COSMIC REIONIZATION AFTER PLANCK: COULD QUASARS DO IT ALL?

The reionization of the all-pervading intergalactic medium (IGM) marks a turning point in the history of structure formation in the universe. The details of this process reflect the nature of the first astrophysical sources of radiation and heating as well as the early thermodynamics of cosmic baryons and continue to be the subject of considerable observational and theoretical efforts (for a recent review, see Robertson et al. 2010). Studies of resonant absorption in the spectra of distant quasars show that while inhomogeneous hydrogen reionization may still be ongoing at z ∼ 6, it has fully completed by redshift 5 (e.g., Fan et al. 2006; Becker et al. 2015; McGreer et al. 2015). It is generally agreed that the IGM is kept ionized by the integrated UV emission from active nuclei and star-forming galaxies, but there is still no consensus on the relative contributions of these sources as a function of epoch. The high ionization threshold (4 ryd) and small photoionization cross section of He ii, the rapid recombination rate of He iii, and the fact that most hot stars lack 4 ryd emission, all delay the double ionization of helium. This is expected to be completed by hard UV-emitting quasars and other active galactic nuclei (AGNs) around the apparent peak of their activity at z ≈ 2.5 (e.g., McQuinn et al. 2009; Haardt & Madau 2012), later than the reionization of H i and He i. It is the traditional view that at z > 3, as the declining population of optically bright quasars makes an increasingly small contribution to the 1 ryd radiation background (e.g., Shapiro & Giroux 1987), massive stars in early galactic halos take over and provide the additional ionizing flux needed (e.g., Madau et al. 1999; Meiksin 2005; Faucher-Giguère et al. 2009).

While plausible, this "two-component" picture for cosmic reionization is not necessarily correct and warrants further investigation for validation. In particular, star-forming galaxies at z > 5 can keep the universe substantially ionized only if one extrapolates the steep UV luminosity function (LF) well below the observed limits and assumes a globally averaged absolute escape fraction of Lyman-continuum (LyC) radiation into the early IGM, ${\bar{f}}_{\mathrm{esc}}$, that exceeds 20% (e.g., Bouwens et al. 2012; Finkelstein et al. 2012; Haardt & Madau 2012). Such leakage values are higher than typically inferred from observations of luminous galaxies at z ∼ 3–4 once contamination by foreground low-redshift interlopers is accounted for (Vanzella et al. 2012). Despite significant efforts and the examination of hundreds of galaxies, there exists only a handful of robust detections as of today (e.g., Mostardi et al. 2015; Siana et al. 2015).

The Planck Collaboration (2015) has reported a new, smaller value, τ = 0.066 ± 0.016, of the integrated reionization optical depth from low-multipole polarization, lensing, and high-multipole temperature cosmic microwave background data. This corresponds to a sudden reionization event at $z={8.8}_{-1.4}^{+1.7}$ and reduces the need for a large LyC background at very early times. Together with the recent claim by Giallongo et al. (2015; see also Glikman et al. 2011) of a significant population of faint AGNs at 4 < z < 6.5, these facts have prompted us to reassess a scenario in which quasars and active galaxies may actually dominate the cosmic reionization process at all epochs, with normal star-forming galaxies making only a negligible contribution due to their small leakages. We explore this intriguing possibility below, assuming a (Ω_M, Ω_Λ, Ω_b) =(0.3, 0.7, 0.045) flat cosmology throughout with ${H}_{0}=70\;\;\mathrm{km}\;{{\rm{s}}}^{-1}\;{\mathrm{Mpc}}^{-1}.$

