EMPRESS. II. Highly Fe-enriched Metal-poor Galaxies with ∼1.0 (Fe/O)_⊙ and 0.02 (O/H)_⊙: Possible Traces of Supermassive (>300 M_⊙) Stars in Early Galaxies^*†‡

Izotov et al. (2006b) report that Ne/O and Ar/O ratios little depend on metallicity because the neon, argon, and oxygen are all α-elements, which are produced by the nuclear fusion of α-particles inside stars. As shown in panels (a) and (b) of Figure 3, we find that our metal-poor galaxy sample shows almost constant values of log(Ne/O) ∼ −0.8 and log(Ar/O) ∼ −2.5 within a scatter of ± 0.2 dex, which are almost consistent with the solar abundance ratios. We also find that the Ne/O and Ar/O ratios are consistent with those of local galaxies reported by Izotov et al. (2006b) within the scatter. The consistency suggests that our metal-poor galaxy sample also shows no metallicity dependence in Ne/O and Ar/O ratios.

Note that the Ar/O ratio might slightly decrease in the range of 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ≳ 8.2 in our sample and the Izotov et al. (2006b) sample, in contrast to the Ne/O ratios. The Ar/O ratio is expected to be constant and consistent with the solar abundance, log(Ar/O)_⊙ = −2.29, because there seems to be no physical reason for the Ar/O (i.e., α-element ratio) decrease at 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ≳ 8.2. This may be explained by the underlying unknown systematics in the Ar/O estimation at 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ≳ 8.2. We do not discuss it further in this paper because the Ar/O abundances at relatively higher metallicities are out of the scope of this paper.

As suggested by previous studies (Pérez-Montero & Contini 2009; Andrews & Martini 2013; Pérez-Montero et al. 2013), N/O ratios of SFGs present a plateau at log(N/O) ∼ −1.6 in the range of 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ≲ 8.0 and a positive slope at higher metallicities as a function of metallicity. Panel (c) of Figure 3 presents model calculations of the N/O evolution (Vincenzo et al. 2016), which also show the plateau and positive slope. The plateau basically results from the primary nucleosynthesis of massive stars, while the positive slope is mainly attributed to the secondary nucleosynthesis of low- and intermediate-mass stars (e.g., Vincenzo et al. 2016). We briefly describe the two nitrogen production processes below.

• 1.  
  Primary nucleosynthesis: Inside a metal-poor star, protons are burned through the proton–proton (p–p) chain reaction, and little nitrogen is produced at this stage. Nitrogen elements are mainly produced after the formation of a heavy-element core (e.g., O and C) and ejected into the ISM by SNe, for stars more massive than ∼8 M_⊙.
• 2.  
  Secondary nucleosynthesis: Metal-rich stars efficiently burn hydrogen through the carbon-nitrogen-oxygen (CNO) cycle, where nitrogen elements accumulate because ¹⁴N fusion (¹⁴N+p → ¹⁵O+γ) is the slowest process in the CNO cycle. Then, nitrogen is ejected through stellar winds during the asymptotic giant branch phase, ∼1 Gyr after the birth of low- and intermediate-mass stars.

As shown in panel (c) of Figure 3, most of our metal-poor galaxies have N/O ratios of log(N/O) < −1.5 (i.e., less than ∼30% of the solar N/O ratio). Especially, HSC J1631+4426 has a strong, 2σ upper limit of log(N/O) < −1.71, and J0811+4730 shows a low N/O ratio of log(N/O) = −1.53. The N/O values of the two EMPGs (HSC J1631+4426 and J0811+4730) will be discussed again in Section 5.1.3. These low N/O ratios suggest that our metal-poor galaxies have not yet started the secondary nucleosynthesis owing to their low metallicities and young stellar ages.

We also find that several galaxies of our metal-poor galaxy sample have relatively low N/O ratios compared to the model lines of the Vincenzo et al. (2016) and Izotov et al. (2006b) SFGs at 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 8.0. Vincenzo et al. (2016) find that the N/O plateau is lowered under the assumption of high sSFR or the top-heavy initial mass function (IMF). Indeed, the members of our metal-poor galaxy sample have high sSFRs (∼300 Gyr⁻¹) compared to those of the Izotov et al. (2006b) SFGs (∼1–10 Gyr⁻¹) because we aim to obtain galaxies with high sSFRs in this study. Thus, the N/O differences between our metal-poor galaxy sample and the Izotov et al. (2006b) SFG sample are explained by the sample selection.

