EDP Sciences
Free Access
Issue
A&A
Volume 591, July 2016
Article Number L8
Number of page(s) 4
Section Letters
DOI https://doi.org/10.1051/0004-6361/201628943
Published online 21 June 2016

© ESO, 2016

1. Introduction

In the quest for identifying the main sources of cosmic reionization and understanding this early epoch of the Universe, three important factors need to be quantified. First, sources emitting Lyman continuum (LyC) photons into the intergalactic medium (IGM) must be identified and their emission quantified. Second, an average escape fraction of ionizing photons must be estimated or assumed, and third, the total ionizing photon production of galaxies (or other sources) needs to be related to a statistical quantity such as a luminosity function to compute the total amount of ionizing photons emitted and escaping from such a population.

Because the galaxy UV luminosity function at high-z is fairly well determined (e.g., Bouwens et al. 2015, Finkelstein et al. 2015), it is convenient to write the rate of ionizing photons escaping from galaxies as(1)where NLyc is the Lyman continuum photon production rate, fesc the LyC escape fraction, Lν the monochromatic UV luminosity, and therefore ξion = NLyc/Lν the ionizing photon production per unit UV luminosity, or in other words, the “efficiency”. Knowing fesc and ξion, we can thus compute the total photon rate at which a given galaxy population ionizes the IGM (e.g., Robertson et al. 2013).

The production efficiency ξion of a given stellar population is a simple prediction from synthesis models, from which canonical values of log (ξion) ≈ 25.2–25.3 erg-1 Hz are adopted for high-z studies (e.g., Robertson et al. 2013), corresponding to constant star-formation and slightly sub-solar metallicity. Higher values may be obtained by stellar population models with young ages, non-constant star formation histories, lower metallicities, or when binary stars are included (see, e.g., Schaerer 2003; Robertson et al. 2013; Wilkins et al. 2016). Observationally, ξion has recently been estimated by Bouwens et al. (2015) for a sample of high-z Lyman-break galaxies (LBGs) by combining indirect measurements of Hα from photometry with the observed UV luminosity, and in another study for a lensed z = 7 galaxy (Stark et al. 2015), finding values of ξion compatible with canonical values or somewhat higher. However, the ionizing photon production of galaxies known to be LyC leakers (i.e., with fesc > 0) has not been measured so far.

Selecting star-forming galaxies for their compactness and high emission line ratio [O iii]λ5007/[O ii] λ3727 = O32> 5 (Izotov et al. 2016a, hereafter I16a) and Izotov et al. (2016b, I16b) have recently found five z ~ 0.3 sources out of five showing a clear detection in the LyC with corresponding absolute escape fractions fesc ~ 6–13%. This breakthrough in the identification of LyC leakers at low-z now for the first time provides the opportunity of determining their ionizing photon production and other properties and of examining how representative these sources might be for galaxies at high redshift, close to and within the epoch of reionization. In this Letter we report the results from this analysis and comparison.

The ionizing properties of our sources are discussed in Sect. 2. In Sect. 3 we show that the main observed and derived properties of our z ~ 0.3 sources are very similar to those of “typical” galaxies at high-z. Our main results are summarized in Sect. 4. We adopt a Λ-cold dark matter cosmological model with H0 = 70 km s-1 Mpc-1, Ωm = 0.3 and ΩΛ = 0.7. Magnitudes are given in the AB system.

2. UV and ionizing properties of z ~ 0.3 leakers

We used the GALEX and SDSS photometry as well as emission line measurements of the five Lyman continuum leakers reported in I16a,b to determine their ionizing photon production efficiency and other UV properties. Since the luminosity in the optical hydrogen recombination lines is proportional to the number of LyC photons absorbed in the galaxy, we determined NLyc for our sources from(2)where L(Hβ) is the (extinction-corrected) Hβ luminosity from I16a,b, and the numerical coefficient is derived from Storey & Hummer (1995) for typical conditions in Hii regions. Cast in terms of the absolute UV magnitude, we have log (ξion) = log (NLyc) + 0.4 × MUV−20.64. The absolute UV magnitude M1500, uncorrected for extinction, was determined from the best-fit spectral energy distribution (SED) to the broad-band photometry of our sources using the fitting tool described below. For comparison with high-z galaxy observations (cf. below) we also used the best-fit SED to determine the UV slope β15001. Other data were taken from I16a,b. The most important derived quantities are summarized in Table 1. The main uncertainty on ξion comes from the aperture correction for the Hβ luminosity (I16a,b), which we estimate is <30–40%. We estimate the uncertainty in the extinction correction of the UV flux to be ~30%. We therefore adopted a typical error of ±0.1 (0.15) dex for ξion ().

