Lyman alpha and Lyman continuum emission of Mg ii-selected star-forming galaxies

ABSTRACT

1 INTRODUCTION

It was established during last decade that Lyman continuum (LyC) emission, which is produced in copious amount in both the high redshift star-forming galaxies (SFGs) at z ∼2–4 (Vanzella et al. 2015; de Barros et al. 2016; Shapley et al. 2016; Bian et al. 2017; Marchi et al. 2017, 2018; Steidel et al. 2018; Vanzella et al. 2018; Fletcher et al. 2019; Rivera-Thorsen et al. 2019; Meštric et al. 2020; Saha et al. 2020; Vielfaure et al. 2020) and the low-redshift SFGs at |$z\, \lesssim$| 0.4 (Leitet et al. 2013; Borthakur et al. 2014; Leitherer et al. 2016; Chisholm et al. 2017; Izotov et al. 2016a,b, 2018a,b, 2021a; Flury et al. 2022a,b; Xu et al. 2022), can escape from the galaxies resulting in ionization of the intergalactic medium (IGM). These galaxies are considered as analogues of the galaxies at redshifts 6–8, which are presumably the main sources of the reionization of the Universe (Ouchi et al. 2009; Wise & Cen 2009; Yajima, Choi & Nagamine 2011; Mitra, Ferrara & Choudhury 2013; Bouwens et al. 2015; Finkelstein et al. 2019; Lewis et al. 2020; Meyer et al. 2020; Naidu et al. 2020).

It was also found that f_esc(LyC) in many discovered galaxies is of the order of 10–20 per cent or higher. This could be sufficient for efficient reionization of the IGM at |$z\, \gtrsim$| 6 (e.g. Ouchi et al. 2009; Robertson et al. 2013, 2015; Dressler et al. 2015; Khaire et al. 2016).

Direct LyC observations of high-redshift galaxies are difficult because of their faintness, the increasing of IGM opacity, and contamination by lower-redshift interlopers (e.g. Vanzella et al. 2010, 2012; Inoue et al. 2014; Grazian et al. 2016). Furthermore, the knowledge of the galaxy H β or H α luminosity is needed to derive the production rate of ionizing photons and thus the f_esc(LyC). This is not possible yet for most of high-z LyC emitters. Low-redshift galaxies are brighter, but observations from space, with the aid of Hubble Space Telescope (HST), are needed for the detection of LyC emission in |$z\, \gtrsim$| 0.3 galaxies. This can be done only for limited samples of low-z galaxies. On the other hand, the H β and H α emission lines can easily be observed in low-z galaxies from the ground. In fact, many such galaxies were observed in the course of the Sloan Digital Sky Survey (SDSS). This survey was successfully used to select promising LyC leaking candidates and their subsequent observations with the HST (this paper, Izotov et al. 2016a,b, 2018a,b, 2021a; Wang et al. 2021; Flury et al. 2022a; Xu et al. 2022).

Due to difficulties of direct detection of LyC emission in both the high- and low-redshift SFGs indirect indicators for the determination of the f_esc(LyC) can be used. However, at present, it cannot be very reliably determined from most indicators due to the large scatter in their correlations with f_esc(LyC).

The shape of the Lyα line can be considered as the prime indicator of the f_esc(LyC) value, since it depends on the distribution of the neutral hydrogen around the galaxy, which also determines the escape of ionizing radiation (e.g. Verhamme et al. 2015). In most galaxies with the Lyα emission line, it has a two-peak shape due to scattering in the neutral gas with a relatively high-column density of H i, with a weaker blue peak and a stronger red peak. The offset of the peaks from the line centre serves as a measure of the neutral hydrogen optical depth along the line of sight (e.g. Verhamme et al. 2015). In particular, a tight correlation between the Lyα blue and red peak separation and the escape fraction of ionizing radiation was found (Izotov et al. 2018b). More complex Lyα profiles with three or more peaks are rarely observed (Rivera-Thorsen et al. 2017, 2019; Izotov et al. 2018b; Vanzella et al. 2018). They show significant central line emission, an indication of direct escape through porous channels, in addition to escape via scattering. In these cases, the separation of the Lyα emission peaks is a poor tracer of f_esc(LyC) because of a combination of two distinct modes of Lyα escape (Naidu et al. 2022). We also note that at redshifts |$z\, \gtrsim$| 6 the detection of Lyα is difficult because of declining Lyα transmission with redshift (Gronke et al. 2021). This decline with redshift is sharper on the blue side of Lyα, making it more difficult to detect the blue peak.

Therefore, other indirect indicators are needed, for example, those, which use strong emission lines in the rest-frame optical and UV ranges, or UV absorption lines, including hydrogen lines of the Lyman series and heavy element lines, such as Si ii λ1260, that can measure the LyC escape fraction (e.g. Chisholm et al. 2018; Gazagnes et al. 2018, 2020; Flury et al. 2022a,b; Saldana-Lopez et al. 2022).

Jaskot & Oey (2013) and Nakajima & Ouchi (2014) proposed to use the O₃₂ = [O iii]λ5007/[O ii]λ3727 flux ratio arguing that its high values of up to ∼60 in some low-z galaxies (Stasińska et al. 2015; Izotov, Thuan & Guseva 2021b) may indicate that the ISM is pre-dominantly ionized, allowing the escape of LyC photons. Indeed, Izotov et al. (2016a, b, 2018a, b, 2021a) obtained HST/COS observations of compact SFGs at redshifts |$z\, \sim$| 0.3–0.4 with O₃₂ = 5–28 and an escape fraction in the range of 2–72 per cent. Although they did find some trend of increasing f_esc(LyC) with increasing O₃₂, the dependence is weak, with a large scatter.

It has also been suggested that f_esc(LyC) tends to be higher in low-mass galaxies (Wise et al. 2014; Trebitsch et al. 2017). However, Izotov et al. (2018b, 2021a) added low-mass LyC leakers and found rather a relatively weak anticorrelation between f_esc(LyC) and stellar-mass M_⋆ in a wide range between 10⁷ and 10¹⁰ M_⊙. A similar correlation is also found in the Low-z Lyman Continuum Survey (LzLCS) in Flury et al. (2022b).

