METAL-POOR, COOL GAS IN THE CIRCUMGALACTIC MEDIUM OF A z = 2.4 STAR-FORMING GALAXY: DIRECT EVIDENCE FOR COLD ACCRETION?^*

The bright background QSO J1444535+291905 (g = 16.4) at redshift z = 2.660 has been observed using the HIRES echelle spectrograph on Keck by several groups from 1995 to 2009. We obtained the spectra from the Keck archive and reduced them with makee.⁴ The final combined spectrum covers wavelength ranges 3100–5950 Å (resolution 8.8 km s⁻¹ ) and 7350–9760 Å (resolution 6.7 km s⁻¹ ).

The foreground galaxy's redshift (z_gal = 2.4391) and impact parameter from the background QSO (R_⊥ = 58 kpc) are given by Rudie et al. (2012, their Figure 12). The redshift uncertainty is 160–180 km s⁻¹ , which arises when rest-frame UV transitions such as Lyα emission and metal absorption are used to determine the intrinsic redshift (Rakic et al. 2011). Rudie et al. do not publish further details on the galaxy, but such galaxies typically have a dust-extinction corrected SFR of 30 M_☉ yr⁻¹ (Erb et al. 2006b), a gas-phase interstellar medium (ISM) metallicity of Z/Z_☉ ∼ 0.8 (Erb et al. 2006a) and a dark matter halo mass 5 × 10¹¹ M_☉ (Bielby et al. 2013). This halo mass implies a virial radius of 75 kpc (Bryan & Norman 1998), and thus the projected separation from the background QSO sightline is comparable to the virial radius.

We note that Rudie (2013) indicates a second star-forming galaxy is present at R_⊥ = 50 kpc within 250 km s⁻¹ of the first galaxy, but does not publish the spectroscopic redshift. We adopt R_⊥ = 58 kpc for our analysis, but our conclusions remain unchanged if a second galaxy exists at an even closer impact parameter.

The HIRES spectrum reveals a sub-damped absorption system, originally discovered by (Simcoe et al. 2002), that is coincident with the redshift of the foreground galaxy. Our Voigt profile fit to the damping wings implies a total neutral hydrogen column density $N_{\mathrm{H}^0}=10^{19.85\pm 0.1}$ cm⁻² (Figure 1, top). Absorption from low and high-ionization ionic metal line transitions is present in 16 distinct components spread across 320 km s⁻¹ (Figure 2). The total $N_{\mathrm{H}^0}$ is well-constrained, but as we will demonstrate, the metallicity varies by an order of magnitude among the components. Fortuitously, we can also constrain $N_{\mathrm{H}^0}$ for several individual components.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We measured column densities in the system by fitting Voigt profiles with vpfit,⁵ and by using the apparent optical depth (AOD; Savage & Sembach 1991) method. Voigt profile fits assess the degree to which a single kinematic model can explain all the observed absorption: low-ionization transitions were used to define the main velocity components, and we found that a single velocity structure matches all the low-ionization species well. The resulting models are shown in Figures 1 and 2. The higher ionization transitions Si iv, C iv and O vi do not follow the same velocity structure as the low-ions, indicating there is likely a contribution from a hotter or more highly ionized phase. Similar offsets between low-ions and high-ions have been observed in other high-z absorbers (e.g., Fox et al. 2007). Table 1 gives Voigt profile parameters and our adopted column densities for each component. Our Voigt profile fits to both the high-ions and low-ions show no evidence for partial covering or unresolved saturated components.

