EDP Sciences
Free Access
Issue
A&A
Volume 555, July 2013
Article Number A90
Number of page(s) 14
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/201221010
Published online 05 July 2013

© ESO, 2013

1. Introduction

Magnesium is an α-process element produced during nuclear burning in massive stars, which is similar to oxygen and some other elements, such as noble neon and argon. Its abundance is studied in detail from absorption lines in stars with a wide range of metallicities and shows trends similar to oxygen.

The Mg ii λλ 2797, 2803 Å doublet (h and k) is one of the most studied tracers of the gas-phase medium in planetary nebulae (PNe), in the local interstellar medium (LISM) of the Galaxy, and in the gaseous environment of distant galaxies. Since magnesium, which is similar to silicon and iron among other elements, is a refractory element, its abundance can provide information about the level of interstellar magnesium depletion onto dust.

The magnesium abundance in the LISM can be determined by analysing absorption lines towards Galactic stars. With high-resolution spectra of many Galactic early-type stars observed from Copernicus launched in 1972 it was found that profiles of Mg ii h and k lines are identical with solar profiles except for the presence of narrow absorption components formed in the interstellar medium along the line of sight (Oegerle et al. 1982; Murray 1983). Follow-up observations of early type and cool stars with the International Ultraviolet Explorer (IUE) and the Hubble Space Telescope (HST) provided quantitative characteristics of the interstellar medium towards Galactic stars and planetary nebulae: the distribution of the magnesium abundance in the LISM (Molaro et al. 1986; Middlemass 1988), specifically; the structure of the LISM within 100 pc (Redfield & Linsky 2004); and the abundances and physical conditions in the diffuse interstellar clouds (Welty et al. 1999). These data indicate a moderate level of magnesium depletion.

There are only a few determinations of the magnesium abundance in H ii regions, such as the Orion nebula in the Galaxy and the Tarantula nebula in the Large Magellanic Cloud and in some bright planetary nebulae (Rodríguez & Rubin 2005; Peimbert & Peimbert 2010; Dinerstein et al. 2012). Dinerstein et al. (2012) has investigated the gas-phase magnesium abundances in 25 planetary nebulae using the Mg iiλ4481 Å recombination line. They find that Mg/H is close to solar, implying that Mg is at most minimally depleted, whereas the measurements in the 30 Doradus and the Orion nebulae, with the same recombination line, indicate significantly higher levels of Mg depletion up to 72% and 90%, respectively (Peimbert & Peimbert 2010).

The Mg ii λλ2707, 2803 Å doublet is the strongest absorption feature that is detectable in the optical range at intermediate redshifts (0.3 ≤ z ≤ 2). Therefore, magnesium absorption lines are often detected in spectra of background quasars and galaxies. Spectra of background sources allow us to probe regions of low-ionisation and cool gas present in the outer regions of spiral discs and dwarf irregular galaxies (Dessauges-Zavadsky et al. 2004). Many Mg ii surveys have been carried out for these systems (Sargent et al. 1988; Steidel & Sargent 1992; Dessauges-Zavadsky et al. 2003; York et al. 2006; Prochter et al. 2006; Prochaska et al. 2007b; Lilly et al. 2007; Quider et al. 2011). Owing to large spectroscopic surveys including the Sloan Digital Sky Survey (SDSS; York et al. 2000) hundreds of thousands of quasars have been used to identify large samples of absorption systems along the line of sight using the Mg ii absorption lines (York et al. 2006; Bouché et al. 2006; Prochter et al. 2006; Lundgren et al. 2009; Quider et al. 2011). Recently, based on a fully-automatic method, Zhu & Ménard (2013) compiled a very large sample of ~40 000 Mg ii absorbers from the SDSS Data Release 7 (DR7).

Using the Mg ii absorber samples, many detailed studies on the distribution of column densities, the redshift evolution of densities, and the kinematic signatures (e.g., outflowing material from star-forming regions) have been performed (Steidel & Sargent 1992; Nestor et al. 2005; Prochter et al. 2006). Based on 4000 Mg ii absorbers from zCOSMOS (Lilly et al. 2007), Bordoloi et al. (2011) find that at the same stellar mass, the strength of Mg ii absorption is much higher for blue star-forming galaxies than for red galaxies. Determinations of the Mg abundances from the absorption lines in damped Lyα (DLA) absorbing systems gave Mg/O abundance ratios that are close to the solar values. Meanwhile, the absorption line systems allow investigations of gas properties in outer parts of galaxies and in the intergalactic medium.

Element abundances of the interstellar medium are usually compared with the solar abundances, assuming that the interstellar medium has the same composition as the Sun, which is, however, not the case (Snow & Witt 1996). Instead, the depletion will be different if the reference standard is derived from different types of stars in the solar neighbourhood, or the solar neighbourhood abundances corrected for the chemical evolution of the Galaxy, since formation of the Sun, or taking the galactocentric gradient of the chemical elements into account (Snow & Witt 1996; Peimbert & Peimbert 2010). Thus, Snow & Witt (1996) found systematically lower depletions using different so-called “cosmic” abundances as reference. Different authors also use different solar abundances as reference abundance (e.g. Grevesse & Sauval 1998; Asplund et al. 2005, 2009) or solar system meteoritic values, and different metallicity indicators (O, S, Zn) in DLA, quasi-stellar object (QSO), and gamma-ray burst (GRB) absorption systems. Using such different reference abundances resulted in a Mg depletion that does not exceed ~0.2 dex (throughout the paper we use solar values by Asplund et al. 2009).

