The Hubble Tension

The expansion rate of the universe should be a single number. Measure it one way, you get 67. Measure it another way, you get 73. The discrepancy is now statistically significant at the 5-sigma level, the threshold physicists use for declaring a discovery. Either one of the measurements is wrong, or the standard model of cosmology is incomplete. The Hubble tension is the most important unresolved problem in modern cosmology, and its resolution could be as mundane as a calibration error or as revolutionary as new fundamental physics.

What the Hubble Constant Measures

The Hubble constant H0 quantifies the current expansion rate of the universe. A galaxy at a distance d recedes from us at a velocity v = H0 x d (Hubble's Law). If H0 = 70 km/s/Mpc, a galaxy 100 megaparsecs away is receding at 7,000 km/s. The Hubble constant also determines the age of the universe (inversely proportional to H0, with corrections for the expansion history) and the physical scale of structures observed at cosmological distances.

Measuring H0 requires two things: the distance to an object and its recession velocity. Recession velocity is straightforward (measured from redshift). Distance is the hard part. Astronomical distance measurement relies on the "cosmic distance ladder," a sequence of overlapping methods that calibrate each other: geometric parallax for nearby stars, Cepheid variable stars and the tip of the red giant branch for nearby galaxies, and Type Ia supernovae for distant galaxies.

The tension arises because two fundamentally different approaches to measuring H0 give different answers.

The Local Measurement: 73 km/s/Mpc

The local (or "late universe") measurement of H0 uses the cosmic distance ladder to measure distances to galaxies whose recession velocities are known from redshifts. The most precise local measurement comes from the SH0ES project (Supernovae and H0 for the Equation of State), led by Adam Riess at Johns Hopkins.

The SH0ES methodology calibrates the distance ladder in three steps. First, geometric distances to anchor galaxies are established using Cepheid variables whose parallaxes have been measured by Gaia or whose distances are known from other geometric methods (eclipsing binaries in the Large Magellanic Cloud, the water maser distance to NGC 4258). Second, Cepheid period-luminosity relations calibrated by these anchors are used to measure distances to galaxies that also contain Type Ia supernovae. Third, the calibrated Type Ia supernova luminosities are used to measure distances to more distant galaxies in the Hubble flow (the regime where recession velocity is dominated by cosmic expansion rather than local motions).

The SH0ES result, updated through multiple publications with progressively improved data and analysis, converges on H0 = 73.0 +/- 1.0 km/s/Mpc. Independent local measurements using the tip of the red giant branch (TRGB) as an alternative distance indicator yield somewhat lower values (69.8 +/- 1.7 km/s/Mpc from the Carnegie-Chicago Hubble Program, led by Wendy Freedman), partially alleviating the tension but not fully resolving it. The TRGB calibration itself is debated, with different analyses producing different results depending on the photometric treatment of the red giant branch.

Other independent local methods include surface brightness fluctuations, gravitational lensing time delays (the H0LiCOW/TDCOSMO projects, which measure time delays between multiple images of gravitationally lensed quasars), and the Tully-Fisher relation. Most of these methods yield values in the 70-75 km/s/Mpc range, broadly consistent with SH0ES.

The Early Universe Measurement: 67 km/s/Mpc

The early universe measurement derives H0 from observations of the cosmic microwave background (CMB), particularly ESA's Planck satellite. The CMB encodes the physics of the universe at recombination (380,000 years after the Big Bang) through the pattern of temperature and polarization fluctuations. Fitting the six-parameter Lambda-CDM cosmological model to the observed power spectrum yields precise values for the cosmological parameters, including H0.

Planck's result is H0 = 67.4 +/- 0.5 km/s/Mpc. The uncertainty is small because the CMB constrains the full set of cosmological parameters simultaneously, and the model's internal consistency is excellent. However, this is not a direct measurement of the current expansion rate. It is an inference: given the physics of the early universe encoded in the CMB, and assuming the Lambda-CDM model correctly describes the subsequent 13.8 billion years of cosmic evolution, what must H0 be today?

This distinction is critical. The Planck measurement assumes that the standard model is correct. If there is new physics between recombination and today, unknown components of the universe, modifications to dark energy behavior, additional relativistic particles, then the inferred H0 could be wrong even if the CMB data and its analysis are perfect.

Baryon acoustic oscillation (BAO) measurements from galaxy surveys (SDSS, DESI) provide an independent constraint that is broadly consistent with Planck's H0, reinforcing the early universe value. The DESI Year 1 results (2024) suggested possible hints of evolving dark energy, which if confirmed would have implications for the tension, though the statistical significance was modest.

