CORN -- Chronometers of Relic Nature I: The first estimate of the expansion rate of the Universe using compact relic galaxies

Context. Measuring the expansion rate of the Universe is a central challenge in cosmology, particularly due to the persistent statistically significant discrepancy between the measurements of the Hubble constant, H0, obtained using early- and late-time probes, commonly known as the Hubble tension. Independent and robust methods are therefore essential to validate existing measurements and assess potential systematic effects.

Aims. The authors employ the cosmic chronometer (CC) approach, which uses the differential age evolution of quiescent galaxies to measure the Hubble parameter H(z) without assuming an underlying cosmological model. In contrast to previous CC studies, they restrict the analysis to relics – ultra compact massive galaxies (UCMGs) hosting the oldest stellar populations – thereby minimising uncertainties related to star formation history and merger history.

Methods. The authors select a sample of 189 relic galaxies in the redshift range 0.07 ≤ z ≤ 0.22 from the E-INSPIRE UCMGs catalogue. Using the Dn 4000 spectral index of relic galaxies and its redshift evolution, combined with MILES stellar population synthesis models, they derive the differential age relation required to infer H(z). They account not only for metallicity effects but also for the impact of α-element enhancement, which has not been explicitly propagated into the systematic uncertainty budget of the Dn 4000 cosmic chronometer method.

Results. The authors obtain an independent measurement of H(z = 0.15) = 85 ± 53 km s−1 Mpc−1. Although the total uncertainty remains large and dominated by statistical limitations, the systematic component is 13% (8.7% when α-enrichment is not propagated), among the tightest estimates in current CC studies. They find that α-enrichment plays a major role in the systematic error budget, particularly in the high metallicity regime, and must be properly accounted for in future analyses.

Conclusions. The authors demonstrate the proof of concept that relic galaxies provide a promising pathway to reduce systematic uncertainties in the CC method. The statistical uncertainties, which dominate the error budget, can be significantly reduced by the increase of data volume expected during the next years with Euclid, Vera C. Rubin Observatory, DESI, and 4MOST.

Key results in detail:

• The final sample consists of 189 galaxies with 0.07 ≤ z ≤ 0.22, M⋆ ≥ 1010·8 M⊙, σDoR ≤ 0.2, and DoR ≳ 0.4, selected from the E-INSPIRE UCMGs catalogue.

• The median mass fraction assembled within 3 Gyr after the Big Bang for the final sample is fM⋆tBB=3 = 0.95 ± 0.05, compared to 0.92 ± 0.08 for the full UCMG sample. This is substantially higher than the 0.68 and 0.38 found for typical massive quiescent galaxies from LEGA-C using PROSPECTOR and BagPipes, respectively, and 0.39–0.49 for massive quiescent galaxies from the SIMBA simulation.

• The Dn 4000–redshift slope is measured as dDn 4000/dz = −0.33 ± 0.19 [fit] ± 0.07 [bin], with a total statistical uncertainty of σstat = 0.2, i.e. 61% of the slope value.

• The final sample is characterised by a median metallicity of Z = 0.034 ± 0.003 and a median α-enhancement of [α/Fe] = 0.20 ± 0.12.

• The calibration parameter A(Z, [α/Fe]) = 0.033+0·003−0.002, giving δσZ,α = +9%−6%. When propagating only metallicity uncertainties, A(Z) = 0.0333+0·0002−0.0004, translating into δσZ = +0·6%−1.2%.

• The total relative systematic uncertainty is +13%−11% (including [α/Fe]), or ≤ 8.7% when [α/Fe] variations are neglected.

• The metallicity-related uncertainty of δσZ ≤ 1.2% is lower by approximately a factor of 2.5 than the most restrictive estimates currently available in the literature (Loubser et al. 2025), and almost an order of magnitude smaller than the uncertainties typically expected for the general population of massive passive galaxies (10%; Moresco et al. 2020).

• The origin of this reduction is that the Z of the relic sample occupy a very specific region of the A(Z) relation, close to the turnover point where the slope A ≡ dDn 4000/dt becomes nearly insensitive to Z variations.

• An additional contribution of approximately δσα ≃ 9% should be included whenever [α/Fe] variations are not explicitly modelled.

