The Ekpyrotic Universe: Was the Big Bang a Collision?

In 2001, four physicists published a paper in Physical Review D with a title that reads like science fiction and works like mathematics: “The Ekpyrotic Universe: Colliding Branes and the Origin of the Hot Big Bang.” The authors were Justin Khoury, Burt Ovrut, Paul Steinhardt, and Neil Turok, and the claim they made was not modest. The ekpyrotic universe they described did not begin with a bang. It began with a crash.

In their picture, the hot, dense early cosmos was not born from a single point of infinite density. It was lit when two vast, higher-dimensional surfaces, drifting toward each other across a hidden dimension, collided. Ovrut put the radicalism plainly in an interview the same year. “It’s a very radical idea we have,” he said, adding that “time zero was just a marker,” and that the universe “really existed long before that.”

The word ekpyrotic comes from the Greek ekpyrosis, meaning out of fire, borrowed from the ancient Stoics who imagined a cosmos that periodically dissolved into flame and was reborn. To understand why anyone would revive a Stoic metaphor inside twenty-first-century cosmology, you have to start with the problem the model was built to solve, and with the theory it was built to challenge.

The two problems every origin story must solve

Look at the cosmic microwave background, the faint glow of radiation released about 380,000 years after the Big Bang, and you find something that should not be there. Point an instrument at one edge of the sky, then at the opposite edge, and the two regions agree in temperature to within a few parts in 100,000. Yet in the standard Hot Big Bang those regions have never been in causal contact. Light has not had time since the beginning to travel from one to the other. They are strangers that somehow share a temperature. This is the horizon problem.

There is a second puzzle beside it. The geometry of the universe, as far as we can measure, is flat to extraordinary precision. But flatness is unstable. Any small initial curvature should have grown enormously as the universe aged, so for space to be as flat as we see it now, it had to begin almost perfectly flat. This is the flatness problem.

Inflation’s answer

For more than forty years the mainstream answer to both has been cosmic inflation. In 1981 the physicist Alan Guth proposed that in the first unimaginably small fraction of a second, around 10⁻³⁵ seconds after the beginning, the universe underwent a burst of exponential expansion, stretching a patch smaller than a proton to cosmic scale. Inflation solves the horizon problem because everything we see grew from one tiny region that had been in contact before the stretching began. It solves the flatness problem the way the surface of an inflating balloon looks flatter the larger it grows.

The theory has been a genuine success. It predicts a nearly scale-invariant spectrum of density ripples, and the Planck satellite measured the tilt of that spectrum, the scalar spectral index, at 0.9649 ± 0.0042, close to one but not exactly one, which is what inflation expects. The collaboration reported that the data were fully consistent with the predictions of inflation.

But inflation carries a soft spot its own architects admit. It does not explain its own beginning, and it does not remove the initial singularity; it pushes the singularity backward behind a curtain and leaves it there. The astronomer Mario Livio, writing in 2001, put it bluntly: “Inflation theory has not been tested observationally in any detail.” That sentence is the gap into which the ekpyrotic model stepped.

Inflation smooths the universe by stretching it after the beginning. The ekpyrotic model proposes the opposite: that the universe was smoothed by squeezing, before any beginning at all.

A universe painted on a membrane

The ekpyrotic model is built from string theory and its larger framework, M-theory, which propose that reality’s basic constituents are not point particles but tiny vibrating strings living in eleven dimensions rather than four. Most of the extra dimensions are imagined as curled too small to see. But in one version, worked out by Petr Horava and Edward Witten in the mid-1990s, a single extra dimension is special. It is an interval bounded at each end by a wall, and these walls are not metaphors. They are branes, short for membranes: vast sheets stretching across all of space, the boundaries of a higher-dimensional volume the theorists call the bulk.

Picture two enormous parallel sheets separated by a gap along a fifth dimension. Our entire universe, every galaxy and star and atom of your body, is painted onto one of them. We cannot see the gap because we cannot leave the sheet; light and matter are confined to the brane, and only gravity reaches across the bulk between the two.

The distance between the branes is itself a physical quantity with a name: the radion, a scalar field that measures the separation. When the radion is large, the branes are far apart. When it shrinks to zero, they touch. In the ekpyrotic model that touch is everything. The kinetic energy of the approach is converted, in the instant of collision, into a flood of hot radiation and matter spread across the entire brane at once. That flood is what we have always called the Big Bang.

Notice what the geometry buys you. There is no single point and no center from which everything flies outward. Because two flat, parallel sheets strike face to face rather than corner to corner, the fire ignites everywhere at the same moment, and the early universe is uniform for the same reason a struck pane of glass shatters evenly. The very thing the horizon problem struggles to explain is built into the shape of the collision. The science writer Paul Sutter, summarizing the model, described the branes taking “a long, leisurely time drifting toward each other,” more than enough for everything “to settle into quiet agreement before the bounce.”