Mg ii λ2796, 2803 emission may also provide a constraint of the LyC escape and its doublet ratio can be used to infer the neutral gas column-density (Henry et al. 2018; Chisholm et al. 2020; Katz et al. 2022; Naidu et al. 2022; Xu et al. 2022). These two lines in emission are commonly seen in the spectra of local compact SFGs (Guseva et al. 2013, 2019), including LyC leaking galaxies (Chisholm et al. 2020; Guseva et al. 2020) and might be more likely to leak LyC than similar galaxies without strong Mg ii (Xu et al. 2022). They are also detected in |$z\, \sim$| 1–2 galaxies (Weiner et al. 2009; Erb et al. 2012; Finley et al. 2017; Naidu et al. 2022) and in a |$z\, \sim$| 5 SFG (Witstok et al. 2021). Henry et al. (2018) found that the Mg ii escape fraction correlates with the Lyα escape fraction, and that the Mg ii emission line profiles are broader and more red-shifted in galaxies with low escape fractions. They and Chisholm et al. (2020) pointed out that the link between Lyα and Mg ii can be used for a LyC diagnostic at high redshifts, where Lyα and LyC are difficult to observe. However, Katz et al. (2022) pointed out from the numerical simulations that Mg ii is a useful diagnostic of escaping ionizing radiation only in the optically thin regime.

The goal of this paper is to determine f_esc(LyC) for seven low-mass galaxies with various Mg₂ = Mg ii λ2796/Mg ii λ2803 flux ratios and various O₃Mg₂ = [O iii]λ5007/Mg ii λ2796+2803 flux ratios. The O₃Mg₂ flux ratios range from 10 to 35 in six galaxies and |$\gtrsim$|100 in one galaxy, where Mg ii emission is almost undetected. We aim to study the dependence of leaking LyC emission on the characteristics of Mg ii emission. We also wish to enlarge the known sample of low-redshift LyC leakers, to search for and to improve reliable diagnostics for the indirect estimation of f_esc(LyC). The properties of the selected SFGs derived from observations in the optical range are presented in Section 2. The HST observations and data-reduction are described in Section 3. The surface-brightness profiles in the UV range are discussed in Section 4. In Section 5, we compare the HST/COS spectra with the extrapolation of the SEDs modelled with the SDSS spectra to the UV range. Lyα emission and escaping LyC emission are discussed in Section 6 together with the corresponding escape fractions. The indirect indicators of escaping LyC emission are considered in Section 7. Mg ii diagnostics are discussed in Section 8. We summarize our findings in Section 9.

2 INTEGRATED PROPERTIES OF SELECTED GALAXIES

We selected a sample of local compact low-mass SFGs from the SDSS in the redshift range z = 0.32–0.43 with O₃Mg₂ in a wide range to observe their Lyα and LyC emission with HST/COS. These galaxies are chosen to be sufficiently bright, to have high O₃₂ ratios and high equivalent widths EW(H β) of the H β emission line. This ensures that a galaxy can be acquired and observed with low- and medium-resolution gratings in one visit, consisting of four orbits. Finally, we selected a total sample of seven galaxies with EW(H β) >170 Å and O₃₂ |$\gtrsim$| 4. They are listed in Table 1. All galaxies are nearly unresolved by the SDSS five-band images and have FWHMs of ∼1.0 arcsec, so that all the galaxy’s light falls within the 2.5 arcsec diameter COS aperture and within the 2 arcsec diameter SDSS aperture. This ensures that global quantities can be derived from both the UV and optical spectra. We note, however, that Mg ii lines are located in the noisy parts of SDSS spectra and detected with a low signal-to-noise ratio, at least in some galaxies. As such, their fluxes, and especially the Mg ii flux ratio Mg₂ = Mg ii λ2796/Mg ii λ2803 should only be considered tentatively. We note that follow-up spectroscopy of these galaxies with high signal-to-noise ratio covering the wavelength range with Mg ii emission will be presented in (King et al., in preparation).

The SDSS, GALEX, and WISE apparent magnitudes of the selected galaxies are shown in Table A1, indicating that these SFGs are among the faintest low-redshift LyC leaker candidates selected so far for HST observations.

To derive absolute magnitudes and other integrated parameters, we adopted luminosity and angular size distances [NASA Extragalactic Database (NED); Wright 2006] with the cosmological parameters H₀ = 67.1 km s⁻¹ Mpc⁻¹, |$\Omega _\Lambda$| = 0.682, and Ω_m = 0.318 (Ade et al. 2014). These distances are presented in Table 1.

Internal interstellar extinction A(V)_int has been derived from the observed decrement of hydrogen emission lines in the SDSS spectra after correction for the Milky Way extinction with A(V)_MW from the NED, adopting the Cardelli, Clayton & Mathis (1989) reddening law and R(V)_int = 2.7 and R(V)_MW = 3.1. The motivation of the adopted R(V)_int value is following. Izotov et al. (2017) modelled UV FUV and NUV magnitudes of the large sample of SDSS galaxies and found that the FUV magnitudes of galaxies better match the observed magnitudes with R(V)_int = 2.7 if EW(H β) >150 Å, which is the case for our galaxies, whereas R(V)_int = 3.1 is more appropriate for galaxies with lower EW(H β)s. However, we note that in the optical range, which is used for SED fitting, the determination of intrinsic fluxes of the LyC and of the elemental abundances, extinction does only slightly depend on R(V)_int.

The extinction-corrected emission lines are used to derive ionic and total element abundances following the methods described in Izotov et al. (2006) and Guseva et al. (2013).

The emission-line fluxes I(λ) relative to the H β flux corrected for both the Milky Way and internal extinctions, the rest-frame equivalent widths, the Milky Way (C(H β)_MW) and internal (C(H β)_int) extinction coefficients, and extinction-corrected H β fluxes are shown in Table B1. It is seen in the table that the extinction-corrected fluxes of the H δ, H γ, and H α emission lines in all galaxies are consistent within the errors with theoretical recombination values indicating that C(H β)_int is derived correctly.

The fluxes and the direct T_e method are used to derive the physical conditions (electron temperature and electron number density) and the element abundances in the H ii regions. These quantities are shown in Table B2. The derived oxygen abundances are comparable to those in known low-redshift LyC leakers by Izotov et al. (2016a, b, 2018a, b, 2021a). The ratios of the α-element (neon and magnesium) abundances to oxygen abundance are similar to those in dwarf emission-line galaxies (e.g. Izotov et al. 2006; Guseva et al. 2013). On the other hand, the nitrogen-to-oxygen abundance ratios in some galaxies are somewhat elevated, similar to those in other LyC leakers at |$z\, \gtrsim$| 0.3.

We determine absolute FUV magnitudes from the fluxes of the intrinsic (i.e. extinction-corrected) SEDs at the rest-frame wavelength λ = 1500 Å, which are reddened adopting extinction derived from the observed decrement of hydrogen Balmer lines. The attenuations are, on average, similar to the ones for other |$z\, \sim$| 0.3−0.4 LyC leakers and the M_FUV are similar as observed at high-redshift.