Figure 1 shows the inferred quasar/AGN comoving emissivity at 1 ryd as a function of redshift. Our modeling is based on a limited number of contemporary, optically selected AGN samples (see also Khaire & Srianand 2015 for a similar compilation). All the surveys cited below provide best-fit LF parameters, which are then used to integrate the LF down to the same relative limiting luminosity, L_min/L_⋆ = 0.01. Most of these LFs have faint-end slopes >−1.7, which makes the corresponding volume emissivities rather insensitive to the value of the adopted limiting luminosity. Schulze et al. (2009) combined the Sloan Digital Sky Survey (SDSS) and the Hamburg/ESO survey results into a single z = 0 AGN LF covering four orders of magnitude in luminosity. In the redshift range 0.68 < z < 3.0, the g-band LF of Palanque-Delabrouille et al. (2013) combines SDSS-III and Multiple Mirror Telescope quasar data with the 2SLAQ sample of Croom et al. (2009). The 1 < z < 4 AGN LF by Bongiorno et al. (2007) again merges SDSS data at the bright end with a faint AGN sample from the VIMOS-VLT Deep Survey. The high-redshift quasar LF in the Cosmic Evolution Survey (COSMOS) in the bins 3.1 < z < 3.5 and 3.5 < z < 5 has been investigated by Masters et al. (2012), who find a decrease in the space density of faint quasars by roughly a factor of four from redshift 3–4. A significantly higher number of faint AGNs at z ∼ 4 is found by Glikman et al. (2011) in the NOAO Deep Wide-Field Survey and the Deep Lens Survey and by Giallongo et al. (2015) at z = 4–6 in the CANDELS GOODS-South field. A novel detection criterion is adopted in Giallongo et al. (2015), whereby high-redshift galaxies are first selected in the NIR H-band using photometric redshifts and become AGN candidates if detected in X-rays by Chandra. AGN candidates are found to have X-ray luminosities and rest-frame UV/X-ray luminosity ratios that are typical of Seyfert-like and brighter active nuclei. If correct, these claims suggest that AGNs may be a more significant contributor to the ionizing background radiation than previously estimated.

Zoom In Zoom Out Reset image size

We have converted the integrated optical emissivity inferred from these studies, _λ (in units of $\;\mathrm{erg}\;{{\rm{s}}}^{-1}\;{\mathrm{Mpc}}^{-3}\;{\mathrm{Hz}}^{-1}$), into a 1 ryd emissivity, ₉₁₂, using a power-law spectral energy distribution, ${\epsilon }_{912}={\epsilon }_{\lambda }{(\lambda /912)}^{-{\alpha }_{{\rm{uv}}}}\;{\bar{f}}_{\mathrm{esc}}$, with α_uv = 0.61 following Lusso et al. (2015). We assume an escape fraction of hydrogen-ionizing radiation ${\bar{f}}_{{\rm{esc}}}=1.$ To assess whether a faint AGN population can dominate the cosmic reionization process under reasonable physical assumptions, we adopt in the following an AGN comoving emissivity of the form

$$\begin{eqnarray}&&\mathrm{log}{\epsilon }_{912}(z)=25.15{e}^{-0.0026z}-1.5{e}^{-1.3z},\end{eqnarray} \tag{ 1 }$$

for z < z_QSO, and zero otherwise. Despite the significant scatter in the data points, this function fits reasonably well the z = 0, z < 2.5, and 4 < z < 5 emissivities from Schulze et al. (2009), Bongiorno et al. (2007), and Giallongo et al. (2015), respectively. Note that this emissivity does not drop at high redshift like, e.g., the LyC emissivity of luminous quasars inferred by Hopkins et al. (2007; see Figure 1). It is also higher compared to previous estimates at low redshift, a fact that could contribute to solving the "photon underproduction crisis" of Kollmeier et al. (2014; see also Khaire & Srianand 2015).

Reionization is achieved when ionizing sources have radiated at least one LyC photon per atom and the rate of LyC photon production is sufficient to balance radiative recombinations. Specifically, the time-dependent ionization state of the IGM can be modeled semi-analytically by integrating the "reionization equations" (Shapiro & Giroux 1987; Madau et al. 1999)

