🔭 astrophysics

The oldest Milky Way stars: New constraints on the age of the Universe and the Hubble constant

This paper utilizes robust age estimates from 3,000 ancient stars in the Gaia DR3 catalog to establish a statistically significant lower bound on the age of the Universe, thereby providing a cosmology-independent upper limit on the Hubble constant.

Original authors: Elena Tomasetti, Cristina Chiappini, Samir Nepal, Michele Moresco, Carmela Lardo, Andrea Cimatti, Friedrich Anders, Anna B. A. Queiroz, Guilherme Limberg

Published 2026-03-11

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Elena Tomasetti, Cristina Chiappini, Samir Nepal, Michele Moresco, Carmela Lardo, Andrea Cimatti, Friedrich Anders, Anna B. A. Queiroz, Guilherme Limberg

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

Imagine the Universe as a giant, ancient library. For decades, cosmologists have been trying to figure out exactly when the library opened its doors (the Big Bang) and how fast the shelves are expanding (the Hubble constant). But there's a problem: the two main ways they've tried to measure this time are giving different answers, like two clocks in the same room ticking at different speeds. This disagreement is called the "Hubble Tension."

In this new paper, a team of astronomers decided to try a completely different approach. Instead of looking at the big picture of the whole library, they decided to look at the oldest books on the shelves to see how old the library must be.

Here is the story of their discovery, broken down simply:

1. The Search for the "Oldest Books"

The researchers used data from the Gaia satellite, which is like a super-precise GPS for stars in our Milky Way galaxy. They had a catalog of about 200,000 stars.

Think of stars like people. Just as you can tell a person's age by looking at their wrinkles and hair color, astronomers can tell a star's age by looking at its brightness, temperature, and chemical makeup. However, just like estimating a human's age can be tricky, estimating a star's age is full of "systematic errors" (like using the wrong ruler).

The team didn't just pick any old stars. They went on a treasure hunt for the oldest, most reliable "Main Sequence Turn-off" and "Subgiant Branch" stars.

The Analogy: Imagine a school of students. The "Main Sequence" is the time they spend sitting in class learning. The "Turn-off" is the moment they graduate and leave the classroom to go on a different path. The "Subgiant" is the first step of their new career. These specific stages are the "sweet spot" for age-dating because the stars change visibly and predictably at this moment, making them perfect cosmic clocks.

2. The Great Filter (Cleaning the Data)

The team started with 200,000 stars but needed to be extremely picky. They applied a series of "filters" to remove the bad data, much like a librarian removing damaged or misfiled books.

The "19-Gyr" Glitch: At first, their data showed a weird spike of stars that were 19 billion years old. But we know the Universe is only about 13.8 billion years old! This was a "ghost" in the machine—a mathematical error where the computer got confused. The team realized these were likely "contaminants" (stars that looked old but weren't, or binary stars that had stolen mass from each other).
The Visual Inspection: After running the numbers through strict mathematical tests, the team did something very human: they looked at the charts with their own eyes. They checked the "posterior probability distributions" (a fancy way of saying, "Does this star's age look like a nice, smooth bell curve, or is it a messy, jagged mess?"). If the curve looked weird, they threw the star out.

The Result: From 200,000 stars, they whittled it down to a "Golden Sample" of just 160 stars. These are the absolute most reliable, ancient timekeepers the team could find.

3. The Big Discovery: The Universe is "Old Enough"

When they measured the ages of these 160 stars, the average came out to be 13.6 billion years.

Now, here is the crucial logic step:

The Universe must be older than the oldest things inside it.
If the oldest stars are 13.6 billion years old, the Universe must be at least that old, plus a little bit of time for the stars to actually form after the Big Bang.
Adding a small "formation delay" (about 0.2 billion years), they concluded the Universe is at least 13.8 billion years old.

4. Solving the "Hubble Tension"

Why does this matter? Because the age of the Universe is directly tied to how fast it is expanding (the Hubble constant, or $H_0$ ).