However, despite numerous magnesium abundance determinations, only upper limits of Mg ii abundances have been obtained in the overwhelming majority of the cases. There is also a problem with some O i absorption oxygen lines since those are generally saturated. Therefore there are not so many exact Mg/O abundance ratios derived in DLAs.

No such determinations exist for extragalactic H ii regions, excluding the Tarantula nebula. This is because of the weakness of the Mg iiλ4481 Å recombination line seen in emission, and its blending with other weak emission lines. The only possibility for studying the Mg abundance in extragalactic H ii regions is to use intensities of the resonance doublet Mg iiλ2797, λ2803 lines, which are much brighter than the recombination lines. However, these emission lines, at variance to forbidden lines, are subject to absorption by the weakly ionised interstellar medium. It is also possible that in some galaxies gas outflows can be responsible for an additional underlying contribution to the Mg ii lines producing broad features due to the Doppler effect. However, for our galaxy sample we cannot consider the effects of Mg gas outflows and radiation transfer in the medium surrounding the H ii regions due to the unknown gas velocity distribution and ionisation structure of our objects. Additionally, Mg iiλ2797, λ2803 emission and absorption can be produced by cool giant and supergiant stars, and luminous blue variable (LBV) stars.

Finally, the doublet in spectra of low-redshift galaxies can only be observed from space. However, in higher redshift galaxies, Mg iiλ2797, λ2803 lines can be measured in the optical range. In particular, these lines are seen in the SDSS spectra with a lower wavelength cut-off of ~3800 Å if the galaxy redshift z is greater than ~0.36, allowing the determination of its abundance.

The aim of this paper is to select SDSS spectra of emission-line star-forming galaxies with z ≳ 0.36 showing the Mg iiλ2797, λ2803 doublet in emission and to derive Mg abundances. Our sample of emission-line galaxies extracted from the SDSS is discussed in Sect. 2. The element abundances are derived in Sect. 3. We discuss our results and compare them with other types of objects in Sect. 4. Our main findings are summarised in Sect. 5.

Table 1

General characteristics of galaxies.

thumbnail Fig. 1

Redshift-corrected SDSS spectrum of J0207+0047. The spectral region λλ2750–2850, which includes the Mg iiλ2797, λ2803 emission lines, is shown in the inset. Vertical dotted lines in the inset indicate the nominal wavelengths of the lines.

Open with DEXTER

thumbnail Fig. 2

Redshift-corrected SDSS spectra in the wavelength range λλ2750–2850 Å of all selected galaxies excluding J0207+0047, which is shown in Fig. 1. Vertical dotted lines indicate the nominal wavelengths of the Mg iiλ2797, λ2803 emission lines. For three galaxies, J0812+3200, J1045+3225, and J1716+2744, two spectra are shown that are available in the SDSS database.

Open with DEXTER

2. Sample of SDSS galaxies

We use a sample that is composed of spectra of low-metallicity H ii regions with strong emission lines selected from the SDSS DR7. The SDSS (York et al. 2000) offers a gigantic database of galaxies with well-defined selection criteria that are observed in a homogeneous way. First, we extracted ~15 000 spectra with strong emission lines from the whole database of ~800 000 galaxy spectra. Out of this sample we selected 65 spectra of star-forming galaxies with strong nebular Hβ, Hα, [O ii]λ3727, [O iii]λ4959, λ5007 emission lines and redshifts z ≳ 0.36. The [O iii]λ4363 emission line is also present in many of these spectra (in ~63% of the total sample and ~70% of the galaxies in which Mg ii lines were detected), allowing an accurate determination of the oxygen abundance by the direct Te-method. For three galaxies, two spectra are available in the SDSS database. Therefore, the total number of selected galaxies is 62.

The general characteristics of the selected galaxies are shown in Table 1. The redshift range is 0.36–0.70. In general these galaxies are very faint, with the apparent SDSS g magnitude fainter than 19 mag. The redshift-corrected spectrum of one galaxy, J0207+0047, is shown in Fig. 1. This spectrum resembles the spectrum of high-excitation star-forming H ii region with strong emission lines, including [O iii]λ4363. The Mg iiλ2797, λ2803 doublet is also present in emission and it is shown in more detail in the inset. The remaining 64 redshift-corrected spectra in the wavelength range λλ2750–2850 Å are shown in Fig. 2. Despite the noisy spectra in some cases, the Mg ii emission is detected in about two thirds of the spectra. However, we also note that blue-shifted Mg ii absorption is present in many cases, which may be an indication of stellar winds from cool massive stars, such as LBVs. In some cases broad absorption features are observed at the position of Mg iiλ2797 Å and λ2803 Å emission lines. The most prominent feature of this kind is seen in the spectrum of J0815+3129. In some other cases absorption features are absent in the spectra. We take all these peculiarities into account and measure the emission line intensities by placing a continuum level at the bottom of the absorption profiles if they are present.

