Chemical Abundances in the Leading Arm of the Magellanic Stream^∗

We were awarded six orbits of HST/COS time in Cycle 24 under Program ID 14687 to observe two LA targets: (1) the B5IV star CD14-A05 lying close to region LA I; this star was first identified as a Magellanic System candidate by Casetti-Dinescu et al. (2012), has a photospheric metallicity [Mg/H] = −0.57 ±0.35, and a radial velocity v_r = 133 ± 9 km s⁻¹ (Zhang et al. 2017); (2) the QSO UVQS J1016-3150 at z = 0.2417 (identified as a QSO by Monroe et al. 2016) lying behind LA III. Both of these targets show H i 21 cm detections at LA velocities, making them suitable for abundance analyses. We also searched the HST archive for COS spectra of UV-bright AGN lying either behind or projected near to the LA, and uncovered five more sightlines with H i 21 cm detections from the LA, making a total sample of seven sightlines¹¹ (see Table 1).

The seven directions in our sample are divided into two categories. The first category are the four Confirmed LA Sightlines. These pass through the known H i regions LA I, LA II, or LA III, or (in the case of CD14-A05) are confirmed since the target itself is a member of the LA stellar population (Casetti-Dinescu et al. 2014; Zhang et al. 2017). The second category includes the three Potential LA Sightlines. These are directions that lie within ≈30° of the main LA regions and that show 21 cm H i emission, but have an unconfirmed physical connection to the LA. Two of these sightlines pass through the H i debris field that lies ≈15°–30° northwest of region LA III. The third (PKS 1101-325) lies directly in between LA II and LA III. The HVCs observed in these directions may be physically associated with the LA, and we use their chemical properties and kinematics to explore this possibility.

The location of each direction with respect to the H i emission is shown in Figure 1, where we include both column-density maps and velocity field maps (zeroth and first moment maps). The lack of HST targets at low latitude is due to extinction from the Galactic disk, preventing background UV sources from being detectable, at least with current instrumentation. This precludes the measurement of LA abundances in the latitude range −10° ≲ b ≲ 10°, where the LA crosses (and potentially interacts with) the Galactic disk.

Zoom In Zoom Out Reset image size

For each COS target, we took the extracted one-dimensional spectra from the calcos pipeline (x1d files) and aligned them in wavelength space using customized reduction software that cross-correlates the positions of low-ionization ISM lines (following Wakker et al. 2015). The aligned spectra are co-added and then further calibrated using the velocities of interstellar 21 cm H i components as zero-points. Lines due to intergalactic systems at higher redshift were identified. The data were continuum-normalized around each absorption line of interest. The spectra were binned by five pixels (to 10 km s⁻¹ bins, to give two rebinned pixels per resolution element) for display purposes. A second, night-only reduction was conducted to extract the data taken during orbital nighttime. This reduces geocoronal airglow emission in the range $-100\lesssim {v}_{\mathrm{LSR}}\lesssim 100$ km s⁻¹, which allows us to measure the O i 1302 line crucial to our abundance analysis. The spectral resolution of the COS data is R ≈ 15,000–18,000 (FWHM ≈ 17–20 km s⁻¹), depending on wavelength and detector lifetime position.

A stack of absorption lines for each direction is presented in Figure 2. The suite of lines shown depends on whether the target was observed with both the G130M and G160M gratings (three targets, with coverage from ≈1150 to 1700 Å), or just the G130M grating (four targets, with coverage from ≈1150 to 1450 Å). For the sightlines observed with both gratings, we show O i λ1302 (night-only data), Al ii λ1670, Si ii λλ1260, 1193, 1190, 1526, Si iii λ1206, S ii λλ1250, 1253, Fe ii λλ1144, 1608, C iv λλ1548, 1550, and Si iv λλ1393, 1402. These lines were chosen as the strongest metal lines available in the COS far-ultraviolet (FUV) bandpass. We do not show C ii λ1334 or S ii 1259 since these lines are blended at LA velocities by Galactic C ii* 1335 and Si ii 1260, respectively. Atomic data were taken from Morton (2003) and verified against the recent updates by Cashman et al. (2017).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For the reference H i column densities, we use 21 cm spectra from a variety of telescopes, as summarized in Table 1. For four of our directions, we obtained new high-sensitivity observations from the Robert C. Byrd Green Bank Telescope (GBT). These data were taken under programs GBT12A_206 and GBT17B_424. The spectra were measured over the LSR velocity range −450 and +550 km s⁻¹ at a velocity resolution of 0.15 km s⁻¹ using frequency switching. The spectra were smoothed, converted to brightness temperature, and corrected for stray radiation following the procedures described in Boothroyd et al. (2011). A third-order polynomial was fit to emission-free regions of the spectra to achieve a final rms noise in brightness temperature of ≈11 mK in a 0.6 km s⁻¹ channel. The GBT has an angular resolution of 91 at 1420 MHz.