The H β luminosities L(H β) and the corresponding star formation rates, SFR, were obtained from the extinction-corrected H β fluxes, using the relation from Kennicutt (1998) for the SFR and adopting I(H α)/I(H β) from Table B1. SFRs are increased by a factor 1/[1−f_esc(LyC)] to take into account the escaping ionizing radiation which is discussed later. The SFRs corrected for escaping LyC radiation are shown in Table 2. They are somewhat below the range of 14–80 M_⊙ yr⁻¹ for the other LyC leakers studied by Izotov et al. (2016a, b, 2018a, b, 2021a).

We use the SDSS spectra of our LyC leakers to fit the SED in the optical range and derive their stellar masses. The fitting method, using a two-component model with a young burst and older continuously formed stellar population, is described for example in Izotov et al. (2018a, b). Spectral energy distributions of instantaneous bursts in the range between 0 and 10 Gyr with evolutionary tracks of non-rotating stars by Girardi et al. (2000) and a combination of stellar atmosphere models (Schmutz, Leitherer & Gruenwald 1992; Lejeune, Buser & Cuisiner 1997) are used to produce the integrated SED for each galaxy. The star formation history is approximated by a young burst with a randomly varying age t_b in the range <10 Myr, and a continuous star formation for older ages between times t₁ and t₂, randomly varying in the range 10 Myr–10 Gyr, and adopting a constant SFR. The contribution of the two components is determined by randomly varying the ratio of their stellar masses, b = M_o/M_y, in the range 0.1–1000, where M_o and M_y are the masses of the old and young stellar populations.

The nebular continuum emission, including free–free and free–bound hydrogen and helium emission, and two-photon emission, is also taken into account using the observed H β flux (i.e. not corrected for escaping LyC emission), the ISM temperature, and density. The fraction of nebular continuum emission in the observed spectrum near H β is determined by the ratio of the observed H β equivalent width EW(H β)_obs, shifted to the rest frame, to the equivalent width EW(H β)_rec for pure nebular emission. EW(H β)_rec varies from ∼900 to ∼1100 Å for electron temperatures in the range T_e = 10 000–20 000 K. We note that non-negligible nebular emission in the continuum is produced only by the young burst with ages of a few Myr.

The Salpeter (1955) initial mass function (IMF) is adopted, with a slope of −2.35, upper and lower mass limits M_up and M_low of 100 and 0.1 M_⊙, respectively. Izotov et al. (2016a) compared differences in SEDs obtained with two different IMFs, by Salpeter (1955) and Kroupa (2001). They concluded that the effect is minor. A |$\chi ^2$| minimization technique was used (1) to fit the continuum in such parts of the rest-frame wavelength range 3600–6500 Å, where the SDSS spectrum is least noisy and free of nebular emission lines, and (2) to reproduce the observed H β and H α equivalent widths.

The total stellar masses (M_⋆ = M_y +M_o) of our LyC leakers derived from SED fitting are presented in Table 2. They are derived in exactly the same way as the stellar masses of the other LyC leakers studied by Izotov et al. (2016a, b, 2018a, b, 2021a), permitting a direct comparison.

3 HST/COS OBSERVATIONS AND DATA REDUCTION

HST /COS spectroscopy of the seven selected galaxies was obtained in program GO 15845 (PI: Y. I. Izotov) during the period 2020 October–2021 May. The observational details are presented in Table 3. As in our previous programs (Izotov et al. 2016a,b, 2018a,b, 2021a), the galaxies were directly acquired by COS near-ultraviolet (NUV) imaging. All these galaxies are compact (as compact as all the other targets from our previous programs) and they have accurate SDSS astrometry for direct imaging acquisition. The NUV-brightest region of each target was centred in the 2.5 arcsec diameter spectroscopic aperture (Fig. 1). We note, however, that the acquisition exposure failed for J1014+5501 and J1352+5617 due to guide star acquisition failure in both cases because the acquisition of the guide stars was delayed. This is a frequent HST gyro issue. For safety reasons, the shutter remained closed and no acquisition image was obtained. Therefore, both galaxies were blindly acquired. The blind acquisition accuracy is ∼0.3 arcsec, which will result in very modest vignetting for a compact galaxy, possibly introducing uncertainties in the wavelength and flux calibration in the partially vignetted COS aperture. For J1352+5617, the vignetting is negligible, because the COS spectrophotometric magnitude (FUV = 21.90 mag) agrees well with the GALEX FUV = 21.83 ± 0.17 mag. For J1014+5501, the spectrophotometry (FUV = 22.69 mag) is still consistent with the GALEX magnitude (FUV = 21.88 ± 0.59 mag), considering the significant Eddington bias for the latter. The wavelength calibration was confirmed with Lyman series absorption lines of the galaxies.

Figure 1.

The HST/COS NUV acquisition images of the candidate LyC leaking galaxies in a log SB scale. The COS spectroscopic aperture with a diameter of 2.5 arcsec is shown in all panels by a circle. The linear scale in each panel is derived adopting an angular size distance.

Open in new tabDownload slide

The spectra were obtained with the low-resolution grating G140L and medium-resolution grating G160M, applying all four focal-plane offset positions. The 800 Å setup was used for the G140L grating (sensitive wavelength range 1100–1950 Å, resolving power R ≃ 1050 at 1150 Å) to include the redshifted LyC emission for all targets. We obtained resolved spectra of the galaxies’ Lyα emission lines with the G160M grating (R ∼ 16 000 at 1600 Å), varying the G160M central wavelength with galaxy redshift to cover the emission line and the nearby continuum on a single detector segment.

The individual exposures were reduced with the calcos pipeline v3.3.10, followed by accurate background subtraction and co-addition as required for our Poisson-limited data with faintcos v1.09 (Makan et al. 2021). We used the same methods and extraction aperture sizes as in Izotov et al. (2018a, b, 2021a) to achieve a homogeneous reduction of the galaxy sample observed in multiple programmes.We corrected for scattered geocoronal Lyα according to Worseck et al. (2016). The accuracy of our custom correction for scattered light in COS G140L data was checked by comparing the LyC fluxes obtained in the total exposure and in orbital night, respectively. We find that the differences in LyC fluxes for five galaxies are less or similar to the 1σ errors. Due to insufficient time spent in orbital night, this check was not possible for J1157+5801 and J1352+5617. However, we verified that the detected LyC flux of J1352+5617 (Section 6) is insignificantly affected by residual uncertainties in the G140L scattered light model.

