The COS/UVES Absorption Survey of the Magellanic Stream: II. Evidence for a complex enrichment history of the Stream from the Fairall 9 sightline
Philipp Richter, Andrew J. Fox, Bart P. Wakker, Nicolas Lehner, J. Christopher Howk, Joss Bland-Hawthorn, Nadya Ben Bekhti, Cora Fechner
aa r X i v : . [ a s t r o - ph . GA ] J un Accepted version from May 29, 2013
Preprint typeset using L A TEX style emulateapj v. 5/2/11
THE COS/UVES SURVEY OF THE MAGELLANIC STREAM:II. EVIDENCE FOR A COMPLEX ENRICHMENT HISTORY OF THE STREAM FROM THE FAIRALL 9SIGHTLINE
Philipp Richter , Andrew J. Fox , Bart P. Wakker , Nicolas Lehner , J. Christopher Howk , JossBland-Hawthorn , Nadya Ben Bekhti , Cora Fechner Institut f¨ur Physik und Astronomie, Universit¨at Potsdam, Haus 28, Karl-Liebknecht-Str. 24/25, 14476 Golm (Potsdam), Germany Leibniz-Institut f¨ur Astrophysik Potsdam (AIP), An der Sternwarte 16, 14482 Potsdam, Germany Space Telescope Science Institute, Baltimore, MD 21218, USA Department of Astronomy, University of Wisconsin-Madison, 475 N. Charter Street, Madison, WI 53706, USA Department of Physics, University of Notre Dame, 225 Nieuwland Science Hall, Notre Dame, IN 46556, USA Institute of Astronomy, School of Physics, University of Sydney, NSW 2006, Australia and Argelander-Institut f¨ur Astronomie, Universit¨at Bonn, Auf dem H¨ugel 71, 53121 Bonn, Germany
Accepted version from May 29, 2013
ABSTRACTWe present a multi-wavelength a study of the Magellanic Stream (MS), a massive gaseous structurein the Local Group that is believed to represent material stripped from the Magellanic Clouds. We useultraviolet, optical and radio data obtained with HST /COS, VLT/UVES,
FUSE , GASS, and ATCAto study metal abundances and physical conditions in the Stream toward the quasar Fairall 9. Lineabsorption in the MS from a large number of metal ions and from molecular hydrogen is detected inup to seven absorption components, indicating the presence of multi-phase gas. From the analysisof unsaturated S ii absorption, in combination with a detailed photoionization model, we obtain asurprisingly high α abundance in the Stream toward Fairall 9 of [S/H] = − . ± .
04 (0 .
50 solar).This value is 5 times higher than what is found along other MS sightlines based on similar COS/UVESdata sets. In contrast, the measured nitrogen abundance is found to be substantially lower ([N/H] = − . ± . α ] ratio of − .
85 dex. The substantial differences in thechemical composition of MS toward Fairall 9 compared to other sightlines point toward a complexenrichment history of the Stream. We favour a scenario, in which the gas toward Fairall 9 was locallyenriched with α elements by massive stars and then was separated from the Magellanic Clouds beforethe delayed nitrogen enrichment from intermediate-mass stars could set in. Our results support (butdo not require) the idea that there is a metal-enriched filament in the Stream toward Fairall 9 thatoriginates in the LMC. Subject headings:
ISM: abundances – Galaxy: halo – Galaxy: evolution – Magellanic Clouds – quasars:absorption lines INTRODUCTION
The distribution of neutral and ionized gas in the cir-cumgalactic environment of galaxies is known to be animportant indicator of the past and present evolutionof galaxies. Both the infall of metal-poor gas from inter-galactic space and from satellite galaxies and the outflowof metal-rich gaseous material through galactic windsrepresent key phenomena that determine the spatial dis-tribution and the physical state of the circumgalactic gasaround massive galaxies.From observations and theoretical studies, it is knownthat galaxy interactions between gas-rich galaxies cantransport large amounts of neutral and ionized gas intothe circumgalactic environment of galaxies. In the localUniverse, the most massive of these extended tidal gasfeatures can be observed in the 21 cm line of neutral hy-drogen (H i ). The most prominent nearby example of atidal gas stream produced by the interaction of galaxies is a Based on observations obtained with the NASA/ESA HubbleSpace Telescope, which is operated by the Space Telescope Sci-ence Institute (STScI) for the Association of Universities for Re-search in Astronomy, Inc., under NASA contract NAS5D26555,and on observations collected at the European Organisation forAstronomical Research in the Southern Hemisphere, Chile underProgram ID 085.C − the Magellanic Stream (MS), a massive ( ∼ − M ⊙ )stream of neutral and ionized gas in the outer halo of theMilky Way at a distance of ∼ −
60 kpc (e.g., Wan-nier & Wrixon 1972; Gardiner & Noguchi 1996; Weiner& Williams 1996; Putman et al. 2003; Br¨uns et al. 2005;Fox et al. 2005, 2010; Koerwer 2009; Besla et al. 2007,2010, 2012). The Magellanic Stream spans over 200 ◦ on the sky (e.g., Nidever et al. 2010) and has a (mean)metallicity that is lower than that of the Milky Way, butcomparable with the metallicity found in the SMC andLMC (0 . − . ;Sembach et al. 2001; Richter et al. 2001a). A numberof theoretical studies, including tidal models and ram-pressure stripping models, have been carried out to de-scribe the Stream’s motion in the extended halo of theMilky Way and pinpoint its origin in one of the two Mag-ellanic Clouds (Gardiner & Noguchi 1996; Mastropietroet al. 2005; Connors et al. 2006; Besla et al. 2010; Diaz &Bekki 2011).The origin and fate of the Magellanic Stream is closelyrelated to the trajectories of LMC and SMC (e.g., Con-nors et al. 2004, 2006; Besla et al. 2007), and any real-istic model of the MS thus needs to consider the dy- P. Richter et al.namical and physical state of the Milky Way/MagellanicClouds system as a whole (see also Bland-Hawthorn etal. 2007; Heitsch & Putman 2009). While early tidalmodels have assumed that the Magellanic Stream is aproduct from the tidal interaction between LMC andSMC as they periodically orbit the Milky Way (e.g., Gar-diner & Noguchi 1996), more recent proper motion mea-surements of the Magellanic Clouds (MCs) (Kallivayalilet al. 2006a, 2006b, 2013) indicate that the MCs maybe on their first passage around the Milky Way. Somesubsequent tidal models (Besla et al. 2010; Diaz & Bekki2011) thus favour a first-infall scenario for the MagellanicStream. Moreover, while many models (e.g., Connors etal. 2006) place the origin of the Stream’s gaseous materialin the SMC, other, more recent studies trace back at leastpart of the Stream’s gaseous material in the LMC (e.g.,Nidever et al. 2008). The latter study also highlights therole of energetic blowouts from star-forming regions inthe LMC for the formation of the Stream. Clearly, fur-ther theoretical studies and observations are required topinpoint the origin of the MS based on different (inde-pendent) methods.In the first paper in our series analyzing the chem-ical and physical conditions in the Magellanic Stream(Fox et al. 2013; hereafter Paper I), we have investi-gated MS absorption in the UV and optical along thelines of sight toward RBS 144, NGC 7714 PHL 2525, andHE 0056 − z em = 0 . l = 295 . b = − . ◦ on thesky from the SMC. This sightline is the best-studied inabsorption of all MS directions (Songaila 1981; York etal. 1982; Lu, Savage & Sembach 1994; Gibson et al. 2000;Richter et al. 2001a; Sembach et al. 2003), largely becausethe Fairall 9 is bright in both the optical and the UVand the Stream’s H i column in this direction is large(log N (H i ) ≈
20; see Gibson et al. 2000). The high col-umn of neutral gas ensures that a wide range of low-ionization UV metal lines are detectable in the Stream,and even molecular hydrogen was observed in the MS to-ward Fairall 9 data from the
Far Ultraviolet SpectroscopicExplorer ( FUSE ; Richter et al. 2001a; Wakker 2006). Us-ing a spectrum of Fairall 9 obtained with the GoddardHigh Resolution Spectrograph (GHRS) onboard the
Hub-ble Space Telescope (HST) together with Parkes 21 cmH i data Gibson et al. (2000) derived a metallicity of theStream toward Fairall 9 of [S/H] = − . ± . +0 . − . ( ∼ . ∼ − HST , both data sets providingabsoption spectra with excellent S/N ratios. The combi-nation of these data sets, as desribed in this study, there-fore provides a particular promising strategy to study ingreat detail the chemical and physical conditions in theMagellanic Stream in this direction.This paper is organized as follows: in Sect. 2 we de-scribe the observations and the data reduction. The col-umn density measurements and the profile modeling areexplained in Sect. 3. In Sect. 4 we derive chemical andphysical properties of the gas in the MS. We discuss ourresults in Sect. 5. Finally, a summary of our study isgiven in Sect. 6. OBSERVATIONS AND SPECTRAL ANALYSIS
VLT/UVES observations
Fairall 9 was observed with the VLT/UVES spectro-graph (Dekker et al. 2010) in 2010 under ESO programID 085.C-0172(A) (PI: A. Fox). The observations weretaken in Service Mode using Dichroic 1 in the 390+580setting, a 0.6 ′′ slit, and no rebinning. The observationswere carried out under good seeing conditions ( < ′′ ).The raw data were reduced with the standard UVESpipeline, using calibration frames taken close in time tothe corresponding science frames. The reduction stepsinvolve subtraction of the bias level, inter-order back-ground, sky background, night sky emission lines, andcosmic ray hits. The frames were then flat-fielded, opti-mally extracted and merged. The wavelength scale wascorrected for atmospheric dispersion and heliocentric ve-locity and then placed into the local standard of rest(LSR) velocity frame. Multiple exposures on the sametarget were registered onto a common wavelength gridand then added. The final spectra have a very highspectral resolution of R ≈
70 000 corresponding to aFWHM of 4 . − . They cover the wavelength rangebetween 3300 and 6800 ˚A. The S/N ratio per resolutionelement is 40 at 3500 ˚A (Ti ii ), 65 at 4000 ˚A (Ca ii ), and83 at 6000 ˚A (Na i ). The UVES data thus provide muchhigher sensitivity and substantially higher spectral res-olution than previous optical measurements of Fairall 9(Songaila 1981).he Magellanic Stream toward Fairall 9 3 G a l a c t i c l a t i t ude N ( H ) [ c m ] I - SMC - o
30´ Galactic longitude ( 10 ) x - o
00´ 295 o o o - o G a l a c t i c l a t i t ude ( 10 ) x SMCLMC N ( H ) [ c m ] I - Galactic longitude285 o - o - o - o Fig. 1.—
Upper panel: H i column-density map of the Magel-lanic Stream in the general direction of Fairall 9, based on 21cmdata from GASS (the angular resolution is 15 . ′ − . The directions to LMC and SMC are in-dicated with the white arrows. Lower panel: H i column-densitymap of the Fairall 9 filament in the Stream, based on 21cm datafrom ATCA+Parkes (the angular resolution is 1 . ′ × . ′ HST /COS observations
Fairall 9 was observed with the
HST /COS spectro-graph (Green et al. 2012) in 2012 under
HST programID 12604 (PI: A. Fox). A four-orbit visit provided atotal of 5378 s of exposure time with the G130M/1291wavelength setting, and 6144 s with the G160M/1589 set-ting. With each grating, all four FP-POS positions wereused to dither the position of the spectrum on the de-tector to reduce the fixed-pattern noise. The raw datawere processed and combined with the CALCOS pipeline(v2.17.3). For the coaddition of the individual exposureswe used interstellar absorption lines as wavelength refer-ence. The final, co-added spectra then were transformedinto the LSR velocity frame. The COS data cover theUV wavelength range between 1131 and 1767 ˚A. Thespectra have a resolution of R ≈
16 000 (FWHM ≈ − ) and a pixel size of ∼ − . The S/N ratioper resolution element is 37 at 1200 ˚A and 26 at 1550 ˚A.In order to minimize geocoronal emission, which con-taminates the absorption lines of O i λ ii λ i Ly α in the velocity range −
200 km s − . v LSR . +200 km s − during orbital daytime, we re-reduced thedata with a night-only extraction. For this, data were ex-tracted from those time intervals when the Sun’s altitudewas less than 20 degrees. FUSE observations
As part of our study we also re-analyze archival
FUSE spectra of Fairall 9. These spectra were obtained in 2000under
FUSE program ID P101 (PI: K.R. Sembach) witha total exposure time of 34 827 s. They show strongmolecular hydrogen absorption arising in the MS, as pre-sented by Richter et al. (2001a). The
FUSE spectra havea resolution of ∼
20 km s − (FWHM), and cover thewavelength range 912 − absorption in the MS, we mea-sure a S/N of ∼ − ∼
15 per resolutionelement due to the increased background flux from thebroad, redshifted Ly β emission from Fairall 9. GASS
The H i . ′
6, lead-ing to an RMS of 57 mK per spectral channel (∆ v =0 . − ). This value translates to an H i column den-sity detection limit of N (H i ) lim = 4 . × cm − , as-suming a Gaussian-like emission line with a width of20 km s − FWHM. Fig. 1, upper panel, shows an H i col-umn density map of the local environment of the Mag-ellanic Stream centered on Fairall 9 based on the GASS21cm data. ATCA 21cm data
We supplement our measurements with higher-resolution H i data obtained with the Australia Com-pact Telescope Array (ATCA). The data in the direc-tion of Fairall 9 were observed in 1998 and 1999 by MaryPutman using the 750A, 750B, and 750D configurations.The ATCA is an east-west interferometer with six an-tennas. Each antenna has a diameter of 22 m. For theobservations a correlator band width of 4 MHz was cho-sen, resulting in a velocity resolution of about 0 . − .Since the Fairall 9 cloud is too large to fit within a singlepointing a mosaic consisting of three fields was made.The observing time for each of these three fields was12 hours leading to a full u-v coverage. The FWHM ofthe synthesized ATCA beam is 1 . ′ × . ′