For the other three directions, we use H i 21 cm spectra from either (1) the Galactic All-Sky Survey (GASS; 161 angular resolution; McClure-Griffiths et al. 2009), (2) the Effelsberg-Bonn H i Survey (EBHIS; 108 resolution; Winkel et al. 2016), or (3) the Australia Telescope Compact Array (ATCA; 1'-resolution; Wakker et al. 2002). The choice of data set adopted in each direction depends on the target declination and availability of data; we present the highest angular resolution spectra available for each target.

We use the known 21 cm velocity field of the LA (shown in Figure 1(a)) to identify LA absorption components. The identification step is important since non-Magellanic absorption components (from foreground or background structures) could be present in the data at other velocities, and without the 21 cm velocity field one could misidentify them as LA components.

The H i velocity field in the LA is complex, extending from below ≈150 to ≈350 km s⁻¹ (Figure 1(a)). The traditionally defined regions LA I, LA II, and LA III lie in the higher-velocity end (≈200–350 km s⁻¹) of this range, though gas at ≈150 km s⁻¹ is found near each LA I, LA II, and LA III. For the seven sightlines in our survey, the LSR velocities of the LA components we identify range from +120 to +289 km s⁻¹. For context, the velocity field for high-positive-velocity gas in this quadrant of the sky is given in Richter et al. (2017). Their Figure 8 shows an extended population of high-positive-velocity UV absorbers (halo clouds) extending to the Galactic northwest beyond the boundaries of the LA, in the direction of Complexes WA and Complex M, but without 21 cm counterparts. Therefore, there are no known 21 cm HVCs other than the LA in this part of the sky.

The two ions we focus on for chemical abundance measurements are S ii and O i, traced via S ii λλ1250, 1253 (the λ1259 region is blended) and O i λ1302. These ions are the best metallicity indicators available in the FUV because sulfur and oxygen are relatively undepleted onto dust grains (Jensen et al. 2005; Jenkins 2009) and have relatively small ionization corrections (ICs; particularly for oxygen; see Section 3.1). They are also both α-elements so they have similar nucleosynthetic origins. Column densities for these two ions were determined via the apparent optical depth (AOD) technique (Savage & Sembach 1991), which returns accurate measures of the column density provided the line profiles are resolved and unsaturated. The AOD in each pixel is given by ${\tau }_{a}(v)=\mathrm{ln}[{F}_{c}(v)/F(v)]$, where F(v) and F_c(v) are the observed flux and the estimated continuum flux, respectively, as a function of velocity. The integrated AOD is ${\tau }_{a}={\int }_{{v}_{\min }}^{{v}_{\max }}{\tau }_{a}(v){dv}$, and the apparent column density is ${N}_{{\rm{a}}}{(v)=3.768\times {10}^{14}(f\lambda )}^{-1}{\tau }_{a}(v)$ cm⁻². The velocity range of absorption, ${v}_{\min }$ to ${v}_{\max }$, is selected by eye to encompass the range of the UV metal-line absorption and the H i emission, though we include a contribution to the error budget to account for varying these limits by ±5 km s⁻¹.

The AOD measurements in each direction are presented in Table 2. In lines that are clearly or potentially saturated (those with normalized flux $F(v)/{F}_{c}(v)\lt 0.1$) we present a lower limit on N_a(v). In lines that are not detected at 3σ significance, we present an upper limit on N(v), based on measuring a 3σ limit on the equivalent width and converting it to column density assuming a linear curve of growth.