The extinction-corrected line fluxes I(λ), normalised to I(Hβ), are given in Table 2. The line fluxes were obtained using the IRAF1 SPLOT routine. The line flux errors include statistical errors in addition to errors introduced by the standard star absolute flux calibration, which we set to 1% of the line fluxes. These errors will be propagated later into the calculation of abundance errors. The line fluxes were corrected for two effects: (1) reddening using the extinction curve of Cardelli et al. (1989) and (2) underlying hydrogen stellar absorption derived simultaneously by an iterative procedure as described in Izotov et al. (1994). Since the redshifts of the selected galaxies are high, the correction for extinction was done in two steps. First, emission-line intensities with observed wavelengths were corrected for the Milky Way extinction, using values of the extinction A(V) in the V band from the NASA/IPAC extragalactic database (NED). Then, the internal extinction was derived from the Balmer hydrogen emission lines, corrected for the Milky Way extinction. The internal extinction was applied to correct line intensities at non-redshifted wavelengths. The extinction coefficients in both cases of the Milky Way and the internal extinction are defined as C(Hβ) = 1.47E(B − V), where E(B − V) = A(V)/3.2 (Aller 1984). The mean value of the internal extinction coefficient C(Hβ) ~ 0.2 is typical of star-forming galaxies.

Additionally, the Mg iiλ2797, λ2803 emission lines were corrected for the underlying stellar absorption. For this, we used the Bruzual & Charlot (2003) population synthesis models of single stellar populations. Spectra of these models do not have sufficient spectral resolution to separate Mg iiλ2797 and λ2803 absorption lines. Therefore, we measured the equivalent width of the blend. We find that in a wide range of starburst ages of 3–10 Myr, the equivalent width EWabs(Mg iiλ2797 + λ2803) is constant and is equal to ~–1 Å. The “–” sign means that the line is in absorption. Then, for separate Mg ii absorption lines we adopt equal EWabs’s of − 0.5 Å and correct Mg ii emission lines multiplying their intensities by a factor (EW+|EWabs|)/EW, where EW is the equivalent width of the emission line. Equivalent widths EW(Hβ), extinction coefficients C(Hβ)(MW) and C(Hβ)(int), and EWabs of the hydrogen absorption stellar lines are also given in Table 2, along with the uncorrected Hβ fluxes.

3. Element abundances

3.1. Oxygen abundance

To determine element abundances, we generally follow the procedures of Izotov et al. (1994, 1997) and Thuan et al. (1995). We adopt a two-zone photoionised H ii region model: a high-ionisation zone with temperature Te(O iii), where the [O iii] lines originate, and a low-ionisation zone with temperature Te(O ii), where the [O ii] lines originate. In the H ii regions with a detected [O iii] λ4363 emission line, the temperature Te(O iii) is calculated using the direct method based on the [O iii] λ4363/(λ4959+λ5007) line ratio. In H ii regions where the [O iii] λ4363 emission line is not detected, we used a semi-empirical method described by Izotov & Thuan (2007) to derive Te(O iii). For Te(O ii), we use the relation between the electron temperatures Te(O iii) and Te(O ii) obtained by Izotov et al. (2006) from the H ii photoionisation models (Stasińska & Izotov 2003).

Ionic and total oxygen abundances are derived using expressions for ionic abundances obtained by Izotov et al. (2006). For magnesium, we assume that the Mg ii emission has a nebular origin (see text below and Fig. 3). Then the magnesium abundance can be derived if the electron temperature Te(Mg ii) in the Mg+ zone and the ionisation correction factor ICF(Mg+) for unseen stages of ionisation are known. To derive Te(Mg ii) and ICF(Mg+) we use a grid of the photoionised H ii region models that also predict Mg iiλ2797, λ2803 emission-line intensities.

thumbnail Fig. 3

Dependence of I(Mg iiλ2797)/I(Hβ) on the excitation parameter x = O2+/O. Red open circles are data from CLOUDY models calculated by adopting 12 + log O/H = 8.0 and 8.3. Large black filled circles are extinction-corrected intensities for the H ii regions with 12 + log O/H ≥ 7.9 and small light blue filled circles with 12 + log O/H < 7.9.

Open with DEXTER

Table 3

Input parameters for the grid of the photoionised H ii region models.

3.2. Grid of photoionisation CLOUDY models

Using the version v10.00 of the CLOUDY code (Ferland et al. 1998) we calculated a grid of 432 spherical ionisation-bounded H ii region models with parameters shown in Table 3, which cover the entire range of parameters in real high-excitation, low-metallicity H ii regions. In particular, the range of oxygen abundances is 12 + log O/H ≈ 7.5–8.4 for our sample. The abundances of other heavy elements relative to oxygen are kept constant and correspond to the typical value obtained for low-metallicity emission-line galaxies (e.g. Izotov et al. 2006). We also include dust, scaling it according to the oxygen abundance. The characteristics adopted for the dust are those offered by CLOUDY as “Orion nebula dust”.