4. P. Richter et al.
Fig. 2.—
Optical absorption profiles of Ca ii , Na i , and Ti ii fromVLT/UVES data of Fairall 9 are shown. Absorption from the Mag-ellanic Stream is seen at LSR velocities between +130 and +240km s − . The solid red line display the best possible Voigt profilefit to the data. The individual velocity components are indicatedby the tic marks. In the upper two panels the H i The ATCA data set was reduced by Christian Br¨unswith the MIRIAD software. To circumvent the missingshort-spacings problem, single-dish data from the Parkestelescope were used. Image deconvolution and combina-tion with the single-dish data were performed with theMiriad-Task MOSMEM. The Parkes data used for theshort-spacings correction were obtained in the frameworkof an H i survey of the Magellanic System (see Br¨uns etal. 2005 for details). Spectral analysis methods
Our strategy for the analysis of the optical and UVabsorption-line data of Fairall 9 combines different tech-niques to optimally account for the different spectral res-olutions and S/N in the data. The reduced and coad-ded spectra from UVES, COS, and
FUSE first werecontinuum-normalized using low-order polynomials thatwere were fit locally to the spectral regions of interest. For the high-resolution VLT/UVES data we have usedVoigt-profile fitting to decompose the MS absorption pat-tern in the optical lines of Ca ii , Na i , and Ti ii into indi-vidual absorption components (Voigt components) andto derive column densities ( N ) and Doppler parameters( b values) for the individual components. For the fitting,we made use of the FITLYMAN package implemented inthe ESO-MIDAS analysis software (Fontana & Ballester1995). Laboratory wavelengths and oscillator strengthshave been taken from the compilation of Morton (2003).Note that with our fitting technique we are able to mea-sure b values smaller than the intrumental resolution aswe simultaneously fit the line doublets of Ca ii and Na i ,so that relative strengths of the lines are taken into ac-count for the determination of both b and N . From thisfitting procedure we obtain a component model for theFairall 9 sightline, in which the LSR velocity centroids ofthe individual absorption components and the b valuesfor the low ions are defined.In addition to Voigt-profile fitting, we have used theapparent optical depth method (AOD method; Savage& Sembach 1991) to derive total gas column densitiesfor the (unsaturated) optical absorption profiles of Ca ii ,Na i , and Ti ii . The AOD analysis was made using thecustom-written MIDAS code span that allows us to mea-sure equivalent widths and AOD column densities (andtheir errors) in absorption spectra from a direct pixelintegration.For the medium-resolution HST /COS data we haveused profile modeling and the AOD method to derivecolumn densities and column density limits for thevarious different low, intermediate, and high ions thathave detectable transitions in the COS wavelengthrange. The ion transitions in the COS data consid-ered in this study include C ii λ .
5, C ii ⋆ λ . iv λλ . , .
8, N i λλ . , . , . v λλ . , .
8, O i λ .
2, Al ii λ . ii λλ . , . , . , . , .
7, Si iii λ .
5, Si iv λλ . , .
8, P ii λ .
8, S ii λλ . , . , .
5, Fe ii λλ . , . , . ii λ .
1. For the profile modeling of neutraland singly-ionized species that trace predominantlyneutral gas in the MS we have used as input the compo-nent model defined by the optical Ca ii absorption (seeabove). In this model, the LSR velocities and b valuesof the seven absorption components are constrained bythe Ca ii fitting results, while the column density foreach component is the main free parameter that can bevaried for each ion listed above. Our previous studiesof Ca ii in the Galactic halo (Richter et al. 2005, 2009;Ben Bekhti et al. 2008, 2011; Wakker et al. 2007, 2008)and in intervening absorbers at low redshift (Richter etal. 2011) have demonstrated that Ca ii is an an excellenttracer for the distribution of neutral and partly ionizedgas and its velocity-component structure, even in regionswhere Ca ii is not the dominant ionization state.Based on the Ca ii model, and using a modified ver-sion of FITLYMAN , we have calculated for each individualUV line a synthetic absorption profile, for which we haveconvolved an initial Voigt profile with input parameters( v i , b i , N i ) for i = 1 ... N i were varied to minimize the differences between thesynthetic absorption profile and the observed COS data.This method delivers reliable (total) column densities forthose ions that have multiple transitions with substan-tially different oscillator strengths in the COS wavelengthrange and for individual lines that are not fully saturated.More details about the accuracy of this method are pre-sented in Sect. 3.1 and in the Appendix.For the intermediate and high ions (Si iii , Si iv , C iv )the component structure and b values are expected tobe different from that of the low ions (as the gas phasetraced by these ions often is spatially distinct from thegas phase traced by the low ions), but no informationis available on the true component structure of theseions from the optical data. Therefore, we did not try tomodel the absorption profiles of the high-ion lines, butestimated total column densities (and limits) for theseions solely from the AOD method. Similarly, we usedsolely the AOD method to determine the column densityof C ii ⋆ .For the analysis of H in the Magellanic Stream de-tected in the Fairall 9 FUSE data (Richter et al. 2001a),we have modeled the H absorption using synthetic spec-tra generated with FITLYMAN . As model input we takeinto accout the component structure and line widthsseen in the optical Ca ii /Na i absorption, together witha Gaussian LSF according to the FUSE spectral resolu-tion. H wavelengths and oscillator strengths have beenadopted from Abgrall & Roueff (1989). Details on theH modeling are presented in Sect. 3.3. COLUMN-DENSITY MEASUREMENTS
Metal absorption in the optical
Optical absorption related to gas in the MagellanicStream in the velocity range 120 −
230 km s − (this veloc-ity range is defined by the H i
21 cm data of the Stream)is detected in the UVES spectrum of Fairall 9 in thelines of Ca ii ( λλ . , . i ( λλ . , . ii ( λ . i ii absorption is detectedin seven individual absorption components centered at v LSR = +143 , +163 , +172 , +188 , +194 , +204 and +218km s − with logarithmic column densities in the rangelog N (Ca ii ) = 11 . − .
68 (where N is in units [cm − ]throughout the paper). The red solid line in Fig. 2 indi-cates the best Voigt-profile fit to the data. All Ca ii , Na i and Ti ii column-density measurements are summarizedin Table 1.The strongest Ca ii absorption component in the MS(in terms of the absorption depth) is component 4 at+188 km s − ; this very narrow component is also de-tected in both Na i lines (Fig. 2). From the simultaneousfit of the Ca ii and Na i doublets we obtain a very smallDoppler parameter for component 4 of b = 1 . − .The small b value for this component is confirmed by fitsto the individual lines of Ca ii and Na i , which all imply b < − . The detection of Na i together with thesmall b value indicates that the gas in component 4 isrelatively cold and dense and possibly is confined in adense core with little turbulence. Components 3,4, and Fig. 3.—
Velocity profiles of various low ions in the COS dataof Fairall 9 are shown. The FUV data for Si ii λ . FUSE . The velocity components related to gas in the MagellanicStream, as identified in the optical Ca ii data, are indicated with thetic marks. The red solid line indicates the best-fitting absorptionmodel for the UV absorption, as described in Sect. 3.2. The single,saturated lines of C ii λ . i λ . ii (Fig. 2). Since Ti ii and H i have almost identical ionization potentials (Table 2), thedetection of Ti ii suggests that most of the neutral gascolumn density is contained in these three components.This conclusion is supported by the H i − . Weak and relativelybroad ( b = 7 . − ) Na i absorption is also detectedin component 1 at +143 km s − .As mentioned in Sect. 2.5, we also have used the AODmethod to determine the total column densities for Ca ii ,Na i and Ti ii by integrating over the velocity range rel-evant for MS absorption (i.e., +130 km s − ≤ v LSR ≤ +230 km s − .) The column densities derived by thesetwo different methods agree very well within their 1 σ error ranges (see Table 1, rows 10 and 11). Metal absorption in the UV
P. Richter et al.
TABLE 1Summary of ion column-density measurements
Component 1 2 3 4 5 6 7 v [km s − ] +143 +163 +172 +188 +194 +204 +218 b [km s − ] 7 . . . . . . . σ b [km s − ] 2 . . . . . . . N log N log N log N log N log N log N log N tot , fit log N tot , AOD H i GASS ... ... ... ... ... ... ... ... 19 . i VLT/UVES 10 .
92 ... ... 11 .
05 ... ... ... 11 . ± .
04 11 . ± . ii VLT/UVES 11 .
58 11 .
29 11 .
63 11 .
68 11 .
58 11 .
43 11 .
35 12 . ± .
02 12 . ± . ii VLT/UVES ... ... 11 .
11 11 .
15 11 .
24 ... ... 11 . ± . ≤ . i HST /COS ... ... ... ... ... ... ... ... ≤ . ii HST /COS ... ... ... ... ... ... ... ... ≥ . ii ⋆ HST /COS ... ... ... ... ... ... ... ... 13 . ± . iv HST /COS ... ... ... ... ... ... ... ... 13 . ± . i HST /COS 13 .
65 13 .
42 13 .
84 14 .
10 13 .
98 13 .
54 13 .
46 14 . ± . ≥ . i HST /COS ... ... ... ... ... ... ... ... ≥ . ii HST /COS 12 .
55 12 .
53 12 .
92 12 .
58 12 .
69 12 .
37 12 .