Table 2.  Measurements of Leading Arm Absorption and Emission

Target	${v}_{\min }$ a	${v}_{\max }$ a	log N(H i)b	log N(S ii)c	log N(O i)d	[S ii/H i]e	[O i/H i]f
	( km s−1)	( km s−1)	(N in cm−2)	(N in cm−2)	(N in cm−2)
Confirmed LA Sightlines
CD14-A05	75	170	18.83 ± 0.10	14.46 ± 0.22g	14.64 ± 0.10	+0.51 ± 0.26	−0.88 ± 0.17
NGC 3783	160	300	19.92 ± 0.10	14.41 ± 0.12	>14.95h	−0.63 ± 0.16	>−1.66h
NGC 3125	170	245	19.05 ± 0.05	<14.90	14.37 ± 0.28	<+0.73	−1.37 ± 0.30
UVQS J1016-3150	175	255	18.47 ± 0.05	<14.62	13.96 ± 0.31	<+1.03	−1.20 ± 0.33
Potential LA Sightlines
PKS 1101-325	80	150	19.30 ± 0.05	14.52 ± 0.17	>15.04	+0.10 ± 0.20	>−0.95
IRAS F09539-0439	110	250	19.21 ± 0.05	<14.93	14.94 ± 0.09	<+0.60	−0.96 ± 0.14
SDSS J0959+0503	230	330	18.74 ± 0.10	<14.82	<14.54	<+0.96	<−0.89

Notes.

^aMinimum and maximum LSR velocities of LA emission and absorption. ^bH i column is integrated using N(H i) = $1.823\times {10}^{18}\,{\mathrm{cm}}^{-2}{\int }_{{v}_{\min }}^{{v}_{\max }}{T}_{B}\,{dv}$. ^cColumn integrated by the AOD technique (Savage & Sembach 1991). S ii columns derived from the 1253 line. Upper limits are 3σ (nondetections). ^dO i 1302 measurement is made on night-only data, to remove geocoronal emission. Lower limits given for saturated lines. ^e[S ii/H i] = [log N(S ii)–log N(H i)]–(S/H)_⊙. The beam-smearing error of 0.10 dex has been added in quadrature with the statistical error on the S ii and H i column densities. ^f[O i/H i] = [log N(O i)–log N(H i)]–(O/H)_⊙. The beam-smearing error of 0.10 dex has been added in quadrature with the statistical errors. ^gAOD profile analysis indicates that this line may be contaminated by an unidentified blend. ^hUnresolved saturation is possible in O i λ1302, so N(O i) and [O i/H i] are lower limits. See P. Richter et al. (2018, in preparation) for further analysis.

Download table as:  ASCIITypeset image

Due to differences in ionization potentials (IPs) between various ions and H i, ionization effects can lead to differences between the ion abundances and the true elemental abundances. These differences are known as ICs, where

$$\begin{eqnarray*}[{\rm{O}}/{\rm{H}}] & = & [{\rm{O}}\,{\rm{I}}/{\rm{H}}\,{\rm{I}}]+\mathrm{IC}({\rm{O}})\ \mathrm{and}\\ [{\rm{S}}/{\rm{H}}] & = & [{\rm{S}}\,{\rm{II}}/{\rm{H}}\,{\rm{I}}]+\mathrm{IC}({\rm{S}}).\end{eqnarray*}$$

The magnitude of the IC depends on the element. For oxygen, the close similarity in the first IPs of hydrogen (13.60 eV) and oxygen (13.62 eV) coupled with charge-exchange reactions make the ICs very small (Field & Steigman 1971; Viegas 1995), except in cases of intense radiation fields. For sulfur, the (larger) IP of S ii (23.34 eV) means that the ion can exist in regions where the hydrogen is largely ionized, and so the IC is larger.