Table 1

Observed and derived UV and ionizing properties of our sample.

thumbnail Fig. 1

Ionizing photon production per unit UV luminosity, ξion, as a function of the absolute UV magnitude (top panel) and the UV slope (bottom) of the five Lyman continuum leakers of I16a,b. Large red symbols show ξion, large blue symbols after correction for UV attenuation (two blue triangles are indistinguishable). The cyan band illustrates canonical values for the intrinsic . Recent determinations of for LBGs at z = 3.8–5 and z = 5.1–5.4 from Bouwens et al. (2015) are shown by small black and magenta symbols with error bars, respectively.

Open with DEXTER

In Fig. 1 we show the ionizing photon production efficiency (i.e., per UV luminosity) of our Lyman continuum leakers as a function of UV magnitude and compare them with the canonical value log (ξion) ≈ 25.2–25.3 erg-1 Hz (cf. above) and the recent estimate from observations of high-redshift galaxies (Bouwens et al. 2015, hereafter B15). Normalized to the observed UV luminosity, the ionizing photon production efficiency of our sources is found to be log (ξion) ≈ 25.5–26 erg-1 Hz, which is higher by a factor 2–6 than the canonical value that is generally applied to translate the observed UV luminosity density into a global ionizing photon production rate (e.g., Robertson et al. 2013). This implies that the contribution of relatively bright galaxies, for instance, at ~ for z ~ 6–8 (cf. Bouwens et al. 2015; Finkelstein et al. 2015), to the cosmic ionizing photon production might be larger than commonly thought.

The UV flux of our leaking galaxies is attenuated by a factor 1.8–3.8 with a median of 2.6 (AUV ≈ 1; cf. Table 1). After correction for dust attenuation, the resulting intrinsic ionizing photon production efficiency , also listed in the table, is log (ξion) ≈ 25.1–25.5 (erg Hz-1)-1, close to the canonical value.

The behavior of the observed and dust-corrected values of ξion and , respectively, now as a function of the observed UV slope, is shown in the bottom panel of Fig. 1. Our sources have UV slopes of about β1500 ~ −1.7 to −2. Broadly speaking, our results are comparable to those derived by B15, who found values of compatible with the canonical one for the bulk of their sources, and a possible increase for the bluest sources (β< −2.3). However, an important point to keep in mind is that the determinations of by B15 rely on using the UV slope to estimate the UV attenuation2. For example, for sources with β = −2 and an intrinsic slope β0 = −2.23, this implies AUV = 0.25 for the SMC law, whereas our sources show a median AUV ≈ 1 for the same UV slope, a factor ~2 higher than the UV attenuation applied by B15. In reality the UV attenuation of the high-z galaxies analyzed by B15 might be underestimated since their true UV slope is expected to be bluer than β0 = −2.23, as already stressed by de Barros et al. (2014) and Castellano et al. (2014). If correct, this would imply the same factor 2 downward revision of the value of of B15.

Independently of dust corrections, the ionizing emissivity needed to match cosmic reionization is estimated to correspond to log (fescξion) = 24.5 (24.9) erg-1 Hz if the UV luminosity function extends down to M1500 = −13 (−17) (cf. Robertson et al. 2013; Bouwens et al. 2015). Our sources show log (fescξion) = 24.24−24.83 erg-1 Hz with a median of 24.67, a factor 1.5 higher than the above value log (fescξion) = 24.5 erg-1 Hz. If the escape fraction of our sources were higher, fesc = 0.2 as assumed in these studies, they would emit log (fescξion) = 24.8–25.7 erg-1 Hz, with a median of 25.1. We now compare other observed and derived physical properties of our sources to those of high-redshift star-forming galaxies.

