INVESTIGATING NEARBY STAR-FORMING GALAXIES IN THE ULTRAVIOLET WITH HST/COS SPECTROSCOPY. I. SPECTRAL ANALYSIS AND INTERSTELLAR ABUNDANCE DETERMINATIONS

The galaxies of our sample were selected on the basis of the following criteria. The most important one was the existence of archival FUSE data (e.g., I Zw 18, whose FUSE spectra have already been analyzed by Aloisi et al. 2003). Selecting galaxies with FUSE spectra provided three advantages: (1) it ensured that the sample galaxies are strong emitters in the FUV, (2) it guaranteed the detection of metal absorption lines, and (3) the FUSE data provided a direct estimate of the flux in the FUV spectral region used for the COS observations. Although our sample of nine nearby SFGs is by no means a rigorously defined, statistically complete one, it does span a broad range in the parameter space. More specifically, the sample covers a range of galaxy spectral types, metallicities as inferred from O in the H ii regions, and SFRs as indicated by log(L_UV+L_FIR).

The galaxies in the samples ordered by increasing metallicity and basic properties are listed in Table 1. The galaxy right ascension and declination, morphological type, systemic velocity, and foreground galactic extinction E(B − V) were all taken from the NASA Extragalactic Database (NED). The metallicities (Z) are given in terms of the oxygen abundance by number (Z = 12+log(O/H), where Z_☉ = 8.69, Asplund et al. 2009). Further information about these metallicities (including references) are given in the table and its associated notes. The majority of distances were compiled from the HyperLeda database⁴ (Paturel et al. 2003) and represent those calculated from the tip of the red giant branch (RGB) obtained from resolved stellar population studies with HST. For galaxies where this method cannot be applied, distances adopted were calculated using the standard linear Virgo-centric infall model of Mould et al. (2000), as listed in NED.

The far infrared (FIR) luminosity (L_FIR) was calculated from the "FIR" parameter (the flux between 60 and 100 μm) published in the IRAS survey (Fullmer & Lonsdale 1989). The UV luminosity (L_UV) was calculated from the flux at 1900 Å corrected for reddening using the E(B − V) values listed in the table. See Heckman et al. (1998) for more details on how these luminosities were calculated.

Because the primary motivation for this work was to study the abundances in the neutral ISM of nearby SFGs by using absorption lines arising from the COS spectra of UV background sources, it was necessary to select within each galaxy the optimal UV "point-like" sources to target with COS. In order to do this, we performed ACS/solar blind channel (SBC) FUV imaging of each galaxy in the F125LP filter (Figure 1). The only exception was SBS 0335-052, for which an ACS/SBC F140LP archival image already existed and was used for this purpose. With λ_eff = 1440 Å, the F125LP filter has a bandwidth that covers as much as possible of the wavelength range of the COS G130M grating (∼1150–1450 Å) and at the same time avoids contamination from Lyα. Table 2 lists the details of the ACS/SBC observations for all galaxies within our sample. Reduction of the images was performed via the standard ACS reduction pipeline CALACS v5.0.5, which is part of the STSDAS (Space Telescope Science Data Analysis System) software package within IRAF (Image Reduction and Analysis Facility).

Zoom In Zoom Out Reset image size

Along with acting as 30'' × 30'' early-acquisition images for the spectroscopy, these images were crucial in guiding the selection of the COS targets. The point sources were selected based on the following criteria. First, in order to optimize the signal-to-noise ratio (S/N) within the allocated amount of time, we needed to select the brightest point sources within the ACS/SBC images. Second, because the 25 COS aperture is larger than the extension of a point source, in order to avoid resolution degradation we had to consider the spatial extent of these point sources. Based on the spectral resolution of FUSE observations of very similar targets (e.g., I Zw 18, Aloisi et al. 2003), which was sufficient in performing the ISM analysis we present here, we found that a value as high as 60 km s⁻¹ would work. Combining this with the line-spread function (LSF) of the COS spectrograph (which is linearly related to the spatial extent of the source within the aperture), this lower limit implied that we could observe sources that are ∼four times wider than a point source, i.e., with a maximum diameter of ∼0.5 arcsec compared to the angular extent of 0.13 arcsec in the dispersion direction for a point source⁵. Third, to avoid selecting red targets instead of UV targets as a result of the ACS/SBC "red leak," the optical color of each point source was derived using multiband images of the sample galaxies in the HST archive. The final consideration was to only select sources that were in relatively uncrowded regions, to allow the search procedure in the COS target acquisition (TA) to correctly select the desired point source. Thus the selected target sources were as compact as possible, no larger than ∼0.5 arcsec in diameter, bright in FUV (i.e., ⩽16 STMAG), and in uncrowded regions. The coordinates of the UV point-like sources observed within each galaxy are listed in Table 3.

For two galaxies within the sample (NGC 5253 and M83), we obtained observations along two separate sight lines. This was done with an aim of testing for spatial variations of the ISM metal abundances within the galaxy.

We obtained HST/COS FUV observations of 11 point-like sources in nine nearby SFGs using the G130M medium-resolution grating with the 1291 central-wavelength setting. Each observation utilized the four FP-POS positions (small dithers in the dispersion direction) to get the best data quality achievable without a flat-field correction. This corresponded to a final wavelength range of ∼1130–1420 Å. Full details of each COS observation are given in Table 3. For each COS observation, the telescope was initially pointed at the coordinates listed in Table 3, and NUV imaging TA was performed to properly center each target within the COS aperture. A TA configuration of primary science aperture (PSA) and Mirror A was used in most cases, except for the brightest targets, where Mirror B was instead adopted.

After retrieval from the archive, all data were reduced locally using the COS data-reduction package CALCOS v.2.14.4. This calibration pipeline consists of three main components that calibrate COS data by (1) correcting for instrumental effects such as thermal drifts, geometric distortion corrections, Doppler corrections, and pixel-to-pixel variations in sensitivity (this latter step is not performed for FUV data because of a lack of a good two-dimensional flat field), (2) generating an exposure-specific wavelength-calibrated scale, and (3) extracting and producing a final (one-dimensional) flux-calibrated (summed) spectrum for the entire observation (Shaw et al. 2009).

In consideration of the absence of a two-dimensional flat-field correction for COS FUV observations within the CALCOS pipeline, it was necessary to investigate whether the quality of the data would be improved by applying something similar to such a correction. To do this, we derived two sets of one-dimensional flat-field templates using an iterative procedure that selects regular features in pixel space (e.g., the shadows of the ion-repeller grid wires of the COS FUV detector) within the four FP-POS positions of a spectroscopic observation⁶. We used both our spectra and high S/N data from white dwarf stars for this purpose. In both cases, due to the moderately high S/N of our data (S/N ∼ 10–30 per six-pixel resolution element in the final combined spectra; see Table 3), the imperfect one-dimensional flat-field correction introduced noise rather than removing any spurious detector features. We therefore decided that in the case of our observations taken with all four FP-POS positions available for the setting selected, removing the grid wires by flagging them through the bad-pixel table used by CALCOS during the data reduction was sufficient to remove the major flat-field features.

Each spectrum was then inspected in pixel space for any erroneous features due to effects of gain sag around pixels 7200 and 9100 (the position on the FUV detector where the geocoronal Lyα emission line falls when the COS spectrograph is used in two of its most popular observing modes). If these features were present, the data were reduced with pulse-height amplitude settings of 2-31 (instead of the default 4-31) to recover any flux lost due to this gain–sag effect. The data were finally binned by a factor of six in order to improve the identification of spectral features without degrading the spectral resolution.

The resultant 11 COS spectra of the nine targets in the rest-frame wavelengths are shown in Figure 2. The radial velocity inferred from the C iii λ1176 stellar absorption line arising in the spectrum of the UV background source within each galaxy has been used to convert observed into rest-frame wavelengths (see Table 4). Up to three absorption line systems at three different radial velocities are present in each spectrum: the radial velocity of the galaxy, the Milky Way (MW), and, if present, a high-velocity cloud (HVC). Zoomed-in spectra for each of the galaxies in the observed wavelengths are shown separately in Figures 3–8.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Because of the passage of sunlight through the Earth's atmosphere, each observation is subject to contaminating geocoronal emission or "airglow." The magnitude of this contamination depends highly on the limb angle and the altitude of the sun during each observation. The emission spectrum is made up of two major components: the Lyα emission at 1215.67 Å and a triplet of oxygen lines at 1302 Å (O i), 1304 Å (O i*), and 1306 Å (O i**). The removal of the oxygen geocoronal emission is paramount for the low-redshift absorption line spectra presented here, where O i λ1302 galaxy absorption and O i geocoronal emission coincide with one another.

Although the observer cannot completely remove Lyα emission (because it is in a very extended layer in the Earth's atmosphere), the strong dependence of the O i geocoronal lines on limb angle enables the observer to essentially select times when such emission is absent. We were therefore able to remove the O i emission contamination in each of our spectra by creating "night" spectra. This was done in a three-stage process. First, the number of counts in both the 1215.67 Å and 1302 Å regions were analyzed as a function of time from each FP-POS corrtag file and compared against counts within ∼20 Å regions either side of these wavelengths. This essentially allowed us to determine when the emission was minimal or absent. Second, we plotted the altitude of the sun as a function of time throughout each FP-POS observation. Finally, we correlated these two pieces of information to determine the angle at which geocoronal contamination was absent and filtered the observations by angle, i.e., flagging observations above a certain angle as "bad."

By only selecting the observations that were taken below this "airglow angle," where there was no increase in counts on either side of the O i region, we were confident that the airglow had been removed rather than minimized. After recombining the data, this left us with a "night" spectrum, upon which we were able to measure the O i λ1302 absorption line. In some cases, where either the redshift of the galaxy was sufficiently high to shift this line away from the airglow emission or the limb angle was large enough not to contaminate the spectrum, the creation of "night" spectra was not necessary.

The fit of the continuum was performed by interpolating between points of the observed flux free of apparent absorptions (referred to as "nodes" hereafter). Each node was chosen by carefully inspecting by eye the spectra of each target, an average of the flux within a ∼0.2 Å region either side of each node was calculated, and a spline was used for the interpolation of the flux among these nodes.

Systematic errors on the measured column densities (σ_N(X)) due to the uncertainty in the continuum placement were considered as follows. First, the total uncertainty on the continuum placement ($\sigma _{F_C}$) was considered as the error on the node flux (indicated as F_c), i.e., the variance on the mean flux within a 0.4 Å region centered on the node. Then $\sigma _{F_C}$ was calculated for each node across each spectrum and was found to be on the order of ∼3% at most. Second, we needed to assess the effect that $\sigma _{F_C}$ would have on the equivalent width (W) of the absorption line from which column densities, N(X), are derived. Following the calculations of Vollmann & Eversberg (2006), the error on W can be stated as

$$\begin{equation} \sigma _W^2 = \left[-\frac{\Delta \lambda }{F_c}\sigma _F\right]^2+\left[\frac{1}{F_c}(\Delta \lambda -W)\,\sigma _{F_c}\right]^2. \end{equation} \tag{ 1 }$$

Although the first term is dealt with by our line-profile fitting software (which utilizes σ_F, the uncertainty in the flux at wavelength λ, and propagates it to derive σ_N(X)), the second term is not and needs to be considered independently. Third, we needed to propagate σ_W in terms of error on N(X). Because the relationship between W and N(X) varies according to the strength of the line, this had to be assessed for three different line profiles: unsaturated, saturated, and damped. We therefore calculated the magnitude of σ_N(X), taking into account also the contribution of $\sigma _{F_C}$ to σ_W for lines of varying oscillator strength and column density in each of these regimes. The newly estimated σ_N(X) was found to differ by much less than 1% from the value given by our line-profile fitting software. To confirm this, an additional test was performed by measuring with our software a selection of lines on a spectrum whose continuum had been placed $\pm \!1\sigma _{F_c}$ above and below the original continuum nodes. Column densities were found to vary within the uncertainties resulting from σ_F alone. We are therefore confident that σ_N(X) is dominated by σ_F, which is fully taken into account by the line-profile fitting software (described below).

Fitting of the absorption-line profiles requires the spectral resolution of each spectrum as a function of wavelength (FWHM_spec). This was derived using the FWHM of the COS instrument spectral response for a point source in the G130M grating, as measured from the G130M LSF (FWHM_lsf, ∼8 pixels corresponding to ∼0.18 arcsec)⁷, combined in quadrature with the intrinsic FWHM (FWHM_int) of the UV source. The FWHM_int in arcseconds was calculated from the FWHM of the image profile measured in the NUV TA images (FWHM_imag) using the formula FWHM$_{{\rm int}}^2$ = FWHM$_{{\rm imag}}^2-$ FWHM$_{{\rm psf}}^2$, where FWHM_psf is the FWHM of the point-spread function in the NUV TA images (∼2 pixels corresponding to ∼0.05 arcsec). The values of FWHM_spec at ∼1150 Å for each UV point-like source targeted with COS are reported in Table 3. Notice that most of the sources in our sample are only slightly extended (FWHM_spec ∼19 km s⁻¹ for a point-like source).