4 ACQUISITION IMAGES AND SURFACE BRIGHTNESS PROFILES IN THE NUV RANGE

The acquisition images of five galaxies in the NUV range are shown in Fig. 1. All galaxies are very compact with angular diameters considerably smaller than the COS spectroscopic aperture (the circles in Fig. 1) and linear diameters of ∼1–4 kpc. However, two of the most compact galaxies, J0130−0014 and J1157+5801, appear to be non-leaking LyC galaxies, whereas LyC emission is detected in the remaining three galaxies with more extended envelopes (see Section 6). We use these images to derive the surface brightness (SB) profiles of our galaxies, in accordance with previous studies by Izotov et al. (2016b, 2018a, b, 2021a). No SB profiles have been derived for galaxies J1014+5501 and J1352+5617 because their acquisition exposures failed, as noted before. In accordance with Izotov et al. (2016b, 2018a, b, 2021a), we have found that the outer parts of our galaxies are characterized by a linear decrease in SB (in mag per square arcsec scale), characteristic of a disc structure, and by a sharp increase in the central part due to the bright star-forming region (Fig. 2). The scale lengths α of our galaxies, defined in equation (1) of Izotov et al. (2016b), are in the range ∼0.2–0.6 kpc (Fig. 2), lower than α  =  0.6–1.8 kpc in other LyC leakers (Izotov et al. 2016b, 2018a,b), but similar to scale lengths of low-mass LyC leakers with masses < 10⁸ M_⊙ (Izotov et al. 2021a). The corresponding surface densities of star formation rate in the studied galaxies, Σ = SFR/(πα²), are similar to those of other LyC leakers. The half-light radii r₅₀ of our galaxies in the NUV are considerably smaller than α because of the bright compact star-forming regions in the galaxy centres (see Table 2).

Figure 2.

NUV SB profiles of galaxies indicated by the dots with error bars. Straight lines are linear fits of surface brightness μ_NUV in outer galaxy regions, and αs are exponential scale lengths in arcsec and kpc.

Open in new tabDownload slide

5 MODELLED SPECTRAL ENERGY DISTRIBUTIONS IN THE UV RANGE

To derive the fraction of the escaping ionizing radiation, we use the two methods (e.g. Izotov et al. 2018a) based on the comparison between the observed flux in the LyC and its intrinsic flux in the same wavelength range. The intrinsic LyC flux is obtained (1) from SED fitting of the SDSS spectra simultaneously with reproducing the observed H β and H α equivalent widths (and thus corresponding observed H β and H α fluxes) or (2) from the flux of the H β emission line. The attenuated extrapolations of SEDs to the UV range along with the observed COS spectra are shown in Fig. 3. For comparison, we also show the GALEX FUV and NUV fluxes with magenta filled squares and the fluxes in the SDSS u, g, r, i, z filters with blue filled circles. We find that the spectroscopic and photometric data in the optical range are consistent, indicating that almost all the emission of our galaxies is inside the SDSS spectroscopic aperture. Therefore, aperture corrections are not needed.

Figure 3.

A comparison of the COS G140L and SDSS spectra (grey lines), and photometric data together with the modelled SEDs of the optical spectra and their extrapolation to the UV range in the rest-frame wavelength scale. GALEX FUV and NUV fluxes and SDSS fluxes in u, g, r, i, z bands are shown by magenta-filled squares with 1σ deviations and blue-filled circles, respectively. Modelled intrinsic SEDs and their extrapolation to the UV range, which are reddened by the Milky Way extinction with R(V)_MW = 3.1 and internal extinction with R(V)_int = 3.1, 2.7, and 2.4, are shown by red, black, and cyan solid lines, respectively. The black dotted lines show the 1σ spread of the SED fit reddened with the C(H β)_int values and R(V)_int = 2.7. Fluxes are in 10⁻¹⁶ erg s⁻¹ cm⁻²Å⁻¹, wavelengths are in Å.

Open in new tabDownload slide

The attenuated modelled intrinsic SEDs in the optical range and their extrapolations to the UV range (Fig. 3) are obtained by assuming that extinctions for stellar and nebular emission are equal and adopting the extinction coefficients C(H β)_MW from the NED and C(H β)_int derived from the hydrogen Balmer decrement (Table B1), and the reddening law by Cardelli et al. (1989) at |$\lambda \, \ge$| 1250Å and its extension to shorter wavelengths by Mathis (1990) with R(V)_int = 3.1 (red solid lines), R(V)_int = 2.7 (black solid lines), and R(V)_int = 2.4 (cyan solid lines). Mathis (1990) presents the data only for R(V) = 3.1. For practical use, we fit them with polynomials and adjusted in such a way to have the same values at λ = 1250Å and for a variety of R(V)s as the values from the Cardelli et al. (1989) reddening law at the same wavelength and same R(V). The dotted lines indicate the range of attenuated SEDs adopting R(V)_int = 2.7 and varying C(H β) within 1σ errors of its nominal value.

It is seen in Fig. 3 that the SDSS spectra are reproduced by the models quite well. On average, extrapolations of the attenuated SEDs to the UV range with R(V)_int  =  2.7 reproduce the observed COS spectra somewhat better with flux deviations not exceeding ∼10 per cent for most galaxies. An exception is J1014+5501, for which the difference in fluxes is as high as ∼50 per cent. This difference can possibly be caused in part by the uncertain location of the galaxy within the COS spectroscopic aperture as the acquisition exposure was failed. It could also be caused by the underestimation of interstellar extinction, which is derived from the hydrogen Balmer decrement in the SDSS spectrum. The observed FUV shape could be fit by increasing C(H β) by 0.065 from the value in Table B1. This would increase the H β fluxes by ∼15 per cent and decrease the Lyα and LyC escape fractions by a similar amount. However, in this case the extinction-corrected fluxes of H δ, H γ, and H α emission lines are considerably off from their theoretical recombination values. Furthermore, the difference between the models and observations can be caused by the non-perfect absolute flux calibration of the SDSS spectrum.

However, we note that Fig. 3 is used only for the sake of illustration to check whether extrapolation of the SED in the optical range reproduces the observed COS spectrum. But it is not used for the determination of the escaping LyC fraction. Instead the observed LyC flux is measured in COS spectra and the intrinsic LyC flux is determined by two methods mentioned above: from the extinction-corrected flux of the H β emission line I(H β) and from simultaneous fitting of the SED in the optical range and of observed equivalent widths of the H β and H α emission lines. The fluxes of latter lines are also iteratively corrected for the escaping ionizing radiation (e.g. Izotov et al. 2018b) and they determine the level of the intrinsic LyC emission. It is seen in Fig. 3 that the SED in the optical range is almost independent on R(V)_int. Consequently, the LyC escape fraction f_esc(LyC) is also almost independent of R(V)_int. This is because f_esc(LyC) is derived from the ratio of the observed to modelled intrinsic LyC fluxes with the latter fluxes being derived from data in the optical range.