We adopt three values of the number of ionising photons Q and the shape for the ionising radiation spectrum corresponding to the Starburst99 model with the ages of 2.0, 3.5, and 4.0 Myr and different metallicities (Leitherer et al. 1999). Thus, for 12+log O/H = 7.3 and 7.6 we adopt Starburst99 models with the heavy element mass fraction Z = 0.001, for those with 12+log O/H = 8.0 models with Z = 0.004 and for those with 12+log O/H = 8.3 models with Z = 0.008. All Starburst99 models were calculated with the Hillier & Miller (1998) and Pauldrach et al. (2001) stellar atmosphere set and with the stellar tracks from Meynet et al. (1994). We also vary the log of volume – filling factor f between −0.5 and −2.0 to obtain CLOUDY models with different ionisation parameters.

Our range of the number density, Ne = 10–103 cm-3, which is kept constant along a given H ii region radius, covers the whole range expected for the extragalactic H ii regions. Additionally, we calculated a set of 144 H ii region models with parameters from Table 3 (excluding Ne) with a Gaussian density distribution as a function of radius r according to (1)where Ne(0) = 103 cm-3. Thus, the total number of the models, which we use for the subsequent analysis, is 576.

Since the CLOUDY code allows us to predict Mg ii line intensities, we compared these intensities with the observed ones. The result of the comparison is shown in Fig. 3 where we show the Mg iiλ2797 line intensity as a function of the excitation parameter x = O2+/O. The observed H ii regions with 12 + log O/H ≥ 7.9 are shown by large black filled circles and H ii regions with 12 + log O/H < 7.9 by small light blue filled circles. Thus, the range of oxygen abundances in most of our galaxies is 12 + log O/H ~ 7.7–8.4 (see Table 5). Therefore, we show in Fig. 3 only predicted Mg iiλ2797 line intensities in the models with 12 + log O/H = 8.0 and 8.3 (red open circles). Our comparison shows that in general extinction-corrected Mg iiλ2797 observed intensities are by a factor of ~1.3 lower than the predicted ones even for the H ii regions with 12 + log O/H between 7.9 and 8.4, but they follow a trend with x which is similar to the trend for predicted intensities. This implies that Mg iiλ2797 emission is most likely nebular in origin, and the reduced line intensities are due to the combined effect of interstellar absorption and Mg depletion onto dust.

3.3. Magnesium abundance

The electron temperature Te(Mg ii) in the Mg+ zone and Te(O ii) in the O+ zone are obtained from the CLOUDY photoionised H ii region models. In Fig. 4a we show the relation between the electron temperatures in the O+ and Mg+ zones by different symbols for different starburst ages. As expected, these temperatures are very similar. The dispersion of the data at high temperatures Te(O ii) ≥ 15 000 K is mainly due to the different starburst ages. However, at lower temperatures, which is the case for our galaxies, the dispersion is small and there is no evident separation between models with different starburst ages. We fit the relation for models with all ages by an expression (2)which is shown in Fig. 4a by solid line. We will use this fit in our subsequent analysis. However, we note that the fit in Eq. (2) is only applicable to the range Te(O ii) ~ 9000–17 000 K.

Table 4

Coefficients for the ICF(Mg+)a fit in Eq. (5).

The Mg+ abundance is derived from the equation (3)where we use collisional strengths for the Mg iiλ2797 and λ2803 transitions from Mendoza (1983). These collisional strengths are in very good agreement with those obtained more recently by Sigut & Pradhan (1995).

thumbnail Fig. 4

a) Relation between the electron temperatures te(Mg ii) and te(O ii). te are in units of 10-4   Te. Red filled circles, blue open circles, and green stars are data from CLOUDY models with starburst ages of 2.0 Myr, 3.5 Myr, and 4.0 Myr, respectively. The solid line is the fit to data with all ages, defined by Eq. (2). The dotted line is the line of equal temperatures. The range of the electron temperature te(O ii) in the sample galaxies is shown by a horizontal line. The error bar is the dispersion of te(Mg ii) in the range of te(O ii) shown by the vertical line. b) Relation between the ionisation correction factor ICF(Mg+) and the excitation parameter x = O2+/O. Red filled circles, blue open circles, and green stars are data from CLOUDY models with starburst ages of 2.0 Myr, 3.5 Myr, and 4.0 Myr, respectively. The solid line is the fit to data with all ages as defined by Eq. (5) with the coefficients from Table 3. The range of the excitation parameter x in the sample galaxies is shown by a horizontal line. The error bar is the dispersion of ICF(Mg+) (in dex) in the range of x shown by the vertical line.

Open with DEXTER

thumbnail Fig. 5

Dependence of the magnesium-to-oxygen abundance ratio relative to the solar value, [Mg/O] = log (Mg/O) – log (Mg/O), on oxygen abundance 12+log O/H with error bars for all data points. The solar value log (Mg/O) = −1.09 is adopted from Asplund et al. (2009). Dashed line and ⟨[Mg/O]⟩ in both panels are mean values of [Mg/O]. a) The encircled galaxies are those with no clear presence of blue-shifted absorption profiles (Fig. 2). b) The same sample as in a) but with division between bright Mg ii lines with I(λ2797 Å)/I(Hβ) ≥ 0.3 (red filled circles) and weak Mg ii lines with I(λ2797 Å)/I(Hβ) < 0.3 (blue filled circles). Encircled galaxies in b) are those where [O iii]λ4363 Å is measured. Additionally, the mean values of [Mg/O] for bright (⟨[Mg/O]⟩(bright)) and weak (⟨[Mg/O]⟩(weak)) Mg ii lines are denoted.