24 13 . ± . ≥ . ii HST /COS 14 .
10 13 .
75 14 .
17 14 .
13 14 .
24 13 .
92 13 .
89 14 . ± .
04 15 . ± . iii HST /COS ... ... ... ... ... ... ... ... ≥ . iv HST /COS ... ... ... ... ... ... ... ... 13 . ± . ii HST /COS ... ... ... ... ... ... ... ... ≤ . ii HST /COS 13 .
95 13 .
70 14 .
03 14 .
00 14 .
09 13 .
77 13 .
74 14 . ± .
02 14 . ± . ii HST /COS 13 .
47 13 .
61 13 .
94 13 .
85 13 .
94 13 .
67 13 .
65 14 . ± .
04 14 . ± . ii HST /COS ... ... ... ... ... ... ... ... ≤ . In Fig. 3 we show normalized UV absorption profilesof the low ions C ii , N i , O i , Al ii , Si ii , S ii , and Fe ii .The absorption profiles of the intermediate and high ionsSi iii , C iv and Si iv are shown in Fig. 4. Following theprocedure described in Sect. 2.5, we have reconstructedthe absorption pattern of UV metal lines of the low ions,using the component model defined by the optical lines ofCa ii . As already mentioned, the modeling method pro-vides relevant results only for those ions that have mul-tiple transitions in the available COS wavelength bandand for single lines that are not (or at most mildly) sat-urated. In our case, the modeling method could be usedto determine column densities for N i , Al ii , Si ii , S ii ,and Fe ii , while for the fully saturated (single) lines ofC ii and O i the modeling does not yield relevant columndensity limits. In Fig. 3, the best-fitting column-densitymodel for each ion is overlaid with the red solid line. Thebest-fitting model assumes the same b values for the indi-vidual subcomponents as derived for Ca ii from the Voigtfit of the UVES data. The ion column densities for eachvelocity component are listed in Table 1 in rows 2 − ∼ .
30 dex, typically). Much better constrained are the total column densities (Table 1, row 10) for the ion multi- plets N i , S ii , Si ii and Fe ii , which have 1 σ uncertainties < .
05 dex. This is because the integrated (total) col-umn density for each of these ions is determined predom-inantly by the integrated equivalent widths of the weakerlines and the equivalent width ratios of the weaker andstronger transitions. In the Appendix we present addi-tonal model plots to provide a more detailed insight intothe allowed parameter range ( b i , N i ) for the individualsubcomponents and their impact on the shape of the ab-sorption profiles and total column-density estimates.Next to the absorption modeling we have used theAOD method to determine total column densities (orcolumn-density limits) in the COS data for each of theions listed above (Table 1, row 11). For the unsaturatedlines of S ii , Si ii , and Fe ii the total column densities de-rived from the AOD method are in excellent agreement( < σ ) with the total column densities determined fromthe profile modeling. Note that the absorption of C ii ⋆ in the MS will be discussed separately in Sect. 4.4. H i The GASS 21cm velocity profile shown in the upperpanel of Fig. 2 indicates H i emission from gas in the Mag-ellanic Stream in the LSR velocity range between +100and +250 km s − . The MS emission shows an asym-metric pattern with a peak near v LSR ≈ +180 km s − ,surrounded by weaker satellite componets at lower andhigher radial velocities. The overall shape of the emis-sion profile mimics that of the inversed absorption pro-files of the low-ion lines of N i and S ii indicating thatboth the pencil-beam absorption data and the GASSemission data, that have a beam-size of 15 . ′
6, samplethe same physical structure. Integrating the GASS datain the above given velocity range for the single pixelthat covers the position of Fairall 9 we obain a totalH i column density in the Stream of log N (H i ) = 19 . N (H i ) = (8 . ± . × cm − ). The pixel value ishe Magellanic Stream toward Fairall 9 7 Fig. 4.—
COS velocity profiles of the intermediate and high ionsSi iii , C iv and Si iv . The velocity structure in these ions is verydifferent from that of the low ions, suggesting that they trace spa-tially distinct regions within the MS. High-velocity Si iv absorptionbetween +150 and +200 km s − in the Si iv λ . somewhat smaller than the beam-averaged column den-sity, which is log N (H i ) = 19 . − . Because of the relatively low S/N in theATCA data, the spectrum shown in Fig. 2 was binned fordisplaying purposes to 8 . − wide pixels. The over-all shape of the profile from the ATCA data is roughlysimilar to the one from GASS, but in the ATCA spec-trum the MS 21cm emission peaks at a somewhat highervelocity ( v LSR ≈ +190 km s − , thus coinciding with thepeak absorption in Ca ii and Na i ; see Fig. 2) and at aslightly higher brightness temperature. The MS emissionprofile from ATCA is, however, somewhat narrower thanthe GASS profile. Integration over the MS velocity rangein the (unbinned) ATCA data yields a total H i columndensity in the Stream of N (H i ) = (8 . ± . × cm − or log N (H i ) = 19 . +0 . − . , thus in excellent agreementwith the GASS data.For comparison, Gibson et al. (2000) find log N (H i ) =19 .
97 ( N (H i ) = (9 . ± . × cm − ) using Parkes TABLE 2Summary of H column-density modeling J b a [km s − ] log N ( J ) b T J [K] c . . ± .
24 93 +149 − . . ± .
26 93 +149 − . . ± . < . ≤ . < . +0 . − . b -value fixed to 2 . − ; see Sect. 3.4. b Logarithmic H column density for rotational state J . c Equivalent Boltzmann temperature implied by N ( J ); seeSect. 4.4.2. . ′ Leiden-Argentine-Bonn Sur-vey (LAB; Kalberla et al. 2005) with ∼ . i column den-sity in the Stream of log N (H i ) = 19 . i column densities in the MS toward Fairall 9 de-rived from different instruments with very different an-gular resolutions agree with each other within their 1 σ error ranges. Our conclusion is that beam smearing isnot expected to be a critical issue for the determinationof metal abundances in the MS toward Fairall 9 (using21cm data in combination with the pencil-beam UVESand COS data). In the following, we adopt the value oflog N (H i ) = 19 .
95 for the total H i column density in theStream toward Fairall 9 together with an appropriate es-timate of the systematic error of 0 .
03 dex that accountsfor the limited spatial resolution of the 21cm data.As it is shown in the Appendix, this H i column densityin the MS is in agreement with the observed shape of theH i Ly α absorption line in the COS data of Fairall 9,which provides an independent (albeit less stringent)measure for the neutral-gas column density in the Streamin this direction (see also Wakker, Lockman, & Brown2011 for a detailed discussion on this topic). Molecular hydrogen absorption
The detection of H absorption in the MagellanicStream in the FUSE data of Fairall 9 has been reportedpreviously by Richter et al. (2001a) and Wakker (2006).H absorption is seen near v LSR = +190 km s − in therotational levels J = 0 , FUSE data is low ( ≤
15 per resolution ele-ment; see Sect. 2.5), the H absorption is clearly visible inmany lines due to the relatively high H column density(Richter et al. 2001a).We have reanalyzed the Fairall 9 FUSE data basedon an improved data reduction pipeline and the com-ponent model discussed in the previous sections. Richteret al. (2001a) have used a curve-of-growth technique toderive the column densities N ( J ) and b values of the H rotational states J = 0 , ,
2, assuming as single MS ab-sorption component. In contrast, we here model the ab-sorption spectrum of the MS H absorption in the FUSE
Fairall 9 data using synthetic spectra. For our model weassume that the H absorption arises in the region withthe highest gas density in the Stream, which is compo-nent 4 at v LSR = +188 km s − detected in the UVESdata in Ca ii , Na i , and Ti ii (Table 1). In fact, from ab- P. Richter et al. Fig. 5.—
Continuum-normalized
FUSE spectrum of Fairall 9 in the wavelength range between 1075 and 1081 ˚A. Absorption by H fromthe rotational levels J = 0 , absorption is overplotted with the red solid line. sorption studies of Na i and H in interstellar gas it is wellknown that these two species trace the same gas phase(i.e., cold neutral and cold molecular gas; e.g., Welty etal. 2006).We adopt b turb = 1 . − for the H absorptionin component 4, as suggested by the Na i and Ca ii ab-sorption in the same component (Table 1). We furtherassume (and prove later) that the kinetic temperature ofthe H absorbing gas is T kin ≈
100 K in component 4. Itis usually assumed that the Doppler parameter is com-posed of a turbulent and a thermal component, so that b = b + b . For light elements and molecules (suchas H ) b th must not be neglected. We therefore adopt b (H ) = 2 . − for our H model spectrum.Fig. 5 shows our best fitting H model (red solid line)together with the FUSE data for the wavelength rangebetween 1070 and 1080 ˚A. Since the b value is fixed to b (H ) = 2 . − and the velocity to v LSR = +188km s − , the only free parameters in our model are theH column densities N ( J ). Table 2 summarizes the bestfitting H column densities using this method. Thegiven errors represent the uncertainties for log N ( J ) un-der the assumption that b = 2 . − . The result-ing total H column density is log N (H ) = 17 . +0 . − . and the fraction of hydrogen in molecular form is log f H = log [2 N (H ) / ( N (H i ) + 2 N (H ))] = − .
03. Notethat the value for log f H derived in this way representsa sightline-average; the local molecular fraction in the H absorbing region must be higher (because N (H i ) in thisregion is smaller).The H column densities we derive from our spectralmodeling are substantially ( ∼ FUSE data (Richter et al. 2001a; Wakker 2006). The reason forthis discrepancy is the substantially lower b value that weadopt here. This more realistic b value for the H absorb-ing gas is a direct consequence of the previously unknowncomplex absorption pattern in the MS toward Fairall 9,that is visible only in the high-resolution optical data ofCa ii and Na i (see Sect. 3.1). Our study suggests, that,without the support of high-resolution and high S/N op-tical or UV data that can be used to resolve the truesub-component structure of the absorbers and to deliverreliable b values for the subcomponents, the interpreta- tion of interstellar H absorption lines in low-resolutionand low-S/N data from FUSE can be afflicted with largesystematic uncertainties. PHYSICAL AND CHEMICAL CONDITIONS IN THE GAS
Ionization conditions and overall metallicity
The presence of high, intermediate and low ions, aswell as molecular hydrogen, in the Magellanic Streamtoward Fairall 9 in seven individual velocity componentsimplies a complex multi-phase nature of the gas. Toobtain meaningful results for the chemical compositionand physical conditions from the measured ion columndensities it is necessary to consider in detail the ioniza-tion conditions in the absorbing gas structure. For anelement M we define its relative gas-phase abundancecompared to the solar abundance in the standard way[M/H] = log (M/H) − log (M/H) ⊙ . Solar reference abun-dances have been adopted from Asplund et al. (2009).The best ion to study the overall metallicity of neu-tral and weakly ionized interstellar gas is O i . Neutraloxygen and neutral hydrogen have similar ionization po-tentials and there is a strong charge-exchange reactionthat couples both ions in gas with sufficiently high den-sity. For our sightline, however, the O i column density isnot well constrained because of the saturation of the O i λ . i transitionsin the COS wavelength range. Because it is impossibleto measure the O i column density in the MS using ourcomponent models and the AOD method to a satisfac-tory accuracy, we do not further consider the O i λ . ii , which is another useful ion for measuring in-terstellar abundances, as singly-ionized sulfur is an excel-lent tracer for neutral hydrogen without being depletedinto dust grains (e.g., Savage & Sembach 1996). We alsohave a very accurate column-density determination oflog N (S ii ) = 14 . ± .
