The PARADIGM project II: The lifetimes and quenching of satellites in Milky Way-mass haloes

TL;DR

The paper uses the PARADIGM suite to study satellite lifetimes, disruption, and quenching around Milky Way–mass haloes by running two contrasting galaxy formation models (VINTERGATAN and IllustrisTNG) on identical initial conditions across varied halo assembly histories. It finds that disruption fractions are high for satellites accreted early and decline toward present day, with typical disruption timescales of roughly 2–8 Gyr after accretion; pericentric distance and orbital period are strong predictors of disruption, while halo formation time mainly modulates satellite abundance rather than disruption likelihood. A major result is that differences in satellite counts between VG and TNG are predominantly driven by their distinct stellar-to-hhalo mass relations, with VG overproducing satellites at all masses due to early, intense star formation, though disruption rates still show meaningful concordance when accounting for SMHM disparities. The study demonstrates that despite different physical implementations, robust convergent trends emerge for satellite disruption and quenching, underscoring the potential to constrain disruption timescales from MW-like systems and informing future observations of stellar haloes and satellite quenching pathways.

Abstract

The abundance and star-formation histories of satellites of Milky Way (MW)-like galaxies are linked to their hosts' assembly histories. To explore this connection, we use the PARADIGM suite of zoom-in hydrodynamical simulations of MW-mass haloes, evolving the same initial conditions spanning various halo assembly histories with the VINTERGATAN and IllustrisTNG models. Our VINTERGATAN simulations overpredict the number of satellites compared to observations (and to IllustrisTNG) due to a higher $M_{*}$ at fixed $M_{\rm tot}$. Despite this difference, the two models show good qualitative agreement for both satellite disruption fractions and timescales, and quenching. The number of satellites rises rapidly until $z=1$ and then remains nearly constant. The fraction of satellites from each epoch that are disrupted by $z=0$ decreases steadily from nearly 100% to 0% during $4>z>0.1$. These fractions are higher for VINTERGATAN than IllustrisTNG, except for massive satellites ($M_{*}>10^{7}\,M_{\odot}$) at $z>0.5$. This difference is largely due to varying distributions of pericentric distance, orbital period and number of orbits, in turn determined by which sub(haloes) are populated with galaxies by the two models. The time between accretion and disruption also remains approximately constant over $2>z>0.3$ at $6-8$ Gyr. For surviving satellites at $z=0$, both models recover the observed trend of massive satellites quenching more recently ($<8$ Gyr ago) and within $1.5\,r_{\rm 200c}$ of the host, while low mass satellites quench earlier and often outside the host. Our results provide constraints on satellite accretion, quenching and disruption timescales, while highlighting the convergent trends from two very different galaxy formation models.

Figures (16)

• Figure 1: Evolution of the host mass $M_{\rm 200c}$ until $z=1$ for all haloes in our sample. After this epoch, the masses grow gradually by 0.1-0.25 dex to their final values. The EF, FM and LF haloes have starkly different growth histories. The EF grows rapidly at early times and has a quiet merger history after $z=4$, while LF grows more slowly and through multiple mergers. The FM halo has an intermediate growth history, with a halo formation time intermediate to the former two haloes, significant mergers occurring at $z\approx2$ and a quiet merger history after. The four GM haloes behave similarly to the FM halo until $z\approx3$, when the impact of the GMs are seen on two significant mergers. Finally, the FM-EC halo, whose ICs were modified from the FM ICs to have an earlier collapse time has a similar growth history to the EF halo at early times, but slower growth after $z=3$.
• Figure 2: Mock stellar images of the EF, FM and LF haloes at $z=0$ from both sets of simulations in our sample. Each galaxy has been rotated to be face-on. The colours correspond to Johnsons U, V and I bands and encompass $20-32$mag/arcsec$^{2}$ in surface brightness. We use this faint lower limit to highlight the satellites of the haloes. A simple dust correction has been applied following the Calzetti2000 law. Each image is 300 kpc on a side and uses a projected image of depth 300 kpc. The TNG central galaxies are always larger than the VG ones. Each VG halo has significantly more satellites than its TNG counterpart, and VG bright satellites appear more concentrated than TNG ones.
• Figure 3: Luminosity functions (LFs) of satellites within 300 kpc of each simulation at $z=0$, indicated by the different colours. V-band magnitudes were calculated within a radius $R_{\rm SB=30}$, where the V-band surface brightness of the satellite is 30 mag/arcsec$^{-2}$. The corresponding stellar mass (according to a linear fit to our data) is indicated by the top x-axis. The VG simulations have significantly more satellites at all masses/luminosities compared to the TNG simulations. For comparison, the dashed and dotted black curves show the observed MW and M31 LFs from McConnachie2012, while the shaded region and solid lines indicate the spread and median LFs of a matched sample from the SAGA Survey DR3 Mao2024. The TNG LFs are in good agreement with the observed MW LF and within the spread of LFs from the SAGA survey, albeit with marginally fewer bright satellites than average. The VG simulations overpredict the number of satellites at all but the brightest magnitudes. This is found to be due to the formation of several satellites at early times that produce most if not all of their stellar mass through a single burst of SF (see Section \ref{['sec:resultsZ0LFs']} for details).
• Figure 4: Numbers of satellites within $r_{\rm 200c}$ as a function of epoch across all assembly histories. The data points show the median value for all simulations generated with the specified code (blue diamonds for TNG and gold circles for VG), while the corresponding errorbars show the 16$^{\rm th}$-$84^{\rm th}$ percentile ranges. The left and right panels show the results for low- and high-mass satellites respectively. Note that the TNG/VG datapoints have been shifted to the left/right by a small amount for clarity. The VG simulations have significantly higher numbers of satellites at all times compared to the TNG ones, and have more massive satellites than lower mass ones, while the opposite is true for the TNG satellites. In both sets of simulations however, there is an initial increase in numbers of satellites at early times between $z=4-1.5$, after which the total numbers of satellites remains approximately constant on average.
• Figure 5: Fraction of satellites at a given epoch that have been disrupted by $z=0$, across all assembly histories. Here we define the time of disruption as the time when $R_{50}/R_{\rm 50,ini}>5$, where $R_{50}$ is the stellar 3D half-mass radius and $R_{\rm 50,ini}$ is this radius at the selection epoch. Colours and symbols are as in Fig. \ref{['fig:Nsats1Rvir']}. Over 70 per cent, if not all, satellites from $z\geq1$ ($\gtrsim 8$ Gyr ago) have been disrupted by $z=0$, regardless of mass and in both sets of simulations. After $z=2$, this fraction decreases gradually, to less than 30 per cent by $z=0.2$ i.e. $\sim 2.5$ Gyr ago. High mass satellites are less likely to disrupt compared to low-mass ones and VG low mass satellites are more likely to be disrupted than TNG ones, with the exception of high-mass satellites at $z>0.5$, where both these trends are reversed. The decreasing disruption fractions after $z\sim1$ are to be expected due to the increasingly limited time available, assuming the disruption timescales are comparable.