Probing the Metal Enrichment of the Intergalactic Medium at z = 5–6 Using the Hubble Space Telescope

The C iv absorbers at z = 5–6 were selected from the sample of Simcoe et al. (2011). From this sample, we select four secure, strong C iv absorbers at z = 4.87–5.74 in two QSO sightlines: SDSS J1030+0524 (z_QSO  = 6.30) and SDSS J1306+0356 (z_QSO  = 6.00). The C iv column densities range from 13.8 to 14.8 cm⁻². These two QSO fields have broadband (F775W, F850LP) images available from Stiavelli et al. (2005). The sample is summarized in Table 1.

Table 1.  Summary of the C iv-associated Lyα Emitter Candidates

R.A.	Decl.	zC iv	$\mathrm{Log}\left[\tfrac{{N}_{{\rm{C}}{\rm{IV}}}}{{\mathrm{cm}}^{2}}\right]$	EWC iv	Filter	NB	F775W	F850LP	LLyα	SFRLyα	SFRUV	b	EWLyα
				(Å)					(1042 erg s−1)	(M⊙ yr−1)	(M⊙ yr−1)	(kpc)	(Å)
10:30:26.746	+5:24:59.76	5.744	14.00 ± 0.04	1.3	FR853	25.90 ± 0.21	2σ ≥ 28.0	27.10 ± 0.13	1.75 ± 0.44	2.0	4.1	42	44 Å
		5.517	14.02 ± 0.04	0.6	FR782	2σ ≥ 26.8	2σ ≥ 27.9	2σ ≥ 28.0	3σ ≤ 1.3	3σ ≤ 1.2	3σ ≤ 1.5	≥125	⋯
10:30:27.486	+05:25:20.15	4.948	13.76 ± 0.11	0.5	FR716	25.24 ± 0.12	26.21 ± 0.06	26.04 ± 0.06	3.27 ± 0.56	3.0	8.7	163	28 Å
13:06:06.161	+03:56:32.94	4.866	14.80 ± 0.01	2.6	FR716	25.70 ± 0.17	27.23 ± 0.12	27.56 ± 0.19	2.72 ± 0.55	2.5	2.1	205	76 Å

Download table as:  ASCIITypeset image

We devoted 15 orbits to observe four C iv absorber fields in different filters (HST-GO-12860). We used five orbits using the FR853N ramp filter for the z = 5.7 system, four orbits using F782N for the z = 5.5 absorber, and three orbits using F716N for each of the two z = 4.9 systems. For each ramp filter, the central wavelength is tuned to the C iv absorption to search for the Lyα emission associated with C iv absorption. The ramp filter has a peak efficiency (70%) spanning 3000 km s⁻¹, much larger than the systematic velocity offset of Lyα to the C iv absorber (450 km s⁻¹; Steidel et al. 2010). Our HST observations allow us to measure: (1) SFR_Lyα, (2) projected distances between the galaxies and C iv absorbers, and (3) UV-based star formation rates (SFR_UV) or their upper limits. The observations probe SFR_Lyα = 2 M_⊙ yr⁻¹ at the 4σ level, corresponding to a Lyα flux of 2.2 × 10⁴² erg s⁻¹ (Kennicutt 1998). Data reduction is performed with Multidrizzle (Koekemoer et al. 2013). Using a uniform 06 diameter aperture, we calculated the depth of our observations in different bands. The narrowband magnitude for point sources reached the 5σ level of m_F716N = 26.1, m_F782N = 26.0, and m_F853N = 26.1. The broadband observations reached m_F850LP = 27.8 and m_775W = 27.7 at the 5σ level. Photometry is performed with SExtractor (Bertin & Arnouts 1996) using the rms map converted from the inverse variance image generated by Multidrizzle (Casertano et al. 2000).