02 in the Magellanic Stream,based on the absorption modeling of the two S ii linesat 1250 . . ii column density of log N (S ii ) = 14 . ± . HST /GHRS data.From the measured (S ii /H i ) ratio we obtain an initialhe Magellanic Stream toward Fairall 9 9estimate of the sulfur abundance in the Stream towardFairall 9 of [S/H] ≈ − .
30. Since the ionization poten-tial of S ii (23.2 eV; see Table 3) is higher than that ofH i (13.6 eV), we have to model the ionization condi-tions in the gas to obtain a more precise estimate for thesulfur abundance (and the abundances of the other de-tected metals) in the Stream toward Fairall 9. While pre-vious studies of Galactic halo clouds (Wakker et al. 1999;Gibson et al. 2000) have demonstrated that the (S ii /H i )ratio is a very robust measure for the sulfur-to-hydrogenratio, hardly being affected by the local ionizing radiationfield, the ionization modeling is important to obtain a re-libale estimate of the systematic uncertainty that arisesfrom the use of S ii as reference indicator for the overall α abundance in the gas. For this task, we use the photoion-ization code Cloudy (v10.00; Ferland et al. 1998), whichcalculates the expected column densities for different ionsas a function of the ionization parameter U = n γ /n H fora gas slab with a given neutral gas column density andmetallicity, assuming that the slab is illuminated by anexternal radiation field. As radiation field we adopt acombined Milky Way plus extragalactic ionizing radia-tion field based on the models by Fox et al. (2005) andBland-Hawthorn & Maloney (1999,2002), appropriatelyadjusted to the position of the MS relative to the MilkyWay disk.While our absorption modeling provides rough esti-mates for the ion column densities in the seven sub-components, we refrain from trying to model the ioniza-tion conditions in the individual components. This is be-cause the unknown H i column densities in these compo-nents, together with the unknown geometry of the over-all gas structure, would lead to large systematic uncer-tainties for such models. Instead, we have used Cloudyto obtain integrated elemental abundances in the gas ofthe Magellanic Stream toward Fairall 9 (representing theoptical-depth weighted mean of the individual elementabundances in all subcomponents). It is, however, impor-tant to emphasize that meaningful results for integratedelemental abundances from Cloudy can be obtained onlyfor those ions for which the column densities in the ion-ization model do not strongly depend on the gas density(and ionization parameter), because the latter quantitiesare expected to vary substantially among the individualsubcomponents.For our Cloudy model we assume a neutral gas col-umn density of log N (H i ) = 19 .
95, an overall metallicityof [Z/H] = − .
3, and solar relative abundances of themetals from Asplund et al. (2009), based on our resultsdescribed above. In Fig. 6, upper panel, we display theexpected logarithmic column densities for low and inter-mediate ions as a function of log U and log n H for thismodel. The measured total column densities for theseions (Table 1, 10th column) are plotted on the right handside of the panel. The total Ca ii column density of log N (Ca ii ) = 12 .
37 sets a limit for the (averaged) ionisationparameter and density in the gas of log h U i ≤ − .
55 andlog h n H i ≥ − .
95. The corresponding density range thatwe consider as relevant for our Cloudy model (2 . ≤ log n H ≤ − .
95) is indicated in Fig. 6 with the gray-shadedarea. For this range, the expected column densities ofN i , Si ii , Fe ii , S ii , Al ii , Ni ii , and Ti ii are nearly inde-pendent of log U and log n H . For S ii , in particular, thecolumn density varies within only 0 .
03 dex in the above D Fig. 6.—
Upper panel:
Cloudy photoionization model of low-ionization species in the MS toward Fairall 9. The colored linesindicate the predicted ion column densities as a function of theionization parameter U and gas density n H for a gas slab with a(total) neutral hydrogen column density of log N (H i ) = 19 .
95 anda metallicity of [Z/H] = − .
30. Observed ion column densities areindicated on the right-hand side with the filled colored circles. Thegray-shaded area marks the range for U and n H that is relevant forour study based on the oberved Ca ii column density (see Sect. 4.1). Lower panel: differences between the observed and the predictedion column densities, ∆ N = N Cloudy − N obs . ; they are discussed inSect. 4.1. For Ni ii , the bar and the arrow indicate an upper limitfor ∆ N (Ni ii ), while for Ca ii the bar and the double-arrow indicatethe allowed range for ∆ N (Ca ii ). given density range, supporting the previous conclusionsfrom Wakker et al. (1999) and Gibson et al. (2000). In thelower panel of Fig. 6 we show the differences (∆ N ) be-tween the measured column densities and the mean col-umn densities predicted by Cloudy. For Ni ii and Ca ii ,the arrows indicate the upper limit and the allowed rangefor ∆ N , respectively.From the comparison between the predicted and mea-sured ion column densities, together with the solar ref-erence abundances from Asplund et al. (2009), we obtainthe following gas-phase abundances: [N/H] = − . ± .
06, [Si/H] = − . ± .
06, [Fe/H] = − . ± . − . ± .
04, [Al/H] = − . ± .
08, [Ti/H] = − . ± .
07, [Ni/H] ≤ − .
81, and [Ca/H] ≤
0. Thelisted errors include the uncertainties from the columndensity measurements for the metal ions and H i (Ta-ble 1), the systematic uncertainty due to the beam sizeof the 21cm measurements (Sect. 3.3), and the system-atic uncertainty for the ionization correction from theCloudy model (see above). The relative contributions ofthese uncertainties to the total error budget are roughlyidentical. Note that we here do not consider the system-atic errors that come with the solar reference abundances0 P. Richter et al. TABLE 3Summary of integrated gas-phase abundances and dust depletion values
Ion (X) Z I.P. a log (M/H) b ⊙ log (X/H i ) c [M/H] d log δ (M) e [eV] +12C i . ± . ≤ − .
93 ... ...C ii . ± . ≥ − .
02 ... ...N i . ± . − . ± . − . ± .
06 ...O i . ± . ≥ − .
87 ... ...Na i
11 5.14 6 . ± . − . ± .
04 ... ...Al ii
13 18.83 6 . ± . − . ± . − . ± . − . ii
14 16.35 7 . ± . − . ± . − . ± . − . ii
15 19.77 5 . ± . ≤ − .
08 ... ...S ii
16 23.34 7 . ± . − . ± . − . ± .
04 (0)Ca ii
20 11.87 6 . ± . − . ± . ≤ − .
30 0 to − . ii
22 13.58 4 . ± . − . ± . − . ± . − . ii
26 16.12 7 . ± . − . ± . − . ± . − . ii
28 18.17 6 . ± . ≤ − . ≤ − . ≤ − . a I.P.= ionization potential. b Solar reference abundance for element M from Asplund et al. (2009). c Listed errors include uncertainties from the column-density measurements of X and H i . d Gas phase abundance for element M, defined as [M/H] = log (M/H) − log (M/H) ⊙ , derived from the CLOUDY model. The listederrors include uncertainties from the column-density measurements, the beam-smearing of the 21cm data, and the ionizationcorrection from the Cloudy modeling. e Depletion value, defined as log δ (M)= [M/H] MS − [S/H] MS . (Asplund et al. 2009; see Table 3, third row), becausethese errors are irrelevant for the comparison betweenour results and other abundance measurements, if theidentical reference values are used. All gas-phase abun-dances derived with Cloudy are listed in Table 3, sixthrow.The measured S/H ratio of [S/H] = − . ± .
04 corre-sponds to an overall sulfur abundance in the MagellanicStream toward Fairall 9 of 0 . +0 . − . solar. The value for[S/H] is 0 .
25 dex higher than the value derived for the MStoward Fairall 9 by Gibson et al. (2000) based on GHRSdata. Several factors contribute to this discrepancy: (a)the somewhat higher (+0 .
08 dex) S ii column densitythat we derive from the COS data, (b) the slightly lowerH i column density ( − .
02 dex) that we obtain from theGASS data, and (c) the substantially higher (0 .
15 dex)solar reference abundance of sulfur (Asplund et al. 2009)that we use to derive [S/H] compared to the value fromAnders & Grevesse (1989) used by Gibson et al. (2000).Remarkably, the sulfur abundance we derive here is higher than the present-day stellar sulfur abundancesin the SMC ([S/H] = − . ± . and in the LMC([S/H] = − . ± .
09; Russell & Dopita 1992), butmatches the present-day interstellar S abundances foundin H ii regions in the Magellanic Clouds (Russell & Do-pita 1990). This interesting result will be further dis-cussed in Sect. 5. Because of the large abundance scatterof sulfur in the Magellanic Clouds and because S ap-pears to be underabundant compared to oxygen in theLMC and overabundant compared to oxygen in the SMC(Russel & Dopita 1992), the measured (S/H) ratio alonedoes not provide a constraining parameter to pinpointthe origin of the gaseous material in the Stream towardFairall 9 in either SMC or LMC.From our Cloudy model it follows that all other ele-ments listed above have gas-phase abundances (or abun-dance limits) that are lower than that of S. Next to theintrinsic nucleosynthetic MS abundance pattern, dust de- pletion is another important effect that contributes tothe deficiency of certain elements (e.g., Ti, Al, Fe, Si) inthe gas phase. This aspect will be discussed in Sect. 4.3.Note, again, that we do not attempt to constrain a singlevalue for log U or log n H with our Cloudy model, as oursightline passes a complex multiphase structure that isexpected to span a substantial range in gas densities andionization parameters.In this paper, we do not further analyze the gas phasethat is traced by the intermediate and high ions Si iii ,C iv and Si iv , which probably is arising in the ion-ized boundary layer between the neutral gas body ofthe Stream and the ambient hot coronal halo gas of theMilky Way (see Fox et al. 2010). For completeness, weshow in Fig. 4 the velocity profiles of the available tran-sitions of Si iii , C iv and Si iv in the COS spectrum ofFairall 9. The velocity structure in these ions clearly isdifferent from that of the low ions, suggesting that theytrace spatially distinct regions in the MS. The high-ionabsorption toward Fairall 9 will be included in a forth-coming paper that will discuss those ions in several linesof sight intersecting the Magellanic Stream (A.J. Fox etal. 2014, in preparation). Nitrogen abundance
The measured gas phase abundance of nitrogen in theMagellanic Stream toward Fairall 9 is [N/H] = − . ± .
06, which is 0 .
85 dex lower than the abundance of sul-fur. In diffuse neutral gas the depletion of nitrogen intodust grains is expected to be very small (Savage & Sem-bach 1996). Consequently, the low nitrogen abundancein the Stream reflects the nucleosynthetic enrichment his-tory of the gas and thus provides important informationabout the origin of the gas.For the SMC, Russell & Dopita (1992) derive amean present-day stellar nitrogen abundance of [N/H] = − .
20. This value is close to the value derived for theMS toward Fairall 9, although the gas in the Stream hasnot been enriched with nitrogen since it was strippedhe Magellanic Stream toward Fairall 9 11
Fig. 7.—
COS velocity profile of C ii ⋆ λ . ii ⋆ absorption from the Magellanic Stream is visible between+120 and +200 km s − . For comparison, the velocity profile ofS ii λ . off its parent galaxy. The nitrogen abundances found inSMC H ii regions and supernova remnants span a ratherwide range of − . ≤ [N/H] ≤ − .
95 (Russell & Do-pita 1990). Thus, the observed nitrogen abundance inthe MS toward Fairall 9 would be in accordance with anSMC origin of the gas only if the (mean) nitrogen abun-dance in the SMC has not increased substantially afterthe Stream was separated ∼ − − . ii regions andsupernova remnants is [N/H] = − .
86 to − .