In panel (d) of Figure 3, we find that our metal-poor galaxies show a decreasing trend in Fe/O ratio as metallicity increases. The same decreasing Fe/O trend is found in the SFG sample of Izotov et al. (2006b). Most of our metal-poor galaxies have Fe/O ratios comparable to the SFG sample of Izotov et al. (2006b). Three EMPGs, HSC J1631+4426, SDSS J2115−1734, and J0811+4730 (encircled by a red or orange circle), show relatively high Fe/O ratios, log(Fe/O) > −1.7, among our metal-poor galaxies. Especially, we find that HSC J1631+4426 and J0811+4730, two of the lowest-metallicity galaxies with 0.016 and 0.019 (O/H)_⊙, have high Fe/O ratios of log(Fe/O) = $-{1.25}_{-0.53}^{+0.17}$ and log(Fe/O) = $-{1.06}_{-0.31}^{+0.09}$, respectively, which are comparable to the solar Fe/O ratio, log(Fe/O)_⊙ = −1.19. In this paper, we mainly focus on the two representative EMPGs, HSC J1631+4426 and J0811+4730, which interestingly show high Fe/O ratios. Note again that J0811+4730 is an EMPG reported by Izotov et al. (2018). Table 5 summarizes the Fe/O ratios and He ii λ4686/Hβ ratios (discussed in Section 5.2.2) of the two representative EMPGs (HSC J1631+4426 and J0811+4730), which play an important role in this paper. We also show the fluxes of the two key emission lines of [Fe iii] λ4658 and He ii λ4686 in Table 5.

Table 5. Parameters of Two Representative EMPGs

ID	12+log(O/H)	log(Fe/O)	log(He ii/Hβ)	F([Fe iii])	F(He ii)	Reference
				(erg s−1 cm−2)	(erg s−1 cm−2)
(1)	(2)	(3)	(4)	(5)	(6)	(7)
HSC J1631+4426	6.90 ± 0.03	$-{{1.25}_{-0.31}^{+0.17}}_{-0.22}$ a	$-{1.58}_{-0.08}^{+0.07}$	12.0 ± 5.02	30.5 ± 5.04	This paper
J0811+4730	6.98 ± 0.02	$-{{1.06}_{-0.09}^{+0.09}}_{-0.22}$ a	−1.64 ± 0.03	6.55 ± 1.26	28.6 ± 1.89	I18

Notes. Column (1): ID. Column (2): gas-phase metallicity. Column (3): abundance ratio of log(Fe/O). Column (4): emission-line ratio of log(He ii/Hβ). Columns (5)–(6): emission-line fluxes of [Fe iii] λ4658 and He ii λ4686 in units of 10⁻¹⁸ erg s⁻¹ cm⁻². Column (7): reference. I18 represents Izotov et al. (2018).

^a We show two kinds of Fe/O lower errors. The first term is the statistical error propagated from spectral noise, and the second one is the systematic error originating from ICF uncertainties explained in Section 4.2.

Download table as:  ASCIITypeset image

To characterize the two EMPGs (HSC J1631+4426 and J0811+4730) with a high Fe/O ratio, we also compare our metal-poor galaxies with Galactic stars (Gratton et al. 2003; Cayrel et al. 2004; Bensby et al. 2013) in Figure 4. The Galactic star samples are composed of dwarf or subdwarf stars. The solid line here represents a stellar Fe/O evolution model under the assumption that gas is enriched by massive stars with 9–100 M_⊙ (Suzuki & Maeda 2018). The gaseous abundance ratios at the time of the star formation are imprinted in the stellar abundance patterns because stars are formed from gas. As explained in Section 1, the Fe/O ratio increases at 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ≳ 8.0 owing to the contribution of SNe Ia ∼1 Gyr after the start of the star formation. Surprisingly, we find in Figure 4 that the two EMPGs (HSC J1631+4426 and J0811+4730) deviate from the observational results of Galactic stars and the Fe/O evolution model. Below, we mainly focus on the discussion of the two EMPGs HSC J1631+4426 and J0811+4730, unless described explicitly. Note that our metal-poor galaxies present gas-phase Fe/O ratios in nebulae, while the Galactic stars show the Fe/O ratio obtained from stellar atmospheric absorption lines. The nebular abundance ratios are subject to change by the effects of SNe, stellar wind, and galactic inflow in a short timescale (i.g., ≲10 Myr). By contrast, the abundance ratios of Galactic stars (i.e., dwarf and subdwarf stars) change little across the cosmic time because the element production proceeds very slowly and heavy elements such as oxygen and iron are produced little in low-mass stars such as dwarf and subdwarf stars. Thus, the abundance ratios of the dwarf and subdwarf stars are almost fixed at the time of the star formation and can be regarded as tracers of the past chemical evolution. The dwarf and subdwarf stars with low metallicities of 0.01–0.1 Z_⊙ are as old as ∼12 Gyr (e.g., Bensby et al. 2013), which is right after the Milky Way formation. The two EMPGs (HSC J1631+4426 and J0811+4730), whose Fe/O ratios deviate from the Galactic stars and the Fe/O model, suggest that their Fe/O ratios have increased for some reason after the galaxy formation.