6 Lyα AND LYC EMISSION

A resolved Ly|$\alpha \, \lambda$|1216 Å emission line is detected in the G160M medium-resolution spectra of five out of seven galaxies (Fig. 4). Its shape is similar to that observed in most known LyC leakers (Izotov et al. 2016a,b, 2018a,b, 2021a) and in some other galaxies at lower redshift (Jaskot & Oey 2014; Henry et al. 2015; Yang et al. 2017a; Izotov et al. 2020). Profiles with two peaks are detected in the spectra of four galaxies from the present sample with detected LyC emission, J0141−0304, J0844+5312, J1137+3605, J1352+5617, and in one galaxy with non-detected LyC emission, J1014+5501. The blue Lyα component in the latter galaxy is ∼2.5 times brighter than the red component (Fig. 4d). This fact is at variance with that for other galaxies, where the blue component is considerably weaker than the red component, and may be indicative of a gas inflow. The Lyα emission line is very weak in the spectra of two galaxies, J0130−0014 and J1157+5801. The parameters of Lyα emission are shown in Table 4.

Figure 4.

Lyα profiles. Vertical dashed lines indicate the rest-frame wavelength of 1215.67Å for Lyα. Fluxes are in 10⁻¹⁶ erg s⁻¹ cm⁻²Å⁻¹ and rest-frame wavelengths are in Å.

Open in new tabDownload slide

The observed G140L total-exposure spectra with the LyC spectral region (grey lines) and extrapolations to the UV range of predicted intrinsic SEDs in the optical range (blue dash–dotted lines) are shown in Fig. 5. Additionally, we include the attenuated extrapolations of the intrinsic SEDs (black solid lines), the same as those with R(V) = 2.7 that are shown in Fig. 3 but with different flux and wavelength scales.

Figure 5.

COS G140L spectra of our sources (grey lines). The LyC fluxes shown by the solid (detections) and dotted (upper limits of non-detections) red horizontal lines are measured in the wavelength ranges determined by their location. The extrapolations to the UV range of intrinsic SEDs and of attenuated SEDs in the optical range adopting a R(V) = 2.7 are represented by blue dash–dotted lines and black solid lines, respectively. The Lyman limit at the rest-frame wavelength 912 Å is indicated by dotted vertical lines. Strong emission lines at the observed wavelengths 1216 Å (all panels) and 1303 Å (panels c and d) are geocoronal Lyα and O i lines. Zero flux is represented by dotted horizontal lines. Insets in all panels show expanded parts of spectra with LyC emission. Fluxes are in 10⁻¹⁶ erg s⁻¹ cm⁻²Å⁻¹, wavelengths are in Å.

Open in new tabDownload slide

The level of the observed LyC continuum is indicated by horizontal red lines. The vertical dotted lines show the Lyman limit. The LyC emission is detected in the spectra of four galaxies, J0141−0304, J0844+5312, J1137+3605, and J1352+5617 (solid red lines), and only 1σ upper limits are derived in the spectra of the remaining three galaxies (dotted red lines). The measurements are summarized in Table 5.

The escape fraction f_esc(LyC) ranges between 3.1 and 4.6 per cent in four out of the seven galaxies and the 1σ upper limits of f_esc(LyC) for the remaining galaxies are shown in Table 5. We find that f_esc(LyC) obtained by the two methods are similar.

7 INDIRECT DETERMINATION OF THE LYC ESCAPE FRACTION

The direct measurement of LyC emission is the best way to derive the LyC escape fraction. However, LyC emission in most cases is weak and it difficult to detect in both the high-z and low-z galaxies. Furthermore, only HST can be used for the observation of the LyC wavelength range in galaxies with |$z\, \sim$| 0.3–1.0. Therefore, reasonable indirect indicators of LyC leakage at low and high redshift are needed, namely those which can more easily be derived from observations, to build a larger sample for statistical studies. Several possible indicators have been proposed, which are based mainly on observations of strong emission lines in the UV and optical ranges. For the analysis of possible indirect indicators, we use a sample of ∼30–50 galaxies with Mg ii emission in their SDSS spectra from Izotov et al. (2016a, b, 2018a, b, 2021a), Borthakur et al. (2014), Chisholm et al. (2017), Flury et al. (2022a), Xu et al. (2022), and this paper. The number of galaxies varies for different indicators because not all indicators are determined for all galaxies in the sample.

The Lyα escape fraction f_esc(Lyα), which is derived from the observed Lyα/H β emission line ratio, can potentially be linked with the LyC escape fraction. However, there are differences between mechanisms controlling the escape of LyC and Lyα. The LyC photons can efficiently be absorbed by neutral hydrogen and/or dust. On the other hand, Lyα photons can be ceased only via absorption by dust and via inefficient two-photon transitions. Thus, the fraction of escaping Lyα photons is expected to be higher than that of escaping LyC photons, in agreement with theoretical predictions (Jaskot & Oey 2013; Nakajima & Ouchi 2014; Dressler et al. 2015). This is seen in Fig. 6a, where almost all LyC leaking galaxies are located below the line of equal escape fractions (black solid line). There is a tendency for f_esc(LyC) to increase with increasing f_esc(Lyα) but with a large spread (see also e.g. Izotov et al. 2018b, 2021a; Flury et al. 2022b). New data do not contradict with this conclusion.

Figure 6.

Relations between the LyC escape fraction f_esc(LyC) in low-redshift LyC leaking galaxies derived by the method with the use of SED fits and constraints from the observed H β and H α equivalent widths, and (a) the Lyα escape fraction f_esc(Lyα), (b) the separation V_sep between the Lyα profile peaks, (c) the stellar mass M_⋆, (d) the [O iii]λ5007/[O ii]λ3727 emission-line flux ratios. In all panels, the galaxies from this paper are shown by red symbols and from Izotov et al. (2016a, b, 2018a, b, 2021a), Borthakur et al. (2014), Chisholm et al. (2017), Flury et al. (2022a), and Xu et al. (2022) are represented by blue symbols. LyC leakers and galaxies with upper limits of LyC emission are shown by filled circles and open circles with downward arrows, respectively. The solid line in (a) is the equality line and the solid line in (b) represents the relation from Izotov et al. (2018b).