We derived column densities of H i and heavy elements from the COS spectra of each galaxy by line-profile fitting of the observed absorption lines. Standard theoretical Voigt profiles were convolved with the spectral resolution inferred in Section 3.2 and fitted to the data. The package FITLYMAN (Fontana & Ballester 1995) for multicomponent fitting in the MIDAS software was used for this purpose. This method is more powerful than a simple curve-of-growth (COG) analysis (based on the equivalent width of the absorption lines) because it allows for (1) the deblending of multiple components along the line of sight contributing to the same absorption and (2) the simultaneous and independent fit of contaminating absorption components. We were able to fit absorption lines of all of the ions considered in each galaxy with very simple velocity-component models (i.e., one component, except for NGC 4214, NGC 3690, and M83-Pos 1, where two components were necessary, and NGC 4449, where three components were necessary).

The real situation described by the far-UV spectra of these galaxies is, however, more complex. The COS detects a nonlinear average absorption over the full extent of the stellar background sources covered by its 25 aperture. This implies that the observed lines arise from a combination of many unresolved velocity components from different absorbing clouds along the many sight lines within such an aperture. Moreover, as discussed in Section 5.1.1, some lines of sight may have saturated absorption even if the composite profile does not go to zero intensity (e.g., Savage & Sembach 1991).

Jenkins (1986) has demonstrated that the single-velocity approximation applied to complex blends of features gives nearly the correct answer (the simulated-to-true column density ratio rarely goes below 0.8) if the distribution function of the line characteristics is not irregular. This result holds also if different lines have different saturation levels or Doppler parameters, b. Because our COS data sample many sight lines toward each galaxy over the 25 aperture, we expect a quite regular distribution of the kinematic properties of the single absorbing components. We thus believe that we amply fall within the regime where the single-velocity approximation is valid.

In light of these considerations, we preferred to maintain a simple approach in the determination of the column densities with the line-profile fitting method. We thus avoided the introduction of additional free parameters, i.e., the number of intervening clouds and their velocity distribution, because we believe the resolution of the data does not allow us to correctly constrain the real physical situation represented by this type of observation. In the single-velocity approximation, the fitting parameter b has no precise physical meaning but is rather the result of the combination of both the various line Doppler widths present (due, e.g., to turbulent or thermal broadening) and the various velocity separations among the different line components (Hobbs 1974). On the other hand, according to Jenkins (1986), the column density is well constrained. In addition, the column density of a certain ion is even better constrained if several lines with different values of fλ (λ is the rest-frame wavelength and f is the oscillator strength of the absorption) are available for simultaneous fitting, and the results are independent of saturation problems affecting the strongest lines (see Section 5.1 for a detailed discussion of saturation issues).

However, in order to check the reliability of the line-profile fitting results and the single-velocity approximation, following the method outlined in Aloisi et al. (2003), we also determined total column densities of heavy elements by applying the apparent optical depth method (AODM; Savage & Sembach 1991). This method is very powerful for determining column densities. Its strength lies in the fact that no assumption needs to be made concerning the velocity distribution of absorbers; that is, it does not depend on the number of intervening clouds along the single or multiple lines of sight. However, it does not provide a correct answer for quite strong lines that are not fully resolved. In the case of our resolution, ∼18 km s⁻¹, the line-profile fitting should give a more accurate measure of the ionic column densities compared with the AODM, especially for stronger lines. As mentioned previously, because AODM can only be applied to unblended lines, we were limited to comparing only a few lines within each spectrum and only lines known to be free from saturation issues (Section 5.1). For the handful of lines that were suitable, column densities resulting from the two methods are in excellent agreement, with an average difference of ∼0.1 dex.

In our analysis, the similarities between the column densities obtained with the line-profile fitting and AODM for (unsaturated) lines strengthen our results and justify the single-velocity approximation.

Absorption line profiles can suffer from two types of saturation: "classical" saturation, where a line is optically thick and line width increases logarithmically with column density, and "hidden" saturation, where multiple lines of sight containing both large and small column densities combine to make a line appear, e.g., unsaturated, when in fact it is saturated. In the following sections, we describe the methods used to check for, assess, and treat both types of saturation.

5.1.1. Hidden Saturation

Although there are strong lines that are clearly affected by saturation (namely C ii λ1134 and O ii λ1302, see below), cases of hidden saturation are also present within the COS spectra. This effect can be recognized when fitting multiple lines of the same ion because the strongest lines are too weak (i.e., saturated) compared to the other weaker lines, as illustrated in Figure 9. For the spectra in our study, this effect was evident in the Fe ii triplet (at 1142, 1143, and 1144 Å) and the N i triplet (at 1134.1, 1134.4, and 1134.9 Å) because in both cases several lines are available of varying oscillator strength. Here we discuss the variable amounts of hidden saturation observed within our spectra and assess the roles that various parameters have in causing this effect.

Zoom In Zoom Out Reset image size

Table 7 lists the difference in inferred column density between the weakest and strongest lines within the above-mentioned Fe ii and N i triplets for the majority of galaxies within our sample (in some cases we were unable to constrain column densities from the weaker lines only because of blending). In the case of Fe ii we constrained the column density using Fe ii λ1133, λ1142, or λ1143 (the weakest lines available) and excluded the strongest line, Fe ii λ1144, from the fit. On average, we found that Fe ii λ1144 underpredicts N(Fe ii) by 0.6 dex compared to using only the weakest lines. The difference was less extreme for N i, where N i λ1134.9 underpredicts N(N i) by ∼0.2 dex compared to the λλ1134.1, 1134.4 lines (the N i λ1200 triplet was not used because these lines are very strong and almost certainly suffer from "classical" saturation).

Table 7. Amount of Hidden Saturation in the Strongest Fe ii and N i Absorption Lines

	Fe ii	N i
Galaxy	log[N(1142 or 1143)]−log[N(1144)]	log[N(1134.1, 1134.4)]−log[N(1134.9)]	log[N(H i)]	12+log(O/H)
	(dex)	(dex)	(cm−2)
I Zw 18	0.29 ± 0.06	0.02 ± 0.05	21.28	7.2
SBS 0335−052	0.97 ± 0.04	0.25 ± 0.04	21.70	7.3
SBS 1415+437	0.50 ± 0.04	0.28 ± 0.12	21.09	7.6
NGC 4214	0.19 ± 0.08	0.08 ± 0.08	21.12	8.2
NGC 5253-1	0.92 ± 0.07	0.46 ± 0.06	21.20	8.2
NGC 5253-2	0.96 ± 0.04	0.03 ± 0.04	20.65	8.2
NGC 4670	0.22 ± 0.05	0.03 ± 0.03	21.07	8.2
NGC 4449	...	...	21.14	8.3
NGC 3690v1	...	...	20.62	8.8
NGC 3690v2	...	...	19.82	8.8
M83-1	...	...	19.60	9.2
M83-2	...	0.07 ± 0.06	18.44	9.2

Notes. This table reports the amount of hidden saturation seen within the Fe ii λ1144 and N i λ1134.9 absorption lines. Columns 2 (Fe ii) and 3 (N i) list the difference in the logarithm of the column densities N₁ and N₂ from lines with different oscillator strengths such as f₁ < f₂. For Fe ii, this corresponds to λ₁ = λ1142 or λ1143, λ₂ = λ1144 and for N i, λ₁ = λλ1134.1, 1134.4 and λ₂ = λ1134.9. Ellipses indicate galaxies for which we were unable to constrain N₁ or N₂ because of blending. Columns 4 and 5 list the H i column density (as listed in Table 4) and metallicity (as listed in Table 1), respectively. The values reported in this table can be seen in graphical format in Figures 10–12.

Download table as:  ASCIITypeset image

Figure 10 shows the values listed in Table 7 as a function of neutral hydrogen column density, log[N(H i)/cm⁻²]. It can be seen that, overall, the amount of hidden saturation is greatest for galaxies with log[N(H i)/cm⁻²] > 20.6. For N i alone, saturation of the 1134.9 Å line only occurs when log[N(H i)/cm⁻²] > 21.1. Unfortunately, in the case of Fe ii we are unable to provide a definitive limit to the amount of N(H i) where hidden saturation becomes an issue because several of the galaxies with log[N(H i)/cm⁻²] < 20.6 have multiple velocity components and column densities from the Fe ii λ1142 or Fe ii λ1143 absorption lines could not be constrained.

Zoom In Zoom Out Reset image size

However, this is a multiparameter problem, affected by not only the amount of gas within the sight line but also the physical amount of that species within the gas. Figure 11 shows the magnitude of N(X) underestimation as a function of N(X). Although the effects on nitrogen are smaller and practically negligible in about half of the cases because of the lower abundance of this species in our sample, a clear increase in the amount of hidden saturation can be seen with increasing N(Fe ii), where the Fe ii λ1144 line is very strong compared to the others and hidden saturation can be easily assessed.

Zoom In Zoom Out Reset image size

The main contributing parameters to hidden saturation are combined in Figure 12, which shows the amount of saturation as a function of N(X), N(H i), and metallicity as inferred from the nebular gas (see Table 1). For Fe ii (top panel of Figure 12) it can be seen that, as expected, the maximum amount of hidden saturation occurs within the galaxies with highest N(Fe ii), metallicity, and N(H i). Although there are galaxies with N(H i) comparable to those showing maximum saturation, the physical amount of Fe ii as assessed from N(Fe ii) in the neutral ISM (and the galaxy metallicity to a lesser extent) is lower, effectively resulting in weaker lines and decreasing the overall effect of hidden saturation. For nitrogen (lower panel of Figure 12), we can see the effects of low N(H i) counteracting the metallicity of the galaxy, where the highest metallicity galaxy displays a negligible amount of saturation, despite having N(N i) comparable to galaxies with log[N(H i)/cm⁻²] > 20.6. This illustrates that it is a combination of the amount of the ion and the amount of the absorbing gas that results in hidden saturation.

Zoom In Zoom Out Reset image size

To conclude, the amount of hidden saturation of a certain ion X from its absorption lines within a spectrum is a function of two main parameters, N(X) and N(H i). For both Fe ii and N i we find that hidden saturation effects occur when log[N(X)/cm⁻²] > 14.5, but only when log[N(H i)/cm⁻²] > 20.6 for Fe ii; for N i the saturation effects are only experienced at higher H i column densities, i.e., when log[N(H i)/cm⁻²] > 21.1, corresponding to abundance limits of [Fe/H] >−6.1 and [N/H] >−6.6.

As a result of these findings, and for the purpose of this paper, column density measurements of ions that have multiple lines within our wavelength range (e.g., Fe ii, N i, S ii, Si ii, Si ii*) were made using all of the "weak" lines that were consistent with the weakest line available and avoiding strong lines that may be affected by hidden saturation. For galaxies where multiple velocity components exist in their absorption lines, we were often unable to isolate the weakest lines because they were merged with velocity components from the stronger lines. For these cases, column densities are listed as lower limits in Table 6.

5.1.2. Classical Saturation and Curve-of-growth Analysis

In order to determine whether our measurements were affected by "classical" saturation, for each ion in each galaxy we checked the column density measurements from the line-profile fitting against the COG suited for that ion. This method is particularly useful for those ions where only one absorption line exists, preventing us from identifying this kind of saturation by simply performing a line-profile fitting. In fact, the fitting always gives the lowest column density possible in the "saturated" COG regime, which despite the same equivalent width may correspond to a wide range of column densities and b parameters. Each ion has a specific COG that expresses how the equivalent width, log(W/λ), changes with column density, log(fN), where f is the oscillator strength for the transition undertaken to produce a photon of wavelength λ. As the strength of the absorption line profile increases, the curve takes on three separate parts: (1) a linear part, where W grows linearly with N and is independent of the Doppler parameter b; (2) a saturated part, where the wings continue to grow deeper and broader but the core gets flatter (W increases much more slowly, as (ln N)^1/2, so it is only weakly dependent on N but is strongly dependent on b); and (3) a square-root part, where the wings grow even deeper and W increases as N^1/2 and is independent of b.

In order to compare our measurements with the best-suited COG, we had to adopt the appropriate b parameter for each ion considered in the spectra of each galaxy. Due to the degeneracy between b and N in the saturated part of the COG (this degeneracy implies that for a certain W value of an absorption line the column density N cannot be unequivocally determined, and only an interval of values can be given, unless the b parameter is known), we could not merely rely on our b measurements from the line-profile fitting. This is particularly true for those cases where only one absorption line was available. We thus tried to estimate the most likely b value based on the following considerations. For each galaxy, the b parameters of all of the ions considered were plotted as a function of atomic mass to assess whether the broadening of the lines was due to thermal rather than turbulent motion of the atoms. In the case of thermal motion, the b parameter would decrease with the atomic mass of the ion, and we would be able to infer the b value specific to the ion under consideration by taking this trend into account. Alternatively, for turbulent motions the b parameters would be relatively constant compared to the atomic mass of the ions, and an average b parameter would suffice. The latter case was found to be true for all of the galaxies within our sample. Note that the reasoning above implies that we fall within the regime where the single-velocity approximation is valid (see Section 3.3).