Zoom In Zoom Out Reset image size

We discuss a possibility that the Fe/O ratios might be overestimated by the contribution of hard EUV radiation (e.g., AGNs) or shock heating (e.g., SNe). Collisional excitation lines of low-ionization ions such as [N ii], [S ii], and [Fe iii] are sensitive to hard EUV radiation or shock heating. For example, the strong [N ii] λ6584 and [S ii] λλ6717, 6731 lines are often used in the Baldwin–Phillips–Terlevich (BPT) diagram (Baldwin et al. 1981) as indicators of hard EUV radiation from an AGN. The AGN photoionization models (e.g., Groves et al. 2004a, 2004b) actually predict [N ii] and [S ii] line intensities stronger than the stellar photoionization models. The [N ii] and [S ii] lines are enhanced by the power-law radiation of an AGN because a partially ionized zone is formed at the edge of ionized gas. In addition, shock gas models (Allen et al. 2008) also demonstrate that the [N ii] and [S ii] lines are boosted when the shock heating contributes to the line emission. The [N ii] and [S ii] lines are enhanced by the shock heating because low-ionization ions such as N⁺ and S⁺ are abundant in the recombination zone behind a shock front (Allen et al. 2008). Thus, abundances of low-ionization ions can be overestimated by the hard EUV radiation or shock heating. Especially, because the ionization potentials of N⁰ (14.5 eV) and Fe⁺ (16.2 eV) are very close, the N/O and Fe/O ratios can be overestimated simultaneously if the hard EUV radiation or shock heating contributes. However, the N/O ratios of HSC J1631+4426 and J0811+4730 are as low as galaxy chemical evolution models at log(N/O) ∼ −1.6. Only Fe/O ratios of the two EMPGs deviate from the chemical evolution models. Thus, we rule out the possibility that the Fe/O ratios are overestimated by the hard EUV radiation or shock heating.

We also discuss another possibility that the [Fe iii] λ4658 emission line might be contaminated by a C iv λ4659 emission line, which can lead to the overestimation of the Fe/O ratio. If the C iv λ4659 line exists, the [Fe iii] λ4658 and C iv λ4659 lines may be unresolved owing to the low spectral resolutions of the FOCAS and MODS spectroscopy. The C iv λ4659 line is a C³⁺ recombination line, which is characterized by stellar winds from Wolf-Rayet (W-R) stars. Typical galaxies with W-R features (so-called "W-R galaxies") show broad emission lines such as C iv λ4659, He ii λ4686, and C iv λ5808 with FWHMs ∼ 3000 km s⁻¹ (e.g., López-Sánchez & Esteban 2010). Stellar wind velocities are expected to decrease with decreasing metallicity (e.g., Nugis & Lamers 2000; Crowther & Hadfield 2006) because atmospheric opacity of metal-poor stars becomes smaller than that of metal-rich stars. However, W-R galaxies/stars at low metallicities are very rare and have not yet been studied very well. One of the most metal-poor galaxies, I Zw 18 (12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ = 7.16, 0.030 Z_⊙), shows prominent broad emission lines of C iv λ4659, He ii λ4686, and C iv λ5808 with FWHMs ∼ 2600–3600 km s⁻¹ (Δλ =50–55 Å, Legrand et al. 1997), which are as broad as typical metal-rich W-R galaxies (∼3000 km s⁻¹; López-Sánchez & Esteban 2010). This result may suggest that line widths of C iv λ4659, He ii λ4686, and C iv λ5808 may be fairly large even when a galaxy has a few percent solar metallicity. In contrast to I Zw 18, the two EMPGs discussed in this paper (HSC J1631+4426 and J0811+4730 with ∼0.02 Z_⊙) show no detection of broad emission lines of C iv λ4659, He ii λ4686, and C iv λ5808. The detected [Fe iii] λ4658 line of HSC J1631+4426 and J0811+4730 (FWHMs ∼ 4 Å, i.e., ∼260 km s⁻¹) are much narrower than the broad lines of I Zw 18 (FWHMs ∼ 2600–3600 km s⁻¹). Thus, the C iv λ4659 line that originated from stellar winds may not contaminate the [Fe iii] λ4658 line of HSC J1631+4426 and J0811+4730 significantly.