• Only 30% of the final relic sample satisfies all classical CC criteria, although this fraction increases to 87% when the velocity dispersion requirement is omitted. The principal difference arises from the velocity dispersion requirement, which excludes more than half of the relic sample.

• The measurement is fully consistent, within uncertainties, with previous CC measurements as well as with the prediction of the flat ΛCDM cosmology inferred from Planck.

• The result should be interpreted primarily as a proof-of-concept demonstrating that relic galaxies constitute one of the most promising populations currently available for the CC method.

Improvements to AI systems:

1. Uncertainty-Aware Stellar Population Modeling: Enhance AI models used for galaxy spectral fitting to explicitly incorporate α-element enhancement ([α/Fe]) as a free parameter alongside metallicity, rather than assuming solar-scaled abundances. This reduces systematic biases in age and mass estimates for high-metallicity, old stellar populations.

2. Calibration Curve Exploitation: Train AI systems to identify and leverage the turnover region in calibration relations (e.g., A(Z) vs. Z) where observables become insensitive to nuisance parameters. This allows automatic selection of galaxy samples that minimize systematic uncertainties without manual tuning.

3. Selection Criterion Optimization: Develop AI classifiers that predict which galaxies are optimal cosmic chronometers based on multi-dimensional criteria (mass, velocity dispersion, Dn4000, star formation history), rather than relying on rigid thresholds. This can increase usable sample sizes by 30% while maintaining systematic control.

4. Differential Age Inference: Implement neural networks that directly infer H(z) from differential age measurements of quiescent galaxies, with built-in propagation of correlated uncertainties from spectral indices, metallicity, and α-enhancement. This avoids model-dependent assumptions and provides robust error bars.

5. Systematic Error Budgeting: Create AI frameworks that automatically decompose total uncertainty into statistical, metallicity, α-enhancement, and binning components, enabling real-time identification of dominant error sources and guiding future observational strategies.

What the improved AI system can do:

• Predict H(z) with 8–13% systematic precision from small samples of relic galaxies, rivaling current cosmic chronometer methods but with explicit handling of α-enhancement.

• Automatically select optimal galaxy samples for cosmological measurements by scoring galaxies on their suitability as cosmic chronometers, reducing manual vetting and increasing yield.

• Provide calibrated uncertainty estimates that distinguish between reducible (statistical, sample size) and irreducible (α-enhancement, metallicity) errors, enabling targeted follow-up observations.

• Forecast the impact of upcoming surveys (Euclid, DESI, Rubin) on H(z) precision by simulating sample growth and propagating expected statistical improvements.

• Cross-validate cosmological models (e.g., ΛCDM vs. alternative) using independent, model-agnostic H(z) measurements with quantified systematics, directly addressing the Hubble tension.

Measuring the expansion rate of the Universe is a central challenge in cosmology. Independent and robust methods are essential to validate existing measurements and assess potential systematic effects. We employ the cosmic chronometer (CC) approach, which uses the differential age evolution of quiescent galaxies to measure the Hubble parameter H(z) without assuming an underlying cosmological model. In contrast to previous CC studies, we restrict our analysis to relics -- ultra compact massive galaxies (UCMGs) hosting the oldest stellar populations -- thereby minimising uncertainties related to star formation history and merger history. We select a sample of 189 relic galaxies in the redshift range 0.07< z <0.22 from the E-INSPIRE UCMGs catalogue. Using the D n4000 spectral index of relic galaxies and its redshift evolution, combined with MILES stellar population synthesis models, we derive the differential age relation required to infer H(z). We account not only for metallicity effects but also for the impact of alpha-element enhancement, which has not been explicitly propagated into the systematic uncertainty budget of the D n4000 cosmic chronometer method. We obtain an independent measurement of H(z=0.15) = 85 plus or minus53 km/s/Mpc. The total uncertainty is dominated by statistical limitations. The systematic component is 13% (8.7% when alpha-enrichment is not propagated), among the tightest estimates in current CC studies. We find that alpha-enrichment plays a major role in the systematic error budget, particularly in the high metallicity regime, and must be properly accounted for in future analyses. We demonstrate the proof of concept that relic galaxies provide a promising pathway to reduce systematic uncertainties in the CC method. The statistical uncertainties, which dominate the error budget, can be significantly reduced by the increase of data volume expected during the next years.