To calculate the ICs, we computed a grid of photoionization simulations using the Cloudy radiative transfer code (Ferland et al. 2013). The simulations use a plane–parallel geometry and assume that the plasma has uniform density, so they do not account for clumpiness in the gas. For a given set of input parameters, Cloudy calculates the ionization breakdown of all chemical elements up to zinc. The ICs are a function of the ionization parameter $U\equiv {n}_{\gamma }/{n}_{{\rm{H}}}$ (the ratio of the ionizing photon density to the gas density) and the H i column density N(H i) in the LA component. These parameters are constrained as follows:

• (1)  
  To derive n_γ, we use a 3D model of the Galactic ionizing radiation field from Fox et al. (2014), which is based on Bland-Hawthorn & Maloney (1999) and updated by Barger et al. (2013) to include the radiation from the Magellanic Clouds. This is combined with the Haardt & Madau (2001) model of the extragalactic radiation field at z = 0. The LA is taken to be at d = 20 kpc, based on the existing distance constraints toward LA I (Casetti-Dinescu et al. 2014; Zhang et al. 2017), though see Section 3.2 for a discussion of the effect of changing the distance.
• (2)  
  U is derived from the observed Si iii/Si ii column-density ratio in the LA; this ratio is a monotonic function of U for various ionizing radiation fields (Fox et al. 2016; Bordoloi et al. 2017). Knowing U and n_γ allows ${n}_{{\rm{H}}}$ to be calculated. LA directions, where both Si iii λ1206 and one of the Si ii lines (λ1260, 1190, 1193, 1190) are unsaturated (and thus measurable) have the best constrained models. When Si iii is saturated, only a lower limit on the Si iii/Si ii ratio can be measured, which translates to a lower limit on log U and a limit on IC(S). Fortunately, IC(O) is independent of log U when log N(H i) > 18 (Bordoloi et al. 2017), as is the case for all seven sightlines in our sample. Therefore, uncertainty in the knowledge of log U does not affect the reliability of our oxygen abundances.
• (3)  
  N(H i) is directly measured in the 21 cm data by integrating over the velocity range of the LA component.

By running Cloudy models in this manner, we derived the ICs and the corrected sulfur and oxygen abundances presented in Table 3. Notice that for the range of log N(H i) in our LA sample (18.47–19.92), the ICs for oxygen are negligible $(| \mathrm{IC}({\rm{O}})| \lesssim 0.04$ dex) but the ICs for sulfur are substantial (up to −0.90 dex), particularly in the cases where log N(H i) < 19.0. The oxygen abundances should therefore be considered more robust.

Table 3.  Ionization Corrections and Corrected Abundances in the Leading Arm

		Input to Cloudy			Output from Cloudy			Corrected Abundances
Target	Region	log N(H i)	log nγa	log $\tfrac{\mathrm{Si}\,{\rm{III}}}{\mathrm{Si}\,{\rm{II}}}$ b	log Uc	IC(S)d	IC(O)d	[S/H]e	[O/H]e
		(cm−2)	(cm−3)
Confirmed LA Sightlines
CD14-A05	LA I	18.83 ± 0.10	−5.66	>−0.42	>−2.9	<−0.58	−0.043 ± 0.004	<−0.07	−0.90 ± 0.17
NGC 3783	LA II	19.92 ± 0.10	−5.55	−1.95 ± 0.15	−3.4 ± 0.1	+0.003 ± 0.001	−0.006 ± 0.001	−0.62 ± 0.20	>−1.67f
NGC 3125	LA III	19.05 ± 0.05	−5.78	−0.56 ± 0.36	−3.0 ± 0.1	−0.37 ± 0.02	−0.011 ± 0.002	<+0.24	−1.40 ± 0.30
UVQS J1016- 3150	LA III	18.47 ± 0.05	−5.73	−0.16 ± 0.17	−2.7 ± 0.1	−0.90 ± 0.03	−0.014 ± 0.001	<+0.26	−1.21 ± 0.33
Potential LA Sightlines
PKS 1101-325	LA II/III	19.30 ± 0.05	−5.54	>−0.66	>−3.0	<−0.57	−0.014 ± 0.001	<−0.47	>−0.96
IRAS F09539-0439	LA Ext.	19.21 ± 0.05	−5.66	>−0.51	>−2.9	<−0.29	−0.015 ± 0.002	<+0.07	−0.99 ± 0.14
SDSS J0959+0503	LA Ext.	18.74 ± 0.10	−5.48	>−0.83	>−3.2	<−0.57	−0.036 ± 0.002	<+0.41	<−0.93

Notes.