We briefly discuss an extreme case of a very narrow C iv λ4659 line (≲260 km s⁻¹), although the case is unlikely because the line width of ≲260 km s⁻¹ is one order of magnitude smaller than the broad C iv λ4659, He ii λ4686, and C iv λ5808 lines of I Zw 18 (FWHMs ∼ 2600–3600 km s⁻¹). Even if the C iv λ4659 line can be extremely narrow, nondetection of the C iv λ5808 line suggests that the C iv λ4659 line is very faint in HSC J1631+4426 and J0811+4730. Note that the C iv λ5808 intensity is almost comparable to that of C iv λ4659 (López-Sánchez & Esteban 2010). Thus, we conclude that the [Fe iii] λ4658 line is not contaminated by the C iv λ4659 line significantly even in the extreme case.

As described above, we have found that the Fe/O ratios of the two EMPGs (HSC J1631+4426 and J0811+4730) deviate from the Galactic stars and the Fe/O chemical evolution models and have ruled out the possibility that the Fe/O ratios are overestimated. Below, we discuss three scenarios (i)–(iii) that might be able to explain the Fe/O deviation of the two EMPGs.

(i) Preferential Dust Depletion: The first scenario is the preferential dust depletion of iron, suggested by Rodriguez & Rubin (2005) and Izotov et al. (2006b). Rodriguez & Rubin (2005) and Izotov et al. (2006b) explain that gas-phase Fe/O ratios decrease as a function of metallicity in the range of 12 + $\mathrm{log}({\rm{O}}/{\rm{H}})$ ≲ 8.5 because iron elements are depleted into dust more effectively than oxygen. The depletion becomes dominant in a higher metallicity range, where the dust production becomes more efficient. For dust-free (i.e., metal-poor) galaxies, gas-phase Fe/O ratios are expected to become comparable to the observational results of Galactic stars and the Fe/O evolution model. Although the dust depletion may explain the negative Fe/O slope, it does not explain the fact that the two EMPGs (HSC J1631+4426 and J0811+4730) show higher Fe/O ratios than the Galactic stars and models at fixed metallicity. In addition, as we have seen in Section 4.1 and Table 2, most of our metal-poor galaxies show E(B − V) ∼ 0 (i.e., less dusty). At least, we do not find evidence that galaxies with a larger metallicity show larger color excesses (i.e., dustier). This means that the Fe/O decrease of our sample is not attributed to the dust depletion. Based on these facts, we rule out the first scenario.

(ii) Metal Enrichment and Gas Dilution: The second scenario is a combination of metal enrichment and gas dilution caused by inflow. In this scenario, we assume that EMPGs are formed from metal-enriched gas with the solar metallicity and solar Fe/O ratio. The Fe/O evolution models suggest that the Fe/O ratio increases at 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ≳ 8.0 owing to the contribution of SNe Ia ∼1 Gyr after the start of the star formation. Such mature galaxies also tend to have solar metallicity (i.e., O/H) at the same time. If primordial gas (i.e., almost metal free) falls into the metal-enriched galaxies, the metallicity (i.e., O/H) decreases while the Fe/O ratio does not change. At first glance, this scenario seems to explain the Fe/O deviation of the two EMPGs. However, if the second scenario is true, both the Fe/O and N/O ratios should match solar abundances because the N/O ratio also reaches the solar N/O ratio, log(N/O)_⊙ = −0.86, at solar metallicity (Sections 1 and 5.1.2). As we have seen in panel (c) of Figure 3, the two deviating EMPGs, HSC J1631+4426 and J0811+4730 (encircled by a red or orange circle at 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 7.0), have a strong 2σ upper limit of <0.14 (N/O)_⊙ and a low value of 0.21 (N/O)_⊙, respectively. These low N/O ratios suggest that the two deviating EMPGs are experiencing the primary nucleosynthesis, not the secondary nucleosynthesis expected to start ∼1 Gyr after the onset of the star formation. This also means that the high Fe/O ratios are not attributed to the SNe Ia, which arise ∼1 Gyr after the onset of the star formation. This conclusion is also consistent with the fact that the two EMPGs are very young, ≲ 50 Myr (Paper I). We exclude the second scenario because the second scenario does not explain the observed Fe/O and N/O ratios simultaneously.