$$\begin{eqnarray}&&\displaystyle \frac{{{dQ}}_{{\rm{H}}\;{\rm{II}}}}{{dt}}=\displaystyle \frac{{\dot{n}}_{{\rm{ion,H}}}}{\langle {n}_{{\rm{H}}}\rangle }-\displaystyle \frac{{Q}_{{\rm{H}}\;{\rm{II}}}}{{t}_{{\rm{rec,H}}}}\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&\displaystyle \frac{{{dQ}}_{{\rm{He}}\;{\rm{III}}}}{{dt}}=\displaystyle \frac{{\dot{n}}_{{\rm{ion,He}}}}{\langle {n}_{\mathrm{He}}\rangle }-\displaystyle \frac{{Q}_{{\rm{H}}\;{\rm{III}}}}{{t}_{{\rm{rec,He}}}}\end{eqnarray} \tag{ 3 }$$

for the volume fractions Q of ionized hydrogen and doubly ionized helium. Here, the angle brackets denote a volume average, gas densities are expressed in comoving units, t_rec is a characteristic recombination timescale, and ${\dot{n}}_{{\rm{ion}}}=\int d\nu ({\epsilon }_{\nu }/h\nu )$ is the injection rate density of ionizing radiation, i.e., photons between 1 and 4 ryd in the case of H i (${\dot{n}}_{{\rm{ion,H}}}$) and above 4 ryd for He ii (${\dot{n}}_{{\rm{ion,He}}}$). We do not explicitly follow the transition from neutral to singly ionized helium, as this occurs nearly simultaneously to and cannot be readily decoupled from the reionization of hydrogen.

The ODEs above assume that the mean free path of UV radiation is always much smaller than the horizon and that the absorption of photons above 4 ryd is dominated by He ii. Because only a negligible amount of recombinations occurs in mostly neutral gas, these equations do not explicitly account for the presence of optically thick absorbers that reduce the mean free path of LyC radiation and may further delay reionization (Bolton et al. 2009). They allow, mathematically, for values of Q that are >1, which is physically impossible.

Following the results of cosmological hydrodynamical simulations by Shull et al. (2012; see also Finlator et al. 2012), we define the characteristic hydrogen recombination timescale as

$$\begin{eqnarray}&&{t}_{{\rm{rec,H}}}={[(1+\chi )\langle {n}_{{\rm{H}}}\rangle {(1+z)}^{3}\langle {\alpha }_{{\rm{B}}}(T)\rangle \;{C}_{{\rm{RR}}}]}^{-1},\end{eqnarray} \tag{ 4 }$$

where α_B is the case-B radiative recombination rate coefficient, χ ≡ Y/[4(1 − Y)] = 0.083 allows for the presence of photoelectrons from He ii (here, Y is the primordial helium mass fraction), and ${C}_{{\rm{RR}}}=2.9{[(1+z)/6]}^{-1.1}$ is the clumping factor of ionized hydrogen that accounts for both density and temperature effects on the average recombination rate. Similarly, the recombination timescale of doubly ionized helium is

$$\begin{eqnarray}&&{t}_{{\rm{rec,He}}}={[(1+2\chi )\langle {n}_{{\rm{H}}}\rangle {(1+z)}^{3}Z\langle {\alpha }_{{\rm{B}}}(T/{Z}^{2})\rangle \;{C}_{{\rm{RR}}}]}^{-1},\end{eqnarray} \tag{ 5 }$$

where Z = 2 is the ionic charge and we have assumed that H ii and He iii have similar clumping factors. We have numerically integrated the reionization equations from z_QSO = 12 onward, extrapolating the AGN emissivity in Equation (1) to z_QSO and assuming an EUV power-law spectrum $\propto {\nu }^{-{\alpha }_{{\rm{euv}}}}$ with α_euv = 1.7 (Lusso et al. 2015) and a gas temperature of T = 20,000 K. The integrated electron scattering optical depth can be calculated as

$$\begin{eqnarray}&&\tau (z)=c{\sigma }_{{\rm{T}}}\langle {n}_{{\rm{H}}}\rangle {\displaystyle \int }_{0}^{z}\displaystyle \frac{{(1+{z}^{\prime })}^{2}{{dz}}^{\prime }}{H({z}^{\prime })}[{Q}_{{\rm{H}}\;{\rm{II}}}(1+\chi )+\chi {Q}_{{\rm{H}}\;{\rm{III}}}],\end{eqnarray} \tag{ 6 }$$

where c is the speed of light, σ_T the Thomson cross section, H(z) is the Hubble parameter, and we assumed Q_{He ii} = Q_{H ii} − Q_{He iii}.