For each species in each target, we generated the appropriate COG by using both the b parameter inferred from the simultaneous line-profile fitting of all of the available lines of that species (even if this value may not be reliable as the line(s) may be saturated) and the average b parameter estimated from the analysis of the trends versus atomic mass (different velocity components were analyzed separately, under the assumption that each of them is real and arises from a separate volume of gas along the line of sight). We then placed on this curve each line of that species by taking into account its W and N as inferred from the fitting. Because W is just the integral of the fitted line, it is a good approximation of the real equivalent width. In this way we could easily determine whether or not classical saturation was affecting our measurements, e.g., because all of the lines used were in the saturated part of the COG.

Figure 13 shows a selection of COGs illustrating each of the line strength regimes encountered, each of which is discussed below.

Zoom In Zoom Out Reset image size

5.1.3. Assessing Saturation Effects

Another major concern in abundance determinations from UV absorption-line analyses is represented by ionization and dust-depletion effects. Our approach to each of these issues is described in the following sections.

5.2.1. Ionization Corrections

Ionization effects are usually negligible in studies of the neutral ISM, and abundances of a certain element are inferred from the amount of that element in the primary ionization state at the temperature of the neutral gas. From Galactic interstellar studies it is well known that the singly ionized stage is the dominant one for most elements because their first ionization potential is below 13.6 eV (the H⁰ ionization threshold) and their second one is above it. The neutral stage instead prevails for those elements having the first ionization potential above 13.6 eV. The reason for this is that the bulk of the H i gas with N_H i ≳ 10¹⁹ cm⁻² is self-shielded from hv > 13.6 eV photons but transparent to hv < 13.6 eV photons. This means that C ii, Si ii, P ii, S ii, Fe ii, and Ni ii, as well as O i and N i will be the dominant ionization stages of these elements in the neutral gas of our sample of nearby SFGs. However, our sample also includes a few targets with N_H i ≲ 10¹⁹ cm⁻² for which ionization effects can be noticeable and need to be properly taken into account in order to avoid an underestimate of the real amount of a certain element in the cold gas.

There is also another kind of ionization effect that needs to be considered for our sample. Some of the ions that are dominant ionization states in H i regions may also be produced in photoionized clouds (or H ii regions) where H i is a small fraction of the total hydrogen content. These clouds are likely photoionized by the same UV background sources that we used to detect absorption lines in the neutral ISM of our targets, or they could simply be along the same sight line. The formation of metal absorption lines in both ionized and neutral regions can have a significant impact on element abundance determinations of the neutral ISM, resulting in an overestimate of the real amount of a certain element in the cold gas. The size of the impact depends on the ionizing source and the geometry of the system. For example, the ionization corrections are in general smaller when a stellar spectrum dominates over an external UV background, i.e., for damped Lyα (DLA) systems (see, e.g., Howk & Sembach 1999). Fortunately, this is the case for our sample of SFGs, where the radiation field is predominantly made up of young massive stars within the galaxies themselves.

The ionization correction factors (ICFs) that are due to contaminating ionized gas along the line of sight of the COS targets in our sample are usually larger than the so-called classical ICFs needed to properly take into account all of the ions of a certain element present in the neutral ISM. From now onward we will refer to these two corrections as ICF_ionized and ICF_neutral, respectively, and to the total ionization correction as

$$\begin{equation} {\rm ICF}_{{\rm total}} = {\rm ICF}_{{\rm ionized}} - {\rm ICF}_{{\rm neutral}} \end{equation} \tag{ 2 }$$

(see Table 8 for a list of the adopted values for these corrections in each SFG of our sample). The final column density of each element X, corrected for both ionization effects, is defined as

$$\begin{equation} \log [N(X)_{{\rm ICF}}]=\log [N(X)] - {\rm ICF}_{{\rm total}}. \end{equation} \tag{ 3 }$$

Table 8. Ionization Correction Factors for the Metallicity and H i Column Density of the Galaxies in the Sample

Ionized-gas ICFs, ICFionized
Galaxy	log[N(H i)]	Z/Z☉	H i	C ii	N i	O i	Si ii	P ii	S ii	Fe ii	Ni ii
SBS 0335−052	21.70	0.04	0.00	0.02 ± 0.02	0.00	0.00	0.00	0.01 ± 0.01	0.00	0.00	0.00
I Zw 18	21.28	0.03	0.00	0.05 ± 0.05	0.00	0.00	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.00	0.01 ± 0.01
NGC 5253-1	21.20	0.32	0.00	0.05 ± 0.05	0.00	0.00	0.02 ± 0.02	0.01 ± 0.01	0.01 ± 0.01	0.00	0.01 ± 0.01
NGC 4449	21.14	0.41	0.00	0.06 ± 0.06	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01
NGC 4214	21.12	0.32	0.00	0.06 ± 0.06	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.01 ± 0.01	0.00	0.01 ± 0.01
SBS 1415+437	21.09	0.08	0.00	0.07 ± 0.07	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02	0.00	0.01 ± 0.01
NGC 4670	21.07	0.32	0.00	0.07 ± 0.07	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02	0.01 ± 0.01	0.01 ± 0.01
NGC 5253-2	20.65	0.32	0.00	0.14 ± 0.14	0.00	0.00	0.06 ± 0.06	0.05 ± 0.05	0.04 ± 0.04	0.01 ± 0.01	0.02 ± 0.02
NGC 3690v1	20.62	1.29	0.00	0.13 ± 0.13	0.00	0.00	0.08 ± 0.08	0.04 ± 0.04	0.03 ± 0.03	0.02 ± 0.02	0.02 ± 0.02
NGC 3690v2	19.82	1.29	0.02 ± 0.02	0.41 ± 0.41	0.01 ± 0.01	0.02 ± 0.02	0.29 ± 0.29	0.19 ± 0.19	0.16 ± 0.16	0.12 ± 0.12	0.10 ± 0.10
M83-1	19.60	3.24	0.04 ± 0.04	0.47 ± 0.47	0.03 ± 0.03	0.04 ± 0.04	0.41 ± 0.41	0.26 ± 0.26	0.23 ± 0.23	0.18 ± 0.18	0.12 ± 0.12
M83-2	18.44	3.24	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Neutral-gas ICFs, ICFneutral
Galaxy	log[N(H i)]	Z/Z☉	H ii	C iii	N ii	O ii	Si iii	P iii	S iii	Fe iii	Ni iii
SBS 0335−052	21.70	0.04	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
I Zw 18	21.28	0.03	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NGC 5253-1	21.20	0.32	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NGC 4449	21.14	0.41	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NGC 4214	21.12	0.32	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
SBS 1415+437	21.09	0.08	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NGC 4670	21.07	0.32	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NGC 5253-2	20.65	0.32	0.00	0.00	0.01 ± 0.01	0.00	0.00	0.00	0.00	0.00	0.00
NGC 3690v1	20.62	1.29	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NGC 3690v2	19.82	1.29	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01
M83-1	19.60	3.24	0.03 ± 0.03	0.01 ± 0.01	0.02 ± 0.02	0.01 ± 0.01	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02
M83-2	18.44	3.24	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Total ICFs, ICFtotal
Galaxy	log[N(H i)]	Z/Z☉	H	C	N	O	Si	P	S	Fe	Ni
SBS 0335−052	21.70	0.04	0.00	0.02 ± 0.02	0.00	0.00	0.00	0.01 ± 0.01	0.00	0.00	0.00
I Zw 18	21.28	0.03	0.00	0.05 ± 0.05	0.00	0.00	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01	0.00	0.01 ± 0.01
NGC 5253-1	21.20	0.32	0.00	0.05 ± 0.05	0.00	0.00	0.02 ± 0.02	0.01 ± 0.01	0.01 ± 0.01	0.00	0.01 ± 0.01
NGC 4449	21.14	0.41	0.00	0.06 ± 0.06	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.01 ± 0.01	0.01 ± 0.01	0.01 ± 0.01
NGC 4214	21.12	0.32	0.00	0.06 ± 0.06	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.01 ± 0.01	0.00	0.01 ± 0.01
SBS 1415+437	21.09	0.08	0.00	0.07 ± 0.07	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02	0.00	0.01 ± 0.01
NGC 4670	21.07	0.32	0.00	0.07 ± 0.07	0.00	0.00	0.02 ± 0.02	0.02 ± 0.02	0.02 ± 0.02	0.01 ± 0.01	0.01 ± 0.01
NGC 5253-2	20.65	0.32	0.00	0.14 ± 0.14	-0.01 ± 0.01	0.00	0.06 ± 0.06	0.05 ± 0.05	0.04 ± 0.04	0.01 ± 0.01	0.02 ± 0.02
NGC 3690v1	20.62	1.29	0.00	0.13 ± 0.13	0.00	0.00	0.08 ± 0.08	0.04 ± 0.04	0.03 ± 0.03	0.02 ± 0.02	0.02 ± 0.02
NGC 3690v2	19.82	1.29	0.01 ± 0.02	0.40 ± 0.41	0.00 ± 0.01	0.01 ± 0.02	0.28 ± 0.29	0.18 ± 0.19	0.15 ± 0.16	0.11 ± 0.12	0.09 ± 0.10
M83-1	19.60	3.24	0.01 ± 0.05	0.46 ± 0.47	0.01 ± 0.04	0.03 ± 0.04	0.39 ± 0.41	0.24 ± 0.26	0.21 ± 0.23	0.16 ± 0.18	0.10 ± 0.12
M83-2	18.44	3.24	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

Download table as:  ASCIITypeset image

Column density ratios relative to H and abundances derived from ionization-corrected column densities are listed in Table 9 as log (X/H)_ICF and [X/H]_ICF, respectively.