(iii) Supermassive Star beyond 300 M_⊙: The third scenario is the contribution of supermassive stars beyond 300 M_⊙. Supermassive stars beyond 300 M_⊙ eject much iron at the time of core-collapse SN explosion. Ohkubo et al. (2006) have calculated yields from core-collapse SNe under the assumption of the progenitor stellar mass with 500–1000 M_⊙, obtaining ∼2–40 (Fe/O)_⊙. In the supermassive stars beyond 300 M_⊙, an iron core grows until the iron core occupies more than 20% of the stellar mass. Although massive stars with 140–300 M_⊙ undergo thermonuclear explosions triggered by pair-creation instability (PISNe; Barkat et al. 1967), supermassive stars beyond 300 M_⊙ are too massive to trigger PISNe and thus continue the iron core growth. The supermassive stars beyond 300 M_⊙ eject a large amount of iron by a jet stream from the massive iron core during the SN explosion. On the other hand, the core-collapse SNe of typical-mass stars (10–50 M_⊙) eject gas with an average of ∼0.4 (Fe/O)_⊙ (Tominaga et al. 2007; IMF integrated in the range of 10–50 M_⊙), which is below the solar Fe/O ratio. Yields of SNe Ia calculated by Iwamoto et al. (1999) show ∼40 (Fe/O)_⊙. Of the three types of SNe, only the SNe Ia and the SNe of supermassive stars (>300 M_⊙) contribute to the iron enrichment larger than the solar Fe/O ratio. As we have discussed in the second scenario above, the low N/O ratios of the two EMPGs suggest that their high Fe/O ratios are not explained by SNe Ia. Ruling out the SNe Ia, we find that the remaining possibility is the contribution from the SNe of supermassive stars beyond 300 M_⊙. We also confirm that SNe of the supermassive stars (>300 M_⊙) do not change N/O ratios in comparison with the core-collapse SNe of typical massive stars (Iwamoto et al. 1999; Ohkubo et al. 2006), strengthening the reliability of the supermassive star (>300 M_⊙) scenario.

In addition to the two EMPGs (HSC J1631+4426 and J0811+4730) with ∼2% solar metallicity, a metal-poor galaxy, SBS 0335−052, shows 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ = 7.31 ± 0.01 (i.e., 4.2% solar metallicity) and $\mathrm{log}(\mathrm{Fe}/{\rm{O}})=-{1.42}_{-0.28}^{+0.06}$, which is 59% of the solar Fe/O ratio (Izotov et al. 2006a). In Figure 4, SBS 0335−052 also deviates from the observational results of Galactic stars and the Fe/O evolution model, which supports the results of this paper. Note that Izotov & Thuan (1999) suggest that the Fe/O value of SBS 0335−052 can be overestimated by the C iv λ4659 contamination on the [Fe iii] λ4658 line, based on the low-resolution spectroscopy of Multiple Mirror Telescope (MMT) spectrophotometry (R ∼ 700). However, conducting the VLT/GIRAFFE spectroscopy with high spectral resolutions (R ∼ 10,000), Izotov et al. (2006a) find that the [Fe iii] λ4658 line is very narrow (∼1 Å, i.e., ∼60 km s⁻¹), which is not explained by the C iv λ4659 line that originated from stellar winds. Nondetection of the C iv λ5808 line also suggests that the C iv λ4659 line is not strong enough to contaminate the [Fe iii] λ4658 line significantly. Thus, SBS 0335−052 (Izotov et al. 2006a) is another example of a metal-poor galaxy that significantly shows a higher Fe/O ratio than the chemical evolution models. Izotov et al. (2006a) attribute the higher Fe/O ratio of SBS 0335−052 to low dust depletion of iron, which has been ruled out in this paper (i.e., the first scenario in this section).

In summary of this subsection, we have discussed the three scenarios that might be able to explain the high Fe/O ratios of the two EMPGs (HSC J1631+4426 and J0811+4730). We suggest that the high Fe/O ratios of the two EMPGs are attributed to the contribution from core-collapse SNe of supermassive stars beyond 300 M_⊙. The contribution of supermassive stars beyond 300 M_⊙ to the iron enhancement has never been discussed by previous studies, including Izotov et al. (2006b) and Izotov et al. (2018). Many previous studies (e.g., Fragos et al. 2013a, 2013b; Stanway et al. 2016; Suzuki & Maeda 2018; Xiao et al. 2018) assume the IMF maximum stellar mass (M_max) at M_max = 100, 120, or 300 M_⊙, ignoring supermassive stars beyond 300 M_⊙, so this paper sheds light on the supermassive stars beyond 300 M_⊙ in metal-poor galaxies undergoing the early-phase galaxy formation.

We investigate ionizing radiation of our metal-poor galaxy sample by comparing emission-line ratios of various ions. Figure 5 shows four emission-line ratios of [O ii] λ λ3727, 3729/Hβ, [Ar iii] λ4740/Hβ, [O iii] λ5007/Hβ, and [Ar iv] λ7136/Hβ as a function of metallicity. Among many emission lines detected in our spectroscopy, we choose the [O ii] λ λ3727, 3729, [Ar iii] λ4740, [O iii] λ5007, and [Ar iv] λ7136 emission lines for two reasons below. The first reason is that oxygen and argon are both α-elements, and thus the Ar/O abundance ratio is almost constant as we confirm in Section 5.1. Thus, emission-line ratios are simply interpreted by ionizing radiation intensity and/or hardness, free from variance of element abundance ratio. The second reason is that the four lines are sensitive to ionization photons in a wide energy range from 13.6 to 40.7 eV. The [O ii] λ λ3727, 3729, [Ar iii] λ4740, [O iii] λ5007, and [Ar iv] λ7136 lines are emitted via spontaneous emission after collisional excitation of O⁺, Ar²⁺, O²⁺, and Ar³⁺, respectively. Table 6 summarizes these emission-line processes and the corresponding photon energy required to emit these lines.

