C III] EMISSION IN STAR-FORMING GALAXIES NEAR AND FAR

We measure the equivalent widths of Lyα and the C iii] doublet in the rest-frame UV spectra of 11 gravitationally lensed galaxies at 1.6 < z < 3.1. Spectra were obtained using the MagE spectrograph (Marshall et al. 2008) on the Clay Magellan II telescope, as part of a large study of the rest-frame ultraviolet spectral properties of bright lensed galaxies (J. R. Rigby et al. 2016, in preparation).

Intrinsic (corrected for lensing magnification) stellar masses and star formation rates have been published for three of these galaxies (Wuyts et al. 2012a) and are (3–7) × 10⁹ M_⊙ and 20–100 M_⊙ yr⁻¹. Since final lensing models are in development for the MagE sample, we cannot yet quote stellar masses and star formation rates for the others. Metallicities from [N ii]/Hα have been published for four of the sample galaxies: SGAS 090003.3+223408 (Bian et al. 2010), the Cosmic Horseshoe (Hainline et al. 2009), SGAS J152745.1+065219 (Wuyts et al. 2012a), and RCSGA 032727−132609 (hereafter RCS0327; Whitaker et al. 2014; Wuyts et al. 2014). We measure the metallicity for a fifth galaxy, SGAS J000451.7−010321, from the [N ii]/Hα ratio in FIRE/Magellan spectra. We use the second-order polynomial calibration of Pettini & Pagel (2004).

From the literature, we take C iii] equivalent widths for six galaxies from Erb et al. (2010), Christensen et al. (2012), Stark et al. (2014), and Bayliss et al. (2014) with measured metallicities. Four galaxies have metallicities derived by Stark et al. (2014) using a Bayesian approach that relies on the strong emission lines used by the R₂₃ method; the R₂₃ metallicity measured by Christensen et al. (2012) for one of those galaxies agrees with the Bayesian method. Erb et al. (2010) measured an R₂₃ metallicity for BX418 at z = 2.30, as did Bayliss et al. (2014) for SGAS J105039.6+001730 at z = 3.625.

We take spectra of nearby galaxies from three spectral atlases: Leitherer et al. (2011), M. Peña-Guerrero et al. (2015, in preparation), and Giavalisco et al. (1996).

For 17 galaxies from Leitherer et al. (2011), we take metallicities from their tabulation and measurements of the C iii] equivalent width from Bayliss et al. (2014). Four of these galaxies have spectra of multiple regions, totaling 27 measurements of C iii]. Thirteen of these galaxies have archival International Ultraviolet Explorer (IUE) spectra that cover C iii], many of which show C iii] emission, even in favorable cases lines of moderate equivalent width, for example, IRAS 08339+6517. For the 18 galaxies of Peña-Guerrero et al. (2015, in preparation), we measure C iii] equivalent widths from the HST spectra and take direct gas-phase metallicity measurements from their Table 9. For the 11 galaxies (out of 22) from Giavalisco et al. (1996) that have a continuum signal-to-noise ratio ≳1 pixel⁻¹, we measure equivalent widths from archival low dispersion (6 Å) IUE spectra.

We measure Lyα equivalent widths for the subset of those galaxies that have archival spectra covering Lyα and redshifts sufficient, given the spectral resolution of each spectrum, to cleanly separate in the galaxy's Lyα emission from the much brighter geocoronal Lyα emission (Biretta et al. 2015). These criteria are met for: Tol-1214–277 from Leitherer et al. (2011; IUE spectra); two galaxies from Peña-Guerrero et al. (2015, in preparation), IRAS 08339+6517 from HST and IUE spectra, and NGC 1741 from an IUE spectrum; and the 11 galaxies from Giavalisco et al. (1996; IUE spectra).

These are spectra from published atlases, not from a complete galaxy sample. Further, they were obtained using a range of aperture size: 10'' × 20'' with IUE, 1''–2'' with FOS and GHRS, and 02 with STIS. As such, some observations capture bright star-forming regions within a galaxy, whereas others capture the integrated galaxy spectrum. These are necessary limitations of current ultraviolet spectroscopic atlases.

When fitting equivalent widths of C iii] and Lyα, our uncertainties incorporate statistical uncertainty as well as systematic uncertainty due to continuum estimation. Each reported equivalent width is the mean of measurements assuming a linear continuum, power-law continuum, and (if the spectrum is locally flat) median continuum. We take the standard deviation as the systematic uncertainty due to continuum fitting; it is typically less than half the measurement uncertainty.

The spectra analyzed here vary widely in resolution and signal-to-noise ratio, resulting in emission line equivalent width measurements ranging from strong to weak detections. For sources with strong emission lines and a weak or non-detected continuum, the measurement is a lower limit on the equivalent width; the associated uncertainty may be large even though the emission line is very well detected and reflects the continuum uncertainty.

