LYRA ultra-faints: The emergence of faint dwarf galaxies in the presence of an early Lyman-Werner background

TL;DR

This study addresses how ultra-faint dwarf galaxies form and survive under uncertain early Lyman-Werner backgrounds by performing 65 high-resolution zoom-in hydrodynamical simulations with $4\,\mathrm{M_{\odot}}$ baryons using the LYRA model. By varying the high-redshift LW background, the authors show that molecular hydrogen cooling enables star formation in halos as small as $M_{200c}\sim10^{5.5}$–$10^{6}\,\mathrm{M_{\odot}}$, while the strength and evolution of the LW field shift the dark-to-luminous transition to higher halo masses, producing distinct SMHM relations and halo occupation fractions for ultra-faint dwarfs. The results reveal a floor in stellar mass around $M_{\ast}\sim10^{3}$–$10^{4}\,\mathrm{M_{\odot}}$ set by early, single bursts of star formation, and a sharp LW-driven break in SMHM near $M_{200c}\sim10^{9}\,\mathrm{M_{\odot}}$ in the strong-background scenario. Overall, the work demonstrates that the faint end of the dwarf galaxy population is a sensitive probe of the early universe, offering a path to constrain Population III star formation and the pre-reionisation SFR through future deep surveys.

Abstract

We present a suite of zoom-in cosmological hydrodynamical simulations of dwarf galaxies using the LYRA galaxy formation model with an extremely high mass resolution of $4\, \mathrm{M_{\odot}}$, evolved to $z=0$. The suite contains 65 haloes selected from Local Group like environments, spanning $M_{\mathrm{200c}}=10^7$ to $5\times10^9\, \mathrm{M_{\odot}}$. The sample includes small ultra-faints with $M_\ast\sim100\, \mathrm{M_{\odot}}$ through to classical dwarfs with $M_\ast \sim 5\times10^6 \mathrm{M_{\odot}}$, as well as haloes that remain dark to the present day. We explore two prescriptions for the high-redshift ($z>7$) Lyman-Werner background (LWB), differing in intensity and redshift evolution. Star formation begins early ($z\gtrsim8$) in progenitors with $M_{\mathrm{200c}}\sim10^5$-$10^6 \mathrm{M_{\odot}}$, where molecular hydrogen enables warm moderate-density gas to efficiently cool. The LWB strongly influences the $z=0$ halo occupation fraction, shifting the dark-to-luminous transition from $M_{\mathrm{200c}}\sim10^7 \mathrm{M_{\odot}}$ (weaker LWB) to $M_{\mathrm{200c}}\sim10^8 \mathrm{M_{\odot}}$ (stronger LWB). Galaxies with $M_\ast\gtrsim10^5 \mathrm{M_{\odot}}$ are mostly insensitive to the LWB choice, whereas lower mass systems respond strongly, producing markedly different stellar mass-halo mass (SMHM) relations. The weaker LWB yields a very shallow SMHM slope with nearly constant scatter, while the stronger LWB introduces a pronounced break at $M_{\mathrm{200c}}\sim10^9 \mathrm{M_{\odot}}$, where haloes of similar mass host galaxies with $M_\ast\sim10^3$ to $10^5 \mathrm{M_{\odot}}$ or remain dark. Both models produce a minimum stellar mass floor at $M_\ast\sim10^3 \mathrm{M_{\odot}}$, originating from galaxies that undergo a single burst of star formation at high redshift before self-quenching from their first supernovae.

Figures (11)