Zoom In Zoom Out Reset image size

In panels (a)–(d) of Figure 5, we show local, average SFGs of Andrews & Martini (2013, AM13 hereafter) with black circles. We regard the AM13 SFGs as local averages because the AM13 sample is obtained by the SDSS composite spectra in bins of wide SFR and stellar mass ranges. In panels (a)–(d), the AM13 SFGs form sequences as a function of metallicity. The sequences of [O ii] λ λ3727, 3729/Hβ and [Ar iii] λ4740/Hβ show peaks at around 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 8.7 and 8.3, respectively. The [O iii] λ5007/Hβ and [Ar iv] λ7136/Hβ ratios may also have peaks around 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 8.0 and 12 + $\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 7.2–7.7 by interpolating AM13 SFGs and our metal-poor galaxies. Recalling that the [O ii] λ λ3727, 3729, [O iii] λ5007, [Ar iii] λ4740, and [Ar iv] λ7136 lines are sensitive to ionizing photon above 13.6, 35.1, 27.6, and 40.7 eV, respectively, we find that the peak metallicities decrease with increasing ionizing potentials of the corresponding emission lines. The peak transition demonstrates that ISM is irradiated by more intense or harder ionizing radiation in lower metallicity, as suggested by previous studies (e.g., Nakajima & Ouchi 2014; Nakajima et al. 2016; Steidel et al. 2016; Kojima et al. 2017). We also find that our metal-poor galaxies fall on the sequences of AM13 SFGs within a scatter. Thus, we infer that our metal-poor galaxies and the AM13 SFGs have a similar spectral shape in the energy range of 13.6–40.7 eV for a given metallicity.

Figure 6 shows He ii λ4686/Hβ ratios of our metal-poor galaxies as a function of metallicity, as well as the AM13 SFGs (black circles). The AM13 SFGs show almost constant He ii λ4686/Hβ ratios around log(He ii λ4686/Hβ) ∼ −2.0 in the range of 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ = 8.1–8.6, while the He ii λ4686/Hβ ratios increase with decreasing metallicity below 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ =8.1. Our metal-poor galaxies show a wide range of He ii λ4686/Hβ ratios between log(He ii λ4686/Hβ) ∼ −1.6 and −2.6. The distributions of our metal-poor galaxies are similar to those of SFGs of Schaerer et al. (2019, S19 hereafter). Among our metal-poor galaxies, three EMPGs (encircled by red and orange circles) show highest He ii λ4686/Hβ ratios around log(He ii λ4686/Hβ) ∼ −1.6, including the two representative EMPGs, HSC J1631+4426 and J0811+4730 (Izotov et al. 2018).

Zoom In Zoom Out Reset image size

As described in Section 1, previous studies also find large He ii λ4686/Hβ ratios of log(He ii λ4686/Hβ) > −1.8. In a study of W-R galaxies, López-Sánchez & Esteban (2010) find four galaxies that show large He ii λ4686/Hβ ratios of log(He ii λ4686/Hβ) = −1.7 to −1.5 with no W-R features in the metallicity range of 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 7.6–8.1. Shirazi & Brinchmann (2012) construct an SDSS galaxy sample with an He ii λ4686 detection. Shirazi & Brinchmann (2012) find 68 galaxies that have log(He ii λ4686/Hβ) = −2.4 to −1.8 with no W-R features in the metallicity range of 12 + $\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼7.7–8.2. Senchyna et al. (2017) conduct HST/COS spectroscopy for five SDSS galaxies with a nebular He ii λ4686 detection and no W-R features. The five galaxies show log(He ii λ4686/Hβ) = −2.1 to −1.4 in the metallicity range of 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 7.8–8.0. Our EMPGs, HSC J1631+4426 and J0811+4730, have 12+$\mathrm{log}({\rm{O}}/{\rm{H}})$ ∼ 6.90 and 6.98, respectively, which are much lower than those of these samples. The He ii λ4686/Hβ ratios of HSC J1631+4426 and J0811+4730 are log(He ii λ4686/Hβ) ∼ −1.6, which is comparable to those of the three samples of López-Sánchez & Esteban (2010), Shirazi & Brinchmann (2012), and Senchyna et al. (2017).

