Black holes in the low-mass galaxy regime: imprint of AGN feedback on the circumgalactic medium of central dwarf galaxies

TL;DR

This paper investigates how AGN feedback affects the circumgalactic medium of central dwarf galaxies in the TNG50-1 simulation. By selecting dwarfs with 8 ≤ log(M*/M⊙) ≤ 9.5 and identifying AGN via a threshold on the Eddington ratio, the authors compare active dwarfs to carefully matched inactive controls to isolate the impact of black hole activity. They find that AGN hosts have significantly less neutral gas (≈3.9× less M_neutral and ≈4.8× less M_HI) and more extended gas halos (gas half-mass radius > ~10 kpc larger), with a modest reduction in sSFR and little change in local environment; these effects stem from the high-accretion thermal feedback mode in IllustrisTNG. However, the AGN fraction in the simulation is highly sensitive to the chosen λ_Edd threshold (ranging from ~1% to ~24%), and TNG50-1 may overpredict active-dwarf fractions relative to observations, motivating observational tests to constrain BH seeding and feedback in the low-mass regime.

Abstract

Active galactic nuclei (AGN) have been observed in dwarf galaxies, yet the impact of black hole feedback in these low-mass systems remains unclear. To uncover the potential effects of AGN in the low-mass galaxy regime, we study the properties and demographics of active dwarf galaxies at $z=0$, using the IllustrisTNG simulations. We use data from the TNG50-1 simulation, selecting central galaxies with stellar masses in the range $8 \leq \log(M_\ast/{\rm M_\odot}) \leq 9.5$, and selecting AGN based on their Eddington ratio ($λ_{\rm Edd}$). We analyzed the properties and environment of AGN host galaxies and compared them with inactive control galaxies. The AGN fractions found in the simulation depend strongly on the threshold for $λ_{\rm Edd}$ in the AGN selection, ranging from $\sim$ 1\% ($λ_{\rm Edd} \geq 0.05$) to $\sim$ 24\% ($λ_{\rm Edd} \geq 0.01$). In comparison with non-AGN galaxies of similar stellar and halo mass, dwarf AGN hosts are deficient in neutral gas, having $\sim$ 3.9 times less neutral mass, in qualitative agreement with observations. The dearth in neutral gas is stronger beyond two stellar half-mass radii ($r \gtrsim 3$ kpc), and AGN hosts have more extended gas components than non-AGN galaxies, with a gas half-mass radius, on average, $\gtrsim$ 10 kpc larger. AGN hosts are also slightly less star-forming, but have no differences in local environment. We found that AGN can significantly decrease the neutral gas component of dwarf galaxies, a direct effect of the high-accretion feedback mode employed in IllustrisTNG. However, it is important to test our findings with observations to unveil the complete role of AGN in dwarf galaxies. In TNG50, dwarf AGN fractions are an order of magnitude larger than those observed, motivating a detailed investigation to precisely quantify the mismatch between simulations and observations.

Figures (21)

• Figure 1: Joint distribution of stellar ($M_{\ast}$) and halo masses ($M_{\rm 200c}$) for TNG50-1 central dwarf galaxies at $z=0$. All 3297 dwarf galaxies are shown as blue squares, while the 789 dwarf galaxies hosting AGN are shown as red diamonds. Density histograms represent the marginal distributions of samples in each axis.
• Figure 2: Scheme to illustrate the samples referred to in this work. Numbers in parentheses indicate the number of galaxies. From the main sample of dwarf central galaxies in TNG50-1 (white ellipse), we create a sample of AGN hosts (red ellipse). We also create three parent samples of non-AGN galaxies (grey ellipses), one sample of galaxies with no restriction (NR) on their properties, another of galaxies with the restriction of having gas (WG), and another of galaxies with the restriction of having black holes (WBH). Specific non-AGN control samples (pink, green, and blue boxes) are thus drawn from these parent samples, while the corresponding AGN samples (orange boxes) are drawn from the sample of active dwarf galaxies (red ellipse). Dashed lines indicate the pairing of a specific non-AGN control sample and its respective AGN samples, following the method described in Section \ref{['sec:control_sample']}. The colors of the boxes of control samples indicate the variables used as control, and correspond to the same colors used in Figure \ref{['fig:boxplots1']}.
• Figure 3: Fraction of central dwarf galaxies hosting AGN according to different minimum thresholds for variables used in the sample selection. Top panel: AGN fraction as a function of the minimum threshold for $\lambda_{\rm Edd}$. Bottom panel: AGN fraction as a function of the minimum threshold for stellar mass. In each panel, the solid lines show the fraction for different values of $\lambda_{\rm Edd, min}$ or $M_{\ast, \rm min}$, with the shaded regions representing the 95% confidence intervals. The red dashed lines indicate the overall fraction of dwarf AGN candidates from the DESI survey Pucha2025, for the whole mass range of $8 \lesssim \log (M_\ast/{\rm M_\odot}) \leq 9.5$. This value ($\sim$ 1.2%) is derived from the values presented in their figure 7 (right panel), and is the fraction of dwarf galaxies identified as AGN/Composite in the [Nii]-BPT diagram, over the total number of dwarf galaxies (including dwarf galaxies without emission lines).
• Figure 4: Box plots of AGN (orange) and non-AGN control (pink, green, blue) samples properties. Each box extends from the first to the third quartile of the data, with a line at the median. The whiskers extend from the box edges to the farthest data point lying within 1.5 times the inter-quartile range from the box. Data past the end of the whiskers are shown as empty black circles. AGN samples are represented by orange boxes, control samples paired by $M_\ast$, $M_\ast$ & SFR or $M_\ast$ & $M_{\rm 200c}$ are represented by pink, green, and blue boxes, respectively. Different hatch patterns indicate the different parent samples of non-AGN central galaxies from which the control samples were constructed: no pattern, "x" pattern, and dot pattern refer to the "No Restriction" (NR), "With Gas" (WG), and "With BH" (WBH) samples, respectively. The legend above the upper panel indicates the color and hatch pattern for each control sample shown in the figure. In each panel, above the boxes, we show the number of AGN hosts ($N_{\rm AGN}$), the difference between the median of the AGN and control ($\Delta_{\rm med}$), and the p-value of an Anderson-Darling two-sample test performed with the control and AGN paired samples. In the cases where the difference was considered statistically significant (p-value $< 0.05$), the upper text is highlighted as bold and enclosed in a red box. From top to bottom, the panels show box plots for host halo mass, neutral-to-total gas ratio, and gas half-mass radius.
• Figure 5: Same as Figure \ref{['fig:boxplots1']}. From top to bottom, the panels show box plots for: specific star formation rate, stellar metallicity, median stellar formation time, and distance to the 1st nearest neighbor (with $M_\ast \geq 10^8 \ \rm M_\odot$).