Open in new tabDownload slide

The shape of the Lyα profile provides the best indirect indicator of the LyC leakage due to the fact that it depends on the column density of the neutral hydrogen, along the line of sight, which determines the optical depth in both the Lyα emission line and the LyC continuum. In particular, a non-zero intensity at the centre of Lyα or a small offset of its brighter red component from the centre of the line indicate low optical depth in the H i cloud. However, these indicators may be influenced by insufficient spectral resolution and uncertainties in the wavelength calibration. On the other hand, the separation between its blue and red components in medium-resolution COS spectra is less subject to these limitations. Previously, Verhamme et al. (2017) and Izotov et al. (2018b) found a tight dependence of f_esc(LyC) on the separation V_sep between the peaks of the Lyα emission line in LyC leakers. This dependence has been updated in the later paper by Izotov et al. (2021a) and in this paper. The new data also follow the relation discussed by Izotov et al. (2018b; see the solid line in Fig. 6b). There is no new galaxy in our present sample having a peak separation less than ∼300 km s⁻¹, which is considerably higher compared to the lowest peak separation of ∼150 km s⁻¹ in the sample of low-z leakers shown in Fig. 6b. The relation by Izotov et al. (2018b) is likely not applicable for complex Lyα profiles with three or more peaks, indicating considerable direct Lyα escape, in addition to escape through scattering in the neutral gas (Naidu et al. 2022). The Lyα profile in only one galaxy, J1243+4646, from the Izotov et al. (2018b) sample consists of three peaks with the peak separations of 143 and 164 km s⁻¹. This galaxy does not change significantly the shape of the relation shown in Fig. 6b by the solid line because most of the galaxies in the sample have two Lyα peaks.

The new observations of Mg ii-selected galaxies (red symbols) support previous findings on the existence of the tight relation between f_esc(LyC) and V_sep. However, the application of this relation for galaxies observed during epoch of reionization is limited because of incomplete ionization of the IGM and thus high optical depth for Lyα emission.

Low galaxy stellar masses are also considered as a possible indicator of high f_esc(LyC) (Wise et al. 2014; Trebitsch et al. 2017). Indeed, there is a trend of decreasing f_esc(LyC) with increasing stellar mass in galaxies with the detected LyC continuum (filled circles in Fig. 6c). However, Izotov et al. (2021a) found several SFGs with |$M_\star \, \lt $| 10⁸ M_⊙ and non-detected LyC (blue open circles in Fig. 6c), considerably weakening the anticorrelation between f_esc(LyC) and M_⋆. New data in the present paper are in agreement with the conclusion of no or only a weak correlation between f_esc(LyC) and M_⋆.

Jaskot & Oey (2013) and Nakajima & Ouchi (2014) proposed a high O₃₂ ratio as an indication of escaping ionizing radiation. However, the increase of this ratio is caused not only by decreasing optical depth of the neutral hydrogen around the H ii region, but also by increasing ionization parameter and/or decreasing metallicity. These effects are difficult to separate. O₃₂ in low-redshift galaxies can easily be derived from their spectra in the optical range. This quantity is known for all low-z LyC leakers.

The relation between f_esc(LyC) and O₃₂ has been discussed by Faisst (2016), Izotov et al. (2018b, 2021a), and Flury et al. (2022b). Its updated version from Izotov et al. (2021a) is presented in Fig. 6d, which shows a trend of increasing f_esc(LyC) with increasing of O₃₂, but with a substantial scatter. This scatter, in part, can be caused by a variety of scenarios with leakage through channels with low optical depth and their orientation relative to the observer. Similar conclusion can be drawn from the Flury et al. (2022b) data. Therefore, a high O₃₂ can be used for selection of the LyC leaking candidates, but it is not a very certain indicator of high f_esc(LyC) (Izotov et al. 2018b; Nakajima et al. 2020).

8 MG ii DIAGNOSTICS

Henry et al. (2018) and Chisholm et al. (2020) have proposed to use the double resonance line of Mg ii λ2796, 2803 in emission as an indicator of escaping LyC emission based on the fact that its escape fraction correlates with the Lyα escape fraction. Later, Xu et al. (2022) also proposed Mg ii as low-z tracer of Lyα and LyC, Naidu et al. (2022) pointed out that Mg ii λ2796/λ2803 line ratio is higher in |$z\, \sim$| 2 galaxies with higher f_esc(LyC). Following these papers, we consider the properties of Mg ii emission and their relations with the Lyα and LyC escape fractions. For many low-redshift LyC leaking galaxies (Izotov et al. 2016a,b, 2018a,b, 2021a; Flury et al. 2022a; Xu et al. 2022, this paper), the wavelength range with the redshifted Mg ii λ2796, 2803 emission lines is covered by the SDSS spectra (Figs 7–8). However, these redshifted lines are outside the wavelength range of SDSS spectra from the releases earlier than DR10 of some LyC leakers with lowest redshifts of z ≈ 0.3 (for example, J0925+1403, J1011+1947, J1442−0209). XShooter spectra covering the Mg ii emission (Fig. 9) are also available for some LyC galaxies (Guseva et al. 2020), including those with z ≈ 0.3.

Figure 7.

Mg ii λ2796, 2803 emission lines in SDSS spectra of new galaxies discussed in this paper. Red dotted lines indicate the rest-frame centres of the lines.

Open in new tabDownload slide

Figure 8.

Mg ii λ2796, 2803 emission lines in SDSS spectra of LyC leaking galaxies from Izotov et al. (2016a, b, 2018b, c, 2021a). Red dotted lines indicate the rest-frame centres of the lines.

Open in new tabDownload slide

Figure 9.

Mg ii λ2796, 2803 emission lines in XShooter spectra of LyC leaking galaxies from Guseva et al. (2020). Red dotted lines indicate the rest-frame centres of the lines.

Open in new tabDownload slide

We note that Mg ii emission is located in the noisy parts of the SDSS spectra. Because of weakness of these lines, they cannot be measured with high accuracy. The spectral resolution of SDSS spectra is insufficient to determine the Mg ii emission line profiles. On the other hand, the accuracy of measurements and spectral resolution are better for XShooter spectra. Because of the limitations for the SDSS sample, we consider only two characteristics for the entire SDSS+XShooter sample, the extinction-corrected O₃Mg₂ = [O iii]λ5007/Mg ii λ2796+2803 and Mg₂ = Mg ii λ2796/Mg ii λ2803 flux ratios, which are less subject to the uncertainties compared to those in fitting of the Mg ii emission-line profiles.

Mg ii emission is detected in most LyC leaking galaxies if it falls in the wavelength range of SDSS spectra, as expected in the case of low neutral gas column densities. The two galaxies with very little (or no) Mg ii detections in Fig. 7 (J0130−0014 and J1157+5801) also do not have LyC detections, illustrating how a non-detection of Mg ii can also lead to a non-detection of LyC. However, there is one possible exception. The galaxy J1121+3806 has f_esc(LyC) ∼35 per cent and strong and narrow Lyα emission line (Izotov et al. 2021a). On the other hand, Mg ii emission in this galaxy is barely seen (Fig. 8d). Thus, the high LyC leakage is possibly not always associated with the presence of strong Mg ii emission. However, the SNR of SDSS spectrum is low and this galaxy merits deeper observations (King et al. in preparation).

