THE MAGELLANIC BRIDGE AS A DAMPED LYMAN ALPHA SYSTEM: PHYSICAL PROPERTIES OF COLD GAS TOWARD PKS 0312–770^*

We briefly summarize our modeling procedure, although it is similar to previous studies (e.g., Churchill & Charlton 1999). Using the photoionization code Cloudy, version 07.02.00 (Ferland et al. 1998), for each of the five clouds (the fifth one is introduced later) we search for the best combinations of the fit parameters: (1) ionization parameter (log U = log [n_γ/n_H], defined as the ratio of ionizing photons to the number density of hydrogen in the absorbing gas) and (2) metallicity (log(Z/Z_☉)) in solar units.¹⁰ The MB is likely to be confined and compressed by its interaction with the Galactic halo, but this does not produce shock ionization (Bland-Hawthorn et al. 2007). Therefore, we consider only photoionization. We optimize on the observed column densities of H i or other metal ions (i.e., O i and Si iv) that are listed in the first column of Table 3, that is we require the Cloudy models to produce those column densities. For example, in the case of the component at V_☉ ∼176 km s⁻¹, we always fix the H i column density to log N(H i/[cm⁻²]) = 19.59, and search for the best values of log U and log(Z/Z_☉) in a grid with steps of 0.1 dex, to reproduce as many absorption lines from other transitions as possible. We repeat this procedure for the other four components, individually. We assume that the absorbers are plane–parallel structures of constant density, and that they are in photoionization equilibrium. We initially adopt a solar abundance pattern, but also explore some variations based on the observational constraints. For the incident radiation field, we first consider a pure extragalactic background radiation with contributions from quasars and star-forming galaxies (with a photon escape fraction of 0.1), following Haardt & Madau (1996, 2001). We also explore the effects of other incident radiation fields, including fluxes from our Galaxy and the LMC, as described in the following section.

For a given log U and log(Z/Z_☉), Cloudy computes the column density of each element in various ionization stages, and the equilibrium gas temperature, T. We can calculate the Doppler parameter for each element using $b = \sqrt{b_{\rm T}^2 + b_{\rm turb}^2}$, where b_T is the thermal broadening defined as $b_{\rm T} = \sqrt{2kT/m}$, and b_turb is the broadening from gas turbulence and bulk motion. We use the measured b for the transition on which we optimized (listed in Table 3) in order to calculate b_turb, which is then applied for other elements. Using these derived line parameters from the model, we synthesize a spectrum, after convolving the instrumental line spread function of HST/STIS, and compare it to the observed spectrum. The synthesized spectra are compared to the observed spectrum by eye, because models with metallicity and ionization parameters that differ only slightly from the best values (e.g., 0.1 or 0.2 dex) would deviate significantly from the observed spectrum (see Figure 4 of Misawa et al. 2008). Moreover, a formal procedure like χ² fitting is not applicable, because the system in the MB has various (more than 10) transitions that we must consider simultaneously, and it is very difficult to determine how they should be weighted in a χ² calculation (e.g., Misawa et al. 2008).

The transitions we optimize are H i and O i, and both have very low ionization potentials (IP = 13.6 eV). As frequently reported for photoionization models of Mg ii absorbers, the low-ionization phase clouds that produce Mg ii absorption lines also produce other low-ionization transitions (e.g., C ii, Si ii, and Fe ii), but not high-ionization transitions (e.g., Si iv and C iv). An additional high-ionization phase is almost always required to reproduce these transitions. Therefore, we repeat line fitting and photoionization modeling for the high-ionization phase by optimizing high-ionization transitions (Si iv in this study, as described below). Finally, we synthesize a model spectrum including both low- and high-ionization phases, and compare it to the observed spectrum.

