Revised AGN Spectral Model Reveals a More Significant Role in Cosmic Reionization

TL;DR

This study revisits the AGN contribution to cosmic reionization by implementing a physically motivated disk–corona AGN spectral energy distribution (SED) and comparing it with the traditional broken power-law SED. By combining this SED with updated ultraviolet luminosity functions (UVLFs) from JWST-era surveys, it finds that AGNs could contribute a substantially larger share of the ionizing photon budget than previously thought, with faint AGNs playing a pivotal role and the potential to approach or exceed the total budget under plausible assumptions. The enhancement arises from the disk–corona SED’s mass- and accretion-rate–dependent peak, which yields 2–4× higher ionizing efficiency for certain black hole regimes. The findings remain robust across a range of escape fractions and accretion histories but are sensitive to the clumping factor and high-energy photon contributions, underscoring the need for joint constraints on galaxies and faint AGNs during the Epoch of Reionization.

Abstract

Reionization marks one of the most important phase transitions in the history of the Universe, during which neutral baryonic matter was transformed into ionized plasma. While star-forming galaxies are widely regarded as the primary drivers of this process, the extent to which active galactic nuclei (AGNs) contribute remains a subject of ongoing investigation. In this study, we integrate a physically motivated AGN spectral energy distribution (SED) model with state-of-the-art observations to reassess the contribution of AGNs to cosmic reionization. Our findings indicate that adopting a more sophisticated AGN SED model could substantially increase the predicted ionizing photon output by a factor of 3$\sim$4, elevating AGNs to a more significant role ($\approx$20\%) in maintaining reionization than previously estimated. The inclusion of abundant faint AGNs further amplifies this contribution by a factor of a few. These conclusions remain robust across a wide range of accretion rates and ionizing photon escape fractions. Collectively, our results suggest that AGNs may have played a more prominent and previously underestimated role in the reionization of the Universe.

Figures (7)

• Figure 1: SED comparison and ionizing photon number ratio.Left panel: Spectral energy distributions (SEDs) of the broken power-law model Lusso2015 (dashed curves) and the disk-corona model Su2025 (solid curves), color-coded by black hole mass. An Eddington accretion rate is assumed for the disk-corona model. The broken power-law SEDs are normalized to match the UV magnitude ($M_{1450}$) of the corresponding disk-corona cases. The vertical black dash-dotted line marks 1450Å. Right panel: Ionizing photon number ratio between the two models, $\dot{n}_\mathrm{ion,diskcor}/\dot{n}_\mathrm{ion,PL}$, indicated by the color scale. The peak frequency of the disk-corona model shifts away from 1450 Å with decreasing black hole mass and increasing accretion rate. The ratio reaches its maximum at the lowest black hole mass and the highest accretion rate.
• Figure 2: The AGN contribution to reionization. The solid curves represent the AGN contributions to the ionizing photon budget, calculated using the disk-corona model Su2025. An Eddington accretion rate is assumed in computing the disk-corona SEDs. The escape fraction adopted throughout this panel is $f_\mathrm{esc} = 0.75$. The blue and orange curves correspond to the AGN contributions based on the Harikane2023 luminosity functions at $z = 5.76$ and $z = 6.94$, respectively. The green shaded region indicates the range of contributions derived from the Kokorev2024 luminosity function across the redshift interval $6.5 < z < 8.5$. For comparison, the gray dash-dotted curve reproduces the result from Jiang2022, which estimates an AGN contribution of approximately 7% using the ionizing efficiency derived from integrating the broken power-law SED model Lusso2015, also assuming $f_\mathrm{esc} = 0.75$.
• Figure 3: The AGN contribution to reionization: dependence on escape fraction and accretion rate.Left panel: Same as Fig. \ref{['fig:fagn_Edd']}, but adopting a lower escape fraction of $f_\mathrm{esc} = 0.3$. Under this assumption, the type-1 AGN population reported in Harikane2023 still contributes approximately 36-88% to the ionizing photon budget. In contrast, AGN contributions based on the Jiang2022 and Kokorev2024 luminosity functions are reduced to around 10% due to their lower faint-end number densities. The broken power-law model, combined with the Jiang2022 luminosity function, predicts an even smaller contribution of roughly 3%. Right panel: AGN contributions computed using the $z = 5.76$ and $z = 6.94$ luminosity functions from Harikane2023, assuming escape fraction $f_\mathrm{esc} = 0.75$ and two different accretion rates: $\dot{m} = 0.7$ and $\dot{m} = 0.1$. Here, $\dot{m} = 0.7$ represents the lower limit of the Eddington-normalized accretion rate required to reach typical LRD properties, as discussed in the main text; $\dot{m} = 0.1$ corresponds to the lower bound of accretion rates inferred in Harikane2023.
• Figure 4: Comparison between model-predicted and observed SEDs. Observed SEDs of CBS126, Mrk 493, RXJ1007.1+2203, Mrk 1018, Mrk 705, and Ton 1388 are shown as symbols with error bars, with data compiled from Cheng2019Cheng2020. Our disk-corona SED models, with the inferred SMBH mass and Eddington-normalized accretion rate indicated in each panel, are overplotted for comparison Su2025. The gray solid curves represent the broken power-law SED rescaled to match the UV luminosity at 1450 Å. The vertical black dotted line marks 1450 Å, and the cyan-shaded region highlights the wavelength range (2500-600 Å) over which the broken power-law slopes were defined Lusso2015. Overall, the disk-corona models provide a closer match to the observed SED shapes than the broken power-law approximation.
• Figure 5: UV luminosity functions and UV magnitude vs. SMBH mass relation.Left panel: AGN UV luminosity functions compiled from previous observational studies Jiang2022Harikane2023Kokorev2024. The black curve shows the pre-JWST constraints from Jiang2022 (95% confidence level); the green curve corresponds to the luminosity function of JWST-identified "Little Red Dots" reported by Kokorev2024; the blue and orange curves represent the faint type-1 AGN population identified by Harikane2023 at two redshifts. Right panel: UV magnitude versus black hole mass relation for the disk-corona model (solid curve) and the broken power-law model (dash-dotted curve). Black hole masses are derived from the bolometric luminosities of each model, assuming Eddington-limited accretion in both cases and a radiative efficiency of 0.1 for the power-law model. The scatter points show observed UV magnitude–black hole mass measurements compiled from Harikane2023, Maiolino2024, and Kokorev2024. Black hole masses in the Harikane2023 and Maiolino2024 samples are estimated using H$\alpha$ line widths and luminosities. In the Kokorev2024 sample, black hole masses are inferred from bolometric luminosities - calculated assuming a radiative efficiency of 0.1 and Eddington-limited accretion - where the bolometric luminosities are estimated from the rest-frame 5100 Å continuum luminosities obtained via SED fitting.