The Age of the Universe from the Galaxy's Oldest Stars ✦
Using spectroscopy from LAMOST DR7 and parallaxes from Gaia eDR3 for 247,103 Milky Way stars, ages were estimated with YY isochrones up to 20 billion years. Strict criteria (metallicity, α-enrichment, agreement with FLAME) narrowed the sample to 155,600 objects within a 5 kpc radius. MCMC reconstruction of the underlying distribution gave an oldest star age of 13.73+0.18−0.15 billion years, with a spread of 13.31–14.02 billion years under selection variation. The result is consistent with 13.6 billion years in the ΛCDM model calibrated by the relic radiation (first stars after 0.2 billion years). This challenges Hubble tension solutions via new physics before recombination, as such models predict a universe age of ~12.9 billion years, which is hard to reconcile with observations of old metal-poor stars and asteroseismic constraints.
Context
The age of the Universe is a fundamental constant, but its precise value depends on the cosmological model. Direct measurements of redshift via Hubble's law (discovered by Edwin Hubble) conflict with estimates from the cosmic microwave background (CMB). This 'Hubble tension' motivates the search for alternatives to the standard ΛCDM cosmology. 'Early' solutions, which modify physics before the recombination epoch, are of particular interest. They predict a younger Universe—about 12.9 billion years. This can be tested through 'galactic archaeology': the study of the most ancient stars in our Milky Way, whose ages serve as cosmic clocks, almost independent of theoretical models.
Methods
The foundation was a catalog of 247,103 stars, for which ages were determined from LAMOST spectroscopy and Gaia eDR3 astrometry using YY isochrones (up to 20 Gyr) and likelihood functions. To filter out stars with unreliable estimates (e.g., inexplicably small errors for ages >16 Gyr), additional population criteria were introduced: 'old' stars must be metal-poor ([Fe/H] < 2.2 – 0.21×age) and enriched in α-elements ([α/Fe] > –0.5 + 0.05×age); their ages must agree with independent FLAME estimates from Gaia within 3σ. After selection, 155,600 stars remained, with age uncertainties rising monotonically toward high values. To determine the true age of the oldest star (A★), a non-parametric MCMC reconstruction of the hidden population age distribution was performed in 100 bins, using a gradient penalty for smoothing. Then, by analyzing the decline of the probability-to-uncertainty ratio (P/σ) with age, a break was identified and interpreted as the A★ boundary.
Results
The reconstructed distribution shows an almost linear decline in P/σ over 12–14 Gyr, followed by a 'noise plateau' beyond 14 Gyr. The intersection of these two regimes gives an oldest star age of A★ = 13.73 (+0.18/–0.15) Gyr. Adding 0.2 Gyr for the formation of the first long-lived stars after the Big Bang, a concept pioneered by Georges Lemaître, yields a cosmic age of 13.93 (+0.18/–0.15) Gyr, which agrees with the ΛCDM prediction of 13.8 Gyr (calibrated by the CMB) within 1σ. Meanwhile, a model with an 'early' solution to the Hubble tension, using Friedmann's equations and modern parameter values, gives a Universe age of 12.91 ± 0.18 Gyr and consequently A★ = 12.71 ± 0.18 Gyr. Our result rules out this scenario at >3σ tension, even with the most stringent sample cuts. Analysis of 1,556 stars with extreme metallicity ([Fe/H] ≤ –0.9) and high [α/Fe] (≥0.2) confirms a close age of 13.85 (+0.05/–0.04) Gyr, almost independent of assumptions about the initial helium abundance.
Implications
This age limit calls into question a whole class of ΛCDM modifications that attribute accelerated expansion to new physics in the early Universe. If the expansion rate had been 9% higher throughout most of cosmic history—as required by such models—the age of the cosmos would have dropped by ~1 Gyr, incompatible with the observed age of the oldest stars. This strengthens the case for 'late' solutions (e.g., a local void or modifications of gravity) and points to the need to reinterpret redshifts in the nearby Universe without revising general relativity A. Einstein.
Future development
Further progress requires improvements in five directions: larger samples and finer age grids; reducing systematics in stellar models (especially regarding helium enrichment and the mixing length parameter αML); refining the time of first star formation (tf); increasing the precision of cosmographic parameters from low-redshift observations; and more fully accounting for uncertainties in the MCMC reconstruction method.
Impact
The results impact cosmology (Hubble tension, nature of dark energy), galactic archaeology, and the physics of stellar evolution.
Next steps
The next step is to repeat the analysis with LAMOST DR8 and Gaia DR4 data, and to apply more modern isochrone grids that account for individual α-element abundances.
Key open problems
The work is directly linked to unsolved problems: the nature of dark energy and dark matter, the first moments after the Big Bang (tf), and the origin of stellar populations in the early Galaxy.
🎯 The oldest star in the selected sample, Gaia DR3 3656896440426302592, has an observed age of ~13.5 Gyr—only 0.3 Gyr younger than the entire Universe. If its age were underestimated by the typical 7.5%, the true value could reach 14 Gyr!
Key numbers
- oldest star (A★): 13.73 (+0.18/–0.15) Gyr
- Universe age from CMB (ΛCDM): 13.8 Gyr
- Universe age from 'early' solutions: 12.91 ± 0.18 Gyr
- final sample size: 155,600 stars
- initial catalog size: 247,103 stars