Mergers lighting the early Universe: enhanced star formation, AGN triggering, and Ly$α$ emission in close pairs at $z=3-9$

TL;DR

This study uses JWST/JADES data to quantify how close companions influence star formation, AGN activity, and Lyα emission in galaxies at z=3–9. By constructing a mass-complete sample of major-merger pairs (rp 5–100 kpc, μ ≥ 0.25) and a rigorously matched control in redshift, mass, and environment, the authors reveal a modest sSFR enhancement (~12% at rp ≲ 20 kpc) on 50–100 Myr timescales, with stronger effects at higher redshift and lower mass. SFH analysis indicates two interaction stages (pre- and post-pericentre) contributing to the observed trends. They find no robust excess of AGN or Lyα emitters in pairs at the studied separations, suggesting that AGN triggering and Lyα visibility require closer encounters or specific environments. Overall, the results support a picture where mergers contribute to mass growth and modest star formation fueling in the early universe, without acting as the dominant driver of AGN activity or Lyα production at the probed scales.

Abstract

Galaxy mergers and interactions are often invoked to explain enhanced star formation, black hole growth, and mass build-up of galaxies at later cosmic times, but their effect is poorly understood at high redshift ($z>2$). We use JADES data to analyse a mass-complete sample of 2095 galaxies at $z=3-9$ with ${\rm log}(M_\star/{\rm M_\odot}) = [8, 10]$, identifying major merger pairs (projected separation of $5-100$ pkpc, mass ratio $\geq 1/4$) using a probabilistic method. To look for signatures of enhancement in multiple physical properties, we carefully build a control sample of non-pairs that are simultaneously matched in redshift, stellar mass, isolation, and environment to the pair sample. We find a moderate enhancement in specific star formation rate (sSFR) of $1.12 \pm 0.05$ at separations $\lesssim 20$ kpc, which is weakly detectable out to $\sim50$ kpc. We find that at longer averaging timescales (50-100 Myr) the sSFR is more affected by interactions and environment, whereas at shorter timescales (5-10 Myr) it is dominated by internal feedback and burstiness. By averaging star formation histories, we find two distinct populations: pre-first passage/coalescence (monotonically rising SFR) and post-pericentre pairs (earlier peak in SFR). Finally, we find no significant excess of AGN in pairs, suggesting galaxy interactions are not effectively triggering black hole activity at separations $>5$ kpc. Similarly, we also do not detect an excess in the fraction of Lyman-$α$ emitters in pairs, implying that at the probed separations, galaxy interactions are not efficient at enhancing Lyman-$α$ photon production and escape, which may only become important at the smallest scales.

Figures (11)

• Figure 1: An example galaxy pair system with its environment (top left), and three closely matched control systems. The primary galaxy is in the centre of each cutout, indicated by a red circle. The secondary galaxy in the pair system is indicated by a blue circle (only present on the top left cutout). The second closest companion in this case is indicated by an orange circle, while in the case of the three control systems, the orange circles indicate the closest companions (at comparable distances to the pair system's second closest neighbour). All other galaxies that are counted in the local density measurement $N_{\rm env}$ are indicated by a white circle. The cutouts shown above are limited to a size of $200 \times 200$ kpc for display purposes (the environment is analysed within a projected physical radius of 1 Mpc). While the galaxies in question are not easily visible, our focus here is on their relative projected positions, using a real example, rather than on their individual appearance.
• Figure 2: Histograms of the distributions of stellar mass ($M_{\star}$), isolation ($r_2$), redshift and local density ($N_{\rm env}$) of the selected galaxy pairs and the matched control sample. These distributions and the corresponding KS test probabilities displayed suggest that the two samples are statistically indistinguishable, meaning that we achieved a robust and representative control sample for our selected galaxy pairs. This allows us to study the differences in the physical properties of the paired galaxies compared to the controls, which is driven by the presence of the close companions.
• Figure 3: Left: Scatter plot of the sSFR excess of the paired galaxies compared to the median of the controls corresponding to each respective pair as a function of projected pair separation. This is calculated for our entire sample of high-probability pairs at redshift $z=3-9$ and stellar masses of $\log_{10}(M_\star / {\rm M_\odot}) = [8, 10]$. The uncertainties on the data points represent the propagated $16^{\rm th}$ and $84^{\rm th}$ percentile values of the ${\rm SFR_{100}}$ posteriors from Prospector. The dashed grey line represents no excess compared to the control median. Right: Bin plot of the ratio between the sSFRs of the paired galaxies and the control medians, where the bins are defined by an adaptive method such that each contains approximately the same number of galaxies, and the associated uncertainties are the standard errors on the median of the points in the bin. As the projected separation between the primary galaxy and its closest companion gets smaller, the positive excess in the sSFR of the paired galaxy increases, up to a factor of $1.12 \pm 0.05$ at $r_{\rm p} \lesssim 20 \ {\rm kpc}$. This suggests that the presence of a close neighbour induces star formation in the primary galaxy. We also display the virial radius of dark matter halo at $z=6$ (approximately the median redshift of our pair sample) of a typical galaxy in our sample, having stellar mass of $\log_{10}(M_\star / {\rm M_\odot}) = 9$ (central value of our stellar mass range).
• Figure 4: Ratio of the stellar age of galaxies in pairs versus controls as a function of projected pair separation. As the separation gets smaller, the stellar age gets shorter compared to the control median. This suggests that the paired galaxies grew more quickly in stellar mass compared to the isolated controls, due to the induced star formation by the approach of the companion galaxies (Fig. \ref{['fig:sSFR_enhancement']}).
• Figure 5: Median sSFR enhancement in paired galaxies relative to controls as a function of projected separation, for sSFRs estimated over different averaging timescales (5, 10, 50, 100, and 2000 Myr). Points are colour–coded from light to dark blue with increasing $t_{\rm avg}$. Adaptive bins contain approximately equal numbers of galaxies; vertical error bars show the standard error on the median in each bin. A clear radial trend (stronger enhancement at smaller $r_{\rm p}$) is observed for $t_{\rm avg}=100$ Myr (our fiducial choice) and is present but weaker at 50 Myr. On very short timescales ($5-10$ Myr), the signal is consistent with zero even at $r_{\rm p}<20$ kpc, consistent with stochastic, bursty star formation driven by internal processes and not influenced by the pair interaction. For $t_{\rm avg}=2000$ Myr, the trend disappears, as this long baseline approximates a constant sSFR $z>3$ by construction.