To make that work, the contraction cannot be ordinary. It has to be governed by an unusually steep field, with an equation of state physicists write as w much greater than one, so that the field’s energy rises faster than anything else as the universe shrinks. This same steepness does a second job. In a contracting universe, tiny differences in how fast it shrinks along different directions tend to grow, a chaos known as the Belinsky-Khalatnikov-Lifshitz instability that long seemed fatal to any bouncing cosmology. The steep ekpyrotic field outpaces that chaos, its energy climbing as the scale factor falls faster than the disorder can, so the contraction becomes an attractor toward smoothness rather than a descent into mayhem. Remarkably, only about ten to twelve e-folds of contraction are needed to do what inflation requires more than sixty e-folds of expansion to accomplish.

The whole smoothness of the ekpyrotic universe rests on the steepness of a single field. Make the slope gentler and the chaos wins.

From one bounce to an eternal cycle

A single collision raises an obvious question. If the universe existed before the crash, and if it can bounce once, why not again? In 2002 Steinhardt and Turok answered with a cyclic model, published in Science, that embedded the brane collision in an endless sequence: contraction, bounce, expansion, and a slow return to contraction, repeating without end. We are not living in the aftermath of the beginning, on this view. We are living in one cycle of an unbounded series, far from any first cause and far from any last.

The early cyclic model had three honest problems, and its builders named all three. The bounce still passed through a singular state where general relativity stops giving answers. The growing anisotropy threatened to shatter the contraction. And entropy, which always increases, seemed to forbid cycles that had run forever, an objection raised against cyclic cosmologies as far back as the 1930s by the physicist Richard Tolman.

In 2019 Anna Ijjas and Paul Steinhardt published “A New Kind of Cyclic Universe” in Physics Letters B, and it answered all three at once. The new model dropped the literal branes and worked in ordinary three-dimensional space with two scalar fields. One drives the ekpyrotic contraction. The other behaves like dark energy, the accelerating expansion discovered in 1998 by the teams of Saul Perlmutter, Brian Schmidt, and Adam Riess, work that won the Nobel Prize in 2011. Here the discovery that embarrassed the standard model becomes the engine of the cycle. The dark-energy phase accelerates the universe for an immense span, somewhere between 10¹⁴ and 10¹⁷ years, diluting the debris of the previous cycle across an enormous volume before winding down so the next contraction can begin. The acceleration we observe today is, in this reading, a clock counting down to the next bounce.

The dark energy that astronomers once treated as a cosmic embarrassment is, in the cyclic model, the keystone: the slow timer that keeps the branes apart and then, when it fades, triggers the fall.

The bounce itself demanded something ordinary physics forbids. A smooth, non-singular bounce requires briefly violating the Null Energy Condition, the rule that energy density plus pressure stays non-negative, and the fields that usually do this carry runaway instabilities called ghosts. In 2016, Anna Ijjas, Paul Steinhardt, and Abraham Loeb demonstrated a stable, ghost-free bounce using a special class of fields known as Galileons, part of the broader Horndeski family. It was a genuine advance. It is also undeniably exotic, a kind of matter never observed in any laboratory, doing the one thing the standard energy conditions forbid.

The ekpyrotic idea did not arrive without ancestors, and it does not stand without rivals. In the early 1990s, Gabriele Veneziano and Maurizio Gasperini proposed a Pre-Big Bang scenario, also rooted in string theory, in which the universe began in a cold, nearly empty vacuum and ran through a long history before the hot phase. The ekpyrotic authors acknowledged the kinship while noting that their own model avoids inflation or deflation altogether. A sharper contrast comes from Roger Penrose, whose Conformal Cyclic Cosmology, proposed around 2007, also makes the universe repeat, but with entirely different physics: no branes and no bounce, only an infinite sequence of aeons in which the remote future of one universe becomes the Big Bang of the next, after every massive particle has decayed and the cosmos has forgotten its own scale. Both Penrose and the ekpyrotic theorists refuse the single creation event. They disagree about almost everything else.

The modern cyclic model answers its old objections on paper. What it cannot supply is direct evidence that any of its machinery is real.

The missing signal in the oldest light

A cosmological model is judged by whether it reproduces the universe we see, and the decisive test runs through the cosmic microwave background. When the early universe ripples, it ripples in two ways: density ripples, and gravitational waves, ripples in spacetime itself. Inflation, because it stretches space so violently, generically produces a sea of primordial gravitational waves, which leave a faint swirling fingerprint in the polarization of the background called B-mode polarization. The strength of that signal is measured by the tensor-to-scalar ratio, written r. Many inflationary models predicted r around one tenth.