Table 9. Interstellar Abundances

Element	Ion	log[N(X)/cm−2]	log(X/H)	log(X/H)ICFa	log(X/H)☉b	[X/H]c	[X/H]ICFd
I Zw 18
H	H i	21.28 ± 0.03
C†	C ii	14.41 ± 0.01	−6.87 ± 0.03	−6.92 ± 0.06	−3.57 ± 0.05	−3.30 ± 0.06	−3.35 ± 0.08
N	N i	14.38 ± 0.03	−6.90 ± 0.04	−6.90 ± 0.04	−4.17 ± 0.05	−2.73 ± 0.06	−2.73 ± 0.06
O†	O i	14.80 ± 0.02	−6.48 ± 0.04	−6.48 ± 0.04	−3.31 ± 0.05	−3.17 ± 0.06	−3.17 ± 0.06
Si†	Si ii	14.36 ± 0.01	−6.92 ± 0.03	−6.93 ± 0.03	−4.49 ± 0.03	−2.43 ± 0.04	−2.44 ± 0.04
S	S ii	14.73 ± 0.05	−6.55 ± 0.06	−6.56 ± 0.06	−4.88 ± 0.03	−1.67 ± 0.07	−1.68 ± 0.07
Fe	Fe ii	14.52 ± 0.06	−6.76 ± 0.07	−6.76 ± 0.07	−4.50 ± 0.04	−2.26 ± 0.08	−2.26 ± 0.08
SBS 0335−052
H	H i	21.70 ± 0.05
C†	C ii	14.41 ± 0.01	−7.29 ± 0.05	−7.31 ± 0.05	−3.57 ± 0.05	−3.72 ± 0.07	−3.74 ± 0.07
N	N i	14.83 ± 0.02	−6.87 ± 0.05	−6.87 ± 0.05	−4.17 ± 0.05	−2.70 ± 0.07	−2.70 ± 0.07
O†	O i	14.77 ± 0.01	−6.93 ± 0.05	−6.93 ± 0.05	−3.31 ± 0.05	−3.62 ± 0.07	−3.62 ± 0.07
Si†	Si ii	14.38 ± 0.02	−7.32 ± 0.05	−7.32 ± 0.05	−4.49 ± 0.03	−2.83 ± 0.06	−2.83 ± 0.06
P	P ii	12.97 ± 0.15	−8.73 ± 0.16	−8.74 ± 0.16	−6.59 ± 0.03	−2.14 ± 0.16	−2.15 ± 0.16
S	S ii	14.96 ± 0.03	−6.74 ± 0.06	−6.74 ± 0.06	−4.88 ± 0.03	−1.86 ± 0.07	−1.86 ± 0.07
Fe	Fe ii	15.32 ± 0.04	−6.38 ± 0.06	−6.38 ± 0.06	−4.50 ± 0.04	−1.88 ± 0.07	−1.88 ± 0.07
Ni	Ni ii	13.96 ± 0.06	−7.74 ± 0.08	−7.74 ± 0.08	−5.78 ± 0.04	−1.96 ± 0.09	−1.96 ± 0.09
SBS 1415+437
H	H i	21.09 ± 0.03
C†	C ii	14.51 ± 0.01	−6.58 ± 0.03	−6.65 ± 0.08	−3.57 ± 0.05	−3.01 ± 0.06	−3.08 ± 0.09
N	N i	14.57 ± 0.04	−6.52 ± 0.05	−6.52 ± 0.05	−4.17 ± 0.05	−2.35 ± 0.07	−2.35 ± 0.07
O†	O i	14.87 ± 0.09	−6.22 ± 0.09	−6.22 ± 0.09	−3.31 ± 0.05	−2.91 ± 0.10	−2.91 ± 0.10
Si†	Si ii	14.42 ± 0.01	−6.67 ± 0.03	−6.69 ± 0.04	−4.49 ± 0.03	−2.18 ± 0.04	−2.20 ± 0.05
P	P ii	13.43 ± 0.08	−7.66 ± 0.09	−7.68 ± 0.09	−6.59 ± 0.03	−1.07 ± 0.09	−1.09 ± 0.09
S	S ii	14.96 ± 0.02	−6.13 ± 0.04	−6.15 ± 0.04	−4.88 ± 0.03	−1.25 ± 0.05	−1.27 ± 0.05
Fe†	Fe ii	14.83 ± 0.04	−6.26 ± 0.05	−6.26 ± 0.05	−4.50 ± 0.04	−1.76 ± 0.06	−1.76 ± 0.06
Ni	Ni ii	13.26 ± 0.10	−7.83 ± 0.10	−7.84 ± 0.10	−5.78 ± 0.04	−2.05 ± 0.11	−2.06 ± 0.11
NGC 4214
H	H i	21.12 ± 0.03
N	N i	15.20 ± 0.08	−5.92 ± 0.09	−5.92 ± 0.09	−4.17 ± 0.05	−1.75 ± 0.10	−1.75 ± 0.10
O†	O i	15.25 ± 0.02	−5.87 ± 0.04	−5.87 ± 0.04	−3.31 ± 0.05	−2.56 ± 0.06	−2.56 ± 0.06
Si†	Si ii	14.87 ± 0.02	−6.25 ± 0.04	−6.27 ± 0.04	−4.49 ± 0.03	−1.76 ± 0.05	−1.78 ± 0.05
P	P ii	13.76 ± 0.03	−7.36 ± 0.04	−7.38 ± 0.04	−6.59 ± 0.03	−0.77 ± 0.05	−0.79 ± 0.05
S	S ii	15.56 ± 0.03	−5.56 ± 0.04	−5.57 ± 0.04	−4.88 ± 0.03	−0.68 ± 0.05	−0.69 ± 0.05
Fe	Fe ii	14.80 ± 0.07	−6.32 ± 0.08	−6.32 ± 0.08	−4.50 ± 0.04	−1.82 ± 0.09	−1.82 ± 0.09
Ni	Ni ii	13.48 ± 0.04	−7.64 ± 0.05	−7.65 ± 0.05	−5.78 ± 0.04	−1.86 ± 0.06	−1.87 ± 0.06
NGC 5253-Pos. 1
H	H i	21.20 ± 0.01
C†	C ii	15.04 ± 0.01	−6.16 ± 0.01	−6.21 ± 0.05	−3.57 ± 0.05	−2.59 ± 0.05	−2.64 ± 0.07
N	N i	15.24 ± 0.03	−5.96 ± 0.03	−5.96 ± 0.03	−4.17 ± 0.05	−1.79 ± 0.06	−1.79 ± 0.06
O†	O i	15.44 ± 0.14	−5.76 ± 0.14	−5.76 ± 0.14	−3.31 ± 0.05	−2.45 ± 0.15	−2.45 ± 0.15
Si†	Si ii	14.79 ± 0.02	−6.41 ± 0.02	−6.43 ± 0.03	−4.49 ± 0.03	−1.92 ± 0.04	−1.94 ± 0.04
P	P ii	13.74 ± 0.08	−7.46 ± 0.08	−7.47 ± 0.08	−6.59 ± 0.03	−0.87 ± 0.09	−0.88 ± 0.09
S	S ii	15.44 ± 0.02	−5.76 ± 0.02	−5.77 ± 0.02	−4.88 ± 0.03	−0.88 ± 0.04	−0.89 ± 0.04
Fe	Fe ii	15.67 ± 0.07	−5.53 ± 0.07	−5.53 ± 0.07	−4.50 ± 0.04	−1.03 ± 0.08	−1.03 ± 0.08
Ni	Ni ii	13.68 ± 0.08	−7.52 ± 0.08	−7.53 ± 0.08	−5.78 ± 0.04	−1.74 ± 0.09	−1.75 ± 0.09
NGC 5253-Pos. 2
H	H i	20.65 ± 0.05
C†	C ii	14.92 ± 0.01	−5.73 ± 0.05	−5.87 ± 0.15	−3.57 ± 0.05	−2.16 ± 0.07	−2.30 ± 0.16
N	N i	14.98 ± 0.02	−5.67 ± 0.05	−5.66 ± 0.05	−4.17 ± 0.05	−1.50 ± 0.07	−1.49 ± 0.07
O†	O i	15.31 ± 0.05	−5.34 ± 0.07	−5.34 ± 0.07	−3.31 ± 0.05	−2.03 ± 0.09	−2.03 ± 0.09
Si†	Si ii	14.81 ± 0.01	−5.84 ± 0.05	−5.90 ± 0.08	−4.49 ± 0.03	−1.35 ± 0.06	−1.41 ± 0.09
P	P ii	13.65 ± 0.11	−7.00 ± 0.12	−7.05 ± 0.13	−6.59 ± 0.03	−0.41 ± 0.12	−0.46 ± 0.13
S	S ii	15.37 ± 0.02	−5.28 ± 0.05	−5.32 ± 0.06	−4.88 ± 0.03	−0.40 ± 0.06	−0.44 ± 0.07
Fe	Fe ii	15.66 ± 0.04	−4.99 ± 0.06	−5.00 ± 0.06	−4.50 ± 0.04	−0.49 ± 0.08	−0.50 ± 0.08
Ni	Ni ii	13.77 ± 0.13	−6.88 ± 0.14	−6.90 ± 0.14	−5.78 ± 0.04	−1.10 ± 0.15	−1.12 ± 0.15
NGC 4670
H	H i	21.07 ± 0.08
C†	C ii	15.02 ± 0.01	−6.05 ± 0.08	−6.12 ± 0.11	−3.57 ± 0.05	−2.48 ± 0.09	−2.55 ± 0.12
N	N i	15.42 ± 0.01	−5.65 ± 0.08	−5.65 ± 0.08	−4.17 ± 0.05	−1.48 ± 0.09	−1.48 ± 0.09
O†	O i	15.26 ± 0.02	−5.81 ± 0.08	−5.81 ± 0.08	−3.31 ± 0.05	−2.50 ± 0.09	−2.50 ± 0.09
Si†	Si ii	14.97 ± 0.01	−6.10 ± 0.08	−6.12 ± 0.08	−4.49 ± 0.03	−1.61 ± 0.09	−1.63 ± 0.09
P	P ii	13.82 ± 0.05	−7.25 ± 0.09	−7.27 ± 0.09	−6.59 ± 0.03	−0.66 ± 0.09	−0.68 ± 0.09
S†	S ii	15.60 ± 0.01	−5.47 ± 0.08	−5.49 ± 0.08	−4.88 ± 0.03	−0.59 ± 0.09	−0.61 ± 0.09
Fe	Fe ii	15.07 ± 0.05	−6.00 ± 0.09	−6.01 ± 0.09	−4.50 ± 0.04	−1.50 ± 0.10	−1.51 ± 0.10
Ni	Ni ii	13.76 ± 0.05	−7.31 ± 0.09	−7.32 ± 0.09	−5.78 ± 0.04	−1.53 ± 0.10	−1.54 ± 0.10
NGC 4449
H	H i	21.14 ± 0.03
O†	O i	15.54 ± 0.02	−5.60 ± 0.04	−5.60 ± 0.04	−3.31 ± 0.05	−2.29 ± 0.06	−2.29 ± 0.06
Si†	Si ii	15.25 ± 0.10	−5.89 ± 0.10	−5.91 ± 0.10	−4.49 ± 0.03	−1.40 ± 0.10	−1.42 ± 0.10
P	P ii	14.11 ± 0.09	−7.03 ± 0.09	−7.05 ± 0.09	−6.59 ± 0.03	−0.44 ± 0.09	−0.46 ± 0.09
S	S ii	15.61 ± 0.09	−5.53 ± 0.09	−5.54 ± 0.09	−4.88 ± 0.03	−0.65 ± 0.09	−0.66 ± 0.09
Fe†	Fe ii	15.04 ± 0.06	−6.10 ± 0.07	−6.11 ± 0.07	−4.50 ± 0.04	−1.60 ± 0.08	−1.61 ± 0.08
NGC 3690v1
H	H i	20.62 ± 0.02
N	N i	15.11 ± 0.02	−5.51 ± 0.03	−5.51 ± 0.03	−4.17 ± 0.05	−1.34 ± 0.06	−1.34 ± 0.06
O†	O i	15.48 ± 0.02	−5.14 ± 0.03	−5.14 ± 0.03	−3.31 ± 0.05	−1.83 ± 0.06	−1.83 ± 0.06
Si†	Si ii	15.09 ± 0.02	−5.53 ± 0.03	−5.61 ± 0.09	−4.49 ± 0.03	−1.04 ± 0.04	−1.12 ± 0.09
S	S ii	15.42 ± 0.04	−5.20 ± 0.04	−5.23 ± 0.05	−4.88 ± 0.03	−0.32 ± 0.05	−0.35 ± 0.06
Fe	Fe ii	14.84 ± 0.03	−5.78 ± 0.04	−5.80 ± 0.04	−4.50 ± 0.04	−1.28 ± 0.06	−1.30 ± 0.06
NGC 3690v2
H	H i	19.82 ± 0.08
N	N i	13.63 ± 0.71	−6.19 ± 0.71	−6.18 ± 0.71	−4.17 ± 0.05	−2.02 ± 0.71	−2.01 ± 0.71
O†	O i	14.85 ± 0.05	−4.97 ± 0.09	−4.97 ± 0.09	−3.31 ± 0.05	−1.66 ± 0.10	−1.66 ± 0.10
Si†	Si ii	14.33 ± 0.09	−5.49 ± 0.12	−5.76 ± 0.31	−4.49 ± 0.03	−1.00 ± 0.12	−1.27 ± 0.31
Fe	Fe ii	14.64 ± 0.04	−5.18 ± 0.09	−5.28 ± 0.15	−4.50 ± 0.04	−0.68 ± 0.10	−0.78 ± 0.15
M83-Pos. 1
H	H i	19.60 ± 0.32
N†	N i	15.29 ± 0.02	−4.31 ± 0.32	−4.31 ± 0.33	−4.17 ± 0.05	−0.14 ± 0.32	−0.14 ± 0.33
Si†	Si ii	15.05 ± 0.04	−4.55 ± 0.32	−4.93 ± 0.52	−4.49 ± 0.03	−0.06 ± 0.32	−0.44 ± 0.52
P	P ii	14.08 ± 0.02	−5.52 ± 0.32	−5.75 ± 0.41	−6.59 ± 0.03	+1.07 ± 0.32	+0.84 ± 0.41
S	S ii	15.60 ± 0.09	−4.00 ± 0.33	−4.20 ± 0.41	−4.88 ± 0.03	+0.88 ± 0.33	+0.68 ± 0.41
Fe†	Fe ii	14.83 ± 0.03	−4.77 ± 0.32	−4.92 ± 0.37	−4.50 ± 0.04	−0.27 ± 0.32	−0.42 ± 0.37
Ni	Ni ii	13.89 ± 0.06	−5.71 ± 0.33	−5.80 ± 0.35	−5.78 ± 0.04	+0.07 ± 0.33	−0.02 ± 0.35
M83-Pos. 2
H	H i	18.44 ± 0.30
C†	C ii	15.09 ± 0.01	−3.35 ± 0.30	NaN	−3.57 ± 0.05	+0.22 ± 0.30	NaN
N	N i	14.90 ± 0.03	−3.54 ± 0.30	NaN	−4.17 ± 0.05	+0.63 ± 0.30	NaN
Si†	Si ii	14.69 ± 0.01	−3.75 ± 0.30	NaN	−4.49 ± 0.03	+0.74 ± 0.30	NaN
P	P ii	13.73 ± 0.04	−4.71 ± 0.30	NaN	−6.59 ± 0.03	+1.88 ± 0.30	NaN
S	S ii	15.39 ± 0.02	−3.05 ± 0.30	NaN	−4.88 ± 0.03	+1.83 ± 0.30	NaN
Fe†	Fe ii	14.71 ± 0.02	−3.73 ± 0.30	NaN	−4.50 ± 0.04	+0.77 ± 0.30	NaN

Notes. ^aIonization-corrected ratios in logarithmic scale, calculated after applying the ionization correction factors listed in Table 8 with Equation (3). ^bSolar photospheric abundances from Asplund et al. (2009). ^c[X/H] = log(X/H)−log(X/H)_☉. ^d[X/H]_ICF = log(X/H)_ICF−log(X/H)_☉.