As described above, our method for deriving constraints on metallicities and ionization parameters of the absorbing gas relies on Voigt-profile fitting of separate components of the absorption profile of a particular transition which we feel is best constrained. We then use photoionization modeling to infer the column densities and Doppler parameters of other transitions in these Voigt-profile component clouds. Synthesized profiles from these models are compared to the data in order that we can place constraints on the parameters. Although this method does rely on the assumptions behind the Cloudy models, it has some advantages over simply taking the ratios of apparent column densities of selected transitions in order to determine metallicity. First of all, we can separately examine properties of individual clouds along the line of sight at different velocities. We are not averaging these components together, which the apparent column density method requires, since it does not distinguish, for example, how much of the hydrogen is associated with the separate clouds. Second, we can use the components determined from unsaturated lines in order to constrain the properties of those clouds using other saturated or partially saturated components as well. The measured Doppler parameters for the unsaturated lines can be used along with the temperature given by a given Cloudy model in order to determine model Doppler parameters of other lines that may be saturated or blended. Comparing the model to the shapes of the observed profiles of these lines, particularly the shapes of the sides of the profiles, often yields meaningful constraints on parameters. Models also take into account the appropriate ionization corrections in each case. In this way, we can determine the range of acceptable parameters, considering all observational constraints. Finally, we are able to consider separate phases of gas, having different densities and velocities along the line of sight, by comparing model predictions to the observations, component by component.

Because the Milky Way (MW) and the LMC are located within several tens of kiloparsec of the MB, they also contribute as additional ionizing radiation sources. Therefore, we consider three alternative incident radiation fields, (1) extra-galactic background (EGB; Haardt & Madau 1996, 2001) plus radiation from the MW at a distance of D = 50 kpc from the MW (Fox et al. 2005), (2) a maximum flux model, with the EGB, the MW radiation, plus radiation from the LMC with a 30% escape fraction, and (3) an intermediate case, with EGB, the MW radiation, plus radiation from the LMC with a 15% escape fraction.¹¹ We construct these radiation fields following the procedure of Bland-Hawthorn & Maloney (1999, 2002). Figure 6 shows the strength of the ionizing radiation from the MW and the LMC as a function of a distance from the center of the MW. At D ∼50 kpc, at which the MB is located, the contribution from the LMC to the radiation field is significant. On the other hand, the contribution from the SMC is negligible because of its low luminosity.

Zoom In Zoom Out Reset image size

The spectral shapes for each of the four incident radiation field models on the MB gas are plotted in Figure 7. The radiation from the MW and the LMC start to dominate the EBR at log (ν/[Hz]) < 16.1, and strongly dominate over the EBR at log (ν/[Hz]) < 15.5. Normalizations for these fields, in units of the density of ionizing photons, and the contribution from the local flux sources (MW and LMC) compared to the EBR are summarized in Table 4.

Zoom In Zoom Out Reset image size

At first, we seek the best photoionization model parameters for the three H i 21 cm clouds at Δv = −34, −3, and 26 km s⁻¹ from the system center (i.e., V_☉ = 176, 207, and 236 km s⁻¹) that we determined in Section 3 by optimizing a fit to the H i 21 cm absorption line. We require our model to reproduce the column density of H i from that fit. Because the cloud at Δv = −3 km s⁻¹ has clear detections in C i, N i, Ni ii, and S ii, without blending with other absorption features, we begin with this cloud. We can constrain the metallicity to be −0.9 ⩽ log(Z/Z_☉) ⩽ −0.5/−0.5 ⩽ log(Z/Z_☉) ⩽−0.2 to avoid over/underproduction of Ni ii and S ii. Therefore, the acceptable metallicity is log(Z/Z_☉) ∼−0.5. With this metallicity, the ionization parameter must be −6.0 < log U< −4.8 to avoid over/underproduction of C i. Even our favored model, with log(Z/Z_☉) ∼−0.5 and log U ∼−5.8, for this Δv = −3 km s⁻¹ cloud substantially overproduces the N i λ1201 absorption. To reconcile the model with the observed N i λ1201 absorption, we decrease the nitrogen abundance by a factor of 0.7 dex compared to its solar abundance pattern. The production mechanisms of nitrogen are poorly understood. However, a nitrogen deficiency in the SMC has already been reported (e.g., Mallouris et al. 2001), and similar deficiencies are seen in some DLA systems (e.g., Pettini et al. 2002) and weak Mg ii systems (Zonak et al. 2004). Since O i and H i are strongly coupled in terms of their ionization, their ratio is often used as a measure of metallicity (Lehner et al. 2008). In our case, the O i is highly saturated so that the Voigt-profile components cannot be measured directly from the profile. However, once we have determined the metallicity from the weaker Ni ii and S ii profiles, we can verify that the favored model produces an O i/H i ratio consistent with these metallicities. For the −3 km s⁻¹ component, we find log [N(O i)/N(H i)] = −3.81, corresponding to log(Z/Z_☉)= −0.5. The fact that this is consistent with the metallicity inferred from the weaker profiles confirms that ionization corrections are inferred properly from our models, and that the saturated O i profile has been separated self-consistently into Voigt-profile components.