3. Comparison with high-redshift galaxies

3.1. Comparison with the z = 3.218 LyC leaking galaxy Ion2

In many respects, the properties of the compact z ~ 0.3 leaking sources are very similar to those of the z = 3.218 Lyman continuum leaker Ion2 found by Vanzella et al. (2015) and de Barros et al. (2016). First Ion2 is also bright in the UV, MUV = −21, which is . Its low stellar mass, 1.6 × 109 M, is similar to M = (0.2−4) × 109 M (a median of 1 × 109 M) of our five sources (I16a,b). The metallicity of Ion2, determined from rest-frame UV and optical emission lines, is ~1/6 solar, compared to ~(0.1–0.2) solar.

The non-detection of [O ii] λ3727 translates into a 2σ lower limit of a high ratio O32> 10 for Ion2, even higher than for our objects. Furthermore, Ion2 is also a compact source with a size ~300 ± 70 pc (cf. de Barros et al. 2016). Interestingly, the two latter properties are found a posteriori, since the source was selected from a peculiar color selection. Different selection criteria finding LyC leakers may thus pick up sources with similar properties.

Finally, Ion 2 shows strong rest-frame optical emission lines, for example, EW(5007) =1103 ± 60 Å or even larger by a factor 2 if corrected for a large escape fraction of ionizing photons, compared to EW(5007) = 900–1260 Å of the z ~ 0.3 leakers (I16a,b). Such high equivalent widths seem fairly typical for star-forming galaxies at z ≳ 6, as we show below (cf. Fig. 3).

3.2. Comparison with typical high-z galaxies

As is clear from Table 1, our five Lyman continuum leakers show, at the same absolute UV magnitude, a UV slope that agrees well with the average slope observed in Lyman-break galaxies at z ~ 4–7 (cf., e.g., Dunlop et al. 2012, Bouwens et al. 2014). Our z ~ 0.3 sources are also in line with the average relation between stellar mass and UV magnitude derived at high redshifts, which indicates a stellar mass M ~ 109 M for M1500 ~ −20 (cf. Duncan et al. 2014; de Barros et al. 2014; Grazian et al. 2015).

Figure 2 shows for illustration the SED fit of the first of our LyC leakers, J0925+1403 from I16a, to the broad-band photometry from the SDSS and the two GALEX bands. The fits are compatible with the more detailed SED fits to the observed COS and SDSS spectra discussed in I16a,b. The fit was obtained with a version of the Hyperz code including nebular emission, described in Schaerer & de Barros (2009, 2010), which has been used extensively to fit large samples of high-z LBGs (cf. de Barros et al. 2014). Since the attenuation law and the metallicity are constrained or measured (I16a), we used the SMC law and a metallicity =1/5 solar, the closest value available for the Bruzual & Charlot (2003) models. Clearly, the SDSS photometry is dominated by strong emission lines in bands at λ ≳ 5500 Å, and the SED is well fit with the average emission line ratios taken from Anders & Fritze-v. Alvensleben (2003) (here for 1/5 solar metallicity) that is adopted in our models. This demonstrates that the SED of extreme, rare objects of the nearby Universe with very strong emission lines can also be well reproduced with typical line ratios of low-z galaxies.

thumbnail Fig. 2

Observed broad band photometry and best-fit SED (black curve) of the compact Lyman continuum leaker J09 from I16a. Red crosses indicate the synthetic flux in the corresponding filters, showing that most of the optical bands are dominated by nebular emission.

Open with DEXTER

A salient feature of LBGs at high redshift, which has been clearly established during recent years, is the presence of strong optical emission lines, whose signature is detected in broad-band photometry and whose (average) strength increases strongly with redshift (e.g., Shim et al. 2011; Labbé et al. 2013; de Barros et al. 2014). We now compare the line strengths (equivalent widths) of our z ~ 0.3 LyC leakers with these observations, summarized in Fig. 3 for Hα, [O iii]λ5007, and [O ii] with fits derived by Khostovan et al. (2016) for the range of stellar mass 9.5 < log (M/M) < 10, where these quantities can be derived over a wide redshift domain. Overplotted is the range of EWs measured in our five z ~ 0.3 LyC leakers reported in I16a,b. With rest-frame EW(Hα) ~ 560–1060 Å and EW([O iii]λ5007) ~ 900–1260 Å our sources are comparable to typical star-forming galaxies at z ≳ 63. This shows that the nebular properties of extreme and rare objects of the low-redshift Universe, selected by compactness and high O32 ratios, appear to be very similar of those of average star-forming galaxies at high-z. By analogy with our leakers, this also suggests that Lyman continuum leaking may be frequent in high-redshift galaxies.

