🔭 astrophysics

The age of the Universe from a large sample of the oldest Galactic stars

By analyzing a large sample of 155,600 old, metal-poor, and $\alpha$ -enriched Milky Way stars, the authors estimate the oldest stellar age to be approximately 13.73 Gyr, a result consistent with the standard $\Lambda$ CDM cosmological model and challenging Hubble tension solutions that rely on new physics prior to recombination to lower the cosmic age.

Original authors: Indranil Banik, Thenujaya Kudakolawa Kaluarachchige, Stephen Cookson, Harry Desmond

Published 2026-07-02

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Indranil Banik, Thenujaya Kudakolawa Kaluarachchige, Stephen Cookson, Harry Desmond

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Cosmic Birthday Party

Imagine the Universe as a giant, ancient house. Cosmologists (scientists who study the Universe) have been trying to figure out exactly when the house was built. They have two main ways to guess the age:

The Blueprint: They look at the "blueprints" of the early Universe (the Cosmic Microwave Background) to calculate how long it has been expanding. This suggests the house is about 13.8 billion years old.
The Fossils: They look for the oldest furniture inside the house. In our case, the "furniture" is stars. If you find a star that is 13 billion years old, the house must be at least that old.

The Problem: A Tension in the Room

Recently, a conflict has arisen. Some measurements of how fast the Universe is expanding today (the "Hubble Tension") suggest that if we use certain new physics theories to fix the math, the Universe might actually be much younger—only about 12.9 billion years old.

If the Universe is only 12.9 billion years old, but we find a star that is 13.5 billion years old, we have a problem. It's like finding a 100-year-old man living in a house that was only built 90 years ago. Something is wrong with the math, the measurements, or the story.

The Investigation: Cleaning the Data

The authors of this paper decided to investigate the "fossils" (the stars) to see if they could settle the argument. They started with a massive list of 247,000 stars from a telescope called LAMOST and a space map called Gaia.

However, they knew this list was messy. It was like a library where some books had been mislabeled.

The Glitch: They noticed that some stars were listed as being incredibly old (over 16 billion years) but with suspiciously perfect, tiny error bars. It was as if a librarian claimed a book was 100% accurate but had never actually read it.
The Cleanup: To fix this, they applied strict "quality filters." They looked for stars that made sense chemically.
- The Metal Rule: The oldest stars in the galaxy should be "metal-poor" (like a raw, unrefined rock). If a star looks ancient but is "metal-rich" (like a shiny, processed coin), it's likely a mistake.
- The Alpha Rule: The oldest stars should be rich in "alpha elements" (specific chemical ingredients). If an ancient star lacks these, it's probably a fake.
- The Double-Check: They cross-referenced their star ages with a different database (Gaia FLAME). If the two databases disagreed too much, they threw the star out.

After this rigorous cleaning, they were left with 155,600 stars. The "glitch" where error bars got weirdly small disappeared. The data looked clean.

The Discovery: Finding the Oldest Star

With their clean list, they used advanced statistics (like a super-powered detective) to find the true age of the oldest star in the sample.

The Result: They found the oldest star is approximately 13.73 billion years old.
The Margin of Error: Even accounting for uncertainty, the star is almost certainly older than 13.5 billion years.

The Verdict: Who is Right?

This result is a big deal because it acts as a referee in the cosmic argument:

It supports the "Blueprint" (Standard Model): The age of 13.73 billion years fits perfectly with the standard theory that the Universe is 13.8 billion years old (assuming the first stars formed 0.2 billion years after the Big Bang).
It challenges the "New Physics" theories: The theories that try to solve the Hubble Tension by changing the early Universe predict a Universe that is only ~12.9 billion years old.
- The Analogy: If the Universe is only 12.9 billion years old, then a 13.73-billion-year-old star is a "time traveler." It shouldn't exist.
- The Conclusion: The authors argue that it is very difficult to explain how a star could be older than the Universe itself. Therefore, the theories that predict a younger Universe are likely incorrect.

Summary

The paper says: "We took a huge, messy list of stars, cleaned it up using strict chemical rules, and found the oldest star is about 13.7 billion years old. This confirms that the Universe is likely the standard 13.8 billion years old and casts doubt on theories that try to make the Universe younger to fix other cosmic problems."

They did not find a new way to travel through time, nor did they suggest new medical uses for stars. They simply used the stars as a clock to check the age of the Universe, and the clock says the standard story is likely the correct one.

Technical Summary: The Age of the Universe from a Large Sample of the Oldest Galactic Stars

Problem Statement
The paper addresses the determination of the age of the Universe ( $A_U$ ) by estimating the latent age of the oldest stars in the Milky Way ( $A_\star$ ). This measurement serves as a critical lower limit for $A_U$ (where $A_U = A_\star + t_f$ , with $t_f \approx 0.2$ Gyr being the formation time of the first long-lived stars). The study is motivated by the "Hubble tension," a statistically significant discrepancy between the Hubble constant ( $H_0$ ) derived from the Cosmic Microwave Background (CMB) under the standard $\Lambda$ CDM model ( $H_0^{CMB} \approx 67.4$ km/s/Mpc) and local distance-ladder measurements ( $H_0^{local} \approx 73.5$ km/s/Mpc).

Solutions to this tension involving new physics prior to recombination generally predict a younger Universe ( $A_U \approx 12.9 \pm 0.2$ Gyr) to accommodate the higher $H_0$ while maintaining consistency with CMB data. Conversely, the standard CMB-calibrated $\Lambda$ CDM model predicts $A_U \approx 13.8$ Gyr. The authors aim to test these competing predictions by deriving $A_\star$ from a large, high-quality sample of Galactic stars without imposing strong cosmological priors that would bias the result toward either model.