Next, we consider the H i cloud at Δv = 26 km s⁻¹. Because this component has the largest recessional velocity, it should reproduce at the high-velocity side of all absorption features seen in the O i, Mg ii, Fe ii, and Si ii lines. We also require that the component reproduces the weak Ni ii absorption. The Si ii provides the best lower limit on metallicity, log(Z/Z_☉) > −0.8, because the less saturated Si ii λ 1304 profile is available. Similarly, an upper limit of log(Z/Z_☉)< −0.6 applies in order that Ni ii is not overproduced. At the preferred value of log(Z/Z_☉) ∼−0.7, there is a strict lower limit on the ionization parameter of log U >−6.1 under which C i would be overproduced. We do not place any formal upper limit on the ionization parameter of this cloud, since the constraints from the relatively low signal-to-noise spectrum were not significant. However, we can place a marginal upper limit, log U < −5.0 to avoid an underproduction of Si ii, once we adopt log(Z/Z_☉) = −0.7. We favor values toward the lower end of this range, log U ∼−6.0, because the fits to O i, Mg ii, and Fe ii are slightly better. We also find that this cloud must have a deficiency of nitrogen of 1.4 dex compared to the solar value, although this value could be smaller (Section 6).

Because the third H i cloud at Δv = −34 km s⁻¹ is heavily blended with the adjacent clouds, it is difficult to place strict constraints on its physical conditions. Therefore, we simply note that its metallicity and ionization parameter could be very close to those of the cloud at Δv = 26 km s⁻¹. Several transitions, Mg i, Ni ii, and C i, are overproduced for metallicities higher than log(Z/Z_☉) > −0.6 or −0.7. The metallicity should also be greater than −1.0, to avoid an underproduction of Si ii. Although we cannot place strong constraints on the ionization parameter, the log U should be between −6.0 and −5.0 in the acceptable range of the metallicity, −1.0 < log(Z/Z_☉) < −0.7, to reproduce the observed profiles of C i, Si ii, and O i. We adopt a model with log U ∼−5.7 and log(Z/Z_☉) ∼−0.7, the same as for the Δv = 26 km s⁻¹ cloud, because it is consistent with the observations. We also find that this cloud must have a deficiency of nitrogen of 1.0 dex compared to the solar value.

In addition to the three H i clouds above, one high-ionization cloud, at Δv = 7 km s⁻¹ (i.e., V_☉ = 217 km s⁻¹), is necessary to reproduce absorption in the high-ionization lines, Si iv and C iv. To determine the photoionization model parameters for this cloud, we optimize on the Si iv column density, since the Voigt-profile fitting is better in this region than it is for the C iv doublet. The result of a Voigt-profile fit to the Si iv is given in the last row of Table 3. Comparing to the observed C iv absorption, we find an ionization parameter of −2.7 ⩽ log U ⩽ −2.4. The metallicity of the high-ionization cloud is constrained to be log(Z/Z_☉) > −4.0 by the requirement that the corresponding H i 21 cm line is not detected at that velocity. The metallicity could actually be much higher, as high as that of the low-ionization clouds, but we have no way to place further constraints.