From Fig. 3 we note also that the observed EW([O ii] λ3727) ~ 60–130 Å of our sources also agrees well with the behavior of [O ii] λ3727 extrapolated to high redshift by Khostovan et al. (2016) for the same redshift (z ~ 6–8), as indicated by the other emission lines. If true, this would indicate that the average O32 ratio continues to increase beyond z ≳ 4, continuing the trend previously observed from z ~ 0 to 3, as has been shown by Khostovan et al. (2016). If the current small samples are representative, high O32 ratios (e.g., O32> 4) imply a LyC escape fraction fesc> 5% with a possible trend of fesc increasing with O32 (I16b). This would imply that the average star-forming galaxy at z> 4 would also be a leaker with fesc> 5%, since the average O32 ratio exceeds 4 above this redshift (cf. Khostovan et al. 2016).

thumbnail Fig. 3

Top, middle, and bottom panels show the redshift evolution of the average (rest-frame) equivalent widths of Hα, [O iii]λ5007, and [O ii] λ3727, respectively, for galaxies with stellar masses 9.5 < log (M/M) < 10 as fitted by Khostovan et al. (2016), and the range of the EWs observed in our five z ~ 0.3 Lyman continuum leakers (blue horizontal bands). The leakers from our study show line equivalent widths typical of star-forming galaxies at z ≳ 6.

Open with DEXTER

3.3. Discussion

All the above mentioned properties show that the five z ~ 0.3 galaxies recently identified as Lyman continuum leakers by I16a,b are very similar to both the arguably most reliable high-z leaker, the z = 3.218 galaxy found by Vanzella et al. (2015) and de Barros et al. (2016), and to typical star-forming galaxies at z ≳ 6.

In terms of their very high equivalent widths of [O iii] λλ4959, 5007 and Hα, our z ~ 0.3 sources are very rare for low-z sources. By this measure they correspond to the <10-3 tail of the high EW distribution of SDSS DR12 galaxies. On the other hand, these high equivalent widths appear to be common, possibly even typical, for z> 6 LBGs, as shown in Fig. 3. This analogy with our LyC leakers suggests that the typical LBG at high redshift may also be leaking Lyman continuum radiation.

The recent work of Sharma et al. (2016) provides indirectly further support for this hypothesis. Based on simple, but plausible assumptions about local LyC escape, these authors predicted the escape fraction for galaxies from simulations, finding an increasing fesc with redshift and significant escape from most high-z galaxies. They traced these trends back to the mean surface density of star formation, which is found to be very high in most of their simulated galaxies at high redshift. Interestingly, all five z ~ 0.3 LyC leakers from I16a,b also show a very high surface density of star formation, ΣSFR ~ 2–50 M yr-1kpc-2, comparable to observations of high-z galaxies and to the predictions of Sharma et al. (2016). The success of the selection method of I16a,b in finding LyC leakers at high O32 ratios may thus be related to compactness, which together with very strong emission lines indicating a high specific SFR implies a high surface density of star formation. Heckman et al. (2001, 2011) and Borthakur et al. (2014) have suggested that strong star formation like this results in strong outflows that clear channels in the ISM, allowing thus the escape of Lyman continuum photons.

4. Conclusion

We have analyzed the properties of five low-redshift Lyman continuum leaking galaxies observed with the COS spectrograph onboard HST that have recently been reported by Izotov et al. (2016a,b). The z ~ 0.3 sources were selected for compactness and for showing a high emission line ratio O32, which has previously been suggested as a possible diagnostic for Lyman continuum escape (Jaskot & Oey 2013; Nakajima & Ouchi 2014).