As described in Section 1, the physical mechanism of the He ii λ4686 emission from SFGs is still under debate. One of the possible sources of He⁺ ionizing photons is the very hot star produced via binary evolution. Xiao et al. (2018) have created nebular emission models with the combination of the photoionization code cloudy (Ferland et al. 2013) and the bpass code (Stanway et al. 2016; Eldridge et al. 2017). The bpass code calculates the stellar binary evolution, including atmosphere stripping, stellar rotation, and stellar mergers. The binary stellar evolution models of Xiao et al. (2018) predict low values of He ii λ4686/Hβ ≲ 1/1000, which are well below the observed He ii λ4686/Hβ ratios of our galaxies and S19 galaxies (He ii λ4686/Hβ ∼ 1/300–1/30). This result suggests that the main contributors of He ii λ4686 are not hot stars produced via binary evolution themselves.

Another possible explanation is an HMXB, where X-ray is emitted from a binary system of a compact object and a star through gas accretion. S19 aim to explain the He ii λ4686 emission from local SFGs with HMXB models of Fragos et al. (2013a, 2013b). HMXBs are binary systems consisting of a compact object (such as BH) and a companion star. The companion star provides gas onto the compact object and creates a hot accretion disk around the compact object. The hot accretion disk radiates very hard, power-law radiation ranging from UV to X-ray. Fragos et al. (2013a, 2013b) carefully calculate the HMXB evolution along the star formation history and predict total X-ray luminosities (L_X) from a galaxy as functions of metallicity and age. S19 convert an L_X/SFR ratio to the He ii λ4686/Hβ ratio, under the simple assumptions of He ii λ4686/Hβ = 1.74×Q(He⁺)/Q(H) (Case B recombination of 20,000 K; Stasińska et al. 2015), Q(H)/SFR =9.26 × 10⁵² photons s⁻¹/(M_⊙ yr⁻¹) (Kennicutt 1998), and hardness of Q(He⁺)/L_X = 2 × 10¹⁰ photon erg⁻¹. Here, the Q(He⁺) and Q(H) are defined by ionizing photon production rates above 54.4 and 13.6 eV, respectively. S19 also use the bpass binary stellar synthesis models of Xiao et al. (2018) to associate stellar ages with EW₀(Hβ).

Figure 7 compares He ii λ4686/Hβ ratios of our metal-poor galaxies and those obtained by the S19 HMXB models (solid lines) as a function of EW₀(Hβ). The solid lines trace the time evolution of He ii λ4686/Hβ and EW₀(Hβ) with different metallicities of 0.05, 0.10, 0.20, and 0.50 Z_⊙. The EW₀(Hβ) decreases and the He ii λ4686/Hβ ratio increases as time passes owing to the stellar evolution and HMXB evolution, respectively. The HMXB models (especially 0.05 and 0.10 Z_⊙) show a rapid increase of He ii λ4686/Hβ around EW₀(Hβ) ∼ 100–300 Å. The EW₀(Hβ) ∼ 100–300 Å corresponds to ∼5 Myr in the bpass binary stellar synthesis models. The rapid increase is triggered by the first compact object formation (i.e., the first HMXB formation) after ∼5 Myr of the starburst. As shown in Figure 7, the HMXB models of S19 have quantitatively explained the He ii λ4686/Hβ ratios of half of our metal-poor galaxies. However, we find that the other five metal-poor galaxies are not explained by the HMXB model, which fall in the ranges of EW₀(Hβ) > 100 Å and log(He ii λ4686/Hβ) > −2.0. Interestingly, three out of the five metal-poor galaxies are EMPGs (i.e., Z < 0.1 Z_⊙), which are marked with red and orange circles in Figure 7. Especially, HSC J1631+4426 (0.016 Z_⊙) and J0811+4730 (0.019 Z_⊙) are the representative EMPGs with the two lowest metallicities reported to date, showing high He ii λ4686/Hβ ratios of log(He ii λ4686/Hβ) ∼ −1.6. Furthermore, the S19 SFG sample also includes galaxies in the same ranges of EW₀(Hβ) >100 Åand log(He ii λ4686/Hβ) > −2.0. S19 have argued that other X-ray sources are likely to appear fairly soon after the onset of the star formation (≲5 Myr) in galaxies with high values of EW₀(Hβ) and He ii λ4686/Hβ. Although W-R stars might contribute to the strong He ii λ4686 emission, we do not find broad He ii λ4686 emission lines typical of the W-R stars (Brinchmann et al. 2008; López-Sánchez & Esteban 2009) in our spectra. Instead, S19 suggest that an underlying older population or shocks could also contribute to the high He ii λ4686/Hβ ratios. S19 do not discuss the metallicity in the explanation of the large He ii λ4686/Hβ ratios at EW₀(Hβ) > 100 Å. However, the two EMPGs (HSC J1631+4426 and J0811+4730) in this paper emphasize that both the EW₀(Hβ) and metallicity may be keys to explaining the He ii λ4686/Hβ ratios being higher than the HMXB models.

