There is one universe, but two ways of measuring how fast it is expanding, and they disagree. When cosmologists read the expansion rate from the cosmic microwave background through the standard ΛCDM model, they find a Hubble constant near 67 kilometres per second per megaparsec. When they build the distance ladder from Cepheid variables to Type Ia supernovae to distant galaxies, the leading measurements arrive near 73. The disagreement has survived years of scrutiny, including new Cepheid observations from the James Webb Space Telescope.

This is the Hubble tension, and it is strange because neither measurement is obviously flawed. The early-universe value is not observed directly. It is inferred, and the inference depends on a physical scale imprinted roughly 380,000 years after the Big Bang. If our physics for that epoch is incomplete, the inferred expansion rate today could be wrong even if every photon has been measured perfectly. The error would not be in the telescope. It would be in the story we use to interpret the light.

One of the most studied proposals is early dark energy: a component that briefly made up perhaps a tenth of the universe's energy shortly before recombination, then vanished almost without trace. The idea is elegant, and it is also uncomfortable. Following it carefully shows how tightly the cosmic microwave background constrains the young universe, and how hard it is to change one piece of cosmic history without disturbing everything connected to it.

Before recombination, photons and baryons formed a tightly coupled plasma. Radiation pressure resisted gravitational compression, so every overdensity launched a sound wave moving at a substantial fraction of the speed of light. When the universe cooled enough for neutral hydrogen to form, photons decoupled and the waves froze in place. The distance a wave could travel by then is the sound horizon, about 147 megaparsecs in comoving units. It became a standard ruler stamped into both the microwave sky and the clustering of galaxies.

What the cosmic microwave background measures with extraordinary precision is not the sound horizon itself. It measures the ruler's angular size on the sky, which Planck pinned down to a few parts in ten thousand. That angle is the ratio of the ruler's physical length to the distance to the last-scattering surface. Fix the angle and you fix the ratio. If the ruler shrinks, the distance must shrink in proportion, and a shorter distance to recombination requires faster expansion in the late universe. That means a higher Hubble constant.

This is where early dark energy enters. The sound horizon is the sound speed divided by the expansion rate, summed over every moment from the Big Bang to recombination. Raise the expansion rate during that window and the waves have less time to travel before the plasma releases its light. The ruler contracts. A reduction of roughly seven to eight percent would bring the microwave-background inference into line with the distance ladder.

When the extra energy appears matters enormously. The sound horizon is most sensitive to the final epochs before recombination, so energy injected near matter–radiation equality, around redshift 3,500, has the greatest leverage. Energy added much earlier barely moves the ruler. Energy added in the form of extra relativistic particles is spread throughout the radiation era. That approach also changes the scale at which photon diffusion smooths small fluctuations, and it conflicts with the fine-scale microwave-background data and with primordial nucleosynthesis.

Analyses such as Knox and Millea's Hubble Hunter's Guide argued that changes before recombination are among the few plausible routes to reconciliation. Changes to late-time expansion are tightly pinned by baryon acoustic oscillations and supernova distances, leaving little room to adjust anything after recombination. Early dark energy is therefore a targeted intervention. It is a brief, carefully timed injection designed to shorten the ruler while leaving the rest of the plasma's physics nearly intact.

Takeaway

The cosmic microwave background never measures the Hubble constant directly; it measures an angle, and the expansion rate follows only once we assume how long the young universe's sound waves were allowed to travel.

The best-known version, developed by Poulin, Smith, Karwal and Kamionkowski, uses a scalar field resembling an axion. In the earliest epochs, cosmic expansion acts like friction and holds the field frozen high on its potential. While frozen, it behaves like a cosmological constant: its energy density stays fixed as radiation and matter thin out. Its share of the total therefore grows steadily, a dormant component quietly gaining influence.

Once the expansion rate falls below the field's effective mass, that friction releases its grip. The field rolls toward its minimum and begins to oscillate. For this to work, the release must happen near matter–radiation equality, and the field's peak contribution must reach roughly ten percent of the total energy. Too little, and the sound horizon barely changes. Too much, and the acoustic peaks of the microwave background deform beyond what the data allow.