We determined the ionizing photon flux production of these galaxies, which are metal poor (~1/6 solar), dominated by young stellar populations (<10 Myr; cf. I16a,b), and are relatively UV bright (M1500 ~ −20 to −20.8, cf. Table 1). Finally we compared the observed and derived physical properties of these rare, extreme objects from the nearby Universe to those of high-redshift galaxies. Our main results can be summarized as follows:

  • The ionizing photon production efficiencyper observed UV luminosity, ξion, of the leakers islog (ξion) ≈ 25.6–26 erg-1 Hz, which is higher by a factor 2–6 than the canonical value (cf. Robertson et al. 2013; Wilkins et al. 2016).

  • Although our sources show a low extinction in the optical (AV ~ 0.15–0.4), their UV attenuation is AUV ~ 0.6–1.4, which implies that the intrinsic, extinction-corrected ionizing photon flux production efficiency is –25.6 erg-1 Hz, close to the canonical value.

  • The five z ~ 0.3 sources of I16a,b share many properties with the best established high-z leaking galaxy Ion2 (de Barros et al. 2016): absolute UV magnitude, stellar mass, metallicity, line equivalent widths, high O32 ratio, and other parameters are very similar.

  • The high rest-frame equivalent widths of Hα and [O iii]λ5007 of the z ~ 0.3 leakers are very similar to those inferred for typical star-forming galaxies at z ≳ 6 from broad-band photometry. This shows that the rare, extreme galaxies selected by I16a,b from the Sloan survey might very well be fairly representative of average galaxies in the early Universe.

  • Our results also suggest that UV bright galaxies at high-z such as Lyman break galaxies can be Lyman continuum leakers and that their contribution to cosmic reionization, based on canonical assumptions for ξion, is probably underestimated.


1

β1500 is defined as the slope of the spectrum Fλλβ between 1300 and 1800 Å. β2000, measured over 1800–2200 Å, is typically bluer by ~0.2–0.4 for our sources.

2

For the Small Magellanic Cloud (SMC) law their relation is AUV = 1.1(ββ0) = 1.1(β + 2.23), where β0 = −2.23 is the intrinsic UV slope corresponding to solar metallicity and constant SFR with age >100 Myr.

3

For galaxies with a median mass ~ 109 M such as our sources, these EWs may be typical even at somewhat lower redshift, since EWs increase on average with decreasing stellar mass (cf. Khostovan et al. 2016).

Acknowledgments

We thank various colleagues, including Eros Vanzella, Andrea Grazian, and Tom Theuns, for stimulating discussions, Ali Khostovan for kindly sharing python scripts, and Yves Revaz for help with python.

References

All Tables

Table 1

Observed and derived UV and ionizing properties of our sample.

All Figures

thumbnail Fig. 1

Ionizing photon production per unit UV luminosity, ξion, as a function of the absolute UV magnitude (top panel) and the UV slope (bottom) of the five Lyman continuum leakers of I16a,b. Large red symbols show ξion, large blue symbols after correction for UV attenuation (two blue triangles are indistinguishable). The cyan band illustrates canonical values for the intrinsic . Recent determinations of for LBGs at z = 3.8–5 and z = 5.1–5.4 from Bouwens et al. (2015) are shown by small black and magenta symbols with error bars, respectively.

Open with DEXTER
In the text
thumbnail Fig. 2

Observed broad band photometry and best-fit SED (black curve) of the compact Lyman continuum leaker J09 from I16a. Red crosses indicate the synthetic flux in the corresponding filters, showing that most of the optical bands are dominated by nebular emission.

Open with DEXTER
In the text
thumbnail Fig. 3

Top, middle, and bottom panels show the redshift evolution of the average (rest-frame) equivalent widths of Hα, [O iii]λ5007, and [O ii] λ3727, respectively, for galaxies with stellar masses 9.5 < log (M/M) < 10 as fitted by Khostovan et al. (2016), and the range of the EWs observed in our five z ~ 0.3 Lyman continuum leakers (blue horizontal bands). The leakers from our study show line equivalent widths typical of star-forming galaxies at z ≳ 6.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.