Open with DEXTER

The total Mg abundance is obtained from the relation (4)where the ionisation correction factor ICF(Mg+) takes all unseen stages of magnesium ionisation inside the H ii region into account. To derive ICF(Mg+) we use our grid CLOUDY models and define it as the ratio x(H+)/x(Mg+), where x(H+) = H+/H and x(Mg+) = Mg+/Mg are volume-averaged fractions of H+ and Mg+, respectively.

In Fig. 4b we show the dependence of ICF(Mg+) on the excitation parameter O2+/O. Ionisation potentials of Mg of the three ionisation stages Mg, Mg+, and Mg++ are 7.6, 15.0 and 80.1 eV, respectively. Therefore, most of magnesium in a photoionised region is in the Mg+ and Mg++ stages. However, because Mg+ is not the dominant ionic species for most nebulae and because Mg++ is not observed in the visible range, a rather large ICF(Mg+) is needed. Models with different starburst ages in Fig. 4b are shown by different symbols. There is a clear offset between the models with younger and older starburst ages, indicating differences in the spectral energy distribution of the ionising radiation. The fit to the data with all ages defined by equation (5)where coefficients ai, bi, and ci are given in Table 4. The fit is produced for a wide range of x = 0.2–1.0 where ICF increases from ~3 to more than 100. However, the range of x in the sample galaxies (horizontal line) is smaller, corresponding to ICFs in the range ~3.5–~20 with an average value of ~6. The fit closely follows the models with starburst age of 3.5 Myr corresponding to the equivalent width EW(Hβ) ~ 100 Å that is typical for our galaxies.

Table 5

Spectroscopic parameters.

4. Results and discussion

In Figs. 5a and b the dependences of [Mg/O] = log (Mg/O) – log (Mg/O) on 12+log O/H are shown for the 45 spectra where the Mg ii emission lines was detected. We note also that in many cases blue-shifted Mg ii absorption near the Mg iiλ2797 Å emission line is present which may indicate of stellar winds from cool massive stars, such as LBVs (see, for instance, J0207+0047, J0925+2709, J1045+3225 in Fig. 2). The encircled galaxies are those with no clear presence of blue-shifted absorption profiles (Fig. 5a). There is no obvious offset of the encircled points as compared to other data. This testifies to the correctness of Mg ii λ2797 and λ2803 line measurements in cases when the continuum is placed at the bottom of absorption profiles. In Fig. 5b we divide the sample between objects with bright Mg ii lines (I(λ2797 Å)/I(Hβ) ≥ 0.3) and weak Mg ii lines (I(λ2797 Å)/I(Hβ) < 0.3). Encircled galaxies in (b) are those where [O iii]λ4363 Å emission is detected. A slight shift of H ii regions with bright Mg ii emission line λ2797 to lower metallicity is seen in Fig. 5b.

Table 5 summarises spectroscopic parameters for all 65 spectra, where we show the galaxy name, the oxygen abundance 12+log O/H, the quantity [Mg/O], and the equivalent width EW(Hβ) of the Hβ emission line in Å. We adopt log (Mg/O) = −1.09 (Asplund et al. 2009). We note that both O and Mg are subject to depletion onto dust. However, oxygen is by a factor of ~10 more abundant than magnesium. Therefore, the quantity [Mg/O] is mainly a characteristic of the magnesium depletion.

The dependences of [Mg/O] on EW(Hβ) are shown in Fig. 6. Three H ii regions from our sample have EW(Hβ) > 250 Å and are outside of the figure. Their [Mg/O] values range between –0.02 and –0.24 (see Table 5), similar to the range of [Mg/O] for the entire sample. The galaxies with measured [O iii]λ4363 Å emission line are encircled. There is no trend in [Mg/O] with EW(Hβ) in Fig. 6. The EW(Hβ) is a characteristic of the starburst age, hence of the excitation parameter x. Therefore, the absence of any correlation in Fig. 6 indicates the correctness of the correction for unseen stages of Mg ionisation.

thumbnail Fig. 6

Dependence of the magnesium-to-oxygen abundance ratio relative to the solar value [Mg/O] on the equivalent width EW(Hβ) of the Hβ emission line. The galaxies with the measured [O iii]λ4363 Å line are encircled.