Figure 2 shows the resulting ionization history, quantified by Q_{H ii}, Q_{He iii}, and τ(z). The shading shows the effects of changing clumping factor (${C}_{{\rm{RR}}}=9.25-7.21\mathrm{log}(1+z)$; Finlator et al. 2012), IGM temperature (T = 15,000 K), and EUV spectral slope (α_euv = 1.57; Telfer et al. 2002; Stevans et al. 2014). With our default parameters, hydrogen reionization is completed by z ≃ 5.7, helium is doubly ionized by z ≃ 4.2, and the Thomson scattering optical depth is τ ≃ 0.056. The last is consistent with the value reported by the Planck collaboration at the 1σ level. The first agrees with an updated analysis by Becker et al. (2015) of the line of sight variance in the intergalactic Lyα opacity at 4 < z < 6, showing that the data near z = 5.6–5.8 require fluctuations in the volume-weighted hydrogen neutral fraction that are higher than expected from density variations alone. These fluctuations are most likely driven by large-scale variations in the mean free path, a signature of the trailing edge of the cosmic reionization epoch. As shown in Figure 2, our reionization history is also consistent with the dark pixel fraction observed in the Lyα and Lyβ forest of z > 6 quasars, which provides a model-independent upper limit of 1 − Q_{H ii} < 0.11 at z = 5.9 (McGreer et al. 2015).

Zoom In Zoom Out Reset image size

The redshift-dependent fraction of color-selected galaxies revealing Lyα emission has become a valuable constraint on the evolving neutrality of the early IGM. We notice here that in our late reionization model, the hydrogen neutral fraction evolves quite rapidly from redshift 6 (1 − Q_{H ii} = 0.15) to redshift 8 (1 − Q_{H ii} = 0.65). This will cause a fast decline of the mean Lyα line transmissivity of the IGM and may explain the swift drop observed in the space density of Lyα emitting galaxies at z > 6 (e.g., Schenker et al. 2014; Choudhury et al. 2015).

While a late hydrogen reionization epoch is not a unique feature of an AGN-dominated scenario, the early reionization of He ii is. Traditionally, the peak in luminous quasar activity observed at z ≈ 2.5 has been expected to coincide with the end of helium reionization. And while observations of patchy absorption in the He ii Lyα forest at these epochs have been interpreted as signaling the tail end of He ii reionization at z ≤ 2.7 (e.g., Shull et al. 2010 and references therein), the large fluctuations in the background radiation field above 4 ryd predicted from the rarity of bright quasars and the relatively short attenuation lengths of EUV photons make the interpretation of the He ii data still controversial. Indeed, rather than complete Gunn–Peterson absorption at higher redshifts, recent observations have revealed high-transmission regions out to z = 3.5 (Worseck et al. 2015), in conflict with numerical models of He ii reionization driven by luminous quasars (McQuinn et al. 2009; Compostella et al. 2013). According to Worseck et al. (2015), the observed mild evolution with redshift of He ii absorption demands that the bulk of intergalactic helium was already doubly ionized at z > 3.5 by a population of early EUV sources. The high AGN comoving emissivity present at z > 4 in our model may accomplish just that. We find that singly and doubly ionized helium contribute about 13% to τ, and the He iii fraction is already 50% when hydrogen becomes fully reionized at redshift 5.7.

Only the most luminous distant quasars can be detected in surveys such as the SDSS, leaving the contribution of faint AGNs to the early ionizing background highly uncertain. Recent multiwavelength deep surveys have suggested the presence of a hitherto unknown population of faint AGNs (Fiore et al. 2012; Giallongo et al. 2015) and have prompted a reexamination of the role played by AGNs in the reionization of the IGM. We have expanded on previous studies and assessed a model in which the UV radiation responsible for the reionization of hydrogen and helium in the universe arises entirely from quasars and active galaxies. We have assumed here that normal star-forming galaxies will make only a negligible contribution to the AGN LyC emissivity given in Equation (1), which is true provided their globally averaged escape fraction does not exceed a few percent or so (e.g., Madau & Dickinson 2014).

