Dissecting Reionisation with the Cosmic Star Formation and Active Galactic Nuclei Luminosity History

TL;DR

This work connects the cosmic star formation history (CSFH) and cosmic AGN luminosity history (CAGNH) to the cosmic spectral energy distribution (CSED) using the ProSpect SED model to study reionisation. By bracket­ing the ionising emissivity with escape fractions for both stars and AGN, it shows that stellar photons could reionise the IGM by $z\approx6$ if $f_{\mathrm{esc}}$ is sufficiently large (roughly $10-30\%$ depending on metallicity), while AGN alone are unlikely to achieve this even at $f_{\mathrm{esc}}=1$. A hybrid model with $f_{\mathrm{esc}}^{\mathrm{stars}}\approx12\%$ and $f_{\mathrm{esc}}^{\mathrm{AGN}}\approx63\%$ provides a consistent reionisation history, reconciling the CSFH/CAGNH with the Planck $\tau_{\mathrm{CMB}}$ and Gunn-Peterson constraints. The analysis emphasizes that stars likely dominated early reionisation but AGN played a non-negligible role in sustaining ionisation at $z\lesssim6$, while highlighting systematic uncertainties in escape fractions and the clumping factor $\mathcal{C}$ as key limitations for precise inferences.

Abstract

The combination of the $z=0-13.5$ cosmic star formation history and active galactic nuclei (AGN) luminosity history as inferred by the James Webb Space Telescope is connected to the cosmic spectral energy distribution (CSED) to explore the sources of reionisation. We compute the redshift evolution of the corresponding cosmic ionising photon emissivity, the neutral fraction and the cosmic microwave background optical depth. We use the generative SED modelling code ProSpect to bracket the ionising emissivity between escape fractions of $f_{\mathrm{esc}} = 1 - 100\%$ for both the stars and AGN. Stars alone could have achieved reionisation by $z\approx 6$ with $f_{\mathrm{esc}} \gtrsim 30\%$ for solar metallicity ($Z=0.02$) stars or $f_{\mathrm{esc}} \gtrsim 10\%$ for metal-poor ($Z=10^{-4}$) stars. On the other hand, AGN by themselves would have struggled to produce sufficiently many ionising photons even with $f_{\mathrm{esc}} = 100\%$. A hybrid model containing both stars and AGN is explored where we find best fit (median$\pm 1σ$) $f_{\mathrm{esc}}=$ $12\%$ ($14^{+9}_{-7}\%$) for the stars and $f_{\mathrm{esc}}=$ $63\%$ ($60^{+28}_{-32}\%$) for the AGN, maintained at all redshifts. In essence, the joint growth of stellar mass and supermassive black holes produces neither more nor fewer ionising photons than needed to reionise $\gtrsim 99\%$ of the intergalactic medium by $z\approx 6$.

Figures (11)

• Figure 1: Top: CSFH as a function of redshift. The circles with error bars show the results and $1\sigma$ uncertainties from dsilvaGAMADEVILSCosmic2023 at $z<5$. The stars with error bars show the results and $1\sigma$ uncertainties at $z \gtrsim 5$ from dsilvaSelfConsistentJWSTCensus2025. The horizontal error bars show both the $1\sigma$ spread of the redshift distribution in each bin and the width of the bin. The hatched region shows the $16-\nth{84}$ percentile spread of the model fit to the data points from dsilvaSelfConsistentJWSTCensus2025. Middle: CBHARH as a function of redshift. Points and hatches have the same meaning as the top panel. The upward (downward) facing arrows show the lower (upper) limits at $z\gtrsim5.5$ from dsilvaSelfConsistentJWSTCensus2025. The last points of the CBHARH are highlighted with white to signify that they are uncertain and were not used in the model fitting. Bottom: percentage ratio of the model fits to the CBHARH and CSFH. The solid line shows the median and the hatched regions show the $16-\nth{84}$ percentiles propagated from the fits.
• Figure 2: Top left: cosmic ionising emissivity as a function of redshift, $\dot{n}_{\mathrm{ion}}(z)$, from stars only and assuming primordial metallicity, $Z=10^{-4}$. The blue filled region shows minimum/maximum bounds by considering $f_{\mathrm{esc}}=1\%$/$100\%$ with the lighter blue showing the $1\sigma$ uncertainty of the CSFH fit propagated through. Top right: the same as the top left only for solar metallicity, $Z=0.02$. Bottom left: the neutral fraction as per Equation \ref{['eq:madaureionize']} for primordial metallicity. The filled regions have the same meaning as the top panels. The gold star with the horizontal error bar shows the mid-point and $1\sigma$ uncertainty of reionisation inferred from aghanimPlanck2018Results2020b. Bottom right: the same as the bottom left but for solar metallicity.
• Figure 3: Top: cosmic ionising emissivity as a function of redshift, $\dot{n}_{\mathrm{ion}}(z)$, from AGN only. The red filled region minimum/maximum bounds by considering $f_{\mathrm{esc}}=1\%$/$100 \%$ with the lighter red showing the $1 \sigma$ uncertainty of the CAGNH fit propagated through. Bottom: the neutral fraction as per Equation \ref{['eq:madaureionize']}. The filled regions have the same meaning as the top panels. The gold star with the horizontal error bar shows the mid-point and $1\sigma$ uncertainty of reionisation inferred from aghanimPlanck2018Results2020b.
• Figure 4: Optical depth of CMB photons as a function of redshift, $\tau_{\mathrm{CMB}}(<z)$. The black hatched region shows the results for the hybrid model as in Figure \ref{['fig:reionisation_hybrid']}. The red filled region shows the same AGN limits as in Figure \ref{['fig:reionisation_AGN']}. The blue filled region shows the same stellar limits as in Figure \ref{['fig:reionisation_stars']} except that the upper/lower bounds correspond to primordial/solar metallicity to encompass the entire range of stellar chemistry. The filled horizontal band shows the $1\sigma$ range obtained from the Planck analysis of the CMB aghanimPlanck2018Results2020aaghanimPlanck2018Results2020b.
• Figure 5: Corner plot of the posterior distribution of $f_{\mathrm{esc}}$ for both stars (top left) and AGN (lower right) when fitting to the beckerNewMeasurementsIonizing2013$\dot{n}_{\mathrm{ion}}$ at $z\approx4.75$, the CMB optical depth at $z>15$ and the unity ratio of $\dot{n}_{\mathrm{ion}}(z=3)$ from AGN to stars found by smithLymanContinuumEscape2020. The green shaded regions show the uniform prior distributions on $f_{\mathrm{esc}}$. Vertical dashed lines for the 1D histograms are the median and $16-\nth{84}$ percentiles. The median and $16-\nth{84}$ are quoted on top of the histograms, while the best fit values are quoted in the legend. Black contours (bottom left) show the 16, 50 and 84 percentile levels of the posterior samples. The green square in the bottom left panel and the green lines in all panels show the maximum likelihood/best fit values.