Galactic bars and active galactic nucleus fuelling in the second half of cosmic history

TL;DR

This study tests whether galactic bars in disc galaxies fuel Active Galactic Nuclei (AGN) through secular processes across the second half of cosmic history, up to $z\sim0.8$. It combines a Deep Learning bar classifier trained on Galaxy Zoo classifications with multi-wavelength AGN diagnostics (MIR, X-ray, and SED-fitting) to compare barred and carefully matched unbarred discs. The results show a robust bar-associated enhancement in AGN incidence for MIR and X-ray tracers, but only a modest effect for SED-selected AGN, and a near-absence of bars in the most luminous or dominant accretion regimes ($f_{AGN}>0.75$ or $L_{disc}>10^{44.5}$ erg s$^{-1}$). Overall, bars appear to fuel low-to-moderate luminosity AGN, while major mergers remain the primary mechanism for triggering the most powerful AGN, supporting a dual-mode fueling scenario with significant implications for galaxy–SMBH co-evolution and future surveys such as Euclid.

Abstract

We investigate the role of galactic bars in fuelling and triggering Active Galactic Nucleus (AGN) in disc galaxies up to $z\sim 0.8$. We utilise a Deep Learning model, fine-tuned on Galaxy Zoo volunteer classifications, to identify (strongly and weakly) barred and unbarred disc galaxies in Hyper Suprime-Cam Subaru Strategic Program $i$-band images. We select AGN using three independent diagnostics: mid-infrared colours, X-ray detections, and spectral energy distribution (SED) fitting. The SED analysis, performed using CIGALE, quantifies the relative AGN contribution to the total galaxy luminosity ($f_{\rm AGN}$) and the AGN luminosity ($L_{\rm disc}$). We assess the impact of bars by comparing AGN incidence and properties in barred galaxies against carefully constructed redshift-, stellar mass-, and colour-matched unbarred control samples. Our binary AGN classification experiment demonstrates that barred disc galaxies host a statistically detectable higher fraction of AGN compared to their unbarred counterparts, suggesting a contributing role for bars in the global AGN budget. The contribution of bars to AGN fuelling appears confined to systems where the AGN has a lower relative contribution to the host galaxy's emission ($f_{\rm AGN} < 0.75$). Crucially, we find a significant dearth of barred disc galaxies hosting AGN with $f_{\rm AGN} > 0.75$, independent of bar strength. Consistent with this, the fraction of barred galaxies among AGN hosts decreases with increasing $L_{\rm disc}$. Combined with previous results, we suggest that bars contribute to fuelling the population of low-to-moderate luminosity AGN, but major mergers are the principal mechanism for triggering the most powerful and dominant accretion events.

Figures (19)

• Figure 1: 2D histogram (linear scaling) representation of the stellar mass--redshift distribution for the lamarcaDustPowerUnravelling2024 parent sample used in this work. The marginal histograms display the individual $z$ and $M_{\star}$ distributions.
• Figure 2: A simplified representation of the relevant questions from the Galaxy Zoo decision tree willett_galaxy_2013 used to classify galaxies for the training set.
• Figure 3: Examples of GZ training galaxies for each class. Top row shows galaxies with $z<0.5$, bottom row shows galaxies at $z\geq0.5$. Each cutout has a size of $8\times 8$ Kron radius (measured in the $i$-band). Images are displayed using a logarithmic scaling.
• Figure 4: Confusion matrix for the fine-tuned Zoobot model, evaluated on the balanced test set. The matrix is colour-coded relative to the total number of galaxies in each true morphological class (rows). Each cell displays the raw count of galaxies, the column-wise percentage (precision, yellow boldface text on the diagonal), the row-wise percentage (recall, orange text in brackets on the diagonal).
• Figure 5: Confusion matrices for the sample selected using the criteria described in Sect. \ref{['sect:bar-selection']}, colour-coded according to the total number of galaxies in each row. The content of each cell is the same as in Fig. \ref{['fig:confusion_all_classes']}. Top: Confusion matrix considering all bars as a single class ($A_{\rm bar}$). We selected the same number of $A_{\rm bar}$ and $U_{\rm bar}$ examples. Lower: Confusion matrix dividing the bars into $S_{\rm bar}$ and $W_{\rm bar}$.