The departure is as important as the arrival. After release, the field's energy must fade faster than radiation, or it lingers and distorts later expansion and the growth of structure. The simplest potential fails here, because a field oscillating in it behaves like ordinary matter and fades too slowly. Model builders instead use steeper, axion-like potentials. With the usual choice, the field averages a pressure of half its energy density, and its energy falls as the inverse 4.5 power of the scale factor. Radiation falls only as the inverse fourth power, so the field gets out of the way.

Fluctuations in the field impose further requirements. They travel at an effective sound speed that depends on scale, and they act back on the photon–baryon fluid through gravity. The polarization of the microwave background is sensitive to these details, so smooth stand-ins that only mimic the field's average behaviour often fit the data worse than the field itself. Competing variants include rock 'n' roll fields, acoustic dark energy, new early dark energy triggered by a phase transition, and mechanisms tied to neutrinos becoming non-relativistic.

The deepest discomfort is the coincidence. Why should an otherwise unrelated field wake up at matter–radiation equality, the one moment when it can help? Dark energy today already poses a similar timing puzzle, and early dark energy adds a second one. Efforts to link the trigger to equality itself, or to neutrino masses, are not cosmetic. A model whose parameters must be set by its own purpose explains less than it appears to.

Takeaway

When a model fits only because its parameters sit exactly where the data need them, the right response is curiosity and suspicion in equal measure, and the real question becomes what would make that timing natural.

Early dark energy first looked compelling. Fits combining Planck data with the SH0ES distance-ladder measurement found a peak contribution near ten percent and a Hubble constant around 71 to 72. That result, however, leaned on the SH0ES measurement included in the fit, which pulls the answer toward higher values. Planck data analysed on their own show no detection, only upper limits that keep the contribution below roughly nine percent at 95 percent confidence.

The cost shows up elsewhere. Faster expansion before recombination changes how gravitational potentials decay around last scattering, and the fit compensates by raising the density of cold dark matter. More dark matter means more clustering, which drives up the amplitude of matter fluctuations and sharpens the existing mild tension with weak-lensing surveys such as KiDS and DES. Hill and collaborators argued that once large-scale-structure data are included, the case for early dark energy largely fades. The fit also pushes the spectral index of primordial fluctuations closer to one, a shift with consequences for inflationary models.

Ground-based microwave-background experiments have sharpened the picture. An earlier Atacama Cosmology Telescope release showed an intriguing mild preference for early dark energy without any distance-ladder input, while the South Pole Telescope did not. More recent ACT analyses, combined with Planck and DESI baryon acoustic oscillation data, find no significant preference. In these combined fits, the Hubble constant stays well short of 73.

Statistical subtleties complicate the verdict. As the early dark energy contribution approaches zero, the field's onset time and starting value stop being constrained by data. Standard Bayesian averaging then counts a large, empty region of parameter space against the model, an artefact known as the prior volume effect. Profile likelihood analyses, which ask only how well the best configuration fits, recover larger contributions and higher Hubble constants. Whether the data disfavour early dark energy depends partly on which statistical question is posed.

The emerging consensus is cautious. Early dark energy can ease the tension but has not convincingly resolved it, and it trades one discrepancy for strain elsewhere. The distance ladder is also being re-examined: tip-of-the-red-giant-branch and carbon-star calibrations from the CCHP team have yielded values closer to 70, while SH0ES maintains a value near 73. Simons Observatory polarization data, Euclid lensing and the full DESI survey will test the model's distinctive signatures far more sharply than current data can.

Takeaway

A solution that relieves one tension by worsening another may not be the answer, but it can still trace the outline of what our model is missing.

Early dark energy is a remarkable piece of theoretical engineering: a field that wakes up at a precise moment, does just enough work to shorten the cosmic ruler, then fades faster than light itself. That it can be built at all shows how much freedom the young universe might still hide.

Yet the microwave sky is an unforgiving witness. Every change before recombination echoes through acoustic peaks, polarization patterns and the later growth of galaxies. So far, those echoes suggest that early dark energy, at least in its simplest forms, cannot carry the full weight of the Hubble tension alone.

Perhaps the tension will dissolve as distance-ladder systematics are understood. Or perhaps it marks the first real crack in ΛCDM, pointing toward physics active in that brief, luminous era before the universe became transparent. Either way, how fast the universe is expanding today depends on what happened 380,000 years after the Big Bang.