^aLogarithm of density of ionizing photons in this direction at d = 20 kpc given our 3D radiation field model (see Section 3.1). ^bLogarithm of Si iii/Si ii column-density ratio in the LA component, measured using AOD integrations to Si iii λ1206 and Si ii λ1193 over the velocity intervals given in Table 2. Limits are 3σ. These limits propagate to a limit on [S/H]. ^cIonization parameter, $U\equiv {n}_{\gamma }/{n}_{{\rm{H}}}$. ^dIonization corrections for sulfur and oxygen at the inferred value of log U. ^eIonization-corrected sulfur and oxygen abundances, calculated by applying ICs to ion abundances presented in Table 2. Errors include statistical errors on the H i and metal column densities, a beam-smearing error of 0.10 dex, a velocity integration range error, a distance error, and an error on the ionization parameter (see Section 3.2). ^fLower limit given due to saturation in O i λ1302. See P. Richter et al. (2018, in preparation) for further analysis.

Download table as:  ASCIITypeset image

We investigated the uncertainty in the abundance measurements caused by six sources of error: the distance to the LA, the velocity integration range, the beam-size mismatch between radio and UV observations, the error on log U, the statistical error in the measurements of UV absorption, and the statistical error in the measurements of H i emission. The first four of these are now discussed in turn (the last two are self-explanatory). The six sources of error were added in quadrature to produce the final errors on the abundance measurements presented in Table 3. The statistical errors on the UV absorption measurements dominate the errors.

(1) Uncertainty in distance: to investigate the uncertainty in the ICs caused by uncertainty in the distance to the LA, we ran Cloudy models for the NGC 3125 sightline for five heliocentric distances: 10, 20, 30, 40, and 50 kpc, where 20 kpc is the nominal case because of existing distance constraints. This sightline (through LA III) was chosen to be representative of the LA directions, given its intermediate value of log N(H i) = 19.05. All other properties of the models were held constant; increasing the distance decreases the Galactic contribution to the ionizing radiation field, while keeping the UV background contribution fixed. For each distance, we computed the best-fit value of log U, IC(S), and IC(O). We found that IC(S) varies from −0.42 at 10 kpc to −0.35 at 50 kpc, whereas IC(O) remains flat at −0.01 at all distances. The distance error is therefore negligible for the oxygen abundances and reaches −0.07 dex for the sulfur abundances.

(2) Uncertainty in velocity integration range: in each direction, the velocity integration range ${v}_{\min }$ to ${v}_{\max }$ was chosen to encompass the H i emission and the UV absorption from the LA. By varying ${v}_{\min }$ and ${v}_{\max }$ by ±5 km s⁻¹, we quantify the error on N(H i), which propagates directly to the abundance measurements. The magnitude of this error varies from 0.01 to 0.06 dex for our seven sightlines.

(3) Uncertainty due to beam smearing: because of the mismatched beam-size between pencil-beam UV observations and finite-beam radio observations, a beam-smearing error should be included in abundance measurements to account for small-scale structure. In our case, the radio beam-sizes range from 1' (ATCA), 91 (GBT), 108 (EBHIS), to 161 (GASS) The magnitude of the beam-smearing effect on abundance measurements across these angular scales has been shown to be ≈0.10 dex (Wakker et al. 2001; Fox et al. 2010), so we adopt an error on the abundances of 0.10 dex to account for this. Nonetheless, small-scale structure on sub-beam scales is difficult to quantify and cannot be ruled out, particularly given the highly fragmentary nature of the LA emission.

(4) Uncertainty in log U: the error in the value of log U derived from the Si iii/Si ii ratio is ≈0.1 dex. This translates to errors in IC(S) ranging from 0.01 to 0.03, and errors in IC(O) ranging from 0.001 to 0.004 dex. The errors on IC(O) are an order of magnitude smaller since IC(O) is a flat function of log U, whereas IC(S) increases with log U.