Figs 10a and b show the dependencies of the Lyα escape fraction f_esc(Lyα) on the O₃Mg₂ and Mg₂, respectively. It is seen that f_esc(Lyα) is almost independent of both the O₃Mg₂ and Mg₂ ratios.

Figure 10.

(a) and (b) Relations between the Lyα escape fraction f_esc(Lyα) in low-redshift LyC leaking galaxies and the O₃Mg₂ = [O iii]λ5007/Mg ii λ2796+2803 ratio and the Mg₂ = Mg ii λ2796/Mg ii λ2803 ratio, respectively. (c) and (d) Relations between the LyC escape fraction f_esc(LyC) in low-redshift LyC leaking galaxies derived from the SED fits and the O₃Mg₂ and Mg₂, respectively. The galaxies from this paper and from Izotov et al. (2016a, b, 2018a, b, 2021a), Borthakur et al. (2014), Chisholm et al. (2017), Flury et al. (2022a), and Xu et al. (2022) are represented by red and blue symbols, excluding objects observed with the XShooter (Fig. 9, Guseva et al. 2020), which are shown by black symbols. LyC leakers and galaxies with upper limits of LyC emission are shown by filled circles and open circles with downward arrows, respectively. Error bars in black (all panels) represent average 1σ deviations, whereas error bars in cyan and magenta [panels (b) and (d)] are minimal and maximal 1σ errors of Mg₂ for the SDSS+XShooter sample, respectively. The values of O₃Mg₂ and Mg₂ for all galaxies are calculated in this paper.

Open in new tabDownload slide

Mg₂ in two galaxies, J1127+4610 and J1455+6107 (Figs 8e and n, Izotov et al. 2021a), in Fig. 10b is considerably above the value of 2 in the case of zero optical depth in Mg ii lines, which is unlikely. However, we note that Mg₂s in these two galaxies are measured with the largest errors, approximately two times higher than typical errors for objects shown in Fig. 10b. Furthermore, these lines in all galaxies were not corrected for interstellar or stellar photospheric Mg ii absorption. Equivalent widths of these absorption lines are somewhat uncertain. Guseva et al. (2019) adopted equal equivalent widths of ∼0.5 Å for each of Mg ii absorption lines, whereas Pérez-Ràfols et al. (2015) derived 2.33 Å for both lines, which are consistent with the value of ∼1Å for Mg ii λ2796 absorption line in SFGs with stellar masses |$\lesssim$|10^9.5 M_⊙ (Martin et al. 2012) and the values adopted by Prochaska, Kasen & Rubin (2011). All these values are lower than equivalent widths of Mg ii emission lines (Table B1). Assuming that equivalent widths of Mg ii λ2796 and λ2803 absorption lines are equal and correcting emission lines by multiplying with (EW_em+EW_abs)/EW_em results in a reduction of Mg₂ ratio if this ratio is above 1. This is because the equivalent width of the Mg ii λ2796 emission line is greater than that of the Mg ii λ2803 emission line. The effect is larger for higher values of Mg₂ reducing the number of galaxies with Mg₂ above 2.

Using the analytic work of Chisholm et al. (2020), a Mg₂ of 1.3 would correspond to an Mg ii 2803Å optical depth of 0.43 (or a 2796 optical depth near 1). For the typical abundances 12 + log(O/H) = 7.9 - 8.0 of the sample, that would lead to H   i column densities near 9.4 × 10¹⁶ cm⁻², which is very close to being optically thin for the LyC emission. It is notable that Mg₂ in all five galaxies with high f_esc(LyC) observed with the high SNR at the XShooter by Guseva et al. (2020) is very close to 2 (black symbols in Fig. 10b), in agreement with expectations for the low optical depth (e.g. Chisholm et al. 2020).

In Figs 10c and d we show the relations of f_esc(LyC) with the O₃Mg₂ and Mg₂ flux ratios, respectively. We note an interesting feature in Figs 10b and d that the LyC leakers have preferentially Mg|$_2\, \gtrsim$| 1.3, as expected because high values of Mg₂ indicate low optical depth (Chisholm et al. 2020). Similarly, Naidu et al. (2022) found that galaxies with low f_esc(LyC) have preferentially low Mg|$_2\, \sim$| 0.9. Possibly, a tendency of increasing f_esc(LyC) with increasing of the O₃Mg₂ and Mg₂ is present albeit scatter of the data is large.

The statistics in Fig. 10 are small and subject to large errors of individual measurements. Therefore, for a comparison, we selected ∼6000 galaxies with z ≥ 0.3 from the sample of compact SFGs by Izotov et al. (2021c) in which both the Mg ii λ2796 and 2803 emission lines were observed. The errors of Mg ii line fluxes in this sample are also large. However, large statistics in each bin of the O₃Mg₂ and Mg₂ flux ratios considerably reduce the impact of uncertain individual values. These galaxies constitute 60 per cent of the total number of galaxies in the catalogue of Izotov et al. (2021c) with z ≥ 0.3. Mg ii in the remaining galaxies is either in absorption or only one of the two lines is detected.

The distribution of Mg₂ for selected galaxies is shown in Fig. 11a. This distribution is broad and approximately 1/3 galaxies have Mg|$_2\, \gt $| 2. The scatter is likely caused not only by errors of measurements. It remains even if only brightest galaxies with well measured Mg ii fluxes are considered (compare Figs 11c and d). On the other hand, correction for underlying absorption can make the distribution narrower together with the decreasing number of galaxies with Mg|$_2\, \gt $| 2. We find that nearly 2/3 of the sample is characterized by a Mg₂ > 1.3, implying that most of selected compact SFGs could possibly be LyC leakers.

Figure 11.

(a) Distribution of the flux ratio Mg₂ = Mg ii λ2796/λ2803 in the ∼6000 SDSS compact SFGs with redshifts |$z\, \ge$| 0.3. (b) Distribution of ionizing photon production efficiency ξ_ion for the same sample as in (a). Dotted line in (a) separates expected LyC leakers (Mg|$_2\, \ge$| 1.3) from non-LyC leakers (Mg|$_2\, \lt $| 1.3), whereas the dotted line in (b) is the canonical value of ionizing photon production efficiency commonly adopted to complete reionization. (c) Relation between Mg₂ and ξ_ion. The LyC leaking galaxies with detected LyC emission and its upper limits, the same as in Figs 6 and 10, are represented by black filled and open circles, respectively. Objects with the fluxes of Mg ii λ2796, 2803 emission lines obtained from the XShooter spectra (Guseva et al. 2020) are encircled. (d) Same as in (c), but are shown only galaxies from the SDSS with H β fluxes above 5 × 10⁻¹⁶ erg s⁻¹cm⁻² and equivalent widths of the Mg ii λ2796 emission line above 10 Å. In all panels, SDSS compact SFGs with equivalent widths EW(H β) ≥100 Å and <100 Å are shown in blue and red, respectively. Error bars in (c) and (d) represent average 1σ errors.