Compared to the standard picture widely discussed in the literature, the AGN-dominated scenario examined in this work completes hydrogen reionization late (z ≲ 6) and double helium reionization early (z ≳ 4) and produces a low electron scattering optical depth that is consistent with the Planck value at the 1σ level. It may provide an explanation to some otherwise puzzling recent findings, from the rapid decline of the space density of Lyα emitting galaxies observed at z > 6 to an IGM whose temperature is found to increase from redshift 5 to 2 and where He ii appears to be predominantly ionized at z ≃ 3.5. It is, of course, a model that is also plagued by a number of uncertainties regarding the properties of faint AGNs (their space densities and EUV spectra at high-z) and of the early IGM (its temperature and clumpiness). As already pointed out by Giallongo et al. (2015), a critical assumption is the high escape fraction of ionizing radiation needed for the global AGN population to dominate reionization. While no discernible continuum edge at 912 Å is seen in a composite FUV spectrum of 159 0 < z < 1.5 quasars and active galaxies obtained with the Cosmic Origins Spectrograph (Stevans et al. 2014), it is unclear whether escape fractions of order unity are also typical of fainter, higher-redshift AGNs. The model is also somewhat sensitive to z_QSO, the "formation epoch" of the earliest AGNs. Here, we have assumed z_QSO = 12 and a comoving AGN emissivity that is only a factor of 3 smaller at z_QSO than at redshift 4. In a model with z_QSO = 9, for example, hydrogen reionization would only be completed by z ≃ 5.4.

As a corroboration of the slowly evolving LyC emissivity adopted in this work, we have run a modified version of our radiative transfer code CUBA (Haardt & Madau 2012) to estimate the intensity and spectrum of the filtered ionizing background—specifically the hydrogen photoionization rate Γ_{H i}—predicted by our AGN-dominated scenario,

$$\begin{eqnarray}&&{{\rm{\Gamma }}}_{{\rm{H}}\;{\rm{I}}}\equiv {\displaystyle \int }_{{\nu }_{{\rm{H}}\;{\rm{I}}}}^{\infty }d\nu \;\displaystyle \frac{4\pi {J}_{\nu }}{h\nu }{\sigma }_{{\rm{H}}\;{\rm{I}}}(\nu ).\end{eqnarray} \tag{ 7 }$$

Here, J_ν is the angle-averaged monochromatic intensity, h is the Planck constant, and σ_{H i} and ν_{H i} are the hydrogen photoionization cross section and ionizing threshold frequency. We have computed the background ionizing intensity J using the emissivity in Equation (1), an EUV spectrum with α_euv = 1.7 and a new a piecewise parameterization of the distribution in redshift and column density of intergalactic absorbers that fits the measurements of the mean free path of 1 ryd photons by Worseck et al. (2014). The results, shown in Figure 3, are in formal agreement with empirical determinations of Γ_{H i} in the interval 2 < z < 5 based on the Lyα opacity of the IGM (Becker & Bolton 2013). Note that, in the traditional view, the slowly evolving emissivity needed to reproduce the Lyα opacity data can only be achieved by carefully tuning the escape fraction from star-forming galaxies to increase rapidly with lookback time, so as to compensate for the decline in the star formation activity toward early epochs (e.g., Haardt & Madau 2012; Kuhlen & Faucher-Giguère 2012).