A model with three low-ionization clouds and one high-ionization cloud reproduces the observed spectrum very well except for the low-velocity side of O I, Fe ii, and Si ii. As noted in Section 3, we add an additional cloud at Δv = −49 km s⁻¹ (i.e., V_☉ = 161 km s⁻¹) in order to fully produce the observed O i absorption. For this cloud, we can constrain the metallicity by the requirement that the optical depth of H i 21 cm at the line center must not exceed τ = 0.025. We obtain log(Z/Z_☉) ⩾ −1.0 for log U ⩾ −5.1, and log(Z/Z_☉) ⩾ −0.9 for log U⩽ −5.2. The ionization parameter should be log U >−6.0, below which Mg i would be overproduced. An upper limit on the ionization parameter of this component is −3.5 to avoid an overproduction of Si iv. We again see that nitrogen is deficient in this cloud by 0.6 dex. We give a sample of an acceptable model for this cloud, with log(Z/Z_☉) = −1.0 and log U = −5.1 in Table 3. However, we note that the parameters could instead be more similar to those of the other clouds.

Table 3 lists the best parameters of our photoionization models: Column 1 is the optimized transition, Column 2 is the heliocentric velocity of the absorption line, Column 3 is relative velocity from the system center, Columns 4 and 5 are absorption optical depth at the line center and spin temperature measured from the H i 21 cm emission line (only for H i 21 cm lines), Columns 6 and 7 are the best Voigt-profile fit values of column density and Doppler parameter, Columns 8 and 9 are the best-model parameters (assuming a solar abundance pattern) of metallicity and ionization parameter, Column 10 is gas temperature from our model, Columns 11 and 12 are the gas volume density and the thickness of the absorber assuming a plane–parallel structure, and Column 13 notes changes in the nitrogen abundance (compared to the Solar abundance pattern) that are necessary to reproduce the observed spectrum. The synthesized spectrum, using the best-model parameters, is overplotted on the observed spectrum for each transition in Figure 8. If we compare the models with/without nitrogen deficiency to the observed spectrum, it is obvious that the nitrogen deficiency is necessary, as shown in Figure 9. A summary of how specific transitions were used to constrain log U and log(Z/Z_☉) for the MB system is given in Table 5. We also compare the equivalent widths of the observed and the modeled spectra in Table 2, and confirmed they are in good agreement.

Zoom In Zoom Out Reset image size

Because this absorber is located in the MB, close to the MW and the LMC, the radiation field around it could be enhanced by contributions from these galaxies, as mentioned in Section 4.2. We perform photoionization analysis using the various radiation fields listed in Table 4. Because the purpose of this analysis is to examine how the shape of radiation field affects the results of the photoionization modeling, we always use the best-fit parameters from Table 3. We compare the model spectra using various incident radiations to the observed spectrum in Figure 10. On the observed spectrum, we do not see any significant differences between the results for the different radiation field models, except for C i λ1329, for which the model is slightly improved if we take a radiation from the LMC into account. However, the quality of spectrum around this line is not very high. Apparently the same fit parameters are still acceptable for the alternative radiation fields. However, once we estimate the line parameters of various transitions in the strongest component at Δv = −3 km s⁻¹ when using these incident fields, there are noticeable differences for some transitions (i.e., C iv, N iii, Al iii, and Si iv) that are very weak and/or positioned at low S/N regions. This is because the number of hydrogen ionizing photon is basically same by our assumption (because we assume the same ionization parameter); however, the ionizing photons of some ions (especially ions with low ionization potentials) would be increased significantly once the additional radiation is considered. Although the ionization parameter that we infer is the same for these alternative radiation fields, the number of ionizing photons is substantially increased so that this same ionization parameter corresponds to a higher density. For the Δv = −3 km s⁻¹ component, the density has increased from 0.56 cm⁻³ to 19 cm⁻³, and the size decreased from 25 pc to 0.72 pc in the most extreme case. The other clouds are similarly smaller, with the Si iv cloud reduced in size to 0.57 kpc under the more intense LMC radiation field. Since the differences in the ionization parameters that we infer are so small between the different models, it is not practical to distinguish between the different possible radiation fields using these data, and thus the inferred densities and sizes are uncertain. Hereafter, we discuss the results using only the EGB radiation, as summarized in Table 3.

