The Age of the Universe with Globular Clusters IV: Multiple Stellar Populations

By analyzing 69 globular clusters with a hierarchical framework that explicitly accounts for multiple stellar populations, this study confirms that such complexity has a negligible impact on age estimates, yielding a robust, model-independent age for the Universe of $13.81 \pm 0.32$ Gyr.

The Gist

Imagine the Universe as a giant, ancient library. For decades, astronomers have been trying to figure out exactly how old this library is. One of the best ways to do this is to look at the "oldest books" in the collection: Globular Clusters. These are massive, spherical groups of hundreds of thousands of stars that have been orbiting our galaxy since the dawn of time.

For a long time, scientists treated these clusters like a choir of identical twins: they assumed every star in a cluster was born at the exact same moment, from the exact same cloud of gas, and was exactly the same age.

This paper is about realizing that the choir isn't made of identical twins, but of two distinct generations of siblings.

Here is the story of what the researchers did, explained simply:

1. The Old Assumption: "One Big Birthday Party"

Previously, when scientists looked at a globular cluster, they assumed it was a single "batch" of stars. They would look at the stars' colors and brightness (like looking at a photo of a crowd) and calculate: "Okay, if everyone was born at the same time, this cluster is 13.5 billion years old."

This method worked well, but it ignored a known fact: Stars in these clusters often have multiple generations. Some stars are the "firstborn" (Population I), and others are the "secondborn" (Population II), born later from the leftover gas of the first generation.

2. The New Approach: "The Two-Generation Family Reunion"

The authors of this paper decided to stop pretending the stars are all the same age. They built a new, more complex model that says: "Okay, let's assume there are two groups of stars here. Group A is older, and Group B is slightly younger. They might also have slightly different chemical recipes (like different amounts of helium)."

They took high-quality photos of 69 of these star clusters from the Hubble Space Telescope and ran them through a super-smart computer program. This program didn't just guess; it used a "hierarchical" method. Think of this like a detective solving a mystery:

• Step 1: It looked at each cluster individually to figure out the age of its two star groups.
• Step 2: It looked at all 69 clusters together to find the "big picture" pattern of how old the oldest stars really are.

3. The Big Surprise: "It Doesn't Matter Much!"

The researchers were worried. They thought, "If we add this extra complexity (two ages instead of one), maybe our calculation of the Universe's age will change drastically. Maybe we'll find the Universe is much younger or much older."

The result? They were wrong to worry.

The paper found that allowing for two generations of stars changed almost nothing about the final answer.

• The Old Way: The oldest stars were about 13.6 billion years old.
• The New Way: The oldest stars are still about 13.6 billion years old.

The difference was so tiny (less than the width of a human hair compared to the size of a football field) that it didn't change the conclusion. This is great news because it means the old, simpler method was actually very robust. It also means the Universe's age estimate is very reliable, even if the stars inside the clusters are a bit more complicated than we thought.

4. The Final Verdict: How Old is the Universe?

By taking the age of the oldest stars (13.61 billion years) and adding a tiny "setup time" (the time it took for the very first stars to form after the Big Bang), the team calculated the age of the Universe:

13.81 Billion Years.

This number is a perfect match for other ways we measure the age of the Universe (like looking at the afterglow of the Big Bang). It also helps settle a debate in astronomy:

• Some measurements suggest the Universe is younger (around 13.5 billion years).
• Some suggest it's older (around 13.8 billion years).

This study confirms that the 13.8 billion year estimate is correct and robust, even when we account for the messy, complex reality of how stars are born in clusters.

The Takeaway

Think of this paper as a quality control check. The scientists said, "We thought our clock was accurate, but let's check if it breaks if we look at the gears more closely." They looked at the gears (the multiple star populations), and the clock didn't break. It kept ticking at the same time.

In short: The Universe is about 13.8 billion years old, and our method of figuring that out is solid, even if the stars inside our cosmic clocks are a bit more complicated than a simple "all born at once" story.

Technical Summary

Here is a detailed technical summary of the paper "The Age of the Universe with Globular Clusters IV: Multiple Stellar Populations" by Valcin et al.

1. Problem Statement

The determination of the age of the Universe ($t_U$) using Galactic Globular Clusters (GCs) serves as a critical, cosmology-independent "chronometer." Previous analyses (e.g., Valcin et al. 2025, "V25") successfully constrained the age of the Universe to $\sim 13.57$ Gyr using a single-stellar-population (SSP) assumption. However, it is well-established that GCs host Multiple Stellar Populations (MSPs)—distinct generations of stars with different chemical abundances (specifically Helium and light elements) and potentially different ages.

The central question addressed in this work is: Does relaxing the SSP assumption to explicitly model two independent stellar populations significantly alter the inferred ages of GCs and, consequently, the derived age of the Universe? There is a concern that ignoring MSP complexity could introduce systematic biases in age determination, potentially exacerbating tensions with the $\Lambda$CDM model (which predicts $t_U \approx 13.8$ Gyr) or the Hubble Tension.

2. Methodology

The authors extend the hierarchical Bayesian framework established in V25 to accommodate a two-population model. The analysis utilizes high-quality Hubble Space Telescope (HST) color-magnitude diagrams (CMDs) for 69 Galactic GCs in the F606W and F814W filters.

