The Hubble Tension Has Survived a Decade of Scrutiny. What If It Is Not a Mistake?

Imagine two ways to measure how fast the universe is expanding. One reads the cosmic microwave background (CMB), the relic light released when the universe was about 380,000 years old. The other builds a distance ladder that uses nearby stars to calibrate exploding white dwarfs called Type Ia supernovae. For a decade, these two approaches have refused to agree.

The first, when interpreted within the standard Lambda cold dark matter (ΛCDM) model, gives a present-day expansion rate—the Hubble constant—of about 67 kilometers per second per megaparsec. The second gives about 73. That roughly nine percent gap is the Hubble tension. A newly posted review on arXiv by Rong-Gen Cai and Shao-Jiang Wang argues that after ten years of scrutiny it has hardened into a crisis for modern cosmology. The review is a preprint and synthesizes published work rather than presenting new observations.

Neither Planck Nor SH0ES Stands Alone

It once seemed possible that the low value was an artifact of one satellite. The Planck mission mapped the CMB and, assuming ΛCDM, inferred H0 = 67.27 ± 0.60. The SH0ES team, using Cepheid-calibrated supernovae, found 73.04 ± 1.04, a disagreement at roughly five standard deviations.

The review shows the low value is not unique to Planck. The Atacama Cosmology Telescope (ACT) and the South Pole Telescope (SPT), which observe the CMB with different instruments and analyses, also prefer H0 near 67 to 68; the source notes that some non-Planck constraints still use Planck-derived priors or calibration. A combined Planck, ACT, and SPT analysis gives 67.19 ± 0.38. Even before adding any primary CMB map, the combination of baryon acoustic oscillations (BAO, frozen sound waves imprinted in galaxy clustering) and big bang nucleosynthesis (BBN, light-element abundances) gives values mostly in the 67 to 68 range, with some as high as about 68.6.

The high value is not a single-team fluke either. Cepheids, the tip of the red giant branch (TRGB), surface brightness fluctuations (SBF), Mira stars, Type II supernovae, and gravitational lenses have produced late-universe values often in the low 70s to mid-70s, though individual measurements range lower or higher and carry different systematics. A recent community consensus, the Local Distance Network, reported 73.50 ± 0.81. The authors note that if one ordinary systematic error were responsible, it would have to masquerade as new physics.

Sound-Horizon and Absolute-Magnitude Fixes Run Into No-Go Theorems

If the answer lies in the early universe, the goal is to shrink the sound horizon: the maximum distance sound waves could travel before the CMB was released (more precisely, the sound horizon at the baryon-drag epoch), a physical ruler imprinted on the CMB and on galaxy clustering. A smaller ruler would let the same CMB pattern correspond to a faster expansion today.

The review describes several ways to do this: add early dark energy, change the electron mass or fine-structure constant, or invoke primordial magnetic fields that speed up recombination. These ideas face what the authors call early-time no-go theorems. They must also change the primordial fluctuation spectrum, pushing its tilt toward the scale-invariant Harrison-Zeldovich value, and they tend to worsen a separate tension over how quickly matter clumps, known as S8. Some are tightly constrained by small-scale CMB data. The review treats them as live but not clean solutions.

If the answer lies in the late universe, one can try to change the absolute brightness of Type Ia supernovae or alter dark energy so that distances shift at low redshift. But the inverse distance ladder works in the opposite direction: it calibrates supernovae using the sound horizon inferred from the CMB or BBN, then propagates that calibration to lower redshifts. That leaves little room for a sudden late-time transition. A local void around us has also been ruled out as a significant contributor to the full discrepancy, because it is not large and deep enough.

The review adds another obstacle: the cosmic distance duality relation, which links distances measured by brightness and by angular size. Late-universe models that violate it too strongly are already in tension with data. Homogeneous late-time fixes, the authors conclude, cannot do the job alone.

Intercepts, Densities, and a Chameleon-Like Doorway

The review highlights two newer tensions around the scale where the universe becomes homogeneous. The first is an intercept tension. In the relation between supernova brightness and redshift, local and late-time samples disagree by more than five standard deviations in the intercept, a quantity that mixes the supernova absolute magnitude and the Hubble constant. That suggests the problem may not be a simple calibration error in either quantity.

The second is a slope tension. When supernovae are grouped by the density of their environments, those in denser regions seem to yield a larger Hubble constant. That is the opposite of the simplest expectation from non-local sample variance—the expected correlation between measured H0 and the typical density around the supernova hosts rather than only around the observer. The effect is near three sigma, so it is not decisive, but it points toward local physics or a systematic effect not yet modeled.

The authors argue that the most promising paths are combined early-and-late modifications, especially interacting dark energy-dark matter models (in which dark energy and dark matter exchange energy), or local-scale new physics that disguises itself as a systematic. In one chameleon dark energy scenario, denser regions would have a larger effective cosmological constant, making the local expansion rate higher. In that picture, the Hubble tension is not a mistake to be erased but a phenomenon that depends on where and how we measure.

The review is cautious. DESI's hint of evolving dark energy may be real, but it may also be an apparent phantom crossing—a fit in which dark energy appears to cross the value -1 in its equation of state—produced by a dark-sector interaction misread through a simpler model. It does not by itself solve the Hubble tension and can even worsen it. No proposed solution yet has decisive support.

Still, the decade-long disagreement has become more than a nuisance. It is a persistent signal that two well-tested ways of reading cosmic history do not match. The authors compare the moment to the birth of quantum theory, when conflicting measurements of blackbody radiation forced a new idea. The Hubble tension may be cosmology's version of that clue: not an error to hide, but a doorway to whatever lies beyond the standard model.