Download table as:  ASCIITypeset images: 1 2

5.2.1.1. Ionization Corrections due to Contaminating Ionized Gas.

To estimate the magnitude of ionized gas contributing to our absorption spectra, we used the cloudy photoionization code (Ferland et al. 1998) to model the SFGs as spherically symmetric gas surrounding single ionizing sources (the real geometry of each system may be more complex than this simple assumption). Our aim is to model both the ionized and neutral gas in order to understand the ionization structure along a single line of sight toward the galaxy. Models were run over a grid of metallicities (0.02, 0.5, 1.0, and 4.0 Z/Z_☉) and neutral hydrogen column densities (log[N(H i)/cm⁻²] = 18.2, 18.7, 19.1, 19.7, 20.2, 20.7, 21.1, and 21.7) that encompass the range of metallicities and neutral hydrogen column densities measured within our sample. For each combination of Z and N(H i) over the grid, we modeled the galaxy as a spherical volume of constant-density gas over a large range of volume densities (−6 < log[n(H)/cm⁻³] < 6). The gas is ionized by a central star at a certain distance from the cloud with an effective temperature of 36,000 K (the average estimated temperature for the stellar population targeted in our sample; see Section 4) and a UV luminosity of log[L_⋆] = 40.05 erg s⁻¹ (derived from the average number of counts measured within the COS aperture on the ACS/SBC images; see Figure 1). The magnitude and shape of the ionizing spectrum was found to have a minimal effect on the relative mixture of neutral and ionized gas within our models when compared to the effects of differing metallicity and column density of neutral hydrogen. Models were also run for a variety of radii from the central ionizing source (0.1–100 pc). Significant differences were only seen when the radii were large enough for the cloud to be considered as a plane-parallel slab, i.e., consisting of only neutral gas and therefore not applicable to the galaxies within our sample. A radius of 10 pc was ultimately selected for our modeling.

The modeling process consisted of two phases. In the first phase, for each volume density considered and for each combination of Z and N(H i) over the grid, the models are run and stopped once the desired H i column density is reached, at which point we output the simulated ion column densities of the sphere. By stopping the models at the desired N(H i) (rather than a fixed radius) we ensure that they properly account for the correct amount of neutral gas along the line of sight, including all neutral gas outside the ionized gas volume.

In order to proceed with the second phase, one needs to pinpoint the appropriate volume density of the gas. This is routinely done by using the ratio of successive ion stages. One of the best indicators in this sense is the Si iii/Si ii ratio. However, this ratio cannot be used for the galaxies in our sample because of saturation effects affecting both Si ii and Si iii determinations. Another indicator of the gas volume density is the Fe iii/Fe ii ratio, but because of the low redshift of our galaxies, the Fe iii λ1122 absorption line is only available in the spectrum of SBS 0335-052, giving in this case a ratio of log[N(Fe iii)/N(Fe ii)] = −1.08 ± 0.08. In light of the above, we decided to take a conservative approach. We assumed the maximum Fe iii/Fe ii ratio possible as determined by the preliminary modeling in the first phase and run models that represent the worst-case scenario for a maximum amount of ionized gas along the line of sight for each combination of N(H i) and metallicity. The results of this modeling are the column density of each element associated with both the ionized and neutral gas, i.e., N(X)>_H i+H ii, and the determination of the location of the photodissociation region (PDR), the latter obtained by plotting the ionization stages as a function of depth through the cloud and assessing the depth at which the ionic fraction of H ii is equal to that of H i.

The second phase of the modeling process consisted of running the worst-case scenario model again using the same input parameters, but stopping the depth of the model at the location of the PDR for each combination of N(H i) and metallicity. This allowed us to estimate for each element the column density that is associated with the ionized gas only, i.e., N(X)_H ii. Using Equation (4), we can then assess the magnitude of the column densities arising from the neutral gas only, N(X)_H i:

$$\begin{equation} N(X)_{{\rm H\,\scriptsize{I}}} = {N(X)_{{\rm H\,\scriptsize{I}+H\,\scriptsize{II}}}}-{N(X)_{{\rm H\,\scriptsize{II}}}}. \end{equation} \tag{ 4 }$$

The relative amount in a logarithmic scale of the column density of the total (neutral+ionized) gas, N(X)_H i+H ii, (as seen from the outer edge of the gas cloud) compared to the column density of the neutral gas only, N(X)_H i, gives the ionization correction factor ICF_ionized. This is the amount of column density N(X)_H ii that is within the ionized gas as seen from the edge of the PDR:

$$\begin{equation} {\rm ICF}(X)_{{\rm ionized}} = \log [N(X)_{{\rm H\,\scriptsize{I}+H\,\scriptsize{II}}}] - \log [N(X)_{{\rm H\,\scriptsize{I}}}]. \end{equation} \tag{ 5 }$$

The ionized-gas ICFs obtained in this way are upper limits associated to the "maximum" Fe iii/Fe ii ratio possible as suggested by the modeling itself. In order to estimate the final ICFs, we averaged these upper limits with what would be the best-case scenario of no ionization along the line of sight (no ICF corrections). Figure 14 shows the trends of these "average" ICF_ionized values as a function of metallicity and for specific values of log[N(H i)]. Only the dominant ions within the neutral gas phase are considered in these plots. By interpolating between the grid of average ICFs in metallicity and N(H i) space, we derived the ICFs specific to each galaxy. In this case, the errors on the final ICFs are given by the difference between the average and the maximum ionization ICFs and are equal to the average. The final ionized-gas ICFs and associated errors adopted in our analysis are listed in the first part of Table 8.

Zoom In Zoom Out Reset image size

Our models show that the amount of correction required is small and comparable to the 1σ errors for galaxies with log[N(H i)/cm⁻²]  ≳ 21 if Z/Z_☉ ≲ 0.5, a finding in agreement with models of metal-poor DLAs by Cooke et al. (2011). This correction becomes more important as the column density gets lower and the metallicity higher, i.e., for log[N(H i)/cm⁻²] ≲ 21 and Z/Z_☉ ≳ 0.5, to the point that for the lowest H i column densities considered, our worst-case scenario models predict that only ionized gas remains within the system. This is why the models at log[N(H i)/cm⁻²] = 18.2 and 18.7 are not plotted in Figure 14 and no ionized-gas ICFs are given for position 2 of M83 in Table 8. It should be noted, however, that for this sight line we observe Lyα and many other neutral species in absorption in the COS spectrum, so the lack of neutral gas in the worst-case scenario model is just the result of our cautious approach.

There are two main caveats to the approach presented here. First, the models assume that the same abundances apply to both the ionized and neutral gas. Second, the models assume that the solar abundance ratios apply. Both of these assumptions are routinely made in these types of studies and, despite the complexity of our models, we are aware that they may not be fully representative of the real situation, which may be far more complex. These two assumptions may not hold if, for example, chemical inhomogeneities exist where the galaxy halo has a different chemical imprint due to inefficient mixing.

Our models are analogous to those adopted in the study of DLAs with the exception that in the latter case the ionized gas (clearly present in our study because the background source is a hot UV-bright target within our SFGs) is not considered during the modeling phase. This is because the DLAs are considered as a plane-parallel slab of purely neutral gas illuminated by a background source (e.g., QSO). The existence and amount of ionized gas (including ICF corrections) is then inferred by comparing the model input abundances to the observed abundances.

An alternative method to estimate the amount of ionized gas contaminating the absorption spectra is to assess the ionization fractions within the ionized gas, i.e., X^j:Xⁱ (where i > j) for each element X with Xⁱ only present in the ionized gas phase and X^j present in both the neutral and ionized gas phases, and use N(Xⁱ) to calculate the magnitude N(X^j) contaminating the neutral gas column densities (see Sembach et al. 2000). However, in our case we do not have any element where we detect the ions at the higher ionization state needed for this kind of correction (except for Fe iii and Fe ii in SBS 0335-052). This is why we adopted the approach of using cloudy modeling to assess the ionization corrections needed for all of the ions measured in our spectra. Despite the differences in approach, both methods yield similar answers. For example, Sembach et al. (2000) predicts that the N(S ii)_H ii/N(S ii)_H i ratio can be 0.3–0.8 for log[N(H i)/cm⁻²] = 18.8–19.7, which is not inconsistent with our predictions of ≳0.2 for galaxies within a similar N(H i) range. It should be noted that all of our galaxies within this N(H i) range have supersolar metallicities (i.e., Z/Z_☉ > 1), whereas the models of Sembach et al. (2000) are for Z/Z_☉ ∼ 0.3.

5.2.1.2. "Classical" Ionization Corrections in the Neutral Gas.

The cloudy models were also used to check the magnitude of the "classical" ionization effects due to the lack of sampling of absorption lines relative to higher ionization stages of a certain element in the neutral gas. This was accomplished by looking at the ratio of ions in the radial plots of the models and by checking the relative column densities of a higher ionization state compared to the dominant lower ionization state of a certain element in the neutral gas only. Because the total amount of a certain element X is given by the following equation,

$$\begin{equation} N(X) = N(X^i)+N(X^{i+1}), \end{equation} \tag{ 6 }$$

where N(Xⁱ) is the dominant stage of ionization in the neutral gas for element X, the ionization correction is defined as the following in a logarithmic scale:

$$\begin{equation} {\rm ICF}(X)_{{\rm neutral}}=\log \left[1+\frac{N(X^{i+1})}{N(X^i)}\right]. \end{equation} \tag{ 7 }$$

Similar to the ionization corrections due to contaminating gas along the line of sight, we run the worst-case scenario models corresponding to the maximum amount of ionized gas for each combination of metallicity and N(H i) and averaged these upper limits with the values from the best-case scenario of no ionization (i.e., no ICF corrections). In Figure 15, we plot these average ICF_neutral values as a function of metallicity and for specific values of log[N(H i)]. Only the higher ionization states of the dominant ions within the neutral gas phase are considered in these plots. By interpolating between the grid of average ICFs in metallicity and N(H i) space, neutral-gas ICFs specific to each galaxy were derived. In this case, the errors on the final ICFs are given by the difference between the "average" and the "maximum" ionization ICFs and are equal to the average. The final neutral-gas ICFs and associated errors adopted in our analysis are listed in the second part of Table 8. We find that no corrections are needed for the majority of the targets, except for those where log [N(H i)/cm⁻²]  ≲  20. The PDR is defined as the point at which the ionic fraction of H i is the same as the ionic fraction of H ii, and at this point onward all of the photons able to photoionize, e.g., Fe ii, S ii, or Si ii, have already been absorbed, so the transition from, e.g., Si iii to Si ii occurs just before the PDR, leaving negligible column densities of Si iii within the neutral gas. As for the ionized-gas ICFs, because the models at log[N(H i)/cm⁻²] = 18.2 and 18.7 are fully ionized, they are not plotted in Figure 15, and no neutral-gas ICFs are given for position 2 of M83 in Table 8.

Zoom In Zoom Out Reset image size

5.2.2. Dust Depletion

Within our spectra, we observe a selection of fine-structure lines, which arise from fine-structure to ground-state transitions (such as O i^*, Si ii^*, and C ii^*). The excitation of fine-structure levels can occur via three main mechanisms: collisions with electrons, radiative pumping by the local radiation field, or excitation by photons from cosmic microwave background radiation (although the latter mechanism is not so important for the Si⁺ fine-structure levels because they are too far apart; Silva & Viegas 2002). Much attention has been paid to these transitions because ratios with their ground-state counterparts (e.g., N(Si ii^*)/N(Si ii)) can be used to determine the physical conditions of the gas from which the lines originate (Bahcall & Wolf 1968; Silva & Viegas 2002; Savaglio & Fall 2004; Berger et al. 2005; Howk et al. 2005; Jenkins & Tripp 2011). According to Bahcall & Wolf (1968), if observations suggest that the higher fine-structure states of a ground-state multiplet are occupied, then there are two conditions that the gas will meet. Either the density of the absorbing material is large enough that rates for collisional excitation of fine-structure levels are greater than the photon decay rates, or the absorbing material is being subjected to a strong photon flux that populates the higher fine-structure states. The presence of such lines is therefore an important indicator of the physical conditions of the gas. Here we describe the information gained from three such lines that could be studied within our COS spectra: C ii^*, O i^*, and Si ii^*.

The O i^* λ1304.85 absorption line originates from the middle ground-state fine-structure level of O i, connected to the bottom fine-structure level by the 63.2 μm [O i]  line. To get the middle level populated enough to see this line requires fairly dense, relatively warm (100–200 K) gas (e.g., Morton 1975). Due to its blend with Si ii λ1304.37, it is important to determine if O i^* is present because it must be taken into consideration when constraining the Si ii line. One way to do this is to assess the presence of O i^** λ1306.03, which is connected to the O i^* line. The O i^** originates from the topmost of the three ground-state, fine-structure levels of O i and is connected to the middle fine-structure line by the 146 μm [O i]  line. The additional excitation to get level three populated is not much more than that required to get level two populated, from which the O i^* line originates, and thus the two lines are often seen as a pair. For example, in Dinerstein (2006) both O i^* and O i^** are seen in absorption in a photodissociated region around a planetary nebula. Similarly, Jenkins & Tripp (2011, their Appendix B) see both lines toward the star HD 210839 and use them to deduce a gas temperature of ∼400 K. On inspection of our data we see no evidence of O i^** λ1306, which lies in an uncontaminated region of the spectra in all cases. In addition to this, if we were to attribute the absorption around 1304 Å to this line, the line would be systematically offset from the other absorption lines by ∼100–150 km s⁻¹. The lack of O i^* and O i^** is to be expected because they are rarely detected in Galactic ISM sight lines. In particular, ruling out the presence of the O i^* line is extremely important. This line, if present, would contaminate Si ii λ1304, the weakest Si ii line within our wavelength range and thus the most suited to better constrain an upper limit for the Si ii column density, considering that all of the other Si ii lines are much stronger and suffer even more from saturation effects.