Table 1. Absorption Parameters

Ion	z	b	log10N (VPFIT)	log10N (AOD)	Adopted log10N	AOD Transition
		(km s−1)	(cm−2)	(cm−2)	(cm−2)	(λrest, Å)
D component (one component at −218 km s−1 , AOD range −248.5 to −193.5 km s−1 )
H i	2.436599(01)	<18	19.50 ± 0.16	–	19.50 ± 0.16
D i		10.60 ± 0.53	14.93 ± 0.02	$14.976^{+0.061}_{-0.064}$	14.93 ± 0.06	937
Al ii		10.42a	11.96 ± 0.02	$11.980^{+0.071}_{-0.075}$	$11.980^{+0.071}_{-0.075}$	1670
C ii		11.4 ± 0.2	14.04 ± 0.01	$14.014^{+0.018}_{-0.018}$	$14.014^{+0.018}_{-0.018}$	1334
C ii*			<13.1	$13.863^{+0.018}_{-0.018}$	<13.1	1335.7
C iv			<12.15	$12.115^{+0.377}_{-0.862}$	<12.15	1548
Fe ii		7.40a	12.98 ± 0.04	$13.015^{+0.088}_{-0.082}$	$13.015^{+0.088}_{-0.082}$	2600
Mg i			–	<13.9	<13.9	1683
Mg ii		13.7 ± 0.8	13.20 ± 0.02	$13.198^{+0.102}_{-0.094}$	$13.198^{+0.102}_{-0.094}$	2803
N ii			<13.3	<13.901	<13.3	1084
N v			<12.3	<12.558	<12.3	1242
O i	2.436590(01)b	8.6 ± 0.13	14.303 ± 0.005	$14.261^{+0.014}_{-0.013}$	$14.261^{+0.047}_{-0.013}$	1302
O vi			<13.1	<13.323	<13.1	1031
Si ii		10.43 ± 0.17a	13.36 ± 0.006	$13.329^{+0.049}_{-0.051}$	$13.329^{+0.049}_{-0.051}$	1526
Si iv			<12.30	$12.323^{+0.118}_{-0.146}$	<12.30	1393
Partial LLS 1 (one component at −21 km s−1 , AOD range −38 to −6 km s−1 )
H i	2.438862(04)	14.4 ± 1.7	16.30 ± 0.20	–	16.30 ± 0.20
Al ii		4.91a	11.63 ± 0.03	$11.642^{+0.087}_{-0.107}$	$11.642^{+0.087}_{-0.107}$	1670
C ii		7.35a	13.66 ± 0.02	$13.674^{+0.015}_{-0.015}$	$13.674^{+0.015}_{-0.015}$	1334
C ii*			<11.9	<12.021	<11.9	1335.7
C iv			>14.3c	>14.29c	>14.3	1550
Fe ii			<12.08	$12.067^{+0.335}_{-0.712}$	$12.067^{+0.335}_{-0.712}$	2600
Mg i			–	<13.8	<13.8	1683
Mg ii		5.17a	12.61 ± 0.04	$12.681^{+0.084}_{-0.087}$	$12.681^{+0.084}_{-0.087}$	2796
N ii			<13.56	<13.787	<13.56	1084
N v			<13.5	<13.541	<13.5	1238
O i			<12.18	<12.349	<12.18	1302
O vi			<14.5	$14.290^{+0.019}_{-0.018}$	<14.5	1037
Si ii		4.81 ± 0.27a	12.83 ± 0.01	$12.811^{+0.096}_{-0.101}$	$12.811^{+0.096}_{-0.101}$	1526
Si iv			<13.82	$13.480^{+0.019}_{-0.019}$	<13.82	1402
Partial LLS 2 (three components at 18, 28, 41 km s−1 , AOD range 8 to 57.5 km s−1 )
H i			16.43 ± 0.30	–	16.43 ± 0.30
Al ii			$11.859^{+0.032}_{-0.034}$	$11.851^{+0.087}_{-0.104}$	$11.851^{+0.087}_{-0.104}$	1670
C ii			$13.880^{+0.025}_{-0.026}$	$13.856^{+0.015}_{-0.015}$	$13.856^{+0.015}_{-0.015}$	1334
C ii*			<12.1	<12.114	<12.1	1335.7
C iv			>14.4c	>14.4c	>14.4	1550
Fe ii			<12.01	<12.307	<12.01	2600
Mg i			–	<13.9	<13.9	1683
Mg ii			$12.865^{+0.041}_{-0.046}$	$12.823^{+0.076}_{-0.076}$	$12.823^{+0.076}_{-0.076}$	2796
N ii			<13.4	$13.316^{+0.072}_{-0.089}$	<13.4	1084
N v			<13.45	$13.358^{+0.037}_{-0.041}$	<13.45	1238
O i			$12.395^{+0.184}_{-0.327}$	$12.439^{+0.383}_{-0.698}$	$12.439^{+0.383}_{-0.698}$	1302
O vi			<14.5	$14.311^{+0.020}_{-0.020}$	<14.5	1037
Si ii			$13.080^{+0.016}_{-0.017}$	$12.968^{+0.098}_{-0.114}$	$12.968^{+0.098}_{-0.114}$	1526
Si iv			<13.9	$13.609^{+0.019}_{-0.019}$	<13.9	1402