Open in new tabDownload slide

The distribution of ionizing photon production efficiency ξ_ion for the same galaxies is shown in Fig. 11b. Here, ξ_ion = N(LyC)/L_ν, where N(LyC) and L_ν are the production rate of the LyC radiation in photons s⁻¹ and the intrinsic monochromatic luminosity at the rest-frame wavelength of 1500 Å in erg s⁻¹ Hz⁻¹. It is seen that log ξ_ion in the sample galaxies is high. In most galaxies, it is above the threshold of 25.2, adopted in models of reionization (e.g. Robertson et al. 2013). Finally, we show the relations between log ξ_ion and Mg₂ for all selected SDSS galaxies (Fig. 11c) and brightest SDSS galaxies in the sense that the H β fluxes in these galaxies are above 5 × 10⁻¹⁶ erg s⁻¹cm⁻² and equivalent widths of the Mg ii λ2796 emission line are above 10 Å (Fig. 11d). The unshaded region in Figs 11c and d is populated by the galaxies with Mg|$_2\, \ge$| 1.3 and log |$\xi _{\rm ion}\, \ge$| 25.2, which constitute nearly half of the total sample and somewhat more for the brightest galaxies. Most of low-z LyC leakers (black filled circles) are located in this region. The few galaxies with log ξ_ion below 25.2 are only from the LzLCS sample by Flury et al. (2022a), Flury et al. (2022b), which contains, in general, lower-excitation H ii regions compared e.g. with the galaxies from the Izotov et al. (2016a, b, 2018a, b, 2021a) sample. Thus, the criterion Mg|$_2\, \lt $| 1.3 can be a useful cut to selected LyC leaker candidates at low- and high-redshifts due to the fact that strong Mg ii emission is present in most LyC leaking galaxies.

9 CONCLUSIONS

We present new HST COS low- and medium-resolution spectra of seven compact SFG in the redshift range z = 0.3161–0.4276, with various O₃Mg₂ = [O iii]λ5007/Mg ii λ2796+2803 and Mg₂ = Mg ii λ2796/Mg ii λ2803 emission-line ratios. We aim to obtain properties of leaking LyC and resolved Lyα emission and to study the dependence of leaking LyC emission on the characteristics of Mg ii emission along with other indirect indicators of escaping ionizing radiation. This study is an extension of the work reported earlier in Izotov et al. (2016a, b, 2018a, b, 2021a). Our main results are summarized as follows:

• Emission of LyC is detected in four out of the seven galaxies with the escape fraction f_esc(LyC) in the range between 3.1 per cent (J1137+3605) and 4.6 per cent (J0844+5312). Only upper limits f_esc(LyC) ∼ 1–3 per cent are obtained for the remaining three galaxies.

• A Lyα emission line with two peaks is observed in the spectra of five galaxies. The Lyα emission line in two galaxies, J0130−0014 and J1157+5801, is very weak. Our new observations support a strong anticorrelation between f_esc(LyC) and the peak velocity separation V_sep of the Lyα profile, confirming the finding of Izotov et al. (2018b, 2021a) and making V_sep the most robust indirect indicator of LyC radiation leakage.

• Other characteristics such as O₃₂ ratio, escape fraction of the Lyα emission line f_esc(Lyα) and the stellar mass M_⋆ show weak or no correlations with f_esc(LyC), with a high spread of values, in agreement with earlier studies by e.g. Izotov et al. (2016b, 2018a, 2021a) and Flury et al. (2022b).

• We study the characteristics of Mg ii λ2796+2803 emission, such as O₃Mg₂ and Mg₂ ratios, as possible indirect indicators of escaping LyC emission. We find that galaxies with detected LyC emission have preferentially Mg|$_2\, \ge$| 1.3, the latter indicating low optical depths. A high Mg₂ ratio of ≥1.3 can be used to select LyC leaker candidates. A tendency of an increase of f_esc(LyC) with increasing of both the O₃Mg₂ and Mg₂ is possibly present. However, there is substantial scatter in these relations due to the low signal-to-noise ratio in the blue part of the SDSS spectra near the observed Mg ii emission not allowing their use for reliable prediction of f_esc(LyC).

• We find that galaxies with Mg|$_2\, \ge$| 1.3 and ionizing photon production efficiency ξ_ion greater than the value of 10^25.2 erg⁻¹ Hz used in modelling of the process of reionization of the Universe (e.g. Robertson et al. 2013) constitute ∼40 per cent of all compact SFGs at redshift |$z\, \ge$| 0.3, which were selected by Izotov et al. (2021c) from the Data Release 16 of the SDSS.

• A bright compact star-forming region superimposed on a low-surface-brightness component is seen in the COS NUV acquisition images of five galaxies (two images are missing due to technical problems). The SB at the outskirts of our galaxies can be approximated by an exponential disc, with a scale length of ∼0.20–0.63 kpc. This is approximately four times lower than the scale lengths of the LyC leakers observed by Izotov et al. (2016b, 2018a, b), but is similar to that in low-mass galaxies with |$M_\star \, \lt $| 10⁸ M_⊙ by Izotov et al. (2021a). Part of this difference may be explained by acquisition exposure times that are approximately two times shorter compared to those used by Izotov et al. (2016b, 2018a, b), resulting in less deep images.

• The star formation rates in the range SFR ∼ 4–36 M_⊙ yr⁻¹ and the metallicities of our new galaxies, ranging from 12 +logO/H = 7.81 to 8.06, are overlapping with those in the LyC leakers studied by Izotov et al. (2016a, b, 2018a, b, 2021a).

ACKNOWLEDGEMENTS

Based on observations made with the National Aeronautics and Space Agency (NASA)/European Space Agency (ESA) Hubble Space Telescope, obtained from the data archive at the Space Telescope Science Institute. Support for this work was provided by NASA through grant number HST-GO-15845 from the Space Telescope Science Institute. STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. YI and NG acknowledge support from the National Academy of Sciences of Ukraine by its priority project no. 0122U002259 ‘Fundamental properties of the matter and its manifestation in micro world, astrophysics, and cosmology’. Funding for Sloan Digital Sky Survey-III (SDSS-III) has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the U.S. Department of Energy Office of Science. The SDSS-III website is http://www.sdss3.org/. SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration. Galaxy Evolution Explorer (GALEX) is a NASA mission managed by the Jet Propulsion Laboratory. This research has made use of the NASA/Infrared Processing and Analysis Center (IPAC) Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.

DATA AVAILABILITY

The data underlying this article will be shared on reasonable request to the corresponding author.

REFERENCES