Our analysis is framed by the following questions. Did the LA originate in the LMC, SMC, or both (like the MS)? Did it form in a single gas-removal event or in multiple episodes? And was the event or process that created the Stream the same event or process that created the LA? The observed chemical enrichment pattern along the LA directly addresses these questions. In a single removal event, we expect uniform chemical composition along the Stream, since once the gas is stripped from the Magellanic Clouds its chemical abundances are frozen in. Conversely, in a multiple-removal scenario, the gas in different regions would have different metallicities, with the older regions having lower abundances. In this event, we would expect the region furthest from the Clouds (LA III) to have lower metallicity than the region closest to them (LA I). Any metal mixing between the LA and the surrounding medium would complicate this picture, since it could dilute or enrich the LA gas over time (Gritton et al. 2014; Henley et al. 2017) depending on the contrast between the LA and the halo metallicity. This contrast is not well constrained observationally.

With that introduction, our first observation is that all regions of the LA show metallicities below the current-day metallicity of the LMC (46% solar) and the SMC (22% solar; Russell & Dopita 1992; Trundle et al. 2007), with the single exception of the sulfur abundance of 34 ± 7% solar in the high N(H i) core of LA II. Our three LA measurements of O/H range from ${4.0}_{-2.0}^{+4.0} \%$ solar to ${12.6}_{-4.1}^{+6.2} \%$ solar and are therefore well below ${Z}_{\mathrm{LMC}}$ and ${Z}_{\mathrm{SMC}}$. This simple point is worth emphasizing, because it indicates that the LA was formed either (1) long ago (several gigayears ago), when the LMC and SMC abundances were lower according to their age–metallicity relations (Pagel & Tautvaivsienė 1998; Harris & Zaritsky 2004, 2009; Meschin et al. 2014), or (2) from the outer regions of the LMC or SMC where the metallicity is lower than in the inner regions because of radial abundance gradients. Evidence for radial abundance gradients in the Magellanic Clouds based on stellar and nebular metallicities is mixed: Toribio San Cipriano et al. (2017) find no strong gradient in H ii region abundances in either galaxy, and Cioni (2009) find no significant gradient in AGB-star abundances in the SMC. However, Cioni (2009) report a gradient in AGB-star abundances in the LMC of −0.047 ± 0.003 dex kpc⁻¹ out to 8 kpc, and Dobbie et al. (2014) find a gradient of −0.075 ± 0.011 dex deg⁻¹ for red giant stars in the inner 5° of the SMC. Therefore, both the LA age and Magellanic abundance gradients may be factors in explaining the LA's low metallicity.

Our second observation is that there is spatial variation of the oxygen abundances. The two directions through region LA III have identical (low) oxygen abundances, given the observational errors, with [O/H] = −1.40 ± 0.30 toward NGC 3125 and [O/H] = −1.21 ± 0.33 toward UVQS J1016-3150. For region LA II, we only report a limit on [O/H], but P. Richter et al. (2018, in preparation) find a higher value of [O/H] =−0.54 ± 0.23 in this sightline based on the analysis of a high-resolution HST/STIS spectrum. Region LA I also has a slightly higher abundance, [O/H] = −0.90 ± 0.17 (13% solar), based on the measurement toward CD14-A05. This observed abundance gradient, with LA III having lower metallicity than LA I and II, is consistent with LA III being older than LA I and II, which is important considering that LA III lies far away from the Magellanic Clouds on the sky (see Figure 1) and thus has a longer travel time to reach its current location. The gradient supports a multiple-episode formation scenario for the LA.

However, the existence of this gradient should be considered tentative until more data are available, especially considering the statistical uncertainties. Another caveat is that our single abundance measurement in LA I (toward CD14-A05) is derived from a stellar sightline, not an AGN sightline, and we cannot rule out the possibility of some self-enrichment of the LA in this direction by recent star formation. Moreover, some of the H i emission toward CD14-A05 may exist behind the star, whereas the metal absorption lies in front. This possibility, which is suggested by the mismatch in velocity centroid between H i emission and UV metal absorption in this direction (Figure 2), could potentially lead to an underestimate of the LA oxygen abundance, since it would cause an overestimation of the foreground H i column.