The Si ii^* is detected in all of our targets, apart from the three most metal-poor targets (I Zw 18, SBS 0335−052, SBS 1415+437) and NGC 3690, where it cannot be constrained due to contamination. Although its detection is no surprise, having been detected in varying strengths in UV spectra of gamma-ray bursts (GRBs; Savaglio & Fall 2004), DLAs (e.g., Howk et al. 2005), and Lyman-break galaxies (e.g., Pettini et al. 2002), its strength compared to Si ii is of utmost importance. The Si ii^* absorption is associated with the local environment of the starburst because the presence of this state requires large densities or intense IR/UV radiation fields, which are not common in interstellar environments (Berger et al. 2005). For example, ionized gas along the line of sight of QSOs has been associated to a relative strength of the Si ii^* and Si ii column densities of −2.36 < N(Si ii^*)/N(Si ii) <−1.58 for a covering factor in the range 0.3–1.0 (Srianand & Petitjean 2001). A value of N(Si ii^*)/N(Si ii) ∼ −1.7 has instead been associated with high-density neutral gas in a GRB afterglow (Vreeswijk et al. 2004). Only an upper limit to the relative strength of the Si ii^* and Si ii column densities can be inferred for each galaxy in our sample because of the lower limit values affecting the determination of N(Si ii) because of saturation effects in the Si ii lines. No conclusive results can thus be drawn related to the real physical conditions of the gas from which Si ii^* originates in our case. An upper limit in the range between −2.28 and −0.87 was estimated for the N(Si ii^*)/N(Si ii) in our sample, with an average upper limit of ∼1.5.

Unlike Si ii^*, C ii^* may arise in a cold (T ∼ 100 K) or warm (T ∼ 8000 K) neutral medium or in the ionized gas (T ∼ 10, 000 K) (Wolfe et al. 2003; Srianand et al. 2005). This is because the upper fine-structure levels in Si ii and C ii have excitation energies that differ by a factor of four. As with Si ii, the ratio of the excited- to ground-state column densities of C⁺ can also be used to probe the physical properties of the absorbing gas. However, the measurement of N(C ii) in all of the galaxies of our sample is not possible because C ii λ1344 is highly saturated, deeming the N(C ii^*)/N(C ii) ratio unusable for the purpose of probing the real conditions of the gas from which C ii^* originates.

With a metallicity of ∼1/30, I Zw 18 is an object well known as having the lowest metal content measured in the local universe. The system comprises two separate stellar components that share the same H i envelope, the so-called main body and secondary body (Zwicky 1966). The main body itself consists of two star-forming regions, with a brighter northwest component (containing the bright UV source targeted within this study) separated from a fainter southeast component by ∼8''. Because of its distance of ∼18 Mpc, much debate has surrounded I Zw 18 regarding its true age, with ages <40 Myr being predicted because of its record-low oxygen abundance (Izotov & Thuan 1999). However, the clearest insight into the age of I Zw 18 has been through HST-resolved stellar population studies. The most recent HST/ACS imaging by Aloisi et al. (2007) confirmed the presence of an RGB with stars older than ∼1 Gyr and thus ruled out the possibility that I Zw 18 is a truly primordial galaxy formed recently in the local universe.

The object I Zw 18 has been studied extensively in the UV, first using GHRS (Kunth et al. 1994) and more recently using FUSE (Aloisi et al. 2003). Both of the previous UV studies considered a foreground H i column density in the direction of I Zw 18 due to the MW of log[N(H i) cm⁻²] = 20.30 (Stark et al. 1992). In addition to the interstellar absorptions at the systemic velocity of I Zw 18 of ∼760 km s⁻¹, both studies witnessed absorptions at velocities around ∼−160 km s⁻¹ corresponding to a H i HVC in the direction of I Zw 18 detected by Hulsbosch & Wakker (1988), Stark et al. (1992), and Hartmann (1994). Both studies assumed a H i column density of log[N(H i) cm⁻²] = 19.32 for this cloud. Figure 16 (top panel) shows the Lyα absorption-line profile for I Zw 18. Although we consider the same column density adopted in previous studies for the HVC component, here we adopt a MW column density of log[N(H i) cm⁻²] = 20.49 based on the latest H i Leiden/Dwingeloo survey (Hartmann & Burton 1997). Keeping the column density of the MW profile fixed, we estimate a column density of log[N(H i) cm⁻²] = 21.28 ± 0.03 for the H i within I Zw 18.

Zoom In Zoom Out Reset image size

A selection of the metal lines seen within the COS spectrum of I Zw 18 are shown below the Lyα profile in Figure 16. Because of the low metallicity of this object, some species that are common to other galaxies within the sample (such as P ii, Mn ii, and Ni ii) were not present. However, we were able to constrain all other high S/N lines, each best fitted with a single velocity component profile at an average redshift of z_abs = 0.0025464 (v ∼ 760 km s⁻¹). This is in agreement with the radial velocity of the stellar population (as measured from the C iii λ1176 photospheric absorption line), which also lies at v ∼ 760 km s⁻¹. The higher ionization transition Si iii λ1206, which originates predominantly in the ionized gas, was detected in I Zw 18 with two separate velocity components. The first velocity component (the only one listed in Table 6) aligns with that of the ions arising from the neutral ISM (e.g., O i and N i), suggesting a mix of ionized and neutral gas within the same absorbing cloud. Because Si ii^* was not detected within the spectrum, we are not able to better constrain the physical conditions of this more highly ionized gas along the sight line. The second velocity component of Si iii at a similar b parameter and slightly lower column density is instead blueshifted by ∼160 km s⁻¹, suggesting an outflow of ionized gas or an additional H ii region along the sight line to I Zw 18. All b parameters from the measured absorption line profiles are in good agreement with one another (∼80–90 km s⁻¹), suggesting that turbulent motion is dominating within this gas compared to thermal broadening. Some amount of hidden saturation was affecting the strongest Fe ii λ1144 line, and one of the weakest lines (Fe ii λ1143) was used to constrain the ion column density, which turned out to be higher by 0.29 dex (Table 7). Because of geocoronal contamination in the region of O i λ1302 and Si ii λ1304, it was necessary to fit these two lines in the "night" spectra (see Figure 17).

Zoom In Zoom Out Reset image size

Comparing our results with those reported by Lebouteiller et al. (2013) and obtained from our (HST GO program 11579; PI: A. Aloisi) and additional (HST GTO programs 11523 and 12028; PI: J. Green) COS observations, we find that on average our column densities are in agreement with those published in the above-mentioned paper within the uncertainties (e.g., Si ii, S ii, C ii^*, N i, and Fe ii), except for those ions that have only one saturated line in our COS spectra and for which we have only estimated a lower limit (i.e., C ii and O i).

Along with I Zw 18, SBS 0335−052 is known as one of the most metal-poor galaxies in the local universe. Object SBS 0335−052 was found by Pustilnik et al. (2001) to consist of two prominent H i peaks separated in the east–west direction by 84'' (22 kpc at an assumed distance of ∼54 Mpc), a separation large enough for them to be treated as two distinct BCDs: SBS 0335−052E (considered here and simply called SBS 0335−052 hereafter for simplicity) and SBS 0335−052W. The combination of the low metallicity of SBS 0335−052 (∼1/25 Z_☉) and an undetectable underlying old stellar population from HST imaging suggests that this may be a young galaxy undergoing its first burst of SF (Thuan et al. 1996). However, at the distance of this galaxy, the detection of an RGB is almost impossible with current imaging capabilities, and thus its "young" status remains unconfirmed.

The relatively high redshift of SBS 0335−052 (z  =  0.0134685 or v ∼ 4038 km s⁻¹, the highest value among the galaxies covered within this study) allowed for the separation of the MW and galactic Lyα absorption-line profiles, as shown in the top panel of Figure 18. We estimate log[N(H i) cm⁻²] = 21.70 ± 0.05 for H i within SBS 0335−052 and log[N(H i)_MW cm⁻²] = 20.4 ± 0.1 for the MW H i contribution along the same sight line. The H i envelope of SBS 0335−052 has been previously studied using the large (30'' × 30'') FUSE aperture by Thuan et al. (2005), who found a slightly larger H i column density of log[N(H i) cm⁻²] = $21.86^{+0.08}_{-0.05}$. The velocity of the stellar population, ∼4070 km s⁻¹ (as measured from the C iii λ1176 photospheric absorption line), is slightly redshifted from that of the H i gas by ∼30 km s⁻¹, implying that the gas is outflowing compared to the UV background source.

Zoom In Zoom Out Reset image size

A selection of metal lines seen within the COS spectrum of SBS 0335−052 is shown below the Lyα profile in Figure 18. Each absorption line was best fitted with a single velocity component profile (overlaid). The average redshift of all absorption profiles is z_abs = 0.0134907 (v ∼ 4044 km s⁻¹). This was sufficient enough to shift the O i λ1302 and Si ii λ1304 absorption lines out of the geocoronal contamination, and we were therefore able to use the original 'day+night" spectrum to fit these lines. We were able to constrain all of the expected ions (i.e., those seen within the other spectra in this study) with the exception of Si ii^* (its strongest line Si ii^* λ1264 was redshifted into the spectral break, whereas the next available strongest line Si ii^* λ1194 was not detected). All b parameters are in relatively good agreement with each other, at b ∼ 30–40 km s⁻¹. The Si iii line, which originates predominantly in ionized gas, displays a slightly higher than average b parameter at ∼55 km s⁻¹, along with a velocity blueshifted by ∼55 km s⁻¹, which suggests that the ionized gas cloud is separate from the neutral ISM. Because of the large redshift of this target, Fe iii λ1122 is visible within our COS wavelength range. As with Si iii its velocity is blueshifted with respect to velocities of the species that reside in the neutral ISM. A considerable amount of hidden saturation (Section 5.1.1) was present within this spectrum for both the Fe ii and N i lines. In both cases, the weakest lines available were used to constrain the ion column densities, which amounted to a column density higher by 0.97 and 0.25 dex in the case of Fe ii and N i, respectively (Figure 9, Table 7).

Comparing our results with those reported by Thuan et al. (2005) from FUSE observations of the ions in common, we find that on average our column densities are in agreement with the FUSE values within the uncertainties (e.g., N i, P ii, and Fe ii), except for those ions that have only one saturated line in our COS spectra and for which we can only estimate a lower limit (i.e., Si ii, C ii, and O i).

SBS 1415+437 is classified as a metal-deficient, BCD galaxy, with an oxygen metallicity of ∼1/12 Z_☉. Initially thought to have formed stars only ≲ 100 Myr ago (Thuan et al. 1999b), SBS 1415+437 was considered as one of the best candidate primeval galaxies in the local universe. However, as with several other local metal-poor galaxies, deep HST/ACS imaging was able to reveal a population of much older stars, in this case older than ∼1.3 Gyr (Aloisi et al. 2005).

With a redshift sufficiently high enough to clearly separate the centroid of its Lyα absorption profile from that of the MW, we estimate log[N(H i) cm⁻²] = 21.09 ± 0.03 for the H i within SBS 1415+437 along with a MW component of log[N(H i_MW) cm⁻²] = 20.29 in its blue wing (see top panel of Figure 19). The MW N(H i) component was adopted from the latest H i Leiden/Dwingeloo survey (Hartmann & Burton 1997) and kept fixed during the fitting of the Lyα absorption line. The velocity of the stellar population (∼610 km s⁻¹ as measured from the C iii λ1176 photospheric absorption line) is slightly redshifted compared to that of the H i gas by ∼30 km s⁻¹.

Zoom In Zoom Out Reset image size

In the lower panels of Figure 19, we present a selection of the metal lines observed within the COS spectrum of SBS 1415+437. The lines were best fitted with a single velocity component at an average redshift of z_abs = 0.0019377 (v ∼ 580 km s⁻¹). This average excludes ions that typically originate from ionized gas, such as Si iii, which is blueshifted by ∼50 km s⁻¹ relative to the neutral ISM. The C ii^* is also blueshifted by ∼40 km s⁻¹, suggesting that in this case it is associated with the ionized more than with the neutral gas. Rather than the Doppler width b scaling inversely with ion mass, the b parameters of all ions are mostly in agreement (within errors) around ∼60–70 km s⁻¹. This suggests that turbulent broadening, rather than thermal broadening, is in effect within the neutral gas of SBS 1415+437. Similar to SBS 0335−052 (and many other galaxies within this sample), hidden saturation (Section 5.1.1) was present in the Fe ii λ1144 and N i λ1134.9 absorption lines, which gave rise to column densities that were lower by 0.50 and 0.28 dex, respectively, compared to the column densities inferred when only using the weakest Fe ii and N i lines (Table 7). Unlike most other galaxies within our sample (but similar to I Zw 18, another low-metallicity galaxy), the strongest Si ii^* line within our wavelength range, Si ii^* λ1264, was not present. Because of geocoronal contamination in the region of O i λ1302 and Si ii λ1304, it was necessary to fit these two lines in the "night" spectra (see Figure 20).