Zoom In Zoom Out Reset image size

There have been three scenarios posed to explain the current chemical composition of the MB. Tidal stripping from the SMC, purported to explain the origin of the MB (Gardiner & Noguchi 1996) would imply a chemical composition similar to the SMC itself, or perhaps slightly more metal-poor if the MB is preferentially formed from material drawn from the comparatively pristine outskirts of the SMC. The presence of any chemical gradient in the SMC is not yet well defined, but the recent results of Carrera et al. (2008) suggest it may be non-negligible. A galactic wind from the SMC could enrich the MB with metals, and such a wind would likely be enhanced in alpha elements, given the composition of supernova-driven winds in other dwarf galaxies where such signatures are observed several kpc from their origin (Martin et al. 2002; Strickland et al. 2004). The final possibility is in situ enrichment from stars formed in the MB (Demers & Battinelli 1998). Given the ∼200 Myr age of the MB (Gardiner & Noguchi 1996), there has been ample time for several generations of the most massive B stars to evolve and contribute their nucleosynthetic products to their surroundings. Some combinations of these processes are likely to play a role, but the balance of these has not yet been determined.

Rolleston et al. (1999) found an underabundance (∼0.5 dex) in the light metals of a B-type star in the MB, DGIK 975, compared to the SMC H ii regions. Similarly, metallicity of the interstellar medium in the MB toward a young star, DI 1388, was measured to be ∼0.2 dex lower than the SMC H ii regions (Lehner et al. 2008). The sightline toward our target, PKS 0312 − 770, has an angular separation of ∼41 from DI 1388, which corresponds to about ∼4.4 kpc in physical scale at the distance of the SMC (d ∼ 60 kpc). However, we have found that the metallicity in the MB toward PKS 0312 − 770 is higher than that measured toward DI 1388 by ∼0.5 dex, and even ∼0.1–0.2 dex more metal-rich than the SMC H ii regions. What is the source of this difference? One possible idea is that the sightline toward PKS 0312 − 770 is not mixed with metal-poor gas, like the sightline toward DI 1388. If this is the case, absorbers with lower metallicities should have higher total hydrogen column densities. Such a trend (i.e., a gradual increase in metallicity with decreasing hydrogen column density) has already been pointed out for various objects at higher redshift, including galaxies, (sub)DLA systems, and the intergalactic medium (e.g., Boisse et al. 1998; Péroux et al. 2003; York et al. 2006; Misawa et al. 2008).

There are observational results that support the scenario above. Column densities of sulfur (a modestly dust-depleted element) are similar between these two sightlines. Each of the three clouds toward PKS 0312 − 770 (with log N(S/[cm⁻²])is 14.2, 14.4, and 14.2)¹² has a similar column density to the total (log N(S/[cm⁻²]) = 14.35) in DI 1388. However, the PKS 0312 − 770 clouds have total hydrogen column densities (log N(H/[cm⁻²]) = 19.6, 19.6, 19.7)¹³ 0.6 dex smaller than the total in DI 1388 (log N(H/[cm⁻²]) = 20.2). This implies a higher metallicity and less dilution in PKS 0312 − 770. These results are consistent with the scenario above, i.e., metal-enriched gas flowed out toward the MB from the SMC almost isotropically, but the MB absorbers toward PKS 0312 − 770 were only weakly diluted by metal-free material, while the absorber toward DI 1388 is significantly diluted.