Following the definition of equivalent width, equivalent widths have a positive sign for absorption and a negative sign for emission.

Table 1.  Galaxy Samples, Properties, and Equivalent Widths

Name	z	Z	Ref	Wr(Lyα)	σ	Ref	Wr(C iii])	σ	Ref	Sample	mAB	Ref
				(Å)	(Å)		(Å)	(Å)
RCSGA 032727−132609 Knot E	1.703745	8.34 ± 0.02	W14	>−1.2	−99	R14	−2.0	0.14	R14	MagE	g = 19.15	W10
SGAS J000451.7–010321	1.6811	${8.30}_{-0.08}^{+0.06}$	FIRE	−3.3	0.4	R14	−0.45	0.14	R14	MagE	g = 19.91	A09
Pox 120	0.020748	7.83	G96	−70.2	4.11	IUE	−14.4	1.8	IUE	G96	⋯	⋯
IRAS 08339+6517	0.019113	8.31	PG15	−4.45	0.5	STIS	−0.7	0.1	STIS	PG15	⋯	⋯
NGC 4861	0.002785	7.9	L11	⋯	⋯	⋯	−8.1	1.0	FOS	L11	⋯	⋯

Note. Columns: (1) source name; (2) redshift; (3) $12+\mathrm{log}({\rm{O}}/{\rm{H}})$ metallicity; (4) reference; (5) rest-frame Lyα equivalent width; (6) uncertainty; (7) reference; (8) rest-frame C iii] equivalent width, sum of both transitions when resolved; (9) uncertainty; (10) reference; (11) sample; (12) magnitude for MagE sample; (13) reference. Redshifts for MagE galaxies are from Bordoloi et al. (2015) for RCS0327, else from Rigby et al. (2014) and in preparation. Redshifts for z ∼ 0 galaxies are from NED.

References. A09—Abazajian et al. (2009), B07—Belokurov et al. (2007), B10—Bian et al. (2010), G96—Giavalisco et al. (1996), H09—Hainline et al. (2009), K10—Koester et al. (2010), L11—Leitherer et al. (2011), PG15—M. Peña-Guerrero et al. (2015, in preparation), R14—Rigby et al. (2014), S07—Smail et al. (2007), W10—Wuyts et al. (2010), W12—Wuyts et al. (2012b), W14—Wuyts et al. (2014).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

We measure C iii] strengths in bright lensed, star-forming galaxies at z ∼ 2–4, as well as in local starbursting galaxies with archival ultraviolet spectra from HST and IUE. Figure 1 compares histograms of C iii] equivalent width in these samples, as well as to Stark et al. (2014) and the composite spectrum of Shapley et al. (2003). Figure 1 demonstrates that star-forming galaxies in the local universe and at z ∼ 2–4 show a large range of C iii] equivalent widths. The mode for the distant galaxies and the mode for the z ∼ 0 galaxies are consistent with the equivalent width measured for the Shapley et al. (2003) composite. A significant minority of galaxies show strong C iii] emission with Wr (C iii]) < −5 Å: 22% of the z ∼ 0 sample and 45% of the z ∼ 2–4 sample.

Zoom In Zoom Out Reset image size

Indeed, the local sample includes a tail of strong C iii] emitters with equivalent widths to −25 Å, widths that are beyond that yet observed at z ∼ 2. These results are consistent with a scenario in which strong C iii] emission is not a particular feature of z ≳ 2 galaxies, but rather occurs in a minority of starbursting galaxies independent of redshift.

The MagE sample has systematically lower equivalent widths than the Stark et al. (2014) sample. The MagE sample galaxies were selected as lensed galaxies that are most amenable to deep rest-frame UV spectroscopy; they have typical apparent magnitudes of g_AB ∼ 20–21 and should be biased toward high surface brightness, high magnification, high luminosity, and compact morphology. Wuyts et al. (2012a) examined in detail the spectral energy distributions of three galaxies from the MagE sample, finding them typical of Lyman break galaxies, though younger than average. By contrast, the Stark et al. (2014) sample is selected to be much fainter, with an average apparent V-band magnitude of m_AB = 23.9.

Figure 2 plots the C iii] equivalent width as a function of gas-phase metallicity for the z ∼ 0 samples (black symbols) and the z ∼ 2–4 samples (blue and purple symbols). Stark et al. (2014) suggest that low metallicity may be largely responsible for large equivalent widths. Figure 2 reveals that the relationship with metallicity is complex. Strong emission (≲−5 Å) is observed in galaxies over a wide range of metallicity, from 6% to 40% of solar. Strong emission appears not to occur at metallicities above $12+\mathrm{log}({\rm{O}}/{\rm{H}})=8.4.$ Indeed, only one of the six lowest-metallicity z ∼ 0 galaxies shows strong emission. Thus, low metallicity may be a necessary but not sufficient condition for C iii] emission.