Open with DEXTER

To additionally check the accuracy of our measurements and our magnesium and oxygen abundance determinations, we show in Fig. 7 the dependence of the magnesium-to-oxygen abundance ratio relative to solar value [Mg/O] (Fig. 7a) and I(Mg iiλ2797)/I(Hβ) (Fig. 7b) on the observed fluxes of Hβ emission lines corrected for both the Milky Way and internal extinctions for all sample galaxies. The galaxies with measured [O iii]λ4363 Å emission line are encircled. No differences in the distributions of the galaxies with measured [O iii]λ4363 Å line and without this line is found in Figs. 5b, 6, and 7a,b. Thus, in the following discussion we use all the data together obtained with the direct and semi-empirical methods. There is also no clear trend in [Mg/O] and I(Mg iiλ2797)/I(Hβ) with corrected fluxes of Hβ. This is also evidence for the accuracy of our determinations.

thumbnail Fig. 7

Dependence of the magnesium-to-oxygen abundance ratio relative to the solar value [Mg/O] a) and the I(Mg iiλ2797)/I(Hβ) ratio b) on observed fluxes of the Hβ emission line corrected for both the Milky Way and internal extinctions. a) All sample galaxies are shown. b) The same as in a) but H ii regions with [Mg/O] ≥ –0.3 being shown in red and those with [Mg/O] < –0.3 in black. The galaxies with measured [O iii]λ4363 Å line are encircled.

Open with DEXTER

We do not find any obvious trend in [Mg/O] with oxygen abundance in the entire range of 12+log O/H = 7.52–8.44 (Fig. 5). The mean value of the Mg/O ratio is by a factor of ~2 lower than the solar value, implying moderate Mg depletion of ~50% in the dust phase. This depletion is significantly higher than the mean value obtained by Dinerstein et al. (2012) for PNe. They derived Mg/O close to the solar value using the recombination Mg iiλ4481 emission line. These differences can be explained by the known discrepancy between values derived from recombination lines (RL) and collisionally excited lines (CEL; Esteban et al. 2009; Peimbert et al. 2005; Davey et al. 2000; Guseva et al. 2011). Abundances obtained from RLs tend to be higher than those derived from CEL lines. The origin of the nebular abundance discrepancy problem is currently not known. On the other hand, Barlow et al. (2003) and Wang & Liu (2007) do not empirically find any such discrepancy for Mg/H. Moreover, the Mg fraction in dust also obtained using the recombination Mg iiλ4481 emission line is much higher in 30 Doradus (72%) and in the Orion nebula (91%; Peimbert & Peimbert 2010) than in the PNe analysed by Dinerstein et al. (2012).

Table 6

Magnesium abundances collected from the literature.

The negative values of [Mg/O] in our galaxies can be due to the absorption of the Mg iiλ2797, λ2803 emission by the interstellar gas outside the H ii regions. The spectral resolution of the SDSS spectra is insufficient to separate the Mg ii absorption and emission to estimate the effect of the interstellar absorption and to correct emission-line intensities for this effect. An additional source of uncertainties may arise due to the presence of broad blue-shifted absorption lines in many spectra of our sample superposed with the emission lines (Fig. 2). One of the most evident cases is the galaxy J1045+3225. These lines are likely to be broad lines with P Cygni profiles produced by cool massive stars (red supergiants, LBVs) with a stellar wind (e.g. Hillier et al. 2001). The presence of these lines will introduce uncertainties into the placement of the continuum for measurements of the intensities of nebular emission lines. We measure magnesium emission lines placing the continuum at the bottom of the absorption lines when it is obviously seen in the observed spectra as described above. In spectra of some other galaxies there is no clear evidence of the blue-shifted broad absorption (see Fig. 2). To verify the effect of this absorption on the derived Mg abundance, we divided the objects in Fig. 5a into two samples, those with and those without evident blue-shifted broad absorption. It is very important that the encircled galaxies without blue-shifted absorption have the same [Mg/O] and dispersion as other galaxies.

thumbnail Fig. 8

a) Comparison of magnesium depletion for data available from the literature (denoted by different colours and symbols) and b) comparison of the magnesium depletion averaged for objects of each of four types LISM (local interstellar medium), H ii regions, our Mg ii sample (this paper), and DLA-QSO-GRB (damped Lyα absorption – quasi-stellar object – gamma-ray burst) absorption systems. [α/H] is [O/H] for LISM, H ii regions, and our sample galaxies, and [S/H] for DLA-QSA-GRB absorption systems.

Open with DEXTER

Summarising, we find that the Mg depletion in our galaxies is probably present but at a relatively low level with roughly half of the Mg in dust. This level of Mg depletion onto dust can be considered as an upper limit because the interstellar absorption is not known and is difficult to be taken it into account. For comparison, Jenkins (2009) found the relative proportions of different elements that are incorporated into dust at different stages of grain growth based on gas-phase element abundances for 17 different elements over 243 sight lines in the local part of our Galaxy. According to them ~70% (::tag lxir empty(lxir-formule)id=5505 content=!/sim50% on Fig. 5 by Jenkins 2009) of magnesium is locked onto dust grains and ~95% (::tag lxir empty(lxir-formule)id=5513 content=!/sim90% on Fig. 7 by Jenkins 2009) of iron is also in solid form (see also Fig. 1 by Jenkins 2009).