The overall metallicity and the chemical abundance pattern in the MB can be used to infer the origin of this inter-Cloud material. Hambly et al. (1994) and Rolleston et al. (1999) measured He, O, Si, Mg, and N abundances for several early B stars in the MB, assuming that their composition reflects the present-day interstellar medium (ISM), and concluded that these stars were 0.5 dex more metal-poor than SMC stars. These authors suggested that the MB stars formed from a mixture of SMC and unenriched gas. Dufour (1984) analyzed the H ii region He, C, N, O, and Si abundances for the SMC, finding that the overall metallicity, as measured by the alpha elements O and Si, is ∼0.5 dex more metal-poor than the Sun. The N/Si ratio in H ii regions is 0.5 dex lower than the Sun, consistent with the trend observed in other low-metallicity dwarf galaxies (e.g., Kobulnicky & Skillman 1996; Nava et al. 2006). We summarize these measurements from the literature along with ours in Table 6. Locations of the MB stars are also shown in Figure 1.

Figure 11 shows the abundance ratio log (N/Si) versus log (Si/H) for the aforementioned measurements from Table 6, assuming a solar abundance pattern if another element (instead of silicon) is used to estimate these parameters. The figure includes our 3-velocity components toward PKS 0312 − 770 (open circles), the Rolleston et al. (1999) MB stars (filled stars), the gas toward the MB star DI 1388 (Lehner et al. 2008; open star), the SMC star AV 304 (open square), the Sun (solar symbol), the range of SMC H ii regions (filled square), the dwarf galaxies from Nava et al. (2006; crosses; values computed from O abundance measurements assuming a solar Si/O ratio), and damped Lyα systems from Henry & Prochaska (2007; dots). The abscissa reflects the overall alpha element abundance of the systems in question and shows that the metallicities of the SMC star AV 304, the MB stars, and the MB gas toward PKS 0312 − 770 are generally consistent with the range of metallicities in SMC H ii regions. The MB star DGIK 975 is a possible exception, appearing ∼0.5 dex more metal poor than the rest of these measurements. Figure 3 of Rolleston et al. (1999) shows that this star also lies the furthest from the SMC, in a low H i column density region equidistant between the SMC and LMC.

Zoom In Zoom Out Reset image size

The N/Si ratio shown on the ordinate of Figure 11 is a measure of the chemical enrichment timescale. Si is synthesized in massive stars and returned to the ISM through supernovae on timescales of ∼10 Myr.¹⁴ Nitrogen, by contrast, is thought to be produced in most galaxies by low- and intermediate-mass stars (LIMS) and released on timescales of >100 Myr (van den Hoek & Groenewegen 1997; Marigo 2001; Renzini & Voli 1981).¹⁵ Thus, the N/Si ratio may drop during prolonged starbursts and rise during prolonged periods of quiescence, functioning as a kind of "clock," marking the time since the most recent major episode of star formation (Edmunds & Pagel 1978; Pantelaki 1988; Garnett 1990; Kobulnicky & Skillman 1998). Henry & Prochaska (2007) used this scenario to model the chemical evolution of DLA systems in the N/Si versus Si/H plane and found that they could reproduce the properties of most DLA systems as a function of two variables: the star formation efficiency and the age of the system. They concluded that small ages (i.e., less than 250 Myr) are required to produce the low N/Si systems, while high star formation efficiencies lead to higher Si/H and lower N/Si ratios. However, a number of other processes may contribute to shaping the chemical evolution of a system. The arrows in Figure 11 show qualitatively the four possible evolutionary vectors attributed to N enrichment (vertically upward), α enrichment (toward the lower right), dust depletion (toward upper left), and dilution by metal-poor gas (leftward).