38, the min-imum [N/H] still being substantially higher than the ni-trogen abundance in the Stream toward Fairall 9. There-fore, an LMC origin of the MS is plausible only if theLMC has substantially increased its nitrogen abundanceduring the last ∼ Dust depletion pattern
Heavy elements such as Al, Si, Fe, Ca, Ti, Fe and Niare known to be strongly depleted into dust grains in in-terstellar gas in the Milky Way and other galaxies, whileother elements such as S, O, N are only very mildly or notat all affected by dust depletion (e.g., Savage & Sembach1996; Welty et al. 1997).To study the depletion pattern of these elements in theMagellanic Stream we define the depletion values fromthe relative abundance of each depleted element, M, inrelation to the abundance of sulfur. We define accord-ingly log δ (M)= [M/H] MS − [S/H] MS . Using this defini-tion, the depletions values are log δ (Si) = − .
27, log δ (Fe) = − .
56, log δ (Ti) = − .
32, log δ (Al) = − . δ (Ni) ≤ − .
51 (see also Table 3, seventh row).Such depletion values are typical for warm Galactic haloclouds, as derived from UV absorption-line measure-ments (see Savage & Sembach 1996; their Fig. 6).Note that the depletion values (or limits) would be different, if the relative chemical abundances of the de-pleted elements would differ from the relative chemicalcomposition of the sun (Asplund et al. 2009). In viewof the abundance patterns found in the LMC and SMC(Russell & Dopita 1990,1992), this is actually a likelyscenario, but no further conclusions can be drawn at thispoint without knowing the exact intrinsic chemical com-position of the MS gas toward Fairall 9.Since the expected Ca ii column density in MS gas isexpected to depend strongly on the local gas densityand ionization parameter (Fig. 6), the depletion valueof Ca cannot be tightly constrained, but may lie any-where in the range log δ (Ca) ≥ − .
58. Under typi-cal interstellar conditions, Ca is strongly depleted intodust grains with large depletion values that are similarto those of other elements that have condensation tem-peratures above T = 1500 K (e.g., Ti; Savage & Sembach1996). If one assumes that log δ (Ca) ≈ log δ (Ti) for theMS toward Fairall 9, then Fig. 6 would imply that theaverage density in the neutral gas in the Stream in thisdirection is log n H ≈ n H ≈ − . This density,together with the neutral gas column density, would im-ply a characteristic thickness of the absorbing neutral gaslayer of d = N (H i ) /n H ≈
30 pc.At first glance, one may regard this as a relatively smallvalue. The complex velocity-component structure of theabsorber indicates, however, that the absorber is com-posed of several small cloudlets along the line of sightthat (together) are extending over a spatial range that islarger than the value indicated by the mean thickness. Inview of the relatively high metal abundance derived forthe kinematically complex absorber toward Fairall 9, thepresence of a population of metal-enriched gas clumps atsmall scales in the MS is not an unlikely scenario.
Physical conditions in the H absorbing region As discussed in Sect. 3.1, the UVES spectrum ofFairall 9 shows the presence of Na i , Ca ii , and possiblyTi ii in component 4. In the same component, the de-tected H absorption is expected to arise (Sect. 3.3), indi-cating the presence of a cold, dense (and predominantlyneutral) gas phase in the MS in this direction. In thefollowing, we want to combine the information from thedifferent instruments to explore the physical and chemi-cal conditions in the H absorbing gas phase in the MStoward Fairall 9. Gas density
While our Cloudy model does not provide any rele-vant information on the local gas densities in the individ-ual subcomponents, we can use the observed molecularhydrogen fraction in the gas to constrain n H in the H absorbing component 4, assuming that the abundance ofH relative to H i is governed by a formation-dissociationequilibrium. In a formation-dissociation equilibrium theneutral to molecular hydrogen column density ratio in aninterstellar gas cloud is given by N (H I) N (H ) = h k i βR n (H I) φ , (1)where h k i ≈ .
11 is the probability that the moleculeis dissociated after photo absorption, β is the photo-absorption rate per second within the cloud, and R is2 P. Richter et al.the H formation rate on dust grains in units cm s − .For low molecular hydrogen fractions we can write n H = n (H I) + 2 n (H ) ≈ n (H I). The parameter φ ≤ i that is physically related to the H absorbing gas, i.e.,the fraction of the neutral gas atoms that can be trans-formed into H molecules (see Richter et al. 2003 for de-tails). Our absorption modeling of the undepleted lowions S ii and N i indicates that component 4 contains ∼ −
30 percent of neutral gas column, so that weassume φ = 0 . − .
30 in lack of any more precise infor-mation.For interstellar clouds that are optically thick in H (i.e., for log N (H ) ≫
14) H line self-shielding needs tobe considered in equation (1), because the self-shieldingreduces the photo-absorption rate ( β ) in the cloud inte-rior. Draine & Bertoldi (1996) find that the H self-shielding can be expressed by the relation β = S β ,where S = ( N H / cm − ) − . < β is the photo absorption rate at the edge ofthe cloud. The parameter β is directly related to theintensity of the ambient UV radiation field at the edgeof the H -bearing clump. Compared to the Milky Waydisk, where UV bright stars dominate the mean photo-absorption rate of β , MW = 5 . × − s − (e.g., Spitzer1978), the value for β in the Magellanic Stream is ex-pected to be substantially smaller. This is because thesolid angle of the Milky Way disk is relatively small at50 kpc distance and the contribution of the extragalacticUV background to β , MS is expected to be (relatively)small, too. From the model by Fox et al. (2005) followsthat the UV flux between 900 and 1100 ˚A is reduced by afactor of 160 compared to the mean flux within the MilkyWay disk, so that we assume β , MS = 3 . × − s − .With a total H column density of 8 . × cm − in theStream the self-shielding factor S becomes 1 . × − , sothat the photo-absorption rate in the cloud core is esti-mated as β = 3 . × − s − . For the H grain formationrate in the Magellanic Stream we adopt the value derivedfor the SMC based on FUSE H absorption-line data(Tumlinson et al. 2002), i.e., R MS = 3 × − cm s − .This value is 10 times smaller than R in the disk of theMilky Way (Spitzer 1978).If we solve equation (3) for φn H = φn (H I) and includethe values given above, we obtain a density of φn H ≈ − or n H ≈ − − for φ = 0 . − .
30. Thesedensities are very close to the density derived for theH absorbing gas in the Leading Arm of the MS towardNGC 3783 ( n H ≈
10 cm − ; Sembach et al. 2001). Notethat H absorption has also been detected in the Magel-lanic Bridge (Lehner 2002). Gas temperature
In Fig. 8 we plot the measured logarithmic H columndensity for each rotational state, log N ( J ), divided byits statistical weight, g J , against the rotational excitationenergy, E J . Rotational excitation energies and statisticalweights have been adopted from the compilation of Mor-ton & Dinerstein (1976). The data points in Fig. 8 canbe fit by a Boltzmann distribution (i.e, a straight line inthis plot), where the slope characterizes the (equivalent)excitation temperature for the rotational levels consid-ered for the fit. For the two rotational ground states Fig. 8.—
Rotational excitation of H in the Magellanic Stream.The logarithmic H column density for each rotational state, log N ( J ), divided by its statistical weight, g J , is plotted against therotational excitation energy, E J . The fits for the excitation tem-peratures (assuming a Boltzmann distribution) for J = 0 , J = 2 , J = 0 , N (1) /N (0) = g /g exp ( − E /kT ). Us-ing this equation, together with our measured columndensities for J = 0 ,
1, we find an excitation tempera-ture of T = 93 +149 − K (Fig. 8 solid line). Because thetwo rotational ground states most likely are collisionallyexcited, T represents a robust measure for the kinetictemperature in the H absorbing gas (e.g., Spitzer 1978);however, because of the saturation of the H lines andthe resulting relatively large errors for N (0 , T is substantial. In a similar manner, wederive for the rotational states J = 2 , T ≤ J ≥ formation pumping(Spitzer 1978). However, the relatively low temperaturelimit of T ≤
172 K reflects the low intensity of the am-bient dissociating UV radiation field at the position ofthe MS.The above given excitation temperatures deviate fromour earlier estimates (Richter et al. 2001a). This is notsurprising, however, since the values for N ( J ) have sub-stantially changed, too (see Sect. 3.3). The derived valueof T = 93 K lies in a temperature regime that is typicalfor diffuse H gas in the disk of the Milky Way and in theMagellanic Clouds (e.g., Savage et al. 1977; Tumlinson etal. 2002; de Boer et al. 1998; Richter et al. 1998; Richter2000).From the density and temperature estimate in the H absorbing gas we now are able to provide a direct es-timate of the thermal pressure, P/k = nT , in the coldneutral gas in the MS. With n H = 8 cm − and T = 93 Kwe obtain a pressure of P/k = 744 cm − K. This value isin good agreement with previous estimates for the ther-mal pressure in the cold neutral phase of the MS fromH i absorbing structure in the MShe Magellanic Stream toward Fairall 9 13is d H = φ N (H i ) /n H ≈ . − . b = 1 . − ; see Table 1) that ismeasured for this component. Electron density and C + cooling rate from C ii ⋆ absorption In Fig. 7 (upper panel) we show the LSR velocity pro-file of the C ii ⋆ λ . ii ; Fig. 7, lower panel). Because we could notidentify any other (e.g., intergalactic) origin for this ab-sorption feature, we assume that it is caused by C ii ⋆ inthe Magellanic Stream. The C ii ⋆ feature is characterizedby a stronger and broader absorption peak near +150km s − , falling together with components 1 and 2 definedin Table 1, and another weak and narrow component thatis seen near +190 km s − , most likely associated withcomponent 4, in which H , Na i , and Ca ii is detected.Therefore, the relative strengths of the C ii ⋆ componentsappear to be inverted from those of the ground-statespecies. This indicates multi-phase gas, in which the ion-ization fractions and electron densities vary among thedifferent absorption components.Under typical interstellar conditions, the relative pop-ulation of the fine-structure levels of ionized carbon (C + )are governed by the balance between collisions with elec-trons and the radiative decay of the upper level into theground state 2 s p P / . The C ii ⋆ λ . s p P / state, which has an energy of ∼ × − eV above the ground state. Measurements ofthe column-density ratios N (C ii ⋆ )/ N (C ii ) thus can beused to estimate the electron densities n e in different in-terstellar environments, including Galactic high-velocityclouds (HVCs) and circumgalactic gas structures (e.g.,Zech et al. 2008; Jenkins et al. 2005).For the C ii ⋆ absorption in the MS toward Fairall 9 wemeasure a total column density of log N (C ii ⋆ ) = 13 . ± .
07 for the velocity range v LSR = 120 −
220 km s − usingthe AOD method. For the weak absorption associatedwith component 4 we derive log N (C ii ⋆ ) = 12 . ± . ii λ . N (C ii ) (see Table 1). We therefore use S ii as a proxy,because the S ii /C ii ratio is expected to be constant overa large density range (Fig. 6) and both elements are notexpected to be depleted into dust grains in the MS. Us-ing a solar (S/C) ratio (Asplund et al. 2009), we estimatefrom the values listed in Table 1 that the total C ii col-umn density in the MS is log N tot (C ii ) ≈ .
10 and thecolumn density in component 4 is log N (C ii ) ≈ . + fine-structure population dependsstrongly on the gas temperature (see, e.g., Spitzer 1978),we do not attempt to estimate a mean value for n e from N tot (C ii ⋆ )/ N tot (C ii ). We know that the temperature inthis multi-phase absorber spans a large range between ∼
100 K and probably few 1000 K, and therefore an esti-mate for h n e i would be meaningless. Instead, we concen-trate on component 4, for which we know the gas temper-ature and density from the analysis of the H rotationalexcitation (Sect. 4.4). Keenan et al. (1986) have calcu-lated electron excitation rates for the C + fine-structure transitions for a broad range of physical conditions ininterstellar gas. Using our temperature and density es-timates for the gas in component 4 ( T = 93 K and n H = 4 − − ) together with their predicted pop-ulation rate ratios (2 s p P / /2 s p P / ) for T = 100K and n H = 5 −
10 cm − (their Fig. 2), the measuredcolumn-density ratio log[ N (C ii ⋆ )/ N (C ii )] = − .