4.1. Comparison of magnesium depletion in objects of different types

We compared magnesium depletion obtained for our low-metallicity, emission-line star-forming galaxies with available data from the literature for objects of different types. There are many DLA and QSO absorption systems with magnesium abundance determination from Mg ii λ2797,λ2803 lines. However, only an upper limit to the Mgii abundance has been obtained in the majority of the cases. We include in Table 6 only those objects for which the exact values of number densities for hydrogen, magnesium, and oxygen (or sulphur) have been collected. We recalculated all data with solar abundances by Asplund et al. (2009) where it was necessary. To be more specific, we display in Fig. 8a all data from Table 6, indicating the different type of objects by different colours and symbols. The value of [α/H] is in terms of [S/H] for DLA-QSA-GRB absorption systems and [O/H] for LISM, H ii regions, and our sample galaxies.

Following Lebouteiller et al. (2008) we adopted the oxygen abundance of 30 Dor in the LMC to be a factor of ~0.6 lower than the solar value. The mean oxygen abundance of our Mg sample is 12 + log O/H = 8.02 for 45 H ii regions. A solar abundance of 12 + log O/H = 8.69 (Asplund et al. 2009) was adopted for the LISM (clouds toward Galactic stars and PNe). Following Peimbert & Peimbert (2010) an oxygen abundance of 12 + log O/H = 8.65 was adopted for the Orion nebula (determination based on the observations by Esteban et al. 2004). In all comparisons for DLA absorption systems we used the non-refractory element sulphur for metallicity estimation and for the determination of magnesium depletion because oxygen absorption lines are generally saturated (e.g. Prochaska et al. 2007b).

In Fig. 8b we compare [Mg/O] and [α/H] averaged for each of four types of objects (LISM, H ii regions, our sample, and DLA-QSO-GRB absorption systems). The averaged values of [Mg/O] and [α/H] are obtained from the logarithms of the average ⟨Mg/O⟩ and ⟨α/H⟩.

This is to be compared to iron depletion. Rodríguez & Rubin (2005) discovered the trend of increasing Fe depletion at higher oxygen abundance. For the Galactic H ii regions and PNe they obtained high Fe depletion, with fewer than 5% of their Fe atoms in the gas phase, whereas the metal-deficient blue compact galaxy SBS 0335–052 could have only from 13% to 40% of Fe in the gas phase. Izotov et al. (2006) also obtained the trend in Fe depletion with metallicity. Considering the iron depletion in low-metallicity, star-forming emission-line galaxies they find that ~80% of iron is confined in dust in galaxies with 12+log O/H ~ 8.0 (their Fig. 11l).

Despite a wide spread of points for the individual objects (Fig. 8a), there is also a clear trend in averaged [Mg/O] with metallicity in Fig. 8b for different types of objects, indicating that the depletion is higher for higher metallicity objects. Thus, we confirm the previous findings by Jenkins (2009) of increasing Mg depletion with increasing metallicity.

5. Conclusions

We have presented 65 SDSS spectra of low-metallicity, emission-line star-forming galaxies with redshifts z ~ 0.36−0.70, with the aim of studying the interstellar magnesium abundance as derived from the resonance emission-line doublet Mg iiλ2797, λ2803. This emission is detected in 45 galaxies, or in more than two thirds of the sample. Our main results are as follows.

  • 1.

    Using the grid of 576 CLOUDY photoionisedH ii region models, we obtained fits of the electrontemperature Te(Mg ii) in the Mg+ zoneas a function of the electron temperature Te(O ii) inthe O+ zone. Furthermore, using the same CLOUDY models weobtained fits of the ionisation correction factor ICF(Mg+) as a functionof the excitation parameter x = O2+/O.

  • 2.

    We derived oxygen and magnesium abundances. The [O iii] λ4363 emission line is seen in ~63% of the sources of the total sample and ~70% of the galaxies in which the Mg ii lines were detected, allowing an accurate abundance determination with the direct Te method. The element abundances in the remaining galaxies were derived using a semi-empirical method that is based on strong nebular oxygen lines. Data obtained with the two methods are indistinguishable in all distributions. The oxygen abundances 12+log O/H for the sample galaxies are in the relatively wide range from 7.52 to 8.44.

  • 3.

    We found that the Mg/O abundance ratio in our galaxies is lower by a factor of two than in the Sun, implying a moderate Mg depletion onto dust with ~50% of magnesium confined in dust. This level of depletion is higher than in high-redshift, low-metallicity DLA systems with measured Mg abundances. However, it is much lower than in H ii regions with higher metallicity (e.g., in the Orion nebula and 30 Doradus, where the fraction of Mg in dust is 91% and 72%, respectively), in PNe (averaged value) and in interstellar clouds of the LISM.

  • 4.

    Despite a wide spread of points for the individual determinations there is a clear trend in [Mg/O] or [Mg/S] with metallicity when averaged magnesium abundance and metallicity for different types of objects is used. This implies that the magnesium depletion is higher in higher metallicity objects.


1

IRAF is the Image Reduction and Analysis Facility distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under cooperative agreement with the National Science Foundation (NSF).

Acknowledgments

N.G.G. and Y.I.I. acknowledge the hospitality of the Max-Planck Institute for Radioastronomy, Bonn, Germany. This research made use of the NASA/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. Funding for the Sloan Digital Sky Survey (SDSS), and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, and the Max Planck Society, and the Higher Education Funding Council for England.

References

Online material

Table 2

Extinction-corrected emission-line intensities.