We select LAE candidates with rest-frame equivalent width (EW) > 20 Å. The EW of 20 Å corresponds to the color selection criteria of m_broadband–m_narrowband > 0.65. Also, we require the signal-to-noise ratio of the continuum-subtracted flux to be greater than 3. The signal-to-noise is calculated using ${m}_{{\rm{B}}}\mbox{--}{m}_{\mathrm{NB}}\geqslant {\rm{\Sigma }}\times \sqrt{{\sigma }_{\mathrm{NB}}^{2}+{\sigma }_{{\rm{B}}}^{2}}$, where Σ = 3.0 and m_NB, σ_NB, m_B, and σ_B represent the magnitudes and errors of the narrow band and broad band. The selection criteria are shown in Figures 1–3. Galaxies residing above both red and blue dotted lines are selected. One LAE candidate satisfies the selection criteria in each of the three C iv absorber fields at z = 5.74, z = 4.95, and z = 4.87. We visually checked individual exposures of these candidates to make sure that (a) their photometries are not affected by cosmic rays, and (b) sources do not reside at the detector edges where the photometry is less accurate. We do not detect any LAE candidates associated with the C iv system at z = 5.52.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In each figure, we use the following galaxy model to fit the photometry of LAE candidates to derive the flux of Lyα emission and the Lyα-based and UV-based SFR. We use the Bruzual & Charlot (2003) models with a Salpeter initial mass function, a 0.2 solar metallicity (Z = 0.004), a continuous star formation history, and an age of 30 Myr (e.g., Ono et al. 2010). We use the Lyα optical depth derived by Madau (1995) to account for the IGM absorption. Then, we add the dust extinction following Calzetti et al. (2000) with ${R}_{V}\,=A(V)/E(B-V)=3.1$ for the z ≈ 5.7 candidates, and we set R_V as a free parameter for two z ≈ 5.0 candidates. Further, we added the Lyα emission based on our narrowband observations, assuming the FWHM of Lyα is 300 km s⁻¹ (e.g., Jiang et al. 2013).

In Table 1, we provide the physical properties of the three LAE candidates. For the C iv absorber at ${z}_{\mathrm{abs}}=5.744$, we detect a LAE candidate with an impact parameter of 42 kpc. This LAE candidate has an expected Lyα luminosity of ${L}_{\mathrm{Ly}\alpha }=(1.75\pm 0.44)\times {10}^{42}$ erg s⁻¹, corresponding to the SFR_Lyα of 2.0 M_⊙ yr⁻¹ (e.g., Kennicutt 1998). The SFR_UV is estimated to be 4.1 M_⊙ yr⁻¹, using the flux density of the best-fit spectral energy distribution (SED) around λ_rest = 1500 Å (Madau et al. 1998). Díaz et al. (2011) found an LAE at 79 kpc away from the C iv absorber and suggested that this LAE could be the powering source. We do not detect any NB excess of this source in our deeper HST observations. This source has m_F853N > 26.30, m_F775W = 26.75 ± 0.09, m_F850LP = 26.23 ± 0.06, which is inconsistent with an LAE at z ≈ 5.7.

For the C iv absorber at ${z}_{\mathrm{abs}}=4.948$, the impact parameter of the LAE candidate is 163 kpc. The Lyα luminosity is L_Lyα = (3.27 ± 0.56) × 10⁴² erg s⁻¹, corresponding to a SFR_Lyα = 3.0 M_⊙ yr⁻¹. The SFR_UV is 8.7 M_⊙ yr⁻¹. For the C iv absorber at ${z}_{\mathrm{abs}}=4.866$, there is an LAE candidate with an impact parameter of 205 kpc. The Lyα luminosity is L_Lyα = (2.72 ± 0.55) × 10⁴² erg s⁻¹, corresponding to an SFR_Lyα = 2.5 M_⊙ yr⁻¹. The UV-based SFR is estimated to be 2.1 M_⊙ yr⁻¹. If we further assume a stellar mass (M_*)–SFR relation at z = 4–5 (e.g., Sparre et al. 2014), the SFR_UV-derived stellar mass of these three LAE candidates have stellar masses of ∼10^9.0–9.5 M_⊙ and halo masses of ∼10¹⁰ M_⊙.

For the C iv absorber at ${z}_{\mathrm{abs}}=5.517$, we do not detect any LAE associated with this absorber, placing a 3σ upper limit on the Lyα luminosity of 1.60 × 10⁴² erg s⁻¹ and a 3-σ upper limit on the SFR_Lyα ≈ 1.5 M_⊙ yr⁻¹.