Our third observation is that the very low (≈4%–6% solar) oxygen abundances measured in region LA III are too low to support an LMC origin. Even the abundances in regions LA I (≈11% solar) and LA II (≈29% solar) are difficult to reconcile with a recent LMC origin unless the gas was stripped from the outer regions of the LMC. This strongly suggests that LA III, and potentially all the LA, is formed from material from the SMC, not the LMC. This stands in contrast to studies of the LA H i kinematics, which find that the LA arises almost entirely from the LMC (Putman et al. 1998; Nidever et al. 2008; Venzmer et al. 2012), but it supports several tidal and ram-pressure models of MS and LA formation (Gardiner & Noguchi 1996; Connors et al. 2006; Besla et al. 2012; Diaz & Bekki 2012; Yang et al. 2014), which predict that the MS and LA form together from SMC material.

Furthermore, the oxygen abundances in the LA match the α-element abundances measured in most of the trailing MS; oxygen or sulfur abundances of 10% solar or less have been measured in seven directions through the SMC filament of the Stream (Fox et al. 2010, 2013; Kumari et al. 2015; Howk et al. 2017), though there is also an LMC filament with 50% solar abundances (Gibson et al. 2000; Richter et al. 2013). Our LA metallicities can also be compared to those measured in the Magellanic Bridge, the gaseous connection between the LMC and SMC. A Bridge metallicity of 10% solar or lower has been reported in several studies (Lehner et al. 2001; Lehner 2002; Misawa et al. 2009) targeting stars and background quasars. The finding that the Bridge, most of the LA, and most of the Stream have low metallicity (≲10% solar) provides support for an SMC origin for most of the gas in the Magellanic system.

The similarity in oxygen abundances between regions LA III and the SMC filament of the MS (which represents the majority of the MS) supports an origin model in which LA III and the MS were generated in the same event in the past, when the SMC had a lower metallicity. Most tidal and ram-pressure models of the MS date its age to ≈1.5–2.5 Gyr (Moore & Davis 1994; Gardiner & Noguchi 1996; Yoshizawa & Noguchi 2003; Mastropietro et al. 2005; Connors et al. 2006; Besla et al. 2010; Diaz & Bekki 2012) although a direct LMC–SMC collision may have happened more recently (∼100–500 Myr ago; Besla et al. 2012; Hammer et al. 2015). Independently, age–metallicity studies of the stellar population of the SMC find that it had a mean metallicity of 0.1 solar ≈2 Gyr ago (Pagel & Tautvaivsienė 1998; Harris & Zaritsky 2004). These two findings are consistent with the idea that much of the Stream and LA were formed at that time from SMC gas released from an LMC–SMC encounter.

If the LA is at d = 20 kpc, as measured in two separate regions (McClure-Griffiths et al. 2009; Casetti-Dinescu et al. 2014), then it cannot be a purely tidal feature produced at the first passage of the LMC and SMC. The reason is that the Clouds are at ≈50–60 kpc, so in a first-passage scenario (Besla et al. 2007, 2010, 2012) the LA should be close to or beyond that distance, not 20 kpc (e.g., see Figure 10 in Besla et al. 2012). Conversely, in a multiple-passage scenario, the LA may have had time to fall down closer to the MW. Our LA abundance measurements do not resolve this issue but do provide constraints on when the gas was removed from the Clouds, since they indicate that the gas is chemically unenriched.

In addition to tidal forces and ram pressure, another physical process relevant to the formation of the MS and LA may be stellar feedback. The LMC is known to drive an outflow, as seen in UV absorption lines blueshifted with respect to the LMC systemic velocity (Lehner & Howk 2007; Lehner et al. 2009; Barger et al. 2016), although the LMC outflow does not contain enough mass flux to reproduce the mass of the Stream (Barger et al. 2016). In the SMC, evidence for stellar feedback was seen in O vi absorption by Hoopes et al. (2002), who report O vi column densities correlated to proximity to star-forming regions, and O vi kinematics consistent with a galactic fountain. These studies indicate that feedback from star formation drives material out of the Magellanic Clouds into their halos, where ram-pressure and tidal forces may disperse it over larger distances (Olano 2004; Nidever et al. 2008, 2010; Besla et al. 2012). The finding that the LMC filament of the MS can be traced back to a region near 30 Doradus (Nidever et al. 2008), an actively star-forming region, supports this connection between stellar feedback and Stream formation, although the low LA metallicity we measure does not favor an LMC origin. Indeed, the lack of evidence for a half-solar LA filament as a counterpart to the known half-solar MS filament is an important result of our study, since purely tidal models would be expected to produce such a counterpart.