• Figure 1: The left panel shows the redshift evolution of the LWB radiation ($E \approx 11$--$13.6\,\rm{eV}$), quantified through the $\rm{H_2}$ photodissociation rate, $k_{\rm{H_2}}$. In this work we consider two different backgrounds: one using the spectra from FG20 (FG20) that has a negligible LWB prior to reionisation, while the second modifies the FG20 LW amplitude at high redshift to follow the functional form of Inc23, adjusted to match the amplitude of FG20 at $z=7$ (FG20 + modInc23). Throughout the rest of the paper we refer to these two spectra simply as the stronger and weaker LWBs, respectively. The two spectra only differ in the LWB for $z \gtrsim 7$, and share identical ionising spectral intensities (see Section \ref{['section:cooling_tables']} for details). For comparison, we show a number of LWBs from the literature. This includes observationally-inferred backgrounds that are constrained using lower redshift data ($z \lesssim 6$) and extrapolated to higher redshifts HaardtMadauFG20, along with a number of theoretical predictions from (semi-)analytic models and cosmological simulations Ahn_09Wise_12Inc23. There is little consensus in the evolution of the LW at redshifts higher than $z \sim 10$ with significant differences between the models. In the right panels we show the net cooling rates from Ploeckinger_25 at $z=8.2$ for the two LWBs used in this work. Gas in net cooling is shown in the blue colourmap and net heating in red. Overplotted are contours of constant $\rm{H_2}$ mass fraction, and the labels denote $\log\, 2 n_{\mathrm{H_2}}/n_{\mathrm{H}}$. At this redshift, the cooling function is strongly affected, with net cooling occurring at significantly higher densities for cool gas ($T = 10^3$--$10^4$K) in the stronger LWB.
• Figure 2: The present day ($z=0$) stellar mass-halo mass (SMHM) relation from our simulations for the two different LWBs. All central haloes that form stars are shown as circular points, with dark haloes that do not form any stars shown as crosses, and plotted at $M_{\ast} = 100$$\mathrm{M_{\odot}}$. We additionally plot the running median line for the luminous haloes as a solid line. For reference, we plot the extrapolated SMHM relation from Behroozi_13 and Moster_13. The different LWBs have a significant impact on low mass dwarfs ($M_{\ast} \lesssim 10^6$$\mathrm{M_{\odot}}$) and which haloes are able to form stars. In the top panels, we show the star formation histories for four haloes (A, B, C, D, which are shown with larger symbols and a black outline), which encompass the range of observed formation histories. We use these haloes as case studies throughout the paper.
• Figure 3: The halo occupation fraction, defined as the fraction of haloes containing any stars, for central haloes in the two LWBs. For reference, we plot a number of predictions from the literature. The arrows denote the mass where each model predicts $f_{\rm{occ}} = 0.5$. In the left panel we plot models for field haloes today, including the EAGLE and Illustris-TNG50 cosmological simulations and the (semi-)analytic model of BL_20. In the right hand panel we show works modelling Milky Way satellites, which use a satellite's peak halo mass before infall. Additionally in the right panel we show our $z=2$ results for comparison, a typical accretion time for Milky Way satellites. We show the prediction from the semi-analytic models GRUMPY and Galacticus and the purely empirical constraints from Nadler_20 and the Auriga L2 resolution simulations. The line style denotes which cooling channels are considered in these works; solid lines are works that allow for molecular hydrogen cooling while dashed lines are works with only atomic cooling considered. There is a clear dichotomy, with models that only consider atomic cooling predicting the transition from dark to luminous haloes at $M_{\rm{200c}}$$\sim 10^{9}$$\mathrm{M_{\odot}}$, while those modelling H$_2$ cooling and a cold ISM predict $M_{\rm{200c}}$$\lesssim 10^{8}$$\mathrm{M_{\odot}}$. In the absence of a significant LWB prior to reionisation, the transition can be as low as $M_{\rm{200c}}$$\sim 10^{7}$$\mathrm{M_{\odot}}$.
• Figure 4: The halo mass and redshift at which each galaxy first forms stars. This is primarily calculated when the main progenitor first has stars (circular points). For some galaxies, their stars are not born in the main branch and instead accreted much later. For these systems the birth time of the oldest star and its host halo mass is identified, regardless of the merger tree -- we refer to such objects as 'ghost' galaxies (diamond points). The shaded grey region corresponds to the largest halo that has not formed stars at each redshift. We plot the redshift of reionisation and lines of constant virial temperature, $T_{\rm{200c}}$, for reference. In the weaker LWB, SF occurs when haloes reach $M_{\rm{200c}}$$\sim 10^{5.5}$$\mathrm{M_{\odot}}$ up to reionisation, while haloes in the stronger background form stars at similar masses but are restricted to higher redshift ($z \gtrsim 15$) before the LWB becomes sufficiently strong. We plot the halo mass growth for the four case study haloes, along with a smaller halo that remains dark in both models. Grey lines are used when the halo is dark, and black after the halo has formed stars.
• Figure 5: The halo mass growth since reionisation as a function of present day halo mass. The small grey points correspond to dark haloes at $z=0$, while the larger coloured points are haloes that host a galaxy, with the shade denoting the present day stellar mass (see colourbar). The top and bottom panels show the weaker and stronger LWBs, respectively. Overall at a fixed halo mass, earlier forming haloes are more likely to form stars and host a larger galaxy. However, this correlation is not as strong in the weaker LWB.