Zoom In Zoom Out Reset image size

Figure 2 demonstrates that the local universe contains galaxies whose C iii] emission is as strong, indeed stronger, than yet observed at z ∼ 2. The three galaxies from the z ∼ 0 samples with the most extreme C iii], with equivalent widths exceeding −14 Å, are Tololo 1214–277, Markarian 71 (a star-forming region within NGC 2366),¹⁰ and Pox 120. We now explore what is known about these three galaxies and plot their ultraviolet spectra in Figure 3.

Zoom In Zoom Out Reset image size

Tololo 1214–277 is classified as a Wolf–Rayet galaxy, Mrk 71 is classified as a star cluster containing Wolf–Rayet stars within the irregular galaxy NGC 2366, and Pox 120 is classified as a Wolf–Rayet galaxy (Kunth & Joubert 1985; Schaerer et al. 1999). The equivalent width of Hα is extremely high for Mrk 71 (< −1000Å; Buckalew et al. 2005) and for TOL 1214–277 (−1700 Å; Guseva et al. 2011). We could not find a published W_r(Hα) for Pox 120. Only 1 (15) of the 417 nearby galaxies of Moustakas & Kennicutt (2006) shows an Hα equivalent width of < −1000Å (< −200Å). Simple stellar populations with such high equivalent widths have ages of only a few million years (cf. Figure 23 of Leitherer et al. 2014.) Another indication of the extreme star formation occurring in these galaxies is the fact that Tololo 1214–277 is one of very few galaxies in the local universe that shows indications of leaking Lyman continuum photons (Verhamme et al. 2015).

In addition, 7 z ∼ 0 galaxies show strong C iii] emission with −14 < W_r < −5 Å. These are: NGC 5253 (regions HIIR-1 and HIIR-2), NGC 4861, NGC 5457 (region NGC 5471), Tololo 1345–420, Sbs 1319+579, III Zw 107, and Arp 252.

Shapley et al. (2003) noted a correlation between the equivalent widths of C iii] and Lyα emission in their quartile composite spectra. Stark et al. (2014, 2015) further examined this correlation. In Figure 4, we plot W_r(C iii]) versus W_r(Lyα) for galaxies at 1.6 < z < 7, as well as galaxies at z ∼ 0. A correlation is apparent for both redshift ranges, though there is considerable scatter. The correlation is dominated by the strongest emitters, with W(Lyα) ≲ −50 Å, W(C iii]) ≲ −5 Å. Within the dynamic range of the MagE sample, no correlation is apparent.

Zoom In Zoom Out Reset image size

Since the physical conditions of the lensed galaxy RCS0327 at z = 1.70 have been measured precisely, we compare them to the derived physical parameters of the Stark et al. (2014) sample. RCS0327 has a measured ionization parameter $\mathrm{log}U=-2.84\pm 0.06$ (Rigby et al. 2011, using the average extinction from Whitaker et al. 2014.) This value is consistent with ionization parameters measured for other lensed Lyman break galaxies: SGAS J122651.3+215220 and SGAS J152745.1+065219 in the MagE sample (Wuyts et al. 2012a), the Cosmic Horseshoe and the Clone (Hainline et al. 2009), and cB58 (Pettini et al. 2002). The ionization parameters of these LBGs are 0.6–1 dex lower than those inferred by Stark et al. (2014) for four galaxies. RCS0327 has a measured stellar mass of 6.3 ± 0.7 × 10⁹ M_⊙ and a star formation rate of 30–50 M_⊙ yr⁻¹ (Wuyts et al. 2012a). Its specific star formation rate (sSFR) is 5 × 10⁻⁹ yr⁻¹, such that it lies a factor of 3 above the correlation between star formation rate and stellar mass at that redshift. Three objects in Stark et al. (2014) have lower sSFRs than this; all have very strong C iii] emission with W_r ≤ −10 Å. Of the four galaxies in Stark et al. (2014) with highest stellar mass, $8.9\lt \mathrm{log}[{M}_{*}/{M}_{\odot }]\lt 9.2,$ three have C iii] equivalent widths <−4 Å. Thus, RCS0327 is similar to the Stark et al. (2014) sample in terms of sSFR, but has a higher metallicity and lower ionization parameter than inferred for that sample. This comparison provides some support for the hypothesis of Stark et al. (2014) that both low metallicity and a high ionization parameter are required to create strong C iii] emission.

Perhaps more surprising is the fact that the vigorously star-forming, very low metallicity starburst galaxy 1 Zw 18 is not a strong C iii] emitter; its equivalent width in three different regions varies from −1.3 to −4.4 Å. Its ionization parameter is high for the local universe ($\mathrm{log}U\sim -2.5$ from Dufour et al. 1988; inferred as −2.2 as the lowest-metallicity object in Morales-Luis et al. 2014). Any hypothesis that explains strong C iii] emission will need to explain why 1 Zw 18 shows relatively weak C iii] emission.