Zoom In Zoom Out Reset image size

The galaxy NGC 4214 is a nearby (∼3 Mpc) dwarf-barred, irregular galaxy that consists of two main star-forming complexes containing hundreds of O stars as well as a superstar cluster. It has a metallicity of 1/3 Z_☉, placing it in the center of the metallicity scale covered within this study, along with NGC 5253 and NGC 4670. This galaxy has been studied extensively in the last several years, mainly owing to its nearby proximity, which allows HST to spatially resolve individual stars within it. Morphologically, NGC 4214 harbors multiple sites of active SF along a central bar-like structure, with bright stellar clusters and large cavities blown clear of gas by stellar winds (as revealed by recent HST/WFC3 images by R. O'Connell and the WFC3 Scientific Oversight Committee), surrounded by a large disk of H i gas (Allsopp 1979).

The Lyα absorption-line region of NGC 4214 is shown in the top panel of Figure 21. The overlaid best-fit profile has a measured column density of log[N(H i) cm⁻²] = 21.12 ± 0.03 combined with a MW column density of log[N(H i_MW) cm⁻²] = 20.08. The latter was adopted from the latest H i Leiden/Dwingeloo survey (Hartmann & Burton 1997) and was kept fixed during the fitting. The previously mentioned strong stellar winds within this galaxy are evident in the red and blue wings of the Lyα absorption (e.g., the large N v P-Cygni profile at 1243 Å). The stellar population velocity (as measured from the C iii λ1176 stellar absorption line) of ∼370 km s⁻¹ is redshifted by 100 km s⁻¹ relative to the H i gas. However, because the C iii stellar line has a P-Cygni profile, this velocity offset should be regarded as a lower limit.

Zoom In Zoom Out Reset image size

A selection of the metal lines within the NGC 4214 spectrum is shown below the Lyα profile in Figure 21. Similar to all the other galaxies within our sample, hidden saturation was present for the Fe ii λ1144 absorption line and amounted to 0.19 dex compared to the other two weaker Fe ii λ1143 and Fe ii λ1142 lines that were fitted together. On the other hand, N i λ1134.9 did not show any hidden saturation and we were able to fit all the lines of the N i triplet around 1134 Å together. In this galaxy we were unable to constrain the fit to the C ii λ1134 profile due to blending with the C ii absorption from the MW. Because of geocoronal contamination around 1302 Å, the O i λ1302 line was fitted on a "night" spectrum and can be seen in Figure 22. Si ii λ1304 was unaffected by geocoronal contamination and was fitted on the "day+night" spectrum.

Zoom In Zoom Out Reset image size

The ions fitted within this spectrum can be placed into three different groups based on their average redshift and b parameters. First, a number of ions were best fitted using a two-component line structure (N i, O i, Si ii, S ii, and Fe ii), the first with an average redshift of z_{abs, 1} = 0.0009808 (v₁ ∼ 294 km s⁻¹) and the second with z_{abs, 2} = 0.0008936 (v₂ ∼ 268 km s⁻¹) blueshifted compared to the first one. Although the radial velocities differ by only ∼25 km s⁻¹, the two velocity components have very different b parameters with b₁ = 30–40 km s⁻¹ and b₂ = 80–90 km s⁻¹, suggesting that the two clouds of gas along the line of sight have quite different physical properties. Second, for the weaker absorption lines (P ii and Ni ii), we were able only to resolve a single velocity component. Based on the average redshift of these lines (z_abs = 0.0010380, v ∼ 311 km s⁻¹) and relatively low b parameter of ∼40 km s⁻¹, we may assume that we detected these ions only in the gas associated with the first high-redshift velocity component. The third and final group refers to the higher-ionization ions from nebular gas (i.e., Si ii^* and Si iii), which have an average redshift of z_abs = 0.0007989 (v ∼ 240 km s⁻¹) and a b parameter of ∼90 km s⁻¹. These ions are blueshifted by an additional 30 km s⁻¹ from the second velocity component of the lower ionization ions while sharing similar kinematic properties (i.e., b parameter). This may mean that the second cloud of gas (v₂), while containing only low-ionization species, may be more aligned with regions of active SF, where a larger turbulence-driven b parameter is a result of stellar winds and outflows.

The C ii^* ion has also two velocity components, one associated with the first component of the neutral gas with a redshift of z_abs = 0.0010080 (v ∼ 302 km s⁻¹) and a similar b parameter of ∼30 km s⁻¹, and the other with a redshift of z_abs = 0.0008039 (v ∼ 241 km s⁻¹) and a b parameter of ∼100 km s⁻¹ associated instead with the more highly ionized blueshifted gas.

Because we were unable to resolve the H i line profile into two velocity components, all abundances listed in Table 9 are calculated from the total column density of the two velocity components, where applicable.

The object NGC 5253 is a nearby starburst galaxy, at a distance of ∼3.8 Mpc and a metallicity of ∼1/3 Z_☉. This starburst galaxy is famous for hosting the first case of observed localized enrichment in a H ii region, with Welch (1970), Walsh & Roy (1987, 1989), and Kobulnicky et al. (1997) each reporting the presence of a strong nitrogen overabundance in the starbursting nucleus of the galaxy. The presence of Wolf–Rayet (WR) stars within the nucleus is firmly established and thought to be the source of the nitrogen-enriched ionized gas (López-Sánchez et al. 2007, and references therein).

We obtained spectra from two separate pointings within NGC 5253, both located in the nucleus of the galaxy and separated by ∼71 (a projected distance of ∼130 pc). The spectra from the two pointings are overlaid in Figure 23. The presence of a stellar population with strong winds (e.g., O and WR stars) can be clearly seen in the spectra of both pointings in the form of strong P-Cygni profiles at ∼1248 Å. Although there is relatively no difference in the velocity of the absorption lines, there is a difference in the strength of the profiles. This is most clearly seen in the Lyα absorption profiles, where we estimate log[N(H i) cm⁻²] = 21.20 ± 0.01 for position 1 (top panel of Figure 24) and log[N(H i) cm⁻²] = 20.68 ± 0.01 for position 2 (top panel of Figure 25). A MW column density of log[N(H i_MW) cm⁻²] = 20.68 was adopted in both cases from the latest H i Leiden/Dwingeloo survey (Hartmann & Burton 1997) and was kept fixed during the fitting.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

NGC 5253 Position 1. In Figure 24 we present a selection of the lines observed within the COS spectrum of position 1 in NGC 5253. All lines were best fitted with a single velocity component at an average redshift of z_abs = 0.0013217 (v ∼ 400 km s⁻¹). This average redshift is in agreement with the velocity of the stellar population that also lies at v ∼ 400 km s⁻¹ (as measured from the C iii λ1176 photospheric absorption line). As is the case for previously discussed galaxies, the b parameters of the different species do not scale inversely with the mass of the ion, suggesting that turbulent broadening is at play here rather than thermal broadening. Indeed, all ions have a b parameter that is relatively constant around ∼80–100 km s⁻¹. The more highly ionized species Si ii^* and Si iii are also in agreement with the lower ionization species, indicating a mix of neutral and ionized gas along the sight line. A large amount of hidden saturation was present in both the Fe ii λ1144 and N i λ1134.9 absorption profiles (Section 5.1.1) because an increase in column density of 0.92 and 0.46 dex was found for Fe ii and N i, respectively, when fitting the weakest lines (Table 7). The O i λ1302 and Si ii λ1304 absorption lines are shown in Figure 26(a) and were fitted on "night" spectra due to geocoronal contamination within the 1304 Å region.

Zoom In Zoom Out Reset image size

NGC 5253 Position 2. The main absorption lines observed within the second NGC 5253 pointing are presented in Figure 25. All lines were best fitted with a single velocity component at an average redshift of z_abs = 0.0014118 (v ∼ 420 km s⁻¹), which is in agreement with the radial velocity of the stellar population (∼420 km s⁻¹, as measured from the C iii λ1176 photospheric absorption line). The b parameters are reasonably constant around 80–90 km s⁻¹ with a relatively large scatter around the mean mainly due to the large errors incurred from fits to the weakest lines, such as Ni ii and P ii. The line Si iii λ1206, which is thought to originate predominantly in ionized gas, is in line with the lower ionization species, indicating a mix of neutral and ionized gas along the sight line. The line Si ii^* (also typically found in ionized gas) has instead a lower b parameter of ∼40 km s⁻¹, suggesting that we only detect this species in some of the strongest unresolved velocity components that contribute to Si iii. As with position 1, a ∼0.96 dex increase in column density was found when measuring one of weakest Fe ii lines rather than the strongest Fe ii line. However, no hidden saturation was found in N i λ1134.9, and all lines within the N i triplet were fitted together satisfactorily. The O i λ1302 and Si ii λ1304 absorption lines are shown in Figure 26(b) and were fitted on "night" spectra because of geocoronal contamination around 1304 Å.

NGC 4670 is an amorphous, metal-deficient (Z ∼ 1/3 Z_☉), BCD galaxy. It has often been compared to NGC 4214 because it displays a very similar optical spectrum suggestive of vigorous SF, with the young stars superposed on a much older stellar population (Kinney et al. 1993).

The upper panel of Figure 27 displays the Lyα absorption region of NGC 4670, and, similar to NGC 4214 and NGC 5253 discussed above, large wind profiles can be seen in the red wing of Lyα. A best-fit profile is overlaid, corresponding to log[N(H i) cm⁻²] = 21.07 ± 0.08, which was constrained by the red wing of the profile, along with a MW component of log[N(H i) cm⁻²] = 20.25 ± 0.12 constrained by the blue wing of the profile and in excellent agreement with log[N(H i) cm⁻²] = 20.24 as measured in the latest H i Leiden/Dwingeloo survey (Hartmann & Burton 1997).

Zoom In Zoom Out Reset image size

The heavy-element absorption lines (a selection of which are shown in Figure 27) were all fitted with a single velocity component at an average redshift of z_abs = 0.0036046 (v ∼ 1080 km s⁻¹). The velocity of the stellar component (as measured from the C iii λ1176 photospheric absorption line) is blueshifted by ∼35 km s⁻¹ relative to the H i gas. There is no overall decrease in b parameter with ionic mass, with all ions displaying a b parameter around ∼70–80 km s⁻¹, which suggests that turbulent broadening (rather than thermal broadening) is occurring within the neutral gas. The line Si iii λ1206, which is thought to arise in the ionized gas, shows signs of a two-component velocity profile; however, a satisfactory fit to the profile could only be obtained with one component and by fixing the redshift z and the b parameter from the fit to Si ii λ1304. The line Si ii^*, also arising from ionized gas, is in line with the lower ionization species once the large error on its b parameter is taken into account, indicating a mix of neutral and ionized gas along the sight line. A relatively small amount of hidden saturation was seen within the Fe ii triplet, with a 0.22 dex increase in N(Fe ii) when using the weakest lines (Section 5.1.1). No hidden saturation was seen from the strongest line within the N i 1134 triplet. Due to geocoronal contamination around 1302 Å, a "night" spectrum was used to fit the O i λ1302 line, as shown in Figure 28. Because of the high redshift of this target, the Si ii λ1304 was not affected by geocoronal contamination and was fitted in the "day+night" spectrum instead.

Zoom In Zoom Out Reset image size

The object NGC 4449 is a barred, Magellanic-type irregular galaxy (de Vaucouleurs et al. 1991) with SF occurring throughout the galaxy at a rate almost twice that of the Large Magellanic Cloud (Thronson et al. 1987; Hunter et al. 1999). It is a highly studied galaxy, mainly due to its nearby proximity (∼3.8 Mpc), which provides excellent spatial resolution for extragalactic star cluster studies (see Reines et al. 2008, and references therein) and resolved stellar population studies (see, e.g., Annibali et al. 2008). The H i studies of NGC 4449 reveal a rather perturbed structure, with extended streamers wrapping around the galaxy and counter-rotating (inner and outer) gas systems (see, e.g., Hunter et al. 1998). Its H ii region metallicity of ∼1/2 Z_☉ places it in the mid- to high-metallicity range of galaxies covered within this study.

The top panel of Figure 29 displays the Lyα absorption region of NGC 4449. A best-fit profile is overlaid, corresponding to log[N(H i) cm⁻²] = 21.14 ± 0.03 combined with an absorption from within the MW of log[N(H i) cm⁻²] = 20.21. Because the two absorption profiles are heavily blended with one another, the NGC 4449 H i component was fitted by fixing the MW component at the value measured in the latest H i Leiden/Dwingeloo survey (Hartmann & Burton 1997) while allowing the profile fit parameters for the galaxy to vary freely.

Zoom In Zoom Out Reset image size

The heavy-element absorption lines in NGC 4449 (a selection of which are shown in Figure 29) were fitted with up to three velocity components. In O i, Si ii, and Fe ii, three velocity components were visible, at average redshifts of 0.0001732 (v₁ ∼ 52 km s⁻¹), 0.0005096 (v₂ ∼ 153 km s⁻¹), and 0.0010903 (v₃ ∼ 327 km s⁻¹), respectively. For P ii the lowest velocity component was clearly separated, whereas the two highest velocity components were blended into one, at z = 0.0007982 (v_{2 + 3} ∼ 240 km s⁻¹), probably because of the faintness of the line profiles. In contrast, for S ii only the two highest velocity components were present, likely because of the faintness of its lines. The b parameters are somewhat less clearly defined between the three absorbing components; the lowest velocity component has b₁ ∼ 40–50 km s⁻¹, whereas the two higher velocity components are more in agreement with each other, having b₂ ∼ 90–110 km s⁻¹ and b₃ ∼ 120–140 km s⁻¹. The line Si ii^* (which is thought to arise predominantly from ionized gas) also has a two-component velocity structure, with one component aligned with v₂ and a second slightly redshifted by ∼35 km s⁻¹ relative to v₃ but with a consistent b parameter. The radial velocity of the stellar component is in perfect agreement with v_{2 + 3}, lying at v ∼ 240 km s⁻¹ (as measured from the C iii λ1176 photospheric absorption line). The multicomponent velocity structure observed in our COS data may be the result of the above-mentioned complex gas system seen in the H i imaging studies. There is no overall decrease in b parameter with ionic mass, which suggests that turbulent broadening is occurring within the neutral gas.