The Statistical Significance

The discrepancy between the SH0ES local value (73.0 +/- 1.0) and the Planck CMB value (67.4 +/- 0.5) is approximately 5.6 km/s/Mpc, corresponding to a statistical significance of roughly 5 sigma when the uncertainties are combined in quadrature. In particle physics, 5 sigma is the conventional threshold for claiming a discovery. In cosmology, where systematic errors are harder to quantify than statistical ones, the threshold is treated with more caution, but the significance is undeniably high.

The persistence of the tension across multiple independent local and early universe measurements, each with different systematic error profiles, makes it increasingly difficult to attribute the discrepancy to any single measurement error. If SH0ES is wrong, the error must be in the Cepheid calibration or the supernova standardization, and it must be of a specific magnitude and direction that multiple cross-checks have failed to identify. If Planck is wrong, the error must be in the CMB analysis or the foreground modeling, both of which have been scrutinized intensely. If both are correct, the standard model is incomplete.

Possible Resolutions

Proposed resolutions fall into three categories: systematic errors in local measurements, systematic errors in CMB analysis, and new physics.

Systematic errors in local measurements: The most commonly discussed possibility is an error in the Cepheid distance scale, perhaps related to the effects of crowding (in which unresolved stellar neighbors bias Cepheid brightness measurements), metallicity dependence of the period-luminosity relation, or calibration of the LMC anchor distance. JWST observations of Cepheids in SH0ES galaxies have largely confirmed the HST-based distances, reducing (but not eliminating) concerns about crowding. The TRGB discrepancy (different groups get different TRGB distances to the same galaxies) suggests that distance ladder systematics are not fully resolved.

Systematic errors in CMB analysis: The Planck results have been cross-checked by independent CMB experiments (ACT, SPT) with broadly consistent results, making instrumental systematics unlikely. Foreground modeling (particularly the treatment of galactic dust emission) has been questioned but not shown to produce errors of the required magnitude.

New physics: This is the exciting category. Proposed models include early dark energy (a component of dark energy that was significant in the early universe and then faded, altering the sound horizon and thus the inferred H0), additional relativistic species (extra neutrinos or other light particles that increase the expansion rate before recombination), interacting dark matter (dark matter that exchanges energy with dark energy or neutrinos), modified gravity, decaying dark matter, and primordial magnetic fields. Most of these models can reduce the tension to some degree but struggle to resolve it completely without creating new tensions with other datasets (BAO, galaxy clustering, Big Bang nucleosynthesis).

Early dark energy is perhaps the most studied new physics proposal. It postulates a scalar field that contributes a few percent of the total energy density around the time of recombination, effectively shrinking the sound horizon (the characteristic scale of CMB fluctuations) and increasing the inferred H0. The model fits the CMB data comparably to Lambda-CDM (with additional parameters) and reduces the tension with local measurements, but the theoretical motivation is ad hoc and the model introduces fine-tuning concerns.

The Path Forward

Several observational programs are specifically designed to address the Hubble tension.

JWST observations of Cepheids and TRGB stars in distance ladder galaxies are providing independent checks on HST-based distances with superior resolution and reduced crowding effects. Early results have largely confirmed the SH0ES distances, strengthening the case that the local measurement is robust.

The DESI (Dark Energy Spectroscopic Instrument) survey is measuring baryon acoustic oscillations with unprecedented precision across a wide redshift range, providing constraints on the expansion history that are independent of both the CMB and the distance ladder. DESI's full dataset will significantly tighten or relax the tension.

Gravitational wave standard sirens (like GW170817) provide a completely independent distance measurement method. As the catalog of neutron star mergers with electromagnetic counterparts grows, the standard siren measurement of H0 will become competitive with other methods. Current estimates from a small sample are consistent with both the local and CMB values (the uncertainties are still large), but the method is promising precisely because its systematic errors are entirely different from either the distance ladder or the CMB.

The Vera Rubin Observatory's LSST will discover thousands of Type Ia supernovae per year, enabling supernova-based distance measurements with dramatically improved statistics and better control of systematic effects. The Roman Space Telescope will combine infrared observations of Cepheids and supernovae with weak gravitational lensing measurements, providing multiple independent constraints.

The Hubble tension may be resolved by improved measurements revealing a previously unrecognized systematic error. Or it may be resolved by new physics that requires fundamental revision of the standard cosmological model. Either outcome would be significant. The former would illuminate the limitations of precision measurement in astronomy. The latter would open a new chapter in physics. The universe has one expansion rate. We just don't agree on what it is yet.