For the stellar component of the LA, chemical abundances were measured by Zhang et al. (2017) for a sample of early B-stars located in regions LA I, II, and III discovered by Casetti-Dinescu et al. (2014). They found that the five kinematical members of the LA have an average Mg abundance of [Mg/H] = −0.42 ± 0.16, close to the LMC abundance, and significantly lower than that that of non-members. However, individually, our target CD14-A05 has [Mg/H] = −0.57 ± 0.35, compatible within errors with the SMC Mg abundance ${[\mathrm{Mg}/{\rm{H}}]}_{\mathrm{SMC}}=-0.8\pm 0.1$ (Trundle et al. 2007), and also compatible with the gas-phase oxygen abundance we measure in the same sightline, [O/H]_gas =−0.90 ± 0.17. This similarity between the stellar and gas-phase metallicity is interesting given the possibility that some of the H i lies behind the star, such that [O/H]_gas is formally a lower limit on the LA oxygen abundance in this direction. Furthermore, the B-type stars (like CD14-A05) may have formed from gas enriched by previous episodes of star formation. More data are needed to explore the relationship between the stellar and gas-phase abundances in the LA in more depth.

The three sightlines in our sample lying away from the main regions of the LA are those toward PKS 1101-325 (between LA II and LA III), IRAS F09539-0439 (≈15° northwest of LA III), and SDSS J0959+0503 (≈30° northwest of LA III). All three directions show HVCs with similar H i columns, with log N(H i) between 18.74 and 19.30. Two of the three show kinematics consistent with those of the main LA regions (the exception is SDSS J0959+0503, where the HVC velocity of 289 km s⁻¹ appears to be too high for an LA origin). The sulfur and oxygen abundances in these HVCs are given in the lower part of Table 3. Toward PKS 1101-325, IRAS F09539-0439, and SDSS J0959+0503, we find [O/H] > −0.96, [O/H] =−0.99 ± 0.14, and [O/H] < −0.93 (3σ), respectively. These abundances are all close enough to our confirmed LA abundances to suggest that the HVCs are outlying fragments of the LA: they have the right kinematics, H i columns, and oxygen abundances to support this idea. In this interpretation the LA extends at least ≈20° further to the northwest (in Galactic coordinates) than the named regions (LA I, II, and III), and the outer fragments trace the full extent of the debris field. The LA then covers a linear extent of ≈80° and an area on the sky of ≈60° × 80°. Nonetheless, this extended LA is still considerably shorter than the trailing Stream, which is ≈140° long (Nidever et al. 2010). This asymmetry is an important constraint for origin models (e.g., Pardy et al. 2018). The presence of fragmented small-scale structure has long been noticed as a key feature of the neutral gas in the Magellanic System, including the LA (Stanimirović et al. 2002; Putman et al. 2003; Westmeier & Koribalski 2008; Venzmer et al. 2012; For et al. 2013). Our new results extend the region over which such small-scale structure is seen.

The finding that the LA is larger than traditionally thought is reproduced in several simulations (e.g., Yang et al. 2014; Hammer et al. 2015). It is also consistent with results from other regions of the Magellanic System, where UV absorption detections are reported in directions up to 30° away from the H i Stream (Fox et al. 2005, 2014; Kumari et al. 2015; Howk et al. 2017). Furthermore, For et al. (2013) report a bridge of LA cloudlets in between LA II and LA III; our HVC at +150 km s⁻¹ toward PKS 1101-325 (which lies halfway between LA II and LA III) may trace this gaseous bridge. The known presence of LA gas in this region supports our interpretation of the +150 km s⁻¹ HVC as tracing the LA.