Science › Big Bang & cosmos
The Hubble tension
2019· Open question
Open questionNobody knows the answer.
The disagreement over today's expansion rate: about 67 km/s/Mpc from the early universe, about 73 from nearby stars and supernovae.
See it on the Science map →How it works
The rate is quoted in kilometers per second per megaparsec (3.26 million light-years). Projecting forward from the afterglow with ΛCDM gives 67.4; measuring distances directly, rung by rung, gives about 73. The gap is about 5σ: a statistical fluke this large happens by chance about once in 3.5 million tries. Either one method hides an error, or something in the universe's history, perhaps before the afterglow formed, differs from the standard model.
Why it matters
The rate sets the universe's age and size scale. A real gap would be the first hard sign that ΛCDM is missing an ingredient.
Think of it like
Like two appraisals of one building 9% apart: one projects value from the original pro forma, the other uses recent local comps. Either a comp is bad or the pro forma misses something.
How we know
Planck (2018) gives 67.4 ± 0.5; Adam Riess's SH0ES team gives 73.0 ± 1.0 (2022). JWST checks (2023–24) found no error in their star measurements. Wendy Freedman's team, using red giant stars, gets about 70, between the two.
Still a mystery
Is it a measurement problem or new physics? Proposed fixes, such as a brief burst of early dark energy, tend to spoil other fits. No explanation has won broad support.
Connections (2)
▸The universe is expanding at a rate measured two ways, creating The Hubble tensionOpen question
The relationship
Route one builds a distance ladder: parallax, the geometry of Earth's orbit, measures nearby pulsing Cepheid stars; Cepheids calibrate Type Ia supernovae in other galaxies; supernovae reach far enough to read the expansion, giving about 73. Route two measures the sound-wave ruler in the afterglow and projects forward with ΛCDM, giving 67.4. Both look solid; they disagree by about 9%.
How the link was found
Adam Riess's SH0ES team built the modern ladder from 2005; Planck's 2013 results first exposed a clear gap, which grew to about 5σ by 2022.
Is it proven?
The gap is real at about 5σ unless one method hides an error. Methods independent of Cepheids give mixed results, from about 70 (red giant stars) to about 73 (lensed quasars).
Open questions and mysteries
Is the error in nearby distances, in the early-universe model, or nowhere? Gravitational waves from colliding neutron stars offer a third route but are not yet precise enough.
▸The Hubble tension may crack ΛCDM: the standard model of cosmologyOpen question
The relationship
The afterglow route gives 67 only if ΛCDM's history is right. If both numbers are true, the history is wrong. Favored repairs shrink the sound-wave ruler: a brief burst of 'early dark energy' before the afterglow, extra light particles, or slightly different atomic physics then. A smaller ruler at the same angle means the afterglow is closer than assumed, raising the inferred rate.
How the link was found
Vivian Poulin, Tristan Smith, Tanvi Karwal and Marc Kamionkowski proposed early dark energy for this in 2019. Lloyd Knox and Marius Millea's 2020 review showed how few repairs survive the other data.
Is it proven?
Not proven. Each repair raises the rate only partway and strains other data; the 2025 Atacama Cosmology Telescope maps found no sign of early dark energy.
Open questions and mysteries
If ΛCDM cracks, will the fix also explain DESI's fading dark energy or JWST's early galaxies? Or will the tension dissolve with better distance measurements?