67 im-plies that the electron density in component 4 is smallcompared to n H , namely n e ≤ .
05 cm − . This low valuefor n e is in excellent agreement with expectations for astable cold, neutral medium in gas with subsolar metal-licities (Wolfire et al. 1995).Having a robust measure for N tot (C ii ⋆ ), it is also pos-sible to estimate the C + cooling rate in the MagellanicStream toward Fairall 9. Because ionized carbon is a ma-jor cooling agent of interstellar gas in a wide range ofenvironments (e.g., Dalgarno & McCray 1972), the C + cooling rate is an important quantity that governs thethermal state of diffuse gas inside and outside of galax-ies. For the hydrogen and electron densities in the Magel-lanic Stream (see above) the C + cooling rate is governedpredominantly by spontaneous de-excitations, while col-lisional de-excitations can be neglected (e.g., Spitzer1978). Following Lehner, Wakker & Savage (2004), theC + cooling rate per neutral hydrogen atom can be esti-mated as l C = 2 . × − N (C + ) /N (H i ) erg s − . If weadopt log N (H i ) = 19 .
95 and log N (C ii ⋆ ) = 13 .
35 we de-rive a mean (sightline average) C + cooling rate per neu-tral hydrogen atom of log l C = − .
14 for the MS towardFairall 9. This value is almost one order of magnitudelower than the cooling rate derived for the one-tenth-solar metallicity HVC Complex C (log l C ≈ − l C ≈ −
26; see Lehner, Wakker & Savage 2004,their Table 4). This result underlines the importance ofmetal cooling for the thermal state of metal-rich circum-galactic gas absorbers in the local Universe. DISCUSSION
Enrichment history of the Magellanic Stream
Alpha and nitrogen abundances
One major result of our study is the surprisingly highmetallicity of the MS toward Fairall 9. From the mea-sured sulfur abundance of [S/H] = − . ± .
04 fol-lows that the α abundance in the Stream in this di-rection ( l = 295 , b = −
58) is as high as 0 . ∼ α abundance derivedfor other sightlines passing the MS toward NGC 7469( l = 347 , b = − l = 299 , b = − l = 88 , b = −
56; Fox et al. 2010, 2013). Ourdetailed analysis of the ionization conditions in the gas,the comparison between H i i Ly α absorption toward Fairall 9(see Appendix) do not provide any evidence, that thishigh sulfur abundance could be a result from the varioussystematic errors that come along with our analysis (seealso the discussion in Gibson et al. 2000 on this topic).We thus are forced to conclude that the measured highsulfur abundance in the gas reflects the true chemicalcomposition of the Stream in this direction.As a guide to the following discussion, we have plot-4 P. Richter et al.ted in Fig. 9 the derived MS abundances (black filledcircles) together with the SMC and LMC present-daystellar and nebular abundances (Russel & Dopita 1992;Hughes et al. 1998; red (SMC) and green (LMC) filledcircles). With a sulfur abundance of 0 . α abundance that is higher than the average present-day α abundance in both SMCand LMC ( ∼ . ∼ . ∼ α abundance is at most ∼ . ∼ .
25 solar if originating inthe LMC. While the observed MS abundances towardRBS 144, NGC 7469, and NGC 7714 are in line with theseabundance limits (and actually favour an SMC origin forthis part of the Stream; see Paper I), the sulfur abun-dance in the MS toward Fairall 9 is at least twice as highas expected from a simple tidal model that assumes a ho-mogeneous pre-enrichment of the SMC/LMC gas (e.g.,Dufour 1975) before the Stream was stripped off.A second important finding of our study, that adds tothis puzzle, is the relatively low nitrogen abundance inthe MS toward Fairall 9. The [N/S] ratio (= [N/ α ] ratio)is − .
85 dex, which is very low for the relatively high α abundance of [ α /H]= − . ii regions and high-redshift damped Ly-man α (DLA) systems (see Pettini et al. 2008; Jenkinset al. 2005, and references therein). The α -process ele-ments O, S, and Si are believed to be produced by TypeII supernovae from massive progenitor stars, while theproduction of N, as part of the CNO cycle in stars ofdifferent masses, is less simple (and not yet fully under-stood). The so-called “primary” nitrogen production oc-curs when the seed elements C and O are produced withinthe star during the helium burning phase, while for “sec-ondary” nitrogen production these seed elements werealready present when the star condensed out of the ISM(Pettini et al. 2002; Henry & Prochaska 2007). PrimaryN is believed to be produced predominantly by stars ofintermediate masses on the asymtotic giant branch (e.g.,Henry et al. 2000). Because of the longer lifetime ofintermediate-mass stars, primary N therefore is expectedto be released into the surrounding ISM with a time delayof ∼
250 Myr compared to α elements. At low metallici-ties less than ∼ . ∼ − Fig. 9.—
Comparison of present-day metal abundances in theSMC (red dots), LMC (green dots), and the Magellanic Streamtoward Fairall 9 (black dots). For SMC and LMC the metallicitiesare derived from stellar and nebular abundances (from Russell &Dopita 1992; Hughes et al. 1998). primary (see Pettini et al. 2008).The low [N/S] and high [S/H] ratios observed in the MStoward Fairall 9 therefore indicate an abundance patternthat is dominated by the α enrichment from massive starsand Type II supernovae, while only very little (primary)nitrogen was deposited into the gas. Origin of the gas
The most plausible scenario that would explain suchan enrichment history, is, that the gas that later becamethe Stream was locally enriched in the Magellanic Clouds ∼ − α elements by several supernovaexplosions in a star cluster or OB association, and thenseparated/stripped from the stellar body of the parentgalaxy before the primary nitrogen was dumped into thegas and the metals could mix into the ambient interstel-lar gas. Eventually, the supernova explosions may havepushed away the enriched material from the stellar diskso that the gas was already less gravitationally bound atthe time its was stripped and incorporated into the Mag-ellanic Stream. In this scenario, it is the present-day Nabundance in the MS that defines the metallicity floor ofthe parent galaxy and thus provides clues to the originof the gas.Chemical evolution models of the Magellanic Cloudssuggest that the mean metallicities of SMC and LMC1 − . − . predicts that thehe Magellanic Stream toward Fairall 9 15high-velocity gaseous material toward Fairall 9 is part ofthe LMC filament of the Stream und thus should havea different chemical composition than the SMC filamenttraced by the other MS sightlines (Paper I). Another as-pect that may be of relevance in this context is the factthat the Fairall 9 sightline lies only 14.3 ◦ on the skyfrom the SMC. The relative high metal-abundance inthe Stream toward Fairall 9 thus may reflect the increas-ing importance of continuous ram-pressure stripping ofmetal-enriched SMC gas as the MCs get closer to theMilky Way.While the above outlined enrichment scenario appearsfeasible to explain qualitatively the observed trend seenin the abundance pattern in the MS toward Fairall 9,the question arises, how many massive stars and subse-quent supernova explosions would then be required to liftthe [S/N] and [S/H] ratios to such a high level. To an-swer this question, one would first need to have an ideaabout volume and mass of the gas that is enriched inthis manner. The fact that Fox et al. (2010, 2013) havedetermined a much lower α abundance in the Stream of ∼ . d ≈
30 pcfrom the mean gas density ( n H = 1 . − ) and thetotal column density, assuming that the dust depletionvalues of Ti and Ca are similiar in the gas. In the fol-lowing, we regard this value as a realistic lower limit forthe true physical size of the absorbing gas region. If weadopt the SN yields from Kobayashi et al. (2006) and as-sume a spherical symmetry for the absorbing gas regionwith a diameter and density as given above, we calculatethat only a handful of massive stars would be requiredto boost the α abundance from initially 0 . . M ∝ V n H ∝ d n H ), andbecause we use the relation d = N H /n H to estimate d ,we can write M ∝ n − for a fixed (measured) hydrogencolumn density in the gas. Therefore, if the mean gasdensity in the MS toward Fairall 9 would be 0 . − rather than the assumed 1 . − , then the diameter ofthe cloud would be 300 pc instead of 30 pc, but it wouldrequire 100 times more massive stars to enrich this largevolume to the level required. In view of the stellar con-tent of the most massive star forming regions in the Mag-ellanic Clouds (e.g., 30 Doradus; Melnick 1985) and thepossible presence of a major star burst in the MCs at thetime when the Magellanic Stream was formed (Harris &Zaritsky 2009; Weisz et al. 2013), a size of a few dozen upto a few hundred pc together with the proposed enrich-ment scenario is fully consistent with our understandingof the star-formation history of the Magellanic Clouds.In summary, the observed abundance pattern in theMS toward Fairall 9 suggests the presence of a high-metallicity gas filament in the Stream in this direc-tion, possibly originating in a region with enhanced star- formation activity in the Magellanic Clouds ∼ − i velocity profiles of the LAB 21cm all-sky survey(Kalberla et al. 2005). To further explore this scenarioit will be of great importance to identify other sightlinesthat pass the supposed LMC filament in the Stream withbackground QSOs that are bright enough to be observedwith HST /COS. Such observations would also be help-ful to investigate the α abundances in the LMC filamentbased on oxygen rather than sulfur and to further ex-plore element ratios that could help to constrain the en-richment history of the gas.Interestingly, the MS absorber toward Fairall 9 is notthe only example of a circumgalactic gas structure in thenearby Universe that exhibits a high overall metallicitytogether with a low N/ α abundance ratio. For exam-ple, Jenkins et al. (2005) have measured [ α /H]= − . α ] < .
59 in an intervening Lyman-limit system(LLS) at z = 0 .
081 toward the quasar PHL 1811. Thisabsorption system appears to be associated with twonearby spiral galaxies at impact parameters ρ = 34 h − kpc and ρ = 87 h − kpc. It possibly represents gaseousmaterial that has been ejected or stripped from thesegalaxies (or other galaxies nearby) and thus possibly hasan origin that is very similar to that of the MagellanicStream. Note that low [N/ α ] ratios are also found inlow-metallicity HVCs, such as Complex C (e.g., Richteret al. 2001b). Clearly, a systematic study of N/ α ratios incircumgalactic metal absorbers at low z with HST /COSwould be an important project to investigate whethersimilar abundance patterns are typical for circumgalac-tic gaseous structures in the local Universe.