Methodology
The authors utilize a sample of 247,103 subgiant stars from the Xiang & Rix (2022, hereafter X22) catalogue, which combines LAMOST DR7 spectroscopy with Gaia eDR3 astrometry. Stellar ages were initially estimated using Yonsei-Yale (YY) isochrones. To ensure the reliability of the age estimates for cosmological inference, the authors applied a series of rigorous, population-level quality cuts designed to remove stars with problematic age determinations (e.g., those with unrealistically low uncertainties at high ages) without introducing circular cosmological biases.

Key quality control steps included:

Astrometric Cuts: Requiring parallax uncertainty $\sigma_\varpi/\varpi < 0.1$ and $\varpi > 0.2$ mas (limiting the sample to within 5 kpc).
Binary Contamination: Excluding stars with high ipd_frac_multi_peak values in Gaia data to remove marginally resolved binaries.
Chemical Consistency: Applying age-dependent cuts on metallicity ([Fe/H]) and $\alpha$ -element abundance ([ $\alpha$ /Fe]). The authors enforce that old stars must be metal-poor and $\alpha$ -enriched, consistent with Galactic chemical evolution models. Stars appearing old but metal-rich or $\alpha$ -poor were rejected as likely measurement errors.
Cross-Catalogue Validation: Comparing YY ages with FLAME ages derived solely from Gaia DR3 data. A linear trendline was fitted to the age-mass relation to impute missing FLAME ages, followed by a $3\sigma$ outlier rejection between YY and FLAME ages.

After these cuts, the final nominal sample contained 155,600 stars. The authors employed two primary statistical approaches to reconstruct the latent age distribution $P(A)$ and infer $A_\star$ :

Iterative Reconstruction: A non-parametric method where the population distribution serves as a prior to refine individual stellar posteriors, iterated until convergence.
MCMC Reconstruction: A Markov Chain Monte Carlo approach to infer $P(A)$ in 100 age bins (0.1–20.1 Gyr). The authors modeled the log-posterior of the distribution, including a gradient penalty to smooth noise. To infer $A_\star$ , they analyzed the ratio of the reconstructed probability density to its uncertainty ( $P/\sigma$ ), identifying a sharp "crossover" point where the distribution transitions from a declining trend to a flat "bias floor" (where $P > 0$ is enforced by the algorithm).

Key Results
The analysis yields a robust estimate for the age of the oldest star in the sample:
$A_\star = 13.73^{+0.18}_{-0.15} \text{ Gyr}$
This result is derived from the nominal sample and analysis method. The authors tested the robustness of this value by varying the quality cuts:

Metallicity Cuts: Relaxing the [Fe/H] ceiling (allowing more metal-rich stars) increased $A_\star$ slightly (up to $14.02^{+0.18}_{-0.15}$ Gyr), while tightening it decreased $A_\star$ (down to $13.31^{+0.21}_{-0.18}$ Gyr). However, the authors argue that tightening the cut by 0.2 dex is unphysical as it removes stars consistent with the Galactic age-metallicity trend and excludes known populations like the Gaia-Sausage/Enceladus merger.
$\alpha$ -enhancement and Cross-Catalogue Cuts: Variations in the [ $\alpha$ /Fe] floor and the stringency of the YY-FLAME agreement ( $2.5\sigma$ vs $3\sigma$ ) resulted in $A_\star$ values consistent with the nominal result within uncertainties.

A secondary analysis focusing on a "early chemistry" subsample (1,556 stars with low [Fe/H] and high [ $\alpha$ /Fe]) using a parametric delayed- $\tau$ model for $P(A)$ yielded a consistent result of $A_\star = 13.85^{+0.05}_{-0.04}$ Gyr, further supporting the main finding.

Significance and Claims
The paper claims that its inferred $A_\star$ is consistent with the standard CMB-calibrated $\Lambda$ CDM prediction ( $A_\star^{CMB} \approx 13.6$ Gyr, assuming $t_f = 0.2$ Gyr) but in significant tension ( $>2\sigma$ ) with predictions from "early-time" solutions to the Hubble tension ( $A_\star^{ETS} \approx 12.7$ Gyr).

The authors argue that:

Constraints on New Physics: The agreement with the standard $\Lambda$ CDM age casts doubt on solutions to the Hubble tension that rely solely on new physics prior to recombination, as these generally require a cosmic age of $12.9 \pm 0.2$ Gyr. Reconciling such a low age with the authors' result would require stellar modeling uncertainties to be significantly larger than currently understood, which is unlikely given the low metallicities of the sample and independent asteroseismic constraints.
Methodological Robustness: The study demonstrates that large-sample, population-level quality cuts (based on chemistry and cross-catalogue consistency) are essential for removing outliers that would otherwise bias age estimates. The use of MCMC reconstruction allows for a rigorous handling of the latent age distribution and its uncertainties.
Consistency with Previous Work: The result aligns with previous estimates of the oldest globular clusters and individual stars (e.g., HD 140283), reinforcing the view that the Universe is approximately 13.8 Gyr old.

The paper concludes that while stellar modeling uncertainties exist, they are insufficient to reconcile the high inferred age of the oldest stars with the low ages predicted by early-time Hubble tension solutions, thereby supporting the standard cosmological age of the Universe.