The ekpyrotic model predicts the opposite. Because the universe contracts rather than expands violently, it pumps almost no energy into long-wavelength gravitational waves, and its prediction for r at observable scales is effectively zero. One theory predicts a signal; the other predicts silence. The search for B-modes became a direct contest between two pictures of the beginning.

That search produced one of the sharpest episodes in recent cosmology. In March 2014 the BICEP2 experiment at the South Pole announced it had found the signal, with r near 0.2. Within weeks the claim unraveled. A careful joint analysis with Planck showed the swirling pattern was not primordial at all but polarized dust in our own galaxy. The claim was withdrawn. Since then the limits have only tightened: by 2022 the combined BICEP, Keck, and Planck data placed r below 0.032, with no primordial gravitational waves found. Every year the silence deepens and the upper limit falls closer to the ekpyrotic prediction.

Fairness matters here, because this is exactly where enthusiasm outruns evidence. A non-detection is not a confirmation. Inflation is a broad family, and many of its small-field members predict an r too low to have been seen yet. The honest statement is the one the field actually makes: the data are consistent with ekpyrosis and consistent with low-energy inflation, and we cannot yet tell them apart. That is why the next instruments matter. The Simons Observatory, the proposed CMB-S4 array, and the satellite LiteBIRD aim to measure r to a few thousandths or better, which would either find inflation’s gravitational waves or push the limit so low that the ekpyrotic prediction becomes the more natural reading.

The model has already taken one wound. Its early versions predicted large non-Gaussianity, a particular departure from a perfect bell curve in the primordial ripples, and Planck measured the relevant parameter at 2.7 ± 5.8, consistent with zero. The simplest ekpyrotic models were, in effect, falsified, and had to be revised to fit. Inflation predicted near-zero non-Gaussianity and the data agreed; ekpyrosis predicted large non-Gaussianity, the data disagreed, and the theory was repaired. Both now match the measurement, but they arrived at agreement by different roads.

What does it mean that the cleanest test between the two models is a signal that has refused, for two decades, to appear?

What the cycle cannot escape

Strip the machinery away and the ekpyrotic and cyclic models make one claim larger than any equation: that the cosmos may have no beginning, having always existed, cycling without limit. This is the model’s deepest appeal, and it is worth examining honestly rather than celebrating or dismissing.

The appeal is obvious. A universe with a beginning seems to demand a cause for that beginning, and removing the beginning seems to remove the demand. But look at what the model actually contains. The ekpyrotic universe is not nothing arranging itself into something. It is a vast, intricate physical system: two scalar fields, a steep and precisely shaped potential, a ghost-free bounce, a dark-energy epoch tuned to last and then switch off. The model does not derive these ingredients. It assumes them, at every cycle and at every moment of an infinite past. It explains how an already-existing, already-lawful cosmos moves, not why there is a cosmos at all.

So the oldest question returns, untouched by the length of time. Why is there something rather than nothing? Why these fields, these laws, this steep potential rather than some other? Making the past infinite does not answer those questions; it spreads them across an infinite past. And the fine-tuning that inflation was accused of does not vanish in the cyclic alternative so much as change address, from the flatness of inflation’s potential to the steepness of ekpyrosis’s and the timing of its dark energy. The proponents themselves prize the model’s predictability, writing that “all coarse-grain properties of the universe are deterministically set by the governing classical equations,” with “no quantum runaway that leads to the multiverse effect.” That is a confident claim about how the universe behaves. It is not an account of why the equations exist to govern anything.

This is the question both sides of the debate face, and neither cosmology answers it. Whether one reads the unexplained order as a brute fact, as evidence of design, or as a boundary where physics hands the question to philosophy is a choice the data does not force. What the ekpyrotic universe demonstrates is that even the most radical cosmologies do not make the question disappear. They relocate it.

Even the model’s allies speak with that humility. Alan Guth, the father of inflation, said of the ekpyrotic idea: “I don’t think it’s by any means yet a real rival to inflation, but it is a model well worth pursuing.” Paul Sutter, surveying the whole elegant construction with evident affection, said the thing that haunts the field: it is beautiful, and “it doesn’t work,” at least not yet.

The cyclic universe does not abolish the need for an explanation of existence. It conceals it across an unbounded past, which is a different thing.

Whether or not two branes ever collided, the ekpyrotic universe forces a sharper question than the one it set out to answer. Not how the fire was kindled, but why there is anything to burn. If you find that the standard story of a single beginning leaves a cause hanging in the dark, the cyclic alternative offers a strange consolation and a stranger cost: it removes the first moment, and in doing so it makes the universe older than time, without making it any easier to explain. For more on the deep structure of cosmic time, see our pieces on the block universe and the shape of space itself.