We were unable to constrain a N i column density for this galaxy because the two N i triplets (at ∼1134 Å and ∼1200 Å) suffer from blending due to the galaxy's multiple velocity components and the MW N i absorption profiles. For the same reasons, we were unable to constrain profile fits for C ii, C ii^*, and Si iii. Another consequence of this multivelocity component structure is that we were unable to constrain N(Fe ii) using the weaker λλ1142, 1143 lines because of blending. Therefore, the Fe ii column density and abundances are obtained from the Fe ii λ1144 line and are deemed lower limits because of hidden saturation (Section 5.1.1). Despite the mid- to high metallicity of NGC 4449 within our COS sample, Ni ii was not strong enough to be constrained above the noise of the spectrum. There was no geocoronal contamination experienced during the observation of NGC 4449, and thus measurements of the O i λ1302 and Si ii λ1304 absorption profiles were made on the original "day+night" spectrum. Because we were unable to constrain multiple velocity components within the Lyα profile, all abundances were calculated from the total column densities of all velocity components.

Within our sample, NGC 3690 is the galaxy with the second highest metallicity (Z ∼ Z_☉) and the second highest spectroscopic redshift (see below). It is known as an extremely disturbed system that is interacting or merging with the companion galaxy IC 694 (Gehrz et al. 1983) located in the eastern part of the system. Additional interest in this object arises from the fact that it is one of the most extreme known cases of an extended burst of SF (Augarde & Lequeux 1985). The H i studies of this system reveal that there is no disk-like structure but instead an amorphous collection of bright star-forming regions and dust lanes (Nordgren et al. 1997).

The Lyα absorption profile of NGC 3690 is shown in the top panel of Figure 30. A positive consequence of this galaxy's high redshift is that the intrinsic Lyα is completely separated from that of the MW and is therefore not contaminated by it (however, the downside is that its COS spectrum has the lowest signal to noise of our sample with S/N ∼ 10). We were able to constrain a two-component fit to the Lyα profile using in this case the uncontaminated red wing of the absorption line and the redshifts fixed from the average z_{abs, 1} and z_{abs, 2} measured from well-constrained ions, such as O i and Fe ii. The best-fit profile was found to be log[N(H i) cm⁻²] = 20.62 ± 0.02 at z_{abs, 1} = 0.0106978 (v₁ ∼ 3,210 km s⁻¹) and log[N(H i) cm⁻²] = 19.82 ± 0.08 at z_{abs, 2} = 0.0095165 (v₂ ∼ 2,850 km s⁻¹). The radial velocity of the stellar population (as measured from the C iii λ1176 photospheric absorption line) lies at ∼3,050 km s⁻¹, which is the approximate midpoint between the two velocity components of the ISM. Because NGC 3690 is the only target where we were able to separate two H i velocity components in the Lyα profile, we calculated separate abundances for those ions where the column density of the two components was measured independently.

Zoom In Zoom Out Reset image size

Unfortunately, because of the low S/N of this spectrum and the blending of its multiple velocity components, we were unable to constrain several of the ions available in other COS spectra within our sample, such as C ii, C ii^*, Ni ii, P ii, and Si ii^*. In addition, the Si iii λ1206 absorption line could not be constrained because it was redshifted into the MW Lyα absorption-line profile. A consequence of the high redshift presented by NGC 3690 is that geocoronal contamination did not affect the O i λ1302 and Si ii λ1304 lines, allowing the original "day+night" spectrum to be used for their fitting. In the lower panels of Figure 30, we show a selection of the heavy-element ions that we were able to constrain within the spectrum of NGC 3690. With the exclusion of S ii, all of the ions displayed a clear two-component velocity structure in alignment with the Lyα profile velocities: one at an average redshift of z_{abs, 1} = 0.0106992 (v₁ ∼ 3,210 km s⁻¹) and another at z_{abs, 2} = 0.0094576 (v₂ ∼ 2,835 km s⁻¹). This velocity structure could be a signature of NGC 3690 interaction with its neighboring galaxy IC 694.

A number of assumptions were made during the fitting process for those ions that we were able to constrain. These assumptions are as follows. Component 2 (v₂) of O i λ1302 Å and component 1 (v₁) of Si ii λ1304 Å are well isolated and were thus used to get independent measurements of the b parameter and redshift of the individual v₁ and v₂ components. Component 1 (v₁) of O i λ1302 Å and component 2 (v₂) of Si ii λ1304 Å are instead merged together and were thus constrained using the redshift and b parameter determined from their corresponding isolated components. Component 1 in the couple of lines fitted for S ii and in Fe ii λ1144 is well isolated and was fitted independently, giving b parameters and redshift values that agree within the errors with those found for component 1 of Si ii and O i. This suggests that the gas from which v₁ originates is dominated by turbulent motions (i.e., the b parameters are independent of ionic mass). We therefore decided to assume the same for v₂ and to use the average b parameters from the separate components (b₁ ∼ 104 km s⁻¹ and b₂ ∼ 138 km s⁻¹) to constrain both N i components and component 2 of Fe ii. We were unable to detect component 2 for S ii, likely because a much lower column density made this component too faint to be detected.

Unfortunately, the complicated velocity structure described above prevented us from constraining N(Fe ii) from the weaker Fe ii lines alone, and we were instead forced to also use Fe ii λ1144, which is known to suffer from hidden saturation (Section 5.1.1). However, because the fit of this line for both velocity components was done in conjunction with the weaker Fe ii λ1143 absorption and both lines agree quite well with the same fit, hidden saturation should not be at play.

The object M83 is described in the literature as being a remarkable grand-design spiral with Hubble type SAB(s)c and strong dust lanes that are offset along an oval distortion that contains a central starburst (Elmegreen et al. 1998 and references within). It is from within this central starburst that our COS spectra were obtained, using two centrally located UV sources separated by approximately 65 (∼150 pc, see Figure 1). The galaxy M83 is the highest metallicity galaxy within our sample, with an oxygen abundance Z ∼ 3 Z_☉. Inclusion of this object within our sample meant we could fully probe depletion patterns at the higher end of the metallicity scale.

The COS spectra from the two pointings are shown overlaid in Figure 31, where the spectrum of position 1 has been scaled up to that of position 2. This scaling allows us to highlight the relatively similar H i column density between the two positions and also a similar redshift for the Lyα and metal absorption lines. However, there is a velocity offset in the wind lines within the two spectra (e.g., C iii λ1176 and Si iv λ1402), suggesting that although the absorbing gas clouds may be spatially associated with each other, the stellar populations of the two UV background sources are not directly related to each other. In particular, the centroid of the C iii λ1176 photospheric line within each spectrum shows that the UV sources are offset from one another by ∼270 km s⁻¹. The two targeted sources are within the eye of the central spiral (our two pointings correspond to "hot spots" numbers 3 and 8 within Elmegreen et al. 1998) at different points of the inner arm, where a differential rotation would be expected. The expanse of the wind lines within the M83 spectra are greater than within any other galaxy in our sample. The P-Cygni profiles are visible in C iii λ1176, N vλ1245, and Si iv λ1402.

Zoom In Zoom Out Reset image size

M83 Position 1. In the top panel of Figure 32, we present the Lyα region in the COS spectrum of position 1 of M83 and a best-fit profile corresponding to an M83 component of log[N(H i) cm⁻²] = 19.60 ± 0.32 and a MW component of log[N(H i) cm⁻²] = 20.57 ± 0.05. The profile was fitted by fixing the MW component at the fit parameters found for position 2 of M83 (see below) and by allowing the parameters of the galaxy component to vary freely. A strong P-Cygni profile from Si iii λ1206 exists within the Lyα profile. This is attributable to winds from B0-1 supergiants (Prinja et al. 2002). However, the population of B0-1 stars is not large enough to contribute to the Lyα absorption-line profile, as detailed in Section 4.

Zoom In Zoom Out Reset image size

In the lower panels of Figure 32 we present a selection of metal absorption lines and their profile fits. All metal absorption lines within the spectrum were fitted with two velocity components at average redshifts of z_{abs, 1} = 0.0016155 (v₁ ∼ 480 km s⁻¹) and z_{abs, 2} = 0.0011107 (v₂ ∼ 330 km s⁻¹). The velocity of the stellar population (as measured by the C iii λ1176 photospheric line) lies at v ∼ 150 km s⁻¹, i.e., a blueshift of up to ∼330 km s⁻¹ relative to the ISM. In Figure 32 absorption profiles are displayed relative to z_{abs, 2}. The b parameters for the two components are rather different, with b₁ ∼ 60–70 km s⁻¹ and b₂ ∼ 90–100 km s⁻¹, suggesting that the two components of absorbing gas are physically rather different. The lines Si iii and Si ii^*, which are thought to arise in more highly ionized gas, also show a two-velocity component structure that is blueshifted relative to v₁ and v₂ by ∼70 and ∼25 km s⁻¹, suggesting that these species are indeed situated in separate clouds of higher-ionization absorbing gas (as also indicated by slightly different b parameters).

As with other spectra within this sample that contain multivelocity components, we were unable to constrain a number of lines because of blending (namely C ii, C ii^*, and O i). The N(Fe ii) line could only be constrained from Fe ii λ1144, which is known to suffer heavily from hidden saturation (see Section 5.1.1). This is why this column density and the inferred abundance are listed as lower limits in Tables 6 and 9. Because of the blending between the two velocity components, only the higher velocity component of N i λ1134.9 was isolated enough to be fitted separately to constrain its parameters. The latter were then kept fixed, together with the redshift of the second component as inferred from an average of other ions, while fitting the second velocity component by using all of the three lines of the N i triplet at ∼1134 Å. Because we were able in this way to nicely fit this triplet, we do not think that hidden saturation is an issue. Because of geocoronal contamination, Si ii λ1304 was fitted on a "night" spectrum and is shown in Figure 34.

M83 Position 2. The top panel of Figure 33 presents the Lyα region in the COS spectrum of position 2 of M83 and a best-fit profile corresponding to an M83 component of log[N(H i) cm⁻²] = 18.44 ± 0.30 and a MW component of log[N(H i) cm⁻²] = 20.57 ± 0.05. The latter value is similar to the MW H i column density of log[N(H i) cm⁻²] = 20.8 as measured in the latest H i Leiden/Dwingeloo survey (see Hartmann & Burton 1997). The two H i components were fitted simultaneously by fixing the MW redshift at 0.0 and the M83 redshift at the value of a well-constrained ion, such as Fe ii. The b parameter of the MW component was also fixed at the average value of 75 km s⁻¹ as inferred from some isolated MW lines (P ii λ1152, the N i triplet at ∼1134, and S ii λ1253). The lower panels of Figure 33 show the main metal absorption lines fitted within the spectrum. In contrast to position 1, the Si iii λ1206 stellar line does not show the P-Cygni profile due to stellar winds in B0-1 stars affecting the Lyα profile fit. The spectrum of this COS target has the lowest H i column density within our sample by almost ∼1 dex. As a result, ICFs for the column densities measured from it were not constrained (see Table 8).

Zoom In Zoom Out Reset image size

Unlike M83 position 1, all lines are best fitted with a single velocity component at z_abs = 0.0015593 (v ∼ 470 km s⁻¹) and with b parameters of ∼100–110 km s⁻¹. The C iii λ1176 photospheric line shows that the velocity of the stellar population lies at ∼420 km s⁻¹, a blueshift of ∼50 km s⁻¹ relative to the H i gas. As a consequence of there being only one velocity component within the absorbing gas, we were able to constrain more elements than in the spectrum of position 1, namely C ii and C ii^*. However, we were still unable to constrain O i because of contamination from Si ii λ1304 from the MW and Ni ii because of noise in that region of the spectrum.

As with M83 position 1, we were unable to use Fe ii λ1142 and λ1143 to constrain N(Fe ii) because of severe blending with Fe ii absorption from the MW, and Fe ii λ1144 was instead used. Because this line is known to suffer from hidden saturation (Section 5.1.1), the column density and abundance calculated from it are listed as lower limits in Tables 6 and 9. With regards to contamination from ionized gas, Si iii and Si ii^* (both thought to arise from more highly ionized gas) were found to have redshifts and b parameters that are in agreement with the other species, suggesting that ionized gas may be present within the absorbing cloud. The Si ii λ1304 absorption line is shown in Figure 34 and was fitted on "night" spectra because of geocoronal contamination within the 1304 Å region.

Zoom In Zoom Out Reset image size