The Conflict: One group of scientists (using the Cosmic Microwave Background) says the Universe is expanding at a rate that makes it 14.0 billion years old. Another group (using nearby supernovae) says it's expanding faster, making it only 12.9 billion years old.
The Verdict: If the Universe were only 12.9 billion years old, it would be younger than the oldest stars (13.6 billion years). That is impossible! You can't have a child older than their parent.
The Conclusion: This new study acts as a "reality check." It suggests that the Universe is likely older than 13.8 billion years. This leans heavily toward the "slower expansion" measurement (the 14.0 billion year estimate) and rules out the "faster expansion" estimate (the 12.9 billion year one) as being too young to fit the evidence.

The Takeaway

This paper is like finding a fossil that proves the Earth is older than a specific geological theory predicted. By using individual stars as precise clocks, the team has provided a lower limit on the age of the Universe.

They found that the Universe is definitely old enough to hold these ancient stars. This doesn't solve the "Hubble Tension" completely, but it firmly plants a flag saying: "The Universe cannot be younger than 13.8 billion years."

It's a massive step forward, proving that we can use the stars themselves as independent witnesses to the history of the cosmos, independent of the complex models that have been causing the debate. As the Gaia satellite continues to send back more data, we will have even more "ancient books" to read, making our cosmic timeline even more precise.

1. Problem Statement

The paper addresses the "Hubble Tension," a significant discrepancy (4–5 $\sigma$ ) between measurements of the Hubble constant ( $H_0$ ) derived from the early Universe (Cosmic Microwave Background, e.g., Planck) and the late Universe (Cepheids and Supernovae, e.g., Riess et al.).

The Conflict: Planck data implies $H_0 \approx 67.4$ km/s/Mpc (Universe age $t_U \approx 14.0$ Gyr), while local measurements imply $H_0 \approx 73.0$ km/s/Mpc ( $t_U \approx 12.9$ Gyr).
The Need: Independent, cosmology-agnostic constraints on the age of the Universe ( $t_U$ ) are required to act as an "anchor point." Since the Universe must be older than its oldest constituents, precise measurements of the oldest stars can set a lower bound on $t_U$ , which in turn sets an upper bound on $H_0$ .
Previous Limitations: Previous studies relied on small samples (tens of globular clusters or a handful of stars) or were limited by systematic uncertainties in stellar models and the use of cosmological priors (assuming a specific $t_U$ during age derivation).

2. Methodology

Data Source

The study utilizes a massive dataset of ~200,000 Main Sequence Turn-Off (MSTO) and Subgiant Branch (SGB) stars from the Nepal et al. (2024) catalog, based on Gaia DR3 data.

Input Parameters: Atmospheric parameters ( $T_{\text{eff}}$ , $\log g$ , $[M/H]$ ) derived via a hybrid-CNN from Gaia RVS spectra, combined with Gaia photometry ( $G, B_p, R_p$ ) and 2MASS infrared data ( $J, H, K_s$ ).
Age Derivation: The authors used the Bayesian isochrone fitting code StarHorse.
- Crucial Modification: Unlike the original Nepal et al. (2024) work, which imposed a cosmological prior ( $t_U = 13.73$ Gyr), this study removed the upper age bound, allowing the isochrone fitting to explore a uniform range of 0.025 to 20 Gyr. This ensures the age estimates are independent of any specific cosmological model.

Rigorous Selection Process

To isolate a sample of "bona-fide" old stars and minimize systematic biases, the authors applied a multi-stage filtering process to the initial 2,911 candidates (stars with age > 12.5 Gyr and uncertainty < 1 Gyr):

Kiel Diagram Cut: Applied a stricter cut in the $\log g$ – $T_{\text{eff}}$ plane to exclude contaminants like low-luminosity giants, main-sequence stars, and binaries. This removed a spurious peak at ~19 Gyr.
Parameter Consistency: Enforced tight constraints on the difference between input and output stellar parameters ( $\Delta[M/H]$ , $\Delta T_{\text{eff}}$ , $\Delta \log g$ , $\Delta A_V$ ) to mitigate age-metallicity degeneracies.
Posterior Distribution Quality: Discarded stars with asymmetric age/mass uncertainties or posterior probability distributions (PDFs) that deviated significantly from Gaussianity (Kolmogorov-Smirnov test).
Visual Inspection: A "blind" inspection of corner plots to identify double peaks or anomalies, classifying stars into "great," "good," and "bad" tiers.
Contaminant Removal: Used a Gaussian Mixture Model and hierarchical Bayesian analysis to identify a secondary population of "spurious" old stars (likely mass-stripped binaries). These were excluded with a conservative threshold.