Notes. Errors on the AOD column densities include uncertainties in the continuum and zero levels. High-ion upper limits are calculated by measuring the highest column density Voigt profile consistent with the data, assuming thermal broadening. ^ab value is tied to other species assuming purely thermal broadening. ^bA single-component Voigt profile fit to the O i line is offset ∼0.8 km s⁻¹ from the C ii line. This may indicate unresolved component structure, but we do not expect it to significantly affect $N_{\mathrm{O}^0}$. ^cSaturated components.

Download table as:  ASCIITypeset image

We identify three kinematic regions for which the neutral hydrogen column $N_{\mathrm{H}^0}$, and thus the metallicity, can be precisely measured. The first is at −218 km s⁻¹ relative to the galaxy redshift, where absorption from neutral deuterium is detected at Ly-γ, Ly-δ, Ly- and Ly-8. Deuterium detections in QSO absorbers are rare, as a measurement of the weak D absorption requires a narrow linewidth, high $N_{\mathrm{H}^0}$, and the absence of blending with weaker neighboring Lyα forest lines (Webb et al. 1991). Such detections are valuable however, because a precise D/H measurement constrains the cosmic baryon density according to big bang nucleosynthesis theory. For the D component we detect, $N_{\mathrm{D}^0}=10^{14.93\pm 0.06}\,\rm cm^{-2}$ and b(D⁰) = 10.60 ± 0.53 km s⁻¹ , which are consistent with the expected column density and width of the associated H⁰, estimated from the damping wings and the Lyman series, assuming thermal broadening and the cosmic D/H ratio inferred from the Wilkinson Microwave Anisotropy Probe cosmic baryon density (Keisler et al. 2011). This width is much narrower than is typical for a Lyα forest line with the $N_{\mathrm{H}^0}\approx 10^{15}\,\rm cm^{-2}$ necessary to imitate D i, which demonstrates that this is not an interloping Lyα forest line.

We can use the D/H ratio implied by the cosmic baryon density to make a precise measurement of $N_{\mathrm{H}^0}$ in this component. Pettini & Cooke (2012) find log₁₀(D/H) = −4.585 ± 0.02 in QSO absorbers. The standard deviation in these QSO values (0.15) is larger than that expected from the measurement errors, and represents real scatter due to an unidentified astrophysical effect and/or underestimated systematic errors in the D/H analyses. We conservatively assume the scatter represents a physical variation and take log₁₀(D/H) = −4.585 ± 0.15. D⁰/H⁰ is the same as D/H, and therefore $\log _{10}N_{\mathrm{H}^0}=19.50\pm 0.16$ for this component.

Crucially, we also detect O i in this component, which allows a direct measurement of the metallicity from O⁰/H⁰. For $N_{\mathrm{H}^0}\gtrsim 10^{19}\, \rm cm^{-2}$, the ion ratio O⁰/H⁰ is a robust metallicity indicator, because the ionization potentials of O⁰ and H⁰ are very similar and the two species exchange electrons (Field & Steigman 1971), and O does not deplete strongly onto dust grains (Jenkins 2009). We find [O/H] = −2.00 ± 0.17, which is comparable to the background metallicity of the intergalactic medium (IGM) at this redshift (Schaye et al. 2003; Simcoe et al. 2004).

The remaining components for which we can accurately measure $N_{\mathrm{H}^0}$ are partial Lyman limit systems (LLSs) at ∼ − 20 and ∼ + 30 km s⁻¹ (Partial LLS 1 & 2, see Figure 2). These show unsaturated H i absorption at the Ly-10, Ly-11 and Ly-12 transitions, which imply $N_{\mathrm{H}^0}=10^{16.30\pm 0.20}$ (Partial LLS 1) and 10^{16.43 ± 0.30} cm⁻² (Partial LLS 2). Since ionization corrections are significant at these $N_{\mathrm{H}^0}$, we estimate the metallicity of these components using the photoionization models described in the following section.