Zoom In Zoom Out Reset image size

In our AGN-dominated scenario, the heat input from helium reionization will start dominating the thermal balance of the IGM earlier than in the standard picture. In Figure 4, we plot the photoheating rates after the reionization of hydrogen for gas at the mean density,

$$\begin{eqnarray}&&{H}_{i}=\langle {n}_{i}/{n}_{{\rm{H}}}\rangle {\displaystyle \int }_{{\nu }_{i}}^{\infty }{J}_{\nu }{\sigma }_{i}(h\nu -h{\nu }_{i})d\nu /(h\nu ),\end{eqnarray} \tag{ 8 }$$

for species i = H i, He i, or He ii. Here, σ_i and ν_i are the photoionization cross section and threshold frequency for the respective species, and n_i is their abundance (computed integrating the non-equilibrium rate equations for gas at the mean density). The formula above should provide the correct mean heating rate of intergalactic gas once the background intensity J_ν is properly reprocessed while propagating through the IGM (Puchwein et al. 2015). He ii photoheating exceeds the hydrogen term by nearly a factor of three. Measurements of the IGM temperature evolution from redshift 5 to 2 derived from the Lyα forest have been known to be inconsistent with the mononotic decrease with redshift expected after the completion of hydrogen reionization (Becker et al. 2011). This discrepancy may be solved by the additional heating provided by an earlier and more extended period of He ii reionization.

Zoom In Zoom Out Reset image size

The population of faint, high-redshift AGNs invoked here should leave an imprint on the cosmic X-ray background (XRB). Moretti et al. (2012) derived a value of ${J}_{2\;\mathrm{keV}}\;\simeq \;{2.9}_{-1.3}^{+1.6}\times {10}^{-27}\;\;\mathrm{erg}\;{\mathrm{cm}}^{-2}\;{{\rm{s}}}^{-1}\;{\mathrm{Hz}}^{-1}\;{\mathrm{sr}}^{-1}$ to the 2 keV unresolved XRB. The expected contribution at 2 keV from AGNs above redshift z_x can be estimated as (Haardt & Salvaterra 2015)

$$\begin{eqnarray}&&{J}_{2\;\mathrm{keV}}=\displaystyle \frac{c}{4\pi }{\displaystyle \int }_{{z}_{x}}^{\infty }\displaystyle \frac{{dz}}{(1+z)H(z)}{\epsilon }_{2\;\mathrm{keV}}(z){(1+z)}^{-{\alpha }_{x}},\end{eqnarray} \tag{ 9 }$$

where the specific comoving emissivity at 2 keV is related to the LyC emissivity by

$$\begin{eqnarray}&&{\epsilon }_{2\;\mathrm{keV}}(z)={\epsilon }_{912}(z)\times {R}_{{\rm{II}}}{\left(\displaystyle \frac{2500}{912}\right)}^{{\alpha }_{{\rm{uv}}}}{\left(\displaystyle \frac{2\;{\rm{keV}}}{2500\;\mathring{\rm A} }\right)}^{-{\alpha }_{{\rm{ox}}}}.\end{eqnarray} \tag{ 10 }$$

Here, α_ox is the optical-to-X-ray spectral index needed to join the emissivity at 2500 Å with that at 2 keV, and R_II is a correction factor that accounts for the possible contribution of UV obscured ("Type II") AGNs at z ≥ z_x to the observed XRB. Using α_uv = 0.61 as before, α_ox = 1.37 (Lusso et al. 2010), α_x = 0.9, z_x = 5, and R_II = 2 (Merloni et al. 2014), we obtain ${J}_{2\;\mathrm{keV}}=1.64\times {10}^{-27}\;\;\mathrm{erg}\;{\mathrm{cm}}^{-2}\;{{\rm{s}}}^{-1}\;{\mathrm{Hz}}^{-1}\;{\mathrm{sr}}^{-1},$ i.e., a contribution of nearly 60% to the unresolved XRB. We find that z > 5 active galaxies can reionize the universe without overproducing the unresolved XRB provided their properties (i.e., fraction of obscured objects, optical-to-X-ray spectral indices) are similar to those of their lower-redshift counterparts (cf. Dijkstra et al. 2004; Haardt & Salvaterra 2015).

Finally, we conclude by pointing out that in order to promote further testing of this model against new and old observations, we will make the results of our radiative transfer calculations of an AGN-dominated UV background freely available for public use at http://www.ucolick.org/~pmadau/CUBA.

We thank E. Giallongo. Z. Haiman, E. Lusso, and A. Meiksin for helpful discussions on various aspects of this paper. P.M. acknowledges support by the NSF through grant AST-1229745 and NASA through grant NNX12AF87G.