Final Sample: 160 stars (including a "golden sample" of 78 highest-quality stars).

3. Key Contributions

First Statistical Use of Individual Stars: This is the first study to use a statistically significant sample (160 stars) of individual field stars as cosmic clocks, moving beyond small samples of globular clusters.
Cosmology-Independent Derivation: By removing the $t_U$ prior during the StarHorse fitting process, the resulting ages are purely observational constraints, not circular arguments based on a specific cosmological model.
Systematic Error Quantification: The paper provides a detailed breakdown of systematic uncertainties, particularly those arising from stellar models (e.g., mixing length parameter $\alpha_{ML}$ ) and $\alpha$ -enhancement assumptions.

4. Results

Stellar Age Distribution

The age distribution of the final sample peaks at 13.6 ± 1.0 (stat) ± 1.3 (syst) Gyr.
The sample shows a bimodal distribution: a narrow peak at ~13.4 Gyr and a broader, contaminant peak at ~14.8 Gyr. After removing the contaminant tail, the distribution remains robust.
Lower Bound on $t_U$ : Assuming a formation redshift of $z_f = 20$ (a delay of ~0.2 Gyr between the Big Bang and star formation), the lower bound on the age of the Universe is:
$t_U \ge 13.8 \pm 1.0 \text{ (stat)} \pm 1.3 \text{ (syst) Gyr}$
Individual Star Constraints: At the 90% confidence level (CL), 70 stars in the sample indicate $t_U > 13$ Gyr, while no star exceeds 14.1 Gyr.

Constraints on the Hubble Constant ( $H_0$ )

Translating the age constraint into an upper limit for $H_0$ (assuming a flat $\Lambda$ CDM model with $\Omega_m = 0.3$ and $z_f = 20$ ):
$H_0 \le 68.3^{+5.4}_{-4.7} \text{ (stat)} ^{+7.2}_{-5.9} \text{ (syst) km s}^{-1} \text{ Mpc}^{-1}$

This upper limit is consistent with the Planck value ($67.4$) but significantly lower than the local distance ladder value ($73.0$).
The analysis shows that for the data to support the higher $H_0$ values (implying a younger Universe), the full systematic error budget would need to shift ages downward by ~1 Gyr, which is not supported by current stellar models (which often predict older ages with different mixing lengths).

Sensitivity to Cosmological Assumptions

The paper explores how varying $\Omega_m$ and $z_f$ affects the derived $H_0$ .

Higher assumed formation redshifts ( $z_f$ ) lead to higher derived $H_0$ limits.
Higher matter density ( $\Omega_m$ ) leads to lower $H_0$ limits.
The sample only favors an $H_0$ lower than the CMB value if $\Omega_m = 0.35$ and $z_f = 10$ . However, for standard assumptions ( $\Omega_m \ge 0.3$ ), the stellar ages consistently point toward an $H_0$ lower than the local distance ladder measurements.

5. Significance

Independent Cosmological Probe: This work establishes stellar ages as a robust, independent method to constrain cosmological parameters, free from the systematic errors affecting CMB or supernova distance ladders.
Resolution of Tension: The results suggest that the Universe is likely older than 13.8 Gyr, which favors the lower $H_0$ values (Planck) and challenges the higher local $H_0$ values, potentially hinting at new physics or hidden systematics in the late-Universe measurements.
Future Potential: The authors note that while this is a significant first step, future Gaia data releases and high-resolution spectroscopy (e.g., from the Haydn mission) will allow for even larger samples and reduced systematic errors, further tightening these constraints.

In summary, Tomasetti et al. provide the first statistically rigorous, model-independent lower bound on the age of the Universe using individual stars, effectively placing an upper limit on the Hubble constant that aligns more closely with early-Universe measurements.