The main contaminants of high-z LAEs are strong [O ii] emitters at lower redshift. For two C iv absorbers at $z\approx 4.9$, we currently have no imaging observations bluer than the Lyα emission. We expect that the contamination rate due to low-z [O ii] emission is relatively high. [O ii] emitters at z = 0.91 (z = 0.94) with rest-frame EW > 61.3 Å are contaminants of LAEs at z = 4.948 (z = 4.866) with rest-frame EW > 20 Å. From the [O ii] luminosity function at z ≈ 1.0 (Takahashi et al. 2007; Zhu et al. 2009), strong [O ii] emitters that can be detected in our depth (EW > 61.3 Å, ${L}_{[{\rm{O}}{\rm{II}}]}\gt {10}^{41.0}$ erg s⁻¹) have a number density of 0.009 Mpc⁻³, corresponding to ≈0.2 per field. For LAE candidates at z = 5.7, a number of studies have been conducted at similar redshifts (e.g., Shimasaku et al. 2006; Ouchi et al. 2008, 2017). Spectroscopic confirmations suggest that the contamination rate is ≲20% at our depth. Future spectroscopic follow-ups must be conducted to confirm or rule out the LAE candidates conclusively.

One potential issue is that not all star-forming galaxies have Lyα emission that can be selected using the LAE selection technique. Schenker et al. (2012) showed that 60% galaxies with an absolute magnitude of −20.3 < M_UV < −18.7 have Lyα EW greater than 20 Å. Our observations probe galaxies down to M_UV ≈ −18.5 in 3σ, indicating a high completeness of using the LAE technique to select galaxies in our survey. Another issue is whether LAE candidates we detected are physically associated with the C iv absorbers. Observationally, we can evaluate this issue by calculating the number of galaxies expected in random fields: if ≪1 galaxies are expected within an impact parameter in random fields, then galaxies detected within these impact parameters are more likely to be sources of C iv systems (e.g., Díaz et al. 2014). This is because C iv absorbers are expected to be clustered with the physically associated galaxies. The LAE candidate we detected at z = 5.74 ± 0.07 is the first source reported at z > 5 to be projected ≈40 kpc away from a C iv absorber. From the luminosity function at z = 5.7 (Ouchi et al. 2008; Konno et al. 2017), the ${L}_{\mathrm{Ly}\alpha }^{* }=6.8\times {10}^{42}$ erg s⁻¹ and ϕ* =7.7 × 10⁻⁴ Mpc⁻³. Our limiting Lyα luminosity is 0.26 $\times {L}_{\mathrm{Ly}\alpha }^{* }$. The width of the FR853N ramp filter corresponds to a redshift interval of Δz = 0.14. The expected number of finding an LAE in random fields with ${L}_{\mathrm{Lya}}\geqslant 0.26\,\times \,{L}_{\mathrm{Ly}\alpha }^{* }$ and at impact parameters of ≤42 kpc is 0.01. The number of faint galaxies with $L=0.01\mbox{--}0.26\times {L}_{\mathrm{Ly}\alpha }^{* }$ (below our detection limit) and an impact parameter within 42 kpc is 0.2. Thus, there is a relatively small probability of finding such an LAE in random fields. If this LAE candidate is confirmed by future deep spectroscopy, then one can conclude that this galaxy is highly likely to be the source of the C iv absorber at z = 5.744.

For two C iv absorbers at z ≈ 5, the expected number of LAEs with ${L}_{\mathrm{Ly}\alpha }\geqslant 0.26\times {L}_{{\rm{Ly}}\alpha }^{* }$ and an impact parameter of ≈200 kpc is 0.5. The number of faint LAEs (${L}_{\mathrm{Ly}\alpha }\,=0.01\mbox{--}0.26\times {L}_{\mathrm{Ly}\alpha }^{* }$) is expected to be ≈10 in random fields. As a result, it is unlikely that the LAE candidates we detected at z ≈ 5.0 are sources of the C iv absorbers, and thus we treat these two sightlines as nondetections.

From our one detection and three nondetections, we conclude that strong C iv systems at z ≈ 5–6 do not generally lie close to galaxies identified as LAEs with SFR ≳ 2 M_⊙ yr⁻¹. This is surprising because strongly star-forming galaxies are expected to be driving the strongest outflows at these epochs, enriching the early IGM. Models generically predict early, widespread enrichment of metals (e.g., Madau et al. 2001; Oppenheimer & Davé 2008), such that z ≳ 5 represents the first epoch of IGM enrichment. Given this, the most straightforward interpretation of our result is that early enrichment is dominated, at least in a volumetric sense, by small galaxies with SFR ≲ 2 M_⊙ yr⁻¹.

