cosmic scale on which the overdensities are found (and the time it takes the gravitational force to traverse them), and also by the presence of radiation that’s still energetic, which prevents structure from growing any faster. The first stars form from the normal matter within these growing, overdense regions, but exotic dark matter scenarios can lead to earlier, more massive stars.
Credit: Aaron Smith/TACC/UT-Austin
That’s where the new theoretical idea of Jonathan Tan comes in: proposing that there’s an early population of very massive stars that exist even before the “main” population of pristine stars comes into existence. Instead of a very few select stars forming early, just 50-100 million years after the Big Bang, and then with star-formation gradually accelerating (including with an acceleration driven by these young black hole seeds) to form more and more stars over the next few hundred-million years, this new idea posits the existence of an epoch known as “The Flash,” where a substantial number of supermassive stars formed between a cosmic age of 100-180 million years and produced an enormous number of ionizing photons.
Instead of having just:
Population I stars, which are late-time, metal-rich stars formed out of highly processed material like our Sun,
Population II stars, which are metal-poor stars formed out of more pristine material that’s been less enriched by previous star-formation, such as those found in most old globular clusters,
and Population III stars, which are the (still theoretical, not yet observed) first stars formed from the leftover, pristine material from the Big Bang,
this new theory posits an additional stellar population: what they call Population III.1, an idea first put forth in 2019. Theorized to form in dark matter minihaloes, they should create seed black holes of between 100,000 and 1 million solar masses: the first collapsed structures to form in their local region of the Universe.
This illustration shows a computer simulation of the ionized bubble of gas that surrounds a hypothetical Population III.1 star, which would be the progenitor of early, supermassive black hole seeds formed long before the earliest known JWST galaxy. It’s possible that all supermassive black holes formed in this way, leading to a “flash ionization” of much of the Universe at early cosmic times.
Credit: M. Sanati (Chalmers Univ. & Oxford Univ.), J. Tan (U. Virginia & Chalmers Univ.)
The idea behind their formation is that, sure, these dark matter minihaloes form just as they do in a standard ΛCDM scenario, pulling that normal, pristine matter into them. However, if dark matter isn’t just boring old non-interacting cold dark matter, but instead has a self-interaction or can self-annihilate (as suggested here or here), then you wouldn’t just form a “normal” Population III star inside: of tens or a few hundreds of solar masses, maybe all the way up to 1000 solar masses. Instead, with this tweak to dark matter, the early protostar that forms will instead remain in a large, relatively cool state for a long time, which can grow the protostar up to 100,000 solar masses or even more.
A colossal protostar like that will eventually “overheat” on the interior, produce pairs of electrons-and-positrons, and collapse, where it can collapse to a black hole. But prior to that collapse, these stars can exist for a long time and produce enormous quantities of ionizing photons: enough to substantially ionize a few percent of all the atoms in the Universe at that time. This scenario would explain why:
there are so many supermassive black holes so early on,
why they’re all heavy, and why there are no “small seed” black holes alongside the most massive ones,
and can potentially explain the puzzles brought up by CMB, supernova, and large-scale structure data together, including the Hubble tension and the bizarre appearance of what appears to be an evolution in dark energy from DESI data.
This diagram illustrates the pair production process that occur inside stars of very great masses: when high-energy photons spontaneously convert into electron-positron pairs. When this occurs, a tremendous drop in internal pressure can occur, causing a star to collapse, either leading to a pair-instability supernova or, for masses that are too great, direct collapse to a black hole. Scenarios with exotic dark matter, in principle, can create early protostars of ~100,000 solar masses or more, leading to massive black hole seeds even early on.
Credit: NASA/CXC/M. Weiss
While we’re unlikely to be able to directly detect the properties of dark matter anytime soon, this does provide a series of potential observable signatures to look for:
a signature of 21-cm absorption found earlier than the most distant JWST galaxies (which was tentatively reported in a controversial 2018 study),
an enhancement of the abundances of molecular hydrogen and deuterated hydrogen gas in the “normal” Population III stars that arise afterwards (which still haven’t been spotted and measured),
and, importantly, a greater optical depth (τ, which was measured by Planck to be 0.054 ± 0.007) than has been reported so far.
The first two elements, on their own, aren’t of particularly great interest, as the once reported 21-cm absorption feature in 2018 was shown to be impossible four years later. However, three independent research teams have all shown that the optical depth of the Universe, if it were instead higher than Planck or other CMB results indicated (at about τ = 0.09 or so), could potentially resolve the three puzzles of the Hubble tension, evolving dark energy, and negative neutrino masses. Remarkably, this new scenario of early, Population III.1 stars can produce an early period of reionization, increasing the optical depth parameter to exactly this value, and then allowing those ionized electrons to then recombine with their parent nuclei to form neutral atoms once again prior to the “standard” epoch of reionization that occurs later.
This graphs shows the optical depth of the Universe (y-axis) as a function of redshift (x-axis). The blue curve shows the standard structure formation scenario, with the lower gray bar showing the constraints from Planck. The upper gray bar shows scenarios that could resolve the Hubble tension, evolving dark energy, and negative neutrino masses, with the two red curves showing two different flash ionization scenarios.
Credit: J.C. Tan, Astrophysical Journal Letters accepted/arXiv:2506.18490, 2025
It’s a fascinating idea that, importantly, will be testable as our technology for observing the Universe continues to improve. More senior cosmologists will remember that, in the early days of WMAP (the cutting-edge CMB satellite that was the predecessor to Planck), the optical depth was initially reported to be a whopping τ = 0.17, which is about triple the currently best-measured value from Planck. The reason behind it, as the team later identified, was that there was an incorrect subtraction of the galactic foregrounds, and in particular of the emissions of polarized dust in our own galactic plane. If it turns out that the Planck data is still not perfectly calibrated to our own foregrounds, particularly from the E-mode polarizations of our galactic foregrounds, a τ of 0.09 or so is still very much in play.
In our quest to understand the Universe, it’s important to hold in our heads both the standard consensus scenario — which only became that way because of its tremendous success in fitting the full suite of earlier observations — as well as:
all the ways that observations are hinting that our current picture may not fit the data,
and the various theoretical ideas that could lead to differing predictions from the mainstream consensus.
It is only by considering the mainstream theory and its alternatives, and confronting them with high-quality data about the Universe, that we have any hope of advancing our understanding of reality. This novel scenario, of flash ionization in the early Universe, may turn out to be, like most new theoretical ideas, a red herring that only takes us to a possible world that never was. But reality can be very strange, and this wild idea may yet turn out to describe the Universe more successfully than even the standard ΛCDM cosmology that’s brought us up to this point. The key, as always, will be in acquiring more data, and putting the question of what the Universe is like to the only source that can provide the answer: the Universe itself.
Sign up for the Starts With a Bang newsletter
Travel the universe with Dr. Ethan Siegel as he answers the biggest questions of all.