We use cloudy photoionization models (version 8.01 Ferland et al. 1998) to predict the metal ion column densities for the D and partial LLS components given a metallicity and ionization parameter U (defined as the ratio of the densities of ionizing photons to hydrogen atoms). All models assume solar abundance ratios, no dust, and an ionizing spectrum due to the ambient UV background (Haardt & Madau 2012).⁶ We use a maximum likelihood technique to find the best-fitting metallicity and ionization parameter for each component (N. H. M. Crighton et al., in preparation) and their uncertainties. Column densities for low-ions together with upper limits on high-ions are used to constrain the models. Figure 3 shows the models for the D component and partial LLS components, each at the best-fitting metallicity for that component.

Zoom In Zoom Out Reset image size

For all three components we find that the UV background models provide a good match to the data. For the partial LLS components there is a minor discrepancy for $N_{\rm \rm Al^+}$, which is 0.3–0.4 dex lower than predicted by the models. This could be due to uncertainties in the shape of the ionizing spectrum, although using a starburst-dominated ionizing spectrum makes the disagreement more severe. We believe this discrepancy is likely due to an Al underabundance relative to a solar abundance pattern (e.g., Rolleston et al. 2003). The predicted N⁺ column is close to the observed upper limit for Partial LLS 2, which may indicate that nitrogen is underabundant relative to solar. A nitrogen underabundance is often observed in damped systems at a similar redshift (e.g., Prochaska et al. 2002). For all other ionic species there is excellent agreement between the observed column densities and model predictions.

We have used two independent methods to show that the metallicity of the D component is ∼1/100 solar: Z/Z_☉ = 10^{−2.00 ± 0.17} using O⁰/H⁰, and Z/Z_☉ = 10^{−1.92 ± 0.18} from the cloudy models using all the metal transitions except O i. The width of the D i line requires a gas temperature <20, 000 K, conservatively assuming thermal broadening such that $b=\sqrt{2kT/m_{\rm D}}$. From the models we find log₁₀U = −4.2 ± 0.95, which implies that H is ∼75% neutral (N_H ∼ 10^19.6 cm⁻²) and has a temperature of ∼10, 000 K, in agreement with b(D i). [Fe/O] = −0.10 ± 0.09, therefore there is no evidence for strong dust depletion, as is expected from the low metallicity. The maximum U permitted (10^−3.2) corresponds to a neutral fraction of 25% and $n_\mathrm{H}=10^{-1.8}\, \rm cm^{-3}\times F_\nu ^{\,912}/(3\times 10^{-21} \,{\rm erg\ s}^{-1}\ {\rm cm}^{-2}\ {\rm Hz}^{-1})$, where $F_\nu ^{\,912}$ is the magnitude of the ionizing spectrum at 912 Å. This implies a cloud thickness ≲ 3 kpc.

In contrast to the D component, Partial LLS 1 and 2 are metal enriched and highly ionized. Both LLS have similar conditions: the models yield Z/Z_☉ = 10^{−0.44 ± 0.31} ($n_\mathrm{H}\approx 10^{-2}\, \rm cm^{-3}\times F_\nu ^{\,912}/(3\times 10^{-21}\, {\rm erg\ s}^{-1}\ {\rm cm}^{-2}\ {\rm Hz}^{-1})$) and log₁₀U = −3.10 ± 0.15, implying that less than 1% of H is neutral (N_H ∼ 10¹⁹ cm⁻²). Using a starburst-dominated ionizing spectrum allows a lower Z/Z_☉ = 10^{−0.8 ± 0.32}, but results in a 0.6–0.7 dex disagreement between the observed and predicted $N_{\rm \rm Al^+}$. Therefore we prefer the UVB-only models, but our main conclusions are unchanged if we were to instead use a starburst-dominated spectrum. A further difference between the D and partial LLS components is the presence of absorption from higher ionization gas (Si iv, C iv, O vi) over the same velocity range as the low-ions. While some Si iv components align with the low-ions, overall the high-ions have a different velocity structure, and are much stronger than expected from our photoionization modeling of the low-ions. Very broad components are also present in O vi (see Figure 2). Thus these high-ions may trace another gas phase that is shock-heated and collisionally ionized (see, e.g., Simcoe et al. 2002).

Finally, we calculate [O/H] for the entire system, including all low-ion components, assuming O⁰ is associated with the bulk of the H⁰ and that ionization corrections are negligible. This gives [O/H]_total = −1.89 ± 0.11. Therefore all the O i components, spanning ∼140 km s⁻¹, must have a metallicity similar to the D component.