There are two important caveats to this result. First, our observations cannot select LBGs without strong Lyα emission. However, as noted in the previous section, LBGs and LAEs are often coincident at these epochs (60% of LBGs at z = 5–6 are LAEs; e.g., Schenker et al. 2012). If the 40% that are not LAEs are randomly associated with LBGs, then the probability of all four C iv systems being undetected is 0.4⁴ (<3%). Second, our selection disfavors the detection of massive dust-obscured galaxies. Such galaxies are metal-rich and have high SFRs, so potentially we could be missing these sources. However, typical high-SFR objects at these epochs should not be heavily dust-obscured (e.g., Ono et al. 2010; Finkelstein et al. 2012). Future submillimeter observations would be required to rule out this explanation conclusively.

Setting aside the caveats, our conclusion represents an important constraint on models of early metal production and dissemination. Such models began by postprocessing metal distributions onto cosmological simulations (e.g., Davé et al. 1998; Aguirre et al. 2001), but the introduction of outflows into simulations by Springel & Hernquist (2003) enabled self-consistent IGM enrichment. Oppenheimer & Davé (2006) showed that a momentum-driven outflow model in which small galaxies have higher mass loading (i.e., outflow rate relative to SFR) yielded better agreement with C iv enrichment at z ∼ 2–4, as compared to models where the mass loading is independent of galaxy mass. Oppenheimer et al. (2009) extended these results to z ∼ 6 and showed that the high mass loading factors in small galaxies resulted in early enrichment in accord with C iv data, despite a fairly small metal volume filling factor of a couple percent. They further predicted that strong absorbers (N_{C iv} = 10^13.8–14.8 cm⁻²) arise from outflows from galaxies with stellar masses ∼10^7–9 M_⊙ and impact parameters from a few kiloparsecs up to 50 kpc. These predictions are consistent with our observational results.

Using cosmological simulations, Finlator et al. (2013) introduced on-the-fly radiative transfer into cosmological simulations and examined early metal enrichment in conjunction with an inhomogeneous ionizing background to determine the impact of local UVB fluctuations on the ionization state. They argued that the mass scale of halos hosting O i absorbers was $10\mbox{--}100\times$ smaller than typical LBGs at z ∼ 6, which is consistent with our results if OI and C iv absorption arises around similar systems. In contrast, Keating et al. (2016) examined early IGM enrichment in the Sherwood and Illustris simulations and found that galaxies associated with strong (EW > 1 Å) C iv absorbers have halo masses of M > 10¹¹ M_⊙ and impact parameters of 10–30 kpc. These results seem to be less in agreement with our results at face value, although a more careful comparison is warranted. This is interesting because Illustris utilizes a mass loading factor that increases to lower masses even more rapidly than in momentum-driven wind models. Hence, other factors such as local density and ionization may be playing a key role. It could be that the Keating et al. (2016) and Nelson et al. (2015) models are predicting strong C iv from dust-obscured galaxies that would be missed by our LBG/LAE survey.

To summarize, we present the detection of one plausible C iv source at z = 5.7 with a projected distance of 42 kpc and an SFR_Lyα = 1.75 ± 0.44 M_⊙ yr⁻¹. Also, we found two LAE candidates in the field of two C iv absorbers at z = 4.95 and z = 4.87 with projected distances of 163 and 205 kpc, respectively. But the impact parameters are much larger than that predicted by simulations, making them unlikely to be C iv sources. We do not detect C iv-associated LAEs at z = 5.52. From our observations, no C iv-associated LAEs have impact parameters ≲40 kpc with SFR_Lyα ≥ 1.5 M_⊙ yr⁻¹ (3-σ limit). Particularly, for the strongest absorbers with EW > 2 Å, we do not detect LAEs with 3σ SFR_Lyα ≥ 1.4 M_⊙ yr⁻¹. Our observations provide strong constraints on models of early enrichment in simulations. Future facilities, including the James Webb Space Telescope, will probe the IGM–galaxy relation down to a much fainter galaxy population where some models predict that the C iv arises.