All Tables

Table 1

General characteristics of galaxies.

Table 3

Input parameters for the grid of the photoionised H ii region models.

Table 4

Coefficients for the ICF(Mg+)a fit in Eq. (5).

Table 5

Spectroscopic parameters.

Table 6

Magnesium abundances collected from the literature.

Table 2

Extinction-corrected emission-line intensities.

All Figures

thumbnail Fig. 1

Redshift-corrected SDSS spectrum of J0207+0047. The spectral region λλ2750–2850, which includes the Mg iiλ2797, λ2803 emission lines, is shown in the inset. Vertical dotted lines in the inset indicate the nominal wavelengths of the lines.

Open with DEXTER
In the text
thumbnail Fig. 2

Redshift-corrected SDSS spectra in the wavelength range λλ2750–2850 Å of all selected galaxies excluding J0207+0047, which is shown in Fig. 1. Vertical dotted lines indicate the nominal wavelengths of the Mg iiλ2797, λ2803 emission lines. For three galaxies, J0812+3200, J1045+3225, and J1716+2744, two spectra are shown that are available in the SDSS database.

Open with DEXTER
In the text
thumbnail Fig. 3

Dependence of I(Mg iiλ2797)/I(Hβ) on the excitation parameter x = O2+/O. Red open circles are data from CLOUDY models calculated by adopting 12 + log O/H = 8.0 and 8.3. Large black filled circles are extinction-corrected intensities for the H ii regions with 12 + log O/H ≥ 7.9 and small light blue filled circles with 12 + log O/H < 7.9.

Open with DEXTER
In the text
thumbnail Fig. 4

a) Relation between the electron temperatures te(Mg ii) and te(O ii). te are in units of 10-4   Te. Red filled circles, blue open circles, and green stars are data from CLOUDY models with starburst ages of 2.0 Myr, 3.5 Myr, and 4.0 Myr, respectively. The solid line is the fit to data with all ages, defined by Eq. (2). The dotted line is the line of equal temperatures. The range of the electron temperature te(O ii) in the sample galaxies is shown by a horizontal line. The error bar is the dispersion of te(Mg ii) in the range of te(O ii) shown by the vertical line. b) Relation between the ionisation correction factor ICF(Mg+) and the excitation parameter x = O2+/O. Red filled circles, blue open circles, and green stars are data from CLOUDY models with starburst ages of 2.0 Myr, 3.5 Myr, and 4.0 Myr, respectively. The solid line is the fit to data with all ages as defined by Eq. (5) with the coefficients from Table 3. The range of the excitation parameter x in the sample galaxies is shown by a horizontal line. The error bar is the dispersion of ICF(Mg+) (in dex) in the range of x shown by the vertical line.

Open with DEXTER
In the text
thumbnail Fig. 5

Dependence of the magnesium-to-oxygen abundance ratio relative to the solar value, [Mg/O] = log (Mg/O) – log (Mg/O), on oxygen abundance 12+log O/H with error bars for all data points. The solar value log (Mg/O) = −1.09 is adopted from Asplund et al. (2009). Dashed line and ⟨[Mg/O]⟩ in both panels are mean values of [Mg/O]. a) The encircled galaxies are those with no clear presence of blue-shifted absorption profiles (Fig. 2). b) The same sample as in a) but with division between bright Mg ii lines with I(λ2797 Å)/I(Hβ) ≥ 0.3 (red filled circles) and weak Mg ii lines with I(λ2797 Å)/I(Hβ) < 0.3 (blue filled circles). Encircled galaxies in b) are those where [O iii]λ4363 Å is measured. Additionally, the mean values of [Mg/O] for bright (⟨[Mg/O]⟩(bright)) and weak (⟨[Mg/O]⟩(weak)) Mg ii lines are denoted.

Open with DEXTER
In the text
thumbnail Fig. 6

Dependence of the magnesium-to-oxygen abundance ratio relative to the solar value [Mg/O] on the equivalent width EW(Hβ) of the Hβ emission line. The galaxies with the measured [O iii]λ4363 Å line are encircled.

Open with DEXTER
In the text
thumbnail Fig. 7

Dependence of the magnesium-to-oxygen abundance ratio relative to the solar value [Mg/O] a) and the I(Mg iiλ2797)/I(Hβ) ratio b) on observed fluxes of the Hβ emission line corrected for both the Milky Way and internal extinctions. a) All sample galaxies are shown. b) The same as in a) but H ii regions with [Mg/O] ≥ –0.3 being shown in red and those with [Mg/O] < –0.3 in black. The galaxies with measured [O iii]λ4363 Å line are encircled.

Open with DEXTER
In the text
thumbnail Fig. 8

a) Comparison of magnesium depletion for data available from the literature (denoted by different colours and symbols) and b) comparison of the magnesium depletion averaged for objects of each of four types LISM (local interstellar medium), H ii regions, our Mg ii sample (this paper), and DLA-QSO-GRB (damped Lyα absorption – quasi-stellar object – gamma-ray burst) absorption systems. [α/H] is [O/H] for LISM, H ii regions, and our sample galaxies, and [S/H] for DLA-QSA-GRB absorption systems.

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.