Physical conditions and small-scale structure in thegas
In addition to the very interesting chemical properties,the combined UVES/COS/
FUSE /GASS/ATCA dataset provides a deep insight into the physical conditionsin the Stream toward Fairall 9. The data show a com-plex multiphase gas structure that possibly spans a largerange in gas densities and ionization conditions.The presence of H absorption toward Fairall 9 andNGC 3783 (Sembach et al. 2001) indicate that the MSand its Leading Arm hosts a widespread (because of thelarge absorption cross section/detection rate) cold neu-tral gas phase, possibly structured in a large amount ofsmall, dense clumps or filaments. Assuming a total areaof the MS and its Leading Arm of ∼ , and atypical size for these clumps of a few pc, there could bemillions of these dense structures in the neutral gas bodyof the Stream, if the two H detections toward Fairall 9and NGC 3783 reflects the true absorption-cross sectionof this gas phase throughout the Stream’s neutral gas6 P. Richter et al.body.The fact that the MS can maintain significant amountsof H at moderate gas densities ( n H = 1 −
10 cm − )probably is a result of the relatively low intensity ofthe dissociating UV radiation at the Stream’s locationdue to the absence of local UV sources (see also Fox etal. 2005). Note that the low excitation temperature ofH for J ≥ R MS = 3 × − cm s − and a densityof n H = 5 cm − (see Sect. 4.4), the H formation timeis long, t form = ( R MS n H ) − ≈ (or at least some fraction of it) hasalready formed in the parent galaxy and then survivedthe subsequent tidal stripping. A similar conclusion wasdrawn from the H observations in the Leading Arm ofthe MS toward NGC 3783 (Sembach et al. 2001). Elab-orating our idea, that the gas was locally enriched bymassive stars shortly before the Stream was separatedfrom its parent galaxy, one could imagine the presenceof relatively dense, compressed shells and fragments thatformed the seed structures for the formation of H thatthen were carried along with the Stream’s gaseous body.Independent constraints on the physical conditions inthe gas come from the observed Na i /Ca ii ratio. Wemeasure a Na i /Ca ii column-density ratio of 0 .
23 in com-ponent 4, where the H is expected to reside. In theMilky Way ISM, such a ratio is typical for a dust-bearingwarm neutral medium (WNM), where n H ≤
10 cm − and T = 10 − K, typcially, and Ca ii serves a tracespecies (e.g., Crawford 1992). In such gas, and without dust depletion of Ca and Na, a nearly constant Na i /Ca ii of 0 .
025 ratio would be expected from detailed ioniza-tion models of these ions (see Crawford 1992; Welty etal. 1996; Richter et al. 2011). If these numbers would ap-ply also to the conditions in the MS, they would indicaterelative dust depletions of Ca and Na of log δ (Ca) − log δ (Na) ≈ .
9, or log δ (Ca) ≈ . δ (Na) = 0.These values are in very good agreement with the dust-depletion estimates from our Cloudy model discussed inSect. 4.3. Relevance to intervening QSO absorbers
The Magellanic Stream represents a prime example fora high-column density circumgalactic tidal gas streamin the local Universe. If the Stream would be seen asQSO absorber, it would be classified as LLS, sub-DLA(sub-Damped Lyman α absorber), or DLA, dependingon the position of the sightline passing through the gas.The H i column-density maps presented in Fig. 1 providesan estimate on the covering fractions of these column-density regimes in the MS in the direction of Fairall 9.The results from our multi-sightline campaign to studythe properties of the MS therefore are also of relevancefor the interpretation of interevening absorption-line sys-tems at low redshift. Inhomogeneity of absorbers
The first important result from our study that is of rel-evance for QSO absorption-line studies is, that the phys-ical conditions and the chemical composition appear tovary substantially within the Stream.If such inhomogeneities were typical for tidal gas fila-ments around galaxies, then the interpretation of absorp-tion spectra from circumgalactic gas structures aroundmore distant galaxies (in terms of metallicity, molecu-lar content, physical conditions, gas mass, origin, etc.)would be afflicted with large systematic uncertainties. Infact, most studies that aim at exploring the connectionbetween galaxies and their surrouding circumgalactic gasare limited to single sightlines that pass the galaxy en-vironment at a random impact parameter (e.g., Thomet al. 2011; Tumlinson et al. 2011; Ribaudo et al. 2011).Physical and chemical properties derived from single-sightline studies, however, may not be representative atall for the conditions in the general gaseous environment(in the same way as conditions in the MS along theFairall 9 sightline are not representative for the Streamas a whole). The inhomogeneous chemical compositionof the Stream implies that the metals possibly are notwell mixed in the gas. Former abundance variations inthe parent galaxy due to local star-formation events thusmay have been frozen into the Stream’s spatial metaldistribution. This would be not too surprising, however,since the main processes that stir up and mix the in-terstellar gas within galaxies (i.e., supernova explosions,stellar winds) cannot take place in the Stream simply dueto the lack of stars.
Molecular absorption
The two detections of H absorption in the MS with FUSE toward Fairall 9 and NGC 3783 (Sembach etal. 2001) indicate that tidal gas streams around galax-ies may typically host a widespread, cold gas phase thathas a substantial absorption cross section. This scenariois supported by the recent detection of H absorptionin another, more distant circumgalactic tidal gas streambeyond the Local Group (Crighton et al. 2013). Thesefindings also remind us that the presence of H in anintervening absorber with a complex velocity structureand a high neutral gas column density does not neces-sarily mean that one traces a gaseous disk of a galaxy.The relatively large absorption cross section of H in neutral gas structures around galaxies, as found inthe Milky Way halo (see Richter 2006 and referencestherein), can be explained by the circumstance that thephysical conditions in these star-less gas clouds favourthe formation of diffuse molecular structures even at rel-atively moderate gas densities. In particular, the relativelow intensity of the local UV radiation field, the efficientprocess of H line self-shielding, and the lack of the dis-tructive influence of massive stars probably represent im-portant aspects in this context.H observations in circumgalactic gas clouds not onlyprovide important information on the physical and chem-ical conditions in these structures, they also are of highrelevance for our understanding of the physics of molecu-lar hydrogen in diffuse gas under conditions that are verydiffernt from that in the local ISM (see also Sembachet al. 2001). The transition from neutral to molecularhe Magellanic Stream toward Fairall 9 17gas, in particular, represents one of the most importantprocesses that govern the evolutionary state of galaxiesat low and high redshift. Detailed measurements of H fractions and dust-depletion patterns in (star-less) tidalstreams at low redshift can help to constrain the criticalformation rate of molecular hydrogen in low-metallicityenvironments and thus could be of great importance tobetter understand the distribution and cross section ofmolecular gas in and around high-redshift galaxies. SUMMARY
In this second paper of our ongoing series to studythe Magellanic Stream in absorption we have analyzednewly obtained optical and UV absorption-line datafrom
HST /COS and VLT/UVES together with archival
FUSE and H i v LSR = +143 , +163 , +182 , +194 , +204 and+218 km s − , indicating a complex internal velocitystrucutre of the MS in this direction. Detected ions,atoms and molecules in the Stream include C iv , Si iv ,Si iii , C ii , C ii ⋆ , Al ii , Si ii , S ii , Ca ii , Ti ii , Fe ii , Ti ii ,O i , N i , Na i , and H .2. From the unsaturated S ii absorption and a Cloudyphotoionization model we obtain an α abundance inthe Stream of [S/H] = − . ± .
04 (0 .
50 solar), whichis substantially higher than that found in the Streamalong the lines of sight toward NGC 7469, RBS 144,and NGC 7714 (Fox et al. 2010, 2013). Unfortunately,the unresolved, saturated O i λ . α abundancein the MS toward Fairall 9. Contrary to sulfur, wemeasure a very low nitrogen abundance in the gas of[N/H] = − . ± .
06. The resulting [N/S] ratio is − . α abundances. The low [N/S] andhigh [S/H] ratios observed in the MS toward Fairall 9suggest an abundance pattern that is dominated bythe α enrichment from massive stars and Type IIsupernovae, while only very little primary nitrogen wasdeposited into the gas when the Stream was separatedfrom the Magellanic Clouds.3. The detection of very narrow Na i and H absorption(with b ≈ − ) in the component at v LSR = +188km s − indicates the presence of a compact (pc-scale),cold gas stucure in the MS along this sightline. Fromthe analysis of the newly reduced archival FUSE dataof Fairall 9 we measure a total molecular hydrogencolumn density of log N (H ) = 17 . +0 . − . , improvingprevious results from Richter et al. (2001a). From theanalysis of the H rotational excitation we obtain akinetic temperature in the cold neutral gas phase of T ≈
93 K. For the gas density we derive n ≈ − − ,assuming that the H gas is in a formation-dissociationequilibrium. The resulting estimate for the thermal gaspressure is P/k ≈
750 cm − K, thus in good agreement with values derived from previous studies of the Stream.The detection of H absorbing structures in the MS,whose linear and angular sizes must be very small ( ∼ ∼ ii ,Ti ii , Ni ii , Al ii , and Ca ii indicate that the gas phaseabundances of these elements are affected by dust de-pletion. We combine our column-density measurementsfor these ions with a Cloudy photoionization model andderive dust depletion values relative to sulfur of log δ (Si) = − .
27, log δ (Fe) = − .
56, log δ (Ni) ≤ − . δ (Ti) = − .
32, log δ (Al) = − .
62, and log δ (Ca) = 0to − .
