Die Hard: The On-Off-Cycle of Galaxies on the Star Formation Main Sequence

TL;DR

This study uses the Magneticum Pathfinder simulations to trace the full star formation histories of central galaxies relative to the SFMS across cosmic time. By classifying histories into four modes—Main Sequence, Quenched, Rejuvenating, and Caught—the authors show that gigayear-scale cycles of quenching and rejuvenation are common, with only a minority evolving steadily along the SFMS to $z\approx0$. The work links these histories to gas accretion geometry and environment, finding that inflow anisotropy correlates with SFR evolution while local density plays a minor role, and demonstrates that a simple Bathtub framework describes high-redshift behavior but not all low-redshift SFMS evolution. The findings imply that the SFMS-evolved state is the exception rather than the rule, with rejuvenation cycles leaving distinctive imprints on stellar ages, metallicities, and central black hole growth.

Abstract

Our picture of galaxy evolution currently assumes that galaxies spend their life on the star formation main sequence (SFMS) until they are eventually quenched. However, recent observations show indications that the full picture might be more complicated. We reveal typical in-situ star formation histories and their relations to large-scale environment as well as gas accretion across cosmic time. We follow the evolution of central galaxies in the highest-resolution box of the Magneticum Pathfinder cosmological hydrodynamical simulations and classify their evolution scenarios with respect to the SFMS. We find that a major fraction of the galaxies undergoes long-term cycles of quenching and rejuvenation on gigayear timescales. This expands the framework of galaxy evolution from a secular evolution to a sequence of multiple active and passive phases. Only 14% of field galaxies on the SFMS at z=0 actually evolved along the scaling relation, while the bulk of star-forming galaxies in the local Universe have undergone cycles of quenching and rejuvenation. In this work we describe the statistics of these galaxy evolution modes and how this impacts their mean stellar masses, ages, and metallicities today. We further explore possible explanations and find that the geometry of gas accretion at the halo outskirts shows a strong correlation with the star formation rate (SFR) evolution, while the density parameter as a tracer of environment shows no significant correlation. A derivation of SFRs from gas accretion with simple assumptions only works in the high-z universe, where accreted gas is quickly converted into stars. We conclude that an evolution scenario consistently on the SFMS is the exception, when regarding galaxies on the SFMS at z=0. Galaxies with rejuvenation cycles can be distinguished well from SFMS-evolved galaxies, both in their halo accretion modes and in their features at z=0.

Figures (15)

• Figure 1: The star formation main sequence at $z \approx 0$. The left- and right-hand panels show the results with respect to two selection criteria for Magneticum galaxies. By imposing a threshold of $\mathrm{sSFR} > 0.2 / t_\mathrm{Hub}$ (left) as used by carnall:2020 (dashed black line), we find a slightly higher linear fit to the logarithmic values (dashed blue line) compared to selecting for all kinematic disks with $\mathrm{SFR}>0\ \mathrm{M}_\odot / \mathrm{yr}$ (right). Results from observations are displayed in magenta, simulations are shown in blue colors. The solid blue line and shaded region show the running median and 50th-percentile region. The circles represent the galaxies from the Magneticum Pathfinder simulations colored according to their morphology as defined by the b-value teklu:2015. Gray circles are ignored for the running median lines on each panel. SFR values below the range of the figure are fixed to the bottom to represent the full sample in this work. The magenta solid line and shaded region highlights a range of $\pm 0.5\ \mathrm{dex}$ around the main sequence fit by pearson:2018, which we used in this work to distinguish between on and off the main sequence. The other magenta lines are additional observational results as dashed speagle:2014, dash-dotted leslie:2020 and dotted popesso:2023. The dashed cyan line marks the results from IllustrisTNG by donnari:2019err.
• Figure 2: Examples of the four formation scenarios for galaxies with respect to specific star formation rate changes over time. The black lines and shades indicate the main sequence values depending on the redshift and stellar mass of a galaxy according to results by pearson:2018 with a scatter of $\pm 0.5\ \mathrm{dex}$ applied to differentiate between on or off the main sequence. The panels show a main-sequence-type galaxy (cyan), an early-quenched galaxy (dark red), a rejuvenating galaxy (magenta) and a caught galaxy (blue) from top to bottom. Two classes are not explicitly represented here. First, the “late-quenched” types only differ from the “early-quenched” by dropping off the main sequence after redshift $z=1$. Second, rejuvenator-main-sequence-type galaxies recover a stable position on the main sequence at redshift $z=0$, unlike rejuvenator-quenched-type galaxies which also go through rejuvenation cycles but do not end up on the main sequence. The vertical lines highlight the snapshots corresponding to each row in \ref{['fig:mst_p1e10']}.
• Figure 3: Star formation main sequence evolution of the Magneticum galaxies traced back from the selection made at z=0 (bottom-row panels). The rows contain snapshots with decreasing redshift from top to bottom between $6<z<0$. From left to right, the different sub-samples “Main Sequence”, “Quenched”, “Rejuvenator” and “Caught” are highlighted by colors in front of the total sample in gray in the background. Light blue data points represent galaxies that have not yet dropped below the main sequence; each quenching event is counted and encoded in the changing color of the data points as red (once), orange (twice) and yellow (three times). The large points are the cherry-picked representative galaxies from \ref{['fig:cherrytrack']}. Quenched representative galaxies are set as half-circles to the bottom of the panels for visibility. The black line and gray shades correspond the main-sequence found by pearson:2018. Gray contours outline the distribution of the full galaxy sample for comparison in each panel.
• Figure 4: Similar to \ref{['fig:cherrytrack']} but focused on the difference between the subclass pairs of early- and late-quenched galaxies as well as quenched and main-sequence rejuvenators. The colored dashed lines represent the cases also shown in \ref{['fig:cherrytrack']}, while the colored solid lines are the subtypes to differentiate. The gray dashed quiescence criteria carnall:2020) and black-shaded SFMS expectations pearson:2018 are derived from the stellar masses of the latter. This causes the pink dashed main-sequence-rejuvenator to not lie on the SFMS at $z\approx 0$ due to the larger stellar mass (see \ref{['fig:mssubclasses_1e10']}) and therefore lower sSFR. The early-quenched galaxy does recover star-formation after quenching but never consistently makes it back onto the main sequence. The late-quenched galaxy also almost makes a passive rejuvenation akin to caught-type galaxies, since the main sequence drops at $z<0.5$. However, it is still mostly more than 0.5 dex below the main sequence averaged over a smoothing window of one gigayear, ruling out a rejuvenation scenario. The main-sequence rejuvenator undergoes a more pronounced quenched phase than the quenched rejuvenator.
• Figure 5: Similar to \ref{['fig:mst_p1e10']} but focused on the difference between the subclass pairs of early- and late-quenched galaxies as well as quenched and main-sequence rejuvenators. In this version, the colors refer to the subclass types instead of quenching incidents. The large points show the cherry picked representative galaxies from \ref{['fig:subclass_cherries']}, again set to the bottom of the panels as half-circles for visibility in case of low SFR values. By construction, the difference between early- and late-quenched is most pronounced at $z=1$. At $z\approx 0$, the non-zero distributions are barely noticeable. Analogously, the difference between the two rejuvenator subclasses is the sharpest at $z\approx 0$. Both subclasses undergo rejuvenation cycles at different times, but the relation to the SFMS determines the classification.