Key Methodological Updates:

• Photometric Cleaning & Field Contamination:
  • Implemented a robust, local outlier rejection procedure based on magnitude-dependent CMD consistency rather than global thresholds to preserve intrinsic sequence broadening caused by MSPs.
  • Constructed an empirical background model for field contamination using Kernel Density Estimation (KDE) on rejected stars, rather than assuming an analytic field CMD.
• Multi-Pass Ridgeline Extraction:
  • Adopted a staggered binning strategy (two passes with shifted bin grids) to extract the cluster ridgeline. This reduces sensitivity to arbitrary bin-edge placement and provides a smoother, more robust fiducial line for the likelihood evaluation.
• New Parametrization for Helium and Populations:
  • Instead of sampling individual Helium abundances ($Y_1, Y_2$) which are highly degenerate, the authors introduced a reparametrization:
    • $\bar{Y}$: Mean Helium abundance.
    • $\Delta Y = Y_2 - Y_1$: Helium separation.
    • $p$: Population fraction (fraction of the second generation).
  • This reduces parameter degeneracies and improves MCMC sampling efficiency.
• Weighted Likelihood:
  • The likelihood function is a mixture of two stellar populations and a background component, weighted by the richness (star counts) of each CMD component to prevent sparse regions from dominating the inference.
• Refined Hierarchical Analysis:
  • Crucial Change: In V25, cluster age posteriors were approximated as Gaussians. In this multi-population framework, posteriors exhibit non-Gaussian features (skewness, low-age tails, weak multimodality) due to age-helium degeneracies.
  • The authors now model individual cluster age posteriors as Gaussian Mixture Models (GMMs). These GMMs are then fed into the hierarchical analysis to infer the intrinsic age distribution of the GC population, rather than relying on simple mean/variance summaries.
• Priors:
  • Population fraction $p$ is constrained by a Beta prior informed by the radial distribution of stars (second-generation stars are typically more centrally concentrated).
  • Helium priors are physically motivated, with $\bar{Y}$ constrained by literature values.

3. Key Contributions

1. Explicit MSP Modeling: First application of a full hierarchical Bayesian framework to 69 GCs that explicitly models two independent populations with distinct ages, metallicities, and helium abundances.
2. Robustness Test: Demonstrates that the SSP assumption is a valid approximation for age determination, as the inclusion of MSP complexity yields negligible changes in the final age estimates.
3. Non-Gaussian Hierarchical Inference: Introduces a method to propagate non-Gaussian, skewed age posteriors from individual clusters into the global age distribution, avoiding biases introduced by Gaussian approximations.
4. Simultaneous Constraints: Provides simultaneous measurements of Helium enrichment ($\Delta Y$), population fractions, and metallicities, validating these against independent literature (e.g., chromosome maps).

4. Results

A. Impact on Cluster Ages

• Negligible Shift: Allowing for multiple populations has a negligible impact on age estimates. The ages of the oldest populations remain consistent with the SSP results from V25, with differences at the 0.6$\sigma$ level.
• Robustness: The analysis confirms that the SSP assumption does not introduce significant systematic errors in age determination, even though it simplifies the underlying physics.

B. Intrinsic Age Distribution

• Focusing on the metal-poor subsample ($[Fe/H] < -1.5$, $N=36$), the hierarchical model identifies a dominant old component:
  • Mean Age: $t_{GC} = 13.61 \pm 0.25 \text{ (stat)} \pm 0.23 \text{ (sys)}$ Gyr.
  • The intrinsic distribution is narrower than the observed stacked distribution, indicating that much of the apparent spread is due to measurement uncertainty and internal degeneracies.
  • Two additional, younger components were identified ($\sim 13.0$ Gyr and $\sim 11.3$ Gyr), likely tracing different formation epochs or accreted substructures.

C. Age of the Universe ($t_U$)

• By adopting a conservative delay ($\Delta t = 0.2$ Gyr) between the Big Bang and the formation of the first GCs (based on Population III/II formation models), the authors derive:
  $$t_U = 13.81 \pm 0.25 \text{ (stat)} \pm 0.23 \text{ (sys)} \text{ Gyr}$$
• Consistency: This value is fully consistent with the $\Lambda$CDM prediction ($t_U \approx 13.8$ Gyr) derived from the Cosmic Microwave Background (CMB), even when accounting for the "Hubble Tension" (where local $H_0$ measurements imply a younger universe).
• Extreme Age: The 95th percentile of the age distribution corresponds to $t_{U,95} \approx 14.15$ Gyr, providing a conservative upper bound.

D. Validation of MSP Properties

• Helium: The inferred helium spread ($\Delta Y$) and mean abundance are consistent with literature values, though a small systematic offset ($\sim 0.045$ dex) exists compared to Milone et al. (2018), likely due to methodological differences (chromosome maps vs. full CMD modeling).
• Correlations: The study successfully recovers established astrophysical trends:
  • $\Delta Y$ and mean $Y$ positively correlate with cluster mass and central density.
  • First-population fraction ($f_{P1}$) negatively correlates with cluster mass.
• Age-Metallicity Relation (AMR): The reconstructed AMR preserves the classic bifurcation between in-situ and accreted clusters, validating the physical reliability of the derived ages and metallicities.

5. Significance

• Cosmological Model Independence: This work reinforces globular clusters as a precise, largely model-independent probe of the age of the Universe. The results hold even when relaxing the simplifying SSP assumption.
• Resolution of Tensions: The derived age ($13.81$ Gyr) aligns with CMB predictions, suggesting that the "Hubble Tension" (where local distance ladder measurements suggest a younger universe) is not driven by systematic errors in GC age modeling related to stellar population complexity.
• Methodological Advancement: The transition to a hierarchical framework that handles non-Gaussian posteriors and empirical field contamination sets a new standard for precision stellar chronometry.
• Future Outlook: The robustness of these results supports the use of GCs as a complementary probe to Type Ia supernovae and CMB data, particularly as future JWST observations of high-redshift clusters become available to further refine these constraints.

In conclusion, the paper demonstrates that stellar population complexity does not compromise the reliability of globular clusters as cosmic chronometers, solidifying their role in constraining the age of the Universe and testing cosmological models.