58. These depletion values are similar to thosefound in warm, diffuse clouds in the lower Milky Wayhalo (Savage & Sembach 1996).5. Our study indicates that the enrichment his-tory of the Magellanic Stream as well as the physicalconditions in the Stream are more complex than pre-viously known. The abundances and gas-to-dust ratiosmeasured in the Stream along the Fairall 9 sightline aresubstantially higher than what is found along other MSsightlines. The high sulfur abundance in the gas possiblyindicates a substantial α enrichment from massive starsin a region of enhanced star-formation ∼ − REFERENCESAbgrall, H., & Roueff, E. 1989, A&A, 79, 313Anders, E., & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53,197Asplund, M., Grevesse, N., Jacques Sauval, A., & Scott, P. 2009,ARA&A, 47, 481Ben Bekhti, N., Richter, P., Westmeier, T., & Murphy, M.T. 2008,A&A, 487, 583Ben Bekhti, N., Winkel, B., Richter, P., et al. 2011, A&A, 542,110Besla, G., Kallivayalil, N., Hernquist, L., et al. 2007, ApJ, 668,949Besla, G., Kallivayalil, N., Hernquist, L., et al. 2010, ApJ, 721,L97Besla, G., Kallivayalil, N., Hernquist, L., et al. 2012, MNRAS,421, 2109Bland-Hawthorn, J., & Maloney, P. R. 1999, ApJ, 510, L33Bland-Hawthorn, J., Sutherland, R., Agertz, O., & Moore, B. 2007,
ApJ, 670, L109Bland-Hawthorn, J.,& Maloney, P. R. 2002, in ASP Conf. Ser.254, Extragalactic Gas at Low Redshift, ed. J. S. Mulchaey & J.T. Stocke (San Francisco: ASP), 267Br¨uns, C., Kerp, J., Stavely-Smith, L., et al. 2005, A&A, 432, 45Crawford, I.A. 1992, MNRAS, 259, 47Crighton, N. H. M., Bechtold, J., Carswell, R. F. et al. 2013, MN-RAS, in press (astro-ph/1210.0905)Connors, T. W., Kawata, D., Maddison, S. T., & Gibson, B. K.2004, PASA, 21, 222Connors, T. W., Kawata, D., & Gibson, B. K. 2006, MNRAS, 371,108Dalgarno, A., & McCray, R. A. 1972, ARA&A, 10, 375de Boer, K. S., Richter, P., Bomans, D. J., Heithausen, A., &Koornneef, J. A&A, 338, L5Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B., & Kotzlowski,H. 2000, SPIE, 4008, 534Diaz, J. D., & Bekki, K. 2011, ApJ, 413, 2015Draine, B., & Bertoldi, F. 1996, ApJ, 468, 269Dufour, R.J. 1975, ApJ, 195, 315Ferland, G. J., Korista, K. T., Verner, D. A. et al. 1998, PASP,110, 761Fontana, A., & Ballester, P. 1995, ESO Messenger, 80, 37Fox, A. J., Wakker, B. P., Savage, B. D., et al. 2005, ApJ, 630,332Fox, A.J., Wakker, B.P., Smoker, J.V., et al. 2010, ApJ, 718, 1046Fox, A.J., Richter, P., Wakker, B.P. et al. 2013, ApJ, in press(astro-ph/1304.4240)Gardiner, L.T. & Noguchi, M. 1996, MNRAS, 278, 191Gibson, B. K., Giroux, M. L., Penton, S. V., et al. 2000, AJ, 120,1830Green, J. C., Froning, C. S., Osterman, S., et al. 2012, ApJ, 744,60Harris, J., Zaritsky, D. 2004, AJ, 127, 1531Harris, J., Zaritsky, D. 2009, AJ, 138, 1243Heitsch, F. & Putman, M. E. 2009, ApJ, 698, 1485Henry, R. B. C., Edmunds, M. G, K¨oppen, J. 2000, ApJ, 541, 660Henry, R. B. C. & Prochaska, J. X. 2007, PASP, 119, 962Hughes, J. P., Hayashi, I., & Koyama, K. 1998, ApJ, 505, 732Irwin, M. J., Kunkel, W. E., Demers, S. 1985, Nature, 318, 160Jenkins, E. B., Bowen, D. V., Tripp, T. M., & Sembach, K. R.2005, ApJ, 623, 767Kalberla, P. M. W., Burton, W. B., Hartmann, D., et al. 2005,A&A, 440, 775Kalberla, P. M. W., McClure-Griffiths, N. M., Pisano, D. J., et al.2010, A&A, 521, 17Kallivayalil, N., van der Marel, R. P., Alcock, C., et al. 2006a,ApJ, 638, 772Kallivayalil, N., van der Marel, R. P., & Alcock, C. 2006b, ApJ,652, 1213Kallivayalil, N., van der Marel, R. P., Besla, G., Anderson, J., &Alcock, C. 2013, ApJ, 764, 161Keenan, F. P., Lennon, D. J., Johnson, C. T., & Kingston, A. E.1986, MNRAS, 220, 571Kobayashi, C., Umeda, H., Nomoto, K., Tominaga, N., & Ohkubo,T. 2006, ApJ, 653, 1145Koerwer, J.F. 2009, AJ, 138, 1Kriss, G. A. 2011,
COS
Instrument Science Report, 1Lehner, N. 2002, ApJ, 578, 126Lehner, N., Wakker, B.P., & Savage, B.D. 2004, ApJ, 615, 767Lehner, N., Howk, J.C., Keenan, F.P., & Smoker, J.V. 2008, ApJ678, 219Lu, L., Savage, B. D., & Sembach, K. R. 1994, ApJ, 437, L119Mastropietro, C., Moore, B.,Mayer, L.,Wadsley, J., & Stadel, J.2005, MNRAS, 363, 509Mathewson, D. S., Ford, V. L., Schwarz, M. P., & Murray, J. D.1979, IAUS, 84, 547Melnick, J. 1985, A&A, 153, 235Morton, D. C., & Dinerstein, H. L. 1976, ApJ, 204, 1Morton, D. C. 2003, ApJS, 149, 205McClure-Griffiths, N. M., Pisano, D. J., Calabretta, M. R., et al.2009, ApJS, 181, 398Nidever, D. L., Majewski, S. R., & Burton, W. B. 2008, ApJ, 679,432Nidever, D. L., Majewski, S. R., Burton, W. B., & Nigra, L. 2010,ApJ, 723, 1618Pagel, B. E. J., & Tautvaisiene, G. 1998, MNRAS, 299, 535Pettini, M., Ellison, S. L., Bergeron, J., & Petitjean, P. 2002, A&A,391, 21Pettini, M., Zych, B. J., Steidel, C. C., & Chaffee, F. H. 2008, MNRAS, 385, 2011Putman, M. E., Staveley-Smith, L., Freeman, K. C., Gibson, B.K., & Barnes, D. G. 2003, ApJ, 586, 170Ribaudo, J., Lehner, N., Howk, J. C., et al. 2011, ApJ, 743, 207Richter, P., Widmann, H., de Boer, K. S. et al. 1998, A&A, 338,L9Richter, P. 2000, A&A, 359, 1111Richter, P., Sembach, K. R., Wakker, B. P., & Savage, B. D., 2001a,ApJ, 562, L181Richter, P., Savage, B. D., Wakker, B. P., Sembach, K. R., &Kalberla, P. M. W. 2001b, ApJ, 549, 281Richter P., Wakker B. P., Savage B. D., & Sembach K. R. 2003,ApJ, 586, 230Richter, P., Westmeier, T., & Br¨uns 2005, A&A, 442, L49Richter, P. 2006, Rev. Mod. Astron., 19, 31Richter, P., Charlton, J. C., Fangano, A. P. M., Ben Bekhti, N., &Masiero, J. R. 2009, ApJ, 695, 1631Richter, P., Krause, F., Fechner, C., Charlton, J. C., & Murphy,M. T. 2011, A&A, 528, A12Russell, S. C., & Dopita, M. A. 1990, ApJ, 74, 93Russell, S. C., & Dopita, M. A. 1992, ApJ, 384, 508Savage, B. D., Drake, J. F., Budich, W., & Bohlin, R. C. 1977,ApJ, 216, 291Savage, B. D., & Sembach, K. R. 1991, ApJ, 379, 245Savage, B. D. & Sembach, K. R. 1996, ARA&A, 34, 279Sembach, K. R., Howk, J. C., Savage, B. D., & Shull, J. M. 2001,AJ, 121, 992Songaila, A. 1981, ApJ, 243, L19Spitzer, L. 1978,
Physical processes in the interstellar medium ,(New York Wiley-Interscience)Thom, C., Werk, J. K., Tumlinson, J., et al. 2011, ApJ, 736, 1Tumlinson, J., Shull, J. M., Rachford, B. L., et al. 2002, ApJ, 566,857Tumlinson, J., Werk, J. K., Thom, C., et al. 2011, ApJ, 733, 111Wakker, B. P., Howk, J. C., Savage, B. D., et al. 1999, Nature,402, 388Wakker, B. P., Oosterloo, T. A., & Putman, M. E. 2002, AJ, 123,1953Wakker, B.P. 2006, ApJS, 163, 282Wakker, B. P., York, D. G., Howk, J. C., et al. 2007, ApJ, 670,L113Wakker, B. P., York, D. G., Wilhelm, R., et al. 2008, ApJ, 672,298Wakker, B. P., Lockman, F. J., & Brown, J. M. 2011, ApJ, 728,159Wannier, P., & Wrixon, G. T. 1972, ApJ, 173, L119Weiner, B. J., & Williams, T. B. 1996, AJ, 111, 1156Weisz, D. R., Doplhin, A. E., Skillmann, E. D., et al. 2013, MN-RAS, 431, 364Welty, D. E., Morton, D. C., & Hobbs, L.M. 1996, ApJS, 106, 533Welty, D. E., Lauroesch, J. T., Blades, J. C., Hobbs, L.M., & York,D. G. 1997, ApJ, 489, 672Welty, D. E., Federman, S. R., Gredel, R., Thorburn, J. A., &Lambert, D. L. 2006, ApJS, 165, 138Wolfire, M. G., McKee, C. F., Hollenbach, D., & Tielens, A. G. G.M. 1995, ApJ, 453, 673York, D. G., Songaila, A., Blades, J. C., et al. 1982, ApJ, 255, 467Zech, W. F., Lehner, N., Howk, J. C., Dixon, W. V. D., & Brown,T. M. 2008, ApJ, 679, 460 he Magellanic Stream toward Fairall 9 19
Fig. 10.— H i Ly α absorption in the COS spectrum of Fairall 9 in the wavelength range between 1195 and 1235 ˚A. The red solid lineindicates a two-component absorption model with a Milky Way H i column density of log N (H i ) = 20 .
25 and a MS H i column density oflog N (H i ) = 19 .
95. APPENDIX
H I LY α ABSORPTION
In Fig. 10 we show the normalized COS spectrum in the wavelength range between 1195 and 1235 ˚A. Strong H i Ly α absorption centered at 1216 ˚A from neutral gas in the Milky Way (MW) and in the Magellanic Stream withwell-defined damping wings is visible. The 21cm data from various observations and instruments (Sect. 3.3; Gibson etal. 2000) indicate that the Milky Way disk gas dominates the neutral-gas column density along this sightline.The strength and shape of the Ly α absorption toward Fairall 9 supports this conclusion, but the relative contributionof MW and MS cannot be tightly constrained from a Voigt-profile fit of the Ly α absorption. The red solid line in Fig. 10shows a two-component absorption model with a MW H i column density of log N (H i ) = 20 .
25 and a MS H i columndensity of log N (H i ) = 19 .
95. The Ly α absorption profile is fully consistent with the value of log N (H i ) = 19 .
95 inMS, as indicated by the GASS 21cm data. However, the Ly α profile is also consistent with MS column densities in therange log N (H i ) = 19 . − .
15, if the MW contribution is modified accordingly. In conclusion, the Ly α absorptiondoes not provide constraining limits on the H i column density in the Magellanic Stream toward Fairall 9 that couldhelp to minimize the systematic errors for the determination of log N (H i ), as discussed in Sect. 3.3. ABSORPTION MODELING OF UV LINES
In Sect. 2.5 we have outlined the method that we have used to model the UV lines of the low ions in the COS spectrumof Fairall 9 to infer column densities of these species. We here present additional information on the modeling procedureto provide an insight into the allowed parameter range for ( N i , b i ) and N tot for each ion.In Fig. 11 we show velocity plots of the low ions from synthetic spectra that have been generated by us based ondifferent input models, in which we have systematically varied the parameters N i and b i . The resulting model profilesare compared to the best-fitting model presented in Fig. 3 and Table 1.In the left panel of Fig. 11 (a), the best-fitting model is indicated with the black solid line, while the lines in red,green, and blue show models, in which all column densities N i in the individual subcomponents have been increasedby 0 .
3, 0 .
5, and 1 . b values remain unchanged. In the middle panelof Fig. 11 (b), the column densities N i in the individual subcomponents have been decreased by 0 .
3, 0 .
5, and 1 . v i and b i . These two figures indicate that the shapes of the strong andsaturated lines from C ii , O i and Si ii are basically not affected by the change in column density. In contrast, theabsorption depths of the weaker lines of N i , Al ii , Si ii , S ii , and Fe ii are strongly affected by the increase or decreaseof the column densities with respect to the reference model (black), demonstrating that these lines represent sensitivediagnostics to constrain N i and N tot for these ions in the Stream.In the right panel of Fig. 11 (c) we show a model in which we have increased the Doppler parameters b i for theindividual subcomponents by factors of 1 . . . b i , the changes in b i are irrelevant for the shape of theweak, unsaturated lines, while the strong, saturated lines mildly grow in strength in their absorption wings. Therefore,even if we would underestimate the values of b i for the singly-ionized species Al ii , Si ii , S ii , and Fe ii when adoptingthe values from Ca ii , as one could argue because these ions may trace a slightly different (more extended) gas phase,our column-density estimate would not be affected at all by this systematic error.These figures, together with Fig. 2, underline that the total gas column densities of N i , Al ii , Si ii , S ii , and Fe ii inthe MS, as listed in Table 1, are well constrained by the COS data and our absorption model.0 P. Richter et al. Fig. 11.—
Velocity profiles of low ions, as derived from synthetic spectra from different input models for the gas absorption in theMagellanic Stream toward Fairall 9. The best-fitting model is indicated with the black solid line. The lines in red, green, and blue showtest models, in which the input parameters N ii
Velocity profiles of low ions, as derived from synthetic spectra from different input models for the gas absorption in theMagellanic Stream toward Fairall 9. The best-fitting model is indicated with the black solid line. The lines in red, green, and blue showtest models, in which the input parameters N ii , b ii