Figure 11 shows that the 207 km s⁻¹ component toward PKS 0312 − 770 is 0.3–0.4 dex more metal-rich than the bulk of the SMC and other MB stars. This component also exhibits an N/Si ratio that is 0.2–0.3 dex lower than the SMC stars and H ii regions. This departure is consistent with a composition consisting of SMC material augmented by a small amount of α enrichment from supernovae. None of the MB stars in Figure 11 nor the ISM probed toward the MB star DI 1388 share this abundance pattern.¹⁶ The PKS 0312 − 770 line of sight is therefore the first MB location with a metallicity that is similar to, or perhaps slightly higher than, the SMC at large.

Figure 11 shows that the absorbing component at 236 km s⁻¹ has a metallicity similar to the SMC and most MB stars, but the N/Si ratio is about 1.0 dex lower.¹⁷ Such a low N/Si ratio places this material among the most N-deficient DLA systems and suggests a nucleosynthetic history dominated by α-producing massive stars. In the Henry & Prochaska (2007) models, such a low N abundance at such high metallicity can only be achieved by very efficient and very recent star formation such that massive stars dominate the mass-averaged nucleosynthetic contribution with virtually no contribution from longer lived N-producing stars. Given the presence of B stars in the MB (e.g., Rolleston et al. 1999), some supernova activity during the < 200 Myr lifetime of the MB is likely. In situ enrichment appears a plausible explanation for this component on the basis of chemical cloud. However, it is also possible that a Si enrichment of ambient SMC material plus dilution from putative metal-poor halo gas could also explain this data point (i.e., a combination of vectors that together drive evolution in Figure 11 in a downward vertical direction from the SMC H ii regions composition).

Zoom In Zoom Out Reset image size

Finally, the 176 km s⁻¹ component has an Si/H ratio consistent with the SMC and 236 km s⁻¹ component, but the uncertainties on the N/Si ratio are too large to place meaningful constraints on the origin of this material. Moreover, abundance pattern would also be strongly affected by dust depletion and ionization conditions, which can only be explored with very high S/N spectrum. Therefore, this component may either be N-deficient or it may be consistent with the SMC H ii regions.

In addition to various absorption lines we have detected in the HST/STIS spectrum, Smoker et al. (2005) also detected the Ca ii K line in their medium resolution (R = 6000) optical spectra of seven quasars behind the MB including our target (Smoker et al. 2005). By comparing the total column densities of Ca ii and H i (measured from H i 21 cm emission line), they found that the abundance ratio of Ca ii to H i in the MB is systematically higher than that of Galactic gas by a factor of ∼0.5 dex. Smoker et al. (2005) proposed possible scenarios for this difference, such as the higher ionization condition of the hydrogen gas, and weaker dust depletion in the MB. To test the scenario, a higher resolution spectrum will be necessary to deblend an unresolved Ca ii K profile into multiple components (as we did for other transitions in the UV spectrum) in order to constrain the photoionization model.

In Section 5, we found the best-model parameters. However, there is still an ambiguity due to the possible effects of dust, which would lead to different inferred physical conditions. Our data were not of sufficient quality to measure dust depletion, but the estimated low gas temperature of the three H i clouds (T_gas < 1000 K) would imply that dust grains could survive in the absorber. Moreover, the Ca ii K absorption strength, expected from our best model (W_rest = 0.99 Å), is 7 times greater than the observed value toward PKS 0312 − 770 (W_rest = 0.14 Å; Smoker et al. 2005). This also implies that the absorber contains substantial amounts of dust, because calcium is one of the most severely depleted elements (Savage & Sembach 1996). Actually, Lehner et al. (2008) suggested Si and Fe in the MB toward DI 1388 are depleted to dust by factors of −0.45 and −0.61 dex, respectively, although the condition could be different toward PKS 0312 − 770. If depletion onto dust is significant, then the Si abundances and the implied metallicities become even larger, making this sightline significantly more metal-enriched than the SMC itself.