Zoom In Zoom Out Reset image size

In addition to the S19 suggestions, we propose two other possibilities, for the first time, which can explain the high He ii λ4686/Hβ ratios seen in the range of EW₀(Hβ) > 100 Å. First, we propose the possibility of supermassive stars beyond 300 M_⊙. The HMXB models of Fragos et al. (2013a, 2013b) assume the Kroupa IMF (Kroupa 2001; Kroupa & Weidner 2003) with a maximum stellar mass of M_max = 120 M_⊙. Thus, in the HMXB models of Fragos et al. (2013a, 2013b), the first HMXBs emerge ∼5 Myr after the start of the star formation, which corresponds to a lifetime of a star with 120 M_⊙. On the other hand, stars more massive than 120 M_⊙ are expected to have a shorter lifetime than stars with 120 M_⊙. According to the theoretical study of Yungelson et al. (2008), supermassive stars with 300 and 1000 M_⊙ die 2.5 and 2.0 Myr after the onset of the star formation, respectively. As described in Section 5.1.3, stars between 140 and 300 M_⊙ undergo thermonuclear explosions triggered by PISNe (Barkat et al. 1967) and do not leave any compact object (e.g., Heger & Woosley 2002). On the other hand, stars beyond 300 M_⊙ experience core-collapse SNe and form IMBHs (e.g., Ohkubo et al. 2006). Ohkubo et al. (2006) estimate that BH masses become ∼230 and ∼500 M_⊙ for stars with initial masses of 500 and 1000 M_⊙, respectively. Thus, when we assume supermassive stars beyond 300 M_⊙, IMBHs appear as early as ∼2 Myr, and part of the IMBHs may form HMXBs. Accretion disks of IMBHs emit very hard radiation, including ionizing photons above 54.4 eV, which boosts the He ii λ4686 intensity. A galaxy as young as ∼2 Myr has EW₀(Hβ) ∼ 300–400 Å according to the bpass models. Under the assumption of supermassive stars beyond 300 M_⊙, the He ii λ4686/Hβ ratio is expected to start increasing at around EW₀(Hβ) ∼300–400 Å. Such a model may cover the regions of EW₀(Hβ) > 100 Å and log(He ii λ4686/Hβ) > −2.0 shown in Figure 7. Thus, we suggest that supermassive stars beyond 300 M_⊙ would be able to explain the high ratios, log(He ii λ4686/Hβ) > −2.0 in the galaxies with EW₀(Hβ) > 100 Å. Note again that galaxies with log(He ii λ4686/Hβ) > −2.0 and EW₀(Hβ) > 100 Å include HSC J1631+4426, SDSS J2115−1734, and J0811+4730. These EMPGs might form supermassive stars beyond 300 M_⊙ from their extremely metal-poor gas. However, our explanation and the interpretation of S19 are based on some simple assumptions that associate the HMXB models and the bpass stellar synthesis models. We propose to construct self-consistent SED models ranging from X-ray to UV with the HMXB evolution models under the assumption of M_max > 300 M_⊙.

Second, we also suggest the possibility of a metal-poor AGN, which can contribute to boost the He ii λ4686 intensity of the very young galaxies. In Paper I, we have confirmed that all of our metal-poor galaxies fall on the SFG region of the BPT diagram defined by the maximum photoionization models with stellar radiation (Kewley et al. 2001). However, Kewley et al. (2013) suggest that emission-line ratios calculated under the assumption of a metal-poor AGN also fall on the SFG region. Thus, we cannot exclude the possibility of a metal-poor AGN. Groves et al. (2004a, 2004b) have constructed the photoionization models under the assumption of AGN-like, power-law radiation. The models of Groves et al. (2004a, 2004b) predict very strong He ii λ4686 emission represented by log(He ii λ4686/Hβ) ∼ −1.5 to 0.0. On the other hand, photoionization models with stellar radiation (Xiao et al. 2018) predict log(He ii λ4686/Hβ) ≲ −2.5. To explain the observed ratios of log(He ii λ4686/Hβ) ∼ −2.0, the combination of AGN and stellar radiation is required. We have checked the archival data of ROSAT and XMM and found no detection in X-ray. This is because the data are ∼2 orders of magnitudes shallower than the expected X-ray luminosities (∼10⁻¹⁴ erg s⁻¹ cm⁻²) of our metal-poor galaxy sample, which are obtained under the assumption of the L_2keV–M_UV relation of AGNs (Lusso et al. 2010). Deep X-ray observations are required to constrain X-ray sources of metal-poor galaxies.