Revealing the ages of metal-rich RR Lyrae via kinematic label transfer

TL;DR

Metal-rich RR Lyrae stars in the Galactic disc are not restricted to ancient epochs. By transferring age information from O-rich Mira variables—whose ages correlate with pulsation periods—and comparing their on-sky velocity distributions with Gaia DR3 data, the study infers the age distribution of metal-rich RRLs as a function of [Fe/H]. Validated against Auriga simulations and applied to a large, cleaned RR Lyrae and Mira sample, the results reveal an age–metallicity trend: higher metallicity RRLs tend to be younger (3–7 Gyr) or even bimodal (4–6 Gyr and 8–9 Gyr) at the highest metallicities, while moderately metal-poor RRLs align with older ages (≈9–11 Gyr). This challenges the canonical view of RR Lyrae as exclusively old and supports formation channels involving enhanced mass loss or binary evolution, reshaping our understanding of stellar evolution and Galactic disc assembly.

Abstract

RR Lyrae stars have long been considered reliable tracers of old, metal-poor populations, primarily due to their prevalence in globular clusters and the Galactic halo. However, the discovery of a metal-rich subpopulation in the Galactic disc, kinematically colder and more rotationally supported, challenges this classical view. Understanding the age of these metal-rich RR Lyrae stars is crucial for constraining their formation pathways and assessing what Galactic populations they are tracing. In this work, we leverage the unprecedented astrometric precision of Gaia DR3 to infer the age distribution of metal-rich RR Lyrae stars through a kinematic comparison with O-rich Mira variables. Mira variables, with their well-established period-age relation, serve as a natural clock, allowing us to transfer age information to RR Lyrae stars via their phase-space properties. By applying this approach across different metallicity bins, we find that the most metal-rich RR Lyrae stars ($[\rm Fe/H] > -0.5$) exhibit kinematics consistent with long-period ($\rm{period}\approx 150\,\rm{days}$), young Mira variable population; its age corresponds to $\sim 6-7$ Gyr (adopting the period-age relation in Zhang & Sanders 2023) that is significantly younger than typically assumed for RR Lyrae stars. In contrast, those with $-1 < [\rm Fe/H] < -0.5$ show properties more aligned with older ($\approx 9-11$ Gyr) populations. Interestingly we also find evidence of a possible double age populations for the most metal-rich RR Lyrae, one younger with ages between 4 and 6 Gyr, and another one older ranging from 8 to 9 Gyr. These results provide strong evidence that metal-rich RR Lyrae stars in the Galactic field do not exclusively trace ancient populations. This finding challenges the current model of RR Lyrae formation and supports alternative formation scenarios, such as binary evolution.

Figures (17)

• Figure 1: Properties of the RRL sample. Top left: distribution of the fractional distance uncertainty. Top middle: metallicity distribution. Top right:$z$-distribution of RRL in various metallicity bins. Bottom left: the spatial face-on ($x$-$y$) distribution of RRL, where the red contours enclosed $75\%$ of the RRL with ${\rm [Fe/H]}>-1$. The red star with a white edge labels the location of the Sun in this coordinate. Bottom middle: the spatial edge-on ($x$-$z$) distribution of the RRL candidates, where the red contours enclosed $75\%$ of the RRL with ${\rm [Fe/H]}>-1$. Bottom right: the mean metallicity of RRLs in the $x$-$z$ plane.
• Figure 2: Top panel: the column-normalised Galactic latitude velocity$-{\rm age}$ distribution of the Mira variable sample. Bottom panel: the same column-normalised distribution of the RRLs that reside within $2$ kpc above or below the Galactic plane.
• Figure 3: Properties of the O-rich Mira sample. Leftmost: fractional distance uncertainty distribution. Middle left: Period (age) distribution of the O-rich Mira variable sample. The age is computed from the characteristic period using the period-age relation in ZS23. Middle right: the spatial face-on ($x$-$y$) distribution of the O-rich Mira candidates. The red star with a white edge labels the location of the Sun in this coordinate. Rightmost: the spatial edge-on ($x$-$z$) distribution of the O-rich Mira candidates.
• Figure 4: Demonstration of the methodology using an idealised simulated galaxy, Au18. Left: the velocity distances between sample A and B ,$p_\mathrm{AB}(\Delta v)$, approximated with a discrete histogram. Middle: Same as the top left but for $p_\mathrm{B'B}(\Delta v)$. Right: The ratio $p_\mathrm{AB}/p_\mathrm{B'B}(\Delta v)$, where the errorbar is propagated from the Poisson uncertainty for the approximated $p_\mathrm{AB}$ and $p_\mathrm{B'B}$. Sample A is fixed as a mono-age population sampled from particles in Au18 with ages between 7 and 9 Gyr. The results with different sample Bs of various ages are denoted in different colours. The Wasserstein scores (Wasserstein distances) of three example pairs are shown in the corresponding colour in the lower right corner. The ratio $p_\mathrm{AB}/p_\mathrm{B'B}(\Delta v)$ is close to unity for all $\Delta v$ when the ages of samples A and B are the same, which has the lowest Wasserstein score.
• Figure 5: Left panels: Test result of the grid-search for modelling the age distribution of sample A from Au18. Each point on the grid represents a sample B constructed with an age distribution of $\mathcal{N(\tau|\mu_B,\,\sigma_B)}$ and is coloured with the logarithmic of the Wasserstein distances between the resulting $p_\mathrm{AB}(\Delta v)$ and $p_\mathrm{B'B}(\Delta v)$. The dots with black circles are those for which the Wasserstein distance is within the smallest $5\%$, and the histograms on the top and right show the $\mu_B$ and $\sigma_B$ distribution associated with these points. The red dashed and red solid lines on the distributions label the 16th, 84th and 50th percentile of the distribution. The black dot with a red circle is the grid point with the smallest Wasserstein distance, which denotes the best-fitted sample B so that ($\mu_B$, $\sigma_B$)$\approx$($\mu_A$, $\sigma_A$). The blue solid lines and the blue square with red edges denote the parameters for the Gaussian age distribution of sample A, i.e. $\mu_A$ and $\sigma_A$. Right: The same as the left set of the plots but for the test on the Mira variables, which verifies the method in the Milky Way environment.