DLAs are characterized by high H i column densities (i.e., log N(H i) ⩾ 2 × 10²⁰ cm⁻³) and low metallicities; i.e., log(Z/Z_☉) ∼ 0.1–0.01 (e.g., Pettini et al. 1997; Prochaska et al. 2003). DLAs are also known to have lower nitrogen abundance relative to α elements, compared to the solar abundance pattern (e.g., Pettini et al. 2002). DLAs provide plentiful information on the physical condition in gas clouds in the ancient universe that can never be traced by stellar objects. However, so far DLA absorbers have not been identified clearly, especially at higher redshift¹⁸, although at least elliptical galaxies are probably ruled out (Calura et al. 2003). Thus, it is quite helpful to find local counterparts of those systems and study their properties in detail.

With respect to H i column density, the MB absorbers toward PKS 0312 − 770, whose total column density is just below the criterion to be classified as a DLA system, could be analogs of high-z DLA systems. However, their nitrogen abundance relative to α elements (e.g., [N/Si]) tend to be small and positioned at the lowest end of the [N/Si] distribution of DLA systems in Figure 11. Particularly, the nitrogen abundance of the V_☉ = 236 cloud is very small compared to that expected for DLA systems and blue compact galaxies of a similar metallicity. The origin of this difference should be understood before using the MB absorbers as local counterparts of high-z DLA systems.

The nitrogen deficiency may be linked to the synthesis process. It is already known that nitrogen is produced by LIMS (M/M_☉ ⩽ 8), while α elements are produced by massive stars. For DLA systems with low nitrogen (hereafter, low nitrogen DLAs: LNDLAs), there have been two possible scenarios presented (Henry & Prochaska 2007 and references therein), i.e., (1) they are in early production stages of nitrogen released from LIMS (delay scenario; Pettini et al. 2002), and (2) they have an intrinsically flattened or truncated initial mass function (IMF) with fewer LIMS (reduction scenario; Prochaska et al. 2002). However, as summarized in Henry & Prochaska (2007), neither scenario can explain the nitrogen deficiency of LNDLAs perfectly; the former cannot reproduce a possible bimodality of [N/α] distribution of DLAs (Prochaska et al. 2002; Centurión et al. 2003), while the latter produces too much iron compared with the observed amount (Lanfranchi & Matteucci 2003).

Thus, we have not yet identified the origin of the nitrogen deficiency. Nonetheless, it is likely that the production history of nitrogen is not the same between the SMC and the MB. The IMF of a stellar association NGC 602 in the SMC (a single power law with a slope of Γ ∼−1.2 for M/M_☉ = 1–45; Schmalzl et al. 2008) is quite similar to the IMF in the solar neighborhood (Γ ∼−1.35 for M/M_☉ = 0.4–10; Salpeter 1955). This means that the abundance patterns of the MW and the SMC are expected to be similar, while they can be different from that of the MB, whose IMF is not necessarily the same. The MB probably has a local star-forming history independent of the nearby star-forming regions (i.e., the LMC and the SMC), which is consistent with the observed results that the young stars in the MB are not old enough to have escaped from the SMC based on their peculiar motions.

If the MB absorbers are indeed local counterparts of high-z DLAs, our results would have interesting implications: (1) DLA systems along our sightlines to the background quasars do not necessarily represent global properties of absorbing structures; and (2) the large scatter of metallicity and [N/α] values could be due to the internal gradient of each DLA absorber, and they are different from place to place along different sightlines go through. For example, we would underestimate the global metallicity (and nitrogen abundance) of the Magellanic Clouds if our line of sight went through only the MB. Moreover, there might be metallicity gradient in the MB itself—higher in the SMC wing than in the remaining parts of the MB (Lehner et al. 2008 and references therein). As for DLAs, Ellison et al. (2005) and Chen et al. (2005) already suggested that there could be a metallicity gradient as a function of a distance from the galactic center. Ellison et al. (2005) and Wolfe et al. (2003) also proposed that the DLA region is sampling a lower metallicity than the star-forming regions. Thus, high-z DLA systems also may have complex internal structures like the Magellanic Clouds and their neighborhood.